diff options
Diffstat (limited to 'src')
495 files changed, 53346 insertions, 16913 deletions
diff --git a/src/SConscript b/src/SConscript index 5440acd1351..6083fcbec98 100644 --- a/src/SConscript +++ b/src/SConscript @@ -1,5 +1,6 @@ Import('*') +SConscript('glsl/SConscript') SConscript('gallium/SConscript') if 'mesa' in env['statetrackers']: diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 23d8b609e17..14375426ed8 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -35,6 +35,8 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" + +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index bcb860da2e0..0cc2b718641 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -38,6 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" diff --git a/src/gallium/auxiliary/rbug/rbug_connection.c b/src/gallium/auxiliary/rbug/rbug_connection.c index 52acb700af9..ae4e27f9f6b 100644 --- a/src/gallium/auxiliary/rbug/rbug_connection.c +++ b/src/gallium/auxiliary/rbug/rbug_connection.c @@ -87,6 +87,7 @@ rbug_get_message(struct rbug_connection *c, uint32_t *serial) if (!data) { return NULL; } + data->opcode = 0; do { uint8_t *ptr = ((uint8_t*)data) + read; diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 4092f78f4a2..92903fe57f3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -942,3 +942,107 @@ tgsi_default_full_dst_register( void ) return full_dst_register; } +struct tgsi_property +tgsi_default_property( void ) +{ + struct tgsi_property property; + + property.Type = TGSI_TOKEN_TYPE_PROPERTY; + property.NrTokens = 1; + property.PropertyName = TGSI_PROPERTY_GS_INPUT_PRIM; + property.Padding = 0; + + return property; +} + +struct tgsi_property +tgsi_build_property(unsigned property_name, + struct tgsi_header *header) +{ + struct tgsi_property property; + + property = tgsi_default_property(); + property.PropertyName = property_name; + + header_bodysize_grow( header ); + + return property; +} + + +struct tgsi_full_property +tgsi_default_full_property( void ) +{ + struct tgsi_full_property full_property; + + full_property.Property = tgsi_default_property(); + memset(full_property.u, 0, + sizeof(struct tgsi_property_data) * 8); + + return full_property; +} + +static void +property_grow( + struct tgsi_property *property, + struct tgsi_header *header ) +{ + assert( property->NrTokens < 0xFF ); + + property->NrTokens++; + + header_bodysize_grow( header ); +} + +struct tgsi_property_data +tgsi_build_property_data( + unsigned value, + struct tgsi_property *property, + struct tgsi_header *header ) +{ + struct tgsi_property_data property_data; + + property_data.Data = value; + + property_grow( property, header ); + + return property_data; +} + +unsigned +tgsi_build_full_property( + const struct tgsi_full_property *full_prop, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0, i; + struct tgsi_property *property; + + if( maxsize <= size ) + return 0; + property = (struct tgsi_property *) &tokens[size]; + size++; + + *property = tgsi_build_property( + TGSI_PROPERTY_GS_INPUT_PRIM, + header ); + + assert( full_prop->Property.NrTokens <= 8 + 1 ); + + for( i = 0; i < full_prop->Property.NrTokens - 1; i++ ) { + struct tgsi_property_data *data; + + if( maxsize <= size ) + return 0; + data = (struct tgsi_property_data *) &tokens[size]; + size++; + + *data = tgsi_build_property_data( + full_prop->u[i].Data, + property, + header ); + } + + return size; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h index ffea786770c..9de2757fe40 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -127,6 +127,34 @@ tgsi_build_full_immediate( unsigned maxsize ); /* + * properties + */ + +struct tgsi_property +tgsi_default_property( void ); + +struct tgsi_property +tgsi_build_property( + unsigned property_name, + struct tgsi_header *header ); + +struct tgsi_full_property +tgsi_default_full_property( void ); + +struct tgsi_property_data +tgsi_build_property_data( + unsigned value, + struct tgsi_property *property, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_property( + const struct tgsi_full_property *full_prop, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* * instruction */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index d3a5da4b2b0..5e7e5d2ff9c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -101,7 +101,8 @@ static const char *file_names[TGSI_FILE_COUNT] = "ADDR", "IMM", "LOOP", - "PRED" + "PRED", + "SV" }; static const char *interpolate_names[] = @@ -150,6 +151,27 @@ static const char *texture_names[] = "SHADOWRECT" }; +static const char *property_names[] = +{ + "GS_INPUT_PRIMITIVE", + "GS_OUTPUT_PRIMITIVE", + "GS_MAX_OUTPUT_VERTICES" +}; + +static const char *primitive_names[] = +{ + "POINTS", + "LINES", + "LINE_LOOP", + "LINE_STRIP", + "TRIANGLES", + "TRIANGLE_STRIP", + "TRIANGLE_FAN", + "QUADS", + "QUAD_STRIP", + "POLYGON" +}; + static void _dump_register( @@ -274,6 +296,50 @@ tgsi_dump_declaration( } static boolean +iter_property( + struct tgsi_iterate_context *iter, + struct tgsi_full_property *prop ) +{ + int i; + struct dump_ctx *ctx = (struct dump_ctx *)iter; + + assert(Elements(property_names) == TGSI_PROPERTY_COUNT); + + TXT( "PROPERTY " ); + ENM(prop->Property.PropertyName, property_names); + + if (prop->Property.NrTokens > 1) + TXT(" "); + + for (i = 0; i < prop->Property.NrTokens - 1; ++i) { + switch (prop->Property.PropertyName) { + case TGSI_PROPERTY_GS_INPUT_PRIM: + case TGSI_PROPERTY_GS_OUTPUT_PRIM: + ENM(prop->u[i].Data, primitive_names); + break; + default: + SID( prop->u[i].Data ); + break; + } + if (i < prop->Property.NrTokens - 2) + TXT( ", " ); + } + EOL(); + + return TRUE; +} + +void tgsi_dump_property( + const struct tgsi_full_property *prop ) +{ + struct dump_ctx ctx; + + ctx.printf = dump_ctx_printf; + + iter_property( &ctx.iter, (struct tgsi_full_property *)prop ); +} + +static boolean iter_immediate( struct tgsi_iterate_context *iter, struct tgsi_full_immediate *imm ) @@ -493,6 +559,7 @@ tgsi_dump( ctx.iter.iterate_instruction = iter_instruction; ctx.iter.iterate_declaration = iter_declaration; ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.iterate_property = iter_property; ctx.iter.epilog = NULL; ctx.instno = 0; @@ -547,6 +614,7 @@ tgsi_dump_str( ctx.base.iter.iterate_instruction = iter_instruction; ctx.base.iter.iterate_declaration = iter_declaration; ctx.base.iter.iterate_immediate = iter_immediate; + ctx.base.iter.iterate_property = iter_property; ctx.base.iter.epilog = NULL; ctx.base.instno = 0; diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.h b/src/gallium/auxiliary/tgsi/tgsi_dump.h index ad1e647ec90..4cd27317b36 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.h @@ -49,6 +49,7 @@ tgsi_dump( struct tgsi_full_immediate; struct tgsi_full_instruction; struct tgsi_full_declaration; +struct tgsi_full_property; void tgsi_dump_immediate( @@ -63,6 +64,10 @@ void tgsi_dump_declaration( const struct tgsi_full_declaration *decl ); +void +tgsi_dump_property( + const struct tgsi_full_property *prop ); + #if defined __cplusplus } #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 22984c32320..123117cb0a3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -336,6 +336,9 @@ tgsi_exec_machine_bind_shader( numInstructions++; break; + case TGSI_TOKEN_TYPE_PROPERTY: + break; + default: assert( 0 ); } @@ -1158,6 +1161,7 @@ fetch_src_file_channel( break; case TGSI_FILE_INPUT: + case TGSI_FILE_SYSTEM_VALUE: chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; @@ -1302,6 +1306,7 @@ fetch_source( */ switch (reg->Register.File) { case TGSI_FILE_INPUT: + case TGSI_FILE_SYSTEM_VALUE: index.i[0] *= TGSI_EXEC_MAX_INPUT_ATTRIBS; index.i[1] *= TGSI_EXEC_MAX_INPUT_ATTRIBS; index.i[2] *= TGSI_EXEC_MAX_INPUT_ATTRIBS; @@ -1892,7 +1897,8 @@ exec_declaration(struct tgsi_exec_machine *mach, const struct tgsi_full_declaration *decl) { if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - if (decl->Declaration.File == TGSI_FILE_INPUT) { + if (decl->Declaration.File == TGSI_FILE_INPUT || + decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE) { uint first, last, mask; first = decl->Range.First; @@ -3240,6 +3246,25 @@ exec_instruction( case TGSI_OPCODE_NOP: break; + case TGSI_OPCODE_BREAKC: + FETCH(&r[0], 0, CHAN_X); + /* update CondMask */ + if (r[0].u[0] && (mach->ExecMask & 0x1)) { + mach->LoopMask &= ~0x1; + } + if (r[0].u[1] && (mach->ExecMask & 0x2)) { + mach->LoopMask &= ~0x2; + } + if (r[0].u[2] && (mach->ExecMask & 0x4)) { + mach->LoopMask &= ~0x4; + } + if (r[0].u[3] && (mach->ExecMask & 0x8)) { + mach->LoopMask &= ~0x8; + } + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/tgsi_iterate.c index 7b384f5e12a..0ba5fe48419 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_iterate.c +++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.c @@ -66,6 +66,12 @@ tgsi_iterate_shader( goto fail; break; + case TGSI_TOKEN_TYPE_PROPERTY: + if (ctx->iterate_property) + if (!ctx->iterate_property( ctx, &parse.FullToken.FullProperty )) + goto fail; + break; + default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/tgsi_iterate.h index ef5a33ebce9..8d67f22c429 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_iterate.h +++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.h @@ -57,6 +57,11 @@ struct tgsi_iterate_context struct tgsi_full_immediate *imm ); boolean + (* iterate_property)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_property *prop ); + + boolean (* epilog)( struct tgsi_iterate_context *ctx ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index 8f2b6a307d3..fa65ecb9975 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -220,6 +220,22 @@ tgsi_parse_token( break; } + case TGSI_TOKEN_TYPE_PROPERTY: + { + struct tgsi_full_property *prop = &ctx->FullToken.FullProperty; + uint prop_count; + + memset(prop, 0, sizeof *prop); + copy_token(&prop->Property, &token); + + prop_count = prop->Property.NrTokens - 1; + for (i = 0; i < prop_count; i++) { + next_token(ctx, &prop->u[i]); + } + + break; + } + default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h index 3aa1979a63a..439a57269b7 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h @@ -67,6 +67,12 @@ struct tgsi_full_immediate union tgsi_immediate_data u[4]; }; +struct tgsi_full_property +{ + struct tgsi_property Property; + struct tgsi_property_data u[8]; +}; + #define TGSI_FULL_MAX_DST_REGISTERS 2 #define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */ @@ -86,6 +92,7 @@ union tgsi_full_token struct tgsi_full_declaration FullDeclaration; struct tgsi_full_immediate FullImmediate; struct tgsi_full_instruction FullInstruction; + struct tgsi_full_property FullProperty; }; struct tgsi_parse_context diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c index da6ad6da04c..138d2d095bb 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c @@ -293,6 +293,7 @@ emit_fetch(struct gen_context *gen, case TGSI_SWIZZLE_W: switch (reg->Register.File) { case TGSI_FILE_INPUT: + case TGSI_FILE_SYSTEM_VALUE: { int offset = (reg->Register.Index * 4 + swizzle) * 16; int offset_reg = emit_li_offset(gen, offset); @@ -1173,7 +1174,8 @@ emit_declaration( struct ppc_function *func, struct tgsi_full_declaration *decl ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT || + decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE ) { #if 0 unsigned first, last, mask; unsigned i, j; @@ -1339,6 +1341,9 @@ tgsi_emit_ppc(const struct tgsi_token *tokens, } break; + case TGSI_TOKEN_TYPE_PROPERTY: + break; + default: ok = 0; assert( 0 ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index b5d1faa897a..c27579e7942 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -324,6 +324,17 @@ iter_immediate( return TRUE; } + +static boolean +iter_property( + struct tgsi_iterate_context *iter, + struct tgsi_full_property *prop ) +{ + /*struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;*/ + + return TRUE; +} + static boolean epilog( struct tgsi_iterate_context *iter ) @@ -367,6 +378,7 @@ tgsi_sanity_check( ctx.iter.iterate_instruction = iter_instruction; ctx.iter.iterate_declaration = iter_declaration; ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.iterate_property = iter_property; ctx.iter.epilog = epilog; memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c index ebee000c06e..0f48b0dc3a1 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c @@ -97,7 +97,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens, for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) { const struct tgsi_full_src_register *src = &fullinst->Src[i]; - if (src->Register.File == TGSI_FILE_INPUT) { + if (src->Register.File == TGSI_FILE_INPUT || + src->Register.File == TGSI_FILE_SYSTEM_VALUE) { const int ind = src->Register.Index; if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FOG) { if (src->Register.SwizzleX == TGSI_SWIZZLE_X) { @@ -128,7 +129,7 @@ tgsi_scan_shader(const struct tgsi_token *tokens, info->file_count[file]++; info->file_max[file] = MAX2(info->file_max[file], (int)reg); - if (file == TGSI_FILE_INPUT) { + if (file == TGSI_FILE_INPUT || file == TGSI_FILE_SYSTEM_VALUE) { info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.Name; info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.Index; info->input_interpolate[reg] = (ubyte)fulldecl->Declaration.Interpolate; @@ -164,6 +165,19 @@ tgsi_scan_shader(const struct tgsi_token *tokens, info->file_max[file] = MAX2(info->file_max[file], (int)reg); } break; + case TGSI_TOKEN_TYPE_PROPERTY: + { + const struct tgsi_full_property *fullprop + = &parse.FullToken.FullProperty; + + info->properties[info->num_properties].name = + fullprop->Property.PropertyName; + memcpy(info->properties[info->num_properties].data, + fullprop->u, 8 * sizeof(unsigned));; + + ++info->num_properties; + } + break; default: assert( 0 ); @@ -216,6 +230,7 @@ tgsi_is_passthrough_shader(const struct tgsi_token *tokens) /* Do a whole bunch of checks for a simple move */ if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV || src->Register.File != TGSI_FILE_INPUT || + src->Register.File != TGSI_FILE_SYSTEM_VALUE || dst->Register.File != TGSI_FILE_OUTPUT || src->Register.Index != dst->Register.Index || @@ -239,6 +254,8 @@ tgsi_is_passthrough_shader(const struct tgsi_token *tokens) /* fall-through */ case TGSI_TOKEN_TYPE_IMMEDIATE: /* fall-through */ + case TGSI_TOKEN_TYPE_PROPERTY: + /* fall-through */ default: ; /* no-op */ } diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h index 3ce0e88e4c6..dae5376c24a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h @@ -33,7 +33,6 @@ #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" - /** * Shader summary info */ @@ -62,8 +61,13 @@ struct tgsi_shader_info boolean uses_kill; /**< KIL or KILP instruction used? */ boolean uses_fogcoord; /**< fragment shader uses fog coord? */ boolean uses_frontfacing; /**< fragment shader uses front/back-face flag? */ -}; + struct { + unsigned name; + unsigned data[8]; + } properties[TGSI_PROPERTY_COUNT]; + uint num_properties; +}; extern void tgsi_scan_shader(const struct tgsi_token *tokens, diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 76051ea0d8e..d63c75dafb3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1288,6 +1288,7 @@ emit_fetch( break; case TGSI_FILE_INPUT: + case TGSI_FILE_SYSTEM_VALUE: emit_inputf( func, xmm, @@ -2633,7 +2634,8 @@ emit_declaration( struct x86_function *func, struct tgsi_full_declaration *decl ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT || + decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE ) { unsigned first, last, mask; unsigned i, j; @@ -2952,6 +2954,9 @@ tgsi_emit_sse2( num_immediates++; } break; + case TGSI_TOKEN_TYPE_PROPERTY: + /* we just ignore them for now */ + break; default: ok = 0; diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index eb376fa9572..f000958bfc0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -27,6 +27,7 @@ #include "util/u_debug.h" #include "util/u_memory.h" +#include "pipe/p_defines.h" #include "tgsi_text.h" #include "tgsi_build.h" #include "tgsi_info.h" @@ -59,6 +60,21 @@ static boolean uprcase( char c ) return c; } +/* + * Ignore case of str1 and assume str2 is already uppercase. + * Return TRUE iff str1 and str2 are equal. + */ +static int +streq_nocase_uprcase(const char *str1, + const char *str2) +{ + while (*str1 && uprcase(*str1) == *str2) { + str1++; + str2++; + } + return *str1 == *str2; +} + static boolean str_match_no_case( const char **pcur, const char *str ) { const char *cur = *pcur; @@ -110,6 +126,20 @@ static boolean parse_uint( const char **pcur, uint *val ) return FALSE; } +static boolean parse_identifier( const char **pcur, char *ret ) +{ + const char *cur = *pcur; + int i = 0; + if (is_alpha_underscore( cur )) { + ret[i++] = *cur++; + while (is_alpha_underscore( cur )) + ret[i++] = *cur++; + *pcur = cur; + return TRUE; + } + return FALSE; +} + /* Parse floating point. */ static boolean parse_float( const char **pcur, float *val ) @@ -229,7 +259,8 @@ static const char *file_names[TGSI_FILE_COUNT] = "ADDR", "IMM", "LOOP", - "PRED" + "PRED", + "SV" }; static boolean @@ -939,6 +970,106 @@ static boolean parse_immediate( struct translate_ctx *ctx ) return TRUE; } +static const char *property_names[] = +{ + "GS_INPUT_PRIMITIVE", + "GS_OUTPUT_PRIMITIVE", + "GS_MAX_OUTPUT_VERTICES" +}; + +static const char *primitive_names[] = +{ + "POINTS", + "LINES", + "LINE_LOOP", + "LINE_STRIP", + "TRIANGLES", + "TRIANGLE_STRIP", + "TRIANGLE_FAN", + "QUADS", + "QUAD_STRIP", + "POLYGON" +}; + +static boolean +parse_primitive( const char **pcur, uint *primitive ) +{ + uint i; + + for (i = 0; i < PIPE_PRIM_MAX; i++) { + const char *cur = *pcur; + + if (str_match_no_case( &cur, primitive_names[i])) { + *primitive = i; + *pcur = cur; + return TRUE; + } + } + return FALSE; +} + + +static boolean parse_property( struct translate_ctx *ctx ) +{ + struct tgsi_full_property prop; + uint property_name; + uint values[8]; + uint advance; + char id[64]; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (!parse_identifier( &ctx->cur, id )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + for (property_name = 0; property_name < TGSI_PROPERTY_COUNT; + ++property_name) { + if (streq_nocase_uprcase(id, property_names[property_name])) { + break; + } + } + if (property_name >= TGSI_PROPERTY_COUNT) { + debug_printf( "\nError: Unknown property : '%s'", id ); + return FALSE; + } + + eat_opt_white( &ctx->cur ); + switch(property_name) { + case TGSI_PROPERTY_GS_INPUT_PRIM: + case TGSI_PROPERTY_GS_OUTPUT_PRIM: + if (!parse_primitive(&ctx->cur, &values[0] )) { + report_error( ctx, "Unknown primitive name as property!" ); + return FALSE; + } + break; + default: + if (!parse_uint(&ctx->cur, &values[0] )) { + report_error( ctx, "Expected unsigned integer as property!" ); + return FALSE; + } + } + + prop = tgsi_default_full_property(); + prop.Property.PropertyName = property_name; + prop.Property.NrTokens += 1; + prop.u[0].Data = values[0]; + + advance = tgsi_build_full_property( + &prop, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + + static boolean translate( struct translate_ctx *ctx ) { eat_opt_white( &ctx->cur ); @@ -947,7 +1078,6 @@ static boolean translate( struct translate_ctx *ctx ) while (*ctx->cur != '\0') { uint label_val = 0; - if (!eat_white( &ctx->cur )) { report_error( ctx, "Syntax error" ); return FALSE; @@ -955,7 +1085,6 @@ static boolean translate( struct translate_ctx *ctx ) if (*ctx->cur == '\0') break; - if (parse_label( ctx, &label_val )) { if (!parse_instruction( ctx, TRUE )) return FALSE; @@ -968,6 +1097,10 @@ static boolean translate( struct translate_ctx *ctx ) if (!parse_immediate( ctx )) return FALSE; } + else if (str_match_no_case( &ctx->cur, "PROPERTY" )) { + if (!parse_property( ctx )) + return FALSE; + } else if (!parse_instruction( ctx, FALSE )) { return FALSE; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c index 8b8f489b355..ae875f29abf 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.c +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c @@ -79,6 +79,19 @@ emit_immediate(struct tgsi_transform_context *ctx, } +static void +emit_property(struct tgsi_transform_context *ctx, + const struct tgsi_full_property *prop) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_property(prop, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + /** * Apply user-defined transformations to the input shader to produce @@ -110,6 +123,7 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in, ctx->emit_instruction = emit_instruction; ctx->emit_declaration = emit_declaration; ctx->emit_immediate = emit_immediate; + ctx->emit_property = emit_property; ctx->tokens_out = tokens_out; ctx->max_tokens_out = max_tokens_out; @@ -182,6 +196,17 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in, ctx->emit_immediate(ctx, fullimm); } break; + case TGSI_TOKEN_TYPE_PROPERTY: + { + struct tgsi_full_property *fullprop + = &parse.FullToken.FullProperty; + + if (ctx->transform_property) + ctx->transform_property(ctx, fullprop); + else + ctx->emit_property(ctx, fullprop); + } + break; default: assert( 0 ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.h b/src/gallium/auxiliary/tgsi/tgsi_transform.h index a121adbaef4..818478e277a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.h +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.h @@ -53,6 +53,8 @@ struct tgsi_transform_context void (*transform_immediate)(struct tgsi_transform_context *ctx, struct tgsi_full_immediate *imm); + void (*transform_property)(struct tgsi_transform_context *ctx, + struct tgsi_full_property *prop); /** * Called at end of input program to allow caller to append extra @@ -73,6 +75,8 @@ struct tgsi_transform_context const struct tgsi_full_declaration *decl); void (*emit_immediate)(struct tgsi_transform_context *ctx, const struct tgsi_full_immediate *imm); + void (*emit_property)(struct tgsi_transform_context *ctx, + const struct tgsi_full_property *prop); struct tgsi_header *header; uint max_tokens_out; diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c index 3f943845f5b..1e730e53427 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -90,6 +90,11 @@ struct ureg_program unsigned vs_inputs[UREG_MAX_INPUT/32]; struct { + unsigned index; + } gs_input[UREG_MAX_INPUT]; + unsigned nr_gs_inputs; + + struct { unsigned semantic_name; unsigned semantic_index; } output[UREG_MAX_OUTPUT]; @@ -278,6 +283,22 @@ ureg_DECL_vs_input( struct ureg_program *ureg, } +struct ureg_src +ureg_DECL_gs_input(struct ureg_program *ureg, + unsigned index) +{ + if (ureg->nr_gs_inputs < UREG_MAX_INPUT) { + ureg->gs_input[ureg->nr_gs_inputs].index = index; + ureg->nr_gs_inputs++; + } else { + set_bad(ureg); + } + + /* XXX: Add suport for true 2D input registers. */ + return ureg_src_register(TGSI_FILE_INPUT, index); +} + + struct ureg_dst ureg_DECL_output( struct ureg_program *ureg, unsigned name, @@ -964,8 +985,7 @@ static void emit_decls( struct ureg_program *ureg ) emit_decl_range( ureg, TGSI_FILE_INPUT, i, 1 ); } } - } - else { + } else if (ureg->processor == TGSI_PROCESSOR_FRAGMENT) { for (i = 0; i < ureg->nr_fs_inputs; i++) { emit_decl( ureg, TGSI_FILE_INPUT, @@ -974,6 +994,13 @@ static void emit_decls( struct ureg_program *ureg ) ureg->fs_input[i].semantic_index, ureg->fs_input[i].interp ); } + } else { + for (i = 0; i < ureg->nr_gs_inputs; i++) { + emit_decl_range(ureg, + TGSI_FILE_INPUT, + ureg->gs_input[i].index, + 1); + } } for (i = 0; i < ureg->nr_outputs; i++) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h index 94cc70a2082..7e3e7bcf1d3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -133,6 +133,10 @@ struct ureg_src ureg_DECL_vs_input( struct ureg_program *, unsigned index ); +struct ureg_src +ureg_DECL_gs_input(struct ureg_program *, + unsigned index); + struct ureg_dst ureg_DECL_output( struct ureg_program *, unsigned semantic_name, diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 1d8bb55bbd6..3ed90fd1b70 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -9,8 +9,10 @@ C_SOURCES = \ u_debug_symbol.c \ u_debug_stack.c \ u_blit.c \ + u_blitter.c \ u_cache.c \ u_cpu_detect.c \ + u_dl.c \ u_draw_quad.c \ u_format.c \ u_format_access.c \ @@ -30,6 +32,7 @@ C_SOURCES = \ u_stream_stdc.c \ u_stream_wd.c \ u_surface.c \ + u_texture.c \ u_tile.c \ u_time.c \ u_timed_winsys.c \ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 8d99106d0b8..2a546d19dc0 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -23,6 +23,7 @@ util = env.ConvenienceLibrary( source = [ 'u_bitmask.c', 'u_blit.c', + 'u_blitter.c', 'u_cache.c', 'u_cpu_detect.c', 'u_debug.c', @@ -30,6 +31,7 @@ util = env.ConvenienceLibrary( 'u_debug_memory.c', 'u_debug_stack.c', 'u_debug_symbol.c', + 'u_dl.c', 'u_draw_quad.c', 'u_format.c', 'u_format_access.c', @@ -48,6 +50,7 @@ util = env.ConvenienceLibrary( 'u_stream_stdc.c', 'u_stream_wd.c', 'u_surface.c', + 'u_texture.c', 'u_tile.c', 'u_time.c', 'u_timed_winsys.c', diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index abe1de3302b..3f74e2aa8b8 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -42,6 +42,7 @@ #include "util/u_blit.h" #include "util/u_draw_quad.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_simple_shaders.h" @@ -126,7 +127,8 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) } /* fragment shader */ - ctx->fs[TGSI_WRITEMASK_XYZW] = util_make_fragment_tex_shader(pipe); + ctx->fs[TGSI_WRITEMASK_XYZW] = + util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D); ctx->vbuf = NULL; /* init vertex data that doesn't change */ @@ -420,7 +422,9 @@ util_blit_pixels_writemask(struct blit_state *ctx, cso_set_sampler_textures(ctx->cso, 1, &tex); if (ctx->fs[writemask] == NULL) - ctx->fs[writemask] = util_make_fragment_tex_shader_writemask(pipe, writemask); + ctx->fs[writemask] = + util_make_fragment_tex_shader_writemask(pipe, TGSI_TEXTURE_2D, + writemask); /* shaders */ cso_set_fragment_shader_handle(ctx->cso, ctx->fs[writemask]); diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c new file mode 100644 index 00000000000..1f794d39a17 --- /dev/null +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -0,0 +1,719 @@ +/************************************************************************** + * + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Blitter utility to facilitate acceleration of the clear, surface_copy, + * and surface_fill functions. + * + * @author Marek Olšák + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_state.h" + +#include "util/u_format.h" +#include "util/u_memory.h" +#include "util/u_math.h" +#include "util/u_blitter.h" +#include "util/u_draw_quad.h" +#include "util/u_pack_color.h" +#include "util/u_rect.h" +#include "util/u_simple_shaders.h" +#include "util/u_texture.h" + +struct blitter_context_priv +{ + struct blitter_context blitter; + + struct pipe_context *pipe; /**< pipe context */ + struct pipe_buffer *vbuf; /**< quad */ + + float vertices[4][2][4]; /**< {pos, color} or {pos, texcoord} */ + + /* Templates for various state objects. */ + struct pipe_depth_stencil_alpha_state template_dsa; + struct pipe_sampler_state template_sampler_state; + + /* Constant state objects. */ + /* Vertex shaders. */ + void *vs_col; /**< Vertex shader which passes {pos, color} to the output */ + void *vs_tex; /**<Vertex shader which passes {pos, texcoord} to the output.*/ + + /* Fragment shaders. */ + /* FS which outputs a color to multiple color buffers. */ + void *fs_col[PIPE_MAX_COLOR_BUFS]; + + /* FS which outputs a color from a texture, + where the index is PIPE_TEXTURE_* to be sampled. */ + void *fs_texfetch_col[PIPE_MAX_TEXTURE_TYPES]; + + /* FS which outputs a depth from a texture, + where the index is PIPE_TEXTURE_* to be sampled. */ + void *fs_texfetch_depth[PIPE_MAX_TEXTURE_TYPES]; + + /* Blend state. */ + void *blend_write_color; /**< blend state with writemask of RGBA */ + void *blend_keep_color; /**< blend state with writemask of 0 */ + + /* Depth stencil alpha state. */ + void *dsa_write_depth_stencil[0xff]; /**< indices are stencil clear values */ + void *dsa_write_depth_keep_stencil; + void *dsa_keep_depth_stencil; + + /* Sampler state for clamping to a miplevel. */ + void *sampler_state[PIPE_MAX_TEXTURE_LEVELS]; + + /* Rasterizer state. */ + void *rs_state; +}; + +struct blitter_context *util_blitter_create(struct pipe_context *pipe) +{ + struct blitter_context_priv *ctx; + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state *dsa; + struct pipe_rasterizer_state rs_state; + struct pipe_sampler_state *sampler_state; + unsigned i; + + ctx = CALLOC_STRUCT(blitter_context_priv); + if (!ctx) + return NULL; + + ctx->pipe = pipe; + + /* init state objects for them to be considered invalid */ + ctx->blitter.saved_fb_state.nr_cbufs = ~0; + ctx->blitter.saved_num_textures = ~0; + ctx->blitter.saved_num_sampler_states = ~0; + + /* blend state objects */ + memset(&blend, 0, sizeof(blend)); + ctx->blend_keep_color = pipe->create_blend_state(pipe, &blend); + + blend.colormask = PIPE_MASK_RGBA; + ctx->blend_write_color = pipe->create_blend_state(pipe, &blend); + + /* depth stencil alpha state objects */ + dsa = &ctx->template_dsa; + ctx->dsa_keep_depth_stencil = + pipe->create_depth_stencil_alpha_state(pipe, dsa); + + dsa->depth.enabled = 1; + dsa->depth.writemask = 1; + dsa->depth.func = PIPE_FUNC_ALWAYS; + ctx->dsa_write_depth_keep_stencil = + pipe->create_depth_stencil_alpha_state(pipe, dsa); + + dsa->stencil[0].enabled = 1; + dsa->stencil[0].func = PIPE_FUNC_ALWAYS; + dsa->stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE; + dsa->stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE; + dsa->stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE; + dsa->stencil[0].valuemask = 0xff; + dsa->stencil[0].writemask = 0xff; + /* The DSA state objects which write depth and stencil are created + * on-demand. */ + + /* sampler state */ + sampler_state = &ctx->template_sampler_state; + sampler_state->wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler_state->wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler_state->wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + /* The sampler state objects which sample from a specified mipmap level + * are created on-demand. */ + + /* rasterizer state */ + memset(&rs_state, 0, sizeof(rs_state)); + rs_state.front_winding = PIPE_WINDING_CW; + rs_state.cull_mode = PIPE_WINDING_NONE; + rs_state.bypass_vs_clip_and_viewport = 1; + rs_state.gl_rasterization_rules = 1; + ctx->rs_state = pipe->create_rasterizer_state(pipe, &rs_state); + + /* fragment shaders are created on-demand */ + + /* vertex shaders */ + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_COLOR }; + const uint semantic_indices[] = { 0, 0 }; + ctx->vs_col = + util_make_vertex_passthrough_shader(pipe, 2, semantic_names, + semantic_indices); + } + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indices[] = { 0, 0 }; + ctx->vs_tex = + util_make_vertex_passthrough_shader(pipe, 2, semantic_names, + semantic_indices); + } + + /* set invariant vertex coordinates */ + for (i = 0; i < 4; i++) + ctx->vertices[i][0][3] = 1; /*v.w*/ + + /* create the vertex buffer */ + ctx->vbuf = pipe_buffer_create(ctx->pipe->screen, + 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(ctx->vertices)); + + return &ctx->blitter; +} + +void util_blitter_destroy(struct blitter_context *blitter) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + int i; + + pipe->delete_blend_state(pipe, ctx->blend_write_color); + pipe->delete_blend_state(pipe, ctx->blend_keep_color); + pipe->delete_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil); + pipe->delete_depth_stencil_alpha_state(pipe, + ctx->dsa_write_depth_keep_stencil); + + for (i = 0; i < 0xff; i++) + if (ctx->dsa_write_depth_stencil[i]) + pipe->delete_depth_stencil_alpha_state(pipe, + ctx->dsa_write_depth_stencil[i]); + + pipe->delete_rasterizer_state(pipe, ctx->rs_state); + pipe->delete_vs_state(pipe, ctx->vs_col); + pipe->delete_vs_state(pipe, ctx->vs_tex); + + for (i = 0; i < PIPE_MAX_TEXTURE_TYPES; i++) { + if (ctx->fs_texfetch_col[i]) + pipe->delete_fs_state(pipe, ctx->fs_texfetch_col[i]); + if (ctx->fs_texfetch_depth[i]) + pipe->delete_fs_state(pipe, ctx->fs_texfetch_depth[i]); + } + + for (i = 0; i < PIPE_MAX_COLOR_BUFS && ctx->fs_col[i]; i++) + if (ctx->fs_col[i]) + pipe->delete_fs_state(pipe, ctx->fs_col[i]); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (ctx->sampler_state[i]) + pipe->delete_sampler_state(pipe, ctx->sampler_state[i]); + + pipe_buffer_reference(&ctx->vbuf, NULL); + FREE(ctx); +} + +static void blitter_check_saved_CSOs(struct blitter_context_priv *ctx) +{ + /* make sure these CSOs have been saved */ + assert(ctx->blitter.saved_blend_state && + ctx->blitter.saved_dsa_state && + ctx->blitter.saved_rs_state && + ctx->blitter.saved_fs && + ctx->blitter.saved_vs); +} + +static void blitter_restore_CSOs(struct blitter_context_priv *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + + /* restore the state objects which are always required to be saved */ + pipe->bind_blend_state(pipe, ctx->blitter.saved_blend_state); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->blitter.saved_dsa_state); + pipe->bind_rasterizer_state(pipe, ctx->blitter.saved_rs_state); + pipe->bind_fs_state(pipe, ctx->blitter.saved_fs); + pipe->bind_vs_state(pipe, ctx->blitter.saved_vs); + + ctx->blitter.saved_blend_state = 0; + ctx->blitter.saved_dsa_state = 0; + ctx->blitter.saved_rs_state = 0; + ctx->blitter.saved_fs = 0; + ctx->blitter.saved_vs = 0; + + /* restore the state objects which are required to be saved before copy/fill + */ + if (ctx->blitter.saved_fb_state.nr_cbufs != ~0) { + pipe->set_framebuffer_state(pipe, &ctx->blitter.saved_fb_state); + ctx->blitter.saved_fb_state.nr_cbufs = ~0; + } + + if (ctx->blitter.saved_num_sampler_states != ~0) { + pipe->bind_fragment_sampler_states(pipe, + ctx->blitter.saved_num_sampler_states, + ctx->blitter.saved_sampler_states); + ctx->blitter.saved_num_sampler_states = ~0; + } + + if (ctx->blitter.saved_num_textures != ~0) { + pipe->set_fragment_sampler_textures(pipe, + ctx->blitter.saved_num_textures, + ctx->blitter.saved_textures); + ctx->blitter.saved_num_textures = ~0; + } +} + +static void blitter_set_rectangle(struct blitter_context_priv *ctx, + unsigned x1, unsigned y1, + unsigned x2, unsigned y2, + float depth) +{ + int i; + + /* set vertex positions */ + ctx->vertices[0][0][0] = x1; /*v0.x*/ + ctx->vertices[0][0][1] = y1; /*v0.y*/ + + ctx->vertices[1][0][0] = x2; /*v1.x*/ + ctx->vertices[1][0][1] = y1; /*v1.y*/ + + ctx->vertices[2][0][0] = x2; /*v2.x*/ + ctx->vertices[2][0][1] = y2; /*v2.y*/ + + ctx->vertices[3][0][0] = x1; /*v3.x*/ + ctx->vertices[3][0][1] = y2; /*v3.y*/ + + for (i = 0; i < 4; i++) + ctx->vertices[i][0][2] = depth; /*z*/ +} + +static void blitter_set_clear_color(struct blitter_context_priv *ctx, + const float *rgba) +{ + int i; + + for (i = 0; i < 4; i++) { + ctx->vertices[i][1][0] = rgba[0]; + ctx->vertices[i][1][1] = rgba[1]; + ctx->vertices[i][1][2] = rgba[2]; + ctx->vertices[i][1][3] = rgba[3]; + } +} + +static void blitter_set_texcoords_2d(struct blitter_context_priv *ctx, + struct pipe_surface *surf, + unsigned x1, unsigned y1, + unsigned x2, unsigned y2) +{ + int i; + float s1 = x1 / (float)surf->width; + float t1 = y1 / (float)surf->height; + float s2 = x2 / (float)surf->width; + float t2 = y2 / (float)surf->height; + + ctx->vertices[0][1][0] = s1; /*t0.s*/ + ctx->vertices[0][1][1] = t1; /*t0.t*/ + + ctx->vertices[1][1][0] = s2; /*t1.s*/ + ctx->vertices[1][1][1] = t1; /*t1.t*/ + + ctx->vertices[2][1][0] = s2; /*t2.s*/ + ctx->vertices[2][1][1] = t2; /*t2.t*/ + + ctx->vertices[3][1][0] = s1; /*t3.s*/ + ctx->vertices[3][1][1] = t2; /*t3.t*/ + + for (i = 0; i < 4; i++) { + ctx->vertices[i][1][2] = 0; /*r*/ + ctx->vertices[i][1][3] = 1; /*q*/ + } +} + +static void blitter_set_texcoords_3d(struct blitter_context_priv *ctx, + struct pipe_surface *surf, + unsigned x1, unsigned y1, + unsigned x2, unsigned y2) +{ + int i; + float depth = u_minify(surf->texture->depth0, surf->level); + float r = surf->zslice / depth; + + blitter_set_texcoords_2d(ctx, surf, x1, y1, x2, y2); + + for (i = 0; i < 4; i++) + ctx->vertices[i][1][2] = r; /*r*/ +} + +static void blitter_set_texcoords_cube(struct blitter_context_priv *ctx, + struct pipe_surface *surf, + unsigned x1, unsigned y1, + unsigned x2, unsigned y2) +{ + int i; + float s1 = x1 / (float)surf->width; + float t1 = y1 / (float)surf->height; + float s2 = x2 / (float)surf->width; + float t2 = y2 / (float)surf->height; + const float st[4][2] = { + {s1, t1}, {s2, t1}, {s2, t2}, {s1, t2} + }; + + util_map_texcoords2d_onto_cubemap(surf->face, + /* pointer, stride in floats */ + &st[0][0], 2, + &ctx->vertices[0][1][0], 8); + + for (i = 0; i < 4; i++) + ctx->vertices[i][1][3] = 1; /*q*/ +} + +static void blitter_draw_quad(struct blitter_context_priv *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + + /* write vertices and draw them */ + pipe_buffer_write(pipe->screen, ctx->vbuf, + 0, sizeof(ctx->vertices), ctx->vertices); + + util_draw_vertex_buffer(pipe, ctx->vbuf, 0, PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ +} + +static INLINE +void *blitter_get_state_write_depth_stencil( + struct blitter_context_priv *ctx, + unsigned stencil) +{ + struct pipe_context *pipe = ctx->pipe; + + stencil &= 0xff; + + /* Create the DSA state on-demand. */ + if (!ctx->dsa_write_depth_stencil[stencil]) { + ctx->template_dsa.stencil[0].ref_value = stencil; + + ctx->dsa_write_depth_stencil[stencil] = + pipe->create_depth_stencil_alpha_state(pipe, &ctx->template_dsa); + } + + return ctx->dsa_write_depth_stencil[stencil]; +} + +static INLINE +void **blitter_get_sampler_state(struct blitter_context_priv *ctx, + int miplevel) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_sampler_state *sampler_state = &ctx->template_sampler_state; + + assert(miplevel < PIPE_MAX_TEXTURE_LEVELS); + + /* Create the sampler state on-demand. */ + if (!ctx->sampler_state[miplevel]) { + sampler_state->lod_bias = miplevel; + sampler_state->min_lod = miplevel; + sampler_state->max_lod = miplevel; + + ctx->sampler_state[miplevel] = pipe->create_sampler_state(pipe, + sampler_state); + } + + /* Return void** so that it can be passed to bind_fragment_sampler_states + * directly. */ + return &ctx->sampler_state[miplevel]; +} + +static INLINE +void *blitter_get_fs_col(struct blitter_context_priv *ctx, unsigned num_cbufs) +{ + struct pipe_context *pipe = ctx->pipe; + unsigned index = num_cbufs ? num_cbufs - 1 : 0; + + assert(num_cbufs <= PIPE_MAX_COLOR_BUFS); + + if (!ctx->fs_col[index]) + ctx->fs_col[index] = + util_make_fragment_clonecolor_shader(pipe, num_cbufs); + + return ctx->fs_col[index]; +} + +static INLINE +void *blitter_get_fs_texfetch_col(struct blitter_context_priv *ctx, + unsigned tex_target) +{ + struct pipe_context *pipe = ctx->pipe; + + assert(tex_target < PIPE_MAX_TEXTURE_TYPES); + + /* Create the fragment shader on-demand. */ + if (!ctx->fs_texfetch_col[tex_target]) { + switch (tex_target) { + case PIPE_TEXTURE_1D: + ctx->fs_texfetch_col[PIPE_TEXTURE_1D] = + util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_1D); + break; + case PIPE_TEXTURE_2D: + ctx->fs_texfetch_col[PIPE_TEXTURE_2D] = + util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D); + break; + case PIPE_TEXTURE_3D: + ctx->fs_texfetch_col[PIPE_TEXTURE_3D] = + util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_3D); + break; + case PIPE_TEXTURE_CUBE: + ctx->fs_texfetch_col[PIPE_TEXTURE_CUBE] = + util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_CUBE); + break; + default:; + } + } + + return ctx->fs_texfetch_col[tex_target]; +} + +static INLINE +void *blitter_get_fs_texfetch_depth(struct blitter_context_priv *ctx, + unsigned tex_target) +{ + struct pipe_context *pipe = ctx->pipe; + + assert(tex_target < PIPE_MAX_TEXTURE_TYPES); + + /* Create the fragment shader on-demand. */ + if (!ctx->fs_texfetch_depth[tex_target]) { + switch (tex_target) { + case PIPE_TEXTURE_1D: + ctx->fs_texfetch_depth[PIPE_TEXTURE_1D] = + util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_1D); + break; + case PIPE_TEXTURE_2D: + ctx->fs_texfetch_depth[PIPE_TEXTURE_2D] = + util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_2D); + break; + case PIPE_TEXTURE_3D: + ctx->fs_texfetch_depth[PIPE_TEXTURE_3D] = + util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_3D); + break; + case PIPE_TEXTURE_CUBE: + ctx->fs_texfetch_depth[PIPE_TEXTURE_CUBE] = + util_make_fragment_tex_shader_writedepth(pipe,TGSI_TEXTURE_CUBE); + break; + default:; + } + } + + return ctx->fs_texfetch_depth[tex_target]; +} + +void util_blitter_clear(struct blitter_context *blitter, + unsigned width, unsigned height, + unsigned num_cbufs, + unsigned clear_buffers, + const float *rgba, + double depth, unsigned stencil) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + + assert(num_cbufs <= PIPE_MAX_COLOR_BUFS); + + blitter_check_saved_CSOs(ctx); + + /* bind CSOs */ + if (clear_buffers & PIPE_CLEAR_COLOR) + pipe->bind_blend_state(pipe, ctx->blend_write_color); + else + pipe->bind_blend_state(pipe, ctx->blend_keep_color); + + if (clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) + pipe->bind_depth_stencil_alpha_state(pipe, + blitter_get_state_write_depth_stencil(ctx, stencil)); + else + pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil); + + pipe->bind_rasterizer_state(pipe, ctx->rs_state); + pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, num_cbufs)); + pipe->bind_vs_state(pipe, ctx->vs_col); + + blitter_set_clear_color(ctx, rgba); + blitter_set_rectangle(ctx, 0, 0, width, height, depth); + blitter_draw_quad(ctx); + blitter_restore_CSOs(ctx); +} + +void util_blitter_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean ignore_stencil) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_framebuffer_state fb_state; + boolean is_stencil, is_depth; + unsigned dst_tex_usage; + + /* give up if textures are not set */ + assert(dst->texture && src->texture); + if (!dst->texture || !src->texture) + return; + + is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; + is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; + dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : + PIPE_TEXTURE_USAGE_RENDER_TARGET; + + /* check if we can sample from and render to the surfaces */ + /* (assuming copying a stencil buffer is not possible) */ + if ((!ignore_stencil && is_stencil) || + !screen->is_format_supported(screen, dst->format, dst->texture->target, + dst_tex_usage, 0) || + !screen->is_format_supported(screen, src->format, src->texture->target, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { + util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy, + width, height); + return; + } + + /* check whether the states are properly saved */ + blitter_check_saved_CSOs(ctx); + assert(blitter->saved_fb_state.nr_cbufs != ~0); + assert(blitter->saved_num_textures != ~0); + assert(blitter->saved_num_sampler_states != ~0); + assert(src->texture->target < PIPE_MAX_TEXTURE_TYPES); + + /* bind CSOs */ + fb_state.width = dst->width; + fb_state.height = dst->height; + + if (is_depth) { + pipe->bind_blend_state(pipe, ctx->blend_keep_color); + pipe->bind_depth_stencil_alpha_state(pipe, + ctx->dsa_write_depth_keep_stencil); + pipe->bind_fs_state(pipe, + blitter_get_fs_texfetch_depth(ctx, src->texture->target)); + + fb_state.nr_cbufs = 0; + fb_state.zsbuf = dst; + } else { + pipe->bind_blend_state(pipe, ctx->blend_write_color); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil); + pipe->bind_fs_state(pipe, + blitter_get_fs_texfetch_col(ctx, src->texture->target)); + + fb_state.nr_cbufs = 1; + fb_state.cbufs[0] = dst; + fb_state.zsbuf = 0; + } + + pipe->bind_rasterizer_state(pipe, ctx->rs_state); + pipe->bind_vs_state(pipe, ctx->vs_tex); + pipe->bind_fragment_sampler_states(pipe, 1, + blitter_get_sampler_state(ctx, src->level)); + pipe->set_fragment_sampler_textures(pipe, 1, &src->texture); + pipe->set_framebuffer_state(pipe, &fb_state); + + /* set texture coordinates */ + switch (src->texture->target) { + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: + blitter_set_texcoords_2d(ctx, src, srcx, srcy, + srcx+width, srcy+height); + break; + case PIPE_TEXTURE_3D: + blitter_set_texcoords_3d(ctx, src, srcx, srcy, + srcx+width, srcy+height); + break; + case PIPE_TEXTURE_CUBE: + blitter_set_texcoords_cube(ctx, src, srcx, srcy, + srcx+width, srcy+height); + break; + default: + assert(0); + } + + blitter_set_rectangle(ctx, dstx, dsty, dstx+width, dsty+height, 0); + blitter_draw_quad(ctx); + blitter_restore_CSOs(ctx); +} + +void util_blitter_fill(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_framebuffer_state fb_state; + float rgba[4]; + ubyte ub_rgba[4] = {0}; + union util_color color; + int i; + + assert(dst->texture); + if (!dst->texture) + return; + + /* check if we can render to the surface */ + if (util_format_is_depth_or_stencil(dst->format) || /* unlikely, but you never know */ + !screen->is_format_supported(screen, dst->format, dst->texture->target, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { + util_surface_fill(pipe, dst, dstx, dsty, width, height, value); + return; + } + + /* unpack the color */ + color.ui = value; + util_unpack_color_ub(dst->format, &color, + ub_rgba, ub_rgba+1, ub_rgba+2, ub_rgba+3); + for (i = 0; i < 4; i++) + rgba[i] = ubyte_to_float(ub_rgba[i]); + + /* check the saved state */ + blitter_check_saved_CSOs(ctx); + assert(blitter->saved_fb_state.nr_cbufs != ~0); + + /* bind CSOs */ + pipe->bind_blend_state(pipe, ctx->blend_write_color); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil); + pipe->bind_rasterizer_state(pipe, ctx->rs_state); + pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, 1)); + pipe->bind_vs_state(pipe, ctx->vs_col); + + /* set a framebuffer state */ + fb_state.width = dst->width; + fb_state.height = dst->height; + fb_state.nr_cbufs = 1; + fb_state.cbufs[0] = dst; + fb_state.zsbuf = 0; + pipe->set_framebuffer_state(pipe, &fb_state); + + blitter_set_clear_color(ctx, rgba); + blitter_set_rectangle(ctx, 0, 0, width, height, 0); + blitter_draw_quad(ctx); + blitter_restore_CSOs(ctx); +} diff --git a/src/gallium/auxiliary/util/u_blitter.h b/src/gallium/auxiliary/util/u_blitter.h new file mode 100644 index 00000000000..3da5a6ca525 --- /dev/null +++ b/src/gallium/auxiliary/util/u_blitter.h @@ -0,0 +1,230 @@ +/************************************************************************** + * + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_BLITTER_H +#define U_BLITTER_H + +#include "util/u_memory.h" + +#include "pipe/p_state.h" + + +#ifdef __cplusplus +extern "C" { +#endif + +struct pipe_context; + +struct blitter_context +{ + /* Private members, really. */ + void *saved_blend_state; /**< blend state */ + void *saved_dsa_state; /**< depth stencil alpha state */ + void *saved_rs_state; /**< rasterizer state */ + void *saved_fs, *saved_vs; /**< fragment shader, vertex shader */ + + struct pipe_framebuffer_state saved_fb_state; /**< framebuffer state */ + + int saved_num_sampler_states; + void *saved_sampler_states[32]; + + int saved_num_textures; + struct pipe_texture *saved_textures[32]; /* is 32 enough? */ +}; + +/** + * Create a blitter context. + */ +struct blitter_context *util_blitter_create(struct pipe_context *pipe); + +/** + * Destroy a blitter context. + */ +void util_blitter_destroy(struct blitter_context *blitter); + +/* + * These CSOs must be saved before any of the following functions is called: + * - blend state + * - depth stencil alpha state + * - rasterizer state + * - vertex shader + * - fragment shader + */ + +/** + * Clear a specified set of currently bound buffers to specified values. + */ +void util_blitter_clear(struct blitter_context *blitter, + unsigned width, unsigned height, + unsigned num_cbufs, + unsigned clear_buffers, + const float *rgba, + double depth, unsigned stencil); + +/** + * Copy a block of pixels from one surface to another. + * + * You can copy from any color format to any other color format provided + * the former can be sampled and the latter can be rendered to. Otherwise, + * a software fallback path is taken and both surfaces must be of the same + * format. + * + * The same holds for depth-stencil formats with the exception that stencil + * cannot be copied unless you set ignore_stencil to FALSE. In that case, + * a software fallback path is taken and both surfaces must be of the same + * format. + * + * Use pipe_screen->is_format_supported to know your options. + * + * These states must be saved in the blitter in addition to the state objects + * already required to be saved: + * - framebuffer state + * - fragment sampler states + * - fragment sampler textures + */ +void util_blitter_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean ignore_stencil); + +/** + * Fill a region of a surface with a constant value. + * + * If the surface cannot be rendered to or it's a depth-stencil format, + * a software fallback path is taken. + * + * These states must be saved in the blitter in addition to the state objects + * already required to be saved: + * - framebuffer state + */ +void util_blitter_fill(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value); + +/** + * Copy all pixels from one surface to another. + * + * The rules are the same as in util_blitter_copy with the addition that + * surfaces must have the same size. + */ +static INLINE +void util_blitter_copy_surface(struct blitter_context *blitter, + struct pipe_surface *dst, + struct pipe_surface *src, + boolean ignore_stencil) +{ + assert(dst->width == src->width && dst->height == src->height); + + util_blitter_copy(blitter, dst, 0, 0, src, 0, 0, src->width, src->height, + ignore_stencil); +} + + +/* The functions below should be used to save currently bound constant state + * objects inside a driver. The objects are automatically restored at the end + * of the util_blitter_{clear, fill, copy, copy_surface} functions and then + * forgotten. + * + * CSOs not listed here are not affected by util_blitter. */ + +static INLINE +void util_blitter_save_blend(struct blitter_context *blitter, + void *state) +{ + blitter->saved_blend_state = state; +} + +static INLINE +void util_blitter_save_depth_stencil_alpha(struct blitter_context *blitter, + void *state) +{ + blitter->saved_dsa_state = state; +} + +static INLINE +void util_blitter_save_rasterizer(struct blitter_context *blitter, + void *state) +{ + blitter->saved_rs_state = state; +} + +static INLINE +void util_blitter_save_fragment_shader(struct blitter_context *blitter, + void *fs) +{ + blitter->saved_fs = fs; +} + +static INLINE +void util_blitter_save_vertex_shader(struct blitter_context *blitter, + void *vs) +{ + blitter->saved_vs = vs; +} + +static INLINE +void util_blitter_save_framebuffer(struct blitter_context *blitter, + struct pipe_framebuffer_state *state) +{ + blitter->saved_fb_state = *state; +} + +static INLINE +void util_blitter_save_fragment_sampler_states( + struct blitter_context *blitter, + int num_sampler_states, + void **sampler_states) +{ + assert(num_sampler_states <= Elements(blitter->saved_sampler_states)); + + blitter->saved_num_sampler_states = num_sampler_states; + memcpy(blitter->saved_sampler_states, sampler_states, + num_sampler_states * sizeof(void *)); +} + +static INLINE +void util_blitter_save_fragment_sampler_textures( + struct blitter_context *blitter, + int num_textures, + struct pipe_texture **textures) +{ + assert(num_textures <= Elements(blitter->saved_textures)); + + blitter->saved_num_textures = num_textures; + memcpy(blitter->saved_textures, textures, + num_textures * sizeof(struct pipe_texture *)); +} + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c index 40633574b08..4e01123fff1 100644 --- a/src/gallium/auxiliary/util/u_debug.c +++ b/src/gallium/auxiliary/util/u_debug.c @@ -64,11 +64,13 @@ #include "pipe/p_format.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_string.h" #include "util/u_stream.h" #include "util/u_math.h" #include "util/u_tile.h" +#include "util/u_prim.h" #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY @@ -600,6 +602,32 @@ const char *pf_name( enum pipe_format format ) } + +static const struct debug_named_value pipe_prim_names[] = { +#ifdef DEBUG + DEBUG_NAMED_VALUE(PIPE_PRIM_POINTS), + DEBUG_NAMED_VALUE(PIPE_PRIM_LINES), + DEBUG_NAMED_VALUE(PIPE_PRIM_LINE_LOOP), + DEBUG_NAMED_VALUE(PIPE_PRIM_LINE_STRIP), + DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLES), + DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLE_STRIP), + DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLE_FAN), + DEBUG_NAMED_VALUE(PIPE_PRIM_QUADS), + DEBUG_NAMED_VALUE(PIPE_PRIM_QUAD_STRIP), + DEBUG_NAMED_VALUE(PIPE_PRIM_POLYGON), +#endif + DEBUG_NAMED_VALUE_END +}; + + +const char *u_prim_name( unsigned prim ) +{ + return debug_dump_enum(pipe_prim_names, prim); +} + + + + #ifdef DEBUG void debug_dump_image(const char *prefix, unsigned format, unsigned cpp, @@ -670,9 +698,9 @@ void debug_dump_surface(const char *prefix, debug_dump_image(prefix, texture->format, - pf_get_blocksize(texture->format), - pf_get_nblocksx(texture->format, transfer->width), - pf_get_nblocksy(texture->format, transfer->height), + util_format_get_blocksize(texture->format), + util_format_get_nblocksx(texture->format, transfer->width), + util_format_get_nblocksy(texture->format, transfer->height), transfer->stride, data); diff --git a/src/gallium/auxiliary/util/u_dl.c b/src/gallium/auxiliary/util/u_dl.c new file mode 100644 index 00000000000..b42b429d4d7 --- /dev/null +++ b/src/gallium/auxiliary/util/u_dl.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * Copyright 1999-2008 Brian Paul + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + + +#include "pipe/p_config.h" + +#if defined(PIPE_OS_UNIX) +#include <dlfcn.h> +#endif +#if defined(PIPE_OS_WINDOWS) +#include <windows.h> +#endif + +#include "u_dl.h" + + +struct util_dl_library * +util_dl_open(const char *filename) +{ +#if defined(PIPE_OS_UNIX) + return (struct util_dl_library *)dlopen(filename, RTLD_LAZY | RTLD_GLOBAL); +#elif defined(PIPE_OS_WINDOWS) + return (struct util_dl_library *)LoadLibraryA(filename); +#else + return NULL; +#endif +} + + +util_dl_proc +util_dl_get_proc_address(struct util_dl_library *library, + const char *procname) +{ +#if defined(PIPE_OS_UNIX) + return (util_dl_proc)dlsym((void *)library, procname); +#elif defined(PIPE_OS_WINDOWS) + return (util_dl_proc)GetProcAddress((HMODULE)library, procname); +#else + return (util_dl_proc)NULL; +#endif +} + + +void +util_dl_close(struct util_dl_library *library) +{ +#if defined(PIPE_OS_UNIX) + dlclose((void *)library); +#elif defined(PIPE_OS_WINDOWS) + FreeLibrary((HMODULE)library); +#else + (void)library; +#endif +} diff --git a/src/gallium/auxiliary/util/u_dl.h b/src/gallium/auxiliary/util/u_dl.h new file mode 100644 index 00000000000..018b38543b0 --- /dev/null +++ b/src/gallium/auxiliary/util/u_dl.h @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + + +#ifndef U_DL_H_ +#define U_DL_H_ + + +struct util_dl_library; + + +typedef void (*util_dl_proc)(void); + + +/** + * Open a library dynamically. + */ +struct util_dl_library * +util_dl_open(const char *filename); + + +/** + * Lookup a function in a library. + */ +util_dl_proc +util_dl_get_proc_address(struct util_dl_library *library, + const char *procname); + + +/** + * Close a library. + */ +void +util_dl_close(struct util_dl_library *library); + + +#endif /* U_DL_H_ */ diff --git a/src/gallium/auxiliary/util/u_format.c b/src/gallium/auxiliary/util/u_format.c index 98ea13b60b5..e0724a1a8be 100644 --- a/src/gallium/auxiliary/util/u_format.c +++ b/src/gallium/auxiliary/util/u_format.c @@ -32,15 +32,14 @@ const struct util_format_description * util_format_description(enum pipe_format format) { - const struct util_format_description *desc = util_format_description_table; + const struct util_format_description *desc; - while(TRUE) { - if(desc->format == format) - return desc; + if (format >= PIPE_FORMAT_COUNT) { + return NULL; + } - if(desc->format == PIPE_FORMAT_NONE) - return NULL; + desc = &util_format_description_table[format]; + assert(desc->format == format); - ++desc; - }; + return desc; } diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv index b9cc2aa716e..866b18ff160 100644 --- a/src/gallium/auxiliary/util/u_format.csv +++ b/src/gallium/auxiliary/util/u_format.csv @@ -11,6 +11,8 @@ PIPE_FORMAT_A8_UNORM , arith , 1, 1, un8 , , , , 000x, PIPE_FORMAT_I8_UNORM , arith , 1, 1, un8 , , , , xxxx, rgb PIPE_FORMAT_A8L8_UNORM , arith , 1, 1, un8 , un8 , , , xxxy, rgb PIPE_FORMAT_L16_UNORM , arith , 1, 1, un16, , , , xxx1, rgb +PIPE_FORMAT_YCBCR , yuv , 2, 1, x32 , , , , xyz1, yuv +PIPE_FORMAT_YCBCR_REV , yuv , 2, 1, x32 , , , , xyz1, yuv PIPE_FORMAT_Z16_UNORM , array , 1, 1, un16, , , , x___, zs PIPE_FORMAT_Z32_UNORM , array , 1, 1, un32, , , , x___, zs PIPE_FORMAT_Z32_FLOAT , array , 1, 1, f32 , , , , x___, zs @@ -97,13 +99,11 @@ PIPE_FORMAT_B8G8R8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyxw, PIPE_FORMAT_B8G8R8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyx1, srgb PIPE_FORMAT_X8UB8UG8SR8S_NORM , arith , 1, 1, sn8 , sn8 , un8 , x8 , 1zyx, rgb PIPE_FORMAT_B6UG5SR5S_NORM , arith , 1, 1, sn5 , sn5 , un6 , , xyz1, rgb -PIPE_FORMAT_YCBCR , yuv , 2, 1, x32 , , , , xyz1, yuv -PIPE_FORMAT_YCBCR_REV , yuv , 2, 1, x32 , , , , xyz1, yuv -PIPE_FORMAT_DXT1_RGBA , dxt , 4, 4, x64 , , , , xyzw, rgb PIPE_FORMAT_DXT1_RGB , dxt , 4, 4, x64 , , , , xyz1, rgb +PIPE_FORMAT_DXT1_RGBA , dxt , 4, 4, x64 , , , , xyzw, rgb PIPE_FORMAT_DXT3_RGBA , dxt , 4, 4, x128, , , , xyzw, rgb PIPE_FORMAT_DXT5_RGBA , dxt , 4, 4, x128, , , , xyzw, rgb -PIPE_FORMAT_DXT1_SRGBA , dxt , 4, 4, x64 , , , , xyzw, srgb PIPE_FORMAT_DXT1_SRGB , dxt , 4, 4, x64 , , , , xyz1, srgb +PIPE_FORMAT_DXT1_SRGBA , dxt , 4, 4, x64 , , , , xyzw, srgb PIPE_FORMAT_DXT3_SRGBA , dxt , 4, 4, x128, , , , xyzw, srgb PIPE_FORMAT_DXT5_SRGBA , dxt , 4, 4, x128, , , , xyzw, srgb diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h index 19b902db983..090183fb174 100644 --- a/src/gallium/auxiliary/util/u_format.h +++ b/src/gallium/auxiliary/util/u_format.h @@ -32,6 +32,10 @@ #include "pipe/p_format.h" +#ifdef __cplusplus +extern "C" { +#endif + /** * Describe how to best pack/unpack pixels into/from the prescribed format. @@ -147,6 +151,250 @@ const struct util_format_description * util_format_description(enum pipe_format format); +/* + * Format query functions. + */ + +static INLINE boolean +util_format_is_compressed(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return FALSE; + } + + return desc->layout == UTIL_FORMAT_LAYOUT_DXT ? TRUE : FALSE; +} + +static INLINE boolean +util_format_is_depth_or_stencil(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return FALSE; + } + + return desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS ? TRUE : FALSE; +} + +static INLINE boolean +util_format_is_depth_and_stencil(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return FALSE; + } + + if (desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS) { + return FALSE; + } + + return (desc->swizzle[0] != UTIL_FORMAT_SWIZZLE_NONE && + desc->swizzle[1] != UTIL_FORMAT_SWIZZLE_NONE) ? TRUE : FALSE; +} + + +/** + * Return total bits needed for the pixel format per block. + */ +static INLINE uint +util_format_get_blocksizebits(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 0; + } + + return desc->block.bits; +} + +/** + * Return bytes per block (not pixel) for the given format. + */ +static INLINE uint +util_format_get_blocksize(enum pipe_format format) +{ + uint bits = util_format_get_blocksizebits(format); + + assert(bits % 8 == 0); + + return bits / 8; +} + +static INLINE uint +util_format_get_blockwidth(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 1; + } + + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_YUV: + return 2; + case UTIL_FORMAT_LAYOUT_DXT: + return 4; + default: + return 1; + } +} + +static INLINE uint +util_format_get_blockheight(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 1; + } + + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_DXT: + return 4; + default: + return 1; + } +} + +static INLINE unsigned +util_format_get_nblocksx(enum pipe_format format, + unsigned x) +{ + unsigned blockwidth = util_format_get_blockwidth(format); + return (x + blockwidth - 1) / blockwidth; +} + +static INLINE unsigned +util_format_get_nblocksy(enum pipe_format format, + unsigned y) +{ + unsigned blockheight = util_format_get_blockheight(format); + return (y + blockheight - 1) / blockheight; +} + +static INLINE unsigned +util_format_get_nblocks(enum pipe_format format, + unsigned width, + unsigned height) +{ + return util_format_get_nblocksx(format, width) * util_format_get_nblocksy(format, height); +} + +static INLINE size_t +util_format_get_stride(enum pipe_format format, + unsigned width) +{ + return util_format_get_nblocksx(format, width) * util_format_get_blocksize(format); +} + +static INLINE size_t +util_format_get_2d_size(enum pipe_format format, + size_t stride, + unsigned height) +{ + return util_format_get_nblocksy(format, height) * stride; +} + +static INLINE uint +util_format_get_component_bits(enum pipe_format format, + enum util_format_colorspace colorspace, + uint component) +{ + const struct util_format_description *desc = util_format_description(format); + enum util_format_colorspace desc_colorspace; + + assert(format); + if (!format) { + return 0; + } + + assert(component < 4); + + /* Treat RGB and SRGB as equivalent. */ + if (colorspace == UTIL_FORMAT_COLORSPACE_SRGB) { + colorspace = UTIL_FORMAT_COLORSPACE_RGB; + } + if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) { + desc_colorspace = UTIL_FORMAT_COLORSPACE_RGB; + } else { + desc_colorspace = desc->colorspace; + } + + if (desc_colorspace != colorspace) { + return 0; + } + + switch (desc->swizzle[component]) { + case UTIL_FORMAT_SWIZZLE_X: + return desc->channel[0].size; + case UTIL_FORMAT_SWIZZLE_Y: + return desc->channel[1].size; + case UTIL_FORMAT_SWIZZLE_Z: + return desc->channel[2].size; + case UTIL_FORMAT_SWIZZLE_W: + return desc->channel[3].size; + default: + return 0; + } +} + +static INLINE boolean +util_format_has_alpha(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return FALSE; + } + + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_SCALAR: + case UTIL_FORMAT_LAYOUT_ARITH: + case UTIL_FORMAT_LAYOUT_ARRAY: + /* FIXME: pf_get_component_bits( PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_COMP_A ) should not return 0 right? */ + if (format == PIPE_FORMAT_A8_UNORM || + format == PIPE_FORMAT_A8L8_UNORM || + format == PIPE_FORMAT_A8L8_SRGB) { + return TRUE; + } + return util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3) != 0; + case UTIL_FORMAT_LAYOUT_YUV: + return FALSE; + case UTIL_FORMAT_LAYOUT_DXT: + switch (format) { + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + case PIPE_FORMAT_DXT1_SRGBA: + case PIPE_FORMAT_DXT3_SRGBA: + case PIPE_FORMAT_DXT5_SRGBA: + return TRUE; + default: + return FALSE; + } + default: + assert(0); + return FALSE; + } +} + + +/* + * Format access functions. + */ + void util_format_read_4f(enum pipe_format format, float *dst, unsigned dst_stride, @@ -171,4 +419,8 @@ util_format_write_4ub(enum pipe_format format, void *dst, unsigned dst_stride, unsigned x, unsigned y, unsigned w, unsigned h); +#ifdef __cplusplus +} // extern "C" { +#endif + #endif /* ! U_FORMAT_H */ diff --git a/src/gallium/auxiliary/util/u_format_access.py b/src/gallium/auxiliary/util/u_format_access.py index eeb1a9657fd..0b05ddb9312 100644 --- a/src/gallium/auxiliary/util/u_format_access.py +++ b/src/gallium/auxiliary/util/u_format_access.py @@ -325,14 +325,14 @@ def generate_format_read(format, dst_type, dst_native_type, dst_suffix): elif swizzle == SWIZZLE_0: value = '0' elif swizzle == SWIZZLE_1: - value = '1' + value = get_one(dst_type) else: assert False elif format.colorspace == 'zs': if i < 3: value = 'z' else: - value = '1' + value = get_one(dst_type) else: assert False print ' *dst_pixel++ = %s; /* %s */' % (value, 'rgba'[i]) diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py index 2cd0f956786..571cab55dc8 100755 --- a/src/gallium/auxiliary/util/u_format_table.py +++ b/src/gallium/auxiliary/util/u_format_table.py @@ -89,6 +89,15 @@ def write_format_table(formats): print 'const struct util_format_description' print 'util_format_description_table[] = ' print "{" + print " {" + print " PIPE_FORMAT_NONE," + print " \"PIPE_FORMAT_NONE\"," + print " {0, 0, 0}," + print " 0," + print " {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}}," + print " {0, 0, 0, 0}," + print " 0" + print " }," for format in formats: print " {" print " %s," % (format.name,) @@ -119,15 +128,6 @@ def write_format_table(formats): print " }," print " %s," % (colorspace_map(format.colorspace),) print " }," - print " {" - print " PIPE_FORMAT_NONE," - print " \"PIPE_FORMAT_NONE\"," - print " {0, 0, 0}," - print " 0," - print " {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}}," - print " {0, 0, 0, 0}," - print " 0" - print " }," print "};" diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 83263d9fe64..76023794dcd 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -41,11 +41,13 @@ #include "pipe/p_shader_tokens.h" #include "pipe/p_state.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_draw_quad.h" #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" #include "util/u_math.h" +#include "util/u_texture.h" #include "cso_cache/cso_context.h" @@ -61,7 +63,7 @@ struct gen_mipmap_state struct pipe_sampler_state sampler; void *vs; - void *fs; + void *fs2d, *fsCube; struct pipe_buffer *vbuf; /**< quad vertices */ unsigned vbuf_slot; @@ -996,7 +998,7 @@ reduce_2d(enum pipe_format pformat, { enum dtype datatype; uint comps; - const int bpt = pf_get_blocksize(pformat); + const int bpt = util_format_get_blocksize(pformat); const ubyte *srcA, *srcB; ubyte *dst; int row; @@ -1035,7 +1037,7 @@ reduce_3d(enum pipe_format pformat, int dstWidth, int dstHeight, int dstDepth, int dstRowStride, ubyte *dstPtr) { - const int bpt = pf_get_blocksize(pformat); + const int bpt = util_format_get_blocksize(pformat); const int border = 0; int img, row; int bytesPerSrcImage, bytesPerDstImage; @@ -1159,8 +1161,8 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, const uint zslice = 0; uint dstLevel; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; @@ -1204,8 +1206,8 @@ make_3d_mipmap(struct gen_mipmap_state *ctx, struct pipe_screen *screen = pipe->screen; uint dstLevel, zslice = 0; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; @@ -1317,7 +1319,8 @@ util_create_gen_mipmap(struct pipe_context *pipe, } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs2d = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D); + ctx->fsCube = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_CUBE); /* vertex data that doesn't change */ for (i = 0; i < 4; i++) { @@ -1383,59 +1386,9 @@ set_vertex_data(struct gen_mipmap_state *ctx, static const float st[4][2] = { {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f} }; - float rx, ry, rz; - uint i; - - /* loop over quad verts */ - for (i = 0; i < 4; i++) { - /* Compute sc = +/-scale and tc = +/-scale. - * Not +/-1 to avoid cube face selection ambiguity near the edges, - * though that can still sometimes happen with this scale factor... - */ - const float scale = 0.9999f; - const float sc = (2.0f * st[i][0] - 1.0f) * scale; - const float tc = (2.0f * st[i][1] - 1.0f) * scale; - - switch (face) { - case PIPE_TEX_FACE_POS_X: - rx = 1.0f; - ry = -tc; - rz = -sc; - break; - case PIPE_TEX_FACE_NEG_X: - rx = -1.0f; - ry = -tc; - rz = sc; - break; - case PIPE_TEX_FACE_POS_Y: - rx = sc; - ry = 1.0f; - rz = tc; - break; - case PIPE_TEX_FACE_NEG_Y: - rx = sc; - ry = -1.0f; - rz = -tc; - break; - case PIPE_TEX_FACE_POS_Z: - rx = sc; - ry = -tc; - rz = 1.0f; - break; - case PIPE_TEX_FACE_NEG_Z: - rx = -sc; - ry = -tc; - rz = -1.0f; - break; - default: - rx = ry = rz = 0.0f; - assert(0); - } - ctx->vertices[i][1][0] = rx; /*s*/ - ctx->vertices[i][1][1] = ry; /*t*/ - ctx->vertices[i][1][2] = rz; /*r*/ - } + util_map_texcoords2d_onto_cubemap(face, &st[0][0], 2, + &ctx->vertices[0][1][0], 8); } else { /* 1D/2D */ @@ -1475,7 +1428,8 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) struct pipe_context *pipe = ctx->pipe; pipe->delete_vs_state(pipe, ctx->vs); - pipe->delete_fs_state(pipe, ctx->fs); + pipe->delete_fs_state(pipe, ctx->fs2d); + pipe->delete_fs_state(pipe, ctx->fsCube); pipe_buffer_reference(&ctx->vbuf, NULL); @@ -1513,6 +1467,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb; + void *fs = (pt->target == PIPE_TEXTURE_CUBE) ? ctx->fsCube : ctx->fs2d; uint dstLevel; uint zslice = 0; uint offset; @@ -1550,7 +1505,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_fragment_shader_handle(ctx->cso, fs); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* init framebuffer state */ diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index b76592d1ec6..81aeb83cbb5 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -583,6 +583,19 @@ do { \ #endif +static INLINE uint32_t util_unsigned_fixed(float value, unsigned frac_bits) +{ + value *= (1<<frac_bits); + return value < 0 ? 0 : value; +} + +static INLINE int32_t util_signed_fixed(float value, unsigned frac_bits) +{ + return value * (1<<frac_bits); +} + + + #ifdef __cplusplus } #endif diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index a2e0f266864..43eb0153ee7 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -37,6 +37,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" +#include "util/u_format.h" #include "util/u_math.h" @@ -128,11 +129,18 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, } return; - /* XXX lots more cases to add */ + /* Handle other cases with a generic function. + */ default: - uc->ui = 0; /* keep compiler happy */ - debug_print_format("gallium: unhandled format in util_pack_color_ub()", format); - assert(0); + { + ubyte src[4]; + + src[0] = r; + src[1] = g; + src[2] = b; + src[3] = a; + util_format_write_4ub(format, src, 0, uc, 0, 0, 0, 1, 1); + } } } @@ -282,11 +290,18 @@ util_unpack_color_ub(enum pipe_format format, union util_color *uc, } return; - /* XXX lots more cases to add */ + /* Handle other cases with a generic function. + */ default: - debug_print_format("gallium: unhandled format in util_unpack_color_ub()", - format); - assert(0); + { + ubyte dst[4]; + + util_format_read_4ub(format, dst, 0, uc, 0, 0, 0, 1, 1); + *r = dst[0]; + *g = dst[1]; + *b = dst[2]; + *a = dst[3]; + } } } @@ -302,7 +317,7 @@ util_pack_color(const float rgba[4], enum pipe_format format, union util_color * ubyte b = 0; ubyte a = 0; - if (pf_size_x(format) <= 8) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) <= 8) { /* format uses 8-bit components or less */ r = float_to_ubyte(rgba[0]); g = float_to_ubyte(rgba[1]); @@ -382,11 +397,11 @@ util_pack_color(const float rgba[4], enum pipe_format format, union util_color * uc->f[2] = rgba[2]; } return; - /* XXX lots more cases to add */ + + /* Handle other cases with a generic function. + */ default: - uc->ui = 0; /* keep compiler happy */ - debug_print_format("gallium: unhandled format in util_pack_color()", format); - assert(0); + util_format_write_4f(format, rgba, 0, uc, 0, 0, 0, 1, 1); } } diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h index a9b533eea70..74343299623 100644 --- a/src/gallium/auxiliary/util/u_prim.h +++ b/src/gallium/auxiliary/util/u_prim.h @@ -135,4 +135,6 @@ static INLINE unsigned u_reduced_prim( unsigned pipe_prim ) } } +const char *u_prim_name( unsigned pipe_prim ); + #endif diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 72725b59d2c..298fbacecba 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -34,6 +34,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "util/u_rect.h" @@ -57,9 +58,9 @@ util_copy_rect(ubyte * dst, { unsigned i; int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; - int blocksize = pf_get_blocksize(format); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); assert(blocksize > 0); assert(blockwidth > 0); @@ -105,9 +106,9 @@ util_fill_rect(ubyte * dst, { unsigned i, j; unsigned width_size; - int blocksize = pf_get_blocksize(format); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); assert(blocksize > 0); assert(blockwidth > 0); @@ -203,9 +204,9 @@ util_surface_copy(struct pipe_context *pipe, PIPE_TRANSFER_WRITE, dst_x, dst_y, w, h); - assert(pf_get_blocksize(dst_format) == pf_get_blocksize(src_format)); - assert(pf_get_blockwidth(dst_format) == pf_get_blockwidth(src_format)); - assert(pf_get_blockheight(dst_format) == pf_get_blockheight(src_format)); + assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format)); + assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format)); + assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format)); src_map = pipe->screen->transfer_map(screen, src_trans); dst_map = pipe->screen->transfer_map(screen, dst_trans); @@ -270,7 +271,7 @@ util_surface_fill(struct pipe_context *pipe, if (dst_map) { assert(dst_trans->stride > 0); - switch (pf_get_blocksize(dst_trans->texture->format)) { + switch (util_format_get_blocksize(dst_trans->texture->format)) { case 1: case 2: case 4: diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 1c8b157d91f..8172ead0201 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -2,6 +2,7 @@ * * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -30,6 +31,7 @@ * Simple vertex/fragment shader generators. * * @author Brian Paul + Marek Olšák */ @@ -87,6 +89,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, */ void * util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, + unsigned tex_target, unsigned writemask ) { struct ureg_program *ureg; @@ -116,20 +119,63 @@ util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, ureg_TEX( ureg, ureg_writemask(out, writemask), - TGSI_TEXTURE_2D, tex, sampler ); + tex_target, tex, sampler ); ureg_END( ureg ); return ureg_create_shader_and_destroy( ureg, pipe ); } void * -util_make_fragment_tex_shader(struct pipe_context *pipe ) +util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target ) { return util_make_fragment_tex_shader_writemask( pipe, + tex_target, TGSI_WRITEMASK_XYZW ); } +/** + * Make a simple fragment texture shader which reads an X component from + * a texture and writes it as depth. + */ +void * +util_make_fragment_tex_shader_writedepth(struct pipe_context *pipe, + unsigned tex_target) +{ + struct ureg_program *ureg; + struct ureg_src sampler; + struct ureg_src tex; + struct ureg_dst out, depth; + struct ureg_src imm; + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; + + sampler = ureg_DECL_sampler( ureg, 0 ); + + tex = ureg_DECL_fs_input( ureg, + TGSI_SEMANTIC_GENERIC, 0, + TGSI_INTERPOLATE_PERSPECTIVE ); + + out = ureg_DECL_output( ureg, + TGSI_SEMANTIC_COLOR, + 0 ); + + depth = ureg_DECL_output( ureg, + TGSI_SEMANTIC_POSITION, + 0 ); + + imm = ureg_imm4f( ureg, 0, 0, 0, 1 ); + + ureg_MOV( ureg, out, imm ); + + ureg_TEX( ureg, + ureg_writemask(depth, TGSI_WRITEMASK_Z), + tex_target, tex, sampler ); + ureg_END( ureg ); + + return ureg_create_shader_and_destroy( ureg, pipe ); +} /** * Make simple fragment color pass-through shader. @@ -137,9 +183,18 @@ util_make_fragment_tex_shader(struct pipe_context *pipe ) void * util_make_fragment_passthrough_shader(struct pipe_context *pipe) { + return util_make_fragment_clonecolor_shader(pipe, 1); +} + +void * +util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs) +{ struct ureg_program *ureg; struct ureg_src src; - struct ureg_dst dst; + struct ureg_dst dst[8]; + int i; + + assert(num_cbufs <= 8); ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); if (ureg == NULL) @@ -148,12 +203,13 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0, TGSI_INTERPOLATE_PERSPECTIVE ); - dst = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 ); + for (i = 0; i < num_cbufs; i++) + dst[i] = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, i ); + + for (i = 0; i < num_cbufs; i++) + ureg_MOV( ureg, dst[i], src ); - ureg_MOV( ureg, dst, src ); ureg_END( ureg ); return ureg_create_shader_and_destroy( ureg, pipe ); } - - diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index d2e80d6eb4d..6e760942e25 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -51,16 +51,25 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, extern void * util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, - unsigned writemask ); + unsigned tex_target, + unsigned writemask); extern void * -util_make_fragment_tex_shader(struct pipe_context *pipe); +util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target); + + +extern void * +util_make_fragment_tex_shader_writedepth(struct pipe_context *pipe, + unsigned tex_target); extern void * util_make_fragment_passthrough_shader(struct pipe_context *pipe); +extern void * +util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs); + #ifdef __cplusplus } #endif diff --git a/src/gallium/auxiliary/util/u_surface.c b/src/gallium/auxiliary/util/u_surface.c index f828908f0be..35c49782043 100644 --- a/src/gallium/auxiliary/util/u_surface.c +++ b/src/gallium/auxiliary/util/u_surface.c @@ -36,6 +36,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" +#include "util/u_format.h" #include "util/u_surface.h" diff --git a/src/gallium/auxiliary/util/u_texture.c b/src/gallium/auxiliary/util/u_texture.c new file mode 100644 index 00000000000..cd477ab640f --- /dev/null +++ b/src/gallium/auxiliary/util/u_texture.c @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * Copyright 2008 VMware, Inc. All rights reserved. + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Texture mapping utility functions. + * + * @author Brian Paul + * Marek Olšák + */ + +#include "pipe/p_defines.h" + +#include "util/u_texture.h" + +void util_map_texcoords2d_onto_cubemap(unsigned face, + const float *in_st, unsigned in_stride, + float *out_str, unsigned out_stride) +{ + int i; + float rx, ry, rz; + + /* loop over quad verts */ + for (i = 0; i < 4; i++) { + /* Compute sc = +/-scale and tc = +/-scale. + * Not +/-1 to avoid cube face selection ambiguity near the edges, + * though that can still sometimes happen with this scale factor... + */ + const float scale = 0.9999f; + const float sc = (2 * in_st[0] - 1) * scale; + const float tc = (2 * in_st[1] - 1) * scale; + + switch (face) { + case PIPE_TEX_FACE_POS_X: + rx = 1; + ry = -tc; + rz = -sc; + break; + case PIPE_TEX_FACE_NEG_X: + rx = -1; + ry = -tc; + rz = sc; + break; + case PIPE_TEX_FACE_POS_Y: + rx = sc; + ry = 1; + rz = tc; + break; + case PIPE_TEX_FACE_NEG_Y: + rx = sc; + ry = -1; + rz = -tc; + break; + case PIPE_TEX_FACE_POS_Z: + rx = sc; + ry = -tc; + rz = 1; + break; + case PIPE_TEX_FACE_NEG_Z: + rx = -sc; + ry = -tc; + rz = -1; + break; + default: + rx = ry = rz = 0; + assert(0); + } + + out_str[0] = rx; /*s*/ + out_str[1] = ry; /*t*/ + out_str[2] = rz; /*r*/ + + in_st += in_stride; + out_str += out_stride; + } +} diff --git a/src/gallium/auxiliary/util/u_texture.h b/src/gallium/auxiliary/util/u_texture.h new file mode 100644 index 00000000000..93b2f1e4c97 --- /dev/null +++ b/src/gallium/auxiliary/util/u_texture.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_TEXTURE_H +#define U_TEXTURE_H + +#ifdef __cplusplus +extern "C" { +#endif + +/** + * Convert 2D texture coordinates of 4 vertices into cubemap coordinates + * in the given face. + * Coordinates must be in the range [0,1]. + * + * \param face Cubemap face. + * \param in_st 4 pairs of 2D texture coordinates to convert. + * \param in_stride Stride of in_st in floats. + * \param out_str STR cubemap texture coordinates to compute. + * \param out_stride Stride of out_str in floats. + */ +void util_map_texcoords2d_onto_cubemap(unsigned face, + const float *in_st, unsigned in_stride, + float *out_str, unsigned out_stride); + + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 88c9a1f0977..5b8dd1abb94 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -34,6 +34,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_rect.h" @@ -52,7 +53,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt, const void *src; if (dst_stride == 0) - dst_stride = pf_get_stride(pt->texture->format, w); + dst_stride = util_format_get_stride(pt->texture->format, w); if (pipe_clip_tile(x, y, &w, &h, pt)) return; @@ -81,7 +82,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt, enum pipe_format format = pt->texture->format; if (src_stride == 0) - src_stride = pf_get_stride(format, w); + src_stride = util_format_get_stride(format, w); if (pipe_clip_tile(x, y, &w, &h, pt)) return; @@ -1275,7 +1276,7 @@ pipe_get_tile_rgba(struct pipe_transfer *pt, if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format)); + packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format)); if (!packed) return; @@ -1303,7 +1304,7 @@ pipe_put_tile_rgba(struct pipe_transfer *pt, if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format)); + packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format)); if (!packed) return; diff --git a/src/gallium/auxiliary/util/u_upload_mgr.h b/src/gallium/auxiliary/util/u_upload_mgr.h index 745b5834af6..e158bed9d04 100644 --- a/src/gallium/auxiliary/util/u_upload_mgr.h +++ b/src/gallium/auxiliary/util/u_upload_mgr.h @@ -32,6 +32,8 @@ #ifndef U_UPLOAD_MGR_H #define U_UPLOAD_MGR_H +#include "pipe/p_defines.h" + struct pipe_screen; struct pipe_buffer; struct u_upload_mgr; diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index 4952e9c9f8e..ab196c21f87 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@ -29,6 +29,7 @@ #include <assert.h> #include <pipe/p_context.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_math.h> #include <util/u_memory.h> #include <tgsi/tgsi_parse.h> @@ -1443,7 +1444,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby, assert(r); assert(blocks); - tex_pitch = r->tex_transfer[0]->stride / pf_get_blocksize(r->tex_transfer[0]->texture->format); + tex_pitch = r->tex_transfer[0]->stride / util_format_get_blocksize(r->tex_transfer[0]->texture->format); texels = r->texels[0] + mbpy * tex_pitch + mbpx; for (y = 0; y < 2; ++y) { @@ -1482,7 +1483,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby, mbpy /= 2; for (tb = 0; tb < 2; ++tb) { - tex_pitch = r->tex_transfer[tb + 1]->stride / pf_get_blocksize(r->tex_transfer[tb + 1]->texture->format); + tex_pitch = r->tex_transfer[tb + 1]->stride / util_format_get_blocksize(r->tex_transfer[tb + 1]->texture->format); texels = r->texels[tb + 1] + mbpy * tex_pitch + mbpx; if ((cbp >> (1 - tb)) & 1) { diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 77a57aef147..998944f77a3 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -35,6 +35,8 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" + +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -65,11 +67,11 @@ cell_texture_layout(struct cell_texture *ct) w_tile = align(width, TILE_SIZE); h_tile = align(height, TILE_SIZE); - ct->stride[level] = pf_get_stride(pt->format, w_tile); + ct->stride[level] = util_format_get_stride(pt->format, w_tile); ct->level_offset[level] = ct->buffer_size; - size = ct->stride[level] * pf_get_nblocksy(pt->format, h_tile); + size = ct->stride[level] * util_format_get_nblocksy(pt->format, h_tile); if (pt->target == PIPE_TEXTURE_CUBE) size *= 6; else @@ -281,11 +283,11 @@ cell_get_tex_surface(struct pipe_screen *screen, if (pt->target == PIPE_TEXTURE_CUBE) { unsigned h_tile = align(ps->height, TILE_SIZE); - ps->offset += face * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; + ps->offset += face * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { unsigned h_tile = align(ps->height, TILE_SIZE); - ps->offset += zslice * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; + ps->offset += zslice * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else { assert(face == 0); @@ -340,11 +342,11 @@ cell_get_tex_transfer(struct pipe_screen *screen, if (texture->target == PIPE_TEXTURE_CUBE) { unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); - ctrans->offset += face * pf_get_nblocksy(texture->format, h_tile) * pt->stride; + ctrans->offset += face * util_format_get_nblocksy(texture->format, h_tile) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); - ctrans->offset += zslice * pf_get_nblocksy(texture->format, h_tile) * pt->stride; + ctrans->offset += zslice * util_format_get_nblocksy(texture->format, h_tile) * pt->stride; } else { assert(face == 0); @@ -397,8 +399,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) * Create a buffer of ordinary memory for the linear texture. * This is the memory that the user will read/write. */ - size = pf_get_stride(pt->format, align(texWidth, TILE_SIZE)) * - pf_get_nblocksy(pt->format, align(texHeight, TILE_SIZE)); + size = util_format_get_stride(pt->format, align(texWidth, TILE_SIZE)) * + util_format_get_nblocksy(pt->format, align(texHeight, TILE_SIZE)); ctrans->map = align_malloc(size, 16); if (!ctrans->map) @@ -406,7 +408,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) if (transfer->usage & PIPE_TRANSFER_READ) { /* need to untwiddle the texture to make a linear version */ - const uint bpp = pf_get_blocksize(ct->base.format); + const uint bpp = util_format_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = (uint *) (ct->mapped + ctrans->offset); uint *dst = ctrans->map; @@ -449,7 +451,7 @@ cell_transfer_unmap(struct pipe_screen *screen, /* The user wrote new texture data into the mapped buffer. * We need to convert the new linear data into the twiddled/tiled format. */ - const uint bpp = pf_get_blocksize(ct->base.format); + const uint bpp = util_format_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = ctrans->map; uint *dst = (uint *) (ct->mapped + ctrans->offset); diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c index 24e1024aaa3..c693eb30e87 100644 --- a/src/gallium/drivers/i915/i915_surface.c +++ b/src/gallium/drivers/i915/i915_surface.c @@ -32,6 +32,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" +#include "util/u_format.h" #include "util/u_tile.h" #include "util/u_rect.h" @@ -52,15 +53,15 @@ i915_surface_copy(struct pipe_context *pipe, struct pipe_texture *spt = &src_tex->base; assert( dst != src ); - assert( pf_get_blocksize(dpt->format) == pf_get_blocksize(spt->format) ); - assert( pf_get_blockwidth(dpt->format) == pf_get_blockwidth(spt->format) ); - assert( pf_get_blockheight(dpt->format) == pf_get_blockheight(spt->format) ); - assert( pf_get_blockwidth(dpt->format) == 1 ); - assert( pf_get_blockheight(dpt->format) == 1 ); + assert( util_format_get_blocksize(dpt->format) == util_format_get_blocksize(spt->format) ); + assert( util_format_get_blockwidth(dpt->format) == util_format_get_blockwidth(spt->format) ); + assert( util_format_get_blockheight(dpt->format) == util_format_get_blockheight(spt->format) ); + assert( util_format_get_blockwidth(dpt->format) == 1 ); + assert( util_format_get_blockheight(dpt->format) == 1 ); i915_copy_blit( i915_context(pipe), FALSE, - pf_get_blocksize(dpt->format), + util_format_get_blocksize(dpt->format), (unsigned short) src_tex->stride, src_tex->buffer, src->offset, (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); @@ -76,11 +77,11 @@ i915_surface_fill(struct pipe_context *pipe, struct i915_texture *tex = (struct i915_texture *)dst->texture; struct pipe_texture *pt = &tex->base; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); i915_fill_blit( i915_context(pipe), - pf_get_blocksize(pt->format), + util_format_get_blocksize(pt->format), (unsigned short) tex->stride, tex->buffer, dst->offset, (short) dstx, (short) dsty, diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c index b28b413771d..50a9e19094b 100644 --- a/src/gallium/drivers/i915/i915_texture.c +++ b/src/gallium/drivers/i915/i915_texture.c @@ -35,6 +35,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -129,7 +130,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); - tex->image_offset[level][img] = y * tex->stride + x * pf_get_blocksize(tex->base.format); + tex->image_offset[level][img] = y * tex->stride + x * util_format_get_blocksize(tex->base.format); /* printf("%s level %d img %d pos %d,%d image_offset %x\n", @@ -151,7 +152,7 @@ i915_scanout_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) + if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4) return FALSE; i915_miptree_set_level_info(tex, 0, 1, @@ -161,18 +162,18 @@ i915_scanout_layout(struct i915_texture *tex) i915_miptree_set_image_offset(tex, 0, 0, 0, 0); if (pt->width0 >= 240) { - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; } else if (pt->width0 == 64 && pt->height0 == 64) { - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); } else { return FALSE; } debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - pt->width0, pt->height0, pf_get_blocksize(pt->format), + pt->width0, pt->height0, util_format_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -186,7 +187,7 @@ i915_display_target_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) + if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4) return FALSE; /* fallback to normal textures for small textures */ @@ -199,12 +200,12 @@ i915_display_target_layout(struct i915_texture *tex) 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - pt->width0, pt->height0, pf_get_blocksize(pt->format), + pt->width0, pt->height0, util_format_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -217,7 +218,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) unsigned level; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0); /* used for scanouts that need special layouts */ if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) @@ -229,7 +230,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) if (i915_display_target_layout(tex)) return; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { @@ -242,7 +243,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -255,12 +256,12 @@ i915_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0); unsigned stack_nblocksy = 0; /* Calculate the size of a single slice. */ - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); /* XXX: hardware expects/requires 9 levels at minimum. */ @@ -271,7 +272,7 @@ i915_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } /* Fixup depth image_offsets: @@ -296,14 +297,14 @@ i915_miptree_layout_cube(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; unsigned width = pt->width0, height = pt->height0; - const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); + const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0); unsigned level; unsigned face; assert(width == height); /* cubemap images are square */ /* double pitch for cube layouts */ - tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); + tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4); tex->total_nblocksy = nblocks * 4; for (level = 0; level <= pt->last_level; level++) { @@ -366,8 +367,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) unsigned y = 0; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksx = pf_get_nblocksx(pt->format, pt->width0); - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); + unsigned nblocksx = util_format_get_nblocksx(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0); /* used for scanouts that need special layouts */ if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) @@ -379,7 +380,7 @@ i945_miptree_layout_2d(struct i915_texture *tex) if (i915_display_target_layout(tex)) return; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); /* May need to adjust pitch to accomodate the placement of * the 2nd mipmap level. This occurs when the alignment @@ -388,11 +389,11 @@ i945_miptree_layout_2d(struct i915_texture *tex) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align(pf_get_nblocksx(pt->format, u_minify(width, 1)), align_x) - + pf_get_nblocksx(pt->format, u_minify(width, 2)); + = align(util_format_get_nblocksx(pt->format, u_minify(width, 1)), align_x) + + util_format_get_nblocksx(pt->format, u_minify(width, 2)); if (mip1_nblocksx > nblocksx) - tex->stride = mip1_nblocksx * pf_get_blocksize(pt->format); + tex->stride = mip1_nblocksx * util_format_get_blocksize(pt->format); } /* Pitch must be a whole number of dwords @@ -422,8 +423,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksx = pf_get_nblocksx(pt->format, width); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksx = util_format_get_nblocksx(pt->format, width); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -434,16 +435,16 @@ i945_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0); unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; unsigned level; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; pack_y_pitch = MAX2(nblocksy, 2); - pack_x_pitch = tex->stride / pf_get_blocksize(pt->format); + pack_x_pitch = tex->stride / util_format_get_blocksize(pt->format); pack_x_nr = 1; for (level = 0; level <= pt->last_level; level++) { @@ -468,7 +469,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) if (pack_x_pitch > 4) { pack_x_pitch >>= 1; pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr * pf_get_blocksize(pt->format) <= tex->stride); + assert(pack_x_pitch * pack_x_nr * util_format_get_blocksize(pt->format) <= tex->stride); } if (pack_y_pitch > 2) { @@ -478,7 +479,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); depth = u_minify(depth, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -488,7 +489,7 @@ i945_miptree_layout_cube(struct i915_texture *tex) struct pipe_texture *pt = &tex->base; unsigned level; - const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); + const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0); unsigned face; unsigned width = pt->width0; unsigned height = pt->height0; @@ -508,9 +509,9 @@ i945_miptree_layout_cube(struct i915_texture *tex) * or the final row of 4x4, 2x2 and 1x1 faces below this. */ if (nblocks > 32) - tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); + tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4); else - tex->stride = 14 * 8 * pf_get_blocksize(pt->format); + tex->stride = 14 * 8 * util_format_get_blocksize(pt->format); tex->total_nblocksy = nblocks * 4; @@ -840,8 +841,8 @@ i915_transfer_map(struct pipe_screen *screen, return NULL; return map + i915_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); } static void diff --git a/src/gallium/drivers/i965/Makefile b/src/gallium/drivers/i965/Makefile new file mode 100644 index 00000000000..95fd3cd69bd --- /dev/null +++ b/src/gallium/drivers/i965/Makefile @@ -0,0 +1,74 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = i965 + +C_SOURCES = \ + brw_cc.c \ + brw_clip.c \ + brw_clip_line.c \ + brw_clip_point.c \ + brw_clip_state.c \ + brw_clip_tri.c \ + brw_clip_unfilled.c \ + brw_clip_util.c \ + brw_context.c \ + brw_curbe.c \ + brw_disasm.c \ + brw_draw.c \ + brw_draw_upload.c \ + brw_eu.c \ + brw_eu_debug.c \ + brw_eu_emit.c \ + brw_eu_util.c \ + brw_gs.c \ + brw_gs_emit.c \ + brw_gs_state.c \ + brw_misc_state.c \ + brw_pipe_blend.c \ + brw_pipe_depth.c \ + brw_pipe_fb.c \ + brw_pipe_query.c \ + brw_pipe_shader.c \ + brw_pipe_flush.c \ + brw_pipe_misc.c \ + brw_pipe_sampler.c \ + brw_pipe_vertex.c \ + brw_pipe_clear.c \ + brw_pipe_rast.c \ + brw_sf.c \ + brw_sf_emit.c \ + brw_sf_state.c \ + brw_state_batch.c \ + brw_state_debug.c \ + brw_state_cache.c \ + brw_state_upload.c \ + brw_structs_dump.c \ + brw_swtnl.c \ + brw_urb.c \ + brw_util.c \ + brw_vs.c \ + brw_vs_emit.c \ + brw_vs_state.c \ + brw_vs_surface_state.c \ + brw_wm.c \ + brw_wm_debug.c \ + brw_wm_emit.c \ + brw_wm_fp.c \ + brw_wm_iz.c \ + brw_wm_pass0.c \ + brw_wm_pass1.c \ + brw_wm_pass2.c \ + brw_wm_sampler_state.c \ + brw_wm_state.c \ + brw_wm_surface_state.c \ + brw_screen.c \ + brw_screen_buffers.c \ + brw_screen_tex_layout.c \ + brw_screen_texture.c \ + brw_screen_surface.c \ + brw_batchbuffer.c \ + brw_winsys_debug.c \ + intel_decode.c + +include ../../Makefile.template diff --git a/src/gallium/drivers/i965/SConscript b/src/gallium/drivers/i965/SConscript new file mode 100644 index 00000000000..9c2faaf4b49 --- /dev/null +++ b/src/gallium/drivers/i965/SConscript @@ -0,0 +1,77 @@ +Import('*') + +env = env.Clone() + +i965 = env.ConvenienceLibrary( + target = 'i965', + source = [ + 'brw_batchbuffer.c', + 'brw_cc.c', + 'brw_clip.c', + 'brw_clip_line.c', + 'brw_clip_point.c', + 'brw_clip_state.c', + 'brw_clip_tri.c', + 'brw_clip_unfilled.c', + 'brw_clip_util.c', + 'brw_context.c', + 'brw_curbe.c', + 'brw_disasm.c', + 'brw_draw.c', + 'brw_draw_upload.c', + 'brw_eu.c', + 'brw_eu_debug.c', + 'brw_eu_emit.c', + 'brw_eu_util.c', + 'brw_gs.c', + 'brw_gs_emit.c', + 'brw_gs_state.c', + 'brw_misc_state.c', + 'brw_pipe_blend.c', + 'brw_pipe_clear.c', + 'brw_pipe_depth.c', + 'brw_pipe_fb.c', + 'brw_pipe_flush.c', + 'brw_pipe_misc.c', + 'brw_pipe_query.c', + 'brw_pipe_rast.c', + 'brw_pipe_sampler.c', + 'brw_pipe_shader.c', + 'brw_pipe_vertex.c', + 'brw_screen_buffers.c', + 'brw_screen.c', + 'brw_screen_surface.c', + 'brw_screen_tex_layout.c', + 'brw_screen_texture.c', + 'brw_structs_dump.c', + 'brw_sf.c', + 'brw_sf_emit.c', + 'brw_sf_state.c', + 'brw_state_batch.c', + 'brw_state_cache.c', +# 'brw_state_debug.c', + 'brw_state_upload.c', + 'brw_swtnl.c', + 'brw_urb.c', + 'brw_util.c', + 'brw_vs.c', + 'brw_vs_emit.c', + 'brw_vs_state.c', + 'brw_vs_surface_state.c', + 'brw_wm.c', +# 'brw_wm_constant_buffer.c', + 'brw_wm_debug.c', + 'brw_wm_emit.c', + 'brw_wm_fp.c', +# 'brw_wm_glsl.c', + 'brw_wm_iz.c', + 'brw_wm_pass0.c', + 'brw_wm_pass1.c', + 'brw_wm_pass2.c', + 'brw_wm_sampler_state.c', + 'brw_wm_state.c', + 'brw_wm_surface_state.c', + 'intel_decode.c', + ]) + +Export('i965') diff --git a/src/gallium/drivers/i965/brw_batchbuffer.c b/src/gallium/drivers/i965/brw_batchbuffer.c new file mode 100644 index 00000000000..22607dc6083 --- /dev/null +++ b/src/gallium/drivers/i965/brw_batchbuffer.c @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "util/u_memory.h" + +#include "brw_batchbuffer.h" +#include "brw_reg.h" +#include "brw_winsys.h" +#include "brw_debug.h" +#include "brw_structs.h" + +#define ALWAYS_EMIT_MI_FLUSH 1 + +enum pipe_error +brw_batchbuffer_reset(struct brw_batchbuffer *batch) +{ + enum pipe_error ret; + + ret = batch->sws->bo_alloc( batch->sws, + BRW_BUFFER_TYPE_BATCH, + BRW_BATCH_SIZE, 4096, + &batch->buf ); + if (ret) + return ret; + + batch->size = BRW_BATCH_SIZE; + + /* With map_range semantics, the winsys can decide whether to + * inject a malloc'ed bounce buffer instead of mapping directly. + */ + batch->map = batch->sws->bo_map(batch->buf, + BRW_DATA_BATCH_BUFFER, + 0, batch->size, + GL_TRUE, + GL_TRUE, + GL_TRUE); + + batch->ptr = batch->map; + return PIPE_OK; +} + +struct brw_batchbuffer * +brw_batchbuffer_alloc(struct brw_winsys_screen *sws, + struct brw_chipset chipset) +{ + struct brw_batchbuffer *batch = CALLOC_STRUCT(brw_batchbuffer); + + batch->sws = sws; + batch->chipset = chipset; + brw_batchbuffer_reset(batch); + + return batch; +} + +void +brw_batchbuffer_free(struct brw_batchbuffer *batch) +{ + if (batch->map) { + batch->sws->bo_unmap(batch->buf); + batch->map = NULL; + } + + bo_reference(&batch->buf, NULL); + FREE(batch); +} + + +void +_brw_batchbuffer_flush(struct brw_batchbuffer *batch, + const char *file, + int line) +{ + GLuint used = batch->ptr - batch->map; + + if (used == 0) + return; + + /* Post-swap throttling done by the state tracker. + */ + + if (BRW_DEBUG & DEBUG_BATCH) + debug_printf("%s:%d: Batchbuffer flush with %db used\n", + file, line, used); + + if (ALWAYS_EMIT_MI_FLUSH) { + *(GLuint *) (batch->ptr) = MI_FLUSH | BRW_FLUSH_STATE_CACHE; + batch->ptr += 4; + used = batch->ptr - batch->map; + } + + /* Round batchbuffer usage to 2 DWORDs. + */ + if ((used & 4) == 0) { + *(GLuint *) (batch->ptr) = 0; /* noop */ + batch->ptr += 4; + used = batch->ptr - batch->map; + } + + /* Mark the end of the buffer. + */ + *(GLuint *) (batch->ptr) = MI_BATCH_BUFFER_END; + batch->ptr += 4; + used = batch->ptr - batch->map; + + batch->sws->bo_flush_range(batch->buf, 0, used); + batch->sws->bo_unmap(batch->buf); + batch->map = NULL; + batch->ptr = NULL; + + batch->sws->bo_exec(batch->buf, used ); + + if (BRW_DEBUG & DEBUG_SYNC) { + /* Abuse map/unmap to achieve wait-for-fence. + * + * XXX: hide this inside the winsys and export a fence + * interface. + */ + debug_printf("waiting for idle\n"); + batch->sws->bo_wait_idle(batch->buf); + } + + /* Reset the buffer: + */ + brw_batchbuffer_reset(batch); +} + + +/* The OUT_RELOC() macro ends up here, generating a relocation within + * the batch buffer. + */ +enum pipe_error +brw_batchbuffer_emit_reloc(struct brw_batchbuffer *batch, + struct brw_winsys_buffer *buffer, + uint32_t usage, + uint32_t delta) +{ + int ret; + + if (batch->ptr - batch->map > batch->buf->size) { + debug_printf("bad relocation ptr %p map %p offset %d size %d\n", + batch->ptr, batch->map, batch->ptr - batch->map, batch->buf->size); + + return PIPE_ERROR_OUT_OF_MEMORY; + } + + ret = batch->sws->bo_emit_reloc(batch->buf, + usage, + delta, + batch->ptr - batch->map, + buffer); + if (ret != 0) + return ret; + + /* bo_emit_reloc was resposible for writing a zero into the + * batchbuffer if necessary. Just need to update our pointer. + */ + batch->ptr += 4; + + return 0; +} + +enum pipe_error +brw_batchbuffer_data(struct brw_batchbuffer *batch, + const void *data, GLuint bytes, + enum cliprect_mode cliprect_mode) +{ + enum pipe_error ret; + + assert((bytes & 3) == 0); + + ret = brw_batchbuffer_require_space(batch, bytes); + if (ret) + return ret; + + memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; + return 0; +} diff --git a/src/gallium/drivers/i965/brw_batchbuffer.h b/src/gallium/drivers/i965/brw_batchbuffer.h new file mode 100644 index 00000000000..7473f5bea4d --- /dev/null +++ b/src/gallium/drivers/i965/brw_batchbuffer.h @@ -0,0 +1,148 @@ +#ifndef BRW_BATCHBUFFER_H +#define BRW_BATCHBUFFER_H + +#include "util/u_debug.h" + +#include "brw_types.h" +#include "brw_winsys.h" +#include "brw_reg.h" + +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 + +/* All ignored: + */ +enum cliprect_mode { + IGNORE_CLIPRECTS, + LOOP_CLIPRECTS, + NO_LOOP_CLIPRECTS, + REFERENCES_CLIPRECTS +}; + + + + +struct brw_batchbuffer { + + struct brw_winsys_screen *sws; + struct brw_winsys_buffer *buf; + struct brw_chipset chipset; + + /** + * Values exported to speed up the writing the batchbuffer, + * instead of having to go trough a accesor function for + * each dword written. + */ + /*{@*/ + uint8_t *map; + uint8_t *ptr; + size_t size; + struct { + uint8_t *end_ptr; + } emit; + + + size_t relocs; + size_t max_relocs; + /*@}*/ +}; + +struct brw_batchbuffer *brw_batchbuffer_alloc( struct brw_winsys_screen *sws, + struct brw_chipset chipset ); + +void brw_batchbuffer_free(struct brw_batchbuffer *batch); + +void _brw_batchbuffer_flush(struct brw_batchbuffer *batch, + const char *file, int line); + + +enum pipe_error +brw_batchbuffer_reset(struct brw_batchbuffer *batch); + + +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multple + * intel_buffer_dword() calls. + */ +int brw_batchbuffer_data(struct brw_batchbuffer *batch, + const void *data, GLuint bytes, + enum cliprect_mode cliprect_mode); + + +int brw_batchbuffer_emit_reloc(struct brw_batchbuffer *batch, + struct brw_winsys_buffer *buffer, + enum brw_buffer_usage usage, + uint32_t offset); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE GLint +brw_batchbuffer_space(struct brw_batchbuffer *batch) +{ + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} + + +static INLINE void +brw_batchbuffer_emit_dword(struct brw_batchbuffer *batch, GLuint dword) +{ + assert(batch->map); + assert(brw_batchbuffer_space(batch) >= 4); + *(GLuint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE enum pipe_error +brw_batchbuffer_require_space(struct brw_batchbuffer *batch, + GLuint sz) +{ + assert(sz < batch->size - 8); + if (brw_batchbuffer_space(batch) < sz) { + assert(0); + return PIPE_ERROR_OUT_OF_MEMORY; + } +#ifdef DEBUG + batch->emit.end_ptr = batch->ptr + sz; +#endif + return 0; +} + +/* Here are the crusty old macros, to be removed: + */ +#define BEGIN_BATCH(n, cliprect_mode) do { \ + brw_batchbuffer_require_space(brw->batch, (n)*4); \ + } while (0) + +#define OUT_BATCH(d) brw_batchbuffer_emit_dword(brw->batch, d) + +#define OUT_RELOC(buf, usage, delta) do { \ + assert((unsigned) (delta) < buf->size); \ + brw_batchbuffer_emit_reloc(brw->batch, buf, \ + usage, delta); \ + } while (0) + +#ifdef DEBUG +#define ADVANCE_BATCH() do { \ + unsigned int _n = brw->batch->ptr - brw->batch->emit.end_ptr; \ + if (_n != 0) { \ + debug_printf("%s: %d too many bytes emitted to batch\n", \ + __FUNCTION__, _n); \ + abort(); \ + } \ + brw->batch->emit.end_ptr = NULL; \ + } while(0) +#else +#define ADVANCE_BATCH() +#endif + +static INLINE void +brw_batchbuffer_emit_mi_flush(struct brw_batchbuffer *batch) +{ + brw_batchbuffer_require_space(batch, 4); + brw_batchbuffer_emit_dword(batch, MI_FLUSH); +} + +#endif diff --git a/src/gallium/drivers/i965/brw_cc.c b/src/gallium/drivers/i965/brw_cc.c new file mode 100644 index 00000000000..3e070f5591a --- /dev/null +++ b/src/gallium/drivers/i965/brw_cc.c @@ -0,0 +1,111 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + +static enum pipe_error prepare_cc_vp( struct brw_context *brw ) +{ + return brw_cache_data( &brw->cache, + BRW_CC_VP, + &brw->curr.ccv, + NULL, 0, + &brw->cc.reloc[CC_RELOC_VP].bo ); +} + +const struct brw_tracked_state brw_cc_vp = { + .dirty = { + .mesa = PIPE_NEW_VIEWPORT, + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .prepare = prepare_cc_vp +}; + + +/* A long-winded way to OR two unsigned integers together: + */ +static INLINE struct brw_cc3 +combine_cc3( struct brw_cc3 a, struct brw_cc3 b ) +{ + union { struct brw_cc3 cc3; unsigned i; } ca, cb; + ca.cc3 = a; + cb.cc3 = b; + ca.i |= cb.i; + return ca.cc3; +} + + +static int prepare_cc_unit( struct brw_context *brw ) +{ + brw->cc.cc.cc0 = brw->curr.zstencil->cc0; + brw->cc.cc.cc1 = brw->curr.zstencil->cc1; + brw->cc.cc.cc2 = brw->curr.zstencil->cc2; + brw->cc.cc.cc3 = combine_cc3( brw->curr.zstencil->cc3, brw->curr.blend->cc3 ); + + brw->cc.cc.cc5 = brw->curr.blend->cc5; + brw->cc.cc.cc6 = brw->curr.blend->cc6; + brw->cc.cc.cc7 = brw->curr.zstencil->cc7; + + return brw_cache_data_sz(&brw->cache, BRW_CC_UNIT, + &brw->cc.cc, sizeof(brw->cc.cc), + brw->cc.reloc, 1, + &brw->cc.state_bo); +} + +const struct brw_tracked_state brw_cc_unit = { + .dirty = { + .mesa = PIPE_NEW_DEPTH_STENCIL_ALPHA | PIPE_NEW_BLEND, + .brw = 0, + .cache = CACHE_NEW_CC_VP + }, + .prepare = prepare_cc_unit, +}; + + +void brw_hw_cc_init( struct brw_context *brw ) +{ + make_reloc(&brw->cc.reloc[0], + BRW_USAGE_STATE, + 0, + offsetof(struct brw_cc_unit_state, cc4), + NULL); +} + + +void brw_hw_cc_cleanup( struct brw_context *brw ) +{ + bo_reference(&brw->cc.state_bo, NULL); + bo_reference(&brw->cc.reloc[0].bo, NULL); +} diff --git a/src/gallium/drivers/i965/brw_clip.c b/src/gallium/drivers/i965/brw_clip.c new file mode 100644 index 00000000000..58d9e56df27 --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip.c @@ -0,0 +1,224 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "pipe/p_state.h" + +#include "util/u_math.h" + +#include "brw_screen.h" +#include "brw_batchbuffer.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_pipe_rast.h" +#include "brw_clip.h" + + +#define FRONT_UNFILLED_BIT 0x1 +#define BACK_UNFILLED_BIT 0x2 + + +static enum pipe_error +compile_clip_prog( struct brw_context *brw, + struct brw_clip_prog_key *key, + struct brw_winsys_buffer **bo_out ) +{ + enum pipe_error ret; + struct brw_clip_compile c; + const GLuint *program; + GLuint program_size; + GLuint delta; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(brw, &c.func); + + c.func.single_program_flow = 1; + + c.chipset = brw->chipset; + c.key = *key; + c.need_ff_sync = c.chipset.is_igdng; + + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.header_position_offset = ATTR_SIZE; + + if (c.chipset.is_igdng) + delta = 3 * REG_SIZE; + else + delta = REG_SIZE; + + c.offset_hpos = delta + c.key.output_hpos * ATTR_SIZE; + + if (c.key.output_color0) + c.offset_color0 = delta + c.key.output_color0 * ATTR_SIZE; + + if (c.key.output_color1) + c.offset_color1 = delta + c.key.output_color1 * ATTR_SIZE; + + if (c.key.output_bfc0) + c.offset_bfc0 = delta + c.key.output_bfc0 * ATTR_SIZE; + + if (c.key.output_bfc1) + c.offset_bfc1 = delta + c.key.output_bfc1 * ATTR_SIZE; + + if (c.key.output_edgeflag) + c.offset_edgeflag = delta + c.key.output_edgeflag * ATTR_SIZE; + + if (BRW_IS_IGDNG(brw)) + c.nr_regs = (c.key.nr_attrs + 1) / 2 + 3; /* are vertices packed, or reg-aligned? */ + else + c.nr_regs = (c.key.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ + + c.nr_bytes = c.nr_regs * REG_SIZE; + + c.prog_data.clip_mode = c.key.clip_mode; /* XXX */ + + /* For some reason the thread is spawned with only 4 channels + * unmasked. + */ + brw_set_mask_control(&c.func, BRW_MASK_DISABLE); + + + /* Would ideally have the option of producing a program which could + * do all three: + */ + switch (key->primitive) { + case PIPE_PRIM_TRIANGLES: + if (key->do_unfilled) + brw_emit_unfilled_clip( &c ); + else + brw_emit_tri_clip( &c ); + break; + case PIPE_PRIM_LINES: + brw_emit_line_clip( &c ); + break; + case PIPE_PRIM_POINTS: + brw_emit_point_clip( &c ); + break; + default: + assert(0); + return PIPE_ERROR_BAD_INPUT; + } + + + + /* get the program + */ + ret = brw_get_program(&c.func, &program, &program_size); + if (ret) + return ret; + + /* Upload + */ + ret = brw_upload_cache( &brw->cache, + BRW_CLIP_PROG, + &c.key, sizeof(c.key), + NULL, 0, + program, program_size, + &c.prog_data, + &brw->clip.prog_data, + bo_out ); + if (ret) + return ret; + + return PIPE_OK; +} + +/* Calculate interpolants for triangle and line rasterization. + */ +static enum pipe_error +upload_clip_prog(struct brw_context *brw) +{ + const struct brw_vertex_shader *vs = brw->curr.vertex_shader; + struct brw_clip_prog_key key; + enum pipe_error ret; + + /* Populate the key, starting from the almost-complete version from + * the rast state. + */ + + /* PIPE_NEW_RAST */ + key = brw->curr.rast->clip_key; + + /* BRW_NEW_REDUCED_PRIMITIVE */ + key.primitive = brw->reduced_primitive; + + /* XXX: if edgeflag is moved to a proper TGSI vs output, can remove + * dependency on CACHE_NEW_VS_PROG + */ + /* CACHE_NEW_VS_PROG */ + key.nr_attrs = brw->vs.prog_data->nr_outputs; + key.output_edgeflag = brw->vs.prog_data->output_edgeflag; + + /* PIPE_NEW_VS */ + key.output_hpos = vs->output_hpos; + key.output_color0 = vs->output_color0; + key.output_color1 = vs->output_color1; + key.output_bfc0 = vs->output_bfc0; + key.output_bfc1 = vs->output_bfc1; + + /* PIPE_NEW_CLIP */ + key.nr_userclip = brw->curr.ucp.nr; + + /* Already cached? + */ + if (brw_search_cache(&brw->cache, BRW_CLIP_PROG, + &key, sizeof(key), + NULL, 0, + &brw->clip.prog_data, + &brw->clip.prog_bo)) + return PIPE_OK; + + /* Compile new program: + */ + ret = compile_clip_prog( brw, &key, &brw->clip.prog_bo ); + if (ret) + return ret; + + return PIPE_OK; +} + + +const struct brw_tracked_state brw_clip_prog = { + .dirty = { + .mesa = (PIPE_NEW_RAST | + PIPE_NEW_CLIP), + .brw = (BRW_NEW_REDUCED_PRIMITIVE), + .cache = CACHE_NEW_VS_PROG + }, + .prepare = upload_clip_prog +}; diff --git a/src/gallium/drivers/i965/brw_clip.h b/src/gallium/drivers/i965/brw_clip.h new file mode 100644 index 00000000000..80e3a11a370 --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip.h @@ -0,0 +1,199 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#ifndef BRW_CLIP_H +#define BRW_CLIP_H + +#include "pipe/p_state.h" +#include "brw_reg.h" +#include "brw_eu.h" + +#define MAX_VERTS (3+6+6) + +/* Note that if unfilled primitives are being emitted, we have to fix + * up polygon offset and flatshading at this point: + */ +struct brw_clip_prog_key { + GLuint nr_attrs:6; + GLuint primitive:4; + GLuint nr_userclip:3; + GLuint do_flat_shading:1; + GLuint do_unfilled:1; + GLuint fill_cw:2; /* includes cull information */ + GLuint fill_ccw:2; /* includes cull information */ + GLuint offset_cw:1; + GLuint offset_ccw:1; + GLuint copy_bfc_cw:1; + GLuint copy_bfc_ccw:1; + GLuint clip_mode:3; + GLuint output_hpos:6; /* not always zero? */ + + GLuint output_color0:6; + GLuint output_color1:6; + GLuint output_bfc0:6; + GLuint output_bfc1:6; + GLuint output_edgeflag:6; + GLuint pad1:2; + + GLfloat offset_factor; + GLfloat offset_units; +}; + +struct brw_clip_prog_data { + GLuint curb_read_length; /* user planes? */ + GLuint clip_mode; + GLuint urb_read_length; + GLuint total_grf; +}; + +#define CLIP_LINE 0 +#define CLIP_POINT 1 +#define CLIP_FILL 2 +#define CLIP_CULL 3 + + +#define PRIM_MASK (0x1f) + +struct brw_clip_compile { + struct brw_compile func; + struct brw_clip_prog_key key; + struct brw_clip_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_VERTS]; + + struct brw_reg t; + struct brw_reg t0, t1; + struct brw_reg dp0, dp1; + + struct brw_reg dpPrev; + struct brw_reg dp; + struct brw_reg loopcount; + struct brw_reg nr_verts; + struct brw_reg planemask; + + struct brw_reg inlist; + struct brw_reg outlist; + struct brw_reg freelist; + + struct brw_reg dir; + struct brw_reg tmp0, tmp1; + struct brw_reg offset; + + struct brw_reg fixed_planes; + struct brw_reg plane_equation; + + struct brw_reg ff_sync; + } reg; + + /* 3 different ways of expressing vertex size, including + * key.nr_attrs. + */ + GLuint nr_regs; + GLuint nr_bytes; + + GLuint first_tmp; + GLuint last_tmp; + + GLboolean need_direction; + struct brw_chipset chipset; + + GLuint last_mrf; + + GLuint header_position_offset; + GLboolean need_ff_sync; + + GLuint nr_color_attrs; + GLuint offset_color0; + GLuint offset_color1; + GLuint offset_bfc0; + GLuint offset_bfc1; + + GLuint offset_hpos; + GLuint offset_edgeflag; +}; + +#define ATTR_SIZE (4*4) + +/* Points are only culled, so no need for a clip routine, however it + * works out easier to have a dummy one. + */ +void brw_emit_unfilled_clip( struct brw_clip_compile *c ); +void brw_emit_tri_clip( struct brw_clip_compile *c ); +void brw_emit_line_clip( struct brw_clip_compile *c ); +void brw_emit_point_clip( struct brw_clip_compile *c ); + +/* brw_clip_tri.c, for use by the unfilled clip routine: + */ +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ); +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ); +void brw_clip_tri( struct brw_clip_compile *c ); +void brw_clip_tri_emit_polygon( struct brw_clip_compile *c ); +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + GLuint nr_verts ); + + +/* Utils: + */ + +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + GLboolean force_edgeflag ); + +void brw_clip_init_planes( struct brw_clip_compile *c ); + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + GLboolean allocate, + GLboolean eot, + GLuint header); + +void brw_clip_kill_thread(struct brw_clip_compile *c); + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ); +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ); + +void brw_clip_copy_colors( struct brw_clip_compile *c, + GLuint to, GLuint from ); + +void brw_clip_init_clipmask( struct brw_clip_compile *c ); + +struct brw_reg get_tmp( struct brw_clip_compile *c ); + +void brw_clip_project_position(struct brw_clip_compile *c, + struct brw_reg pos ); +void brw_clip_ff_sync(struct brw_clip_compile *c); +void brw_clip_init_ff_sync(struct brw_clip_compile *c); +#endif diff --git a/src/gallium/drivers/i965/brw_clip_line.c b/src/gallium/drivers/i965/brw_clip_line.c new file mode 100644 index 00000000000..54282d975ed --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_line.c @@ -0,0 +1,271 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_debug.h" + +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + + +static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) +{ + GLuint i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < 4; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.t0 = brw_vec1_grf(i, 1); + c->reg.t1 = brw_vec1_grf(i, 2); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dp0 = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp1 = brw_vec1_grf(i, 4); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + if (c->need_ff_sync) { + c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD); + i++; + } + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +/* Line clipping, more or less following the following algorithm: + * + * for (p=0;p<MAX_PLANES;p++) { + * if (clipmask & (1 << p)) { + * GLfloat dp0 = DOTPROD( vtx0, plane[p] ); + * GLfloat dp1 = DOTPROD( vtx1, plane[p] ); + * + * if (IS_NEGATIVE(dp1)) { + * GLfloat t = dp1 / (dp1 - dp0); + * if (t > t1) t1 = t; + * } else { + * GLfloat t = dp0 / (dp0 - dp1); + * if (t > t0) t0 = t; + * } + * + * if (t0 + t1 >= 1.0) + * return; + * } + * } + * + * interp( ctx, newvtx0, vtx0, vtx1, t0 ); + * interp( ctx, newvtx1, vtx1, vtx0, t1 ); + * + */ +static void clip_and_emit_line( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_indirect vtx0 = brw_indirect(0, 0); + struct brw_indirect vtx1 = brw_indirect(1, 0); + struct brw_indirect newvtx0 = brw_indirect(2, 0); + struct brw_indirect newvtx1 = brw_indirect(3, 0); + struct brw_indirect plane_ptr = brw_indirect(4, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *is_negative; + struct brw_instruction *is_neg2 = NULL; + struct brw_instruction *not_culled; + struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD); + + brw_MOV(p, get_addr_reg(vtx0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vtx1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(newvtx0), brw_address(c->reg.vertex[2])); + brw_MOV(p, get_addr_reg(newvtx1), brw_address(c->reg.vertex[3])); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + + /* Note: init t0, t1 together: + */ + brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0)); + + brw_clip_init_planes(c); + brw_clip_init_clipmask(c); + + /* -ve rhw workaround */ + if (c->chipset.is_965) { + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f)); + } + + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + + /* dp = DP4(vtx->position, plane) + */ + brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset_hpos), c->reg.plane_equation); + + /* if (IS_NEGATIVE(dp1)) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset_hpos), c->reg.plane_equation); + is_negative = brw_IF(p, BRW_EXECUTE_1); + { + /* + * Both can be negative on GM965/G965 due to RHW workaround + * if so, this object should be rejected. + */ + if (c->chipset.is_965) { + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_LE, c->reg.dp0, brw_imm_f(0.0)); + is_neg2 = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, is_neg2); + } + + brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 ); + brw_MOV(p, c->reg.t1, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + is_negative = brw_ELSE(p, is_negative); + { + /* Coming back in. We know that both cannot be negative + * because the line would have been culled in that case. + */ + + /* If both are positive, do nothing */ + /* Only on GM965/G965 */ + if (c->chipset.is_965) { + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0)); + is_neg2 = brw_IF(p, BRW_EXECUTE_1); + } + + { + brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 ); + brw_MOV(p, c->reg.t0, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + + if (c->chipset.is_965) { + brw_ENDIF(p, is_neg2); + } + } + brw_ENDIF(p, is_negative); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* while (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); + + brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0)); + not_culled = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, FALSE); + brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, FALSE); + + brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, not_culled); + brw_clip_kill_thread(c); +} + + + +void brw_emit_line_clip( struct brw_clip_compile *c ) +{ + brw_clip_line_alloc_regs(c); + brw_clip_init_ff_sync(c); + + if (c->key.do_flat_shading) + brw_clip_copy_colors(c, 0, 1); + + clip_and_emit_line(c); +} diff --git a/src/gallium/drivers/i965/brw_clip_point.c b/src/gallium/drivers/i965/brw_clip_point.c new file mode 100644 index 00000000000..e0a5330556d --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_point.c @@ -0,0 +1,48 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + +/* Point clipping, nothing to do? + */ +void brw_emit_point_clip( struct brw_clip_compile *c ) +{ + /* Send an empty message to kill the thread: + */ + brw_clip_tri_alloc_regs(c, 0); + brw_clip_init_ff_sync(c); + + brw_clip_kill_thread(c); +} diff --git a/src/gallium/drivers/i965/brw_clip_state.c b/src/gallium/drivers/i965/brw_clip_state.c new file mode 100644 index 00000000000..5c3ccfd8d0d --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_state.c @@ -0,0 +1,209 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + +#include "brw_context.h" +#include "brw_clip.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_debug.h" + +struct brw_clip_unit_key { + unsigned int total_grf; + unsigned int urb_entry_read_length; + unsigned int curb_entry_read_length; + unsigned int clip_mode; + + unsigned int curbe_offset; + + unsigned int nr_urb_entries, urb_size; + + GLboolean depth_clamp; +}; + +static void +clip_unit_populate_key(struct brw_context *brw, struct brw_clip_unit_key *key) +{ + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_CLIP_PROG */ + key->total_grf = brw->clip.prog_data->total_grf; + key->urb_entry_read_length = brw->clip.prog_data->urb_read_length; + key->curb_entry_read_length = brw->clip.prog_data->curb_read_length; + key->clip_mode = brw->clip.prog_data->clip_mode; + + /* BRW_NEW_CURBE_OFFSETS */ + key->curbe_offset = brw->curbe.clip_start; + + /* BRW_NEW_URB_FENCE */ + key->nr_urb_entries = brw->urb.nr_clip_entries; + key->urb_size = brw->urb.vsize; + + /* */ + key->depth_clamp = 0; /* XXX: add this to gallium: ctx->Transform.DepthClamp; */ +} + +static enum pipe_error +clip_unit_create_from_key(struct brw_context *brw, + struct brw_clip_unit_key *key, + struct brw_winsys_reloc *reloc, + struct brw_winsys_buffer **bo_out) +{ + struct brw_clip_unit_state clip; + enum pipe_error ret; + + memset(&clip, 0, sizeof(clip)); + + clip.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1; + /* reloc */ + clip.thread0.kernel_start_pointer = 0; + + clip.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + clip.thread1.single_program_flow = 1; + + clip.thread3.urb_entry_read_length = key->urb_entry_read_length; + clip.thread3.const_urb_entry_read_length = key->curb_entry_read_length; + clip.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; + clip.thread3.dispatch_grf_start_reg = 1; + clip.thread3.urb_entry_read_offset = 0; + + clip.thread4.nr_urb_entries = key->nr_urb_entries; + clip.thread4.urb_entry_allocation_size = key->urb_size - 1; + /* If we have enough clip URB entries to run two threads, do so. + */ + if (key->nr_urb_entries >= 10) { + /* Half of the URB entries go to each thread, and it has to be an + * even number. + */ + assert(key->nr_urb_entries % 2 == 0); + + /* Although up to 16 concurrent Clip threads are allowed on IGDNG, + * only 2 threads can output VUEs at a time. + */ + if (BRW_IS_IGDNG(brw)) + clip.thread4.max_threads = 16 - 1; + else + clip.thread4.max_threads = 2 - 1; + } else { + assert(key->nr_urb_entries >= 5); + clip.thread4.max_threads = 1 - 1; + } + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + clip.thread4.max_threads = 0; + + if (BRW_DEBUG & DEBUG_STATS) + clip.thread4.stats_enable = 1; + + clip.clip5.userclip_enable_flags = 0x7f; + clip.clip5.userclip_must_clip = 1; + clip.clip5.guard_band_enable = 0; + if (!key->depth_clamp) + clip.clip5.viewport_z_clip_enable = 1; + clip.clip5.viewport_xy_clip_enable = 1; + clip.clip5.vertex_position_space = BRW_CLIP_NDCSPACE; + clip.clip5.api_mode = BRW_CLIP_API_OGL; + clip.clip5.clip_mode = key->clip_mode; + + if (BRW_IS_G4X(brw)) + clip.clip5.negative_w_clip_test = 1; + + clip.clip6.clipper_viewport_state_ptr = 0; + clip.viewport_xmin = -1; + clip.viewport_xmax = 1; + clip.viewport_ymin = -1; + clip.viewport_ymax = 1; + + ret = brw_upload_cache(&brw->cache, BRW_CLIP_UNIT, + key, sizeof(*key), + reloc, 1, + &clip, sizeof(clip), + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + +static int upload_clip_unit( struct brw_context *brw ) +{ + struct brw_clip_unit_key key; + struct brw_winsys_reloc reloc[1]; + unsigned grf_reg_count; + enum pipe_error ret; + + clip_unit_populate_key(brw, &key); + + grf_reg_count = align(key.total_grf, 16) / 16 - 1; + + /* clip program relocation + * + * XXX: these reloc structs are long lived and only need to be + * updated when the bound BO changes. Hopefully the stuff mixed in + * in the delta's is non-orthogonal. + */ + assert(brw->clip.prog_bo); + make_reloc(&reloc[0], + BRW_USAGE_STATE, + grf_reg_count << 1, + offsetof(struct brw_clip_unit_state, thread0), + brw->clip.prog_bo); + + + if (brw_search_cache(&brw->cache, BRW_CLIP_UNIT, + &key, sizeof(key), + reloc, 1, + NULL, + &brw->clip.state_bo)) + return PIPE_OK; + + /* Create new: + */ + ret = clip_unit_create_from_key(brw, &key, + reloc, + &brw->clip.state_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_clip_unit = { + .dirty = { + .mesa = 0, + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_CLIP_PROG + }, + .prepare = upload_clip_unit, +}; diff --git a/src/gallium/drivers/i965/brw_clip_tri.c b/src/gallium/drivers/i965/brw_clip_tri.c new file mode 100644 index 00000000000..4cde7294ea0 --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_tri.c @@ -0,0 +1,595 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + +static void release_tmps( struct brw_clip_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + GLuint nr_verts ) +{ + GLuint i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + if (c->key.nr_attrs & 1) { + for (j = 0; j < 3; j++) { + GLuint delta = c->key.nr_attrs*16 + 32; + + if (c->chipset.is_igdng) + delta = c->key.nr_attrs * 16 + 32 * 3; + + brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0)); + } + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.loopcount = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_D); + c->reg.nr_verts = retype(brw_vec1_grf(i, 2), BRW_REGISTER_TYPE_UD); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dpPrev = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp = brw_vec1_grf(i, 4); + i++; + + c->reg.inlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.outlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.freelist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + if (c->key.do_unfilled) { + c->reg.dir = brw_vec4_grf(i, 0); + c->reg.offset = brw_vec4_grf(i, 4); + i++; + c->reg.tmp0 = brw_vec4_grf(i, 0); + c->reg.tmp1 = brw_vec4_grf(i, 4); + i++; + } + + if (c->need_ff_sync) { + c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD); + i++; + } + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + struct brw_instruction *is_rev; + + /* Initial list of indices for incoming vertexes: + */ + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_TRISTRIP_REVERSE)); + + /* XXX: Is there an easier way to do this? Need to reverse every + * second tristrip element: Can ignore sometimes? + */ + is_rev = brw_IF(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[1]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[0]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(-1)); + } + is_rev = brw_ELSE(p, is_rev); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[0]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[1]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(1)); + } + brw_ENDIF(p, is_rev); + + brw_MOV(p, get_element(c->reg.inlist, 2), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0)); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(3)); +} + + + +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_copy_colors(c, 1, 0); + brw_clip_copy_colors(c, 2, 0); + } + is_poly = brw_ELSE(p, is_poly); + { + brw_clip_copy_colors(c, 0, 2); + brw_clip_copy_colors(c, 1, 2); + } + brw_ENDIF(p, is_poly); +} + + + +/* Use mesa's clipping algorithms, translated to GEN4 assembly. + */ +void brw_clip_tri( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_indirect vtx = brw_indirect(0, 0); + struct brw_indirect vtxPrev = brw_indirect(1, 0); + struct brw_indirect vtxOut = brw_indirect(2, 0); + struct brw_indirect plane_ptr = brw_indirect(3, 0); + struct brw_indirect inlist_ptr = brw_indirect(4, 0); + struct brw_indirect outlist_ptr = brw_indirect(5, 0); + struct brw_indirect freelist_ptr = brw_indirect(6, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *vertex_loop; + struct brw_instruction *next_test; + struct brw_instruction *prev_test; + + brw_MOV(p, get_addr_reg(vtxPrev), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + + brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) ); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + /* vtxOut = freelist_ptr++ + */ + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(freelist_ptr) ); + brw_ADD(p, get_addr_reg(freelist_ptr), get_addr_reg(freelist_ptr), brw_imm_uw(c->nr_regs * REG_SIZE)); + + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0)); + + vertex_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* vtx = *input_ptr; + */ + brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0)); + + /* IS_NEGATIVE(prev) */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset_hpos), c->reg.plane_equation); + prev_test = brw_IF(p, BRW_EXECUTE_1); + { + /* IS_POSITIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_GE); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset_hpos), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + + /* Coming back in. + */ + brw_ADD(p, c->reg.t, c->reg.dpPrev, negate(c->reg.dp)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dpPrev); + + /* If (vtxOut == 0) vtxOut = vtxPrev + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, GL_FALSE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + + } + prev_test = brw_ELSE(p, prev_test); + { + /* *outlist_ptr++ = vtxPrev; + * nr_verts++; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxPrev)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + + /* IS_NEGATIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset_hpos), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + /* Going out of bounds. Avoid division by zero as we + * know dp != dpPrev from DIFFERENT_SIGNS, above. + */ + brw_ADD(p, c->reg.t, c->reg.dp, negate(c->reg.dpPrev)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp); + + /* If (vtxOut == 0) vtxOut = vtx + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, GL_TRUE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + } + brw_ENDIF(p, prev_test); + + /* vtxPrev = vtx; + * inlist_ptr++; + */ + brw_MOV(p, get_addr_reg(vtxPrev), get_addr_reg(vtx)); + brw_ADD(p, get_addr_reg(inlist_ptr), get_addr_reg(inlist_ptr), brw_imm_uw(sizeof(short))); + + /* while (--loopcount != 0) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, vertex_loop); + + /* vtxPrev = *(outlist_ptr-1) OR: outlist[nr_verts-1] + * inlist = outlist + * inlist_ptr = &inlist[0] + * outlist_ptr = &outlist[0] + */ + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_w(-2)); + brw_MOV(p, get_addr_reg(vtxPrev), deref_1uw(outlist_ptr, 0)); + brw_MOV(p, brw_vec8_grf(c->reg.inlist.nr, 0), brw_vec8_grf(c->reg.outlist.nr, 0)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* nr_verts >= 3 + */ + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + c->reg.nr_verts, + brw_imm_ud(3)); + + /* && (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); +} + + + +void brw_clip_tri_emit_polygon(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop, *if_insn; + + /* for (loopcount = nr_verts-2; loopcount > 0; loopcount--) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, + c->reg.loopcount, + c->reg.nr_verts, + brw_imm_d(-2)); + + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect vptr = brw_indirect(1, 0); + + brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + + brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END)); + } + brw_ENDIF(p, if_insn); +} + +static void do_clip_tri( struct brw_clip_compile *c ) +{ + brw_clip_init_planes(c); + + brw_clip_tri(c); +} + + +static void maybe_do_clip_tri( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + do_clip_tri(c); + } + brw_ENDIF(p, do_clip); +} + +static void brw_clip_test( struct brw_clip_compile *c ) +{ + struct brw_reg t = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t2 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t3 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + + struct brw_reg v0 = get_tmp(c); + struct brw_reg v1 = get_tmp(c); + struct brw_reg v2 = get_tmp(c); + + struct brw_indirect vt0 = brw_indirect(0, 0); + struct brw_indirect vt1 = brw_indirect(1, 0); + struct brw_indirect vt2 = brw_indirect(2, 0); + + struct brw_compile *p = &c->func; + struct brw_instruction *is_outside; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + + brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2])); + brw_MOV(p, v0, deref_4f(vt0, c->offset_hpos)); + brw_MOV(p, v1, deref_4f(vt1, c->offset_hpos)); + brw_MOV(p, v2, deref_4f(vt2, c->offset_hpos)); + brw_AND(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(~0x3f)); + + /* test nearz, xmin, ymin plane */ + /* clip.xyz < -clip.w */ + brw_CMP(p, t1, BRW_CONDITIONAL_L, v0, negate(get_element(v0, 3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_L, v1, negate(get_element(v1, 3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_L, v2, negate(get_element(v2, 3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* All vertices are outside of a plane, rejected */ + brw_AND(p, t, t1, t2); + brw_AND(p, t, t, t3); + brw_OR(p, tmp0, get_element(t, 0), get_element(t, 1)); + brw_OR(p, tmp0, tmp0, get_element(t, 2)); + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), tmp0, brw_imm_ud(0x1)); + is_outside = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, is_outside); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* some vertices are inside a plane, some are outside,need to clip */ + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + brw_AND(p, t, t, brw_imm_ud(0x1)); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<5))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* test farz, xmax, ymax plane */ + /* clip.xyz > clip.w */ + brw_CMP(p, t1, BRW_CONDITIONAL_G, v0, get_element(v0, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_G, v1, get_element(v1, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_G, v2, get_element(v2, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* All vertices are outside of a plane, rejected */ + brw_AND(p, t, t1, t2); + brw_AND(p, t, t, t3); + brw_OR(p, tmp0, get_element(t, 0), get_element(t, 1)); + brw_OR(p, tmp0, tmp0, get_element(t, 2)); + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), tmp0, brw_imm_ud(0x1)); + is_outside = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, is_outside); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* some vertices are inside a plane, some are outside,need to clip */ + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + brw_AND(p, t, t, brw_imm_ud(0x1)); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<4))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<2))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<0))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + release_tmps(c); +} + + +void brw_emit_tri_clip( struct brw_clip_compile *c ) +{ + struct brw_instruction *neg_rhw; + struct brw_compile *p = &c->func; + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + brw_clip_init_clipmask(c); + brw_clip_init_ff_sync(c); + + /* if -ve rhw workaround bit is set, + do cliptest */ + if (c->chipset.is_965) { + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + neg_rhw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_test(c); + } + brw_ENDIF(p, neg_rhw); + } + /* Can't push into do_clip_tri because with polygon (or quad) + * flatshading, need to apply the flatshade here because we don't + * respect the PV when converting to trifan for emit: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + if ((c->key.clip_mode == BRW_CLIPMODE_NORMAL) || + (c->key.clip_mode == BRW_CLIPMODE_KERNEL_CLIP)) + do_clip_tri(c); + else + maybe_do_clip_tri(c); + + brw_clip_tri_emit_polygon(c); + + /* Send an empty message to kill the thread: + */ + brw_clip_kill_thread(c); +} diff --git a/src/gallium/drivers/i965/brw_clip_unfilled.c b/src/gallium/drivers/i965/brw_clip_unfilled.c new file mode 100644 index 00000000000..aec835b8cec --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_unfilled.c @@ -0,0 +1,497 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + +/* This is performed against the original triangles, so no indirection + * required: +BZZZT! + */ +static void compute_tri_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg e = c->reg.tmp0; + struct brw_reg f = c->reg.tmp1; + struct brw_reg v0 = byte_offset(c->reg.vertex[0], c->offset_hpos); + struct brw_reg v1 = byte_offset(c->reg.vertex[1], c->offset_hpos); + struct brw_reg v2 = byte_offset(c->reg.vertex[2], c->offset_hpos); + + + struct brw_reg v0n = get_tmp(c); + struct brw_reg v1n = get_tmp(c); + struct brw_reg v2n = get_tmp(c); + + /* Convert to NDC. + * NOTE: We can't modify the original vertex coordinates, + * as it may impact further operations. + * So, we have to keep normalized coordinates in temp registers. + * + * TBD-KC + * Try to optimize unnecessary MOV's. + */ + brw_MOV(p, v0n, v0); + brw_MOV(p, v1n, v1); + brw_MOV(p, v2n, v2); + + brw_clip_project_position(c, v0n); + brw_clip_project_position(c, v1n); + brw_clip_project_position(c, v2n); + + /* Calculate the vectors of two edges of the triangle: + */ + brw_ADD(p, e, v0n, negate(v2n)); + brw_ADD(p, f, v1n, negate(v2n)); + + /* Take their crossproduct: + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, vec4(brw_null_reg()), brw_swizzle(e, 1,2,0,3), brw_swizzle(f,2,0,1,3)); + brw_MAC(p, vec4(e), negate(brw_swizzle(e, 2,0,1,3)), brw_swizzle(f,1,2,0,3)); + brw_set_access_mode(p, BRW_ALIGN_1); + + brw_MUL(p, c->reg.dir, c->reg.dir, vec4(e)); +} + + +static void cull_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + GLuint conditional; + + assert (!(c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL)); + + if (c->key.fill_ccw == CLIP_CULL) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, ccw); +} + + + +static void copy_bfc( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + GLuint conditional; + + /* Do we have any colors to copy? + */ + if ((c->offset_color0 == 0 || c->offset_bfc0 == 0) && + (c->offset_color1 == 0 || c->offset_bfc1 == 0)) + return; + + /* In some wierd degnerate cases we can end up testing the + * direction twice, once for culling and once for bfc copying. Oh + * well, that's what you get for setting wierd GL state. + */ + if (c->key.copy_bfc_ccw) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + GLuint i; + + for (i = 0; i < 3; i++) { + if (c->offset_color0 && c->offset_bfc0) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset_color0), + byte_offset(c->reg.vertex[i], c->offset_bfc0)); + + if (c->offset_color1 && c->offset_bfc1) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset_color0), + byte_offset(c->reg.vertex[i], c->offset_bfc0)); + } + } + brw_ENDIF(p, ccw); +} + + + + +/* + GLfloat iz = 1.0 / dir.z; + GLfloat ac = dir.x * iz; + GLfloat bc = dir.y * iz; + offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE; + offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor; + offset *= MRD; +*/ +static void compute_offset( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg off = c->reg.offset; + struct brw_reg dir = c->reg.dir; + + brw_math_invert(p, get_element(off, 2), get_element(dir, 2)); + brw_MUL(p, vec2(off), dir, get_element(off, 2)); + + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + brw_abs(get_element(off, 0)), + brw_abs(get_element(off, 1))); + + brw_SEL(p, vec1(off), brw_abs(get_element(off, 0)), brw_abs(get_element(off, 1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_MUL(p, vec1(off), off, brw_imm_f(c->key.offset_factor)); + brw_ADD(p, vec1(off), off, brw_imm_f(c->key.offset_units)); +} + + +static void merge_edgeflags( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = get_element_ud(c->reg.tmp0, 0); + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + /* Get away with using reg.vertex because we know that this is not + * a _3DPRIM_TRISTRIP_REVERSE: + */ + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<8)); + brw_MOV(p, byte_offset(c->reg.vertex[0], c->offset_edgeflag), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<9)); + brw_MOV(p, byte_offset(c->reg.vertex[2], c->offset_edgeflag), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + brw_ENDIF(p, is_poly); +} + + + +static void apply_one_offset( struct brw_clip_compile *c, + struct brw_indirect vert ) +{ + struct brw_compile *p = &c->func; + struct brw_reg z = deref_1f(vert, c->header_position_offset + + 2 * type_sz(BRW_REGISTER_TYPE_F)); + + brw_ADD(p, z, z, vec1(c->reg.offset)); +} + + + +/*********************************************************************** + * Output clipped polygon as an unfilled primitive: + */ +static void emit_lines(struct brw_clip_compile *c, + GLboolean do_offset) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_edge; + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v1 = brw_indirect(1, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + struct brw_indirect v1ptr = brw_indirect(3, 0); + + /* Need a seperate loop for offset: + */ + if (do_offset) { + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + apply_one_offset(c, v0); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + } + + /* v1ptr = &inlist[nr_verts] + * *v1ptr = v0 + */ + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v0ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v1ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_MOV(p, deref_1uw(v1ptr, 0), deref_1uw(v0ptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_MOV(p, get_addr_reg(v1), deref_1uw(v0ptr, 2)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw edge if edgeflag != 0 */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset_edgeflag), + brw_imm_f(0)); + draw_edge = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, v1, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, draw_edge); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + +static void emit_points(struct brw_clip_compile *c, + GLboolean do_offset ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_point; + + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw if edgeflag != 0 + */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset_edgeflag), + brw_imm_f(0)); + draw_point = brw_IF(p, BRW_EXECUTE_1); + { + if (do_offset) + apply_one_offset(c, v0); + + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END); + } + brw_ENDIF(p, draw_point); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + + + + + +static void emit_primitives( struct brw_clip_compile *c, + GLuint mode, + GLboolean do_offset ) +{ + switch (mode) { + case CLIP_FILL: + brw_clip_tri_emit_polygon(c); + break; + + case CLIP_LINE: + emit_lines(c, do_offset); + break; + + case CLIP_POINT: + emit_points(c, do_offset); + break; + + case CLIP_CULL: + assert(0); + break; + } +} + + + +static void emit_unfilled_primitives( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + + /* Direction culling has already been done. + */ + if (c->key.fill_ccw != c->key.fill_cw && + c->key.fill_ccw != CLIP_CULL && + c->key.fill_cw != CLIP_CULL) + { + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } + ccw = brw_ELSE(p, ccw); + { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + brw_ENDIF(p, ccw); + } + else if (c->key.fill_cw != CLIP_CULL) { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + else if (c->key.fill_ccw != CLIP_CULL) { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } +} + + + + +static void check_nr_verts( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.nr_verts, brw_imm_d(3)); + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, if_insn); +} + + +void brw_emit_unfilled_clip( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + + c->need_direction = ((c->key.offset_ccw || c->key.offset_cw) || + (c->key.fill_ccw != c->key.fill_cw) || + c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL || + c->key.copy_bfc_cw || + c->key.copy_bfc_ccw); + + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + brw_clip_init_ff_sync(c); + + assert(c->offset_edgeflag); + + if (c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL) { + brw_clip_kill_thread(c); + return; + } + + merge_edgeflags(c); + + /* Need to use the inlist indirection here: + */ + if (c->need_direction) + compute_tri_direction(c); + + if (c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL) + cull_direction(c); + + if (c->key.offset_ccw || + c->key.offset_cw) + compute_offset(c); + + if (c->key.copy_bfc_ccw || + c->key.copy_bfc_cw) + copy_bfc(c); + + /* Need to do this whether we clip or not: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + brw_clip_init_clipmask(c); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_init_planes(c); + brw_clip_tri(c); + check_nr_verts(c); + } + brw_ENDIF(p, do_clip); + + emit_unfilled_primitives(c); + brw_clip_kill_thread(c); +} + + + diff --git a/src/gallium/drivers/i965/brw_clip_util.c b/src/gallium/drivers/i965/brw_clip_util.c new file mode 100644 index 00000000000..97a57103105 --- /dev/null +++ b/src/gallium/drivers/i965/brw_clip_util.c @@ -0,0 +1,388 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + + +struct brw_reg get_tmp( struct brw_clip_compile *c ) +{ + struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + + +static struct brw_reg make_plane_ud(GLuint x, GLuint y, GLuint z, GLuint w) +{ + return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x); +} + + +void brw_clip_init_planes( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + + if (!c->key.nr_userclip) { + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0, 0, 0xff, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0, 0, 1, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0, 1, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff, 0, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1, 0, 0, 1)); + } +} + + + +#define W 3 + +/* Project 'pos' to screen space (or back again), overwrite with results: + */ +void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos ) +{ + struct brw_compile *p = &c->func; + + /* calc rhw + */ + brw_math_invert(p, get_element(pos, W), get_element(pos, W)); + + /* value.xyz *= value.rhw + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, brw_writemask(pos, BRW_WRITEMASK_XYZ), pos, brw_swizzle1(pos, W)); + brw_set_access_mode(p, BRW_ALIGN_1); +} + + +static void brw_clip_project_vertex( struct brw_clip_compile *c, + struct brw_indirect vert_addr ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + + /* Fixup position. Extract from the original vertex and re-project + * to screen space: + */ + brw_MOV(p, tmp, deref_4f(vert_addr, c->offset_hpos)); + brw_clip_project_position(c, tmp); + brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp); + + release_tmp(c, tmp); +} + + + + +/* Interpolate between two vertices and put the result into a0.0. + * Increment a0.0 accordingly. + */ +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + GLboolean force_edgeflag) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + GLuint i; + + /* Just copy the vertex header: + */ + /* + * After CLIP stage, only first 256 bits of the VUE are read + * back on IGDNG, so needn't change it + */ + brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1); + + /* Iterate over each attribute (could be done in pairs?) + */ + for (i = 0; i < c->key.nr_attrs; i++) { + GLuint delta = i*16 + 32; + + if (c->chipset.is_igdng) + delta = i * 16 + 32 * 3; + + if (delta == c->offset_edgeflag) { + if (force_edgeflag) + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1)); + else + brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta)); + } + else { + /* Interpolate: + * + * New = attr0 + t*attr1 - t*attr0 + */ + brw_MUL(p, + vec4(brw_null_reg()), + deref_4f(v1_ptr, delta), + t0); + + brw_MAC(p, + tmp, + negate(deref_4f(v0_ptr, delta)), + t0); + + brw_ADD(p, + deref_4f(dest_ptr, delta), + deref_4f(v0_ptr, delta), + tmp); + } + } + + if (i & 1) { + GLuint delta = i*16 + 32; + + if (c->chipset.is_igdng) + delta = i * 16 + 32 * 3; + + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0)); + } + + release_tmp(c, tmp); + + /* Recreate the projected (NDC) coordinate in the new vertex + * header: + */ + brw_clip_project_vertex(c, dest_ptr ); +} + + + + +#define MAX_MRF 16 + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + GLboolean allocate, + GLboolean eot, + GLuint header) +{ + struct brw_compile *p = &c->func; + GLuint start = c->last_mrf; + + brw_clip_ff_sync(c); + + assert(!(allocate && eot)); + + /* Cycle through mrf regs - probably futile as we have to wait for + * the allocation response anyway. Also, the order this function + * is invoked doesn't correspond to the order the instructions will + * be executed, so it won't have any effect in many cases. + */ +#if 0 + if (start + c->nr_regs + 1 >= MAX_MRF) + start = 0; + + c->last_mrf = start + c->nr_regs + 1; +#endif + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs); + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + + /* Send each vertex as a seperate write to the urb. This + * is different to the concept in brw_sf_emit.c, where + * subsequent writes are used to build up a single urb + * entry. Each of these writes instantiates a seperate + * urb entry - (I think... what about 'allocate'?) + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + start, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response_length */ + eot, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + + +void brw_clip_kill_thread(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + + brw_clip_ff_sync(c); + /* Send an empty message to kill the thread and release any + * allocated urb entry: + */ + brw_urb_WRITE(p, + retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + 0, /* allocate */ + 0, /* used */ + 1, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, + BRW_URB_SWIZZLE_NONE); +} + + + + +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ) +{ + return brw_address(c->reg.fixed_planes); +} + + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ) +{ + if (c->key.nr_userclip) { + return brw_imm_uw(16); + } + else { + return brw_imm_uw(4); + } +} + + +/* If flatshading, distribute color from provoking vertex prior to + * clipping. + */ +void brw_clip_copy_colors( struct brw_clip_compile *c, + GLuint to, GLuint from ) +{ + struct brw_compile *p = &c->func; + + if (c->offset_color0) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset_color0), + byte_offset(c->reg.vertex[from], c->offset_color0)); + + if (c->offset_color1) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset_color1), + byte_offset(c->reg.vertex[from], c->offset_color1)); + + if (c->offset_bfc0) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset_bfc0), + byte_offset(c->reg.vertex[from], c->offset_bfc0)); + + if (c->offset_bfc1) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset_bfc1), + byte_offset(c->reg.vertex[from], c->offset_bfc1)); +} + + + +void brw_clip_init_clipmask( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg incoming = get_element_ud(c->reg.R0, 2); + + /* Shift so that lowest outcode bit is rightmost: + */ + brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26)); + + if (c->key.nr_userclip) { + struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD); + + /* Rearrange userclip outcodes so that they come directly after + * the fixed plane bits. + */ + brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14)); + brw_SHR(p, tmp, tmp, brw_imm_ud(8)); + brw_OR(p, c->reg.planemask, c->reg.planemask, tmp); + + release_tmp(c, tmp); + } +} + +void brw_clip_ff_sync(struct brw_clip_compile *c) +{ + if (c->need_ff_sync) { + struct brw_compile *p = &c->func; + struct brw_instruction *need_ff_sync; + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, brw_null_reg(), c->reg.ff_sync, brw_imm_ud(0x1)); + need_ff_sync = brw_IF(p, BRW_EXECUTE_1); + { + brw_OR(p, c->reg.ff_sync, c->reg.ff_sync, brw_imm_ud(0x1)); + brw_ff_sync(p, + c->reg.R0, + 0, + c->reg.R0, + 1, + 1, /* used */ + 1, /* msg length */ + 1, /* response length */ + 0, /* eot */ + 1, /* write compelete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); + } + brw_ENDIF(p, need_ff_sync); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } +} + +void brw_clip_init_ff_sync(struct brw_clip_compile *c) +{ + if (c->need_ff_sync) { + struct brw_compile *p = &c->func; + + brw_MOV(p, c->reg.ff_sync, brw_imm_ud(0)); + } +} diff --git a/src/gallium/drivers/i965/brw_context.c b/src/gallium/drivers/i965/brw_context.c new file mode 100644 index 00000000000..e67551882dc --- /dev/null +++ b/src/gallium/drivers/i965/brw_context.c @@ -0,0 +1,154 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "pipe/p_context.h" +#include "util/u_simple_list.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_draw.h" +#include "brw_state.h" +#include "brw_batchbuffer.h" +#include "brw_winsys.h" +#include "brw_screen.h" + + +static void brw_destroy_context( struct pipe_context *pipe ) +{ + struct brw_context *brw = brw_context(pipe); + int i; + + brw_context_flush( brw ); + brw_batchbuffer_free( brw->batch ); + brw_destroy_state(brw); + + brw_draw_cleanup( brw ); + + brw_pipe_blend_cleanup( brw ); + brw_pipe_depth_stencil_cleanup( brw ); + brw_pipe_framebuffer_cleanup( brw ); + brw_pipe_flush_cleanup( brw ); + brw_pipe_misc_cleanup( brw ); + brw_pipe_query_cleanup( brw ); + brw_pipe_rast_cleanup( brw ); + brw_pipe_sampler_cleanup( brw ); + brw_pipe_shader_cleanup( brw ); + brw_pipe_vertex_cleanup( brw ); + brw_pipe_clear_cleanup( brw ); + + brw_hw_cc_cleanup( brw ); + + + FREE(brw->wm.compile_data); + + for (i = 0; i < brw->curr.fb.nr_cbufs; i++) + pipe_surface_reference(&brw->curr.fb.cbufs[i], NULL); + brw->curr.fb.nr_cbufs = 0; + pipe_surface_reference(&brw->curr.fb.zsbuf, NULL); + + bo_reference(&brw->curbe.curbe_bo, NULL); + bo_reference(&brw->vs.prog_bo, NULL); + bo_reference(&brw->vs.state_bo, NULL); + bo_reference(&brw->vs.bind_bo, NULL); + bo_reference(&brw->gs.prog_bo, NULL); + bo_reference(&brw->gs.state_bo, NULL); + bo_reference(&brw->clip.prog_bo, NULL); + bo_reference(&brw->clip.state_bo, NULL); + bo_reference(&brw->clip.vp_bo, NULL); + bo_reference(&brw->sf.prog_bo, NULL); + bo_reference(&brw->sf.state_bo, NULL); + bo_reference(&brw->sf.vp_bo, NULL); + + for (i = 0; i < Elements(brw->wm.sdc_bo); i++) + bo_reference(&brw->wm.sdc_bo[i], NULL); + + bo_reference(&brw->wm.bind_bo, NULL); + + for (i = 0; i < Elements(brw->wm.surf_bo); i++) + bo_reference(&brw->wm.surf_bo[i], NULL); + + bo_reference(&brw->wm.sampler_bo, NULL); + bo_reference(&brw->wm.prog_bo, NULL); + bo_reference(&brw->wm.state_bo, NULL); +} + + +struct pipe_context *brw_create_context(struct pipe_screen *screen) +{ + struct brw_context *brw = (struct brw_context *) CALLOC_STRUCT(brw_context); + + if (!brw) { + debug_printf("%s: failed to alloc context\n", __FUNCTION__); + return NULL; + } + + brw->base.screen = screen; + brw->base.destroy = brw_destroy_context; + brw->sws = brw_screen(screen)->sws; + brw->chipset = brw_screen(screen)->chipset; + + brw_pipe_blend_init( brw ); + brw_pipe_depth_stencil_init( brw ); + brw_pipe_framebuffer_init( brw ); + brw_pipe_flush_init( brw ); + brw_pipe_misc_init( brw ); + brw_pipe_query_init( brw ); + brw_pipe_rast_init( brw ); + brw_pipe_sampler_init( brw ); + brw_pipe_shader_init( brw ); + brw_pipe_vertex_init( brw ); + brw_pipe_clear_init( brw ); + + brw_hw_cc_init( brw ); + + brw_init_state( brw ); + brw_draw_init( brw ); + + brw->state.dirty.mesa = ~0; + brw->state.dirty.brw = ~0; + + brw->flags.always_emit_state = 0; + + make_empty_list(&brw->query.active_head); + + brw->batch = brw_batchbuffer_alloc( brw->sws, brw->chipset ); + if (brw->batch == NULL) + goto fail; + + return &brw->base; + +fail: + if (brw->batch) + brw_batchbuffer_free( brw->batch ); + return NULL; +} + diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h new file mode 100644 index 00000000000..56e78074000 --- /dev/null +++ b/src/gallium/drivers/i965/brw_context.h @@ -0,0 +1,853 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + +#include "brw_structs.h" +#include "brw_winsys.h" +#include "brw_reg.h" +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "tgsi/tgsi_scan.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawning a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contiguous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes decisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_MAX_CURBE (32*16) + +struct brw_context; + +struct brw_depth_stencil_state { + /* Precalculated hardware state: + */ + struct brw_cc0 cc0; + struct brw_cc1 cc1; + struct brw_cc2 cc2; + struct brw_cc3 cc3; + struct brw_cc7 cc7; + + unsigned iz_lookup; +}; + + +struct brw_blend_state { + /* Precalculated hardware state: + */ + struct brw_cc2 cc2; + struct brw_cc3 cc3; + struct brw_cc5 cc5; + struct brw_cc6 cc6; + + struct brw_surf_ss0 ss0; +}; + + +struct brw_rasterizer_state; + +struct brw_immediate_data { + unsigned nr; + float (*data)[4]; +}; + +struct brw_vertex_shader { + const struct tgsi_token *tokens; + struct brw_winsys_buffer *const_buffer; /** Program constant buffer/surface */ + + struct tgsi_shader_info info; + struct brw_immediate_data immediates; + + GLuint has_flow_control:1; + GLuint use_const_buffer:1; + + /* Offsets of special vertex shader outputs required for clipping. + */ + GLuint output_hpos:6; /* not always zero? */ + GLuint output_color0:6; + GLuint output_color1:6; + GLuint output_bfc0:6; + GLuint output_bfc1:6; + GLuint output_edgeflag:6; + + unsigned id; +}; + +struct brw_fs_signature { + GLuint nr_inputs; + struct { + GLuint interp:3; /* TGSI_INTERPOLATE_x */ + GLuint semantic:5; /* TGSI_SEMANTIC_x */ + GLuint semantic_index:24; + } input[PIPE_MAX_SHADER_INPUTS]; +}; + +#define brw_fs_signature_size(s) (offsetof(struct brw_fs_signature, input) + \ + ((s)->nr_inputs * sizeof (s)->input[0])) + + +struct brw_fragment_shader { + const struct tgsi_token *tokens; + struct tgsi_shader_info info; + + struct brw_fs_signature signature; + struct brw_immediate_data immediates; + + unsigned iz_lookup; + /*unsigned wm_lookup;*/ + + unsigned uses_depth:1; + unsigned has_flow_control:1; + + unsigned id; + struct brw_winsys_buffer *const_buffer; /** Program constant buffer/surface */ + GLboolean use_const_buffer; +}; + + +struct brw_sampler { + struct brw_ss0 ss0; + struct brw_ss1 ss1; + float border_color[4]; + struct brw_ss3 ss3; +}; + + + +#define PIPE_NEW_DEPTH_STENCIL_ALPHA 0x1 +#define PIPE_NEW_RAST 0x2 +#define PIPE_NEW_BLEND 0x4 +#define PIPE_NEW_VIEWPORT 0x8 +#define PIPE_NEW_SAMPLERS 0x10 +#define PIPE_NEW_VERTEX_BUFFER 0x20 +#define PIPE_NEW_VERTEX_ELEMENT 0x40 +#define PIPE_NEW_FRAGMENT_SHADER 0x80 +#define PIPE_NEW_VERTEX_SHADER 0x100 +#define PIPE_NEW_FRAGMENT_CONSTANTS 0x200 +#define PIPE_NEW_VERTEX_CONSTANTS 0x400 +#define PIPE_NEW_CLIP 0x800 +#define PIPE_NEW_INDEX_BUFFER 0x1000 +#define PIPE_NEW_INDEX_RANGE 0x2000 +#define PIPE_NEW_BLEND_COLOR 0x4000 +#define PIPE_NEW_POLYGON_STIPPLE 0x8000 +#define PIPE_NEW_FRAMEBUFFER_DIMENSIONS 0x10000 +#define PIPE_NEW_DEPTH_BUFFER 0x20000 +#define PIPE_NEW_COLOR_BUFFERS 0x40000 +#define PIPE_NEW_QUERY 0x80000 +#define PIPE_NEW_SCISSOR 0x100000 +#define PIPE_NEW_BOUND_TEXTURES 0x200000 +#define PIPE_NEW_NR_CBUFS 0x400000 +#define PIPE_NEW_FRAGMENT_SIGNATURE 0x800000 + + + +#define BRW_NEW_URB_FENCE 0x1 +#define BRW_NEW_FRAGMENT_PROGRAM 0x2 +#define BRW_NEW_VERTEX_PROGRAM 0x4 +#define BRW_NEW_INPUT_DIMENSIONS 0x8 +#define BRW_NEW_CURBE_OFFSETS 0x10 +#define BRW_NEW_REDUCED_PRIMITIVE 0x20 +#define BRW_NEW_PRIMITIVE 0x40 +#define BRW_NEW_CONTEXT 0x80 +#define BRW_NEW_WM_INPUT_DIMENSIONS 0x100 +#define BRW_NEW_PSP 0x800 +#define BRW_NEW_WM_SURFACES 0x1000 +#define BRW_NEW_xxx 0x2000 /* was FENCE */ +#define BRW_NEW_INDICES 0x4000 + +/** + * Used for any batch entry with a relocated pointer that will be used + * by any 3D rendering. Need to re-emit these fresh in each + * batchbuffer as the referenced buffers may be relocated in the + * meantime. + */ +#define BRW_NEW_BATCH 0x10000 +#define BRW_NEW_NR_WM_SURFACES 0x40000 +#define BRW_NEW_NR_VS_SURFACES 0x80000 +#define BRW_NEW_INDEX_BUFFER 0x100000 + +struct brw_state_flags { + /** State update flags signalled by mesa internals */ + GLuint mesa; + /** + * State update flags signalled as the result of brw_tracked_state updates + */ + GLuint brw; + /** State update flags signalled by brw_state_cache.c searches */ + GLuint cache; +}; + + + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +struct brw_wm_prog_data { + GLuint curb_read_length; + GLuint urb_read_length; + + GLuint first_curbe_grf; + GLuint total_grf; + GLuint total_scratch; + + GLuint nr_params; /**< number of float params/constants */ + GLboolean error; + + /* Pointer to tracked values (only valid once + * _mesa_load_state_parameters has been called at runtime). + */ + const GLfloat *param[BRW_MAX_CURBE]; +}; + +struct brw_sf_prog_data { + GLuint urb_read_length; + GLuint total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + GLuint urb_entry_size; +}; + + +struct brw_clip_prog_data; + +struct brw_gs_prog_data { + GLuint urb_read_length; + GLuint total_grf; +}; + +struct brw_vs_prog_data { + GLuint curb_read_length; + GLuint urb_read_length; + GLuint total_grf; + + GLuint nr_outputs; + GLuint nr_inputs; + + GLuint nr_params; /**< number of TGSI_FILE_CONSTANT's */ + + GLuint output_edgeflag; + + GLboolean writes_psiz; + + /* Used for calculating urb partitions: + */ + GLuint urb_entry_size; +}; + + +/* Size == 0 if output either not written, or always [0,0,0,1] + */ +struct brw_vs_ouput_sizes { + GLubyte output_size[PIPE_MAX_SHADER_OUTPUTS]; +}; + + +/** Number of texture sampler units */ +#define BRW_MAX_TEX_UNIT 16 + +/** Max number of render targets in a shader */ +#define BRW_MAX_DRAW_BUFFERS 4 + +/** + * Size of our surface binding table for the WM. + * This contains pointers to the drawing surfaces and current texture + * objects and shader constant buffers (+2). + */ +#define BRW_WM_MAX_SURF (BRW_MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1) + +/** + * Helpers to convert drawing buffers, textures and constant buffers + * to surface binding table indexes, for WM. + */ +#define BTI_COLOR_BUF(d) (d) +#define BTI_FRAGMENT_CONSTANTS (BRW_MAX_DRAW_BUFFERS) +#define BTI_TEXTURE(t) (BRW_MAX_DRAW_BUFFERS + 1 + (t)) + +/** + * Size of surface binding table for the VS. + * Only one constant buffer for now. + */ +#define BRW_VS_MAX_SURF 1 + +/** + * Only a VS constant buffer + */ +#define SURF_INDEX_VERT_CONST_BUFFER 0 + + +/* Bit of a hack to align these with the winsys buffer_data_type enum. + */ +enum brw_cache_id { + BRW_CC_VP = BRW_DATA_GS_CC_VP, + BRW_CC_UNIT = BRW_DATA_GS_CC_UNIT, + BRW_WM_PROG = BRW_DATA_GS_WM_PROG, + BRW_SAMPLER_DEFAULT_COLOR = BRW_DATA_GS_SAMPLER_DEFAULT_COLOR, + BRW_SAMPLER = BRW_DATA_GS_SAMPLER, + BRW_WM_UNIT = BRW_DATA_GS_WM_UNIT, + BRW_SF_PROG = BRW_DATA_GS_SF_PROG, + BRW_SF_VP = BRW_DATA_GS_SF_VP, + BRW_SF_UNIT = BRW_DATA_GS_SF_UNIT, + BRW_VS_UNIT = BRW_DATA_GS_VS_UNIT, + BRW_VS_PROG = BRW_DATA_GS_VS_PROG, + BRW_GS_UNIT = BRW_DATA_GS_GS_UNIT, + BRW_GS_PROG = BRW_DATA_GS_GS_PROG, + BRW_CLIP_VP = BRW_DATA_GS_CLIP_VP, + BRW_CLIP_UNIT = BRW_DATA_GS_CLIP_UNIT, + BRW_CLIP_PROG = BRW_DATA_GS_CLIP_PROG, + BRW_SS_SURFACE = BRW_DATA_SS_SURFACE, + BRW_SS_SURF_BIND = BRW_DATA_SS_SURF_BIND, + + BRW_MAX_CACHE +}; + +struct brw_cache_item { + /** + * Effectively part of the key, cache_id identifies what kind of state + * buffer is involved, and also which brw->state.dirty.cache flag should + * be set when this cache item is chosen. + */ + enum brw_cache_id cache_id; + /** 32-bit hash of the key data */ + GLuint hash; + GLuint key_size; /* for variable-sized keys */ + const void *key; + struct brw_winsys_reloc *relocs; + GLuint nr_relocs; + + struct brw_winsys_buffer *bo; + GLuint data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + struct brw_context *brw; + struct brw_winsys_screen *sws; + + struct brw_cache_item **items; + GLuint size, n_items; + + enum brw_buffer_type buffer_type; + + GLuint key_size[BRW_MAX_CACHE]; /* for fixed-size keys */ + GLuint aux_size[BRW_MAX_CACHE]; + char *name[BRW_MAX_CACHE]; + + + /* Record of the last BOs chosen for each cache_id. Used to set + * brw->state.dirty.cache when a new cache item is chosen. + */ + struct brw_winsys_buffer *last_bo[BRW_MAX_CACHE]; +}; + + +struct brw_tracked_state { + struct brw_state_flags dirty; + int (*prepare)( struct brw_context *brw ); + int (*emit)( struct brw_context *brw ); +}; + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1<<BRW_CC_VP) +#define CACHE_NEW_CC_UNIT (1<<BRW_CC_UNIT) +#define CACHE_NEW_WM_PROG (1<<BRW_WM_PROG) +#define CACHE_NEW_SAMPLER_DEFAULT_COLOR (1<<BRW_SAMPLER_DEFAULT_COLOR) +#define CACHE_NEW_SAMPLER (1<<BRW_SAMPLER) +#define CACHE_NEW_WM_UNIT (1<<BRW_WM_UNIT) +#define CACHE_NEW_SF_PROG (1<<BRW_SF_PROG) +#define CACHE_NEW_SF_VP (1<<BRW_SF_VP) +#define CACHE_NEW_SF_UNIT (1<<BRW_SF_UNIT) +#define CACHE_NEW_VS_UNIT (1<<BRW_VS_UNIT) +#define CACHE_NEW_VS_PROG (1<<BRW_VS_PROG) +#define CACHE_NEW_GS_UNIT (1<<BRW_GS_UNIT) +#define CACHE_NEW_GS_PROG (1<<BRW_GS_PROG) +#define CACHE_NEW_CLIP_VP (1<<BRW_CLIP_VP) +#define CACHE_NEW_CLIP_UNIT (1<<BRW_CLIP_UNIT) +#define CACHE_NEW_CLIP_PROG (1<<BRW_CLIP_PROG) +#define CACHE_NEW_SURFACE (1<<BRW_SS_SURFACE) +#define CACHE_NEW_SURF_BIND (1<<BRW_SS_SURF_BIND) + +struct brw_cached_batch_item { + struct header *header; + GLuint sz; + struct brw_cached_batch_item *next; +}; + + + +/* Protect against a future where VERT_ATTRIB_MAX > 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define VS_INPUT_BITMASK_DWORDS ((PIPE_MAX_SHADER_INPUTS+31)/32) + + + + +struct brw_vertex_info { + GLuint sizes[VS_INPUT_BITMASK_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */ +}; + + +struct brw_query_object { + /** Doubly linked list of active query objects in the context. */ + struct brw_query_object *prev, *next; + + /** Last query BO associated with this query. */ + struct brw_winsys_buffer *bo; + /** First index in bo with query data for this object. */ + int first_index; + /** Last index in bo with query data for this object. */ + int last_index; + + /* Total count of pixels from previous BOs */ + uint64_t result; +}; + +#define CC_RELOC_VP 0 + + +/** + * brw_context is derived from pipe_context + */ +struct brw_context +{ + struct pipe_context base; + struct brw_chipset chipset; + + struct brw_winsys_screen *sws; + + struct brw_batchbuffer *batch; + + GLuint primitive; + GLuint reduced_primitive; + + /* Active state from the state tracker: + */ + struct { + struct brw_vertex_shader *vertex_shader; + struct brw_fragment_shader *fragment_shader; + const struct brw_blend_state *blend; + const struct brw_rasterizer_state *rast; + const struct brw_depth_stencil_state *zstencil; + + const struct brw_sampler *sampler[PIPE_MAX_SAMPLERS]; + unsigned num_samplers; + + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + unsigned num_vertex_elements; + unsigned num_textures; + unsigned num_vertex_buffers; + + struct pipe_scissor_state scissor; + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state fb; + struct pipe_clip_state ucp; + struct pipe_buffer *vertex_constants; + struct pipe_buffer *fragment_constants; + + struct brw_blend_constant_color bcc; + struct brw_polygon_stipple bps; + struct brw_cc_viewport ccv; + + /** + * Index buffer for this draw_prims call. + * + * Updates are signaled by PIPE_NEW_INDEX_BUFFER. + */ + struct pipe_buffer *index_buffer; + unsigned index_size; + + /* Updates are signalled by PIPE_NEW_INDEX_RANGE: + */ + unsigned min_index; + unsigned max_index; + + } curr; + + struct { + struct brw_state_flags dirty; + + /** + * List of buffers accumulated in brw_validate_state to receive + * dri_bo_check_aperture treatment before exec, so we can know if we + * should flush the batch and try again before emitting primitives. + * + * This can be a fixed number as we only have a limited number of + * objects referenced from the batchbuffer in a primitive emit, + * consisting of the vertex buffers, pipelined state pointers, + * the CURBE, the depth buffer, and a query BO. + */ + struct brw_winsys_buffer *validated_bos[PIPE_MAX_SHADER_INPUTS + 16]; + int validated_bo_count; + } state; + + struct brw_cache cache; /** non-surface items */ + struct brw_cache surface_cache; /* surface items */ + struct brw_cached_batch_item *cached_batch_items; + + struct { + struct u_upload_mgr *upload_vertex; + struct u_upload_mgr *upload_index; + + /* Information on uploaded vertex buffers: + */ + struct { + unsigned stride; /* in bytes between successive vertices */ + unsigned offset; /* in bytes, of first vertex in bo */ + unsigned vertex_count; /* count of valid vertices which may be accessed */ + struct brw_winsys_buffer *bo; + } vb[PIPE_MAX_ATTRIBS]; + + unsigned nr_vb; /* currently the same as curr.num_vertex_buffers */ + } vb; + + struct { + /* Updates to these fields are signaled by BRW_NEW_INDEX_BUFFER. */ + struct brw_winsys_buffer *bo; + unsigned int offset; + unsigned int size; + /* Offset to index buffer index to use in CMD_3D_PRIM so that we can + * avoid re-uploading the IB packet over and over if we're actually + * referencing the same index buffer. + */ + unsigned int start_vertex_offset; + } ib; + + + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + GLuint vsize; /* vertex size plus header in urb registers */ + GLuint csize; /* constant buffer size in urb registers */ + GLuint sfsize; /* setup data size in urb registers */ + + GLboolean constrained; + + GLuint nr_vs_entries; + GLuint nr_gs_entries; + GLuint nr_clip_entries; + GLuint nr_sf_entries; + GLuint nr_cs_entries; + + GLuint vs_start; + GLuint gs_start; + GLuint clip_start; + GLuint sf_start; + GLuint cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + GLuint wm_start; /**< pos of first wm const in CURBE buffer */ + GLuint wm_size; /**< number of float[4] consts, multiple of 16 */ + GLuint clip_start; + GLuint clip_size; + GLuint vs_start; + GLuint vs_size; + GLuint total_size; + + struct brw_winsys_buffer *curbe_bo; + /** Offset within curbe_bo of space for current curbe entry */ + GLuint curbe_offset; + /** Offset within curbe_bo of space for next curbe entry */ + GLuint curbe_next_offset; + + GLfloat *last_buf; + GLuint last_bufsz; + /** + * Whether we should create a new bo instead of reusing the old one + * (if we just dispatch the batch pointing at the old one. + */ + GLboolean need_new_bo; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + struct brw_winsys_buffer *prog_bo; + struct brw_winsys_buffer *state_bo; + + /** Binding table of pointers to surf_bo entries */ + struct brw_winsys_buffer *bind_bo; + struct brw_winsys_buffer *surf_bo[BRW_VS_MAX_SURF]; + GLuint nr_surfaces; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + GLboolean prog_active; + struct brw_winsys_buffer *prog_bo; + struct brw_winsys_buffer *state_bo; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + struct brw_winsys_buffer *prog_bo; + struct brw_winsys_buffer *state_bo; + struct brw_winsys_buffer *vp_bo; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + struct brw_winsys_buffer *prog_bo; + struct brw_winsys_buffer *state_bo; + struct brw_winsys_buffer *vp_bo; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + struct brw_wm_compile *compile_data; + + /** Input sizes, calculated from active vertex program. + * One bit per fragment program input attribute. + */ + /*GLbitfield input_size_masks[4];*/ + + /** Array of surface default colors (texture border color) */ + struct brw_winsys_buffer *sdc_bo[BRW_MAX_TEX_UNIT]; + + GLuint render_surf; + GLuint nr_surfaces; + + GLuint max_threads; + struct brw_winsys_buffer *scratch_bo; + + GLuint sampler_count; + struct brw_winsys_buffer *sampler_bo; + + /** Binding table of pointers to surf_bo entries */ + struct brw_winsys_buffer *bind_bo; + struct brw_winsys_buffer *surf_bo[BRW_WM_MAX_SURF]; + + struct brw_winsys_buffer *prog_bo; + struct brw_winsys_buffer *state_bo; + } wm; + + + struct { + struct brw_winsys_buffer *state_bo; + + struct brw_cc_unit_state cc; + struct brw_winsys_reloc reloc[1]; + } cc; + + struct { + struct brw_query_object active_head; + struct brw_winsys_buffer *bo; + int index; + GLboolean active; + int stats_wm; + } query; + + struct { + unsigned always_emit_state:1; + unsigned always_flush_batch:1; + unsigned force_swtnl:1; + unsigned no_swtnl:1; + } flags; + + /* Used to give every program string a unique id + */ + GLuint program_id; +}; + + + +/*====================================================================== + * brw_queryobj.c + */ +void brw_init_query(struct brw_context *brw); +enum pipe_error brw_prepare_query_begin(struct brw_context *brw); +void brw_emit_query_begin(struct brw_context *brw); +void brw_emit_query_end(struct brw_context *brw); + +/*====================================================================== + * brw_state_dump.c + */ +void brw_debug_batch(struct brw_context *intel); + + +/*====================================================================== + * brw_pipe_*.c + */ +void brw_pipe_blend_init( struct brw_context *brw ); +void brw_pipe_depth_stencil_init( struct brw_context *brw ); +void brw_pipe_framebuffer_init( struct brw_context *brw ); +void brw_pipe_flush_init( struct brw_context *brw ); +void brw_pipe_misc_init( struct brw_context *brw ); +void brw_pipe_query_init( struct brw_context *brw ); +void brw_pipe_rast_init( struct brw_context *brw ); +void brw_pipe_sampler_init( struct brw_context *brw ); +void brw_pipe_shader_init( struct brw_context *brw ); +void brw_pipe_vertex_init( struct brw_context *brw ); +void brw_pipe_clear_init( struct brw_context *brw ); + + +void brw_pipe_blend_cleanup( struct brw_context *brw ); +void brw_pipe_depth_stencil_cleanup( struct brw_context *brw ); +void brw_pipe_framebuffer_cleanup( struct brw_context *brw ); +void brw_pipe_flush_cleanup( struct brw_context *brw ); +void brw_pipe_misc_cleanup( struct brw_context *brw ); +void brw_pipe_query_cleanup( struct brw_context *brw ); +void brw_pipe_rast_cleanup( struct brw_context *brw ); +void brw_pipe_sampler_cleanup( struct brw_context *brw ); +void brw_pipe_shader_cleanup( struct brw_context *brw ); +void brw_pipe_vertex_cleanup( struct brw_context *brw ); +void brw_pipe_clear_cleanup( struct brw_context *brw ); + +void brw_hw_cc_init( struct brw_context *brw ); +void brw_hw_cc_cleanup( struct brw_context *brw ); + + + +void brw_context_flush( struct brw_context *brw ); + + +/* brw_urb.c + */ +int brw_upload_urb_fence(struct brw_context *brw); + +/* brw_curbe.c + */ +int brw_upload_cs_urb_state(struct brw_context *brw); + + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static INLINE struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + + +#define BRW_IS_965(brw) ((brw)->chipset.is_965) +#define BRW_IS_IGDNG(brw) ((brw)->chipset.is_igdng) +#define BRW_IS_G4X(brw) ((brw)->chipset.is_g4x) + + +#endif + diff --git a/src/gallium/drivers/i965/brw_curbe.c b/src/gallium/drivers/i965/brw_curbe.c new file mode 100644 index 00000000000..3f031577d5a --- /dev/null +++ b/src/gallium/drivers/i965/brw_curbe.c @@ -0,0 +1,390 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "brw_batchbuffer.h" +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_util.h" +#include "brw_debug.h" +#include "brw_screen.h" + + +/** + * Partition the CURBE between the various users of constant values: + * Note that vertex and fragment shaders can now fetch constants out + * of constant buffers. We no longer allocatea block of the GRF for + * constants. That greatly reduces the demand for space in the CURBE. + * Some of the comments within are dated... + */ +static int calculate_curbe_offsets( struct brw_context *brw ) +{ + /* CACHE_NEW_WM_PROG */ + const GLuint nr_fp_regs = brw->wm.prog_data->curb_read_length; + + /* BRW_NEW_VERTEX_PROGRAM */ + const GLuint nr_vp_regs = brw->vs.prog_data->curb_read_length; + GLuint nr_clip_regs = 0; + GLuint total_regs; + + /* PIPE_NEW_CLIP */ + if (brw->curr.ucp.nr) { + GLuint nr_planes = 6 + brw->curr.ucp.nr; + nr_clip_regs = (nr_planes * 4 + 15) / 16; + } + + + total_regs = nr_fp_regs + nr_vp_regs + nr_clip_regs; + + /* When this is > 32, want to use a true constant buffer to hold + * the extra constants. + */ + assert(total_regs <= 32); + + /* Lazy resize: + */ + if (nr_fp_regs > brw->curbe.wm_size || + nr_vp_regs > brw->curbe.vs_size || + nr_clip_regs != brw->curbe.clip_size || + (total_regs < brw->curbe.total_size / 4 && + brw->curbe.total_size > 16)) { + + GLuint reg = 0; + + /* Calculate a new layout: + */ + reg = 0; + brw->curbe.wm_start = reg; + brw->curbe.wm_size = nr_fp_regs; reg += nr_fp_regs; + brw->curbe.clip_start = reg; + brw->curbe.clip_size = nr_clip_regs; reg += nr_clip_regs; + brw->curbe.vs_start = reg; + brw->curbe.vs_size = nr_vp_regs; reg += nr_vp_regs; + brw->curbe.total_size = reg; + + if (BRW_DEBUG & DEBUG_CURBE) + debug_printf("curbe wm %d+%d clip %d+%d vs %d+%d\n", + brw->curbe.wm_start, + brw->curbe.wm_size, + brw->curbe.clip_start, + brw->curbe.clip_size, + brw->curbe.vs_start, + brw->curbe.vs_size ); + + brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS; + } + + return 0; +} + + +const struct brw_tracked_state brw_curbe_offsets = { + .dirty = { + .mesa = PIPE_NEW_CLIP, + .brw = BRW_NEW_VERTEX_PROGRAM, + .cache = CACHE_NEW_WM_PROG + }, + .prepare = calculate_curbe_offsets +}; + + + + +/* Define the number of curbes within CS's urb allocation. Multiple + * urb entries -> multiple curbes. These will be used by + * fixed-function hardware in a double-buffering scheme to avoid a + * pipeline stall each time the contents of the curbe is changed. + */ +int brw_upload_cs_urb_state(struct brw_context *brw) +{ + struct brw_cs_urb_state cs_urb; + memset(&cs_urb, 0, sizeof(cs_urb)); + + /* It appears that this is the state packet for the CS unit, ie. the + * urb entries detailed here are housed in the CS range from the + * URB_FENCE command. + */ + cs_urb.header.opcode = CMD_CS_URB_STATE; + cs_urb.header.length = sizeof(cs_urb)/4 - 2; + + /* BRW_NEW_URB_FENCE */ + cs_urb.bits0.nr_urb_entries = brw->urb.nr_cs_entries; + cs_urb.bits0.urb_entry_size = brw->urb.csize - 1; + + assert(brw->urb.nr_cs_entries); + BRW_CACHED_BATCH_STRUCT(brw, &cs_urb); + return 0; +} + +static GLfloat fixed_plane[6][4] = { + { 0, 0, -1, 1 }, + { 0, 0, 1, 1 }, + { 0, -1, 0, 1 }, + { 0, 1, 0, 1 }, + {-1, 0, 0, 1 }, + { 1, 0, 0, 1 } +}; + +/* Upload a new set of constants. Too much variability to go into the + * cache mechanism, but maybe would benefit from a comparison against + * the current uploaded set of constants. + */ +static enum pipe_error prepare_curbe_buffer(struct brw_context *brw) +{ + struct pipe_screen *screen = brw->base.screen; + const GLuint sz = brw->curbe.total_size; + const GLuint bufsz = sz * 16 * sizeof(GLfloat); + enum pipe_error ret; + GLfloat *buf; + GLuint i; + + if (sz == 0) { + if (brw->curbe.last_buf) { + free(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + brw->curbe.last_bufsz = 0; + } + return 0; + } + + buf = (GLfloat *) CALLOC(bufsz, 1); + + /* fragment shader constants */ + if (brw->curbe.wm_size) { + const struct brw_fragment_shader *fs = brw->curr.fragment_shader; + GLuint offset = brw->curbe.wm_start * 16; + GLuint nr_immediate, nr_const; + + nr_immediate = fs->immediates.nr; + if (nr_immediate) { + memcpy(&buf[offset], + fs->immediates.data, + nr_immediate * 4 * sizeof(float)); + + offset += nr_immediate * 4; + } + + nr_const = fs->info.file_max[TGSI_FILE_CONSTANT] + 1; +/* nr_const = brw->wm.prog_data->nr_params; */ + if (nr_const) { + const GLfloat *value = screen->buffer_map( screen, + brw->curr.fragment_constants, + PIPE_BUFFER_USAGE_CPU_READ); + + memcpy(&buf[offset], value, + nr_const * 4 * sizeof(float)); + + screen->buffer_unmap( screen, + brw->curr.fragment_constants ); + } + } + + + /* The clipplanes are actually delivered to both CLIP and VS units. + * VS uses them to calculate the outcode bitmasks. + */ + if (brw->curbe.clip_size) { + GLuint offset = brw->curbe.clip_start * 16; + GLuint j; + + /* If any planes are going this way, send them all this way: + */ + for (i = 0; i < 6; i++) { + buf[offset + i * 4 + 0] = fixed_plane[i][0]; + buf[offset + i * 4 + 1] = fixed_plane[i][1]; + buf[offset + i * 4 + 2] = fixed_plane[i][2]; + buf[offset + i * 4 + 3] = fixed_plane[i][3]; + } + + /* Clip planes: + */ + assert(brw->curr.ucp.nr <= 6); + for (j = 0; j < brw->curr.ucp.nr; j++) { + buf[offset + i * 4 + 0] = brw->curr.ucp.ucp[j][0]; + buf[offset + i * 4 + 1] = brw->curr.ucp.ucp[j][1]; + buf[offset + i * 4 + 2] = brw->curr.ucp.ucp[j][2]; + buf[offset + i * 4 + 3] = brw->curr.ucp.ucp[j][3]; + i++; + } + } + + /* vertex shader constants */ + if (brw->curbe.vs_size) { + GLuint offset = brw->curbe.vs_start * 16; + const struct brw_vertex_shader *vs = brw->curr.vertex_shader; + GLuint nr_immediate, nr_const; + + nr_immediate = vs->immediates.nr; + if (nr_immediate) { + memcpy(&buf[offset], + vs->immediates.data, + nr_immediate * 4 * sizeof(float)); + + offset += nr_immediate * 4; + } + + nr_const = vs->info.file_max[TGSI_FILE_CONSTANT] + 1; + if (nr_const) { + /* XXX: note that constant buffers are currently *already* in + * buffer objects. If we want to keep on putting them into the + * curbe, makes sense to treat constbuf's specially with malloc. + */ + const GLfloat *value = screen->buffer_map( screen, + brw->curr.vertex_constants, + PIPE_BUFFER_USAGE_CPU_READ); + + /* XXX: what if user's constant buffer is too small? + */ + memcpy(&buf[offset], value, nr_const * 4 * sizeof(float)); + + screen->buffer_unmap( screen, brw->curr.vertex_constants ); + } + } + + if (BRW_DEBUG & DEBUG_CURBE) { + for (i = 0; i < sz*16; i+=4) + debug_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, + buf[i+0], buf[i+1], buf[i+2], buf[i+3]); + + debug_printf("last_buf %p buf %p sz %d/%d cmp %d\n", + (void *)brw->curbe.last_buf, (void *)buf, + bufsz, brw->curbe.last_bufsz, + brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1); + } + + if (brw->curbe.curbe_bo != NULL && + brw->curbe.last_buf && + bufsz == brw->curbe.last_bufsz && + memcmp(buf, brw->curbe.last_buf, bufsz) == 0) { + /* constants have not changed */ + FREE(buf); + } + else { + /* constants have changed */ + FREE(brw->curbe.last_buf); + + brw->curbe.last_buf = buf; + brw->curbe.last_bufsz = bufsz; + + if (brw->curbe.curbe_bo != NULL && + (brw->curbe.need_new_bo || + brw->curbe.curbe_next_offset + bufsz > brw->curbe.curbe_bo->size)) + { + bo_reference(&brw->curbe.curbe_bo, NULL); + } + + if (brw->curbe.curbe_bo == NULL) { + /* Allocate a single page for CURBE entries for this + * batchbuffer. They're generally around 64b. We will + * discard the curbe buffer after the batch is flushed to + * avoid synchronous updates. + */ + ret = brw->sws->bo_alloc(brw->sws, + BRW_BUFFER_TYPE_CURBE, + 4096, 1 << 6, + &brw->curbe.curbe_bo); + if (ret) + return ret; + + brw->curbe.curbe_next_offset = 0; + } + + brw->curbe.curbe_offset = brw->curbe.curbe_next_offset; + brw->curbe.curbe_next_offset += bufsz; + brw->curbe.curbe_next_offset = align(brw->curbe.curbe_next_offset, 64); + + /* Copy data to the buffer: + */ + brw->sws->bo_subdata(brw->curbe.curbe_bo, + BRW_DATA_CONSTANT_BUFFER, + brw->curbe.curbe_offset, + bufsz, + buf, + NULL, 0); + } + + brw_add_validated_bo(brw, brw->curbe.curbe_bo); + + /* Because this provokes an action (ie copy the constants into the + * URB), it shouldn't be shortcircuited if identical to the + * previous time - because eg. the urb destination may have + * changed, or the urb contents different to last time. + * + * Note that the data referred to is actually copied internally, + * not just used in place according to passed pointer. + * + * It appears that the CS unit takes care of using each available + * URB entry (Const URB Entry == CURBE) in turn, and issuing + * flushes as necessary when doublebuffering of CURBEs isn't + * possible. + */ + + return 0; +} + +static enum pipe_error emit_curbe_buffer(struct brw_context *brw) +{ + GLuint sz = brw->curbe.total_size; + + BEGIN_BATCH(2, IGNORE_CLIPRECTS); + if (sz == 0) { + OUT_BATCH((CMD_CONST_BUFFER << 16) | (2 - 2)); + OUT_BATCH(0); + } else { + OUT_BATCH((CMD_CONST_BUFFER << 16) | (1 << 8) | (2 - 2)); + OUT_RELOC(brw->curbe.curbe_bo, + BRW_USAGE_STATE, + (sz - 1) + brw->curbe.curbe_offset); + } + ADVANCE_BATCH(); + return 0; +} + +const struct brw_tracked_state brw_curbe_buffer = { + .dirty = { + .mesa = (PIPE_NEW_FRAGMENT_CONSTANTS | + PIPE_NEW_VERTEX_CONSTANTS | + PIPE_NEW_CLIP), + .brw = (BRW_NEW_FRAGMENT_PROGRAM | + BRW_NEW_VERTEX_PROGRAM | + BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */ + BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */ + BRW_NEW_CURBE_OFFSETS | + BRW_NEW_BATCH), + .cache = (CACHE_NEW_WM_PROG) + }, + .prepare = prepare_curbe_buffer, + .emit = emit_curbe_buffer, +}; + diff --git a/src/gallium/drivers/i965/brw_debug.h b/src/gallium/drivers/i965/brw_debug.h new file mode 100644 index 00000000000..ae8e9254a68 --- /dev/null +++ b/src/gallium/drivers/i965/brw_debug.h @@ -0,0 +1,43 @@ +#ifndef BRW_DEBUG_H +#define BRW_DEBUG_H + +/* ================================================================ + * Debugging: + */ + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BLIT 0x8 +#define DEBUG_CURBE 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_BATCH 0x80 +#define DEBUG_PIXEL 0x100 +#define DEBUG_WINSYS 0x200 +#define DEBUG_MIN_URB 0x400 +#define DEBUG_DISASSEM 0x800 +#define DEBUG_unused3 0x1000 +#define DEBUG_SYNC 0x2000 +#define DEBUG_PRIMS 0x4000 +#define DEBUG_VERTS 0x8000 +#define DEBUG_unused4 0x10000 +#define DEBUG_DMA 0x20000 +#define DEBUG_SANITY 0x40000 +#define DEBUG_SLEEP 0x80000 +#define DEBUG_STATS 0x100000 +#define DEBUG_unused5 0x200000 +#define DEBUG_SINGLE_THREAD 0x400000 +#define DEBUG_WM 0x800000 +#define DEBUG_URB 0x1000000 +#define DEBUG_VS 0x2000000 + +#ifdef DEBUG +extern int BRW_DEBUG; +#else +#define BRW_DEBUG 0 +#endif + + + +#endif diff --git a/src/gallium/drivers/i965/brw_defines.h b/src/gallium/drivers/i965/brw_defines.h new file mode 100644 index 00000000000..e201ce4d7ce --- /dev/null +++ b/src/gallium/drivers/i965/brw_defines.h @@ -0,0 +1,847 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_DEFINES_H +#define BRW_DEFINES_H + +/* 3D state: + */ +#define _3DOP_3DSTATE_PIPELINED 0x0 +#define _3DOP_3DSTATE_NONPIPELINED 0x1 +#define _3DOP_3DCONTROL 0x2 +#define _3DOP_3DPRIMITIVE 0x3 + +#define _3DSTATE_PIPELINED_POINTERS 0x00 +#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 +#define _3DSTATE_VERTEX_BUFFERS 0x08 +#define _3DSTATE_VERTEX_ELEMENTS 0x09 +#define _3DSTATE_INDEX_BUFFER 0x0A +#define _3DSTATE_VF_STATISTICS 0x0B +#define _3DSTATE_DRAWING_RECTANGLE 0x00 +#define _3DSTATE_CONSTANT_COLOR 0x01 +#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 +#define _3DSTATE_CHROMA_KEY 0x04 +#define _3DSTATE_DEPTH_BUFFER 0x05 +#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 +#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 +#define _3DSTATE_LINE_STIPPLE 0x08 +#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 +#define _3DCONTROL 0x00 + +#define PIPE_CONTROL_NOWRITE 0x00 +#define PIPE_CONTROL_WRITEIMMEDIATE 0x01 +#define PIPE_CONTROL_WRITEDEPTH 0x02 +#define PIPE_CONTROL_WRITETIMESTAMP 0x03 + +#define PIPE_CONTROL_GTTWRITE_PROCESS_LOCAL 0x00 +#define PIPE_CONTROL_GTTWRITE_GLOBAL 0x01 + +#define _3DPRIM_POINTLIST 0x01 +#define _3DPRIM_LINELIST 0x02 +#define _3DPRIM_LINESTRIP 0x03 +#define _3DPRIM_TRILIST 0x04 +#define _3DPRIM_TRISTRIP 0x05 +#define _3DPRIM_TRIFAN 0x06 +#define _3DPRIM_QUADLIST 0x07 +#define _3DPRIM_QUADSTRIP 0x08 +#define _3DPRIM_LINELIST_ADJ 0x09 +#define _3DPRIM_LINESTRIP_ADJ 0x0A +#define _3DPRIM_TRILIST_ADJ 0x0B +#define _3DPRIM_TRISTRIP_ADJ 0x0C +#define _3DPRIM_TRISTRIP_REVERSE 0x0D +#define _3DPRIM_POLYGON 0x0E +#define _3DPRIM_RECTLIST 0x0F +#define _3DPRIM_LINELOOP 0x10 +#define _3DPRIM_POINTLIST_BF 0x11 +#define _3DPRIM_LINESTRIP_CONT 0x12 +#define _3DPRIM_LINESTRIP_BF 0x13 +#define _3DPRIM_LINESTRIP_CONT_BF 0x14 +#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 + +#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 +#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 + +#define BRW_ANISORATIO_2 0 +#define BRW_ANISORATIO_4 1 +#define BRW_ANISORATIO_6 2 +#define BRW_ANISORATIO_8 3 +#define BRW_ANISORATIO_10 4 +#define BRW_ANISORATIO_12 5 +#define BRW_ANISORATIO_14 6 +#define BRW_ANISORATIO_16 7 + +#define BRW_BLENDFACTOR_ONE 0x1 +#define BRW_BLENDFACTOR_SRC_COLOR 0x2 +#define BRW_BLENDFACTOR_SRC_ALPHA 0x3 +#define BRW_BLENDFACTOR_DST_ALPHA 0x4 +#define BRW_BLENDFACTOR_DST_COLOR 0x5 +#define BRW_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define BRW_BLENDFACTOR_CONST_COLOR 0x7 +#define BRW_BLENDFACTOR_CONST_ALPHA 0x8 +#define BRW_BLENDFACTOR_SRC1_COLOR 0x9 +#define BRW_BLENDFACTOR_SRC1_ALPHA 0x0A +#define BRW_BLENDFACTOR_ZERO 0x11 +#define BRW_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define BRW_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define BRW_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define BRW_BLENDFACTOR_INV_DST_COLOR 0x15 +#define BRW_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define BRW_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define BRW_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define BRW_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define BRW_BLENDFUNCTION_ADD 0 +#define BRW_BLENDFUNCTION_SUBTRACT 1 +#define BRW_BLENDFUNCTION_REVERSE_SUBTRACT 2 +#define BRW_BLENDFUNCTION_MIN 3 +#define BRW_BLENDFUNCTION_MAX 4 + +#define BRW_ALPHATEST_FORMAT_UNORM8 0 +#define BRW_ALPHATEST_FORMAT_FLOAT32 1 + +#define BRW_CHROMAKEY_KILL_ON_ANY_MATCH 0 +#define BRW_CHROMAKEY_REPLACE_BLACK 1 + +#define BRW_CLIP_API_OGL 0 +#define BRW_CLIP_API_DX 1 + +#define BRW_CLIPMODE_NORMAL 0 +#define BRW_CLIPMODE_CLIP_ALL 1 +#define BRW_CLIPMODE_CLIP_NON_REJECTED 2 +#define BRW_CLIPMODE_REJECT_ALL 3 +#define BRW_CLIPMODE_ACCEPT_ALL 4 +#define BRW_CLIPMODE_KERNEL_CLIP 5 + +#define BRW_CLIP_NDCSPACE 0 +#define BRW_CLIP_SCREENSPACE 1 + +#define BRW_COMPAREFUNCTION_ALWAYS 0 +#define BRW_COMPAREFUNCTION_NEVER 1 +#define BRW_COMPAREFUNCTION_LESS 2 +#define BRW_COMPAREFUNCTION_EQUAL 3 +#define BRW_COMPAREFUNCTION_LEQUAL 4 +#define BRW_COMPAREFUNCTION_GREATER 5 +#define BRW_COMPAREFUNCTION_NOTEQUAL 6 +#define BRW_COMPAREFUNCTION_GEQUAL 7 + +#define BRW_COVERAGE_PIXELS_HALF 0 +#define BRW_COVERAGE_PIXELS_1 1 +#define BRW_COVERAGE_PIXELS_2 2 +#define BRW_COVERAGE_PIXELS_4 3 + +#define BRW_CULLMODE_BOTH 0 +#define BRW_CULLMODE_NONE 1 +#define BRW_CULLMODE_FRONT 2 +#define BRW_CULLMODE_BACK 3 + +#define BRW_DEFAULTCOLOR_R8G8B8A8_UNORM 0 +#define BRW_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 + +#define BRW_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 +#define BRW_DEPTHFORMAT_D32_FLOAT 1 +#define BRW_DEPTHFORMAT_D24_UNORM_S8_UINT 2 +#define BRW_DEPTHFORMAT_D16_UNORM 5 + +#define BRW_FLOATING_POINT_IEEE_754 0 +#define BRW_FLOATING_POINT_NON_IEEE_754 1 + +#define BRW_FRONTWINDING_CW 0 +#define BRW_FRONTWINDING_CCW 1 + +#define BRW_SPRITE_POINT_ENABLE 16 + +#define BRW_INDEX_BYTE 0 +#define BRW_INDEX_WORD 1 +#define BRW_INDEX_DWORD 2 + +#define BRW_LOGICOPFUNCTION_CLEAR 0 +#define BRW_LOGICOPFUNCTION_NOR 1 +#define BRW_LOGICOPFUNCTION_AND_INVERTED 2 +#define BRW_LOGICOPFUNCTION_COPY_INVERTED 3 +#define BRW_LOGICOPFUNCTION_AND_REVERSE 4 +#define BRW_LOGICOPFUNCTION_INVERT 5 +#define BRW_LOGICOPFUNCTION_XOR 6 +#define BRW_LOGICOPFUNCTION_NAND 7 +#define BRW_LOGICOPFUNCTION_AND 8 +#define BRW_LOGICOPFUNCTION_EQUIV 9 +#define BRW_LOGICOPFUNCTION_NOOP 10 +#define BRW_LOGICOPFUNCTION_OR_INVERTED 11 +#define BRW_LOGICOPFUNCTION_COPY 12 +#define BRW_LOGICOPFUNCTION_OR_REVERSE 13 +#define BRW_LOGICOPFUNCTION_OR 14 +#define BRW_LOGICOPFUNCTION_SET 15 + +#define BRW_MAPFILTER_NEAREST 0x0 +#define BRW_MAPFILTER_LINEAR 0x1 +#define BRW_MAPFILTER_ANISOTROPIC 0x2 + +#define BRW_MIPFILTER_NONE 0 +#define BRW_MIPFILTER_NEAREST 1 +#define BRW_MIPFILTER_LINEAR 3 + +#define BRW_POLYGON_FRONT_FACING 0 +#define BRW_POLYGON_BACK_FACING 1 + +#define BRW_PREFILTER_ALWAYS 0x0 +#define BRW_PREFILTER_NEVER 0x1 +#define BRW_PREFILTER_LESS 0x2 +#define BRW_PREFILTER_EQUAL 0x3 +#define BRW_PREFILTER_LEQUAL 0x4 +#define BRW_PREFILTER_GREATER 0x5 +#define BRW_PREFILTER_NOTEQUAL 0x6 +#define BRW_PREFILTER_GEQUAL 0x7 + +#define BRW_PROVOKING_VERTEX_0 0 +#define BRW_PROVOKING_VERTEX_1 1 +#define BRW_PROVOKING_VERTEX_2 2 + +#define BRW_RASTRULE_UPPER_LEFT 0 +#define BRW_RASTRULE_UPPER_RIGHT 1 +/* These are listed as "Reserved, but not seen as useful" + * in Intel documentation (page 212, "Point Rasterization Rule", + * section 7.4 "SF Pipeline State Summary", of document + * "Intel® 965 Express Chipset Family and Intel® G35 Express + * Chipset Graphics Controller Programmer's Reference Manual, + * Volume 2: 3D/Media", Revision 1.0b as of January 2008, + * available at + * http://intellinuxgraphics.org/documentation.html + * at the time of this writing). + * + * These appear to be supported on at least some + * i965-family devices, and the BRW_RASTRULE_LOWER_RIGHT + * is useful when using OpenGL to render to a FBO + * (which has the pixel coordinate Y orientation inverted + * with respect to the normal OpenGL pixel coordinate system). + */ +#define BRW_RASTRULE_LOWER_LEFT 2 +#define BRW_RASTRULE_LOWER_RIGHT 3 + +#define BRW_RENDERTARGET_CLAMPRANGE_UNORM 0 +#define BRW_RENDERTARGET_CLAMPRANGE_SNORM 1 +#define BRW_RENDERTARGET_CLAMPRANGE_FORMAT 2 + +#define BRW_STENCILOP_KEEP 0 +#define BRW_STENCILOP_ZERO 1 +#define BRW_STENCILOP_REPLACE 2 +#define BRW_STENCILOP_INCRSAT 3 +#define BRW_STENCILOP_DECRSAT 4 +#define BRW_STENCILOP_INCR 5 +#define BRW_STENCILOP_DECR 6 +#define BRW_STENCILOP_INVERT 7 + +#define BRW_SURFACE_MIPMAPLAYOUT_BELOW 0 +#define BRW_SURFACE_MIPMAPLAYOUT_RIGHT 1 + +#define BRW_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 +#define BRW_SURFACEFORMAT_R32G32B32A32_SINT 0x001 +#define BRW_SURFACEFORMAT_R32G32B32A32_UINT 0x002 +#define BRW_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 +#define BRW_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 +#define BRW_SURFACEFORMAT_R64G64_FLOAT 0x005 +#define BRW_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 +#define BRW_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 +#define BRW_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 +#define BRW_SURFACEFORMAT_R32G32B32_FLOAT 0x040 +#define BRW_SURFACEFORMAT_R32G32B32_SINT 0x041 +#define BRW_SURFACEFORMAT_R32G32B32_UINT 0x042 +#define BRW_SURFACEFORMAT_R32G32B32_UNORM 0x043 +#define BRW_SURFACEFORMAT_R32G32B32_SNORM 0x044 +#define BRW_SURFACEFORMAT_R32G32B32_SSCALED 0x045 +#define BRW_SURFACEFORMAT_R32G32B32_USCALED 0x046 +#define BRW_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 +#define BRW_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 +#define BRW_SURFACEFORMAT_R16G16B16A16_SINT 0x082 +#define BRW_SURFACEFORMAT_R16G16B16A16_UINT 0x083 +#define BRW_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 +#define BRW_SURFACEFORMAT_R32G32_FLOAT 0x085 +#define BRW_SURFACEFORMAT_R32G32_SINT 0x086 +#define BRW_SURFACEFORMAT_R32G32_UINT 0x087 +#define BRW_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 +#define BRW_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 +#define BRW_SURFACEFORMAT_L32A32_FLOAT 0x08A +#define BRW_SURFACEFORMAT_R32G32_UNORM 0x08B +#define BRW_SURFACEFORMAT_R32G32_SNORM 0x08C +#define BRW_SURFACEFORMAT_R64_FLOAT 0x08D +#define BRW_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E +#define BRW_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F +#define BRW_SURFACEFORMAT_A32X32_FLOAT 0x090 +#define BRW_SURFACEFORMAT_L32X32_FLOAT 0x091 +#define BRW_SURFACEFORMAT_I32X32_FLOAT 0x092 +#define BRW_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 +#define BRW_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 +#define BRW_SURFACEFORMAT_R32G32_SSCALED 0x095 +#define BRW_SURFACEFORMAT_R32G32_USCALED 0x096 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 +#define BRW_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 +#define BRW_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 +#define BRW_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 +#define BRW_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA +#define BRW_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB +#define BRW_SURFACEFORMAT_R16G16_UNORM 0x0CC +#define BRW_SURFACEFORMAT_R16G16_SNORM 0x0CD +#define BRW_SURFACEFORMAT_R16G16_SINT 0x0CE +#define BRW_SURFACEFORMAT_R16G16_UINT 0x0CF +#define BRW_SURFACEFORMAT_R16G16_FLOAT 0x0D0 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 +#define BRW_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 +#define BRW_SURFACEFORMAT_R32_SINT 0x0D6 +#define BRW_SURFACEFORMAT_R32_UINT 0x0D7 +#define BRW_SURFACEFORMAT_R32_FLOAT 0x0D8 +#define BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 +#define BRW_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA +#define BRW_SURFACEFORMAT_L16A16_UNORM 0x0DF +#define BRW_SURFACEFORMAT_I24X8_UNORM 0x0E0 +#define BRW_SURFACEFORMAT_L24X8_UNORM 0x0E1 +#define BRW_SURFACEFORMAT_A24X8_UNORM 0x0E2 +#define BRW_SURFACEFORMAT_I32_FLOAT 0x0E3 +#define BRW_SURFACEFORMAT_L32_FLOAT 0x0E4 +#define BRW_SURFACEFORMAT_A32_FLOAT 0x0E5 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC +#define BRW_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED +#define BRW_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE +#define BRW_SURFACEFORMAT_L16A16_FLOAT 0x0F0 +#define BRW_SURFACEFORMAT_R32_UNORM 0x0F1 +#define BRW_SURFACEFORMAT_R32_SNORM 0x0F2 +#define BRW_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 +#define BRW_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 +#define BRW_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 +#define BRW_SURFACEFORMAT_R16G16_SSCALED 0x0F6 +#define BRW_SURFACEFORMAT_R16G16_USCALED 0x0F7 +#define BRW_SURFACEFORMAT_R32_SSCALED 0x0F8 +#define BRW_SURFACEFORMAT_R32_USCALED 0x0F9 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM 0x100 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 +#define BRW_SURFACEFORMAT_R8G8_UNORM 0x106 +#define BRW_SURFACEFORMAT_R8G8_SNORM 0x107 +#define BRW_SURFACEFORMAT_R8G8_SINT 0x108 +#define BRW_SURFACEFORMAT_R8G8_UINT 0x109 +#define BRW_SURFACEFORMAT_R16_UNORM 0x10A +#define BRW_SURFACEFORMAT_R16_SNORM 0x10B +#define BRW_SURFACEFORMAT_R16_SINT 0x10C +#define BRW_SURFACEFORMAT_R16_UINT 0x10D +#define BRW_SURFACEFORMAT_R16_FLOAT 0x10E +#define BRW_SURFACEFORMAT_I16_UNORM 0x111 +#define BRW_SURFACEFORMAT_L16_UNORM 0x112 +#define BRW_SURFACEFORMAT_A16_UNORM 0x113 +#define BRW_SURFACEFORMAT_L8A8_UNORM 0x114 +#define BRW_SURFACEFORMAT_I16_FLOAT 0x115 +#define BRW_SURFACEFORMAT_L16_FLOAT 0x116 +#define BRW_SURFACEFORMAT_A16_FLOAT 0x117 +#define BRW_SURFACEFORMAT_L8A8_UNORM_SRGB 0x118 +#define BRW_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B +#define BRW_SURFACEFORMAT_R8G8_SSCALED 0x11C +#define BRW_SURFACEFORMAT_R8G8_USCALED 0x11D +#define BRW_SURFACEFORMAT_R16_SSCALED 0x11E +#define BRW_SURFACEFORMAT_R16_USCALED 0x11F +#define BRW_SURFACEFORMAT_R8_UNORM 0x140 +#define BRW_SURFACEFORMAT_R8_SNORM 0x141 +#define BRW_SURFACEFORMAT_R8_SINT 0x142 +#define BRW_SURFACEFORMAT_R8_UINT 0x143 +#define BRW_SURFACEFORMAT_A8_UNORM 0x144 +#define BRW_SURFACEFORMAT_I8_UNORM 0x145 +#define BRW_SURFACEFORMAT_L8_UNORM 0x146 +#define BRW_SURFACEFORMAT_P4A4_UNORM 0x147 +#define BRW_SURFACEFORMAT_A4P4_UNORM 0x148 +#define BRW_SURFACEFORMAT_R8_SSCALED 0x149 +#define BRW_SURFACEFORMAT_R8_USCALED 0x14A +#define BRW_SURFACEFORMAT_L8_UNORM_SRGB 0x14C +#define BRW_SURFACEFORMAT_R1_UINT 0x181 +#define BRW_SURFACEFORMAT_YCRCB_NORMAL 0x182 +#define BRW_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 +#define BRW_SURFACEFORMAT_BC1_UNORM 0x186 +#define BRW_SURFACEFORMAT_BC2_UNORM 0x187 +#define BRW_SURFACEFORMAT_BC3_UNORM 0x188 +#define BRW_SURFACEFORMAT_BC4_UNORM 0x189 +#define BRW_SURFACEFORMAT_BC5_UNORM 0x18A +#define BRW_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B +#define BRW_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C +#define BRW_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D +#define BRW_SURFACEFORMAT_MONO8 0x18E +#define BRW_SURFACEFORMAT_YCRCB_SWAPUV 0x18F +#define BRW_SURFACEFORMAT_YCRCB_SWAPY 0x190 +#define BRW_SURFACEFORMAT_DXT1_RGB 0x191 +#define BRW_SURFACEFORMAT_FXT1 0x192 +#define BRW_SURFACEFORMAT_R8G8B8_UNORM 0x193 +#define BRW_SURFACEFORMAT_R8G8B8_SNORM 0x194 +#define BRW_SURFACEFORMAT_R8G8B8_SSCALED 0x195 +#define BRW_SURFACEFORMAT_R8G8B8_USCALED 0x196 +#define BRW_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 +#define BRW_SURFACEFORMAT_R64G64B64_FLOAT 0x198 +#define BRW_SURFACEFORMAT_BC4_SNORM 0x199 +#define BRW_SURFACEFORMAT_BC5_SNORM 0x19A +#define BRW_SURFACEFORMAT_R16G16B16_UNORM 0x19C +#define BRW_SURFACEFORMAT_R16G16B16_SNORM 0x19D +#define BRW_SURFACEFORMAT_R16G16B16_SSCALED 0x19E +#define BRW_SURFACEFORMAT_R16G16B16_USCALED 0x19F +#define BRW_SURFACEFORMAT_INVALID 0xFFF + +#define BRW_SURFACERETURNFORMAT_FLOAT32 0 +#define BRW_SURFACERETURNFORMAT_S1 1 + +#define BRW_SURFACE_1D 0 +#define BRW_SURFACE_2D 1 +#define BRW_SURFACE_3D 2 +#define BRW_SURFACE_CUBE 3 +#define BRW_SURFACE_BUFFER 4 +#define BRW_SURFACE_NULL 7 + +#define BRW_TEXCOORDMODE_WRAP 0 +#define BRW_TEXCOORDMODE_MIRROR 1 +#define BRW_TEXCOORDMODE_CLAMP 2 +#define BRW_TEXCOORDMODE_CUBE 3 +#define BRW_TEXCOORDMODE_CLAMP_BORDER 4 +#define BRW_TEXCOORDMODE_MIRROR_ONCE 5 + +#define BRW_THREAD_PRIORITY_NORMAL 0 +#define BRW_THREAD_PRIORITY_HIGH 1 + +#define BRW_TILEWALK_XMAJOR 0 +#define BRW_TILEWALK_YMAJOR 1 + +#define BRW_VERTEX_SUBPIXEL_PRECISION_8BITS 0 +#define BRW_VERTEX_SUBPIXEL_PRECISION_4BITS 1 + +/* Execution Unit (EU) defines + */ + +#define BRW_ALIGN_1 0 +#define BRW_ALIGN_16 1 + +#define BRW_ADDRESS_DIRECT 0 +#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER 1 + +#define BRW_CHANNEL_X 0 +#define BRW_CHANNEL_Y 1 +#define BRW_CHANNEL_Z 2 +#define BRW_CHANNEL_W 3 + +#define BRW_COMPRESSION_NONE 0 +#define BRW_COMPRESSION_2NDHALF 1 +#define BRW_COMPRESSION_COMPRESSED 2 + +#define BRW_CONDITIONAL_NONE 0 +#define BRW_CONDITIONAL_Z 1 +#define BRW_CONDITIONAL_NZ 2 +#define BRW_CONDITIONAL_EQ 1 /* Z */ +#define BRW_CONDITIONAL_NEQ 2 /* NZ */ +#define BRW_CONDITIONAL_G 3 +#define BRW_CONDITIONAL_GE 4 +#define BRW_CONDITIONAL_L 5 +#define BRW_CONDITIONAL_LE 6 +#define BRW_CONDITIONAL_R 7 +#define BRW_CONDITIONAL_O 8 +#define BRW_CONDITIONAL_U 9 + +#define BRW_DEBUG_NONE 0 +#define BRW_DEBUG_BREAKPOINT 1 + +#define BRW_DEPENDENCY_NORMAL 0 +#define BRW_DEPENDENCY_NOTCLEARED 1 +#define BRW_DEPENDENCY_NOTCHECKED 2 +#define BRW_DEPENDENCY_DISABLE 3 + +#define BRW_EXECUTE_1 0 +#define BRW_EXECUTE_2 1 +#define BRW_EXECUTE_4 2 +#define BRW_EXECUTE_8 3 +#define BRW_EXECUTE_16 4 +#define BRW_EXECUTE_32 5 + +#define BRW_HORIZONTAL_STRIDE_0 0 +#define BRW_HORIZONTAL_STRIDE_1 1 +#define BRW_HORIZONTAL_STRIDE_2 2 +#define BRW_HORIZONTAL_STRIDE_4 3 + +#define BRW_INSTRUCTION_NORMAL 0 +#define BRW_INSTRUCTION_SATURATE 1 + +#define BRW_MASK_ENABLE 0 +#define BRW_MASK_DISABLE 1 + +#define BRW_OPCODE_MOV 1 +#define BRW_OPCODE_SEL 2 +#define BRW_OPCODE_NOT 4 +#define BRW_OPCODE_AND 5 +#define BRW_OPCODE_OR 6 +#define BRW_OPCODE_XOR 7 +#define BRW_OPCODE_SHR 8 +#define BRW_OPCODE_SHL 9 +#define BRW_OPCODE_RSR 10 +#define BRW_OPCODE_RSL 11 +#define BRW_OPCODE_ASR 12 +#define BRW_OPCODE_CMP 16 +#define BRW_OPCODE_CMPN 17 +#define BRW_OPCODE_JMPI 32 +#define BRW_OPCODE_IF 34 +#define BRW_OPCODE_IFF 35 +#define BRW_OPCODE_ELSE 36 +#define BRW_OPCODE_ENDIF 37 +#define BRW_OPCODE_DO 38 +#define BRW_OPCODE_WHILE 39 +#define BRW_OPCODE_BREAK 40 +#define BRW_OPCODE_CONTINUE 41 +#define BRW_OPCODE_HALT 42 +#define BRW_OPCODE_MSAVE 44 +#define BRW_OPCODE_MRESTORE 45 +#define BRW_OPCODE_PUSH 46 +#define BRW_OPCODE_POP 47 +#define BRW_OPCODE_WAIT 48 +#define BRW_OPCODE_SEND 49 +#define BRW_OPCODE_ADD 64 +#define BRW_OPCODE_MUL 65 +#define BRW_OPCODE_AVG 66 +#define BRW_OPCODE_FRC 67 +#define BRW_OPCODE_RNDU 68 +#define BRW_OPCODE_RNDD 69 +#define BRW_OPCODE_RNDE 70 +#define BRW_OPCODE_RNDZ 71 +#define BRW_OPCODE_MAC 72 +#define BRW_OPCODE_MACH 73 +#define BRW_OPCODE_LZD 74 +#define BRW_OPCODE_SAD2 80 +#define BRW_OPCODE_SADA2 81 +#define BRW_OPCODE_DP4 84 +#define BRW_OPCODE_DPH 85 +#define BRW_OPCODE_DP3 86 +#define BRW_OPCODE_DP2 87 +#define BRW_OPCODE_DPA2 88 +#define BRW_OPCODE_LINE 89 +#define BRW_OPCODE_NOP 126 + +#define BRW_PREDICATE_NONE 0 +#define BRW_PREDICATE_NORMAL 1 +#define BRW_PREDICATE_ALIGN1_ANYV 2 +#define BRW_PREDICATE_ALIGN1_ALLV 3 +#define BRW_PREDICATE_ALIGN1_ANY2H 4 +#define BRW_PREDICATE_ALIGN1_ALL2H 5 +#define BRW_PREDICATE_ALIGN1_ANY4H 6 +#define BRW_PREDICATE_ALIGN1_ALL4H 7 +#define BRW_PREDICATE_ALIGN1_ANY8H 8 +#define BRW_PREDICATE_ALIGN1_ALL8H 9 +#define BRW_PREDICATE_ALIGN1_ANY16H 10 +#define BRW_PREDICATE_ALIGN1_ALL16H 11 +#define BRW_PREDICATE_ALIGN16_REPLICATE_X 2 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Y 3 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Z 4 +#define BRW_PREDICATE_ALIGN16_REPLICATE_W 5 +#define BRW_PREDICATE_ALIGN16_ANY4H 6 +#define BRW_PREDICATE_ALIGN16_ALL4H 7 + +#define BRW_ARCHITECTURE_REGISTER_FILE 0 +#define BRW_GENERAL_REGISTER_FILE 1 +#define BRW_MESSAGE_REGISTER_FILE 2 +#define BRW_IMMEDIATE_VALUE 3 + +#define BRW_REGISTER_TYPE_UD 0 +#define BRW_REGISTER_TYPE_D 1 +#define BRW_REGISTER_TYPE_UW 2 +#define BRW_REGISTER_TYPE_W 3 +#define BRW_REGISTER_TYPE_UB 4 +#define BRW_REGISTER_TYPE_B 5 +#define BRW_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ +#define BRW_REGISTER_TYPE_HF 6 +#define BRW_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ +#define BRW_REGISTER_TYPE_F 7 + +#define BRW_ARF_NULL 0x00 +#define BRW_ARF_ADDRESS 0x10 +#define BRW_ARF_ACCUMULATOR 0x20 +#define BRW_ARF_FLAG 0x30 +#define BRW_ARF_MASK 0x40 +#define BRW_ARF_MASK_STACK 0x50 +#define BRW_ARF_MASK_STACK_DEPTH 0x60 +#define BRW_ARF_STATE 0x70 +#define BRW_ARF_CONTROL 0x80 +#define BRW_ARF_NOTIFICATION_COUNT 0x90 +#define BRW_ARF_IP 0xA0 + +#define BRW_AMASK 0 +#define BRW_IMASK 1 +#define BRW_LMASK 2 +#define BRW_CMASK 3 + + + +#define BRW_THREAD_NORMAL 0 +#define BRW_THREAD_ATOMIC 1 +#define BRW_THREAD_SWITCH 2 + +#define BRW_VERTICAL_STRIDE_0 0 +#define BRW_VERTICAL_STRIDE_1 1 +#define BRW_VERTICAL_STRIDE_2 2 +#define BRW_VERTICAL_STRIDE_4 3 +#define BRW_VERTICAL_STRIDE_8 4 +#define BRW_VERTICAL_STRIDE_16 5 +#define BRW_VERTICAL_STRIDE_32 6 +#define BRW_VERTICAL_STRIDE_64 7 +#define BRW_VERTICAL_STRIDE_128 8 +#define BRW_VERTICAL_STRIDE_256 9 +#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF + +#define BRW_WIDTH_1 0 +#define BRW_WIDTH_2 1 +#define BRW_WIDTH_4 2 +#define BRW_WIDTH_8 3 +#define BRW_WIDTH_16 4 + +#define BRW_STATELESS_BUFFER_BOUNDARY_1K 0 +#define BRW_STATELESS_BUFFER_BOUNDARY_2K 1 +#define BRW_STATELESS_BUFFER_BOUNDARY_4K 2 +#define BRW_STATELESS_BUFFER_BOUNDARY_8K 3 +#define BRW_STATELESS_BUFFER_BOUNDARY_16K 4 +#define BRW_STATELESS_BUFFER_BOUNDARY_32K 5 +#define BRW_STATELESS_BUFFER_BOUNDARY_64K 6 +#define BRW_STATELESS_BUFFER_BOUNDARY_128K 7 +#define BRW_STATELESS_BUFFER_BOUNDARY_256K 8 +#define BRW_STATELESS_BUFFER_BOUNDARY_512K 9 +#define BRW_STATELESS_BUFFER_BOUNDARY_1M 10 +#define BRW_STATELESS_BUFFER_BOUNDARY_2M 11 + +#define BRW_POLYGON_FACING_FRONT 0 +#define BRW_POLYGON_FACING_BACK 1 + +#define BRW_MESSAGE_TARGET_NULL 0 +#define BRW_MESSAGE_TARGET_MATH 1 +#define BRW_MESSAGE_TARGET_SAMPLER 2 +#define BRW_MESSAGE_TARGET_GATEWAY 3 +#define BRW_MESSAGE_TARGET_DATAPORT_READ 4 +#define BRW_MESSAGE_TARGET_DATAPORT_WRITE 5 +#define BRW_MESSAGE_TARGET_URB 6 +#define BRW_MESSAGE_TARGET_THREAD_SPAWNER 7 + +#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32 0 +#define BRW_SAMPLER_RETURN_FORMAT_UINT32 2 +#define BRW_SAMPLER_RETURN_FORMAT_SINT32 3 + +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 +#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 + +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG 0 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_IGDNG 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG 0 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_BIAS_IGDNG 1 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG 1 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_IGDNG 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD_IGDNG 2 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD_IGDNG 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG 3 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE_IGDNG 3 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG 3 + +/* for IGDNG only */ +#define BRW_SAMPLER_SIMD_MODE_SIMD4X2 0 +#define BRW_SAMPLER_SIMD_MODE_SIMD8 1 +#define BRW_SAMPLER_SIMD_MODE_SIMD16 2 +#define BRW_SAMPLER_SIMD_MODE_SIMD32_64 3 + +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 +#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS 2 +#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS 3 +#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS 4 + +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 + +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 + +#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 +#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 + +#define BRW_DATAPORT_READ_TARGET_DATA_CACHE 0 +#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE 1 +#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 + +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 + +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 +#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 +#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 +#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 + +#define BRW_MATH_FUNCTION_INV 1 +#define BRW_MATH_FUNCTION_LOG 2 +#define BRW_MATH_FUNCTION_EXP 3 +#define BRW_MATH_FUNCTION_SQRT 4 +#define BRW_MATH_FUNCTION_RSQ 5 +#define BRW_MATH_FUNCTION_SIN 6 /* was 7 */ +#define BRW_MATH_FUNCTION_COS 7 /* was 8 */ +#define BRW_MATH_FUNCTION_SINCOS 8 /* was 6 */ +#define BRW_MATH_FUNCTION_TAN 9 +#define BRW_MATH_FUNCTION_POW 10 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT 12 +#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER 13 + +#define BRW_MATH_INTEGER_UNSIGNED 0 +#define BRW_MATH_INTEGER_SIGNED 1 + +#define BRW_MATH_PRECISION_FULL 0 +#define BRW_MATH_PRECISION_PARTIAL 1 + +#define BRW_MATH_SATURATE_NONE 0 +#define BRW_MATH_SATURATE_SATURATE 1 + +#define BRW_MATH_DATA_VECTOR 0 +#define BRW_MATH_DATA_SCALAR 1 + +#define BRW_URB_OPCODE_WRITE 0 + +#define BRW_URB_SWIZZLE_NONE 0 +#define BRW_URB_SWIZZLE_INTERLEAVE 1 +#define BRW_URB_SWIZZLE_TRANSPOSE 2 + +#define BRW_SCRATCH_SPACE_SIZE_1K 0 +#define BRW_SCRATCH_SPACE_SIZE_2K 1 +#define BRW_SCRATCH_SPACE_SIZE_4K 2 +#define BRW_SCRATCH_SPACE_SIZE_8K 3 +#define BRW_SCRATCH_SPACE_SIZE_16K 4 +#define BRW_SCRATCH_SPACE_SIZE_32K 5 +#define BRW_SCRATCH_SPACE_SIZE_64K 6 +#define BRW_SCRATCH_SPACE_SIZE_128K 7 +#define BRW_SCRATCH_SPACE_SIZE_256K 8 +#define BRW_SCRATCH_SPACE_SIZE_512K 9 +#define BRW_SCRATCH_SPACE_SIZE_1M 10 +#define BRW_SCRATCH_SPACE_SIZE_2M 11 + + + + +#define CMD_URB_FENCE 0x6000 +#define CMD_CS_URB_STATE 0x6001 +#define CMD_CONST_BUFFER 0x6002 + +#define CMD_STATE_BASE_ADDRESS 0x6101 +#define CMD_STATE_INSN_POINTER 0x6102 +#define CMD_PIPELINE_SELECT_965 0x6104 +#define CMD_PIPELINE_SELECT_GM45 0x6904 + +#define CMD_PIPELINED_STATE_POINTERS 0x7800 +#define CMD_BINDING_TABLE_PTRS 0x7801 + +#define CMD_VERTEX_BUFFER 0x7808 +# define BRW_VB0_INDEX_SHIFT 27 +# define BRW_VB0_ACCESS_VERTEXDATA (0 << 26) +# define BRW_VB0_ACCESS_INSTANCEDATA (1 << 26) +# define BRW_VB0_PITCH_SHIFT 0 + +#define CMD_VERTEX_ELEMENT 0x7809 +# define BRW_VE0_INDEX_SHIFT 27 +# define BRW_VE0_FORMAT_SHIFT 16 +# define BRW_VE0_VALID (1 << 26) +# define BRW_VE0_SRC_OFFSET_SHIFT 0 +# define BRW_VE1_COMPONENT_NOSTORE 0 +# define BRW_VE1_COMPONENT_STORE_SRC 1 +# define BRW_VE1_COMPONENT_STORE_0 2 +# define BRW_VE1_COMPONENT_STORE_1_FLT 3 +# define BRW_VE1_COMPONENT_STORE_1_INT 4 +# define BRW_VE1_COMPONENT_STORE_VID 5 +# define BRW_VE1_COMPONENT_STORE_IID 6 +# define BRW_VE1_COMPONENT_STORE_PID 7 +# define BRW_VE1_COMPONENT_0_SHIFT 28 +# define BRW_VE1_COMPONENT_1_SHIFT 24 +# define BRW_VE1_COMPONENT_2_SHIFT 20 +# define BRW_VE1_COMPONENT_3_SHIFT 16 +# define BRW_VE1_DST_OFFSET_SHIFT 0 + +#define CMD_INDEX_BUFFER 0x780a +#define CMD_VF_STATISTICS_965 0x780b +#define CMD_VF_STATISTICS_GM45 0x680b + +#define CMD_DRAW_RECT 0x7900 +#define CMD_BLEND_CONSTANT_COLOR 0x7901 +#define CMD_CHROMA_KEY 0x7904 +#define CMD_DEPTH_BUFFER 0x7905 +#define CMD_POLY_STIPPLE_OFFSET 0x7906 +#define CMD_POLY_STIPPLE_PATTERN 0x7907 +#define CMD_LINE_STIPPLE_PATTERN 0x7908 +#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7909 +#define CMD_AA_LINE_PARAMETERS 0x790a + +#define CMD_PIPE_CONTROL 0x7a00 + +#define CMD_3D_PRIM 0x7b00 + +#define CMD_MI_FLUSH 0x0200 + + +/* Various values from the R0 vertex header: + */ +#define R02_PRIM_END 0x1 +#define R02_PRIM_START 0x2 + +#define URB_SIZES(brw) (BRW_IS_IGDNG(brw) ? 1024 : \ + (BRW_IS_G4X(brw) ? 384 : 256)) /* 512 bit units */ + + + +#endif diff --git a/src/gallium/drivers/i965/brw_disasm.c b/src/gallium/drivers/i965/brw_disasm.c new file mode 100644 index 00000000000..65db27248b1 --- /dev/null +++ b/src/gallium/drivers/i965/brw_disasm.c @@ -0,0 +1,922 @@ +/* + * Copyright © 2008 Keith Packard + * + * Permission to use, copy, modify, distribute, and sell this software and its + * documentation for any purpose is hereby granted without fee, provided that + * the above copyright notice appear in all copies and that both that copyright + * notice and this permission notice appear in supporting documentation, and + * that the name of the copyright holders not be used in advertising or + * publicity pertaining to distribution of the software without specific, + * written prior permission. The copyright holders make no representations + * about the suitability of this software for any purpose. It is provided "as + * is" without express or implied warranty. + * + * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, + * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO + * EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR + * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE + * OF THIS SOFTWARE. + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <getopt.h> +#include <unistd.h> +#include <stdarg.h> + +#include "brw_disasm.h" +#include "brw_structs.h" +#include "brw_reg.h" +#include "brw_defines.h" + +struct { + char *name; + int nsrc; + int ndst; +} opcode[128] = { + [BRW_OPCODE_MOV] = { .name = "mov", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_FRC] = { .name = "frc", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_RNDU] = { .name = "rndu", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_RNDD] = { .name = "rndd", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_RNDE] = { .name = "rnde", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_RNDZ] = { .name = "rndz", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_NOT] = { .name = "not", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_LZD] = { .name = "lzd", .nsrc = 1, .ndst = 1 }, + + [BRW_OPCODE_MUL] = { .name = "mul", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_MAC] = { .name = "mac", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_MACH] = { .name = "mach", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_LINE] = { .name = "line", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_SAD2] = { .name = "sad2", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_SADA2] = { .name = "sada2", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_DP4] = { .name = "dp4", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_DPH] = { .name = "dph", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_DP3] = { .name = "dp3", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_DP2] = { .name = "dp2", .nsrc = 2, .ndst = 1 }, + + [BRW_OPCODE_AVG] = { .name = "avg", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_ADD] = { .name = "add", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_SEL] = { .name = "sel", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_AND] = { .name = "and", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_OR] = { .name = "or", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_XOR] = { .name = "xor", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_SHR] = { .name = "shr", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_SHL] = { .name = "shl", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_ASR] = { .name = "asr", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_CMP] = { .name = "cmp", .nsrc = 2, .ndst = 1 }, + [BRW_OPCODE_CMPN] = { .name = "cmpn", .nsrc = 2, .ndst = 1 }, + + [BRW_OPCODE_SEND] = { .name = "send", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_NOP] = { .name = "nop", .nsrc = 0, .ndst = 0 }, + [BRW_OPCODE_JMPI] = { .name = "jmpi", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_IF] = { .name = "if", .nsrc = 2, .ndst = 0 }, + [BRW_OPCODE_IFF] = { .name = "iff", .nsrc = 1, .ndst = 01 }, + [BRW_OPCODE_WHILE] = { .name = "while", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_ELSE] = { .name = "else", .nsrc = 2, .ndst = 0 }, + [BRW_OPCODE_BREAK] = { .name = "break", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_CONTINUE] = { .name = "cont", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_HALT] = { .name = "halt", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_MSAVE] = { .name = "msave", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_PUSH] = { .name = "push", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_MRESTORE] = { .name = "mrest", .nsrc = 1, .ndst = 1 }, + [BRW_OPCODE_POP] = { .name = "pop", .nsrc = 2, .ndst = 0 }, + [BRW_OPCODE_WAIT] = { .name = "wait", .nsrc = 1, .ndst = 0 }, + [BRW_OPCODE_DO] = { .name = "do", .nsrc = 0, .ndst = 0 }, + [BRW_OPCODE_ENDIF] = { .name = "endif", .nsrc = 2, .ndst = 0 }, +}; + +char *conditional_modifier[16] = { + [BRW_CONDITIONAL_NONE] = "", + [BRW_CONDITIONAL_Z] = ".e", + [BRW_CONDITIONAL_NZ] = ".ne", + [BRW_CONDITIONAL_G] = ".g", + [BRW_CONDITIONAL_GE] = ".ge", + [BRW_CONDITIONAL_L] = ".l", + [BRW_CONDITIONAL_LE] = ".le", + [BRW_CONDITIONAL_R] = ".r", + [BRW_CONDITIONAL_O] = ".o", + [BRW_CONDITIONAL_U] = ".u", +}; + +char *negate[2] = { + [0] = "", + [1] = "-", +}; + +char *_abs[2] = { + [0] = "", + [1] = "(abs)", +}; + +char *vert_stride[16] = { + [0] = "0", + [1] = "1", + [2] = "2", + [3] = "4", + [4] = "8", + [5] = "16", + [6] = "32", + [15] = "VxH", +}; + +char *width[8] = { + [0] = "1", + [1] = "2", + [2] = "4", + [3] = "8", + [4] = "16", +}; + +char *horiz_stride[4] = { + [0] = "0", + [1] = "1", + [2] = "2", + [3] = "4" +}; + +char *chan_sel[4] = { + [0] = "x", + [1] = "y", + [2] = "z", + [3] = "w", +}; + +char *dest_condmod[16] = { + [0] = NULL +}; + +char *debug_ctrl[2] = { + [0] = "", + [1] = ".breakpoint" +}; + +char *saturate[2] = { + [0] = "", + [1] = ".sat" +}; + +char *exec_size[8] = { + [0] = "1", + [1] = "2", + [2] = "4", + [3] = "8", + [4] = "16", + [5] = "32" +}; + +char *pred_inv[2] = { + [0] = "+", + [1] = "-" +}; + +char *pred_ctrl_align16[16] = { + [1] = "", + [2] = ".x", + [3] = ".y", + [4] = ".z", + [5] = ".w", + [6] = ".any4h", + [7] = ".all4h", +}; + +char *pred_ctrl_align1[16] = { + [1] = "", + [2] = ".anyv", + [3] = ".allv", + [4] = ".any2h", + [5] = ".all2h", + [6] = ".any4h", + [7] = ".all4h", + [8] = ".any8h", + [9] = ".all8h", + [10] = ".any16h", + [11] = ".all16h", +}; + +char *thread_ctrl[4] = { + [0] = "", + [2] = "switch" +}; + +char *compr_ctrl[4] = { + [0] = "", + [1] = "sechalf", + [2] = "compr", +}; + +char *dep_ctrl[4] = { + [0] = "", + [1] = "NoDDClr", + [2] = "NoDDChk", + [3] = "NoDDClr,NoDDChk", +}; + +char *mask_ctrl[4] = { + [0] = "", + [1] = "nomask", +}; + +char *access_mode[2] = { + [0] = "align1", + [1] = "align16", +}; + +char *reg_encoding[8] = { + [0] = "UD", + [1] = "D", + [2] = "UW", + [3] = "W", + [4] = "UB", + [5] = "B", + [7] = "F" +}; + +char *imm_encoding[8] = { + [0] = "UD", + [1] = "D", + [2] = "UW", + [3] = "W", + [5] = "VF", + [5] = "V", + [7] = "F" +}; + +char *reg_file[4] = { + [0] = "A", + [1] = "g", + [2] = "m", + [3] = "imm", +}; + +char *writemask[16] = { + [0x0] = ".", + [0x1] = ".x", + [0x2] = ".y", + [0x3] = ".xy", + [0x4] = ".z", + [0x5] = ".xz", + [0x6] = ".yz", + [0x7] = ".xyz", + [0x8] = ".w", + [0x9] = ".xw", + [0xa] = ".yw", + [0xb] = ".xyw", + [0xc] = ".zw", + [0xd] = ".xzw", + [0xe] = ".yzw", + [0xf] = "", +}; + +char *end_of_thread[2] = { + [0] = "", + [1] = "EOT" +}; + +char *target_function[16] = { + [BRW_MESSAGE_TARGET_NULL] = "null", + [BRW_MESSAGE_TARGET_MATH] = "math", + [BRW_MESSAGE_TARGET_SAMPLER] = "sampler", + [BRW_MESSAGE_TARGET_GATEWAY] = "gateway", + [BRW_MESSAGE_TARGET_DATAPORT_READ] = "read", + [BRW_MESSAGE_TARGET_DATAPORT_WRITE] = "write", + [BRW_MESSAGE_TARGET_URB] = "urb", + [BRW_MESSAGE_TARGET_THREAD_SPAWNER] = "thread_spawner" +}; + +char *math_function[16] = { + [BRW_MATH_FUNCTION_INV] = "inv", + [BRW_MATH_FUNCTION_LOG] = "log", + [BRW_MATH_FUNCTION_EXP] = "exp", + [BRW_MATH_FUNCTION_SQRT] = "sqrt", + [BRW_MATH_FUNCTION_RSQ] = "rsq", + [BRW_MATH_FUNCTION_SIN] = "sin", + [BRW_MATH_FUNCTION_COS] = "cos", + [BRW_MATH_FUNCTION_SINCOS] = "sincos", + [BRW_MATH_FUNCTION_TAN] = "tan", + [BRW_MATH_FUNCTION_POW] = "pow", + [BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER] = "intdivmod", + [BRW_MATH_FUNCTION_INT_DIV_QUOTIENT] = "intmod", + [BRW_MATH_FUNCTION_INT_DIV_REMAINDER] = "intdiv", +}; + +char *math_saturate[2] = { + [0] = "", + [1] = "sat" +}; + +char *math_signed[2] = { + [0] = "", + [1] = "signed" +}; + +char *math_scalar[2] = { + [0] = "", + [1] = "scalar" +}; + +char *math_precision[2] = { + [0] = "", + [1] = "partial_precision" +}; + +char *urb_swizzle[4] = { + [BRW_URB_SWIZZLE_NONE] = "", + [BRW_URB_SWIZZLE_INTERLEAVE] = "interleave", + [BRW_URB_SWIZZLE_TRANSPOSE] = "transpose", +}; + +char *urb_allocate[2] = { + [0] = "", + [1] = "allocate" +}; + +char *urb_used[2] = { + [0] = "", + [1] = "used" +}; + +char *urb_complete[2] = { + [0] = "", + [1] = "complete" +}; + +char *sampler_target_format[4] = { + [0] = "F", + [2] = "UD", + [3] = "D" +}; + + +static int column; + +static int string (FILE *file, char *string) +{ + fputs (string, file); + column += strlen (string); + return 0; +} + +static int format (FILE *f, char *format, ...) +{ + char buf[1024]; + va_list args; + va_start (args, format); + + vsnprintf (buf, sizeof (buf) - 1, format, args); + string (f, buf); + return 0; +} + +static int newline (FILE *f) +{ + putc ('\n', f); + column = 0; + return 0; +} + +static int pad (FILE *f, int c) +{ + do + string (f, " "); + while (column < c); + return 0; +} + +static int control (FILE *file, char *name, char *ctrl[], GLuint id, int *space) +{ + if (!ctrl[id]) { + fprintf (file, "*** invalid %s value %d ", + name, id); + return 1; + } + if (ctrl[id][0]) + { + if (space && *space) + string (file, " "); + string (file, ctrl[id]); + if (space) + *space = 1; + } + return 0; +} + +static int print_opcode (FILE *file, int id) +{ + if (!opcode[id].name) { + format (file, "*** invalid opcode value %d ", id); + return 1; + } + string (file, opcode[id].name); + return 0; +} + +static int reg (FILE *file, GLuint _reg_file, GLuint _reg_nr) +{ + int err = 0; + if (_reg_file == BRW_ARCHITECTURE_REGISTER_FILE) { + switch (_reg_nr & 0xf0) { + case BRW_ARF_NULL: + string (file, "null"); + return -1; + case BRW_ARF_ADDRESS: + format (file, "a%d", _reg_nr & 0x0f); + break; + case BRW_ARF_ACCUMULATOR: + format (file, "acc%d", _reg_nr & 0x0f); + break; + case BRW_ARF_MASK: + format (file, "mask%d", _reg_nr & 0x0f); + break; + case BRW_ARF_MASK_STACK: + format (file, "msd%d", _reg_nr & 0x0f); + break; + case BRW_ARF_STATE: + format (file, "sr%d", _reg_nr & 0x0f); + break; + case BRW_ARF_CONTROL: + format (file, "cr%d", _reg_nr & 0x0f); + break; + case BRW_ARF_NOTIFICATION_COUNT: + format (file, "n%d", _reg_nr & 0x0f); + break; + case BRW_ARF_IP: + string (file, "ip"); + return -1; + break; + default: + format (file, "ARF%d", _reg_nr); + break; + } + } else { + err |= control (file, "src reg file", reg_file, _reg_file, NULL); + format (file, "%d", _reg_nr); + } + return err; +} + +static int dest (FILE *file, const struct brw_instruction *inst) +{ + int err = 0; + + if (inst->header.access_mode == BRW_ALIGN_1) + { + if (inst->bits1.da1.dest_address_mode == BRW_ADDRESS_DIRECT) + { + err |= reg (file, inst->bits1.da1.dest_reg_file, inst->bits1.da1.dest_reg_nr); + if (err == -1) + return 0; + if (inst->bits1.da1.dest_subreg_nr) + format (file, ".%d", inst->bits1.da1.dest_subreg_nr); + format (file, "<%d>", inst->bits1.da1.dest_horiz_stride); + err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.da1.dest_reg_type, NULL); + } + else + { + string (file, "g[a0"); + if (inst->bits1.ia1.dest_subreg_nr) + format (file, ".%d", inst->bits1.ia1.dest_subreg_nr); + if (inst->bits1.ia1.dest_indirect_offset) + format (file, " %d", inst->bits1.ia1.dest_indirect_offset); + string (file, "]"); + format (file, "<%d>", inst->bits1.ia1.dest_horiz_stride); + err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.ia1.dest_reg_type, NULL); + } + } + else + { + if (inst->bits1.da16.dest_address_mode == BRW_ADDRESS_DIRECT) + { + err |= reg (file, inst->bits1.da16.dest_reg_file, inst->bits1.da16.dest_reg_nr); + if (err == -1) + return 0; + if (inst->bits1.da16.dest_subreg_nr) + format (file, ".%d", inst->bits1.da16.dest_subreg_nr); + string (file, "<1>"); + err |= control (file, "writemask", writemask, inst->bits1.da16.dest_writemask, NULL); + err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.da16.dest_reg_type, NULL); + } + else + { + err = 1; + string (file, "Indirect align16 address mode not supported"); + } + } + + return 0; +} + +static int src_align1_region (FILE *file, + GLuint _vert_stride, GLuint _width, GLuint _horiz_stride) +{ + int err = 0; + string (file, "<"); + err |= control (file, "vert stride", vert_stride, _vert_stride, NULL); + string (file, ","); + err |= control (file, "width", width, _width, NULL); + string (file, ","); + err |= control (file, "horiz_stride", horiz_stride, _horiz_stride, NULL); + string (file, ">"); + return err; +} + +static int src_da1 (FILE *file, GLuint type, GLuint _reg_file, + GLuint _vert_stride, GLuint _width, GLuint _horiz_stride, + GLuint reg_num, GLuint sub_reg_num, GLuint __abs, GLuint _negate) +{ + int err = 0; + err |= control (file, "negate", negate, _negate, NULL); + err |= control (file, "abs", _abs, __abs, NULL); + + err |= reg (file, _reg_file, reg_num); + if (err == -1) + return 0; + if (sub_reg_num) + format (file, ".%d", sub_reg_num); + src_align1_region (file, _vert_stride, _width, _horiz_stride); + err |= control (file, "src reg encoding", reg_encoding, type, NULL); + return err; +} + +static int src_ia1 (FILE *file, + GLuint type, + GLuint _reg_file, + GLint _addr_imm, + GLuint _addr_subreg_nr, + GLuint _negate, + GLuint __abs, + GLuint _addr_mode, + GLuint _horiz_stride, + GLuint _width, + GLuint _vert_stride) +{ + int err = 0; + err |= control (file, "negate", negate, _negate, NULL); + err |= control (file, "abs", _abs, __abs, NULL); + + string (file, "g[a0"); + if (_addr_subreg_nr) + format (file, ".%d", _addr_subreg_nr); + if (_addr_imm) + format (file, " %d", _addr_imm); + string (file, "]"); + src_align1_region (file, _vert_stride, _width, _horiz_stride); + err |= control (file, "src reg encoding", reg_encoding, type, NULL); + return err; +} + +static int src_da16 (FILE *file, + GLuint _reg_type, + GLuint _reg_file, + GLuint _vert_stride, + GLuint _reg_nr, + GLuint _subreg_nr, + GLuint __abs, + GLuint _negate, + GLuint swz_x, + GLuint swz_y, + GLuint swz_z, + GLuint swz_w) +{ + int err = 0; + err |= control (file, "negate", negate, _negate, NULL); + err |= control (file, "abs", _abs, __abs, NULL); + + err |= reg (file, _reg_file, _reg_nr); + if (err == -1) + return 0; + if (_subreg_nr) + format (file, ".%d", _subreg_nr); + string (file, "<"); + err |= control (file, "vert stride", vert_stride, _vert_stride, NULL); + string (file, ",1,1>"); + err |= control (file, "src da16 reg type", reg_encoding, _reg_type, NULL); + /* + * Three kinds of swizzle display: + * identity - nothing printed + * 1->all - print the single channel + * 1->1 - print the mapping + */ + if (swz_x == BRW_CHANNEL_X && + swz_y == BRW_CHANNEL_Y && + swz_z == BRW_CHANNEL_Z && + swz_w == BRW_CHANNEL_W) + { + ; + } + else if (swz_x == swz_y && swz_x == swz_z && swz_x == swz_w) + { + string (file, "."); + err |= control (file, "channel select", chan_sel, swz_x, NULL); + } + else + { + string (file, "."); + err |= control (file, "channel select", chan_sel, swz_x, NULL); + err |= control (file, "channel select", chan_sel, swz_y, NULL); + err |= control (file, "channel select", chan_sel, swz_z, NULL); + err |= control (file, "channel select", chan_sel, swz_w, NULL); + } + return err; +} + + +static int imm (FILE *file, GLuint type, const struct brw_instruction *inst) { + switch (type) { + case BRW_REGISTER_TYPE_UD: + format (file, "0x%08xUD", inst->bits3.ud); + break; + case BRW_REGISTER_TYPE_D: + format (file, "%dD", inst->bits3.d); + break; + case BRW_REGISTER_TYPE_UW: + format (file, "0x%04xUW", (uint16_t) inst->bits3.ud); + break; + case BRW_REGISTER_TYPE_W: + format (file, "%dW", (int16_t) inst->bits3.d); + break; + case BRW_REGISTER_TYPE_UB: + format (file, "0x%02xUB", (int8_t) inst->bits3.ud); + break; + case BRW_REGISTER_TYPE_VF: + format (file, "Vector Float"); + break; + case BRW_REGISTER_TYPE_V: + format (file, "0x%08xV", inst->bits3.ud); + break; + case BRW_REGISTER_TYPE_F: + format (file, "%-gF", inst->bits3.f); + } + return 0; +} + +static int src0 (FILE *file, const struct brw_instruction *inst) +{ + if (inst->bits1.da1.src0_reg_file == BRW_IMMEDIATE_VALUE) + return imm (file, inst->bits1.da1.src0_reg_type, + inst); + else if (inst->header.access_mode == BRW_ALIGN_1) + { + if (inst->bits2.da1.src0_address_mode == BRW_ADDRESS_DIRECT) + { + return src_da1 (file, + inst->bits1.da1.src0_reg_type, + inst->bits1.da1.src0_reg_file, + inst->bits2.da1.src0_vert_stride, + inst->bits2.da1.src0_width, + inst->bits2.da1.src0_horiz_stride, + inst->bits2.da1.src0_reg_nr, + inst->bits2.da1.src0_subreg_nr, + inst->bits2.da1.src0_abs, + inst->bits2.da1.src0_negate); + } + else + { + return src_ia1 (file, + inst->bits1.ia1.src0_reg_type, + inst->bits1.ia1.src0_reg_file, + inst->bits2.ia1.src0_indirect_offset, + inst->bits2.ia1.src0_subreg_nr, + inst->bits2.ia1.src0_negate, + inst->bits2.ia1.src0_abs, + inst->bits2.ia1.src0_address_mode, + inst->bits2.ia1.src0_horiz_stride, + inst->bits2.ia1.src0_width, + inst->bits2.ia1.src0_vert_stride); + } + } + else + { + if (inst->bits2.da16.src0_address_mode == BRW_ADDRESS_DIRECT) + { + return src_da16 (file, + inst->bits1.da16.src0_reg_type, + inst->bits1.da16.src0_reg_file, + inst->bits2.da16.src0_vert_stride, + inst->bits2.da16.src0_reg_nr, + inst->bits2.da16.src0_subreg_nr, + inst->bits2.da16.src0_abs, + inst->bits2.da16.src0_negate, + inst->bits2.da16.src0_swz_x, + inst->bits2.da16.src0_swz_y, + inst->bits2.da16.src0_swz_z, + inst->bits2.da16.src0_swz_w); + } + else + { + string (file, "Indirect align16 address mode not supported"); + return 1; + } + } +} + +static int src1 (FILE *file, const struct brw_instruction *inst) +{ + if (inst->bits1.da1.src1_reg_file == BRW_IMMEDIATE_VALUE) + return imm (file, inst->bits1.da1.src1_reg_type, + inst); + else if (inst->header.access_mode == BRW_ALIGN_1) + { + if (inst->bits3.da1.src1_address_mode == BRW_ADDRESS_DIRECT) + { + return src_da1 (file, + inst->bits1.da1.src1_reg_type, + inst->bits1.da1.src1_reg_file, + inst->bits3.da1.src1_vert_stride, + inst->bits3.da1.src1_width, + inst->bits3.da1.src1_horiz_stride, + inst->bits3.da1.src1_reg_nr, + inst->bits3.da1.src1_subreg_nr, + inst->bits3.da1.src1_abs, + inst->bits3.da1.src1_negate); + } + else + { + return src_ia1 (file, + inst->bits1.ia1.src1_reg_type, + inst->bits1.ia1.src1_reg_file, + inst->bits3.ia1.src1_indirect_offset, + inst->bits3.ia1.src1_subreg_nr, + inst->bits3.ia1.src1_negate, + inst->bits3.ia1.src1_abs, + inst->bits3.ia1.src1_address_mode, + inst->bits3.ia1.src1_horiz_stride, + inst->bits3.ia1.src1_width, + inst->bits3.ia1.src1_vert_stride); + } + } + else + { + if (inst->bits3.da16.src1_address_mode == BRW_ADDRESS_DIRECT) + { + return src_da16 (file, + inst->bits1.da16.src1_reg_type, + inst->bits1.da16.src1_reg_file, + inst->bits3.da16.src1_vert_stride, + inst->bits3.da16.src1_reg_nr, + inst->bits3.da16.src1_subreg_nr, + inst->bits3.da16.src1_abs, + inst->bits3.da16.src1_negate, + inst->bits3.da16.src1_swz_x, + inst->bits3.da16.src1_swz_y, + inst->bits3.da16.src1_swz_z, + inst->bits3.da16.src1_swz_w); + } + else + { + string (file, "Indirect align16 address mode not supported"); + return 1; + } + } +} + +int brw_disasm_insn (FILE *file, const struct brw_instruction *inst) +{ + int err = 0; + int space = 0; + + if (inst->header.predicate_control) { + string (file, "("); + err |= control (file, "predicate inverse", pred_inv, inst->header.predicate_inverse, NULL); + string (file, "f0"); + if (inst->bits2.da1.flag_reg_nr) + format (file, ".%d", inst->bits2.da1.flag_reg_nr); + if (inst->header.access_mode == BRW_ALIGN_1) + err |= control (file, "predicate control align1", pred_ctrl_align1, + inst->header.predicate_control, NULL); + else + err |= control (file, "predicate control align16", pred_ctrl_align16, + inst->header.predicate_control, NULL); + string (file, ") "); + } + + err |= print_opcode (file, inst->header.opcode); + err |= control (file, "saturate", saturate, inst->header.saturate, NULL); + err |= control (file, "debug control", debug_ctrl, inst->header.debug_control, NULL); + + if (inst->header.opcode != BRW_OPCODE_SEND) + err |= control (file, "conditional modifier", conditional_modifier, + inst->header.destreg__conditionalmod, NULL); + + if (inst->header.opcode != BRW_OPCODE_NOP) { + string (file, "("); + err |= control (file, "execution size", exec_size, inst->header.execution_size, NULL); + string (file, ")"); + } + + if (inst->header.opcode == BRW_OPCODE_SEND) + format (file, " %d", inst->header.destreg__conditionalmod); + + if (opcode[inst->header.opcode].ndst > 0) { + pad (file, 16); + err |= dest (file, inst); + } + if (opcode[inst->header.opcode].nsrc > 0) { + pad (file, 32); + err |= src0 (file, inst); + } + if (opcode[inst->header.opcode].nsrc > 1) { + pad (file, 48); + err |= src1 (file, inst); + } + + if (inst->header.opcode == BRW_OPCODE_SEND) { + newline (file); + pad (file, 16); + space = 0; + err |= control (file, "target function", target_function, + inst->bits3.generic.msg_target, &space); + switch (inst->bits3.generic.msg_target) { + case BRW_MESSAGE_TARGET_MATH: + err |= control (file, "math function", math_function, + inst->bits3.math.function, &space); + err |= control (file, "math saturate", math_saturate, + inst->bits3.math.saturate, &space); + err |= control (file, "math signed", math_signed, + inst->bits3.math.int_type, &space); + err |= control (file, "math scalar", math_scalar, + inst->bits3.math.data_type, &space); + err |= control (file, "math precision", math_precision, + inst->bits3.math.precision, &space); + break; + case BRW_MESSAGE_TARGET_SAMPLER: + format (file, " (%d, %d, ", + inst->bits3.sampler.binding_table_index, + inst->bits3.sampler.sampler); + err |= control (file, "sampler target format", sampler_target_format, + inst->bits3.sampler.return_format, NULL); + string (file, ")"); + break; + case BRW_MESSAGE_TARGET_DATAPORT_WRITE: + format (file, " (%d, %d, %d, %d)", + inst->bits3.dp_write.binding_table_index, + (inst->bits3.dp_write.pixel_scoreboard_clear << 3) | + inst->bits3.dp_write.msg_control, + inst->bits3.dp_write.msg_type, + inst->bits3.dp_write.send_commit_msg); + break; + case BRW_MESSAGE_TARGET_URB: + format (file, " %d", inst->bits3.urb.offset); + space = 1; + err |= control (file, "urb swizzle", urb_swizzle, + inst->bits3.urb.swizzle_control, &space); + err |= control (file, "urb allocate", urb_allocate, + inst->bits3.urb.allocate, &space); + err |= control (file, "urb used", urb_used, + inst->bits3.urb.used, &space); + err |= control (file, "urb complete", urb_complete, + inst->bits3.urb.complete, &space); + break; + case BRW_MESSAGE_TARGET_THREAD_SPAWNER: + break; + default: + format (file, "unsupported target %d", inst->bits3.generic.msg_target); + break; + } + if (space) + string (file, " "); + format (file, "mlen %d", + inst->bits3.generic.msg_length); + format (file, " rlen %d", + inst->bits3.generic.response_length); + } + pad (file, 64); + if (inst->header.opcode != BRW_OPCODE_NOP) { + string (file, "{"); + space = 1; + err |= control(file, "access mode", access_mode, inst->header.access_mode, &space); + err |= control (file, "mask control", mask_ctrl, inst->header.mask_control, &space); + err |= control (file, "dependency control", dep_ctrl, inst->header.dependency_control, &space); + err |= control (file, "compression control", compr_ctrl, inst->header.compression_control, &space); + err |= control (file, "thread control", thread_ctrl, inst->header.thread_control, &space); + if (inst->header.opcode == BRW_OPCODE_SEND) + err |= control (file, "end of thread", end_of_thread, + inst->bits3.generic.end_of_thread, &space); + if (space) + string (file, " "); + string (file, "}"); + } + string (file, ";"); + newline (file); + return err; +} + + +int brw_disasm (FILE *file, + const struct brw_instruction *inst, + unsigned count) +{ + int i, err; + + for (i = 0; i < count; i++) { + err = brw_disasm_insn(stderr, &inst[i]); + if (err) + return err; + } + + fprintf(file, "\n"); + return 0; +} + diff --git a/src/gallium/drivers/i965/brw_disasm.h b/src/gallium/drivers/i965/brw_disasm.h new file mode 100644 index 00000000000..77d402d35e6 --- /dev/null +++ b/src/gallium/drivers/i965/brw_disasm.h @@ -0,0 +1,34 @@ +/* + * Copyright © 2008 Keith Packard + * + * Permission to use, copy, modify, distribute, and sell this software and its + * documentation for any purpose is hereby granted without fee, provided that + * the above copyright notice appear in all copies and that both that copyright + * notice and this permission notice appear in supporting documentation, and + * that the name of the copyright holders not be used in advertising or + * publicity pertaining to distribution of the software without specific, + * written prior permission. The copyright holders make no representations + * about the suitability of this software for any purpose. It is provided "as + * is" without express or implied warranty. + * + * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, + * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO + * EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR + * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE + * OF THIS SOFTWARE. + */ + +#ifndef BRW_DISASM_H +#define BRW_DISASM_H + +struct brw_instruction; + +int brw_disasm_insn (FILE *file, const struct brw_instruction *inst); +int brw_disasm (FILE *file, + const struct brw_instruction *inst, + unsigned count); + +#endif + diff --git a/src/gallium/drivers/i965/brw_draw.c b/src/gallium/drivers/i965/brw_draw.c new file mode 100644 index 00000000000..852fd229828 --- /dev/null +++ b/src/gallium/drivers/i965/brw_draw.c @@ -0,0 +1,291 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "util/u_prim.h" +#include "util/u_upload_mgr.h" + +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_debug.h" +#include "brw_screen.h" + +#include "brw_batchbuffer.h" + + +static uint32_t prim_to_hw_prim[PIPE_PRIM_POLYGON+1] = { + _3DPRIM_POINTLIST, + _3DPRIM_LINELIST, + _3DPRIM_LINELOOP, + _3DPRIM_LINESTRIP, + _3DPRIM_TRILIST, + _3DPRIM_TRISTRIP, + _3DPRIM_TRIFAN, + _3DPRIM_QUADLIST, + _3DPRIM_QUADSTRIP, + _3DPRIM_POLYGON +}; + + + +/* When the primitive changes, set a state bit and re-validate. Not + * the nicest and would rather deal with this by having all the + * programs be immune to the active primitive (ie. cope with all + * possibilities). That may not be realistic however. + */ +static int brw_set_prim(struct brw_context *brw, unsigned prim ) +{ + + if (BRW_DEBUG & DEBUG_PRIMS) + debug_printf("PRIM: %s\n", u_prim_name(prim)); + + if (prim != brw->primitive) { + unsigned reduced_prim; + + brw->primitive = prim; + brw->state.dirty.brw |= BRW_NEW_PRIMITIVE; + + reduced_prim = u_reduced_prim(prim); + if (reduced_prim != brw->reduced_primitive) { + brw->reduced_primitive = reduced_prim; + brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; + } + } + + return prim_to_hw_prim[prim]; +} + + + +static int brw_emit_prim(struct brw_context *brw, + unsigned start, + unsigned count, + boolean indexed, + uint32_t hw_prim) +{ + struct brw_3d_primitive prim_packet; + int ret; + + if (BRW_DEBUG & DEBUG_PRIMS) + debug_printf("%s start %d count %d indexed %d hw_prim %d\n", + __FUNCTION__, start, count, indexed, hw_prim); + + prim_packet.header.opcode = CMD_3D_PRIM; + prim_packet.header.length = sizeof(prim_packet)/4 - 2; + prim_packet.header.pad = 0; + prim_packet.header.topology = hw_prim; + prim_packet.header.indexed = indexed; + + prim_packet.verts_per_instance = count; + prim_packet.start_vert_location = start; + if (indexed) + prim_packet.start_vert_location += brw->ib.start_vertex_offset; + prim_packet.instance_count = 1; + prim_packet.start_instance_location = 0; + prim_packet.base_vert_location = 0; /* prim->basevertex; XXX: add this to gallium */ + + + /* If we're set to always flush, do it before and after the primitive emit. + * We want to catch both missed flushes that hurt instruction/state cache + * and missed flushes of the render cache as it heads to other parts of + * the besides the draw code. + */ + if (0) { + BEGIN_BATCH(1, IGNORE_CLIPRECTS); + OUT_BATCH((CMD_MI_FLUSH << 16) | BRW_FLUSH_STATE_CACHE); + ADVANCE_BATCH(); + } + if (prim_packet.verts_per_instance) { + ret = brw_batchbuffer_data( brw->batch, &prim_packet, + sizeof(prim_packet), LOOP_CLIPRECTS); + if (ret) + return ret; + } + if (0) { + BEGIN_BATCH(1, IGNORE_CLIPRECTS); + OUT_BATCH((CMD_MI_FLUSH << 16) | BRW_FLUSH_STATE_CACHE); + ADVANCE_BATCH(); + } + + return 0; +} + + +/* May fail if out of video memory for texture or vbo upload, or on + * fallback conditions. + */ +static int +try_draw_range_elements(struct brw_context *brw, + struct pipe_buffer *index_buffer, + unsigned hw_prim, + unsigned start, unsigned count) +{ + int ret; + + ret = brw_validate_state(brw); + if (ret) + return ret; + + /* Check that we can fit our state in with our existing batchbuffer, or + * flush otherwise. + */ + ret = brw->sws->check_aperture_space(brw->sws, + brw->state.validated_bos, + brw->state.validated_bo_count); + if (ret) + return ret; + + ret = brw_upload_state(brw); + if (ret) + return ret; + + ret = brw_emit_prim(brw, start, count, index_buffer != NULL, hw_prim); + if (ret) + return ret; + + if (brw->flags.always_flush_batch) + brw_context_flush( brw ); + + return 0; +} + + +static boolean +brw_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *index_buffer, + unsigned index_size, + unsigned min_index, + unsigned max_index, + unsigned mode, unsigned start, unsigned count) +{ + struct brw_context *brw = brw_context(pipe); + int ret; + uint32_t hw_prim; + + hw_prim = brw_set_prim(brw, mode); + + if (BRW_DEBUG & DEBUG_PRIMS) + debug_printf("PRIM: %s start %d count %d index_buffer %p\n", + u_prim_name(mode), start, count, (void *)index_buffer); + + /* Potentially trigger upload of new index buffer. + * + * XXX: do we need to go through state validation to achieve this? + * Could just call upload code directly. + */ + if (brw->curr.index_buffer != index_buffer || + brw->curr.index_size != index_size) { + pipe_buffer_reference( &brw->curr.index_buffer, index_buffer ); + brw->curr.index_size = index_size; + brw->state.dirty.mesa |= PIPE_NEW_INDEX_BUFFER; + } + + /* XXX: do we really care? + */ + if (brw->curr.min_index != min_index || + brw->curr.max_index != max_index) + { + brw->curr.min_index = min_index; + brw->curr.max_index = max_index; + brw->state.dirty.mesa |= PIPE_NEW_INDEX_RANGE; + } + + + /* Make a first attempt at drawing: + */ + ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count ); + + /* Otherwise, flush and retry: + */ + if (ret != 0) { + brw_context_flush( brw ); + ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count ); + assert(ret == 0); + } + + return TRUE; +} + +static boolean +brw_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *index_buffer, + unsigned index_size, + unsigned mode, + unsigned start, unsigned count) +{ + return brw_draw_range_elements( pipe, index_buffer, + index_size, + 0, 0xffffffff, + mode, + start, count ); +} + +static boolean +brw_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count) +{ + return brw_draw_elements(pipe, NULL, 0, mode, start, count); +} + + + +boolean brw_draw_init( struct brw_context *brw ) +{ + /* Register our drawing function: + */ + brw->base.draw_arrays = brw_draw_arrays; + brw->base.draw_elements = brw_draw_elements; + brw->base.draw_range_elements = brw_draw_range_elements; + + /* Create helpers for uploading data in user buffers: + */ + brw->vb.upload_vertex = u_upload_create( brw->base.screen, + 128 * 1024, + 64, + PIPE_BUFFER_USAGE_VERTEX ); + if (brw->vb.upload_vertex == NULL) + return FALSE; + + brw->vb.upload_index = u_upload_create( brw->base.screen, + 32 * 1024, + 64, + PIPE_BUFFER_USAGE_INDEX ); + if (brw->vb.upload_index == NULL) + return FALSE; + + return TRUE; +} + +void brw_draw_cleanup( struct brw_context *brw ) +{ + u_upload_destroy( brw->vb.upload_vertex ); + u_upload_destroy( brw->vb.upload_index ); + + bo_reference(&brw->ib.bo, NULL); +} diff --git a/src/gallium/drivers/i965/brw_draw.h b/src/gallium/drivers/i965/brw_draw.h new file mode 100644 index 00000000000..8dc5dbce622 --- /dev/null +++ b/src/gallium/drivers/i965/brw_draw.h @@ -0,0 +1,39 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_DRAW_H +#define BRW_DRAW_H + +#include "brw_types.h" + +struct brw_context; + +boolean brw_draw_init( struct brw_context *brw ); +void brw_draw_cleanup( struct brw_context *brw ); + + +#endif diff --git a/src/gallium/drivers/i965/brw_draw_upload.c b/src/gallium/drivers/i965/brw_draw_upload.c new file mode 100644 index 00000000000..a27da5f1c17 --- /dev/null +++ b/src/gallium/drivers/i965/brw_draw_upload.c @@ -0,0 +1,542 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_context.h" + +#include "util/u_upload_mgr.h" +#include "util/u_math.h" + +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_screen.h" +#include "brw_batchbuffer.h" +#include "brw_debug.h" + + + + +static unsigned brw_translate_surface_format( unsigned id ) +{ + switch (id) { + case PIPE_FORMAT_R64_FLOAT: + return BRW_SURFACEFORMAT_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return BRW_SURFACEFORMAT_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return BRW_SURFACEFORMAT_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return BRW_SURFACEFORMAT_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return BRW_SURFACEFORMAT_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return BRW_SURFACEFORMAT_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return BRW_SURFACEFORMAT_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return BRW_SURFACEFORMAT_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return BRW_SURFACEFORMAT_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return BRW_SURFACEFORMAT_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return BRW_SURFACEFORMAT_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return BRW_SURFACEFORMAT_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return BRW_SURFACEFORMAT_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return BRW_SURFACEFORMAT_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return BRW_SURFACEFORMAT_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return BRW_SURFACEFORMAT_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return BRW_SURFACEFORMAT_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return BRW_SURFACEFORMAT_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return BRW_SURFACEFORMAT_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return BRW_SURFACEFORMAT_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return BRW_SURFACEFORMAT_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return BRW_SURFACEFORMAT_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return BRW_SURFACEFORMAT_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return BRW_SURFACEFORMAT_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return BRW_SURFACEFORMAT_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return BRW_SURFACEFORMAT_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return BRW_SURFACEFORMAT_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return BRW_SURFACEFORMAT_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED; + + default: + assert(0); + return 0; + } +} + +static unsigned get_index_type(int type) +{ + switch (type) { + case 1: return BRW_INDEX_BYTE; + case 2: return BRW_INDEX_WORD; + case 4: return BRW_INDEX_DWORD; + default: assert(0); return 0; + } +} + + +static int brw_prepare_vertices(struct brw_context *brw) +{ + unsigned int min_index = brw->curr.min_index; + unsigned int max_index = brw->curr.max_index; + GLuint i; + int ret; + + if (BRW_DEBUG & DEBUG_VERTS) + debug_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index); + + + for (i = 0; i < brw->curr.num_vertex_buffers; i++) { + struct pipe_vertex_buffer *vb = &brw->curr.vertex_buffer[i]; + struct brw_winsys_buffer *bo; + struct pipe_buffer *upload_buf = NULL; + unsigned offset; + + if (BRW_DEBUG & DEBUG_VERTS) + debug_printf("%s vb[%d] user:%d offset:0x%x sz:0x%x stride:0x%x\n", + __FUNCTION__, i, + brw_buffer_is_user_buffer(vb->buffer), + vb->buffer_offset, + vb->buffer->size, + vb->stride); + + if (brw_buffer_is_user_buffer(vb->buffer)) { + + /* XXX: simplify this. Stop the state trackers from generating + * zero-stride buffers & have them use additional constants (or + * add support for >1 constant buffer) instead. + */ + unsigned size = (vb->stride == 0 ? + vb->buffer->size - vb->buffer_offset : + MAX2(vb->buffer->size - vb->buffer_offset, + vb->stride * (max_index + 1 - min_index))); + + ret = u_upload_buffer( brw->vb.upload_vertex, + vb->buffer_offset + min_index * vb->stride, + size, + vb->buffer, + &offset, + &upload_buf ); + if (ret) + return ret; + + bo = brw_buffer(upload_buf)->bo; + + assert(offset + size <= bo->size); + } + else + { + offset = vb->buffer_offset; + bo = brw_buffer(vb->buffer)->bo; + } + + assert(offset < bo->size); + + /* Set up post-upload info about this vertex buffer: + */ + brw->vb.vb[i].offset = offset; + brw->vb.vb[i].stride = vb->stride; + brw->vb.vb[i].vertex_count = (vb->stride == 0 ? + 1 : + (bo->size - offset) / vb->stride); + + bo_reference( &brw->vb.vb[i].bo, bo ); + + /* Don't need to retain this reference. We have a reference on + * the underlying winsys buffer: + */ + pipe_buffer_reference( &upload_buf, NULL ); + } + + brw->vb.nr_vb = i; + brw_prepare_query_begin(brw); + + for (i = 0; i < brw->vb.nr_vb; i++) { + brw_add_validated_bo(brw, brw->vb.vb[i].bo); + } + + return 0; +} + +static int brw_emit_vertex_buffers( struct brw_context *brw ) +{ + int i; + + /* If the VS doesn't read any inputs (calculating vertex position from + * a state variable for some reason, for example), just bail. + * + * The stale VB state stays in place, but they don't do anything unless + * a VE loads from them. + */ + if (brw->vb.nr_vb == 0) { + if (BRW_DEBUG & DEBUG_VERTS) + debug_printf("%s: no active vertex buffers\n", __FUNCTION__); + + return 0; + } + + /* Emit VB state packets. + */ + BEGIN_BATCH(1 + brw->vb.nr_vb * 4, IGNORE_CLIPRECTS); + OUT_BATCH((CMD_VERTEX_BUFFER << 16) | + ((1 + brw->vb.nr_vb * 4) - 2)); + + for (i = 0; i < brw->vb.nr_vb; i++) { + OUT_BATCH((i << BRW_VB0_INDEX_SHIFT) | + BRW_VB0_ACCESS_VERTEXDATA | + (brw->vb.vb[i].stride << BRW_VB0_PITCH_SHIFT)); + OUT_RELOC(brw->vb.vb[i].bo, + BRW_USAGE_VERTEX, + brw->vb.vb[i].offset); + if (BRW_IS_IGDNG(brw)) { + OUT_RELOC(brw->vb.vb[i].bo, + BRW_USAGE_VERTEX, + brw->vb.vb[i].bo->size - 1); + } else + OUT_BATCH(brw->vb.vb[i].stride ? brw->vb.vb[i].vertex_count : 0); + OUT_BATCH(0); /* Instance data step rate */ + } + ADVANCE_BATCH(); + return 0; +} + + + + +static int brw_emit_vertex_elements(struct brw_context *brw) +{ + GLuint nr = brw->curr.num_vertex_elements; + GLuint i; + + brw_emit_query_begin(brw); + + /* If the VS doesn't read any inputs (calculating vertex position from + * a state variable for some reason, for example), emit a single pad + * VERTEX_ELEMENT struct and bail. + * + * The stale VB state stays in place, but they don't do anything unless + * a VE loads from them. + */ + if (nr == 0) { + BEGIN_BATCH(3, IGNORE_CLIPRECTS); + OUT_BATCH((CMD_VERTEX_ELEMENT << 16) | 1); + OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) | + BRW_VE0_VALID | + (BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) | + (0 << BRW_VE0_SRC_OFFSET_SHIFT)); + OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) | + (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) | + (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) | + (BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT)); + ADVANCE_BATCH(); + return 0; + } + + /* Now emit vertex element (VEP) state packets. + * + */ + BEGIN_BATCH(1 + nr * 2, IGNORE_CLIPRECTS); + OUT_BATCH((CMD_VERTEX_ELEMENT << 16) | ((1 + nr * 2) - 2)); + for (i = 0; i < nr; i++) { + const struct pipe_vertex_element *input = &brw->curr.vertex_element[i]; + uint32_t format = brw_translate_surface_format( input->src_format ); + uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC; + uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC; + uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC; + uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC; + + switch (input->nr_components) { + case 0: comp0 = BRW_VE1_COMPONENT_STORE_0; + case 1: comp1 = BRW_VE1_COMPONENT_STORE_0; + case 2: comp2 = BRW_VE1_COMPONENT_STORE_0; + case 3: comp3 = BRW_VE1_COMPONENT_STORE_1_FLT; + break; + } + + OUT_BATCH((input->vertex_buffer_index << BRW_VE0_INDEX_SHIFT) | + BRW_VE0_VALID | + (format << BRW_VE0_FORMAT_SHIFT) | + (input->src_offset << BRW_VE0_SRC_OFFSET_SHIFT)); + + if (BRW_IS_IGDNG(brw)) + OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) | + (comp1 << BRW_VE1_COMPONENT_1_SHIFT) | + (comp2 << BRW_VE1_COMPONENT_2_SHIFT) | + (comp3 << BRW_VE1_COMPONENT_3_SHIFT)); + else + OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) | + (comp1 << BRW_VE1_COMPONENT_1_SHIFT) | + (comp2 << BRW_VE1_COMPONENT_2_SHIFT) | + (comp3 << BRW_VE1_COMPONENT_3_SHIFT) | + ((i * 4) << BRW_VE1_DST_OFFSET_SHIFT)); + } + ADVANCE_BATCH(); + return 0; +} + + +static int brw_emit_vertices( struct brw_context *brw ) +{ + int ret; + + ret = brw_emit_vertex_buffers( brw ); + if (ret) + return ret; + + ret = brw_emit_vertex_elements( brw ); + if (ret) + return ret; + + return 0; +} + + +const struct brw_tracked_state brw_vertices = { + .dirty = { + .mesa = (PIPE_NEW_INDEX_RANGE | + PIPE_NEW_VERTEX_BUFFER), + .brw = BRW_NEW_BATCH, + .cache = 0, + }, + .prepare = brw_prepare_vertices, + .emit = brw_emit_vertices, +}; + + +static int brw_prepare_indices(struct brw_context *brw) +{ + struct pipe_buffer *index_buffer = brw->curr.index_buffer; + struct pipe_buffer *upload_buf = NULL; + struct brw_winsys_buffer *bo = NULL; + GLuint offset; + GLuint index_size; + GLuint ib_size; + int ret; + + if (index_buffer == NULL) + return 0; + + if (BRW_DEBUG & DEBUG_VERTS) + debug_printf("%s: index_size:%d index_buffer->size:%d\n", + __FUNCTION__, + brw->curr.index_size, + brw->curr.index_buffer->size); + + ib_size = index_buffer->size; + index_size = brw->curr.index_size; + + /* Turn userbuffer into a proper hardware buffer? + */ + if (brw_buffer_is_user_buffer(index_buffer)) { + + ret = u_upload_buffer( brw->vb.upload_index, + 0, + ib_size, + index_buffer, + &offset, + &upload_buf ); + if (ret) + return ret; + + bo = brw_buffer(upload_buf)->bo; + + /* XXX: annotate the userbuffer with the upload information so + * that successive calls don't get re-uploaded. + */ + } + else { + bo = brw_buffer(index_buffer)->bo; + ib_size = bo->size; + offset = 0; + } + + /* Use CMD_3D_PRIM's start_vertex_offset to avoid re-uploading the + * index buffer state when we're just moving the start index of our + * drawing. + * + * In gallium this will happen in the case where successive draw + * calls are made with (distinct?) userbuffers, but the upload_mgr + * places the data into a single winsys buffer. + * + * This statechange doesn't raise any state flags and is always + * just merged into the final draw packet: + */ + if (1) { + assert((offset & (index_size - 1)) == 0); + brw->ib.start_vertex_offset = offset / index_size; + } + + /* These statechanges trigger a new CMD_INDEX_BUFFER packet: + */ + if (brw->ib.bo != bo || + brw->ib.size != ib_size) + { + bo_reference(&brw->ib.bo, bo); + brw->ib.size = ib_size; + brw->state.dirty.brw |= BRW_NEW_INDEX_BUFFER; + } + + pipe_buffer_reference( &upload_buf, NULL ); + brw_add_validated_bo(brw, brw->ib.bo); + return 0; +} + +const struct brw_tracked_state brw_indices = { + .dirty = { + .mesa = PIPE_NEW_INDEX_BUFFER, + .brw = 0, + .cache = 0, + }, + .prepare = brw_prepare_indices, +}; + +static int brw_emit_index_buffer(struct brw_context *brw) +{ + /* Emit the indexbuffer packet: + */ + if (brw->ib.bo) + { + struct brw_indexbuffer ib; + + memset(&ib, 0, sizeof(ib)); + + ib.header.bits.opcode = CMD_INDEX_BUFFER; + ib.header.bits.length = sizeof(ib)/4 - 2; + ib.header.bits.index_format = get_index_type(brw->ib.size); + ib.header.bits.cut_index_enable = 0; + + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH( ib.header.dword ); + OUT_RELOC(brw->ib.bo, + BRW_USAGE_VERTEX, + brw->ib.offset); + OUT_RELOC(brw->ib.bo, + BRW_USAGE_VERTEX, + brw->ib.offset + brw->ib.size - 1); + OUT_BATCH( 0 ); + ADVANCE_BATCH(); + } + + return 0; +} + +const struct brw_tracked_state brw_index_buffer = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_BATCH | BRW_NEW_INDEX_BUFFER, + .cache = 0, + }, + .emit = brw_emit_index_buffer, +}; diff --git a/src/gallium/drivers/i965/brw_eu.c b/src/gallium/drivers/i965/brw_eu.c new file mode 100644 index 00000000000..a8fcb5f97eb --- /dev/null +++ b/src/gallium/drivers/i965/brw_eu.c @@ -0,0 +1,262 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + + +/* How does predicate control work when execution_size != 8? Do I + * need to test/set for 0xffff when execution_size is 16? + */ +void brw_set_predicate_control_flag_value( struct brw_compile *p, GLuint value ) +{ + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + if (value != 0xff) { + if (value != p->flag_value) { + brw_push_insn_state(p); + brw_MOV(p, brw_flag_reg(), brw_imm_uw(value)); + p->flag_value = value; + brw_pop_insn_state(p); + } + + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } +} + +void brw_set_predicate_control( struct brw_compile *p, GLuint pc ) +{ + p->current->header.predicate_control = pc; +} + +void brw_set_conditionalmod( struct brw_compile *p, GLuint conditional ) +{ + p->current->header.destreg__conditionalmod = conditional; +} + +void brw_set_access_mode( struct brw_compile *p, GLuint access_mode ) +{ + p->current->header.access_mode = access_mode; +} + +void brw_set_compression_control( struct brw_compile *p, GLboolean compression_control ) +{ + p->current->header.compression_control = compression_control; +} + +void brw_set_mask_control( struct brw_compile *p, GLuint value ) +{ + p->current->header.mask_control = value; +} + +void brw_set_saturate( struct brw_compile *p, GLuint value ) +{ + p->current->header.saturate = value; +} + +void brw_push_insn_state( struct brw_compile *p ) +{ + assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); + memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); + p->current++; +} + +void brw_pop_insn_state( struct brw_compile *p ) +{ + assert(p->current != p->stack); + p->current--; +} + + +/*********************************************************************** + */ +void brw_init_compile( struct brw_context *brw, struct brw_compile *p ) +{ + p->brw = brw; + p->nr_insn = 0; + p->current = p->stack; + memset(p->current, 0, sizeof(p->current[0])); + + /* Some defaults? + */ + brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ + brw_set_saturate(p, 0); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_predicate_control_flag_value(p, 0xff); +} + + +enum pipe_error brw_get_program( struct brw_compile *p, + const GLuint **data, + GLuint *sz ) +{ + GLuint i; + + for (i = 0; i < 8; i++) + brw_NOP(p); + + /* Is the generated program malformed for some reason? + */ + if (p->error) + return PIPE_ERROR_BAD_INPUT; + + *sz = p->nr_insn * sizeof(struct brw_instruction); + *data = (const GLuint *)p->store; + return PIPE_OK; +} + + + +/** + * Subroutine calls require special attention. + * Mesa instructions may be expanded into multiple hardware instructions + * so the prog_instruction::BranchTarget field can't be used as an index + * into the hardware instructions. + * + * The BranchTarget field isn't needed, however. Mesa's GLSL compiler + * emits CAL and BGNSUB instructions with labels that can be used to map + * subroutine calls to actual subroutine code blocks. + * + * The structures and function here implement patching of CAL instructions + * so they jump to the right subroutine code... + */ + + +/** + * For each OPCODE_BGNSUB we create one of these. + */ +struct brw_eu_label +{ + GLuint label; /**< the label number */ + GLuint position; /**< the position of the brw instruction for this label */ + struct brw_eu_label *next; /**< next in linked list */ +}; + + +/** + * For each OPCODE_CAL we create one of these. + */ +struct brw_eu_call +{ + GLuint call_inst_pos; /**< location of the CAL instruction */ + GLuint label; + struct brw_eu_call *next; /**< next in linked list */ +}; + + +/** + * Called for each OPCODE_BGNSUB. + */ +void +brw_save_label(struct brw_compile *c, unsigned l, GLuint position) +{ + struct brw_eu_label *label = CALLOC_STRUCT(brw_eu_label); + label->label = l; + label->position = position; + label->next = c->first_label; + c->first_label = label; +} + + +/** + * Called for each OPCODE_CAL. + */ +void +brw_save_call(struct brw_compile *c, GLuint label, GLuint call_pos) +{ + struct brw_eu_call *call = CALLOC_STRUCT(brw_eu_call); + call->call_inst_pos = call_pos; + call->label = label; + call->next = c->first_call; + c->first_call = call; +} + + +/** + * Lookup a label, return label's position/offset. + */ +static GLuint +brw_lookup_label(struct brw_compile *c, unsigned l) +{ + const struct brw_eu_label *label; + for (label = c->first_label; label; label = label->next) { + if (l == label->label) { + return label->position; + } + } + abort(); /* should never happen */ + return ~0; +} + + +/** + * When we're done generating code, this function is called to resolve + * subroutine calls. + */ +void +brw_resolve_cals(struct brw_compile *c) +{ + const struct brw_eu_call *call; + + for (call = c->first_call; call; call = call->next) { + const GLuint sub_loc = brw_lookup_label(c, call->label); + struct brw_instruction *brw_call_inst = &c->store[call->call_inst_pos]; + struct brw_instruction *brw_sub_inst = &c->store[sub_loc]; + GLint offset = brw_sub_inst - brw_call_inst; + + /* patch brw_inst1 to point to brw_inst2 */ + brw_set_src1(brw_call_inst, brw_imm_d(offset * 16)); + } + + /* free linked list of calls */ + { + struct brw_eu_call *call, *next; + for (call = c->first_call; call; call = next) { + next = call->next; + FREE(call); + } + c->first_call = NULL; + } + + /* free linked list of labels */ + { + struct brw_eu_label *label, *next; + for (label = c->first_label; label; label = next) { + next = label->next; + FREE(label); + } + c->first_label = NULL; + } +} diff --git a/src/gallium/drivers/i965/brw_eu.h b/src/gallium/drivers/i965/brw_eu.h new file mode 100644 index 00000000000..af509b2e5f4 --- /dev/null +++ b/src/gallium/drivers/i965/brw_eu.h @@ -0,0 +1,992 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_EU_H +#define BRW_EU_H + +#include "util/u_debug.h" +#include "pipe/p_defines.h" + +#include "brw_structs.h" +#include "brw_defines.h" + +#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6)) +#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3) + +#define BRW_SWIZZLE_NOOP BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XYZW BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XXXX BRW_SWIZZLE4(0,0,0,0) +#define BRW_SWIZZLE_XYXY BRW_SWIZZLE4(0,1,0,1) + +#define BRW_WRITEMASK_NONE 0x00 +#define BRW_WRITEMASK_X 0x01 +#define BRW_WRITEMASK_Y 0x02 +#define BRW_WRITEMASK_XY 0x03 +#define BRW_WRITEMASK_Z 0x04 +#define BRW_WRITEMASK_XZ 0x05 +#define BRW_WRITEMASK_YZ 0x06 +#define BRW_WRITEMASK_XYZ 0x07 +#define BRW_WRITEMASK_W 0x08 +#define BRW_WRITEMASK_XW 0x09 +#define BRW_WRITEMASK_YW 0x0A +#define BRW_WRITEMASK_XYW 0x0B +#define BRW_WRITEMASK_ZW 0x0C +#define BRW_WRITEMASK_XZW 0x0D +#define BRW_WRITEMASK_YZW 0x0E +#define BRW_WRITEMASK_XYZW 0x0F + + +#define REG_SIZE (8*4) + + +/* These aren't hardware structs, just something useful for us to pass around: + * + * Align1 operation has a lot of control over input ranges. Used in + * WM programs to implement shaders decomposed into "channel serial" + * or "structure of array" form: + */ +struct brw_reg +{ + GLuint type:4; + GLuint file:2; + GLuint nr:8; + GLuint subnr:5; /* :1 in align16 */ + GLuint negate:1; /* source only */ + GLuint abs:1; /* source only */ + GLuint vstride:4; /* source only */ + GLuint width:3; /* src only, align1 only */ + GLuint hstride:2; /* align1 only */ + GLuint address_mode:1; /* relative addressing, hopefully! */ + GLuint pad0:1; + + union { + struct { + GLuint swizzle:8; /* src only, align16 only */ + GLuint writemask:4; /* dest only, align16 only */ + GLint indirect_offset:10; /* relative addressing offset */ + GLuint pad1:10; /* two dwords total */ + } bits; + + GLfloat f; + GLint d; + GLuint ud; + } dw1; +}; + + +struct brw_indirect { + GLuint addr_subnr:4; + GLint addr_offset:10; + GLuint pad:18; +}; + + +struct brw_eu_label; +struct brw_eu_call; + + + +#define BRW_EU_MAX_INSN_STACK 5 +#define BRW_EU_MAX_INSN 10000 + +struct brw_compile { + struct brw_instruction store[BRW_EU_MAX_INSN]; + GLuint nr_insn; + + /* Allow clients to push/pop instruction state: + */ + struct brw_instruction stack[BRW_EU_MAX_INSN_STACK]; + struct brw_instruction *current; + + GLuint flag_value; + GLboolean single_program_flow; + struct brw_context *brw; + + struct brw_eu_label *first_label; /**< linked list of labels */ + struct brw_eu_call *first_call; /**< linked list of CALs */ + + boolean error; +}; + + +void +brw_save_label(struct brw_compile *c, unsigned label, GLuint position); + +void +brw_save_call(struct brw_compile *c, unsigned label, GLuint call_pos); + +void +brw_resolve_cals(struct brw_compile *c); + + + +static INLINE int type_sz( GLuint type ) +{ + switch( type ) { + case BRW_REGISTER_TYPE_UD: + case BRW_REGISTER_TYPE_D: + case BRW_REGISTER_TYPE_F: + return 4; + case BRW_REGISTER_TYPE_HF: + case BRW_REGISTER_TYPE_UW: + case BRW_REGISTER_TYPE_W: + return 2; + case BRW_REGISTER_TYPE_UB: + case BRW_REGISTER_TYPE_B: + return 1; + default: + return 0; + } +} + +/** + * Construct a brw_reg. + * \param file one of the BRW_x_REGISTER_FILE values + * \param nr register number/index + * \param subnr register sub number + * \param type one of BRW_REGISTER_TYPE_x + * \param vstride one of BRW_VERTICAL_STRIDE_x + * \param width one of BRW_WIDTH_x + * \param hstride one of BRW_HORIZONTAL_STRIDE_x + * \param swizzle one of BRW_SWIZZLE_x + * \param writemask BRW_WRITEMASK_X/Y/Z/W bitfield + */ +static INLINE struct brw_reg brw_reg( GLuint file, + GLuint nr, + GLuint subnr, + GLuint type, + GLuint vstride, + GLuint width, + GLuint hstride, + GLuint swizzle, + GLuint writemask ) +{ + struct brw_reg reg; + if (type == BRW_GENERAL_REGISTER_FILE) + assert(nr < BRW_MAX_GRF); + else if (type == BRW_MESSAGE_REGISTER_FILE) + assert(nr < BRW_MAX_MRF); + else if (type == BRW_ARCHITECTURE_REGISTER_FILE) + assert(nr <= BRW_ARF_IP); + + reg.type = type; + reg.file = file; + reg.nr = nr; + reg.subnr = subnr * type_sz(type); + reg.negate = 0; + reg.abs = 0; + reg.vstride = vstride; + reg.width = width; + reg.hstride = hstride; + reg.address_mode = BRW_ADDRESS_DIRECT; + reg.pad0 = 0; + + /* Could do better: If the reg is r5.3<0;1,0>, we probably want to + * set swizzle and writemask to W, as the lower bits of subnr will + * be lost when converted to align16. This is probably too much to + * keep track of as you'd want it adjusted by suboffset(), etc. + * Perhaps fix up when converting to align16? + */ + reg.dw1.bits.swizzle = swizzle; + reg.dw1.bits.writemask = writemask; + reg.dw1.bits.indirect_offset = 0; + reg.dw1.bits.pad1 = 0; + return reg; +} + +/** Construct float[16] register */ +static INLINE struct brw_reg brw_vec16_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_16, + BRW_WIDTH_16, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + BRW_WRITEMASK_XYZW); +} + +/** Construct float[8] register */ +static INLINE struct brw_reg brw_vec8_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + BRW_WRITEMASK_XYZW); +} + +/** Construct float[4] register */ +static INLINE struct brw_reg brw_vec4_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_4, + BRW_WIDTH_4, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + BRW_WRITEMASK_XYZW); +} + +/** Construct float[2] register */ +static INLINE struct brw_reg brw_vec2_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_2, + BRW_WIDTH_2, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYXY, + BRW_WRITEMASK_XY); +} + +/** Construct float[1] register */ +static INLINE struct brw_reg brw_vec1_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XXXX, + BRW_WRITEMASK_X); +} + + +static INLINE struct brw_reg retype( struct brw_reg reg, + GLuint type ) +{ + reg.type = type; + return reg; +} + +static INLINE struct brw_reg suboffset( struct brw_reg reg, + GLuint delta ) +{ + reg.subnr += delta * type_sz(reg.type); + return reg; +} + + +static INLINE struct brw_reg offset( struct brw_reg reg, + GLuint delta ) +{ + reg.nr += delta; + return reg; +} + + +static INLINE struct brw_reg byte_offset( struct brw_reg reg, + GLuint bytes ) +{ + GLuint newoffset = reg.nr * REG_SIZE + reg.subnr + bytes; + reg.nr = newoffset / REG_SIZE; + reg.subnr = newoffset % REG_SIZE; + return reg; +} + + +/** Construct unsigned word[16] register */ +static INLINE struct brw_reg brw_uw16_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +/** Construct unsigned word[8] register */ +static INLINE struct brw_reg brw_uw8_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +/** Construct unsigned word[1] register */ +static INLINE struct brw_reg brw_uw1_reg( GLuint file, + GLuint nr, + GLuint subnr ) +{ + return suboffset(retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static INLINE struct brw_reg brw_imm_reg( GLuint type ) +{ + return brw_reg( BRW_IMMEDIATE_VALUE, + 0, + 0, + type, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + 0, + 0); +} + +/** Construct float immediate register */ +static INLINE struct brw_reg brw_imm_f( GLfloat f ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F); + imm.dw1.f = f; + return imm; +} + +/** Construct integer immediate register */ +static INLINE struct brw_reg brw_imm_d( GLint d ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D); + imm.dw1.d = d; + return imm; +} + +/** Construct uint immediate register */ +static INLINE struct brw_reg brw_imm_ud( GLuint ud ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD); + imm.dw1.ud = ud; + return imm; +} + +/** Construct ushort immediate register */ +static INLINE struct brw_reg brw_imm_uw( GLushort uw ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW); + imm.dw1.ud = uw | (uw << 16); + return imm; +} + +/** Construct short immediate register */ +static INLINE struct brw_reg brw_imm_w( GLshort w ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W); + imm.dw1.d = w | (w << 16); + return imm; +} + +/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type + * numbers alias with _V and _VF below: + */ + +/** Construct vector of eight signed half-byte values */ +static INLINE struct brw_reg brw_imm_v( GLuint v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_8; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +/** Construct vector of four 8-bit float values */ +static INLINE struct brw_reg brw_imm_vf( GLuint v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +#define VF_ZERO 0x0 +#define VF_ONE 0x30 +#define VF_NEG (1<<7) + +static INLINE struct brw_reg brw_imm_vf4( GLuint v0, + GLuint v1, + GLuint v2, + GLuint v3) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = ((v0 << 0) | + (v1 << 8) | + (v2 << 16) | + (v3 << 24)); + return imm; +} + + +static INLINE struct brw_reg brw_address( struct brw_reg reg ) +{ + return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr); +} + +/** Construct float[1] general-purpose register */ +static INLINE struct brw_reg brw_vec1_grf( GLuint nr, GLuint subnr ) +{ + return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +/** Construct float[2] general-purpose register */ +static INLINE struct brw_reg brw_vec2_grf( GLuint nr, GLuint subnr ) +{ + return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +/** Construct float[4] general-purpose register */ +static INLINE struct brw_reg brw_vec4_grf( GLuint nr, GLuint subnr ) +{ + return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +/** Construct float[8] general-purpose register */ +static INLINE struct brw_reg brw_vec8_grf( GLuint nr, GLuint subnr ) +{ + return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + + +static INLINE struct brw_reg brw_uw8_grf( GLuint nr, GLuint subnr ) +{ + return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static INLINE struct brw_reg brw_uw16_grf( GLuint nr, GLuint subnr ) +{ + return brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + + +/** Construct null register (usually used for setting condition codes) */ +static INLINE struct brw_reg brw_null_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_NULL, + 0); +} + +static INLINE struct brw_reg brw_address_reg( GLuint subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ADDRESS, + subnr); +} + +/* If/else instructions break in align16 mode if writemask & swizzle + * aren't xyzw. This goes against the convention for other scalar + * regs: + */ +static INLINE struct brw_reg brw_ip_reg( void ) +{ + return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_IP, + 0, + BRW_REGISTER_TYPE_UD, + BRW_VERTICAL_STRIDE_4, /* ? */ + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, /* NOTE! */ + BRW_WRITEMASK_XYZW); /* NOTE! */ +} + +static INLINE struct brw_reg brw_acc_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ACCUMULATOR, + 0); +} + + +static INLINE struct brw_reg brw_flag_reg( void ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_FLAG, + 0); +} + + +static INLINE struct brw_reg brw_mask_reg( GLuint subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_MASK, + subnr); +} + +static INLINE struct brw_reg brw_message_reg( GLuint nr ) +{ + assert(nr < BRW_MAX_MRF); + return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, + nr, + 0); +} + + + + +/* This is almost always called with a numeric constant argument, so + * make things easy to evaluate at compile time: + */ +static INLINE GLuint cvt( GLuint val ) +{ + switch (val) { + case 0: return 0; + case 1: return 1; + case 2: return 2; + case 4: return 3; + case 8: return 4; + case 16: return 5; + case 32: return 6; + } + return 0; +} + +static INLINE struct brw_reg stride( struct brw_reg reg, + GLuint vstride, + GLuint width, + GLuint hstride ) +{ + reg.vstride = cvt(vstride); + reg.width = cvt(width) - 1; + reg.hstride = cvt(hstride); + return reg; +} + + +static INLINE struct brw_reg vec16( struct brw_reg reg ) +{ + return stride(reg, 16,16,1); +} + +static INLINE struct brw_reg vec8( struct brw_reg reg ) +{ + return stride(reg, 8,8,1); +} + +static INLINE struct brw_reg vec4( struct brw_reg reg ) +{ + return stride(reg, 4,4,1); +} + +static INLINE struct brw_reg vec2( struct brw_reg reg ) +{ + return stride(reg, 2,2,1); +} + +static INLINE struct brw_reg vec1( struct brw_reg reg ) +{ + return stride(reg, 0,1,0); +} + + +static INLINE struct brw_reg get_element( struct brw_reg reg, GLuint elt ) +{ + return vec1(suboffset(reg, elt)); +} + +static INLINE struct brw_reg get_element_ud( struct brw_reg reg, GLuint elt ) +{ + return vec1(suboffset(retype(reg, BRW_REGISTER_TYPE_UD), elt)); +} + + +static INLINE struct brw_reg brw_swizzle( struct brw_reg reg, + GLuint x, + GLuint y, + GLuint z, + GLuint w) +{ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x), + BRW_GET_SWZ(reg.dw1.bits.swizzle, y), + BRW_GET_SWZ(reg.dw1.bits.swizzle, z), + BRW_GET_SWZ(reg.dw1.bits.swizzle, w)); + return reg; +} + + +static INLINE struct brw_reg brw_swizzle1( struct brw_reg reg, + GLuint x ) +{ + return brw_swizzle(reg, x, x, x, x); +} + +static INLINE struct brw_reg brw_writemask( struct brw_reg reg, + GLuint mask ) +{ + reg.dw1.bits.writemask &= mask; + return reg; +} + +static INLINE struct brw_reg brw_set_writemask( struct brw_reg reg, + GLuint mask ) +{ + reg.dw1.bits.writemask = mask; + return reg; +} + +static INLINE struct brw_reg negate( struct brw_reg reg ) +{ + reg.negate ^= 1; + return reg; +} + +static INLINE struct brw_reg brw_abs( struct brw_reg reg ) +{ + reg.abs = 1; + return reg; +} + +/*********************************************************************** + */ +static INLINE struct brw_reg brw_vec4_indirect( GLuint subnr, + GLint offset ) +{ + struct brw_reg reg = brw_vec4_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static INLINE struct brw_reg brw_vec1_indirect( GLuint subnr, + GLint offset ) +{ + struct brw_reg reg = brw_vec1_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static INLINE struct brw_reg deref_4f(struct brw_indirect ptr, GLint offset) +{ + return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static INLINE struct brw_reg deref_1f(struct brw_indirect ptr, GLint offset) +{ + return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static INLINE struct brw_reg deref_4b(struct brw_indirect ptr, GLint offset) +{ + return retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B); +} + +static INLINE struct brw_reg deref_1uw(struct brw_indirect ptr, GLint offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW); +} + +static INLINE struct brw_reg deref_1d(struct brw_indirect ptr, GLint offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_D); +} + +static INLINE struct brw_reg deref_1ud(struct brw_indirect ptr, GLint offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD); +} + +static INLINE struct brw_reg get_addr_reg(struct brw_indirect ptr) +{ + return brw_address_reg(ptr.addr_subnr); +} + +static INLINE struct brw_indirect brw_indirect_offset( struct brw_indirect ptr, GLint offset ) +{ + ptr.addr_offset += offset; + return ptr; +} + +static INLINE struct brw_indirect brw_indirect( GLuint addr_subnr, GLint offset ) +{ + struct brw_indirect ptr; + ptr.addr_subnr = addr_subnr; + ptr.addr_offset = offset; + ptr.pad = 0; + return ptr; +} + +/** Do two brw_regs refer to the same register? */ +static INLINE GLboolean +brw_same_reg(struct brw_reg r1, struct brw_reg r2) +{ + return r1.file == r2.file && r1.nr == r2.nr; +} + +static INLINE struct brw_instruction *current_insn( struct brw_compile *p) +{ + return &p->store[p->nr_insn]; +} + +void brw_pop_insn_state( struct brw_compile *p ); +void brw_push_insn_state( struct brw_compile *p ); +void brw_set_mask_control( struct brw_compile *p, GLuint value ); +void brw_set_saturate( struct brw_compile *p, GLuint value ); +void brw_set_access_mode( struct brw_compile *p, GLuint access_mode ); +void brw_set_compression_control( struct brw_compile *p, GLboolean control ); +void brw_set_predicate_control_flag_value( struct brw_compile *p, GLuint value ); +void brw_set_predicate_control( struct brw_compile *p, GLuint pc ); +void brw_set_conditionalmod( struct brw_compile *p, GLuint conditional ); + +void brw_init_compile( struct brw_context *, struct brw_compile *p ); + +enum pipe_error brw_get_program( struct brw_compile *p, + const GLuint **program, + GLuint *sz ); + + +/* Helpers for regular instructions: + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0); + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1); + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(JMPI) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU1(RNDZ) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + +#undef ALU1 +#undef ALU2 + + + +/* Helpers for SEND instruction: + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean eot, + GLboolean writes_complete, + GLuint offset, + GLuint swizzle); + +void brw_ff_sync(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean eot, + GLboolean writes_complete, + GLuint offset, + GLuint swizzle); + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLuint binding_table_index, + GLuint msg_length, + GLuint response_length, + GLboolean eot); + +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLuint binding_table_index, + GLuint sampler, + GLuint writemask, + GLuint msg_type, + GLuint response_length, + GLuint msg_length, + GLboolean eot, + GLuint header_present, + GLuint simd_mode); + +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + GLuint function, + GLuint saturate, + GLuint msg_reg_nr, + struct brw_reg src, + GLuint precision ); + +void brw_math( struct brw_compile *p, + struct brw_reg dest, + GLuint function, + GLuint saturate, + GLuint msg_reg_nr, + struct brw_reg src, + GLuint data_type, + GLuint precision ); + +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + GLuint scratch_offset ); + +void brw_dp_READ_4( struct brw_compile *p, + struct brw_reg dest, + GLboolean relAddr, + GLuint location, + GLuint bind_table_index ); + +void brw_dp_READ_4_vs( struct brw_compile *p, + struct brw_reg dest, + GLuint oword, + GLboolean relAddr, + struct brw_reg addrReg, + GLuint location, + GLuint bind_table_index ); + +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + GLuint scratch_offset ); + +/* If/else/endif. Works by manipulating the execution flags on each + * channel. + */ +struct brw_instruction *brw_IF(struct brw_compile *p, + GLuint execute_size); + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn); + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *if_or_else_insn); + + +/* DO/WHILE loops: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, + GLuint execute_size); + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *patch_insn); + +struct brw_instruction *brw_BREAK(struct brw_compile *p); +struct brw_instruction *brw_CONT(struct brw_compile *p); +/* Forward jumps: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn); + + + +void brw_NOP(struct brw_compile *p); + +/* Special case: there is never a destination, execution size will be + * taken from src0: + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + GLuint conditional, + struct brw_reg src0, + struct brw_reg src1); + +void brw_print_reg( struct brw_reg reg ); + + +/*********************************************************************** + * brw_eu_util.c: + */ + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + GLuint count); + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + GLuint count); + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + GLuint count); + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + GLuint count); + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src); + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ); +#endif diff --git a/src/gallium/drivers/i965/brw_eu_debug.c b/src/gallium/drivers/i965/brw_eu_debug.c new file mode 100644 index 00000000000..5989f5a04ee --- /dev/null +++ b/src/gallium/drivers/i965/brw_eu_debug.c @@ -0,0 +1,94 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_debug.h" + +#include "brw_eu.h" + +void brw_print_reg( struct brw_reg hwreg ) +{ + static const char *file[] = { + "arf", + "grf", + "msg", + "imm" + }; + + static const char *type[] = { + "ud", + "d", + "uw", + "w", + "ub", + "vf", + "hf", + "f" + }; + + debug_printf("%s%s", + hwreg.abs ? "abs/" : "", + hwreg.negate ? "-" : ""); + + if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.nr % 2 == 0 && + hwreg.subnr == 0 && + hwreg.vstride == BRW_VERTICAL_STRIDE_8 && + hwreg.width == BRW_WIDTH_8 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 && + hwreg.type == BRW_REGISTER_TYPE_F) { + /* vector register */ + debug_printf("vec%d", hwreg.nr); + } + else if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.vstride == BRW_VERTICAL_STRIDE_0 && + hwreg.width == BRW_WIDTH_1 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 && + hwreg.type == BRW_REGISTER_TYPE_F) { + /* "scalar" register */ + debug_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); + } + else if (hwreg.file == BRW_IMMEDIATE_VALUE) { + debug_printf("imm %f", hwreg.dw1.f); + } + else { + debug_printf("%s%d.%d<%d;%d,%d>:%s", + file[hwreg.file], + hwreg.nr, + hwreg.subnr / type_sz(hwreg.type), + hwreg.vstride ? (1<<(hwreg.vstride-1)) : 0, + 1<<hwreg.width, + hwreg.hstride ? (1<<(hwreg.hstride-1)) : 0, + type[hwreg.type]); + } +} + + + diff --git a/src/gallium/drivers/i965/brw_eu_emit.c b/src/gallium/drivers/i965/brw_eu_emit.c new file mode 100644 index 00000000000..4fe7b6acc16 --- /dev/null +++ b/src/gallium/drivers/i965/brw_eu_emit.c @@ -0,0 +1,1433 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" +#include "brw_debug.h" +#include "brw_disasm.h" + + + + +/*********************************************************************** + * Internal helper for constructing instructions + */ + +static void guess_execution_size( struct brw_instruction *insn, + struct brw_reg reg ) +{ + if (reg.width == BRW_WIDTH_8 && + insn->header.compression_control == BRW_COMPRESSION_COMPRESSED) + insn->header.execution_size = BRW_EXECUTE_16; + else + insn->header.execution_size = reg.width; /* note - definitions are compatible */ +} + + +static void brw_set_dest( struct brw_instruction *insn, + struct brw_reg dest ) +{ + if (dest.type != BRW_ARCHITECTURE_REGISTER_FILE) + assert(dest.nr < 128); + + insn->bits1.da1.dest_reg_file = dest.file; + insn->bits1.da1.dest_reg_type = dest.type; + insn->bits1.da1.dest_address_mode = dest.address_mode; + + if (dest.address_mode == BRW_ADDRESS_DIRECT) { + insn->bits1.da1.dest_reg_nr = dest.nr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.da1.dest_subreg_nr = dest.subnr; + if (dest.hstride == BRW_HORIZONTAL_STRIDE_0) + dest.hstride = BRW_HORIZONTAL_STRIDE_1; + insn->bits1.da1.dest_horiz_stride = dest.hstride; + } + else { + insn->bits1.da16.dest_subreg_nr = dest.subnr / 16; + insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask; + } + } + else { + insn->bits1.ia1.dest_subreg_nr = dest.subnr; + + /* These are different sizes in align1 vs align16: + */ + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset; + if (dest.hstride == BRW_HORIZONTAL_STRIDE_0) + dest.hstride = BRW_HORIZONTAL_STRIDE_1; + insn->bits1.ia1.dest_horiz_stride = dest.hstride; + } + else { + insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset; + } + } + + /* NEW: Set the execution size based on dest.width and + * insn->compression_control: + */ + guess_execution_size(insn, dest); +} + +static void brw_set_src0( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + if (reg.type != BRW_ARCHITECTURE_REGISTER_FILE) + assert(reg.nr < 128); + + insn->bits1.da1.src0_reg_file = reg.file; + insn->bits1.da1.src0_reg_type = reg.type; + insn->bits2.da1.src0_abs = reg.abs; + insn->bits2.da1.src0_negate = reg.negate; + insn->bits2.da1.src0_address_mode = reg.address_mode; + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + + /* Required to set some fields in src1 as well: + */ + insn->bits1.da1.src1_reg_file = 0; /* arf */ + insn->bits1.da1.src1_reg_type = reg.type; + } + else + { + if (reg.address_mode == BRW_ADDRESS_DIRECT) { + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.da1.src0_subreg_nr = reg.subnr; + insn->bits2.da1.src0_reg_nr = reg.nr; + } + else { + insn->bits2.da16.src0_subreg_nr = reg.subnr / 16; + insn->bits2.da16.src0_reg_nr = reg.nr; + } + } + else { + insn->bits2.ia1.src0_subreg_nr = reg.subnr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset; + } + else { + insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset; + } + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits2.da1.src0_width = BRW_WIDTH_1; + insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits2.da1.src0_horiz_stride = reg.hstride; + insn->bits2.da1.src0_width = reg.width; + insn->bits2.da1.src0_vert_stride = reg.vstride; + } + } + else { + insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits2.da16.src0_vert_stride = reg.vstride; + } + } +} + + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + assert(reg.nr < 128); + + insn->bits1.da1.src1_reg_file = reg.file; + insn->bits1.da1.src1_reg_type = reg.type; + insn->bits3.da1.src1_abs = reg.abs; + insn->bits3.da1.src1_negate = reg.negate; + + /* Only src1 can be immediate in two-argument instructions. + */ + assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE); + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + } + else { + /* This is a hardware restriction, which may or may not be lifted + * in the future: + */ + assert (reg.address_mode == BRW_ADDRESS_DIRECT); + /*assert (reg.file == BRW_GENERAL_REGISTER_FILE); */ + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits3.da1.src1_subreg_nr = reg.subnr; + insn->bits3.da1.src1_reg_nr = reg.nr; + } + else { + insn->bits3.da16.src1_subreg_nr = reg.subnr / 16; + insn->bits3.da16.src1_reg_nr = reg.nr; + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits3.da1.src1_width = BRW_WIDTH_1; + insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits3.da1.src1_horiz_stride = reg.hstride; + insn->bits3.da1.src1_width = reg.width; + insn->bits3.da1.src1_vert_stride = reg.vstride; + } + } + else { + insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits3.da16.src1_vert_stride = reg.vstride; + } + } +} + + + +static void brw_set_math_message( struct brw_context *brw, + struct brw_instruction *insn, + GLuint msg_length, + GLuint response_length, + GLuint function, + GLuint integer_type, + GLboolean low_precision, + GLboolean saturate, + GLuint dataType ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + if (BRW_IS_IGDNG(brw)) { + insn->bits3.math_igdng.function = function; + insn->bits3.math_igdng.int_type = integer_type; + insn->bits3.math_igdng.precision = low_precision; + insn->bits3.math_igdng.saturate = saturate; + insn->bits3.math_igdng.data_type = dataType; + insn->bits3.math_igdng.snapshot = 0; + insn->bits3.math_igdng.header_present = 0; + insn->bits3.math_igdng.response_length = response_length; + insn->bits3.math_igdng.msg_length = msg_length; + insn->bits3.math_igdng.end_of_thread = 0; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_MATH; + insn->bits2.send_igdng.end_of_thread = 0; + } else { + insn->bits3.math.function = function; + insn->bits3.math.int_type = integer_type; + insn->bits3.math.precision = low_precision; + insn->bits3.math.saturate = saturate; + insn->bits3.math.data_type = dataType; + insn->bits3.math.response_length = response_length; + insn->bits3.math.msg_length = msg_length; + insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH; + insn->bits3.math.end_of_thread = 0; + } +} + + +static void brw_set_ff_sync_message( struct brw_context *brw, + struct brw_instruction *insn, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean end_of_thread, + GLboolean complete, + GLuint offset, + GLuint swizzle_control ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.urb_igdng.opcode = 1; + insn->bits3.urb_igdng.offset = offset; + insn->bits3.urb_igdng.swizzle_control = swizzle_control; + insn->bits3.urb_igdng.allocate = allocate; + insn->bits3.urb_igdng.used = used; + insn->bits3.urb_igdng.complete = complete; + insn->bits3.urb_igdng.header_present = 1; + insn->bits3.urb_igdng.response_length = response_length; + insn->bits3.urb_igdng.msg_length = msg_length; + insn->bits3.urb_igdng.end_of_thread = end_of_thread; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB; + insn->bits2.send_igdng.end_of_thread = end_of_thread; +} + +static void brw_set_urb_message( struct brw_context *brw, + struct brw_instruction *insn, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean end_of_thread, + GLboolean complete, + GLuint offset, + GLuint swizzle_control ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + if (BRW_IS_IGDNG(brw)) { + insn->bits3.urb_igdng.opcode = 0; /* ? */ + insn->bits3.urb_igdng.offset = offset; + insn->bits3.urb_igdng.swizzle_control = swizzle_control; + insn->bits3.urb_igdng.allocate = allocate; + insn->bits3.urb_igdng.used = used; /* ? */ + insn->bits3.urb_igdng.complete = complete; + insn->bits3.urb_igdng.header_present = 1; + insn->bits3.urb_igdng.response_length = response_length; + insn->bits3.urb_igdng.msg_length = msg_length; + insn->bits3.urb_igdng.end_of_thread = end_of_thread; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB; + insn->bits2.send_igdng.end_of_thread = end_of_thread; + } else { + insn->bits3.urb.opcode = 0; /* ? */ + insn->bits3.urb.offset = offset; + insn->bits3.urb.swizzle_control = swizzle_control; + insn->bits3.urb.allocate = allocate; + insn->bits3.urb.used = used; /* ? */ + insn->bits3.urb.complete = complete; + insn->bits3.urb.response_length = response_length; + insn->bits3.urb.msg_length = msg_length; + insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB; + insn->bits3.urb.end_of_thread = end_of_thread; + } +} + +static void brw_set_dp_write_message( struct brw_context *brw, + struct brw_instruction *insn, + GLuint binding_table_index, + GLuint msg_control, + GLuint msg_type, + GLuint msg_length, + GLuint pixel_scoreboard_clear, + GLuint response_length, + GLuint end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + if (BRW_IS_IGDNG(brw)) { + insn->bits3.dp_write_igdng.binding_table_index = binding_table_index; + insn->bits3.dp_write_igdng.msg_control = msg_control; + insn->bits3.dp_write_igdng.pixel_scoreboard_clear = pixel_scoreboard_clear; + insn->bits3.dp_write_igdng.msg_type = msg_type; + insn->bits3.dp_write_igdng.send_commit_msg = 0; + insn->bits3.dp_write_igdng.header_present = 1; + insn->bits3.dp_write_igdng.response_length = response_length; + insn->bits3.dp_write_igdng.msg_length = msg_length; + insn->bits3.dp_write_igdng.end_of_thread = end_of_thread; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_WRITE; + insn->bits2.send_igdng.end_of_thread = end_of_thread; + } else { + insn->bits3.dp_write.binding_table_index = binding_table_index; + insn->bits3.dp_write.msg_control = msg_control; + insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear; + insn->bits3.dp_write.msg_type = msg_type; + insn->bits3.dp_write.send_commit_msg = 0; + insn->bits3.dp_write.response_length = response_length; + insn->bits3.dp_write.msg_length = msg_length; + insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE; + insn->bits3.dp_write.end_of_thread = end_of_thread; + } +} + +static void brw_set_dp_read_message( struct brw_context *brw, + struct brw_instruction *insn, + GLuint binding_table_index, + GLuint msg_control, + GLuint msg_type, + GLuint target_cache, + GLuint msg_length, + GLuint response_length, + GLuint end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + if (BRW_IS_IGDNG(brw)) { + insn->bits3.dp_read_igdng.binding_table_index = binding_table_index; + insn->bits3.dp_read_igdng.msg_control = msg_control; + insn->bits3.dp_read_igdng.msg_type = msg_type; + insn->bits3.dp_read_igdng.target_cache = target_cache; + insn->bits3.dp_read_igdng.header_present = 1; + insn->bits3.dp_read_igdng.response_length = response_length; + insn->bits3.dp_read_igdng.msg_length = msg_length; + insn->bits3.dp_read_igdng.pad1 = 0; + insn->bits3.dp_read_igdng.end_of_thread = end_of_thread; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_READ; + insn->bits2.send_igdng.end_of_thread = end_of_thread; + } else { + insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/ + insn->bits3.dp_read.msg_control = msg_control; /*8:11*/ + insn->bits3.dp_read.msg_type = msg_type; /*12:13*/ + insn->bits3.dp_read.target_cache = target_cache; /*14:15*/ + insn->bits3.dp_read.response_length = response_length; /*16:19*/ + insn->bits3.dp_read.msg_length = msg_length; /*20:23*/ + insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; /*24:27*/ + insn->bits3.dp_read.pad1 = 0; /*28:30*/ + insn->bits3.dp_read.end_of_thread = end_of_thread; /*31*/ + } +} + +static void brw_set_sampler_message(struct brw_context *brw, + struct brw_instruction *insn, + GLuint binding_table_index, + GLuint sampler, + GLuint msg_type, + GLuint response_length, + GLuint msg_length, + GLboolean eot, + GLuint header_present, + GLuint simd_mode) +{ + assert(eot == 0); + brw_set_src1(insn, brw_imm_d(0)); + + if (BRW_IS_IGDNG(brw)) { + insn->bits3.sampler_igdng.binding_table_index = binding_table_index; + insn->bits3.sampler_igdng.sampler = sampler; + insn->bits3.sampler_igdng.msg_type = msg_type; + insn->bits3.sampler_igdng.simd_mode = simd_mode; + insn->bits3.sampler_igdng.header_present = header_present; + insn->bits3.sampler_igdng.response_length = response_length; + insn->bits3.sampler_igdng.msg_length = msg_length; + insn->bits3.sampler_igdng.end_of_thread = eot; + insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_SAMPLER; + insn->bits2.send_igdng.end_of_thread = eot; + } else if (BRW_IS_G4X(brw)) { + insn->bits3.sampler_g4x.binding_table_index = binding_table_index; + insn->bits3.sampler_g4x.sampler = sampler; + insn->bits3.sampler_g4x.msg_type = msg_type; + insn->bits3.sampler_g4x.response_length = response_length; + insn->bits3.sampler_g4x.msg_length = msg_length; + insn->bits3.sampler_g4x.end_of_thread = eot; + insn->bits3.sampler_g4x.msg_target = BRW_MESSAGE_TARGET_SAMPLER; + } else { + insn->bits3.sampler.binding_table_index = binding_table_index; + insn->bits3.sampler.sampler = sampler; + insn->bits3.sampler.msg_type = msg_type; + insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32; + insn->bits3.sampler.response_length = response_length; + insn->bits3.sampler.msg_length = msg_length; + insn->bits3.sampler.end_of_thread = eot; + insn->bits3.sampler.msg_target = BRW_MESSAGE_TARGET_SAMPLER; + } +} + + + +static struct brw_instruction *next_insn( struct brw_compile *p, + GLuint opcode ) +{ + struct brw_instruction *insn; + + if (0 && (BRW_DEBUG & DEBUG_DISASSEM)) + { + if (p->nr_insn) + brw_disasm_insn(stderr, &p->store[p->nr_insn-1]); + } + + assert(p->nr_insn + 1 < BRW_EU_MAX_INSN); + + insn = &p->store[p->nr_insn++]; + memcpy(insn, p->current, sizeof(*insn)); + + /* Reset this one-shot flag: + */ + + if (p->current->header.destreg__conditionalmod) { + p->current->header.destreg__conditionalmod = 0; + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } + + insn->header.opcode = opcode; + return insn; +} + + +static struct brw_instruction *brw_alu1( struct brw_compile *p, + GLuint opcode, + struct brw_reg dest, + struct brw_reg src ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + return insn; +} + +static struct brw_instruction *brw_alu2(struct brw_compile *p, + GLuint opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + return insn; +} + + +/*********************************************************************** + * Convenience routines. + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0) \ +{ \ + return brw_alu1(p, BRW_OPCODE_##OP, dest, src0); \ +} + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1) \ +{ \ + return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1); \ +} + + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU1(RNDZ) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + + + + +void brw_NOP(struct brw_compile *p) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP); + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_ud(0x0)); +} + + + + + +/*********************************************************************** + * Comparisons, if/else/endif + */ + +struct brw_instruction *brw_JMPI(struct brw_compile *p, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); + + insn->header.execution_size = 1; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.mask_control = BRW_MASK_DISABLE; + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + +/* EU takes the value from the flag register and pushes it onto some + * sort of a stack (presumably merging with any flag value already on + * the stack). Within an if block, the flags at the top of the stack + * control execution on each channel of the unit, eg. on each of the + * 16 pixel values in our wm programs. + * + * When the matching 'else' instruction is reached (presumably by + * countdown of the instruction count patched in by our ELSE/ENDIF + * functions), the relevent flags are inverted. + * + * When the matching 'endif' instruction is reached, the flags are + * popped off. If the stack is now empty, normal execution resumes. + * + * No attempt is made to deal with stack overflow (14 elements?). + */ +struct brw_instruction *brw_IF(struct brw_compile *p, GLuint execute_size) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) { + assert(execute_size == BRW_EXECUTE_1); + + insn = next_insn(p, BRW_OPCODE_ADD); + insn->header.predicate_inverse = 1; + } else { + insn = next_insn(p, BRW_OPCODE_IF); + } + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.execution_size = execute_size; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.predicate_control = BRW_PREDICATE_NORMAL; + insn->header.mask_control = BRW_MASK_ENABLE; + if (!p->single_program_flow) + insn->header.thread_control = BRW_THREAD_SWITCH; + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn) +{ + struct brw_instruction *insn; + GLuint br = 1; + + if (BRW_IS_IGDNG(p->brw)) + br = 2; + + if (p->single_program_flow) { + insn = next_insn(p, BRW_OPCODE_ADD); + } else { + insn = next_insn(p, BRW_OPCODE_ELSE); + } + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = if_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + if (!p->single_program_flow) + insn->header.thread_control = BRW_THREAD_SWITCH; + + /* Patch the if instruction to point at this instruction. + */ + if (p->single_program_flow) { + assert(if_insn->header.opcode == BRW_OPCODE_ADD); + + if_insn->bits3.ud = (insn - if_insn + 1) * 16; + } else { + assert(if_insn->header.opcode == BRW_OPCODE_IF); + + if_insn->bits3.if_else.jump_count = br * (insn - if_insn); + if_insn->bits3.if_else.pop_count = 0; + if_insn->bits3.if_else.pad0 = 0; + } + + return insn; +} + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *patch_insn) +{ + GLuint br = 1; + + if (BRW_IS_IGDNG(p->brw)) + br = 2; + + if (p->single_program_flow) { + /* In single program flow mode, there's no need to execute an ENDIF, + * since we don't need to do any stack operations, and if we're executing + * currently, we want to just continue executing. + */ + struct brw_instruction *next = &p->store[p->nr_insn]; + + assert(patch_insn->header.opcode == BRW_OPCODE_ADD); + + patch_insn->bits3.ud = (next - patch_insn) * 16; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF); + + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = patch_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + insn->header.thread_control = BRW_THREAD_SWITCH; + + assert(patch_insn->bits3.if_else.jump_count == 0); + + /* Patch the if or else instructions to point at this or the next + * instruction respectively. + */ + if (patch_insn->header.opcode == BRW_OPCODE_IF) { + /* Automagically turn it into an IFF: + */ + patch_insn->header.opcode = BRW_OPCODE_IFF; + patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1); + patch_insn->bits3.if_else.pop_count = 0; + patch_insn->bits3.if_else.pad0 = 0; + } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) { + patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1); + patch_insn->bits3.if_else.pop_count = 1; + patch_insn->bits3.if_else.pad0 = 0; + } else { + assert(0); + } + + /* Also pop item off the stack in the endif instruction: + */ + insn->bits3.if_else.jump_count = 0; + insn->bits3.if_else.pop_count = 1; + insn->bits3.if_else.pad0 = 0; + } +} + +struct brw_instruction *brw_BREAK(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_BREAK); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + /* insn->header.mask_control = BRW_MASK_DISABLE; */ + insn->bits3.if_else.pad0 = 0; + return insn; +} + +struct brw_instruction *brw_CONT(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_CONTINUE); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + /* insn->header.mask_control = BRW_MASK_DISABLE; */ + insn->bits3.if_else.pad0 = 0; + return insn; +} + +/* DO/WHILE loop: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size) +{ + if (p->single_program_flow) { + return &p->store[p->nr_insn]; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO); + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_null_reg()); + brw_set_src0(insn, brw_null_reg()); + brw_set_src1(insn, brw_null_reg()); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = execute_size; + insn->header.predicate_control = BRW_PREDICATE_NONE; + /* insn->header.mask_control = BRW_MASK_ENABLE; */ + /* insn->header.mask_control = BRW_MASK_DISABLE; */ + + return insn; + } +} + + + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *do_insn) +{ + struct brw_instruction *insn; + GLuint br = 1; + + if (BRW_IS_IGDNG(p->brw)) + br = 2; + + if (p->single_program_flow) + insn = next_insn(p, BRW_OPCODE_ADD); + else + insn = next_insn(p, BRW_OPCODE_WHILE); + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + + if (p->single_program_flow) { + insn->header.execution_size = BRW_EXECUTE_1; + + insn->bits3.d = (do_insn - insn) * 16; + } else { + insn->header.execution_size = do_insn->header.execution_size; + + assert(do_insn->header.opcode == BRW_OPCODE_DO); + insn->bits3.if_else.jump_count = br * (do_insn - insn + 1); + insn->bits3.if_else.pop_count = 0; + insn->bits3.if_else.pad0 = 0; + } + +/* insn->header.mask_control = BRW_MASK_ENABLE; */ + + /* insn->header.mask_control = BRW_MASK_DISABLE; */ + p->current->header.predicate_control = BRW_PREDICATE_NONE; + return insn; +} + + +/* FORWARD JUMPS: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn) +{ + struct brw_instruction *landing = &p->store[p->nr_insn]; + GLuint jmpi = 1; + + if (BRW_IS_IGDNG(p->brw)) + jmpi = 2; + + assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); + assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE); + + jmp_insn->bits3.ud = jmpi * ((landing - jmp_insn) - 1); +} + + + +/* To integrate with the above, it makes sense that the comparison + * instruction should populate the flag register. It might be simpler + * just to use the flag reg for most WM tasks? + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + GLuint conditional, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_CMP); + + insn->header.destreg__conditionalmod = conditional; + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + +/* guess_execution_size(insn, src0); */ + + + /* Make it so that future instructions will use the computed flag + * value until brw_set_predicate_control_flag_value() is called + * again. + */ + if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE && + dest.nr == 0) { + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + p->flag_value = 0xff; + } +} + + + +/*********************************************************************** + * Helpers for the various SEND message types: + */ + +/** Extended math function, float[8]. + */ +void brw_math( struct brw_compile *p, + struct brw_reg dest, + GLuint function, + GLuint saturate, + GLuint msg_reg_nr, + struct brw_reg src, + GLuint data_type, + GLuint precision ) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + GLuint msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + GLuint response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* Example code doesn't set predicate_control for send + * instructions. + */ + insn->header.predicate_control = 0; + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(p->brw, + insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + data_type); +} + +/** + * Extended math function, float[16]. + * Use 2 send instructions. + */ +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + GLuint function, + GLuint saturate, + GLuint msg_reg_nr, + struct brw_reg src, + GLuint precision ) +{ + struct brw_instruction *insn; + GLuint msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + GLuint response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* First instruction: + */ + brw_push_insn_state(p); + brw_set_predicate_control_flag_value(p, 0xff); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(p->brw, + insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + /* Second instruction: + */ + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.compression_control = BRW_COMPRESSION_2NDHALF; + insn->header.destreg__conditionalmod = msg_reg_nr+1; + + brw_set_dest(insn, offset(dest,1)); + brw_set_src0(insn, src); + brw_set_math_message(p->brw, + insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + brw_pop_insn_state(p); +} + + +/** + * Write block of 16 dwords/floats to the data port Render Cache scratch buffer. + * Scratch offset should be a multiple of 64. + * Used for register spilling. + */ +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + GLuint scratch_offset ) +{ + GLuint msg_reg_nr = 1; + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + /* set message header global offset field (reg 0, element 2) */ + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + GLuint msg_length = 3; + struct brw_reg dest = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW); + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + + brw_set_dp_write_message(p->brw, + insn, + 255, /* binding table index (255=stateless) */ + BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */ + msg_length, + 0, /* pixel scoreboard */ + 0, /* response_length */ + 0); /* eot */ + } +} + + +/** + * Read block of 16 dwords/floats from the data port Render Cache scratch buffer. + * Scratch offset should be a multiple of 64. + * Used for register spilling. + */ +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + GLuint scratch_offset ) +{ + GLuint msg_reg_nr = 1; + { + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + + /* set message header global offset field (reg 0, element 2) */ + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); /* UW? */ + brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW)); + + brw_set_dp_read_message(p->brw, + insn, + 255, /* binding table index (255=stateless) */ + 3, /* msg_control (3 means 4 Owords) */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 1, /* target cache (render/scratch) */ + 1, /* msg_length */ + 2, /* response_length */ + 0); /* eot */ + } +} + + +/** + * Read a float[4] vector from the data port Data Cache (const buffer). + * Location (in buffer) should be a multiple of 16. + * Used for fetching shader constants. + * If relAddr is true, we'll do an indirect fetch using the address register. + */ +void brw_dp_READ_4( struct brw_compile *p, + struct brw_reg dest, + GLboolean relAddr, + GLuint location, + GLuint bind_table_index ) +{ + /* XXX: relAddr not implemented */ + GLuint msg_reg_nr = 1; + { + struct brw_reg b; + brw_push_insn_state(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + + /* Setup MRF[1] with location/offset into const buffer */ + b = brw_message_reg(msg_reg_nr); + b = retype(b, BRW_REGISTER_TYPE_UD); + /* XXX I think we're setting all the dwords of MRF[1] to 'location'. + * when the docs say only dword[2] should be set. Hmmm. But it works. + */ + brw_MOV(p, b, brw_imm_ud(location)); + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = BRW_PREDICATE_NONE; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + insn->header.mask_control = BRW_MASK_DISABLE; + + /* cast dest to a uword[8] vector */ + dest = retype(vec8(dest), BRW_REGISTER_TYPE_UW); + + brw_set_dest(insn, dest); + brw_set_src0(insn, brw_null_reg()); + + brw_set_dp_read_message(p->brw, + insn, + bind_table_index, + 0, /* msg_control (0 means 1 Oword) */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 0, /* source cache = data cache */ + 1, /* msg_length */ + 1, /* response_length (1 Oword) */ + 0); /* eot */ + } +} + + +/** + * Read float[4] constant(s) from VS constant buffer. + * For relative addressing, two float[4] constants will be read into 'dest'. + * Otherwise, one float[4] constant will be read into the lower half of 'dest'. + */ +void brw_dp_READ_4_vs(struct brw_compile *p, + struct brw_reg dest, + GLuint oword, + GLboolean relAddr, + struct brw_reg addrReg, + GLuint location, + GLuint bind_table_index) +{ + GLuint msg_reg_nr = 1; + + assert(oword < 2); + /* + printf("vs const read msg, location %u, msg_reg_nr %d\n", + location, msg_reg_nr); + */ + + /* Setup MRF[1] with location/offset into const buffer */ + { + struct brw_reg b; + + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + /*brw_set_access_mode(p, BRW_ALIGN_16);*/ + + /* XXX I think we're setting all the dwords of MRF[1] to 'location'. + * when the docs say only dword[2] should be set. Hmmm. But it works. + */ + b = brw_message_reg(msg_reg_nr); + b = retype(b, BRW_REGISTER_TYPE_UD); + /*b = get_element_ud(b, 2);*/ + if (relAddr) { + brw_ADD(p, b, addrReg, brw_imm_ud(location)); + } + else { + brw_MOV(p, b, brw_imm_ud(location)); + } + + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = BRW_PREDICATE_NONE; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + insn->header.mask_control = BRW_MASK_DISABLE; + /*insn->header.access_mode = BRW_ALIGN_16;*/ + + brw_set_dest(insn, dest); + brw_set_src0(insn, brw_null_reg()); + + brw_set_dp_read_message(p->brw, + insn, + bind_table_index, + oword, /* 0 = lower Oword, 1 = upper Oword */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 0, /* source cache = data cache */ + 1, /* msg_length */ + 1, /* response_length (1 Oword) */ + 0); /* eot */ + } +} + + + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLuint binding_table_index, + GLuint msg_length, + GLuint response_length, + GLboolean eot) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_dp_write_message(p->brw, + insn, + binding_table_index, + BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */ + msg_length, + 1, /* pixel scoreboard */ + response_length, + eot); +} + + +/** + * Texture sample instruction. + * Note: the msg_type plus msg_length values determine exactly what kind + * of sampling operation is performed. See volume 4, page 161 of docs. + */ +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLuint binding_table_index, + GLuint sampler, + GLuint writemask, + GLuint msg_type, + GLuint response_length, + GLuint msg_length, + GLboolean eot, + GLuint header_present, + GLuint simd_mode) +{ + GLboolean need_stall = 0; + + if (writemask == 0) { + /*debug_printf("%s: zero writemask??\n", __FUNCTION__); */ + return; + } + + /* Hardware doesn't do destination dependency checking on send + * instructions properly. Add a workaround which generates the + * dependency by other means. In practice it seems like this bug + * only crops up for texture samples, and only where registers are + * written by the send and then written again later without being + * read in between. Luckily for us, we already track that + * information and use it to modify the writemask for the + * instruction, so that is a guide for whether a workaround is + * needed. + */ + if (writemask != BRW_WRITEMASK_XYZW) { + GLuint dst_offset = 0; + GLuint i, newmask = 0, len = 0; + + for (i = 0; i < 4; i++) { + if (writemask & (1<<i)) + break; + dst_offset += 2; + } + for (; i < 4; i++) { + if (!(writemask & (1<<i))) + break; + newmask |= 1<<i; + len++; + } + + if (newmask != writemask) { + need_stall = 1; + /* debug_printf("need stall %x %x\n", newmask , writemask); */ + } + else { + struct brw_reg m1 = brw_message_reg(msg_reg_nr); + + newmask = ~newmask & BRW_WRITEMASK_XYZW; + + brw_push_insn_state(p); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + + brw_MOV(p, m1, brw_vec8_grf(0,0)); + brw_MOV(p, get_element_ud(m1, 2), brw_imm_ud(newmask << 12)); + + brw_pop_insn_state(p); + + src0 = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW); + dest = offset(dest, dst_offset); + response_length = len * 2; + } + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_sampler_message(p->brw, insn, + binding_table_index, + sampler, + msg_type, + response_length, + msg_length, + eot, + header_present, + simd_mode); + } + + if (need_stall) { + struct brw_reg reg = vec8(offset(dest, response_length-1)); + + /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } + */ + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, reg, reg); + brw_pop_insn_state(p); + } + +} + +/* All these variables are pretty confusing - we might be better off + * using bitmasks and macros for this, in the old style. Or perhaps + * just having the caller instantiate the fields in dword3 itself. + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean eot, + GLboolean writes_complete, + GLuint offset, + GLuint swizzle) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + assert(msg_length < BRW_MAX_MRF); + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, brw_imm_d(0)); + + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_urb_message(p->brw, + insn, + allocate, + used, + msg_length, + response_length, + eot, + writes_complete, + offset, + swizzle); +} + +void brw_ff_sync(struct brw_compile *p, + struct brw_reg dest, + GLuint msg_reg_nr, + struct brw_reg src0, + GLboolean allocate, + GLboolean used, + GLuint msg_length, + GLuint response_length, + GLboolean eot, + GLboolean writes_complete, + GLuint offset, + GLuint swizzle) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + assert(msg_length < 16); + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, brw_imm_d(0)); + + insn->header.destreg__conditionalmod = msg_reg_nr; + + brw_set_ff_sync_message(p->brw, + insn, + allocate, + used, + msg_length, + response_length, + eot, + writes_complete, + offset, + swizzle); +} diff --git a/src/gallium/drivers/i965/brw_eu_util.c b/src/gallium/drivers/i965/brw_eu_util.c new file mode 100644 index 00000000000..5405cf17a4e --- /dev/null +++ b/src/gallium/drivers/i965/brw_eu_util.c @@ -0,0 +1,126 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src) +{ + brw_math( p, + dst, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + src, + BRW_MATH_PRECISION_FULL, + BRW_MATH_DATA_VECTOR ); +} + + + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + GLuint count) +{ + GLuint i; + + dst = vec4(dst); + src = vec4(src); + + for (i = 0; i < count; i++) + { + GLuint delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + brw_MOV(p, byte_offset(dst, delta+16), byte_offset(src, delta+16)); + } +} + + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + GLuint count) +{ + GLuint i; + + dst = vec8(dst); + src = vec8(src); + + for (i = 0; i < count; i++) + { + GLuint delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + } +} + + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + GLuint count) +{ + GLuint i; + + for (i = 0; i < count; i++) + { + GLuint delta = i*32; + brw_MOV(p, deref_4f(dst_ptr, delta), deref_4f(src_ptr, delta)); + brw_MOV(p, deref_4f(dst_ptr, delta+16), deref_4f(src_ptr, delta+16)); + } +} + + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + GLuint count) +{ + GLuint i; + + dst = vec4(dst); + + for (i = 0; i < count; i++) + { + GLuint delta = i*32; + brw_MOV(p, byte_offset(dst, delta), deref_4f(ptr, delta)); + brw_MOV(p, byte_offset(dst, delta+16), deref_4f(ptr, delta+16)); + } +} + + + + diff --git a/src/gallium/drivers/i965/brw_gs.c b/src/gallium/drivers/i965/brw_gs.c new file mode 100644 index 00000000000..921b201bae2 --- /dev/null +++ b/src/gallium/drivers/i965/brw_gs.c @@ -0,0 +1,216 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "brw_batchbuffer.h" + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_gs.h" + + + +static enum pipe_error compile_gs_prog( struct brw_context *brw, + struct brw_gs_prog_key *key, + struct brw_winsys_buffer **bo_out ) +{ + struct brw_gs_compile c; + enum pipe_error ret; + const GLuint *program; + GLuint program_size; + + memset(&c, 0, sizeof(c)); + + c.key = *key; + c.need_ff_sync = BRW_IS_IGDNG(brw); + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.nr_attrs = c.key.nr_attrs; + + if (BRW_IS_IGDNG(brw)) + c.nr_regs = (c.nr_attrs + 1) / 2 + 3; /* are vertices packed, or reg-aligned? */ + else + c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ + + c.nr_bytes = c.nr_regs * REG_SIZE; + + + /* Begin the compilation: + */ + brw_init_compile(brw, &c.func); + + c.func.single_program_flow = 1; + + /* For some reason the thread is spawned with only 4 channels + * unmasked. + */ + brw_set_mask_control(&c.func, BRW_MASK_DISABLE); + + + /* Note that primitives which don't require a GS program have + * already been weeded out by this stage: + */ + switch (key->primitive) { + case PIPE_PRIM_QUADS: + brw_gs_quads( &c ); + break; + case PIPE_PRIM_QUAD_STRIP: + brw_gs_quad_strip( &c ); + break; + case PIPE_PRIM_LINE_LOOP: + brw_gs_lines( &c ); + break; + case PIPE_PRIM_LINES: + if (key->hint_gs_always) + brw_gs_lines( &c ); + else { + return PIPE_OK; + } + break; + case PIPE_PRIM_TRIANGLES: + if (key->hint_gs_always) + brw_gs_tris( &c ); + else { + return PIPE_OK; + } + break; + case PIPE_PRIM_POINTS: + if (key->hint_gs_always) + brw_gs_points( &c ); + else { + return PIPE_OK; + } + break; + default: + assert(0); + return PIPE_ERROR_BAD_INPUT; + } + + /* get the program + */ + ret = brw_get_program(&c.func, &program, &program_size); + if (ret) + return ret; + + /* Upload + */ + ret = brw_upload_cache( &brw->cache, BRW_GS_PROG, + &c.key, sizeof(c.key), + NULL, 0, + program, program_size, + &c.prog_data, + &brw->gs.prog_data, + bo_out ); + if (ret) + return ret; + + return PIPE_OK; +} + +static const unsigned gs_prim[PIPE_PRIM_MAX] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINE_LOOP, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_QUADS, + PIPE_PRIM_QUAD_STRIP, + PIPE_PRIM_TRIANGLES +}; + +static void populate_key( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + const struct brw_fs_signature *sig = &brw->curr.fragment_shader->signature; + + memset(key, 0, sizeof(*key)); + + /* PIPE_NEW_FRAGMENT_SIGNATURE */ + key->nr_attrs = sig->nr_inputs + 1; + + /* BRW_NEW_PRIMITIVE */ + key->primitive = gs_prim[brw->primitive]; + + key->hint_gs_always = 0; /* debug code? */ + + key->need_gs_prog = (key->hint_gs_always || + brw->primitive == PIPE_PRIM_QUADS || + brw->primitive == PIPE_PRIM_QUAD_STRIP || + brw->primitive == PIPE_PRIM_LINE_LOOP); +} + +/* Calculate interpolants for triangle and line rasterization. + */ +static int prepare_gs_prog(struct brw_context *brw) +{ + struct brw_gs_prog_key key; + enum pipe_error ret; + + /* Populate the key: + */ + populate_key(brw, &key); + + if (brw->gs.prog_active != key.need_gs_prog) { + brw->state.dirty.cache |= CACHE_NEW_GS_PROG; + brw->gs.prog_active = key.need_gs_prog; + } + + if (!brw->gs.prog_active) + return PIPE_OK; + + if (brw_search_cache(&brw->cache, BRW_GS_PROG, + &key, sizeof(key), + NULL, 0, + &brw->gs.prog_data, + &brw->gs.prog_bo)) + return PIPE_OK; + + ret = compile_gs_prog( brw, &key, &brw->gs.prog_bo ); + if (ret) + return ret; + + return PIPE_OK; +} + + +const struct brw_tracked_state brw_gs_prog = { + .dirty = { + .mesa = PIPE_NEW_FRAGMENT_SIGNATURE, + .brw = BRW_NEW_PRIMITIVE, + .cache = 0, + }, + .prepare = prepare_gs_prog +}; diff --git a/src/gallium/drivers/i965/brw_gs.h b/src/gallium/drivers/i965/brw_gs.h new file mode 100644 index 00000000000..6e616dcb875 --- /dev/null +++ b/src/gallium/drivers/i965/brw_gs.h @@ -0,0 +1,76 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_GS_H +#define BRW_GS_H + + +#include "brw_context.h" +#include "brw_eu.h" + +#define MAX_GS_VERTS (4) + +struct brw_gs_prog_key { + GLuint nr_attrs:8; + GLuint primitive:4; + GLuint hint_gs_always:1; + GLuint need_gs_prog:1; + GLuint pad:18; +}; + +struct brw_gs_compile { + struct brw_compile func; + struct brw_gs_prog_key key; + struct brw_gs_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_GS_VERTS]; + } reg; + + /* 3 different ways of expressing vertex size: + */ + GLuint nr_attrs; + GLuint nr_regs; + GLuint nr_bytes; + GLboolean need_ff_sync; +}; + +#define ATTR_SIZE (4*4) + +void brw_gs_quads( struct brw_gs_compile *c ); +void brw_gs_quad_strip( struct brw_gs_compile *c ); +void brw_gs_tris( struct brw_gs_compile *c ); +void brw_gs_lines( struct brw_gs_compile *c ); +void brw_gs_points( struct brw_gs_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965/brw_gs_emit.c b/src/gallium/drivers/i965/brw_gs_emit.c new file mode 100644 index 00000000000..fd8e2accedd --- /dev/null +++ b/src/gallium/drivers/i965/brw_gs_emit.c @@ -0,0 +1,181 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_batchbuffer.h" + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_gs.h" + +static void brw_gs_alloc_regs( struct brw_gs_compile *c, + GLuint nr_verts ) +{ + GLuint i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->prog_data.urb_read_length = c->nr_regs; + c->prog_data.total_grf = i; +} + + +static void brw_gs_emit_vue(struct brw_gs_compile *c, + struct brw_reg vert, + GLboolean last, + GLuint header) +{ + struct brw_compile *p = &c->func; + GLboolean allocate = !last; + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy8(p, brw_message_reg(1), vert, c->nr_regs); + + /* Send each vertex as a seperate write to the urb. This is + * different to the concept in brw_sf_emit.c, where subsequent + * writes are used to build up a single urb entry. Each of these + * writes instantiates a seperate urb entry, and a new one must be + * allocated each time. + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response length */ + allocate ? 0 : 1, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + +static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim) +{ + struct brw_compile *p = &c->func; + brw_MOV(p, get_element_ud(c->reg.R0, 1), brw_imm_ud(num_prim)); + brw_ff_sync(p, + c->reg.R0, + 0, + c->reg.R0, + 1, + 1, /* used */ + 1, /* msg length */ + 1, /* response length */ + 0, /* eot */ + 1, /* write compelete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + +void brw_gs_quads( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + /* Use polygons for correct edgeflag behaviour. Note that vertex 3 + * is the PV for quads, but vertex 0 for polygons: + */ + if (c->need_ff_sync) + brw_gs_ff_sync(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_quad_strip( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + if (c->need_ff_sync) + brw_gs_ff_sync(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_tris( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 3); + + if (c->need_ff_sync) + brw_gs_ff_sync(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END)); +} + +void brw_gs_lines( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 2); + + if (c->need_ff_sync) + brw_gs_ff_sync(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END)); +} + +void brw_gs_points( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 1); + + if (c->need_ff_sync) + brw_gs_ff_sync(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END)); +} + + + + + + + + diff --git a/src/gallium/drivers/i965/brw_gs_state.c b/src/gallium/drivers/i965/brw_gs_state.c new file mode 100644 index 00000000000..b64ec286cea --- /dev/null +++ b/src/gallium/drivers/i965/brw_gs_state.c @@ -0,0 +1,169 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_debug.h" + +struct brw_gs_unit_key { + unsigned int total_grf; + unsigned int urb_entry_read_length; + + unsigned int curbe_offset; + + unsigned int nr_urb_entries, urb_size; + GLboolean prog_active; +}; + +static void +gs_unit_populate_key(struct brw_context *brw, struct brw_gs_unit_key *key) +{ + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_GS_PROG */ + key->prog_active = brw->gs.prog_active; + if (key->prog_active) { + key->total_grf = brw->gs.prog_data->total_grf; + key->urb_entry_read_length = brw->gs.prog_data->urb_read_length; + } else { + key->total_grf = 1; + key->urb_entry_read_length = 1; + } + + /* BRW_NEW_CURBE_OFFSETS */ + key->curbe_offset = brw->curbe.clip_start; + + /* BRW_NEW_URB_FENCE */ + key->nr_urb_entries = brw->urb.nr_gs_entries; + key->urb_size = brw->urb.vsize; +} + +static enum pipe_error +gs_unit_create_from_key(struct brw_context *brw, + struct brw_gs_unit_key *key, + struct brw_winsys_reloc *reloc, + unsigned nr_reloc, + struct brw_winsys_buffer **bo_out) +{ + struct brw_gs_unit_state gs; + enum pipe_error ret; + + + memset(&gs, 0, sizeof(gs)); + + /* reloc */ + gs.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1; + gs.thread0.kernel_start_pointer = 0; + + gs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + gs.thread1.single_program_flow = 1; + + gs.thread3.dispatch_grf_start_reg = 1; + gs.thread3.const_urb_entry_read_offset = 0; + gs.thread3.const_urb_entry_read_length = 0; + gs.thread3.urb_entry_read_offset = 0; + gs.thread3.urb_entry_read_length = key->urb_entry_read_length; + + gs.thread4.nr_urb_entries = key->nr_urb_entries; + gs.thread4.urb_entry_allocation_size = key->urb_size - 1; + + if (key->nr_urb_entries >= 8) + gs.thread4.max_threads = 1; + else + gs.thread4.max_threads = 0; + + if (BRW_IS_IGDNG(brw)) + gs.thread4.rendering_enable = 1; + + if (BRW_DEBUG & DEBUG_STATS) + gs.thread4.stats_enable = 1; + + ret = brw_upload_cache(&brw->cache, BRW_GS_UNIT, + key, sizeof(*key), + reloc, nr_reloc, + &gs, sizeof(gs), + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + +static enum pipe_error prepare_gs_unit(struct brw_context *brw) +{ + struct brw_gs_unit_key key; + enum pipe_error ret; + struct brw_winsys_reloc reloc[1]; + unsigned nr_reloc = 0; + unsigned grf_reg_count; + + gs_unit_populate_key(brw, &key); + + grf_reg_count = (align(key.total_grf, 16) / 16 - 1); + + /* GS program relocation */ + if (key.prog_active) { + make_reloc(&reloc[nr_reloc++], + BRW_USAGE_STATE, + grf_reg_count << 1, + offsetof(struct brw_gs_unit_state, thread0), + brw->gs.prog_bo); + } + + if (brw_search_cache(&brw->cache, BRW_GS_UNIT, + &key, sizeof(key), + reloc, nr_reloc, + NULL, + &brw->gs.state_bo)) + return PIPE_OK; + + ret = gs_unit_create_from_key(brw, &key, + reloc, nr_reloc, + &brw->gs.state_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_gs_unit = { + .dirty = { + .mesa = 0, + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_GS_PROG + }, + .prepare = prepare_gs_unit, +}; diff --git a/src/gallium/drivers/i965/brw_misc_state.c b/src/gallium/drivers/i965/brw_misc_state.c new file mode 100644 index 00000000000..e4b24229db3 --- /dev/null +++ b/src/gallium/drivers/i965/brw_misc_state.c @@ -0,0 +1,513 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + + +#include "brw_debug.h" +#include "brw_batchbuffer.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_screen.h" +#include "brw_pipe_rast.h" + + + + + +/*********************************************************************** + * Blend color + */ + +static int upload_blend_constant_color(struct brw_context *brw) +{ + BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bcc); + return 0; +} + + +const struct brw_tracked_state brw_blend_constant_color = { + .dirty = { + .mesa = PIPE_NEW_BLEND_COLOR, + .brw = 0, + .cache = 0 + }, + .emit = upload_blend_constant_color +}; + +/*********************************************************************** + * Drawing rectangle - framebuffer dimensions + */ +static int upload_drawing_rect(struct brw_context *brw) +{ + BEGIN_BATCH(4, NO_LOOP_CLIPRECTS); + OUT_BATCH(_3DSTATE_DRAWRECT_INFO_I965); + OUT_BATCH(0); + OUT_BATCH(((brw->curr.fb.width - 1) & 0xffff) | + ((brw->curr.fb.height - 1) << 16)); + OUT_BATCH(0); + ADVANCE_BATCH(); + return 0; +} + +const struct brw_tracked_state brw_drawing_rect = { + .dirty = { + .mesa = PIPE_NEW_FRAMEBUFFER_DIMENSIONS, + .brw = 0, + .cache = 0 + }, + .emit = upload_drawing_rect +}; + + +/*********************************************************************** + * Binding table pointers + */ + +static int prepare_binding_table_pointers(struct brw_context *brw) +{ + brw_add_validated_bo(brw, brw->vs.bind_bo); + brw_add_validated_bo(brw, brw->wm.bind_bo); + return 0; +} + +/** + * Upload the binding table pointers, which point each stage's array of surface + * state pointers. + * + * The binding table pointers are relative to the surface state base address, + * which is 0. + */ +static int upload_binding_table_pointers(struct brw_context *brw) +{ + BEGIN_BATCH(6, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2)); + if (brw->vs.bind_bo != NULL) + OUT_RELOC(brw->vs.bind_bo, + BRW_USAGE_SAMPLER, + 0); /* vs */ + else + OUT_BATCH(0); + OUT_BATCH(0); /* gs */ + OUT_BATCH(0); /* clip */ + OUT_BATCH(0); /* sf */ + OUT_RELOC(brw->wm.bind_bo, + BRW_USAGE_SAMPLER, + 0); /* wm/ps */ + ADVANCE_BATCH(); + return 0; +} + +const struct brw_tracked_state brw_binding_table_pointers = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_BATCH, + .cache = CACHE_NEW_SURF_BIND, + }, + .prepare = prepare_binding_table_pointers, + .emit = upload_binding_table_pointers, +}; + + +/********************************************************************** + * Upload pointers to the per-stage state. + * + * The state pointers in this packet are all relative to the general state + * base address set by CMD_STATE_BASE_ADDRESS, which is 0. + */ +static int upload_pipelined_state_pointers(struct brw_context *brw ) +{ + BEGIN_BATCH(7, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_PIPELINED_STATE_POINTERS << 16 | (7 - 2)); + OUT_RELOC(brw->vs.state_bo, + BRW_USAGE_STATE, + 0); + if (brw->gs.prog_active) + OUT_RELOC(brw->gs.state_bo, + BRW_USAGE_STATE, + 1); + else + OUT_BATCH(0); + OUT_RELOC(brw->clip.state_bo, + BRW_USAGE_STATE, + 1); + OUT_RELOC(brw->sf.state_bo, + BRW_USAGE_STATE, + 0); + OUT_RELOC(brw->wm.state_bo, + BRW_USAGE_STATE, + 0); + OUT_RELOC(brw->cc.state_bo, + BRW_USAGE_STATE, + 0); + ADVANCE_BATCH(); + + brw->state.dirty.brw |= BRW_NEW_PSP; + return 0; +} + + +static int prepare_psp_urb_cbs(struct brw_context *brw) +{ + brw_add_validated_bo(brw, brw->vs.state_bo); + brw_add_validated_bo(brw, brw->gs.state_bo); + brw_add_validated_bo(brw, brw->clip.state_bo); + brw_add_validated_bo(brw, brw->sf.state_bo); + brw_add_validated_bo(brw, brw->wm.state_bo); + brw_add_validated_bo(brw, brw->cc.state_bo); + return 0; +} + +static int upload_psp_urb_cbs(struct brw_context *brw ) +{ + int ret; + + ret = upload_pipelined_state_pointers(brw); + if (ret) + return ret; + + ret = brw_upload_urb_fence(brw); + if (ret) + return ret; + + ret = brw_upload_cs_urb_state(brw); + if (ret) + return ret; + + return 0; +} + +const struct brw_tracked_state brw_psp_urb_cbs = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_URB_FENCE | BRW_NEW_BATCH, + .cache = (CACHE_NEW_VS_UNIT | + CACHE_NEW_GS_UNIT | + CACHE_NEW_GS_PROG | + CACHE_NEW_CLIP_UNIT | + CACHE_NEW_SF_UNIT | + CACHE_NEW_WM_UNIT | + CACHE_NEW_CC_UNIT) + }, + .prepare = prepare_psp_urb_cbs, + .emit = upload_psp_urb_cbs, +}; + + +/*********************************************************************** + * Depth buffer + */ + +static int prepare_depthbuffer(struct brw_context *brw) +{ + struct pipe_surface *zsbuf = brw->curr.fb.zsbuf; + + if (zsbuf) + brw_add_validated_bo(brw, brw_surface(zsbuf)->bo); + + return 0; +} + +static int emit_depthbuffer(struct brw_context *brw) +{ + struct pipe_surface *surface = brw->curr.fb.zsbuf; + unsigned int len = (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? 6 : 5; + + if (surface == NULL) { + BEGIN_BATCH(len, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2)); + OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) | + (BRW_SURFACE_NULL << 29)); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(0); + + if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + OUT_BATCH(0); + + ADVANCE_BATCH(); + } else { + struct brw_winsys_buffer *bo; + unsigned int format; + unsigned int pitch; + unsigned int cpp; + + switch (surface->format) { + case PIPE_FORMAT_Z16_UNORM: + format = BRW_DEPTHFORMAT_D16_UNORM; + cpp = 2; + break; + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT; + cpp = 4; + break; + case PIPE_FORMAT_Z32_FLOAT: + format = BRW_DEPTHFORMAT_D32_FLOAT; + cpp = 4; + break; + default: + assert(0); + return PIPE_ERROR_BAD_INPUT; + } + + bo = brw_surface(surface)->bo; + pitch = brw_surface(surface)->pitch; + + BEGIN_BATCH(len, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2)); + OUT_BATCH(((pitch * cpp) - 1) | + (format << 18) | + (BRW_TILEWALK_YMAJOR << 26) | + ((surface->layout != PIPE_SURFACE_LAYOUT_LINEAR) << 27) | + (BRW_SURFACE_2D << 29)); + OUT_RELOC(bo, + BRW_USAGE_DEPTH_BUFFER, + surface->offset); + OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | + ((pitch - 1) << 6) | + ((surface->height - 1) << 19)); + OUT_BATCH(0); + + if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + OUT_BATCH(0); + + ADVANCE_BATCH(); + } + + return 0; +} + +const struct brw_tracked_state brw_depthbuffer = { + .dirty = { + .mesa = PIPE_NEW_DEPTH_BUFFER, + .brw = BRW_NEW_BATCH, + .cache = 0, + }, + .prepare = prepare_depthbuffer, + .emit = emit_depthbuffer, +}; + + + +/*********************************************************************** + * Polygon stipple packet + */ + +static int upload_polygon_stipple(struct brw_context *brw) +{ + BRW_CACHED_BATCH_STRUCT(brw, &brw->curr.bps); + return 0; +} + +const struct brw_tracked_state brw_polygon_stipple = { + .dirty = { + .mesa = PIPE_NEW_POLYGON_STIPPLE, + .brw = 0, + .cache = 0 + }, + .emit = upload_polygon_stipple +}; + + +/*********************************************************************** + * Line stipple packet + */ + +static int upload_line_stipple(struct brw_context *brw) +{ + const struct brw_line_stipple *bls = &brw->curr.rast->bls; + if (bls->header.opcode) { + BRW_CACHED_BATCH_STRUCT(brw, bls); + } + return 0; +} + +const struct brw_tracked_state brw_line_stipple = { + .dirty = { + .mesa = PIPE_NEW_RAST, + .brw = 0, + .cache = 0 + }, + .emit = upload_line_stipple +}; + + +/*********************************************************************** + * Misc invarient state packets + */ + +static int upload_invarient_state( struct brw_context *brw ) +{ + { + /* 0x61040000 Pipeline Select */ + /* PipelineSelect : 0 */ + struct brw_pipeline_select ps; + + memset(&ps, 0, sizeof(ps)); + if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + ps.header.opcode = CMD_PIPELINE_SELECT_GM45; + else + ps.header.opcode = CMD_PIPELINE_SELECT_965; + ps.header.pipeline_select = 0; + BRW_BATCH_STRUCT(brw, &ps); + } + + { + struct brw_global_depth_offset_clamp gdo; + memset(&gdo, 0, sizeof(gdo)); + + /* Disable depth offset clamping. + */ + gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP; + gdo.header.length = sizeof(gdo)/4 - 2; + gdo.depth_offset_clamp = 0.0; + + BRW_BATCH_STRUCT(brw, &gdo); + } + + + /* 0x61020000 State Instruction Pointer */ + { + struct brw_system_instruction_pointer sip; + memset(&sip, 0, sizeof(sip)); + + sip.header.opcode = CMD_STATE_INSN_POINTER; + sip.header.length = 0; + sip.bits0.pad = 0; + sip.bits0.system_instruction_pointer = 0; + BRW_BATCH_STRUCT(brw, &sip); + } + + /* VF Statistics */ + { + struct brw_vf_statistics vfs; + memset(&vfs, 0, sizeof(vfs)); + + if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + vfs.opcode = CMD_VF_STATISTICS_GM45; + else + vfs.opcode = CMD_VF_STATISTICS_965; + + if (BRW_DEBUG & DEBUG_STATS) + vfs.statistics_enable = 1; + + BRW_BATCH_STRUCT(brw, &vfs); + } + + if (!BRW_IS_965(brw)) + { + struct brw_aa_line_parameters balp; + + /* use legacy aa line coverage computation */ + memset(&balp, 0, sizeof(balp)); + balp.header.opcode = CMD_AA_LINE_PARAMETERS; + balp.header.length = sizeof(balp) / 4 - 2; + + BRW_BATCH_STRUCT(brw, &balp); + } + + { + struct brw_polygon_stipple_offset bpso; + + /* This is invarient state in gallium: + */ + memset(&bpso, 0, sizeof(bpso)); + bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; + bpso.header.length = sizeof(bpso)/4-2; + bpso.bits0.y_offset = 0; + bpso.bits0.x_offset = 0; + + BRW_BATCH_STRUCT(brw, &bpso); + } + + return 0; +} + +const struct brw_tracked_state brw_invarient_state = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .emit = upload_invarient_state +}; + + +/*********************************************************************** + * State base address + */ + +/** + * Define the base addresses which some state is referenced from. + * + * This allows us to avoid having to emit relocations in many places for + * cached state, and instead emit pointers inside of large, mostly-static + * state pools. This comes at the expense of memory, and more expensive cache + * misses. + */ +static int upload_state_base_address( struct brw_context *brw ) +{ + /* Output the structure (brw_state_base_address) directly to the + * batchbuffer, so we can emit relocations inline. + */ + if (BRW_IS_IGDNG(brw)) { + BEGIN_BATCH(8, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2)); + OUT_BATCH(1); /* General state base address */ + OUT_BATCH(1); /* Surface state base address */ + OUT_BATCH(1); /* Indirect object base address */ + OUT_BATCH(1); /* Instruction base address */ + OUT_BATCH(1); /* General state upper bound */ + OUT_BATCH(1); /* Indirect object upper bound */ + OUT_BATCH(1); /* Instruction access upper bound */ + ADVANCE_BATCH(); + } else { + BEGIN_BATCH(6, IGNORE_CLIPRECTS); + OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2)); + OUT_BATCH(1); /* General state base address */ + OUT_BATCH(1); /* Surface state base address */ + OUT_BATCH(1); /* Indirect object base address */ + OUT_BATCH(1); /* General state upper bound */ + OUT_BATCH(1); /* Indirect object upper bound */ + ADVANCE_BATCH(); + } + return 0; +} + +const struct brw_tracked_state brw_state_base_address = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_CONTEXT, + .cache = 0, + }, + .emit = upload_state_base_address +}; diff --git a/src/gallium/drivers/i965/brw_pipe_blend.c b/src/gallium/drivers/i965/brw_pipe_blend.c new file mode 100644 index 00000000000..b759a910b63 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_blend.c @@ -0,0 +1,208 @@ + +#include "util/u_memory.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_debug.h" + +static int translate_logicop(unsigned logicop) +{ + switch (logicop) { + case PIPE_LOGICOP_CLEAR: + return BRW_LOGICOPFUNCTION_CLEAR; + case PIPE_LOGICOP_AND: + return BRW_LOGICOPFUNCTION_AND; + case PIPE_LOGICOP_AND_REVERSE: + return BRW_LOGICOPFUNCTION_AND_REVERSE; + case PIPE_LOGICOP_COPY: + return BRW_LOGICOPFUNCTION_COPY; + case PIPE_LOGICOP_COPY_INVERTED: + return BRW_LOGICOPFUNCTION_COPY_INVERTED; + case PIPE_LOGICOP_AND_INVERTED: + return BRW_LOGICOPFUNCTION_AND_INVERTED; + case PIPE_LOGICOP_NOOP: + return BRW_LOGICOPFUNCTION_NOOP; + case PIPE_LOGICOP_XOR: + return BRW_LOGICOPFUNCTION_XOR; + case PIPE_LOGICOP_OR: + return BRW_LOGICOPFUNCTION_OR; + case PIPE_LOGICOP_OR_INVERTED: + return BRW_LOGICOPFUNCTION_OR_INVERTED; + case PIPE_LOGICOP_NOR: + return BRW_LOGICOPFUNCTION_NOR; + case PIPE_LOGICOP_EQUIV: + return BRW_LOGICOPFUNCTION_EQUIV; + case PIPE_LOGICOP_INVERT: + return BRW_LOGICOPFUNCTION_INVERT; + case PIPE_LOGICOP_OR_REVERSE: + return BRW_LOGICOPFUNCTION_OR_REVERSE; + case PIPE_LOGICOP_NAND: + return BRW_LOGICOPFUNCTION_NAND; + case PIPE_LOGICOP_SET: + return BRW_LOGICOPFUNCTION_SET; + default: + assert(0); + return BRW_LOGICOPFUNCTION_SET; + } +} + + +static unsigned translate_blend_equation( unsigned mode ) +{ + switch (mode) { + case PIPE_BLEND_ADD: + return BRW_BLENDFUNCTION_ADD; + case PIPE_BLEND_MIN: + return BRW_BLENDFUNCTION_MIN; + case PIPE_BLEND_MAX: + return BRW_BLENDFUNCTION_MAX; + case PIPE_BLEND_SUBTRACT: + return BRW_BLENDFUNCTION_SUBTRACT; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BRW_BLENDFUNCTION_REVERSE_SUBTRACT; + default: + assert(0); + return BRW_BLENDFUNCTION_ADD; + } +} + +static unsigned translate_blend_factor( unsigned factor ) +{ + switch(factor) { + case PIPE_BLENDFACTOR_ZERO: + return BRW_BLENDFACTOR_ZERO; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return BRW_BLENDFACTOR_SRC_ALPHA; + case PIPE_BLENDFACTOR_ONE: + return BRW_BLENDFACTOR_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return BRW_BLENDFACTOR_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return BRW_BLENDFACTOR_INV_SRC_COLOR; + case PIPE_BLENDFACTOR_DST_COLOR: + return BRW_BLENDFACTOR_DST_COLOR; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return BRW_BLENDFACTOR_INV_DST_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return BRW_BLENDFACTOR_INV_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return BRW_BLENDFACTOR_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return BRW_BLENDFACTOR_INV_DST_ALPHA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return BRW_BLENDFACTOR_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return BRW_BLENDFACTOR_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return BRW_BLENDFACTOR_INV_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return BRW_BLENDFACTOR_CONST_ALPHA; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return BRW_BLENDFACTOR_INV_CONST_ALPHA; + default: + assert(0); + return BRW_BLENDFACTOR_ZERO; + } +} + +static void *brw_create_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *templ ) +{ + struct brw_blend_state *blend = CALLOC_STRUCT(brw_blend_state); + if (blend == NULL) + return NULL; + + if (templ->logicop_enable) { + blend->cc2.logicop_enable = 1; + blend->cc5.logicop_func = translate_logicop(templ->logicop_func); + } + else if (templ->blend_enable) { + blend->cc6.dest_blend_factor = translate_blend_factor(templ->rgb_dst_factor); + blend->cc6.src_blend_factor = translate_blend_factor(templ->rgb_src_factor); + blend->cc6.blend_function = translate_blend_equation(templ->rgb_func); + + blend->cc5.ia_dest_blend_factor = translate_blend_factor(templ->alpha_dst_factor); + blend->cc5.ia_src_blend_factor = translate_blend_factor(templ->alpha_src_factor); + blend->cc5.ia_blend_function = translate_blend_equation(templ->alpha_func); + + blend->cc3.blend_enable = 1; + blend->cc3.ia_blend_enable = + (blend->cc6.dest_blend_factor != blend->cc5.ia_dest_blend_factor || + blend->cc6.src_blend_factor != blend->cc5.ia_src_blend_factor || + blend->cc6.blend_function != blend->cc5.ia_blend_function); + + /* Per-surface blend enables, currently just follow global + * state: + */ + blend->ss0.color_blend = 1; + } + + blend->cc5.dither_enable = templ->dither; + + if (BRW_DEBUG & DEBUG_STATS) + blend->cc5.statistics_enable = 1; + + /* Per-surface color mask -- just follow global state: + */ + blend->ss0.writedisable_red = (templ->colormask & PIPE_MASK_R) ? 0 : 1; + blend->ss0.writedisable_green = (templ->colormask & PIPE_MASK_G) ? 0 : 1; + blend->ss0.writedisable_blue = (templ->colormask & PIPE_MASK_B) ? 0 : 1; + blend->ss0.writedisable_alpha = (templ->colormask & PIPE_MASK_A) ? 0 : 1; + + return (void *)blend; +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + brw->curr.blend = (const struct brw_blend_state *)cso; + brw->state.dirty.mesa |= PIPE_NEW_BLEND; +} + +static void brw_delete_blend_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + assert((const void *)cso != (const void *)brw->curr.blend); + FREE(cso); +} + + +static void brw_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *blend_color) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_blend_constant_color *bcc = &brw->curr.bcc; + + bcc->blend_constant_color[0] = blend_color->color[0]; + bcc->blend_constant_color[1] = blend_color->color[1]; + bcc->blend_constant_color[2] = blend_color->color[2]; + bcc->blend_constant_color[3] = blend_color->color[3]; + + brw->state.dirty.mesa |= PIPE_NEW_BLEND_COLOR; +} + + +void brw_pipe_blend_init( struct brw_context *brw ) +{ + brw->base.set_blend_color = brw_set_blend_color; + brw->base.create_blend_state = brw_create_blend_state; + brw->base.bind_blend_state = brw_bind_blend_state; + brw->base.delete_blend_state = brw_delete_blend_state; + + { + struct brw_blend_constant_color *bcc = &brw->curr.bcc; + + memset(bcc, 0, sizeof(*bcc)); + bcc->header.opcode = CMD_BLEND_CONSTANT_COLOR; + bcc->header.length = sizeof(*bcc)/4-2; + } + +} + +void brw_pipe_blend_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_clear.c b/src/gallium/drivers/i965/brw_pipe_clear.c new file mode 100644 index 00000000000..211be881789 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_clear.c @@ -0,0 +1,218 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "util/u_pack_color.h" + +#include "pipe/p_state.h" + +#include "brw_batchbuffer.h" +#include "brw_screen.h" +#include "brw_context.h" + +#define MASK16 0xffff +#define MASK24 0xffffff + + +/** + * Use blitting to clear the renderbuffers named by 'flags'. + * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field + * since that might include software renderbuffers or renderbuffers + * which we're clearing with triangles. + * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear + */ +static enum pipe_error +try_clear( struct brw_context *brw, + struct brw_surface *surface, + unsigned value ) +{ + uint32_t BR13, CMD; + int x1 = 0; + int y1 = 0; + int x2 = surface->base.width; + int y2 = surface->base.height; + int pitch = surface->pitch; + int cpp = surface->cpp; + + if (x2 == 0 || y2 == 0) + return 0; + + debug_printf("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + (void *)surface->bo, pitch * cpp, + surface->base.offset, + x1, y1, x2 - x1, y2 - y1); + + BR13 = 0xf0 << 16; + CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_RGB | XY_BLT_WRITE_ALPHA; + + /* Setup the blit command */ + if (cpp == 4) { + BR13 |= BR13_8888; + CMD |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + } + else { + assert(cpp == 2); + BR13 |= BR13_565; + } + + /* XXX: nasty hack for clearing depth buffers + */ + if (surface->tiling == BRW_TILING_Y) { + x2 = pitch; + } + + if (surface->tiling == BRW_TILING_X) { + CMD |= XY_DST_TILED; + pitch /= 4; + } + + BR13 |= (pitch * cpp); + + BEGIN_BATCH(6, 0); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y1 << 16) | x1); + OUT_BATCH((y2 << 16) | x2); + OUT_RELOC(surface->bo, + BRW_USAGE_BLIT_DEST, + surface->base.offset); + OUT_BATCH(value); + ADVANCE_BATCH(); + + return 0; +} + + + + +static void color_clear(struct brw_context *brw, + struct brw_surface *bsurface, + const float *rgba ) +{ + enum pipe_error ret; + unsigned value; + + util_pack_color( rgba, bsurface->base.format, &value ); + + if (bsurface->cpp == 2) + value |= value << 16; + + ret = try_clear( brw, bsurface, value ); + + if (ret != 0) { + brw_context_flush( brw ); + ret = try_clear( brw, bsurface, value ); + assert( ret == 0 ); + } +} + +static void zstencil_clear(struct brw_context *brw, + struct brw_surface *bsurface, + double depth, + unsigned stencil ) +{ + enum pipe_error ret; + unsigned value; + + switch (bsurface->base.format) { + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + value = ((unsigned)(depth * MASK24) & MASK24); + break; + case PIPE_FORMAT_Z16_UNORM: + value = ((unsigned)(depth * MASK16) & MASK16); + break; + default: + assert(0); + return; + } + + switch (bsurface->base.format) { + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + value = value | (stencil << 24); + break; + + case PIPE_FORMAT_Z16_UNORM: + value = value | (value << 16); + break; + + default: + break; + } + + ret = try_clear( brw, bsurface, value ); + + if (ret != 0) { + brw_context_flush( brw ); + ret = try_clear( brw, bsurface, value ); + assert( ret == 0 ); + } +} + + + +/** + * Clear the given surface to the specified value. + * No masking, no scissor (clear entire buffer). + */ +static void brw_clear(struct pipe_context *pipe, + unsigned buffers, + const float *rgba, + double depth, + unsigned stencil) +{ + struct brw_context *brw = brw_context( pipe ); + int i; + + if (buffers & PIPE_CLEAR_COLOR) { + for (i = 0; i < brw->curr.fb.nr_cbufs; i++) { + color_clear( brw, + brw_surface(brw->curr.fb.cbufs[i]), + rgba ); + } + } + + if (buffers & PIPE_CLEAR_DEPTHSTENCIL) { + if (brw->curr.fb.zsbuf) { + zstencil_clear( brw, + brw_surface(brw->curr.fb.zsbuf), + depth, stencil ); + } + } +} + + +void brw_pipe_clear_init( struct brw_context *brw ) +{ + brw->base.clear = brw_clear; +} + + +void brw_pipe_clear_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_depth.c b/src/gallium/drivers/i965/brw_pipe_depth.c new file mode 100644 index 00000000000..e010d76e0d3 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_depth.c @@ -0,0 +1,172 @@ + +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "brw_context.h" +#include "brw_defines.h" + +/* XXX: Fixme - include this to get IZ_ defines + */ +#include "brw_wm.h" + +static unsigned brw_translate_compare_func(unsigned func) +{ + switch (func) { + case PIPE_FUNC_NEVER: + return BRW_COMPAREFUNCTION_NEVER; + case PIPE_FUNC_LESS: + return BRW_COMPAREFUNCTION_LESS; + case PIPE_FUNC_LEQUAL: + return BRW_COMPAREFUNCTION_LEQUAL; + case PIPE_FUNC_GREATER: + return BRW_COMPAREFUNCTION_GREATER; + case PIPE_FUNC_GEQUAL: + return BRW_COMPAREFUNCTION_GEQUAL; + case PIPE_FUNC_NOTEQUAL: + return BRW_COMPAREFUNCTION_NOTEQUAL; + case PIPE_FUNC_EQUAL: + return BRW_COMPAREFUNCTION_EQUAL; + case PIPE_FUNC_ALWAYS: + return BRW_COMPAREFUNCTION_ALWAYS; + default: + assert(0); + return BRW_COMPAREFUNCTION_ALWAYS; + } +} + +static unsigned translate_stencil_op(unsigned op) +{ + switch (op) { + case PIPE_STENCIL_OP_KEEP: + return BRW_STENCILOP_KEEP; + case PIPE_STENCIL_OP_ZERO: + return BRW_STENCILOP_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return BRW_STENCILOP_REPLACE; + case PIPE_STENCIL_OP_INCR: + return BRW_STENCILOP_INCRSAT; + case PIPE_STENCIL_OP_DECR: + return BRW_STENCILOP_DECRSAT; + case PIPE_STENCIL_OP_INCR_WRAP: + return BRW_STENCILOP_INCR; + case PIPE_STENCIL_OP_DECR_WRAP: + return BRW_STENCILOP_DECR; + case PIPE_STENCIL_OP_INVERT: + return BRW_STENCILOP_INVERT; + default: + assert(0); + return BRW_STENCILOP_ZERO; + } +} + +static void create_bcc_state( struct brw_depth_stencil_state *zstencil, + const struct pipe_depth_stencil_alpha_state *templ ) +{ + if (templ->stencil[0].enabled) { + zstencil->cc0.stencil_enable = 1; + zstencil->cc0.stencil_func = + brw_translate_compare_func(templ->stencil[0].func); + zstencil->cc0.stencil_fail_op = + translate_stencil_op(templ->stencil[0].fail_op); + zstencil->cc0.stencil_pass_depth_fail_op = + translate_stencil_op(templ->stencil[0].zfail_op); + zstencil->cc0.stencil_pass_depth_pass_op = + translate_stencil_op(templ->stencil[0].zpass_op); + zstencil->cc1.stencil_ref = templ->stencil[0].ref_value; + zstencil->cc1.stencil_write_mask = templ->stencil[0].writemask; + zstencil->cc1.stencil_test_mask = templ->stencil[0].valuemask; + + if (templ->stencil[1].enabled) { + zstencil->cc0.bf_stencil_enable = 1; + zstencil->cc0.bf_stencil_func = + brw_translate_compare_func(templ->stencil[1].func); + zstencil->cc0.bf_stencil_fail_op = + translate_stencil_op(templ->stencil[1].fail_op); + zstencil->cc0.bf_stencil_pass_depth_fail_op = + translate_stencil_op(templ->stencil[1].zfail_op); + zstencil->cc0.bf_stencil_pass_depth_pass_op = + translate_stencil_op(templ->stencil[1].zpass_op); + zstencil->cc1.bf_stencil_ref = templ->stencil[1].ref_value; + zstencil->cc2.bf_stencil_write_mask = templ->stencil[1].writemask; + zstencil->cc2.bf_stencil_test_mask = templ->stencil[1].valuemask; + } + + zstencil->cc0.stencil_write_enable = (zstencil->cc1.stencil_write_mask || + zstencil->cc2.bf_stencil_write_mask); + } + + + if (templ->alpha.enabled) { + zstencil->cc3.alpha_test = 1; + zstencil->cc3.alpha_test_func = brw_translate_compare_func(templ->alpha.func); + zstencil->cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; + zstencil->cc7.alpha_ref.ub[0] = float_to_ubyte(templ->alpha.ref_value); + } + + if (templ->depth.enabled) { + zstencil->cc2.depth_test = 1; + zstencil->cc2.depth_test_function = brw_translate_compare_func(templ->depth.func); + zstencil->cc2.depth_write_enable = templ->depth.writemask; + } +} + +static void create_wm_iz_state( struct brw_depth_stencil_state *zstencil ) +{ + if (zstencil->cc3.alpha_test) + zstencil->iz_lookup |= IZ_PS_KILL_ALPHATEST_BIT; + + if (zstencil->cc2.depth_test) + zstencil->iz_lookup |= IZ_DEPTH_TEST_ENABLE_BIT; + + if (zstencil->cc2.depth_write_enable) + zstencil->iz_lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; + + if (zstencil->cc0.stencil_enable) + zstencil->iz_lookup |= IZ_STENCIL_TEST_ENABLE_BIT; + + if (zstencil->cc0.stencil_write_enable) + zstencil->iz_lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; + +} + + +static void * +brw_create_depth_stencil_state( struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *templ ) +{ + struct brw_depth_stencil_state *zstencil = CALLOC_STRUCT(brw_depth_stencil_state); + + create_bcc_state( zstencil, templ ); + create_wm_iz_state( zstencil ); + + return (void *)zstencil; +} + + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + brw->curr.zstencil = (const struct brw_depth_stencil_state *)cso; + brw->state.dirty.mesa |= PIPE_NEW_DEPTH_STENCIL_ALPHA; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + assert((const void *)cso != (const void *)brw->curr.zstencil); + FREE(cso); +} + + +void brw_pipe_depth_stencil_init( struct brw_context *brw ) +{ + brw->base.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->base.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->base.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; +} + +void brw_pipe_depth_stencil_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_fb.c b/src/gallium/drivers/i965/brw_pipe_fb.c new file mode 100644 index 00000000000..6b03094f502 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_fb.c @@ -0,0 +1,77 @@ +#include "util/u_math.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "brw_context.h" + +/** + * called from intelDrawBuffer() + */ +static void brw_set_framebuffer_state( struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb ) +{ + struct brw_context *brw = brw_context(pipe); + unsigned i; + + /* Dimensions: + */ + if (brw->curr.fb.width != fb->width || + brw->curr.fb.height != fb->height) { + brw->curr.fb.width = fb->width; + brw->curr.fb.height = fb->height; + brw->state.dirty.mesa |= PIPE_NEW_FRAMEBUFFER_DIMENSIONS; + } + + /* Z/Stencil + */ + if (brw->curr.fb.zsbuf != fb->zsbuf) { + pipe_surface_reference(&brw->curr.fb.zsbuf, fb->zsbuf); + brw->state.dirty.mesa |= PIPE_NEW_DEPTH_BUFFER; + } + + /* Color buffers: + */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + if (brw->curr.fb.cbufs[i] != fb->cbufs[i]) { + brw->state.dirty.mesa |= PIPE_NEW_COLOR_BUFFERS; + pipe_surface_reference(&brw->curr.fb.cbufs[i], fb->cbufs[i]); + } + } + + if (brw->curr.fb.nr_cbufs != fb->nr_cbufs) { + brw->curr.fb.nr_cbufs = MIN2(BRW_MAX_DRAW_BUFFERS, fb->nr_cbufs); + brw->state.dirty.mesa |= PIPE_NEW_NR_CBUFS; + } +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->curr.viewport = *viewport; + brw->curr.ccv.min_depth = 0.0; /* XXX: near */ + brw->curr.ccv.max_depth = 1.0; /* XXX: far */ + + brw->state.dirty.mesa |= PIPE_NEW_VIEWPORT; +} + + +void brw_pipe_framebuffer_init( struct brw_context *brw ) +{ + brw->base.set_framebuffer_state = brw_set_framebuffer_state; + brw->base.set_viewport_state = brw_set_viewport_state; +} + +void brw_pipe_framebuffer_cleanup( struct brw_context *brw ) +{ + struct pipe_framebuffer_state *fb = &brw->curr.fb; + int i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + pipe_surface_reference(&fb->cbufs[i], NULL); + } + + pipe_surface_reference(&fb->zsbuf, NULL); +} diff --git a/src/gallium/drivers/i965/brw_pipe_flush.c b/src/gallium/drivers/i965/brw_pipe_flush.c new file mode 100644 index 00000000000..fdc4814b221 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_flush.c @@ -0,0 +1,83 @@ + +#include "util/u_upload_mgr.h" + +#include "brw_context.h" +#include "brw_screen.h" +#include "brw_batchbuffer.h" + + + +/* All batchbuffer flushes must go through this function. + */ +void brw_context_flush( struct brw_context *brw ) +{ + /* + * + */ + brw_emit_query_end(brw); + + /* Move to the end of the current upload buffer so that we'll force choosing + * a new buffer next time. + */ + u_upload_flush( brw->vb.upload_vertex ); + u_upload_flush( brw->vb.upload_index ); + + _brw_batchbuffer_flush( brw->batch, __FILE__, __LINE__ ); + + /* Mark all context state as needing to be re-emitted. + * This is probably not as severe as on 915, since almost all of our state + * is just in referenced buffers. + */ + brw->state.dirty.brw |= BRW_NEW_CONTEXT; + brw->state.dirty.mesa |= ~0; + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; + + brw->curbe.need_new_bo = GL_TRUE; +} + +static void +brw_flush( struct pipe_context *pipe, + unsigned flags, + struct pipe_fence_handle **fence ) +{ + brw_context_flush( brw_context( pipe ) ); + if (fence) + *fence = NULL; +} + +static unsigned brw_is_buffer_referenced(struct pipe_context *pipe, + struct pipe_buffer *buffer) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_screen *bscreen = brw_screen(brw->base.screen); + + return brw_is_buffer_referenced_by_bo( bscreen, + buffer, + brw->batch->buf ); +} + +static unsigned brw_is_texture_referenced(struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, + unsigned level) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_screen *bscreen = brw_screen(brw->base.screen); + + return brw_is_texture_referenced_by_bo( bscreen, + texture, face, level, + brw->batch->buf ); +} + +void brw_pipe_flush_init( struct brw_context *brw ) +{ + brw->base.flush = brw_flush; + brw->base.is_buffer_referenced = brw_is_buffer_referenced; + brw->base.is_texture_referenced = brw_is_texture_referenced; +} + + +void brw_pipe_flush_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_misc.c b/src/gallium/drivers/i965/brw_pipe_misc.c new file mode 100644 index 00000000000..30359078079 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_misc.c @@ -0,0 +1,54 @@ + +#include "brw_context.h" +#include "brw_structs.h" +#include "brw_defines.h" + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stip ) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_polygon_stipple *bps = &brw->curr.bps; + GLuint i; + + memset(bps, 0, sizeof *bps); + bps->header.opcode = CMD_POLY_STIPPLE_PATTERN; + bps->header.length = sizeof *bps/4-2; + + for (i = 0; i < 32; i++) + bps->stipple[i] = stip->stipple[i]; /* don't invert */ + + brw->state.dirty.mesa |= PIPE_NEW_POLYGON_STIPPLE; +} + + +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->curr.scissor = *scissor; + brw->state.dirty.mesa |= PIPE_NEW_SCISSOR; +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->curr.ucp = *clip; + brw->state.dirty.mesa |= PIPE_NEW_CLIP; +} + + +void brw_pipe_misc_init( struct brw_context *brw ) +{ + brw->base.set_polygon_stipple = brw_set_polygon_stipple; + brw->base.set_scissor_state = brw_set_scissor_state; + brw->base.set_clip_state = brw_set_clip_state; +} + + +void brw_pipe_misc_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_query.c b/src/gallium/drivers/i965/brw_pipe_query.c new file mode 100644 index 00000000000..2eb862635cc --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_query.c @@ -0,0 +1,263 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt <[email protected]> + * + */ + +/** @file support for ARB_query_object + * + * ARB_query_object is implemented by using the PIPE_CONTROL command to stall + * execution on the completion of previous depth tests, and write the + * current PS_DEPTH_COUNT to a buffer object. + * + * We use before and after counts when drawing during a query so that + * we don't pick up other clients' query data in ours. To reduce overhead, + * a single BO is used to record the query data for all active queries at + * once. This also gives us a simple bound on how much batchbuffer space is + * required for handling queries, so that we can be sure that we won't + * have to emit a batchbuffer without getting the ending PS_DEPTH_COUNT. + */ +#include "util/u_simple_list.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_batchbuffer.h" +#include "brw_reg.h" + +/** Waits on the query object's BO and totals the results for this query */ +static boolean +brw_query_get_result(struct pipe_context *pipe, + struct pipe_query *q, + boolean wait, + uint64_t *result) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_query_object *query = (struct brw_query_object *)q; + + /* Map and count the pixels from the current query BO */ + if (query->bo) { + int i; + uint64_t *map; + + if (brw->sws->bo_is_busy(query->bo) && !wait) + return FALSE; + + map = bo_map_read(brw->sws, query->bo); + if (map == NULL) + return FALSE; + + for (i = query->first_index; i <= query->last_index; i++) { + query->result += map[i * 2 + 1] - map[i * 2]; + } + + brw->sws->bo_unmap(query->bo); + bo_reference(&query->bo, NULL); + } + + *result = query->result; + return TRUE; +} + +static struct pipe_query * +brw_query_create(struct pipe_context *pipe, unsigned type ) +{ + struct brw_query_object *query; + + switch (type) { + case PIPE_QUERY_OCCLUSION_COUNTER: + query = CALLOC_STRUCT( brw_query_object ); + if (query == NULL) + return NULL; + return (struct pipe_query *)query; + + default: + return NULL; + } +} + +static void +brw_query_destroy(struct pipe_context *pipe, struct pipe_query *q) +{ + struct brw_query_object *query = (struct brw_query_object *)q; + + bo_reference(&query->bo, NULL); + FREE(query); +} + +static void +brw_query_begin(struct pipe_context *pipe, struct pipe_query *q) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_query_object *query = (struct brw_query_object *)q; + + /* Reset our driver's tracking of query state. */ + bo_reference(&query->bo, NULL); + query->result = 0; + query->first_index = -1; + query->last_index = -1; + + insert_at_head(&brw->query.active_head, query); + brw->query.stats_wm++; + brw->state.dirty.mesa |= PIPE_NEW_QUERY; +} + +static void +brw_query_end(struct pipe_context *pipe, struct pipe_query *q) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_query_object *query = (struct brw_query_object *)q; + + /* Flush the batchbuffer in case it has writes to our query BO. + * Have later queries write to a new query BO so that further rendering + * doesn't delay the collection of our results. + */ + if (query->bo) { + brw_emit_query_end(brw); + brw_context_flush( brw ); + + bo_reference(&brw->query.bo, NULL); + } + + remove_from_list(query); + brw->query.stats_wm--; + brw->state.dirty.mesa |= PIPE_NEW_QUERY; +} + +/*********************************************************************** + * Internal functions and callbacks to implement queries + */ + +/** Called to set up the query BO and account for its aperture space */ +enum pipe_error +brw_prepare_query_begin(struct brw_context *brw) +{ + enum pipe_error ret; + + /* Skip if we're not doing any queries. */ + if (is_empty_list(&brw->query.active_head)) + return PIPE_OK; + + /* Get a new query BO if we're going to need it. */ + if (brw->query.bo == NULL || + brw->query.index * 2 + 1 >= 4096 / sizeof(uint64_t)) { + + ret = brw->sws->bo_alloc(brw->sws, BRW_BUFFER_TYPE_QUERY, 4096, 1, + &brw->query.bo); + if (ret) + return ret; + + brw->query.index = 0; + } + + brw_add_validated_bo(brw, brw->query.bo); + + return PIPE_OK; +} + +/** Called just before primitive drawing to get a beginning PS_DEPTH_COUNT. */ +void +brw_emit_query_begin(struct brw_context *brw) +{ + struct brw_query_object *query; + + /* Skip if we're not doing any queries, or we've emitted the start. */ + if (brw->query.active || is_empty_list(&brw->query.active_head)) + return; + + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_WRITE_DEPTH_COUNT); + /* This object could be mapped cacheable, but we don't have an exposed + * mechanism to support that. Since it's going uncached, tell GEM that + * we're writing to it. The usual clflush should be all that's required + * to pick up the results. + */ + OUT_RELOC(brw->query.bo, + BRW_USAGE_QUERY_RESULT, + PIPE_CONTROL_GLOBAL_GTT_WRITE | + ((brw->query.index * 2) * sizeof(uint64_t))); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + + foreach(query, &brw->query.active_head) { + if (query->bo != brw->query.bo) { + uint64_t tmp; + + /* Propogate the results from this buffer to all of the + * active queries, as the bo is going away. + */ + if (query->bo != NULL) + brw_query_get_result( &brw->base, + (struct pipe_query *)query, + FALSE, + &tmp ); + + bo_reference( &query->bo, brw->query.bo ); + query->first_index = brw->query.index; + } + query->last_index = brw->query.index; + } + brw->query.active = GL_TRUE; +} + +/** Called at batchbuffer flush to get an ending PS_DEPTH_COUNT */ +void +brw_emit_query_end(struct brw_context *brw) +{ + if (!brw->query.active) + return; + + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_WRITE_DEPTH_COUNT); + OUT_RELOC(brw->query.bo, + BRW_USAGE_QUERY_RESULT, + PIPE_CONTROL_GLOBAL_GTT_WRITE | + ((brw->query.index * 2 + 1) * sizeof(uint64_t))); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + + brw->query.active = GL_FALSE; + brw->query.index++; +} + +void brw_pipe_query_init( struct brw_context *brw ) +{ + brw->base.create_query = brw_query_create; + brw->base.destroy_query = brw_query_destroy; + brw->base.begin_query = brw_query_begin; + brw->base.end_query = brw_query_end; + brw->base.get_query_result = brw_query_get_result; +} + + +void brw_pipe_query_cleanup( struct brw_context *brw ) +{ + /* Unreference brw->query.bo ?? + */ +} diff --git a/src/gallium/drivers/i965/brw_pipe_rast.c b/src/gallium/drivers/i965/brw_pipe_rast.c new file mode 100644 index 00000000000..2117e91a9e4 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_rast.c @@ -0,0 +1,161 @@ + +#include "util/u_memory.h" +#include "pipe/p_defines.h" +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_pipe_rast.h" +#include "brw_wm.h" + + +static unsigned translate_fill( unsigned fill ) +{ + switch (fill) { + case PIPE_POLYGON_MODE_FILL: + return CLIP_FILL; + case PIPE_POLYGON_MODE_LINE: + return CLIP_LINE; + case PIPE_POLYGON_MODE_POINT: + return CLIP_POINT; + default: + assert(0); + return CLIP_FILL; + } +} + + +/* Calculates the key for triangle-mode clipping. Non-triangle + * clipping keys use much less information and are computed on the + * fly. + */ +static void +calculate_clip_key_rast( const struct brw_context *brw, + const struct pipe_rasterizer_state *templ, + const struct brw_rasterizer_state *rast, + struct brw_clip_prog_key *key) +{ + memset(key, 0, sizeof *key); + + if (brw->chipset.is_igdng) + key->clip_mode = BRW_CLIPMODE_KERNEL_CLIP; + else + key->clip_mode = BRW_CLIPMODE_NORMAL; + + key->do_flat_shading = templ->flatshade; + + if (templ->cull_mode == PIPE_WINDING_BOTH) { + key->clip_mode = BRW_CLIPMODE_REJECT_ALL; + return; + } + + key->fill_ccw = CLIP_CULL; + key->fill_cw = CLIP_CULL; + + if (!(templ->cull_mode & PIPE_WINDING_CCW)) { + key->fill_ccw = translate_fill(templ->fill_ccw); + } + + if (!(templ->cull_mode & PIPE_WINDING_CW)) { + key->fill_cw = translate_fill(templ->fill_cw); + } + + if (key->fill_cw == CLIP_LINE || + key->fill_ccw == CLIP_LINE || + key->fill_cw == CLIP_POINT || + key->fill_ccw == CLIP_POINT) { + key->do_unfilled = 1; + key->clip_mode = BRW_CLIPMODE_CLIP_NON_REJECTED; + } + + key->offset_ccw = templ->offset_ccw; + key->offset_cw = templ->offset_cw; + + if (templ->light_twoside && key->fill_cw != CLIP_CULL) + key->copy_bfc_cw = 1; + + if (templ->light_twoside && key->fill_ccw != CLIP_CULL) + key->copy_bfc_ccw = 1; +} + + +static void +calculate_line_stipple_rast( const struct pipe_rasterizer_state *templ, + struct brw_line_stipple *bls ) +{ + GLfloat tmp = 1.0f / (templ->line_stipple_factor + 1); + GLint tmpi = tmp * (1<<13); + + bls->header.opcode = CMD_LINE_STIPPLE_PATTERN; + bls->header.length = sizeof(*bls)/4 - 2; + bls->bits0.pattern = templ->line_stipple_pattern; + bls->bits1.repeat_count = templ->line_stipple_factor + 1; + bls->bits1.inverse_repeat_count = tmpi; +} + +static void *brw_create_rasterizer_state( struct pipe_context *pipe, + const struct pipe_rasterizer_state *templ ) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_rasterizer_state *rast; + + rast = CALLOC_STRUCT(brw_rasterizer_state); + if (rast == NULL) + return NULL; + + rast->templ = *templ; + + calculate_clip_key_rast( brw, templ, rast, &rast->clip_key ); + + if (templ->line_stipple_enable) + calculate_line_stipple_rast( templ, &rast->bls ); + + /* Caclculate lookup value for WM IZ table. + */ + if (templ->line_smooth) { + if (templ->fill_cw == PIPE_POLYGON_MODE_LINE && + templ->fill_ccw == PIPE_POLYGON_MODE_LINE) { + rast->unfilled_aa_line = AA_ALWAYS; + } + else if (templ->fill_cw == PIPE_POLYGON_MODE_LINE || + templ->fill_ccw == PIPE_POLYGON_MODE_LINE) { + rast->unfilled_aa_line = AA_SOMETIMES; + } + else { + rast->unfilled_aa_line = AA_NEVER; + } + } + else { + rast->unfilled_aa_line = AA_NEVER; + } + + return (void *)rast; +} + + +static void brw_bind_rasterizer_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + brw->curr.rast = (const struct brw_rasterizer_state *)cso; + brw->state.dirty.mesa |= PIPE_NEW_RAST; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *cso) +{ + struct brw_context *brw = brw_context(pipe); + assert((const void *)cso != (const void *)brw->curr.rast); + FREE(cso); +} + + + +void brw_pipe_rast_init( struct brw_context *brw ) +{ + brw->base.create_rasterizer_state = brw_create_rasterizer_state; + brw->base.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->base.delete_rasterizer_state = brw_delete_rasterizer_state; +} + +void brw_pipe_rast_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_rast.h b/src/gallium/drivers/i965/brw_pipe_rast.h new file mode 100644 index 00000000000..9354f01e18a --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_rast.h @@ -0,0 +1,16 @@ +#ifndef BRW_PIPE_RAST_H +#define BRW_PIPE_RAST_H + +#include "brw_clip.h" + +struct brw_rasterizer_state { + struct pipe_rasterizer_state templ; /* for draw module */ + + /* Precalculated hardware state: + */ + struct brw_clip_prog_key clip_key; + struct brw_line_stipple bls; + unsigned unfilled_aa_line; +}; + +#endif diff --git a/src/gallium/drivers/i965/brw_pipe_sampler.c b/src/gallium/drivers/i965/brw_pipe_sampler.c new file mode 100644 index 00000000000..5ddc63f57ec --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_sampler.c @@ -0,0 +1,233 @@ + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_debug.h" + + + +/* The brw (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static GLuint translate_wrap_mode( unsigned wrap ) +{ + switch( wrap ) { + case PIPE_TEX_WRAP_REPEAT: + return BRW_TEXCOORDMODE_WRAP; + + case PIPE_TEX_WRAP_CLAMP: + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return BRW_TEXCOORDMODE_CLAMP; + + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return BRW_TEXCOORDMODE_CLAMP_BORDER; + + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return BRW_TEXCOORDMODE_MIRROR; + + case PIPE_TEX_WRAP_MIRROR_CLAMP: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + return BRW_TEXCOORDMODE_MIRROR_ONCE; + + default: + return BRW_TEXCOORDMODE_WRAP; + } +} + +static GLuint translate_img_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_FILTER_NEAREST: + return BRW_MAPFILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return BRW_MAPFILTER_LINEAR; + case PIPE_TEX_FILTER_ANISO: + return BRW_MAPFILTER_ANISOTROPIC; + default: + assert(0); + return BRW_MAPFILTER_NEAREST; + } +} + +static GLuint translate_mip_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NONE: + return BRW_MIPFILTER_NONE; + case PIPE_TEX_MIPFILTER_NEAREST: + return BRW_MIPFILTER_NEAREST; + case PIPE_TEX_MIPFILTER_LINEAR: + return BRW_MIPFILTER_LINEAR; + default: + assert(0); + return BRW_MIPFILTER_NONE; + } +} + +/* XXX: not sure why there are special translations for the shadow tex + * compare functions. In particular ALWAYS is translated to NEVER. + * Is this a hardware issue? Does i965 really suffer from this? + */ +static GLuint translate_shadow_compare_func( unsigned func ) +{ + switch (func) { + case PIPE_FUNC_NEVER: + return BRW_COMPAREFUNCTION_ALWAYS; + case PIPE_FUNC_LESS: + return BRW_COMPAREFUNCTION_LEQUAL; + case PIPE_FUNC_LEQUAL: + return BRW_COMPAREFUNCTION_LESS; + case PIPE_FUNC_GREATER: + return BRW_COMPAREFUNCTION_GEQUAL; + case PIPE_FUNC_GEQUAL: + return BRW_COMPAREFUNCTION_GREATER; + case PIPE_FUNC_NOTEQUAL: + return BRW_COMPAREFUNCTION_EQUAL; + case PIPE_FUNC_EQUAL: + return BRW_COMPAREFUNCTION_NOTEQUAL; + case PIPE_FUNC_ALWAYS: + return BRW_COMPAREFUNCTION_NEVER; + default: + assert(0); + return BRW_COMPAREFUNCTION_NEVER; + } +} + + + + +static void * +brw_create_sampler_state( struct pipe_context *pipe, + const struct pipe_sampler_state *template ) +{ + struct brw_sampler *sampler = CALLOC_STRUCT(brw_sampler); + + sampler->ss0.min_filter = translate_img_filter( template->min_img_filter ); + sampler->ss0.mag_filter = translate_img_filter( template->mag_img_filter ); + sampler->ss0.mip_filter = translate_mip_filter( template->min_mip_filter ); + + + /* XXX: anisotropy logic slightly changed: + */ + if (template->max_anisotropy > 1.0) { + sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; + sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC; + + if (template->max_anisotropy > 2.0) { + sampler->ss3.max_aniso = MIN2((template->max_anisotropy - 2) / 2, + BRW_ANISORATIO_16); + } + } + + sampler->ss1.r_wrap_mode = translate_wrap_mode(template->wrap_r); + sampler->ss1.s_wrap_mode = translate_wrap_mode(template->wrap_s); + sampler->ss1.t_wrap_mode = translate_wrap_mode(template->wrap_t); + + /* Set LOD bias: + */ + sampler->ss0.lod_bias = + util_signed_fixed(CLAMP(template->lod_bias, -16, 15), 6); + + + sampler->ss0.lod_preclamp = 1; /* OpenGL mode */ + sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */ + + /* Set shadow function: + */ + if (template->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + + /* Shadowing is "enabled" by emitting a particular sampler + * message (sample_c). So need to recompile WM program when + * shadow comparison is enabled on each/any texture unit. + */ + sampler->ss0.shadow_function = + translate_shadow_compare_func(template->compare_func); + } + + /* Set BaseMipLevel, MaxLOD, MinLOD: + */ + sampler->ss0.base_level = + util_unsigned_fixed(0, 1); + + sampler->ss1.max_lod = + util_unsigned_fixed(CLAMP(template->max_lod, 0, 13), 6); + + sampler->ss1.min_lod = + util_unsigned_fixed(CLAMP(template->min_lod, 0, 13), 6); + + return (void *)sampler; +} + +static void brw_bind_sampler_state(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct brw_context *brw = brw_context(pipe); + int i; + + for (i = 0; i < num; i++) + brw->curr.sampler[i] = sampler[i]; + + for (i = num; i < brw->curr.num_samplers; i++) + brw->curr.sampler[i] = NULL; + + brw->curr.num_samplers = num; + brw->state.dirty.mesa |= PIPE_NEW_SAMPLERS; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *cso) +{ + FREE(cso); +} + +static void brw_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) +{ + struct brw_context *brw = brw_context(pipe); + int i; + + for (i = 0; i < num; i++) + pipe_texture_reference(&brw->curr.texture[i], texture[i]); + + for (i = num; i < brw->curr.num_textures; i++) + pipe_texture_reference(&brw->curr.texture[i], NULL); + + brw->curr.num_textures = num; + brw->state.dirty.mesa |= PIPE_NEW_BOUND_TEXTURES; +} + +static void brw_set_vertex_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) +{ +} + +static void brw_bind_vertex_sampler_state(struct pipe_context *pipe, + unsigned num, void **sampler) +{ +} + + +void brw_pipe_sampler_init( struct brw_context *brw ) +{ + brw->base.create_sampler_state = brw_create_sampler_state; + brw->base.delete_sampler_state = brw_delete_sampler_state; + + brw->base.set_fragment_sampler_textures = brw_set_sampler_textures; + brw->base.bind_fragment_sampler_states = brw_bind_sampler_state; + + brw->base.set_vertex_sampler_textures = brw_set_vertex_sampler_textures; + brw->base.bind_vertex_sampler_states = brw_bind_vertex_sampler_state; + +} +void brw_pipe_sampler_cleanup( struct brw_context *brw ) +{ +} diff --git a/src/gallium/drivers/i965/brw_pipe_shader.c b/src/gallium/drivers/i965/brw_pipe_shader.c new file mode 100644 index 00000000000..31a715ab655 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_shader.c @@ -0,0 +1,299 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" + +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_scan.h" + +#include "brw_context.h" +#include "brw_util.h" +#include "brw_wm.h" + + +/** + * Determine if the given shader uses complex features such as flow + * conditionals, loops, subroutines. + */ +static GLboolean has_flow_control(const struct tgsi_shader_info *info) +{ + return (info->opcode_count[TGSI_OPCODE_ARL] > 0 || + info->opcode_count[TGSI_OPCODE_IF] > 0 || + info->opcode_count[TGSI_OPCODE_ENDIF] > 0 || /* redundant - IF */ + info->opcode_count[TGSI_OPCODE_CAL] > 0 || + info->opcode_count[TGSI_OPCODE_BRK] > 0 || /* redundant - BGNLOOP */ + info->opcode_count[TGSI_OPCODE_RET] > 0 || /* redundant - CAL */ + info->opcode_count[TGSI_OPCODE_BGNLOOP] > 0); +} + + +static void scan_immediates(const struct tgsi_token *tokens, + const struct tgsi_shader_info *info, + struct brw_immediate_data *imm) +{ + struct tgsi_parse_context parse; + boolean done = FALSE; + + imm->nr = 0; + imm->data = MALLOC(info->immediate_count * 4 * sizeof(float)); + + tgsi_parse_init( &parse, tokens ); + while (!tgsi_parse_end_of_tokens( &parse ) && !done) { + tgsi_parse_token( &parse ); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: { + static const float id[4] = {0,0,0,1}; + const float *value = &parse.FullToken.FullImmediate.u[0].Float; + unsigned size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1; + unsigned i; + + for (i = 0; i < size; i++) + imm->data[imm->nr][i] = value[i]; + + for (; i < 4; i++) + imm->data[imm->nr][i] = id[i]; + + imm->nr++; + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + done = 1; + break; + } + } +} + + +static void brw_bind_fs_state( struct pipe_context *pipe, void *prog ) +{ + struct brw_fragment_shader *fs = (struct brw_fragment_shader *)prog; + struct brw_context *brw = brw_context(pipe); + + if (brw->curr.fragment_shader == fs) + return; + + if (brw->curr.fragment_shader == NULL || + fs == NULL || + memcmp(&brw->curr.fragment_shader->signature, &fs->signature, + brw_fs_signature_size(&fs->signature)) != 0) { + brw->state.dirty.mesa |= PIPE_NEW_FRAGMENT_SIGNATURE; + } + + brw->curr.fragment_shader = fs; + brw->state.dirty.mesa |= PIPE_NEW_FRAGMENT_SHADER; +} + +static void brw_bind_vs_state( struct pipe_context *pipe, void *prog ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->curr.vertex_shader = (struct brw_vertex_shader *)prog; + brw->state.dirty.mesa |= PIPE_NEW_VERTEX_SHADER; +} + + + +static void *brw_create_fs_state( struct pipe_context *pipe, + const struct pipe_shader_state *shader ) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_fragment_shader *fs; + int i; + + fs = CALLOC_STRUCT(brw_fragment_shader); + if (fs == NULL) + return NULL; + + /* Duplicate tokens, scan shader + */ + fs->id = brw->program_id++; + fs->has_flow_control = has_flow_control(&fs->info); + + fs->tokens = tgsi_dup_tokens(shader->tokens); + if (fs->tokens == NULL) + goto fail; + + tgsi_scan_shader(fs->tokens, &fs->info); + scan_immediates(fs->tokens, &fs->info, &fs->immediates); + + fs->signature.nr_inputs = fs->info.num_inputs; + for (i = 0; i < fs->info.num_inputs; i++) { + fs->signature.input[i].interp = fs->info.input_interpolate[i]; + fs->signature.input[i].semantic = fs->info.input_semantic_name[i]; + fs->signature.input[i].semantic_index = fs->info.input_semantic_index[i]; + } + + for (i = 0; i < fs->info.num_inputs; i++) + if (fs->info.input_semantic_name[i] == TGSI_SEMANTIC_POSITION) + fs->uses_depth = 1; + + if (fs->info.uses_kill) + fs->iz_lookup |= IZ_PS_KILL_ALPHATEST_BIT; + + if (fs->info.writes_z) + fs->iz_lookup |= IZ_PS_COMPUTES_DEPTH_BIT; + + return (void *)fs; + +fail: + FREE(fs); + return NULL; +} + + +static void *brw_create_vs_state( struct pipe_context *pipe, + const struct pipe_shader_state *shader ) +{ + struct brw_context *brw = brw_context(pipe); + struct brw_vertex_shader *vs; + unsigned i; + + vs = CALLOC_STRUCT(brw_vertex_shader); + if (vs == NULL) + return NULL; + + /* Duplicate tokens, scan shader + */ + vs->tokens = tgsi_dup_tokens(shader->tokens); + if (vs->tokens == NULL) + goto fail; + + tgsi_scan_shader(vs->tokens, &vs->info); + scan_immediates(vs->tokens, &vs->info, &vs->immediates); + + vs->id = brw->program_id++; + vs->has_flow_control = has_flow_control(&vs->info); + + for (i = 0; i < vs->info.num_outputs; i++) { + int index = vs->info.output_semantic_index[i]; + switch (vs->info.output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + vs->output_hpos = i; + break; + case TGSI_SEMANTIC_COLOR: + if (index == 0) + vs->output_color0 = i; + else + vs->output_color1 = i; + break; + case TGSI_SEMANTIC_BCOLOR: + if (index == 0) + vs->output_bfc0 = i; + else + vs->output_bfc1 = i; + break; +#if 0 + case TGSI_SEMANTIC_EDGEFLAG: + vs->output_edgeflag = i; + break; +#endif + } + } + + + + /* Done: + */ + return (void *)vs; + +fail: + FREE(vs); + return NULL; +} + + +static void brw_delete_fs_state( struct pipe_context *pipe, void *prog ) +{ + struct brw_fragment_shader *fs = (struct brw_fragment_shader *)prog; + + bo_reference(&fs->const_buffer, NULL); + FREE( (void *)fs->tokens ); + FREE( fs ); +} + + +static void brw_delete_vs_state( struct pipe_context *pipe, void *prog ) +{ + struct brw_fragment_shader *vs = (struct brw_fragment_shader *)prog; + + /* Delete draw shader + */ + FREE( (void *)vs->tokens ); + FREE( vs ); +} + + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(index == 0); + + if (shader == PIPE_SHADER_FRAGMENT) { + pipe_buffer_reference( &brw->curr.fragment_constants, + buf->buffer ); + + brw->state.dirty.mesa |= PIPE_NEW_FRAGMENT_CONSTANTS; + } + else { + pipe_buffer_reference( &brw->curr.vertex_constants, + buf->buffer ); + + brw->state.dirty.mesa |= PIPE_NEW_VERTEX_CONSTANTS; + } +} + + +void brw_pipe_shader_init( struct brw_context *brw ) +{ + brw->base.set_constant_buffer = brw_set_constant_buffer; + + brw->base.create_vs_state = brw_create_vs_state; + brw->base.bind_vs_state = brw_bind_vs_state; + brw->base.delete_vs_state = brw_delete_vs_state; + + brw->base.create_fs_state = brw_create_fs_state; + brw->base.bind_fs_state = brw_bind_fs_state; + brw->base.delete_fs_state = brw_delete_fs_state; +} + +void brw_pipe_shader_cleanup( struct brw_context *brw ) +{ + pipe_buffer_reference( &brw->curr.fragment_constants, NULL ); + pipe_buffer_reference( &brw->curr.vertex_constants, NULL ); +} diff --git a/src/gallium/drivers/i965/brw_pipe_vertex.c b/src/gallium/drivers/i965/brw_pipe_vertex.c new file mode 100644 index 00000000000..3d87a2853f7 --- /dev/null +++ b/src/gallium/drivers/i965/brw_pipe_vertex.c @@ -0,0 +1,78 @@ +#include "brw_context.h" + + +static void brw_set_vertex_elements( struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy(brw->curr.vertex_element, elements, count * sizeof(elements[0])); + brw->curr.num_vertex_elements = count; + + brw->state.dirty.mesa |= PIPE_NEW_VERTEX_ELEMENT; +} + + +static void brw_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) +{ + struct brw_context *brw = brw_context(pipe); + unsigned i; + + /* Check for no change */ + if (count == brw->curr.num_vertex_buffers && + memcmp(brw->curr.vertex_buffer, + buffers, + count * sizeof buffers[0]) == 0) + return; + + /* Adjust refcounts */ + for (i = 0; i < count; i++) + pipe_buffer_reference(&brw->curr.vertex_buffer[i].buffer, + buffers[i].buffer); + + for ( ; i < brw->curr.num_vertex_buffers; i++) + pipe_buffer_reference(&brw->curr.vertex_buffer[i].buffer, + NULL); + + /* Copy remaining data */ + memcpy(brw->curr.vertex_buffer, buffers, count * sizeof buffers[0]); + brw->curr.num_vertex_buffers = count; + + brw->state.dirty.mesa |= PIPE_NEW_VERTEX_BUFFER; +} + +static void brw_set_edgeflags( struct pipe_context *pipe, + const unsigned *bitfield ) +{ + /* XXX */ +} + + +void +brw_pipe_vertex_init( struct brw_context *brw ) +{ + brw->base.set_vertex_buffers = brw_set_vertex_buffers; + brw->base.set_vertex_elements = brw_set_vertex_elements; + brw->base.set_edgeflags = brw_set_edgeflags; +} + + +void +brw_pipe_vertex_cleanup( struct brw_context *brw ) +{ + + /* Release bound pipe vertex_buffers + */ + + /* Release some other stuff + */ +#if 0 + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + bo_reference(&brw->vb.inputs[i].bo, NULL); + brw->vb.inputs[i].bo = NULL; + } +#endif +} diff --git a/src/gallium/drivers/i965/brw_reg.h b/src/gallium/drivers/i965/brw_reg.h new file mode 100644 index 00000000000..a63403b6afd --- /dev/null +++ b/src/gallium/drivers/i965/brw_reg.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_REG_H +#define BRW_REG_H + +#define CMD_MI (0x0 << 29) +#define CMD_2D (0x2 << 29) +#define CMD_3D (0x3 << 29) + +#define MI_NOOP (CMD_MI | 0) +#define MI_BATCH_BUFFER_END (CMD_MI | 0xA << 23) +#define MI_FLUSH (CMD_MI | (4 << 23)) + +#define _3DSTATE_DRAWRECT_INFO_I965 (CMD_3D | (3 << 27) | (1 << 24) | 0x2) + +/** @{ + * + * PIPE_CONTROL operation, a combination MI_FLUSH and register write with + * additional flushing control. + */ +#define _3DSTATE_PIPE_CONTROL (CMD_3D | (3 << 27) | (2 << 24) | 2) +#define PIPE_CONTROL_NO_WRITE (0 << 14) +#define PIPE_CONTROL_WRITE_IMMEDIATE (1 << 14) +#define PIPE_CONTROL_WRITE_DEPTH_COUNT (2 << 14) +#define PIPE_CONTROL_WRITE_TIMESTAMP (3 << 14) +#define PIPE_CONTROL_DEPTH_STALL (1 << 13) +#define PIPE_CONTROL_WRITE_FLUSH (1 << 12) +#define PIPE_CONTROL_INSTRUCTION_FLUSH (1 << 11) +#define PIPE_CONTROL_INTERRUPT_ENABLE (1 << 8) +#define PIPE_CONTROL_PPGTT_WRITE (0 << 2) +#define PIPE_CONTROL_GLOBAL_GTT_WRITE (1 << 2) + +/** @} */ + +#define XY_SETUP_BLT_CMD (CMD_2D | (0x01 << 22) | 6) +#define XY_COLOR_BLT_CMD (CMD_2D | (0x50 << 22) | 4) +#define XY_SRC_COPY_BLT_CMD (CMD_2D | (0x53 << 22) | 6) + +/* BR00 */ +#define XY_BLT_WRITE_ALPHA (1 << 21) +#define XY_BLT_WRITE_RGB (1 << 20) +#define XY_SRC_TILED (1 << 15) +#define XY_DST_TILED (1 << 11) + +/* BR13 */ +#define BR13_565 (0x1 << 24) +#define BR13_8888 (0x3 << 24) + +#define FENCE_LINEAR 0 +#define FENCE_XMAJOR 1 +#define FENCE_YMAJOR 2 + + + +/* PCI IDs + */ +#define PCI_CHIP_I965_G 0x29A2 +#define PCI_CHIP_I965_Q 0x2992 +#define PCI_CHIP_I965_G_1 0x2982 +#define PCI_CHIP_I946_GZ 0x2972 +#define PCI_CHIP_I965_GM 0x2A02 +#define PCI_CHIP_I965_GME 0x2A12 + +#define PCI_CHIP_GM45_GM 0x2A42 + +#define PCI_CHIP_IGD_E_G 0x2E02 +#define PCI_CHIP_Q45_G 0x2E12 +#define PCI_CHIP_G45_G 0x2E22 +#define PCI_CHIP_G41_G 0x2E32 +#define PCI_CHIP_B43_G 0x2E42 + +#define PCI_CHIP_ILD_G 0x0042 +#define PCI_CHIP_ILM_G 0x0046 + +struct brw_chipset { + unsigned pci_id:16; + unsigned is_965:1; + unsigned is_igdng:1; + unsigned is_g4x:1; + unsigned pad:13; +}; + + +/* XXX: hacks + */ +#define VERT_RESULT_HPOS 0 /* not always true */ +#define VERT_RESULT_PSIZ 10000 /* disabled */ + + +#endif diff --git a/src/gallium/drivers/i965/brw_screen.c b/src/gallium/drivers/i965/brw_screen.c new file mode 100644 index 00000000000..0ecacac9a3a --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen.c @@ -0,0 +1,403 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_inlines.h" +#include "util/u_memory.h" +#include "util/u_string.h" + +#include "brw_reg.h" +#include "brw_context.h" +#include "brw_screen.h" +#include "brw_winsys.h" +#include "brw_debug.h" + +#ifdef DEBUG +static const struct debug_named_value debug_names[] = { + { "tex", DEBUG_TEXTURE}, + { "state", DEBUG_STATE}, + { "ioctl", DEBUG_IOCTL}, + { "blit", DEBUG_BLIT}, + { "curbe", DEBUG_CURBE}, + { "fall", DEBUG_FALLBACKS}, + { "verb", DEBUG_VERBOSE}, + { "bat", DEBUG_BATCH}, + { "pix", DEBUG_PIXEL}, + { "wins", DEBUG_WINSYS}, + { "min", DEBUG_MIN_URB}, + { "dis", DEBUG_DISASSEM}, + { "sync", DEBUG_SYNC}, + { "prim", DEBUG_PRIMS }, + { "vert", DEBUG_VERTS }, + { "dma", DEBUG_DMA }, + { "san", DEBUG_SANITY }, + { "sleep", DEBUG_SLEEP }, + { "stats", DEBUG_STATS }, + { "sing", DEBUG_SINGLE_THREAD }, + { "thre", DEBUG_SINGLE_THREAD }, + { "wm", DEBUG_WM }, + { "urb", DEBUG_URB }, + { "vs", DEBUG_VS }, + { NULL, 0 } +}; + +static const struct debug_named_value dump_names[] = { + { "asm", DUMP_ASM}, + { "state", DUMP_STATE}, + { "batch", DUMP_BATCH}, + { NULL, 0 } +}; + +int BRW_DEBUG = 0; +int BRW_DUMP = 0; + +#endif + + +/* + * Probe functions + */ + + +static const char * +brw_get_vendor(struct pipe_screen *screen) +{ + return "VMware, Inc."; +} + +static const char * +brw_get_name(struct pipe_screen *screen) +{ + static char buffer[128]; + const char *chipset; + + switch (brw_screen(screen)->chipset.pci_id) { + case PCI_CHIP_I965_G: + chipset = "I965_G"; + break; + case PCI_CHIP_I965_Q: + chipset = "I965_Q"; + break; + case PCI_CHIP_I965_G_1: + chipset = "I965_G_1"; + break; + case PCI_CHIP_I946_GZ: + chipset = "I946_GZ"; + break; + case PCI_CHIP_I965_GM: + chipset = "I965_GM"; + break; + case PCI_CHIP_I965_GME: + chipset = "I965_GME"; + break; + case PCI_CHIP_GM45_GM: + chipset = "GM45_GM"; + break; + case PCI_CHIP_IGD_E_G: + chipset = "IGD_E_G"; + break; + case PCI_CHIP_Q45_G: + chipset = "Q45_G"; + break; + case PCI_CHIP_G45_G: + chipset = "G45_G"; + break; + case PCI_CHIP_G41_G: + chipset = "G41_G"; + break; + case PCI_CHIP_B43_G: + chipset = "B43_G"; + break; + case PCI_CHIP_ILD_G: + chipset = "ILD_G"; + break; + case PCI_CHIP_ILM_G: + chipset = "ILM_G"; + break; + } + + util_snprintf(buffer, sizeof(buffer), "i965 (chipset: %s)", chipset); + return buffer; +} + +static int +brw_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + +static float +brw_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + +static boolean +brw_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) +{ + static const enum pipe_format tex_supported[] = { + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_L16_UNORM, + /*PIPE_FORMAT_I16_UNORM,*/ + /*PIPE_FORMAT_A16_UNORM,*/ + PIPE_FORMAT_A8L8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_A1R5G5B5_UNORM, + PIPE_FORMAT_A4R4G4B4_UNORM, + PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + /* video */ + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + /* compressed */ + /*PIPE_FORMAT_FXT1_RGBA,*/ + PIPE_FORMAT_DXT1_RGB, + PIPE_FORMAT_DXT1_RGBA, + PIPE_FORMAT_DXT3_RGBA, + PIPE_FORMAT_DXT5_RGBA, + /* sRGB */ + PIPE_FORMAT_R8G8B8A8_SRGB, + PIPE_FORMAT_A8L8_SRGB, + PIPE_FORMAT_L8_SRGB, + PIPE_FORMAT_DXT1_SRGB, + /* depth */ + PIPE_FORMAT_Z32_FLOAT, + PIPE_FORMAT_X8Z24_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z16_UNORM, + /* signed */ + PIPE_FORMAT_R8G8_SNORM, + PIPE_FORMAT_R8G8B8A8_SNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + static const enum pipe_format render_supported[] = { + PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + static const enum pipe_format depth_supported[] = { + PIPE_FORMAT_Z32_FLOAT, + PIPE_FORMAT_X8Z24_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + const enum pipe_format *list; + uint i; + + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) + list = depth_supported; + else if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) + list = render_supported; + else + list = tex_supported; + + for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { + if (list[i] == format) + return TRUE; + } + + return FALSE; +} + + +/* + * Fence functions + */ + + +static void +brw_fence_reference(struct pipe_screen *screen, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + +static int +brw_fence_signalled(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flags) +{ + return 0; /* XXX shouldn't this be a boolean? */ +} + +static int +brw_fence_finish(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flags) +{ + return 0; +} + + +/* + * Generic functions + */ + + +static void +brw_destroy_screen(struct pipe_screen *screen) +{ + struct brw_screen *bscreen = brw_screen(screen); + + if (bscreen->sws) + bscreen->sws->destroy(bscreen->sws); + + FREE(bscreen); +} + +/** + * Create a new brw_screen object + */ +struct pipe_screen * +brw_create_screen(struct brw_winsys_screen *sws, uint pci_id) +{ + struct brw_screen *bscreen; + struct brw_chipset chipset; + +#ifdef DEBUG + BRW_DEBUG = debug_get_flags_option("BRW_DEBUG", debug_names, 0); + BRW_DEBUG |= debug_get_flags_option("INTEL_DEBUG", debug_names, 0); + BRW_DEBUG |= DEBUG_STATS | DEBUG_MIN_URB | DEBUG_WM; + + BRW_DUMP = debug_get_flags_option("BRW_DUMP", dump_names, 0); +#endif + + memset(&chipset, 0, sizeof chipset); + + chipset.pci_id = pci_id; + + switch (pci_id) { + case PCI_CHIP_I965_G: + case PCI_CHIP_I965_Q: + case PCI_CHIP_I965_G_1: + case PCI_CHIP_I946_GZ: + case PCI_CHIP_I965_GM: + case PCI_CHIP_I965_GME: + chipset.is_965 = TRUE; + break; + + case PCI_CHIP_GM45_GM: + case PCI_CHIP_IGD_E_G: + case PCI_CHIP_Q45_G: + case PCI_CHIP_G45_G: + case PCI_CHIP_G41_G: + case PCI_CHIP_B43_G: + chipset.is_g4x = TRUE; + break; + + case PCI_CHIP_ILD_G: + case PCI_CHIP_ILM_G: + chipset.is_igdng = TRUE; + break; + + default: + debug_printf("%s: unknown pci id 0x%x, cannot create screen\n", + __FUNCTION__, pci_id); + return NULL; + } + + + bscreen = CALLOC_STRUCT(brw_screen); + if (!bscreen) + return NULL; + + bscreen->chipset = chipset; + bscreen->sws = sws; + bscreen->base.winsys = NULL; + bscreen->base.destroy = brw_destroy_screen; + bscreen->base.get_name = brw_get_name; + bscreen->base.get_vendor = brw_get_vendor; + bscreen->base.get_param = brw_get_param; + bscreen->base.get_paramf = brw_get_paramf; + bscreen->base.is_format_supported = brw_is_format_supported; + bscreen->base.fence_reference = brw_fence_reference; + bscreen->base.fence_signalled = brw_fence_signalled; + bscreen->base.fence_finish = brw_fence_finish; + + brw_screen_tex_init(bscreen); + brw_screen_tex_surface_init(bscreen); + brw_screen_buffer_init(bscreen); + + bscreen->no_tiling = debug_get_option("BRW_NO_TILING", FALSE) != NULL; + + + return &bscreen->base; +} diff --git a/src/gallium/drivers/i965/brw_screen.h b/src/gallium/drivers/i965/brw_screen.h new file mode 100644 index 00000000000..7226d9228b7 --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen.h @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_SCREEN_H +#define BRW_SCREEN_H + +#include "pipe/p_state.h" +#include "pipe/p_screen.h" + +#include "brw_reg.h" +#include "brw_structs.h" + +struct brw_winsys_screen; + + +/** + * Subclass of pipe_screen + */ +struct brw_screen +{ + struct pipe_screen base; + struct brw_chipset chipset; + struct brw_winsys_screen *sws; + boolean no_tiling; +}; + +/** + * Subclass of pipe_transfer + */ +struct brw_transfer +{ + struct pipe_transfer base; + + unsigned offset; +}; + +struct brw_buffer +{ + struct pipe_buffer base; + + /* One of either bo or user_buffer will be non-null, depending on + * whether this is a hardware or user buffer. + */ + struct brw_winsys_buffer *bo; + void *user_buffer; + + /* Mapped pointer?? + */ + void *ptr; +}; + + +union brw_surface_id { + struct { + unsigned face:3; + unsigned zslice:13; + unsigned level:16; + } bits; + unsigned value; +}; + + +struct brw_surface +{ + struct pipe_surface base; + + union brw_surface_id id; + unsigned cpp; + unsigned pitch; + unsigned draw_offset; + unsigned tiling; + + struct brw_surface_state ss; + struct brw_winsys_buffer *bo; + struct brw_surface *next, *prev; +}; + + + +struct brw_texture +{ + struct pipe_texture base; + struct brw_winsys_buffer *bo; + struct brw_surface_state ss; + + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + boolean compressed; + unsigned brw_target; + unsigned pitch; + unsigned tiling; + unsigned cpp; + unsigned total_height; + + struct brw_surface views[2]; +}; + + + +/* + * Cast wrappers + */ +static INLINE struct brw_screen * +brw_screen(struct pipe_screen *pscreen) +{ + return (struct brw_screen *) pscreen; +} + +static INLINE struct brw_transfer * +brw_transfer(struct pipe_transfer *transfer) +{ + return (struct brw_transfer *)transfer; +} + +static INLINE struct brw_surface * +brw_surface(struct pipe_surface *surface) +{ + return (struct brw_surface *)surface; +} + +static INLINE struct brw_buffer * +brw_buffer(struct pipe_buffer *buffer) +{ + return (struct brw_buffer *)buffer; +} + +static INLINE struct brw_texture * +brw_texture(struct pipe_texture *texture) +{ + return (struct brw_texture *)texture; +} + + +/* Pipe buffer helpers + */ +static INLINE boolean +brw_buffer_is_user_buffer( const struct pipe_buffer *buf ) +{ + return ((const struct brw_buffer *)buf)->user_buffer != NULL; +} + +unsigned +brw_surface_pitch( const struct pipe_surface *surface ); + +/*********************************************************************** + * Internal functions + */ +GLboolean brw_texture_layout(struct brw_screen *brw_screen, + struct brw_texture *tex ); + +void brw_update_texture( struct brw_screen *brw_screen, + struct brw_texture *tex ); + + +void brw_screen_tex_init( struct brw_screen *brw_screen ); +void brw_screen_tex_surface_init( struct brw_screen *brw_screen ); + +void brw_screen_buffer_init(struct brw_screen *brw_screen); + + +boolean brw_is_texture_referenced_by_bo( struct brw_screen *brw_screen, + struct pipe_texture *texture, + unsigned face, + unsigned level, + struct brw_winsys_buffer *bo ); + +boolean brw_is_buffer_referenced_by_bo( struct brw_screen *brw_screen, + struct pipe_buffer *buffer, + struct brw_winsys_buffer *bo ); + + + +#endif /* BRW_SCREEN_H */ diff --git a/src/gallium/drivers/i965/brw_screen_buffers.c b/src/gallium/drivers/i965/brw_screen_buffers.c new file mode 100644 index 00000000000..d8141a3f5b9 --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen_buffers.c @@ -0,0 +1,202 @@ + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "brw_screen.h" +#include "brw_winsys.h" + + + +static void * +brw_buffer_map_range( struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned offset, + unsigned length, + unsigned usage ) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_winsys_screen *sws = bscreen->sws; + struct brw_buffer *buf = brw_buffer( buffer ); + + if (buf->user_buffer) + return buf->user_buffer; + + return sws->bo_map( buf->bo, + BRW_DATA_OTHER, + offset, + length, + (usage & PIPE_BUFFER_USAGE_CPU_WRITE) ? TRUE : FALSE, + (usage & PIPE_BUFFER_USAGE_DISCARD) ? TRUE : FALSE, + (usage & PIPE_BUFFER_USAGE_FLUSH_EXPLICIT) ? TRUE : FALSE); +} + +static void * +brw_buffer_map( struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned usage ) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_winsys_screen *sws = bscreen->sws; + struct brw_buffer *buf = brw_buffer( buffer ); + + if (buf->user_buffer) + return buf->user_buffer; + + return sws->bo_map( buf->bo, + BRW_DATA_OTHER, + 0, + buf->base.size, + (usage & PIPE_BUFFER_USAGE_CPU_WRITE) ? TRUE : FALSE, + FALSE, + FALSE); +} + + +static void +brw_buffer_flush_mapped_range( struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned offset, + unsigned length ) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_winsys_screen *sws = bscreen->sws; + struct brw_buffer *buf = brw_buffer( buffer ); + + if (buf->user_buffer) + return; + + sws->bo_flush_range( buf->bo, + offset, + length ); +} + + +static void +brw_buffer_unmap( struct pipe_screen *screen, + struct pipe_buffer *buffer ) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_winsys_screen *sws = bscreen->sws; + struct brw_buffer *buf = brw_buffer( buffer ); + + if (buf->bo) + sws->bo_unmap(buf->bo); +} + +static void +brw_buffer_destroy( struct pipe_buffer *buffer ) +{ + struct brw_buffer *buf = brw_buffer( buffer ); + + assert(!p_atomic_read(&buffer->reference.count)); + + bo_reference(&buf->bo, NULL); + FREE(buf); +} + + +static struct pipe_buffer * +brw_buffer_create(struct pipe_screen *screen, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_winsys_screen *sws = bscreen->sws; + struct brw_buffer *buf; + unsigned buffer_type; + enum pipe_error ret; + + buf = CALLOC_STRUCT(brw_buffer); + if (!buf) + return NULL; + + pipe_reference_init(&buf->base.reference, 1); + buf->base.screen = screen; + buf->base.alignment = alignment; + buf->base.usage = usage; + buf->base.size = size; + + switch (usage & (PIPE_BUFFER_USAGE_VERTEX | + PIPE_BUFFER_USAGE_INDEX | + PIPE_BUFFER_USAGE_PIXEL | + PIPE_BUFFER_USAGE_CONSTANT)) + { + case PIPE_BUFFER_USAGE_VERTEX: + case PIPE_BUFFER_USAGE_INDEX: + case (PIPE_BUFFER_USAGE_VERTEX|PIPE_BUFFER_USAGE_INDEX): + buffer_type = BRW_BUFFER_TYPE_VERTEX; + break; + + case PIPE_BUFFER_USAGE_PIXEL: + buffer_type = BRW_BUFFER_TYPE_PIXEL; + break; + + case PIPE_BUFFER_USAGE_CONSTANT: + buffer_type = BRW_BUFFER_TYPE_SHADER_CONSTANTS; + break; + + default: + buffer_type = BRW_BUFFER_TYPE_GENERIC; + break; + } + + ret = sws->bo_alloc( sws, buffer_type, + size, alignment, + &buf->bo ); + if (ret != PIPE_OK) + return NULL; + + return &buf->base; +} + + +static struct pipe_buffer * +brw_user_buffer_create(struct pipe_screen *screen, + void *ptr, + unsigned bytes) +{ + struct brw_buffer *buf; + + buf = CALLOC_STRUCT(brw_buffer); + if (!buf) + return NULL; + + buf->user_buffer = ptr; + + pipe_reference_init(&buf->base.reference, 1); + buf->base.screen = screen; + buf->base.alignment = 1; + buf->base.usage = 0; + buf->base.size = bytes; + + return &buf->base; +} + + +boolean brw_is_buffer_referenced_by_bo( struct brw_screen *brw_screen, + struct pipe_buffer *buffer, + struct brw_winsys_buffer *bo ) +{ + struct brw_buffer *buf = brw_buffer(buffer); + if (buf->bo == NULL) + return FALSE; + + return brw_screen->sws->bo_references( bo, buf->bo ); +} + + +void brw_screen_buffer_init(struct brw_screen *brw_screen) +{ + brw_screen->base.buffer_create = brw_buffer_create; + brw_screen->base.user_buffer_create = brw_user_buffer_create; + brw_screen->base.buffer_map = brw_buffer_map; + brw_screen->base.buffer_map_range = brw_buffer_map_range; + brw_screen->base.buffer_flush_mapped_range = brw_buffer_flush_mapped_range; + brw_screen->base.buffer_unmap = brw_buffer_unmap; + brw_screen->base.buffer_destroy = brw_buffer_destroy; +} diff --git a/src/gallium/drivers/i965/brw_screen_surface.c b/src/gallium/drivers/i965/brw_screen_surface.c new file mode 100644 index 00000000000..e2b9954e596 --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen_surface.c @@ -0,0 +1,262 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" +#include "util/u_simple_list.h" +#include "util/u_math.h" + +#include "pipe/p_screen.h" +#include "brw_screen.h" +#include "brw_defines.h" +#include "brw_winsys.h" + +enum { + BRW_VIEW_LINEAR, + BRW_VIEW_IN_PLACE +}; + + +static boolean need_linear_view( struct brw_screen *brw_screen, + struct brw_texture *brw_texture, + union brw_surface_id id, + unsigned usage ) +{ +#if 0 + /* XXX: what about IDGNG? + */ + if (!BRW_IS_G4X(brw->brw_screen->pci_id)) + { + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i]; + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + + /* The original gen4 hardware couldn't set up WM surfaces pointing + * at an offset within a tile, which can happen when rendering to + * anything but the base level of a texture or the +X face/0 depth. + * This was fixed with the 4 Series hardware. + * + * For these original chips, you would have to make the depth and + * color destination surfaces include information on the texture + * type, LOD, face, and various limits to use them as a destination. + * + * This is easy in Gallium as surfaces are all backed by + * textures, but there's also a nasty requirement that the depth + * and the color surfaces all be of the same LOD, which is + * harder to get around as we can't look at a surface in + * isolation and decide if it's legal. + * + * Instead, end up being pessimistic and say that for i965, + * ... ?? + */ + if (brw_tex->tiling != I915_TILING_NONE && + (brw_tex_image_offset(brw_tex, face, level, zslize) & 4095)) { + if (BRW_DEBUG & DEBUG_VIEW) + debug_printf("%s: need surface view for non-aligned tex image\n", + __FUNCTION__); + return GL_TRUE; + } + } +#endif + + /* Tiled 3d textures don't have subsets that look like 2d surfaces: + */ + + /* Everything else should be fine to render to in-place: + */ + return GL_FALSE; +} + +/* Look at all texture views and figure out if any of them need to be + * back-copied into the texture for sampling + */ +void brw_update_texture( struct brw_screen *brw_screen, + struct brw_texture *tex ) +{ + /* currently nothing to do */ +} + + +/* Create a new surface with linear layout to serve as a render-target + * where it would be illegal (perhaps due to tiling constraints) to do + * this in-place. + * + * Currently not implmented, not sure if it's needed. + */ +static struct brw_surface *create_linear_view( struct brw_screen *brw_screen, + struct brw_texture *tex, + union brw_surface_id id, + unsigned usage ) +{ + return NULL; +} + + +/* Create a pipe_surface that just points directly into the existing + * texture's storage. + */ +static struct brw_surface *create_in_place_view( struct brw_screen *brw_screen, + struct brw_texture *tex, + union brw_surface_id id, + unsigned usage ) +{ + struct brw_surface *surface; + + surface = CALLOC_STRUCT(brw_surface); + if (surface == NULL) + return NULL; + + pipe_reference_init(&surface->base.reference, 1); + + /* XXX: ignoring render-to-slice-of-3d-texture + */ + assert(id.bits.zslice == 0); + + surface->base.format = tex->base.format; + surface->base.width = u_minify(tex->base.width0, id.bits.level); + surface->base.height = u_minify(tex->base.height0, id.bits.level); + surface->base.offset = tex->image_offset[id.bits.level][id.bits.face]; + surface->base.usage = usage; + surface->base.zslice = id.bits.zslice; + surface->base.face = id.bits.face; + surface->base.level = id.bits.level; + surface->id = id; + surface->cpp = tex->cpp; + surface->pitch = tex->pitch; + surface->tiling = tex->tiling; + + bo_reference( &surface->bo, tex->bo ); + pipe_texture_reference( &surface->base.texture, &tex->base ); + + surface->ss.ss0.surface_format = tex->ss.ss0.surface_format; + surface->ss.ss0.surface_type = BRW_SURFACE_2D; + + if (tex->tiling == BRW_TILING_NONE) { + surface->ss.ss1.base_addr = surface->base.offset; + } else { + uint32_t tile_offset = surface->base.offset % 4096; + + surface->ss.ss1.base_addr = surface->base.offset - tile_offset; + + if (brw_screen->chipset.is_g4x) { + if (tex->tiling == BRW_TILING_X) { + /* Note that the low bits of these fields are missing, so + * there's the possibility of getting in trouble. + */ + surface->ss.ss5.x_offset = (tile_offset % 512) / tex->cpp / 4; + surface->ss.ss5.y_offset = tile_offset / 512 / 2; + } else { + surface->ss.ss5.x_offset = (tile_offset % 128) / tex->cpp / 4; + surface->ss.ss5.y_offset = tile_offset / 128 / 2; + } + } + else { + assert(tile_offset == 0); + } + } + +#if 0 + if (region_bo != NULL) + surface->ss.ss1.base_addr += region_bo->offset; /* reloc */ +#endif + + surface->ss.ss2.width = surface->base.width - 1; + surface->ss.ss2.height = surface->base.height - 1; + surface->ss.ss3.tiled_surface = tex->ss.ss3.tiled_surface; + surface->ss.ss3.tile_walk = tex->ss.ss3.tile_walk; + surface->ss.ss3.pitch = tex->ss.ss3.pitch; + + return surface; +} + +/* Get a surface which is view into a texture + */ +static struct pipe_surface *brw_get_tex_surface(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, + unsigned zslice, + unsigned usage ) +{ + struct brw_texture *tex = brw_texture(pt); + struct brw_screen *bscreen = brw_screen(screen); + struct brw_surface *surface; + union brw_surface_id id; + int type; + + id.bits.face = face; + id.bits.level = level; + id.bits.zslice = zslice; + + if (need_linear_view(bscreen, tex, id, usage)) + type = BRW_VIEW_LINEAR; + else + type = BRW_VIEW_IN_PLACE; + + + foreach (surface, &tex->views[type]) { + if (id.value == surface->id.value) + return &surface->base; + } + + switch (type) { + case BRW_VIEW_LINEAR: + surface = create_linear_view( bscreen, tex, id, usage ); + break; + case BRW_VIEW_IN_PLACE: + surface = create_in_place_view( bscreen, tex, id, usage ); + break; + default: + return NULL; + } + + insert_at_head( &tex->views[type], surface ); + return &surface->base; +} + + +static void brw_tex_surface_destroy( struct pipe_surface *surf ) +{ + struct brw_surface *surface = brw_surface(surf); + + /* Unreference texture, shared buffer: + */ + remove_from_list(surface); + bo_reference(&surface->bo, NULL); + pipe_texture_reference( &surface->base.texture, NULL ); + + + FREE(surface); +} + + +void brw_screen_tex_surface_init( struct brw_screen *brw_screen ) +{ + brw_screen->base.get_tex_surface = brw_get_tex_surface; + brw_screen->base.tex_surface_destroy = brw_tex_surface_destroy; +} diff --git a/src/gallium/drivers/i965/brw_screen_tex_layout.c b/src/gallium/drivers/i965/brw_screen_tex_layout.c new file mode 100644 index 00000000000..894f4bea401 --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen_tex_layout.c @@ -0,0 +1,414 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + +#include "pipe/p_format.h" + +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "brw_screen.h" +#include "brw_debug.h" +#include "brw_winsys.h" + +/* Code to layout images in a mipmap tree for i965. + */ + +static int +brw_tex_pitch_align (struct brw_texture *tex, + int pitch) +{ + if (!tex->compressed) { + int pitch_align; + + switch (tex->tiling) { + case BRW_TILING_X: + pitch_align = 512; + break; + case BRW_TILING_Y: + pitch_align = 128; + break; + default: + /* XXX: Untiled pitch alignment of 64 bytes for now to allow + * render-to-texture to work in all cases. This should + * probably be replaced at some point by some scheme to only + * do this when really necessary, for example standalone + * render target views. + */ + pitch_align = 64; + break; + } + + pitch = align(pitch * tex->cpp, pitch_align); + pitch /= tex->cpp; + } + + return pitch; +} + + +static void +brw_tex_alignment_unit(enum pipe_format pf, + GLuint *w, GLuint *h) +{ + switch (pf) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + case PIPE_FORMAT_DXT1_SRGB: + case PIPE_FORMAT_DXT1_SRGBA: + case PIPE_FORMAT_DXT3_SRGBA: + case PIPE_FORMAT_DXT5_SRGBA: + *w = 4; + *h = 4; + break; + + default: + *w = 4; + *h = 2; + break; + } +} + + +static void +brw_tex_set_level_info(struct brw_texture *tex, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, + GLuint w, GLuint h, GLuint d) +{ + + if (BRW_DEBUG & DEBUG_TEXTURE) + debug_printf("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, tex->level_offset[level]); + + assert(tex->image_offset[level] == NULL); + assert(nr_images >= 1); + + tex->level_offset[level] = (x + y * tex->pitch) * tex->cpp; + tex->nr_images[level] = nr_images; + + tex->image_offset[level] = MALLOC(nr_images * sizeof(GLuint)); + tex->image_offset[level][0] = 0; +} + + +static void +brw_tex_set_image_offset(struct brw_texture *tex, + GLuint level, GLuint img, + GLuint x, GLuint y, + GLuint offset) +{ + assert((x == 0 && y == 0) || img != 0 || level != 0); + assert(img < tex->nr_images[level]); + + if (BRW_DEBUG & DEBUG_TEXTURE) + debug_printf("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, + tex->image_offset[level][img]); + + tex->image_offset[level][img] = (x + y * tex->pitch) * tex->cpp + offset; +} + + + +static void brw_layout_2d( struct brw_texture *tex ) +{ + GLuint align_h = 2, align_w = 4; + GLuint level; + GLuint x = 0; + GLuint y = 0; + GLuint width = tex->base.width0; + GLuint height = tex->base.height0; + + tex->pitch = tex->base.width0; + brw_tex_alignment_unit(tex->base.format, &align_w, &align_h); + + if (tex->compressed) { + tex->pitch = align(tex->base.width0, align_w); + } + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (tex->base.last_level > 0) { + GLuint mip1_width; + + if (tex->compressed) { + mip1_width = (align(u_minify(tex->base.width0, 1), align_w) + + align(u_minify(tex->base.width0, 2), align_w)); + } else { + mip1_width = (align(u_minify(tex->base.width0, 1), align_w) + + u_minify(tex->base.width0, 2)); + } + + if (mip1_width > tex->pitch) { + tex->pitch = mip1_width; + } + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + tex->pitch = brw_tex_pitch_align (tex, tex->pitch); + tex->total_height = 0; + + for ( level = 0 ; level <= tex->base.last_level ; level++ ) { + GLuint img_height; + + brw_tex_set_level_info(tex, level, 1, x, y, width, height, 1); + + if (tex->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + tex->total_height = MAX2(tex->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = u_minify(width, 1); + height = u_minify(height, 1); + } +} + + +static boolean +brw_layout_cubemap_idgng( struct brw_texture *tex ) +{ + GLuint align_h = 2, align_w = 4; + GLuint level; + GLuint x = 0; + GLuint y = 0; + GLuint width = tex->base.width0; + GLuint height = tex->base.height0; + GLuint qpitch = 0; + GLuint y_pitch = 0; + + tex->pitch = tex->base.width0; + brw_tex_alignment_unit(tex->base.format, &align_w, &align_h); + y_pitch = align(height, align_h); + + if (tex->compressed) { + tex->pitch = align(tex->base.width0, align_w); + } + + if (tex->base.last_level != 0) { + GLuint mip1_width; + + if (tex->compressed) { + mip1_width = (align(u_minify(tex->base.width0, 1), align_w) + + align(u_minify(tex->base.width0, 2), align_w)); + } else { + mip1_width = (align(u_minify(tex->base.width0, 1), align_w) + + u_minify(tex->base.width0, 2)); + } + + if (mip1_width > tex->pitch) { + tex->pitch = mip1_width; + } + } + + tex->pitch = brw_tex_pitch_align(tex, tex->pitch); + + if (tex->compressed) { + qpitch = ((y_pitch + + align(u_minify(y_pitch, 1), align_h) + + 11 * align_h) / 4) * tex->pitch * tex->cpp; + + tex->total_height = ((y_pitch + + align(u_minify(y_pitch, 1), align_h) + + 11 * align_h) / 4) * 6; + } else { + qpitch = (y_pitch + + align(u_minify(y_pitch, 1), align_h) + + 11 * align_h) * tex->pitch * tex->cpp; + + tex->total_height = (y_pitch + + align(u_minify(y_pitch, 1), align_h) + + 11 * align_h) * 6; + } + + for (level = 0; level <= tex->base.last_level; level++) { + GLuint img_height; + GLuint nr_images = 6; + GLuint q = 0; + + brw_tex_set_level_info(tex, level, nr_images, x, y, width, height, 1); + + for (q = 0; q < nr_images; q++) + brw_tex_set_image_offset(tex, level, q, x, y, q * qpitch); + + if (tex->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + if (level == 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = u_minify(width, 1); + height = u_minify(height, 1); + } + + return TRUE; +} + + +static boolean +brw_layout_3d_cube( struct brw_texture *tex ) +{ + GLuint width = tex->base.width0; + GLuint height = tex->base.height0; + GLuint depth = tex->base.depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + GLuint align_h = 2; + GLuint align_w = 4; + + tex->total_height = 0; + brw_tex_alignment_unit(tex->base.format, &align_w, &align_h); + + if (tex->compressed) { + tex->pitch = align(width, align_w); + pack_y_pitch = (height + 3) / 4; + } else { + tex->pitch = brw_tex_pitch_align(tex, tex->base.width0); + pack_y_pitch = align(tex->base.height0, align_h); + } + + pack_x_pitch = width; + pack_x_nr = 1; + + for (level = 0 ; level <= tex->base.last_level ; level++) { + GLuint nr_images = tex->base.target == PIPE_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + brw_tex_set_level_info(tex, level, nr_images, + 0, tex->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + brw_tex_set_image_offset(tex, level, q, x, y, 0); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + tex->total_height += y; + width = u_minify(width, 1); + height = u_minify(height, 1); + depth = u_minify(depth, 1); + + if (tex->compressed) { + pack_y_pitch = (height + 3) / 4; + + if (pack_x_pitch > align(width, align_w)) { + pack_x_pitch = align(width, align_w); + pack_x_nr <<= 1; + } + } else { + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= tex->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + pack_y_pitch = align(pack_y_pitch, align_h); + } + } + } + + /* The 965's sampler lays cachelines out according to how accesses + * in the texture surfaces run, so they may be "vertical" through + * memory. As a result, the docs say in Surface Padding Requirements: + * Sampling Engine Surfaces that two extra rows of padding are required. + */ + if (tex->base.target == PIPE_TEXTURE_CUBE) + tex->total_height += 2; + + return TRUE; +} + + + +GLboolean brw_texture_layout(struct brw_screen *brw_screen, + struct brw_texture *tex ) +{ + switch (tex->base.target) { + case PIPE_TEXTURE_CUBE: + if (brw_screen->chipset.is_igdng) + brw_layout_cubemap_idgng( tex ); + else + brw_layout_3d_cube( tex ); + break; + + case PIPE_TEXTURE_3D: + brw_layout_3d_cube( tex ); + break; + + default: + brw_layout_2d( tex ); + break; + } + + if (BRW_DEBUG & DEBUG_TEXTURE) + debug_printf("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + tex->pitch, + tex->total_height, + tex->cpp, + tex->pitch * tex->total_height * tex->cpp ); + + return GL_TRUE; +} diff --git a/src/gallium/drivers/i965/brw_screen_texture.c b/src/gallium/drivers/i965/brw_screen_texture.c new file mode 100644 index 00000000000..feb9d5f7657 --- /dev/null +++ b/src/gallium/drivers/i965/brw_screen_texture.c @@ -0,0 +1,573 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" +#include "util/u_simple_list.h" +#include "util/u_format.h" + +#include "brw_screen.h" +#include "brw_defines.h" +#include "brw_structs.h" +#include "brw_winsys.h" + + + +static GLuint translate_tex_target( unsigned target ) +{ + switch (target) { + case PIPE_TEXTURE_1D: + return BRW_SURFACE_1D; + + case PIPE_TEXTURE_2D: + return BRW_SURFACE_2D; + + case PIPE_TEXTURE_3D: + return BRW_SURFACE_3D; + + case PIPE_TEXTURE_CUBE: + return BRW_SURFACE_CUBE; + + default: + assert(0); + return BRW_SURFACE_1D; + } +} + + +static GLuint translate_tex_format( enum pipe_format pf ) +{ + switch( pf ) { + case PIPE_FORMAT_L8_UNORM: + return BRW_SURFACEFORMAT_L8_UNORM; + + case PIPE_FORMAT_I8_UNORM: + return BRW_SURFACEFORMAT_I8_UNORM; + + case PIPE_FORMAT_A8_UNORM: + return BRW_SURFACEFORMAT_A8_UNORM; + + case PIPE_FORMAT_L16_UNORM: + return BRW_SURFACEFORMAT_L16_UNORM; + + /* XXX: Add these to gallium + case PIPE_FORMAT_I16_UNORM: + return BRW_SURFACEFORMAT_I16_UNORM; + + case PIPE_FORMAT_A16_UNORM: + return BRW_SURFACEFORMAT_A16_UNORM; + */ + + case PIPE_FORMAT_A8L8_UNORM: + return BRW_SURFACEFORMAT_L8A8_UNORM; + + case PIPE_FORMAT_R5G6B5_UNORM: + return BRW_SURFACEFORMAT_B5G6R5_UNORM; + + case PIPE_FORMAT_A1R5G5B5_UNORM: + return BRW_SURFACEFORMAT_B5G5R5A1_UNORM; + + case PIPE_FORMAT_A4R4G4B4_UNORM: + return BRW_SURFACEFORMAT_B4G4R4A4_UNORM; + + case PIPE_FORMAT_X8R8G8B8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8X8_UNORM; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + + /* + * Video formats + */ + + case PIPE_FORMAT_YCBCR_REV: + return BRW_SURFACEFORMAT_YCRCB_NORMAL; + + case PIPE_FORMAT_YCBCR: + return BRW_SURFACEFORMAT_YCRCB_SWAPUVY; + + /* + * Compressed formats. + */ + /* XXX: Add FXT to gallium? + case PIPE_FORMAT_FXT1_RGBA: + return BRW_SURFACEFORMAT_FXT1; + */ + + case PIPE_FORMAT_DXT1_RGB: + return BRW_SURFACEFORMAT_DXT1_RGB; + + case PIPE_FORMAT_DXT1_RGBA: + return BRW_SURFACEFORMAT_BC1_UNORM; + + case PIPE_FORMAT_DXT3_RGBA: + return BRW_SURFACEFORMAT_BC2_UNORM; + + case PIPE_FORMAT_DXT5_RGBA: + return BRW_SURFACEFORMAT_BC3_UNORM; + + /* + * sRGB formats + */ + + case PIPE_FORMAT_R8G8B8A8_SRGB: + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB; + + case PIPE_FORMAT_A8L8_SRGB: + return BRW_SURFACEFORMAT_L8A8_UNORM_SRGB; + + case PIPE_FORMAT_L8_SRGB: + return BRW_SURFACEFORMAT_L8_UNORM_SRGB; + + case PIPE_FORMAT_DXT1_SRGB: + return BRW_SURFACEFORMAT_BC1_UNORM_SRGB; + + /* + * Depth formats + */ + + case PIPE_FORMAT_Z16_UNORM: + return BRW_SURFACEFORMAT_I16_UNORM; + + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + return BRW_SURFACEFORMAT_I24X8_UNORM; + + case PIPE_FORMAT_Z32_FLOAT: + return BRW_SURFACEFORMAT_I32_FLOAT; + + /* XXX: presumably for bump mapping. Add this to mesa state + * tracker? + * + * XXX: Add flipped versions of these formats to Gallium. + */ + case PIPE_FORMAT_R8G8_SNORM: + return BRW_SURFACEFORMAT_R8G8_SNORM; + + case PIPE_FORMAT_R8G8B8A8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_SNORM; + + default: + return BRW_SURFACEFORMAT_INVALID; + } +} + + + + + +static struct pipe_texture *brw_texture_create( struct pipe_screen *screen, + const struct pipe_texture *templ ) + +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_texture *tex; + enum brw_buffer_type buffer_type; + enum pipe_error ret; + + tex = CALLOC_STRUCT(brw_texture); + if (tex == NULL) + return NULL; + + memcpy(&tex->base, templ, sizeof *templ); + pipe_reference_init(&tex->base.reference, 1); + tex->base.screen = screen; + + /* XXX: compressed textures need special treatment here + */ + tex->cpp = util_format_get_blocksize(tex->base.format); + tex->compressed = util_format_is_compressed(tex->base.format); + + make_empty_list(&tex->views[0]); + make_empty_list(&tex->views[1]); + + /* XXX: No tiling with compressed textures?? + */ + if (tex->compressed == 0 && + !bscreen->no_tiling) + { + if (bscreen->chipset.is_965 && + util_format_is_depth_or_stencil(templ->format)) + tex->tiling = BRW_TILING_Y; + else + tex->tiling = BRW_TILING_X; + } + else { + tex->tiling = BRW_TILING_NONE; + } + + + + + if (!brw_texture_layout( bscreen, tex )) + goto fail; + + + if (templ->tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_PRIMARY)) { + buffer_type = BRW_BUFFER_TYPE_SCANOUT; + } + else { + buffer_type = BRW_BUFFER_TYPE_TEXTURE; + } + + ret = bscreen->sws->bo_alloc( bscreen->sws, + buffer_type, + tex->pitch * tex->total_height * tex->cpp, + 64, + &tex->bo ); + if (ret) + goto fail; + + tex->ss.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + tex->ss.ss0.surface_type = translate_tex_target(tex->base.target); + tex->ss.ss0.surface_format = translate_tex_format(tex->base.format); + assert(tex->ss.ss0.surface_format != BRW_SURFACEFORMAT_INVALID); + + /* This is ok for all textures with channel width 8bit or less: + */ +/* tex->ss.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ + + + /* XXX: what happens when tex->bo->offset changes??? + */ + tex->ss.ss1.base_addr = 0; /* reloc */ + tex->ss.ss2.mip_count = tex->base.last_level; + tex->ss.ss2.width = tex->base.width0 - 1; + tex->ss.ss2.height = tex->base.height0 - 1; + + switch (tex->tiling) { + case BRW_TILING_NONE: + tex->ss.ss3.tiled_surface = 0; + tex->ss.ss3.tile_walk = 0; + break; + case BRW_TILING_X: + tex->ss.ss3.tiled_surface = 1; + tex->ss.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + break; + case BRW_TILING_Y: + tex->ss.ss3.tiled_surface = 1; + tex->ss.ss3.tile_walk = BRW_TILEWALK_YMAJOR; + break; + } + + tex->ss.ss3.pitch = (tex->pitch * tex->cpp) - 1; + tex->ss.ss3.depth = tex->base.depth0 - 1; + + tex->ss.ss4.min_lod = 0; + + if (tex->base.target == PIPE_TEXTURE_CUBE) { + tex->ss.ss0.cube_pos_x = 1; + tex->ss.ss0.cube_pos_y = 1; + tex->ss.ss0.cube_pos_z = 1; + tex->ss.ss0.cube_neg_x = 1; + tex->ss.ss0.cube_neg_y = 1; + tex->ss.ss0.cube_neg_z = 1; + } + + return &tex->base; + +fail: + bo_reference(&tex->bo, NULL); + FREE(tex); + return NULL; +} + +static struct pipe_texture *brw_texture_blanket(struct pipe_screen *screen, + const struct pipe_texture *templ, + const unsigned *stride, + struct pipe_buffer *buffer) +{ + return NULL; +} + +static void brw_texture_destroy(struct pipe_texture *pt) +{ + struct brw_texture *tex = brw_texture(pt); + bo_reference(&tex->bo, NULL); + FREE(pt); +} + + +static boolean brw_is_format_supported( struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) +{ + return translate_tex_format(format) != BRW_SURFACEFORMAT_INVALID; +} + + +boolean brw_is_texture_referenced_by_bo( struct brw_screen *brw_screen, + struct pipe_texture *texture, + unsigned face, + unsigned level, + struct brw_winsys_buffer *bo ) +{ + struct brw_texture *tex = brw_texture(texture); + struct brw_surface *surf; + int i; + + /* XXX: this is subject to false positives if the underlying + * texture BO is referenced, we can't tell whether the sub-region + * we care about participates in that. + */ + if (brw_screen->sws->bo_references( bo, tex->bo )) + return TRUE; + + /* Find any view on this texture for this face/level and see if it + * is referenced: + */ + for (i = 0; i < 2; i++) { + foreach (surf, &tex->views[i]) { + if (surf->bo == tex->bo) + continue; + + if (surf->id.bits.face != face || + surf->id.bits.level != level) + continue; + + if (brw_screen->sws->bo_references( bo, surf->bo)) + return TRUE; + } + } + + return FALSE; +} + + +/* + * Transfer functions + */ + +static struct pipe_transfer* +brw_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, unsigned x, unsigned y, + unsigned w, unsigned h) +{ + struct brw_texture *tex = brw_texture(texture); + struct brw_transfer *trans; + unsigned offset; /* in bytes */ + + if (texture->target == PIPE_TEXTURE_CUBE) { + offset = tex->image_offset[level][face]; + } else if (texture->target == PIPE_TEXTURE_3D) { + offset = tex->image_offset[level][zslice]; + } else { + offset = tex->image_offset[level][0]; + assert(face == 0); + assert(zslice == 0); + } + + trans = CALLOC_STRUCT(brw_transfer); + if (trans) { + pipe_texture_reference(&trans->base.texture, texture); + trans->base.x = x; + trans->base.y = y; + trans->base.width = w; + trans->base.height = h; + trans->base.stride = tex->pitch * tex->cpp; + trans->offset = offset; + trans->base.usage = usage; + } + return &trans->base; +} + +static void * +brw_transfer_map(struct pipe_screen *screen, + struct pipe_transfer *transfer) +{ + struct brw_texture *tex = brw_texture(transfer->texture); + struct brw_winsys_screen *sws = brw_screen(screen)->sws; + char *map; + unsigned usage = transfer->usage; + + map = sws->bo_map(tex->bo, + BRW_DATA_OTHER, + 0, + tex->bo->size, + (usage & PIPE_TRANSFER_WRITE) ? TRUE : FALSE, + (usage & 0) ? TRUE : FALSE, + (usage & 0) ? TRUE : FALSE); + + if (!map) + return NULL; + + /* XXX: blocksize and compressed textures + */ + return map + brw_transfer(transfer)->offset + + transfer->y /* / transfer->block.height */ * transfer->stride + + transfer->x /* / transfer->block.width */ * brw_texture(transfer->texture)->cpp; +} + +static void +brw_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) +{ + struct brw_texture *tex = brw_texture(transfer->texture); + struct brw_winsys_screen *sws = brw_screen(screen)->sws; + + sws->bo_unmap(tex->bo); +} + +static void +brw_tex_transfer_destroy(struct pipe_transfer *trans) +{ + pipe_texture_reference(&trans->texture, NULL); + FREE(trans); +} + + +/* + * Functions exported to the winsys + */ + +boolean brw_texture_get_winsys_buffer(struct pipe_texture *texture, + struct brw_winsys_buffer **buffer, + unsigned *stride) +{ + struct brw_texture *tex = brw_texture(texture); + + *buffer = tex->bo; + if (stride) + *stride = tex->pitch * tex->cpp; + + return TRUE; +} + +struct pipe_texture * +brw_texture_blanket_winsys_buffer(struct pipe_screen *screen, + const struct pipe_texture *templ, + unsigned pitch, + unsigned tiling, + struct brw_winsys_buffer *buffer) +{ + struct brw_screen *bscreen = brw_screen(screen); + struct brw_texture *tex; + + if (templ->target != PIPE_TEXTURE_2D || + templ->last_level != 0 || + templ->depth0 != 1) + return NULL; + + if (util_format_is_compressed(templ->format)) + return NULL; + + tex = CALLOC_STRUCT(brw_texture); + if (!tex) + return NULL; + + memcpy(&tex->base, templ, sizeof *templ); + pipe_reference_init(&tex->base.reference, 1); + tex->base.screen = screen; + + /* XXX: cpp vs. blocksize + */ + tex->cpp = util_format_get_blocksize(tex->base.format); + tex->tiling = tiling; + + make_empty_list(&tex->views[0]); + make_empty_list(&tex->views[1]); + + if (!brw_texture_layout(bscreen, tex)) + goto fail; + + /* XXX Maybe some more checks? */ + if ((pitch / tex->cpp) < tex->pitch) + goto fail; + + tex->pitch = pitch / tex->cpp; + + tex->bo = buffer; + + /* fix this warning */ +#if 0 + if (tex->size > buffer->size) + goto fail; +#endif + + tex->ss.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + tex->ss.ss0.surface_type = translate_tex_target(tex->base.target); + tex->ss.ss0.surface_format = translate_tex_format(tex->base.format); + assert(tex->ss.ss0.surface_format != BRW_SURFACEFORMAT_INVALID); + + /* This is ok for all textures with channel width 8bit or less: + */ +/* tex->ss.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ + + + /* XXX: what happens when tex->bo->offset changes??? + */ + tex->ss.ss1.base_addr = 0; /* reloc */ + tex->ss.ss2.mip_count = tex->base.last_level; + tex->ss.ss2.width = tex->base.width0 - 1; + tex->ss.ss2.height = tex->base.height0 - 1; + + switch (tex->tiling) { + case BRW_TILING_NONE: + tex->ss.ss3.tiled_surface = 0; + tex->ss.ss3.tile_walk = 0; + break; + case BRW_TILING_X: + tex->ss.ss3.tiled_surface = 1; + tex->ss.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + break; + case BRW_TILING_Y: + tex->ss.ss3.tiled_surface = 1; + tex->ss.ss3.tile_walk = BRW_TILEWALK_YMAJOR; + break; + } + + tex->ss.ss3.pitch = (tex->pitch * tex->cpp) - 1; + tex->ss.ss3.depth = tex->base.depth0 - 1; + + tex->ss.ss4.min_lod = 0; + + return &tex->base; + +fail: + FREE(tex); + return NULL; +} + +void brw_screen_tex_init( struct brw_screen *brw_screen ) +{ + brw_screen->base.is_format_supported = brw_is_format_supported; + brw_screen->base.texture_create = brw_texture_create; + brw_screen->base.texture_destroy = brw_texture_destroy; + brw_screen->base.texture_blanket = brw_texture_blanket; + brw_screen->base.get_tex_transfer = brw_get_tex_transfer; + brw_screen->base.transfer_map = brw_transfer_map; + brw_screen->base.transfer_unmap = brw_transfer_unmap; + brw_screen->base.tex_transfer_destroy = brw_tex_transfer_destroy; +} diff --git a/src/gallium/drivers/i965/brw_sf.c b/src/gallium/drivers/i965/brw_sf.c new file mode 100644 index 00000000000..e1986a9dbbd --- /dev/null +++ b/src/gallium/drivers/i965/brw_sf.c @@ -0,0 +1,216 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "pipe/p_state.h" + +#include "brw_batchbuffer.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_pipe_rast.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" +#include "brw_state.h" + +static enum pipe_error compile_sf_prog( struct brw_context *brw, + struct brw_sf_prog_key *key, + struct brw_winsys_buffer **bo_out ) +{ + enum pipe_error ret; + struct brw_sf_compile c; + const GLuint *program; + GLuint program_size; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(brw, &c.func); + + c.key = *key; + c.nr_attrs = c.key.nr_attrs; + c.nr_attr_regs = (c.nr_attrs+1)/2; + c.nr_setup_attrs = c.key.nr_attrs; + c.nr_setup_regs = (c.nr_setup_attrs+1)/2; + + c.prog_data.urb_read_length = c.nr_attr_regs; + c.prog_data.urb_entry_size = c.nr_setup_regs * 2; + + /* Special case when there are no attributes to setup. + * + * XXX: should be able to set nr_setup_attrs to nr_attrs-1 -- but + * breaks vp-tris.c + */ + if (c.nr_attrs - 1 == 0) { + c.nr_verts = 0; + brw_emit_null_setup( &c ); + } + else { + /* Which primitive? Or all three? + */ + switch (key->primitive) { + case SF_TRIANGLES: + c.nr_verts = 3; + brw_emit_tri_setup( &c, GL_TRUE ); + break; + case SF_LINES: + c.nr_verts = 2; + brw_emit_line_setup( &c, GL_TRUE ); + break; + case SF_POINTS: + c.nr_verts = 1; + if (key->do_point_sprite) + brw_emit_point_sprite_setup( &c, GL_TRUE ); + else + brw_emit_point_setup( &c, GL_TRUE ); + break; + case SF_UNFILLED_TRIS: + c.nr_verts = 3; + brw_emit_anyprim_setup( &c ); + break; + default: + assert(0); + return PIPE_ERROR_BAD_INPUT; + } + } + + /* get the program + */ + ret = brw_get_program(&c.func, &program, &program_size); + if (ret) + return ret; + + /* Upload + */ + ret = brw_upload_cache( &brw->cache, BRW_SF_PROG, + &c.key, sizeof(c.key), + NULL, 0, + program, program_size, + &c.prog_data, + &brw->sf.prog_data, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + +/* Calculate interpolants for triangle and line rasterization. + */ +static enum pipe_error upload_sf_prog(struct brw_context *brw) +{ + const struct brw_fs_signature *sig = &brw->curr.fragment_shader->signature; + struct brw_sf_prog_key key; + enum pipe_error ret; + unsigned i; + + memset(&key, 0, sizeof(key)); + + /* Populate the key, noting state dependencies: + */ + + /* XXX: Add one to account for the position input. + */ + /* PIPE_NEW_FRAGMENT_SIGNATURE */ + key.nr_attrs = sig->nr_inputs + 1; + + + /* XXX: why is position required to be linear? why do we care + * about it at all? + */ + key.linear_attrs = 1; /* position -- but why? */ + + for (i = 0; i < sig->nr_inputs; i++) { + switch (sig->input[i].interp) { + case TGSI_INTERPOLATE_CONSTANT: + break; + case TGSI_INTERPOLATE_LINEAR: + key.linear_attrs |= 1 << (i+1); + break; + case TGSI_INTERPOLATE_PERSPECTIVE: + key.persp_attrs |= 1 << (i+1); + break; + } + } + + /* BRW_NEW_REDUCED_PRIMITIVE */ + switch (brw->reduced_primitive) { + case PIPE_PRIM_TRIANGLES: + /* PIPE_NEW_RAST + */ + if (brw->curr.rast->templ.fill_cw != PIPE_POLYGON_MODE_FILL || + brw->curr.rast->templ.fill_ccw != PIPE_POLYGON_MODE_FILL) + key.primitive = SF_UNFILLED_TRIS; + else + key.primitive = SF_TRIANGLES; + break; + case PIPE_PRIM_LINES: + key.primitive = SF_LINES; + break; + case PIPE_PRIM_POINTS: + key.primitive = SF_POINTS; + break; + } + + key.do_point_sprite = brw->curr.rast->templ.point_sprite; + key.sprite_origin_lower_left = 0; /* XXX: ctx->Point.SpriteOrigin - fix rast state */ + key.do_flat_shading = brw->curr.rast->templ.flatshade; + key.do_twoside_color = brw->curr.rast->templ.light_twoside; + + if (key.do_twoside_color) { + key.frontface_ccw = (brw->curr.rast->templ.front_winding == + PIPE_WINDING_CCW); + } + + if (brw_search_cache(&brw->cache, BRW_SF_PROG, + &key, sizeof(key), + NULL, 0, + &brw->sf.prog_data, + &brw->sf.prog_bo)) + return PIPE_OK; + + ret = compile_sf_prog( brw, &key, &brw->sf.prog_bo ); + if (ret) + return ret; + + return PIPE_OK; +} + + +const struct brw_tracked_state brw_sf_prog = { + .dirty = { + .mesa = (PIPE_NEW_RAST | PIPE_NEW_FRAGMENT_SIGNATURE), + .brw = (BRW_NEW_REDUCED_PRIMITIVE), + .cache = 0 + }, + .prepare = upload_sf_prog +}; + diff --git a/src/gallium/drivers/i965/brw_sf.h b/src/gallium/drivers/i965/brw_sf.h new file mode 100644 index 00000000000..a895c7d2f6a --- /dev/null +++ b/src/gallium/drivers/i965/brw_sf.h @@ -0,0 +1,122 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_SF_H +#define BRW_SF_H + + +#include "brw_context.h" +#include "brw_eu.h" + + +#define SF_POINTS 0 +#define SF_LINES 1 +#define SF_TRIANGLES 2 +#define SF_UNFILLED_TRIS 3 + +struct brw_sf_prog_key { + + /* Bitmask of linear and perspective interpolated inputs, 0..nr + */ + GLuint persp_attrs:32; + GLuint linear_attrs:32; + GLuint point_coord_replace_attrs:32; + + GLuint nr_attrs:8; + GLuint primitive:2; + GLuint do_twoside_color:1; + GLuint do_flat_shading:1; + GLuint frontface_ccw:1; + GLuint do_point_sprite:1; + GLuint sprite_origin_lower_left:1; + GLuint pad:17; + + GLuint attr_col0:8; + GLuint attr_col1:8; + GLuint attr_bfc0:8; + GLuint attr_bfc1:8; +}; + +struct brw_sf_point_tex { + GLboolean CoordReplace; +}; + +struct brw_sf_compile { + struct brw_compile func; + struct brw_sf_prog_key key; + struct brw_sf_prog_data prog_data; + + struct brw_reg pv; + struct brw_reg det; + struct brw_reg dx0; + struct brw_reg dx2; + struct brw_reg dy0; + struct brw_reg dy2; + + /* z and 1/w passed in seperately: + */ + struct brw_reg z[3]; + struct brw_reg inv_w[3]; + + /* The vertices: + */ + struct brw_reg vert[3]; + + /* Temporaries, allocated after last vertex reg. + */ + struct brw_reg inv_det; + struct brw_reg a1_sub_a0; + struct brw_reg a2_sub_a0; + struct brw_reg tmp; + + struct brw_reg m1Cx; + struct brw_reg m2Cy; + struct brw_reg m3C0; + + GLuint nr_verts; + GLuint nr_attrs; + GLuint nr_attr_regs; + GLuint nr_setup_attrs; + GLuint nr_setup_regs; + + GLuint point_coord_replace_mask; +}; + + +void brw_emit_null_setup( struct brw_sf_compile *c ); +void brw_emit_tri_setup( struct brw_sf_compile *c, GLboolean allocate ); +void brw_emit_line_setup( struct brw_sf_compile *c, GLboolean allocate ); +void brw_emit_point_setup( struct brw_sf_compile *c, GLboolean allocate ); +void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate ); +void brw_emit_anyprim_setup( struct brw_sf_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965/brw_sf_emit.c b/src/gallium/drivers/i965/brw_sf_emit.c new file mode 100644 index 00000000000..3b85725e368 --- /dev/null +++ b/src/gallium/drivers/i965/brw_sf_emit.c @@ -0,0 +1,765 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_batchbuffer.h" + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" + + +static struct brw_reg get_vert_attr(struct brw_sf_compile *c, + struct brw_reg vert, + GLuint attr) +{ + GLuint off = attr / 2; + GLuint sub = attr % 2; + + return brw_vec4_grf(vert.nr + off, sub * 4); +} + + +/*********************************************************************** + * Twoside lighting + */ +static void copy_bfc( struct brw_sf_compile *c, + struct brw_reg vert ) +{ + struct brw_compile *p = &c->func; + + if (c->key.attr_col0 && c->key.attr_bfc0) + brw_MOV(p, + get_vert_attr(c, vert, c->key.attr_col0), + get_vert_attr(c, vert, c->key.attr_bfc0)); + + if (c->key.attr_col1 && c->key.attr_bfc1) + brw_MOV(p, + get_vert_attr(c, vert, c->key.attr_col1), + get_vert_attr(c, vert, c->key.attr_bfc1)); +} + + +static void do_twoside_color( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + GLuint backface_conditional = c->key.frontface_ccw ? BRW_CONDITIONAL_G : BRW_CONDITIONAL_L; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + /* XXX: What happens if BFC isn't present? This could only happen + * for user-supplied vertex programs, as t_vp_build.c always does + * the right thing. + */ + if (!(c->key.attr_col0 && c->key.attr_bfc0) && + !(c->key.attr_col1 && c->key.attr_bfc1)) + return; + + /* Need to use BRW_EXECUTE_4 and also do an 4-wide compare in order + * to get all channels active inside the IF. In the clipping code + * we run with NoMask, so it's not an option and we can use + * BRW_EXECUTE_1 for all comparisions. + */ + brw_push_insn_state(p); + brw_CMP(p, vec4(brw_null_reg()), backface_conditional, c->det, brw_imm_f(0)); + if_insn = brw_IF(p, BRW_EXECUTE_4); + { + switch (c->nr_verts) { + case 3: copy_bfc(c, c->vert[2]); + case 2: copy_bfc(c, c->vert[1]); + case 1: copy_bfc(c, c->vert[0]); + } + } + brw_ENDIF(p, if_insn); + brw_pop_insn_state(p); +} + + + +/*********************************************************************** + * Flat shading + */ + +#define VERT_RESULT_COLOR_BITS ((1<<VERT_RESULT_COL0) | \ + (1<<VERT_RESULT_COL1)) + +static void copy_colors( struct brw_sf_compile *c, + struct brw_reg dst, + struct brw_reg src) +{ + struct brw_compile *p = &c->func; + + if (c->key.attr_col0) + brw_MOV(p, + get_vert_attr(c, dst, c->key.attr_col0), + get_vert_attr(c, src, c->key.attr_col0)); + + if (c->key.attr_col1) + brw_MOV(p, + get_vert_attr(c, dst, c->key.attr_col1), + get_vert_attr(c, src, c->key.attr_col1)); + +} + + + +/* Need to use a computed jump to copy flatshaded attributes as the + * vertices are ordered according to y-coordinate before reaching this + * point, so the PV could be anywhere. + */ +static void do_flatshade_triangle( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + GLuint jmpi = 1; + GLuint nr = 0; + + if (c->key.attr_col0) + nr++; + + if (c->key.attr_col1) + nr++; + + if (nr == 0) + return; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + if (BRW_IS_IGDNG(p->brw)) + jmpi = 2; + + brw_push_insn_state(p); + + brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr*2+1))); + brw_JMPI(p, ip, ip, c->pv); + + copy_colors(c, c->vert[1], c->vert[0]); + copy_colors(c, c->vert[2], c->vert[0]); + brw_JMPI(p, ip, ip, brw_imm_d(jmpi*(nr*4+1))); + + copy_colors(c, c->vert[0], c->vert[1]); + copy_colors(c, c->vert[2], c->vert[1]); + brw_JMPI(p, ip, ip, brw_imm_d(jmpi*nr*2)); + + copy_colors(c, c->vert[0], c->vert[2]); + copy_colors(c, c->vert[1], c->vert[2]); + + brw_pop_insn_state(p); +} + + +static void do_flatshade_line( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + GLuint jmpi = 1; + GLuint nr = 0; + + if (c->key.attr_col0) + nr++; + + if (c->key.attr_col1) + nr++; + + if (nr == 0) + return; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + if (BRW_IS_IGDNG(p->brw)) + jmpi = 2; + + brw_push_insn_state(p); + + brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr+1))); + brw_JMPI(p, ip, ip, c->pv); + copy_colors(c, c->vert[1], c->vert[0]); + + brw_JMPI(p, ip, ip, brw_imm_ud(jmpi*nr)); + copy_colors(c, c->vert[0], c->vert[1]); + + brw_pop_insn_state(p); +} + + + +/*********************************************************************** + * Triangle setup. + */ + + +static void alloc_regs( struct brw_sf_compile *c ) +{ + GLuint reg, i; + + /* Values computed by fixed function unit: + */ + c->pv = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_D); + c->det = brw_vec1_grf(1, 2); + c->dx0 = brw_vec1_grf(1, 3); + c->dx2 = brw_vec1_grf(1, 4); + c->dy0 = brw_vec1_grf(1, 5); + c->dy2 = brw_vec1_grf(1, 6); + + /* z and 1/w passed in seperately: + */ + c->z[0] = brw_vec1_grf(2, 0); + c->inv_w[0] = brw_vec1_grf(2, 1); + c->z[1] = brw_vec1_grf(2, 2); + c->inv_w[1] = brw_vec1_grf(2, 3); + c->z[2] = brw_vec1_grf(2, 4); + c->inv_w[2] = brw_vec1_grf(2, 5); + + /* The vertices: + */ + reg = 3; + for (i = 0; i < c->nr_verts; i++) { + c->vert[i] = brw_vec8_grf(reg, 0); + reg += c->nr_attr_regs; + } + + /* Temporaries, allocated after last vertex reg. + */ + c->inv_det = brw_vec1_grf(reg, 0); reg++; + c->a1_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->a2_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->tmp = brw_vec8_grf(reg, 0); reg++; + + /* Note grf allocation: + */ + c->prog_data.total_grf = reg; + + + /* Outputs of this program - interpolation coefficients for + * rasterization: + */ + c->m1Cx = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 1, 0); + c->m2Cy = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 2, 0); + c->m3C0 = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 3, 0); +} + + +static void copy_z_inv_w( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + GLuint i; + + brw_push_insn_state(p); + + /* Copy both scalars with a single MOV: + */ + for (i = 0; i < c->nr_verts; i++) + brw_MOV(p, vec2(suboffset(c->vert[i], 2)), vec2(c->z[i])); + + brw_pop_insn_state(p); +} + + +static void invert_det( struct brw_sf_compile *c) +{ + /* Looks like we invert all 8 elements just to get 1/det in + * position 2 !?! + */ + brw_math(&c->func, + c->inv_det, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + c->det, + BRW_MATH_DATA_SCALAR, + BRW_MATH_PRECISION_FULL); + +} + + +/* Two attributes packed into a wide register. Figure out if either + * or both of them need linear/perspective interpolation. Constant + * regs are left as-is. + */ +static GLboolean calculate_masks( struct brw_sf_compile *c, + GLuint reg, + GLushort *pc, + GLushort *pc_persp, + GLushort *pc_linear) +{ + GLboolean is_last_attr = (reg == c->nr_setup_regs - 1); + GLuint persp_mask = c->key.persp_attrs; + GLuint linear_mask = (c->key.persp_attrs | c->key.linear_attrs); + + *pc_persp = 0; + *pc_linear = 0; + *pc = 0xf; + + if (persp_mask & (1 << (reg*2))) + *pc_persp = 0xf; + + if (linear_mask & (1 << (reg*2))) + *pc_linear = 0xf; + + /* Maybe only processs one attribute on the final round: + */ + if (reg*2+1 < c->nr_setup_attrs) { + *pc |= 0xf0; + + if (persp_mask & (1 << (reg*2+1))) + *pc_persp |= 0xf0; + + if (linear_mask & (1 << (reg*2+1))) + *pc_linear |= 0xf0; + } + + return is_last_attr; +} + + +void brw_emit_null_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + + /* m0 is implicitly copied from r0 in the send instruction: + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), /* r0, will be copied to m0 */ + 0, /* allocate */ + 1, /* used */ + 1, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, /* offset */ + BRW_URB_SWIZZLE_TRANSPOSE); +} + +void brw_emit_tri_setup( struct brw_sf_compile *c, GLboolean allocate) +{ + struct brw_compile *p = &c->func; + GLuint i; + + c->nr_verts = 3; + + if (allocate) + alloc_regs(c); + + invert_det(c); + copy_z_inv_w(c); + + if (c->key.do_twoside_color) + do_twoside_color(c); + + if (c->key.do_flat_shading) + do_flatshade_triangle(c); + + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + struct brw_reg a2 = offset(c->vert[2], i); + GLushort pc, pc_persp, pc_linear; + GLboolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + brw_MUL(p, a2, a2, c->inv_w[2]); + } + + + /* Calculate coefficients for interpolated values: + */ + if (pc_linear) + { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + brw_ADD(p, c->a2_sub_a0, a2, negate(a0)); + + /* calculate dA/dx + */ + brw_MUL(p, brw_null_reg(), c->a1_sub_a0, c->dy2); + brw_MAC(p, c->tmp, c->a2_sub_a0, negate(c->dy0)); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + /* calculate dA/dy + */ + brw_MUL(p, brw_null_reg(), c->a2_sub_a0, c->dx0); + brw_MAC(p, c->tmp, c->a1_sub_a0, negate(c->dx2)); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. m0 is implicitly copied from r0 in + * the send instruction: + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), /* r0, will be copied to m0 */ + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* offset */ + BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */ + } + } +} + + + +void brw_emit_line_setup( struct brw_sf_compile *c, GLboolean allocate) +{ + struct brw_compile *p = &c->func; + GLuint i; + + + c->nr_verts = 2; + + if (allocate) + alloc_regs(c); + + invert_det(c); + copy_z_inv_w(c); + + if (c->key.do_flat_shading) + do_flatshade_line(c); + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + GLushort pc, pc_persp, pc_linear; + GLboolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + } + + /* Calculate coefficients for position, color: + */ + if (pc_linear) { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dx0); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dy0); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate) +{ + struct brw_compile *p = &c->func; + GLuint i; + + c->nr_verts = 1; + + if (allocate) + alloc_regs(c); + + copy_z_inv_w(c); + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* XXX: only seems to check point_coord_replace_attrs for every + * second attribute?!? + */ + boolean coord_replace = !!(c->key.point_coord_replace_attrs & (1<<(2*i))); + struct brw_reg a0 = offset(c->vert[0], i); + GLushort pc, pc_persp, pc_linear; + GLboolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + if (coord_replace) { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + } + } + + if (coord_replace) { + /* Caculate 1.0/PointWidth */ + brw_math(&c->func, + c->tmp, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + c->dx0, + BRW_MATH_DATA_SCALAR, + BRW_MATH_PRECISION_FULL); + + if (c->key.sprite_origin_lower_left) { + brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); + brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0])); + brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); + } + else { + brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); + } + } + else { + brw_MOV(p, c->m1Cx, brw_imm_ud(0)); + brw_MOV(p, c->m2Cy, brw_imm_ud(0)); + } + + { + brw_set_predicate_control_flag_value(p, pc); + if (coord_replace) { + if (c->key.sprite_origin_lower_left) { + brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0)); + brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0)); + } + else { + brw_MOV(p, c->m3C0, brw_imm_f(0.0)); + } + } + else { + brw_MOV(p, c->m3C0, a0); /* constant value */ + } + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +/* Points setup - several simplifications as all attributes are + * constant across the face of the point (point sprites excluded!) + */ +void brw_emit_point_setup( struct brw_sf_compile *c, GLboolean allocate) +{ + struct brw_compile *p = &c->func; + GLuint i; + + c->nr_verts = 1; + + if (allocate) + alloc_regs(c); + + copy_z_inv_w(c); + + brw_MOV(p, c->m1Cx, brw_imm_ud(0)); /* zero - move out of loop */ + brw_MOV(p, c->m2Cy, brw_imm_ud(0)); /* zero - move out of loop */ + + for (i = 0; i < c->nr_setup_regs; i++) + { + struct brw_reg a0 = offset(c->vert[0], i); + GLushort pc, pc_persp, pc_linear; + GLboolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + /* This seems odd as the values are all constant, but the + * fragment shader will be expecting it: + */ + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + } + + + /* The delta values are always zero, just send the starting + * coordinate. Again, this is to fit in with the interpolation + * code in the fragment shader. + */ + { + brw_set_predicate_control_flag_value(p, pc); + + brw_MOV(p, c->m3C0, a0); /* constant value */ + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +void brw_emit_anyprim_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0); + struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0); + struct brw_reg primmask; + struct brw_instruction *jmp; + struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); + + GLuint saveflag; + + c->nr_verts = 3; + alloc_regs(c); + + primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD); + + brw_MOV(p, primmask, brw_imm_ud(1)); + brw_SHL(p, primmask, primmask, payload_prim); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) | + (1<<_3DPRIM_TRISTRIP) | + (1<<_3DPRIM_TRIFAN) | + (1<<_3DPRIM_TRISTRIP_REVERSE) | + (1<<_3DPRIM_POLYGON) | + (1<<_3DPRIM_RECTLIST) | + (1<<_3DPRIM_TRIFAN_NOSTIPPLE))); + jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); + { + saveflag = p->flag_value; + brw_push_insn_state(p); + brw_emit_tri_setup( c, GL_FALSE ); + brw_pop_insn_state(p); + p->flag_value = saveflag; + /* note - thread killed in subroutine, so must + * restore the flag which is changed when building + * the subroutine. fix #13240 + */ + } + brw_land_fwd_jump(p, jmp); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) | + (1<<_3DPRIM_LINESTRIP) | + (1<<_3DPRIM_LINELOOP) | + (1<<_3DPRIM_LINESTRIP_CONT) | + (1<<_3DPRIM_LINESTRIP_BF) | + (1<<_3DPRIM_LINESTRIP_CONT_BF))); + jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); + { + saveflag = p->flag_value; + brw_push_insn_state(p); + brw_emit_line_setup( c, GL_FALSE ); + brw_pop_insn_state(p); + p->flag_value = saveflag; + /* note - thread killed in subroutine */ + } + brw_land_fwd_jump(p, jmp); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE)); + jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); + { + saveflag = p->flag_value; + brw_push_insn_state(p); + brw_emit_point_sprite_setup( c, GL_FALSE ); + brw_pop_insn_state(p); + p->flag_value = saveflag; + } + brw_land_fwd_jump(p, jmp); + + brw_emit_point_setup( c, GL_FALSE ); +} + + + + diff --git a/src/gallium/drivers/i965/brw_sf_state.c b/src/gallium/drivers/i965/brw_sf_state.c new file mode 100644 index 00000000000..25dc2b52e07 --- /dev/null +++ b/src/gallium/drivers/i965/brw_sf_state.c @@ -0,0 +1,333 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + +#include "pipe/p_state.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_debug.h" +#include "brw_pipe_rast.h" + +static enum pipe_error upload_sf_vp(struct brw_context *brw) +{ + const struct pipe_viewport_state *vp = &brw->curr.viewport; + const struct pipe_scissor_state *scissor = &brw->curr.scissor; + struct brw_sf_viewport sfv; + enum pipe_error ret; + + memset(&sfv, 0, sizeof(sfv)); + + /* PIPE_NEW_VIEWPORT, PIPE_NEW_SCISSOR */ + + sfv.viewport.m00 = vp->scale[0]; + sfv.viewport.m11 = vp->scale[1]; + sfv.viewport.m22 = vp->scale[2]; + sfv.viewport.m30 = vp->translate[0]; + sfv.viewport.m31 = vp->translate[1]; + sfv.viewport.m32 = vp->translate[2]; + + sfv.scissor.xmin = scissor->minx; + sfv.scissor.xmax = scissor->maxx - 1; /* ? */ + sfv.scissor.ymin = scissor->miny; + sfv.scissor.ymax = scissor->maxy - 1; /* ? */ + + ret = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0, + &brw->sf.vp_bo ); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_sf_vp = { + .dirty = { + .mesa = (PIPE_NEW_VIEWPORT | + PIPE_NEW_SCISSOR), + .brw = 0, + .cache = 0 + }, + .prepare = upload_sf_vp +}; + +struct brw_sf_unit_key { + unsigned int total_grf; + unsigned int urb_entry_read_length; + unsigned int nr_urb_entries, urb_size, sfsize; + + unsigned scissor:1; + unsigned line_smooth:1; + unsigned point_sprite:1; + unsigned point_attenuated:1; + unsigned front_face:2; + unsigned cull_mode:2; + unsigned flatshade_first:1; + unsigned gl_rasterization_rules:1; + unsigned line_last_pixel_enable:1; + float line_width; + float point_size; +}; + +static void +sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key) +{ + const struct pipe_rasterizer_state *rast = &brw->curr.rast->templ; + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_SF_PROG */ + key->total_grf = brw->sf.prog_data->total_grf; + key->urb_entry_read_length = brw->sf.prog_data->urb_read_length; + + /* BRW_NEW_URB_FENCE */ + key->nr_urb_entries = brw->urb.nr_sf_entries; + key->urb_size = brw->urb.vsize; + key->sfsize = brw->urb.sfsize; + + /* PIPE_NEW_RAST */ + key->scissor = rast->scissor; + key->front_face = rast->front_winding; + key->cull_mode = rast->cull_mode; + key->line_smooth = rast->line_smooth; + key->line_width = rast->line_width; + key->flatshade_first = rast->flatshade_first; + key->line_last_pixel_enable = rast->line_last_pixel; + key->gl_rasterization_rules = rast->gl_rasterization_rules; + + key->point_sprite = rast->point_sprite; + key->point_attenuated = rast->point_size_per_vertex; + + key->point_size = CLAMP(rast->point_size, + rast->point_size_min, + rast->point_size_max); +} + +static enum pipe_error +sf_unit_create_from_key(struct brw_context *brw, + struct brw_sf_unit_key *key, + struct brw_winsys_reloc *reloc, + struct brw_winsys_buffer **bo_out) +{ + struct brw_sf_unit_state sf; + enum pipe_error ret; + int chipset_max_threads; + memset(&sf, 0, sizeof(sf)); + + sf.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1; + /* reloc */ + sf.thread0.kernel_start_pointer = 0; + + sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + + sf.thread3.dispatch_grf_start_reg = 3; + + if (BRW_IS_IGDNG(brw)) + sf.thread3.urb_entry_read_offset = 3; + else + sf.thread3.urb_entry_read_offset = 1; + + sf.thread3.urb_entry_read_length = key->urb_entry_read_length; + + sf.thread4.nr_urb_entries = key->nr_urb_entries; + sf.thread4.urb_entry_allocation_size = key->sfsize - 1; + + /* Each SF thread produces 1 PUE, and there can be up to 24(Pre-IGDNG) or + * 48(IGDNG) threads + */ + if (BRW_IS_IGDNG(brw)) + chipset_max_threads = 48; + else + chipset_max_threads = 24; + + sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + sf.thread4.max_threads = 0; + + if (BRW_DEBUG & DEBUG_STATS) + sf.thread4.stats_enable = 1; + + /* CACHE_NEW_SF_VP */ + /* reloc */ + sf.sf5.sf_viewport_state_offset = 0; + + sf.sf5.viewport_transform = 1; + + if (key->scissor) + sf.sf6.scissor = 1; + + if (key->front_face == PIPE_WINDING_CCW) + sf.sf5.front_winding = BRW_FRONTWINDING_CCW; + else + sf.sf5.front_winding = BRW_FRONTWINDING_CW; + + switch (key->cull_mode) { + case PIPE_WINDING_CCW: + case PIPE_WINDING_CW: + sf.sf6.cull_mode = (key->front_face == key->cull_mode ? + BRW_CULLMODE_FRONT : + BRW_CULLMODE_BACK); + break; + case PIPE_WINDING_BOTH: + sf.sf6.cull_mode = BRW_CULLMODE_BOTH; + break; + case PIPE_WINDING_NONE: + sf.sf6.cull_mode = BRW_CULLMODE_NONE; + break; + default: + assert(0); + sf.sf6.cull_mode = BRW_CULLMODE_NONE; + break; + } + + /* _NEW_LINE */ + /* XXX use ctx->Const.Min/MaxLineWidth here */ + sf.sf6.line_width = CLAMP(key->line_width, 1.0, 5.0) * (1<<1); + + sf.sf6.line_endcap_aa_region_width = 1; + if (key->line_smooth) + sf.sf6.aa_enable = 1; + else if (sf.sf6.line_width <= 0x2) + sf.sf6.line_width = 0; + + /* XXX: gl_rasterization_rules? something else? + */ + sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT; + sf.sf6.point_rast_rule = BRW_RASTRULE_LOWER_RIGHT; + sf.sf6.point_rast_rule = 1; + + /* XXX clamp max depends on AA vs. non-AA */ + + /* _NEW_POINT */ + sf.sf7.sprite_point = key->point_sprite; + sf.sf7.point_size = CLAMP(rint(key->point_size), 1, 255) * (1<<3); + sf.sf7.use_point_size_state = !key->point_attenuated; + sf.sf7.aa_line_distance_mode = 0; + + /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: + */ + if (!key->flatshade_first) { + sf.sf7.trifan_pv = 2; + sf.sf7.linestrip_pv = 1; + sf.sf7.tristrip_pv = 2; + } else { + sf.sf7.trifan_pv = 1; + sf.sf7.linestrip_pv = 0; + sf.sf7.tristrip_pv = 0; + } + + sf.sf7.line_last_pixel_enable = key->line_last_pixel_enable; + + /* Set bias for OpenGL rasterization rules: + */ + if (key->gl_rasterization_rules) { + sf.sf6.dest_org_vbias = 0x8; + sf.sf6.dest_org_hbias = 0x8; + } + else { + sf.sf6.dest_org_vbias = 0x0; + sf.sf6.dest_org_hbias = 0x0; + } + + ret = brw_upload_cache(&brw->cache, BRW_SF_UNIT, + key, sizeof(*key), + reloc, 2, + &sf, sizeof(sf), + NULL, NULL, + bo_out); + if (ret) + return ret; + + + return PIPE_OK; +} + +static enum pipe_error upload_sf_unit( struct brw_context *brw ) +{ + struct brw_sf_unit_key key; + struct brw_winsys_reloc reloc[2]; + unsigned total_grf; + unsigned viewport_transform; + unsigned front_winding; + enum pipe_error ret; + + sf_unit_populate_key(brw, &key); + + /* XXX: cut this crap and pre calculate the key: + */ + total_grf = (align(key.total_grf, 16) / 16 - 1); + viewport_transform = 1; + front_winding = (key.front_face == PIPE_WINDING_CCW ? + BRW_FRONTWINDING_CCW : + BRW_FRONTWINDING_CW); + + /* Emit SF program relocation */ + make_reloc(&reloc[0], + BRW_USAGE_STATE, + total_grf << 1, + offsetof(struct brw_sf_unit_state, thread0), + brw->sf.prog_bo); + + /* Emit SF viewport relocation */ + make_reloc(&reloc[1], + BRW_USAGE_STATE, + front_winding | (viewport_transform << 1), + offsetof(struct brw_sf_unit_state, sf5), + brw->sf.vp_bo); + + + if (brw_search_cache(&brw->cache, BRW_SF_UNIT, + &key, sizeof(key), + reloc, 2, + NULL, + &brw->sf.state_bo)) + return PIPE_OK; + + + ret = sf_unit_create_from_key(brw, &key, + reloc, + &brw->sf.state_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_sf_unit = { + .dirty = { + .mesa = (PIPE_NEW_RAST), + .brw = BRW_NEW_URB_FENCE, + .cache = (CACHE_NEW_SF_VP | + CACHE_NEW_SF_PROG) + }, + .prepare = upload_sf_unit, +}; diff --git a/src/gallium/drivers/i965/brw_state.h b/src/gallium/drivers/i965/brw_state.h new file mode 100644 index 00000000000..d2bbd0123d1 --- /dev/null +++ b/src/gallium/drivers/i965/brw_state.h @@ -0,0 +1,174 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_STATE_H +#define BRW_STATE_H + +#include "pipe/p_defines.h" +#include "util/u_memory.h" + +#include "brw_context.h" + +static INLINE void +brw_add_validated_bo(struct brw_context *brw, struct brw_winsys_buffer *bo) +{ + assert(brw->state.validated_bo_count < Elements(brw->state.validated_bos)); + + if (bo != NULL) { + bo_reference( &brw->state.validated_bos[brw->state.validated_bo_count++], + bo ); + } +} + +const struct brw_tracked_state brw_blend_constant_color; +const struct brw_tracked_state brw_cc_unit; +const struct brw_tracked_state brw_cc_vp; +const struct brw_tracked_state brw_clip_prog; +const struct brw_tracked_state brw_clip_unit; +const struct brw_tracked_state brw_curbe_buffer; +const struct brw_tracked_state brw_curbe_offsets; +const struct brw_tracked_state brw_invarient_state; +const struct brw_tracked_state brw_gs_prog; +const struct brw_tracked_state brw_gs_unit; +const struct brw_tracked_state brw_line_stipple; +const struct brw_tracked_state brw_aa_line_parameters; +const struct brw_tracked_state brw_pipelined_state_pointers; +const struct brw_tracked_state brw_binding_table_pointers; +const struct brw_tracked_state brw_depthbuffer; +const struct brw_tracked_state brw_polygon_stipple; +const struct brw_tracked_state brw_program_parameters; +const struct brw_tracked_state brw_recalculate_urb_fence; +const struct brw_tracked_state brw_sf_prog; +const struct brw_tracked_state brw_sf_unit; +const struct brw_tracked_state brw_sf_vp; +const struct brw_tracked_state brw_state_base_address; +const struct brw_tracked_state brw_urb_fence; +const struct brw_tracked_state brw_vertex_state; +const struct brw_tracked_state brw_vs_surfaces; +const struct brw_tracked_state brw_vs_prog; +const struct brw_tracked_state brw_vs_unit; +const struct brw_tracked_state brw_wm_input_sizes; +const struct brw_tracked_state brw_wm_prog; +const struct brw_tracked_state brw_wm_samplers; +const struct brw_tracked_state brw_wm_constant_surface; +const struct brw_tracked_state brw_wm_surfaces; +const struct brw_tracked_state brw_wm_unit; + +const struct brw_tracked_state brw_psp_urb_cbs; + +const struct brw_tracked_state brw_pipe_control; + +const struct brw_tracked_state brw_drawing_rect; +const struct brw_tracked_state brw_indices; +const struct brw_tracked_state brw_vertices; +const struct brw_tracked_state brw_index_buffer; + + +/*********************************************************************** + * brw_state.c + */ +int brw_validate_state(struct brw_context *brw); +int brw_upload_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + +/*********************************************************************** + * brw_state_cache.c + */ +enum pipe_error brw_cache_data(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *data, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + struct brw_winsys_buffer **bo_out ); + +enum pipe_error brw_cache_data_sz(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *data, + GLuint data_size, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + struct brw_winsys_buffer **bo_out); + +enum pipe_error brw_upload_cache( struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *key, + GLuint key_sz, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + const void *data, + GLuint data_sz, + const void *aux, + void *aux_return , + struct brw_winsys_buffer **bo_out); + +boolean brw_search_cache( struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *key, + GLuint key_size, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + void *aux_return, + struct brw_winsys_buffer **bo_out); + +void brw_state_cache_check_size( struct brw_context *brw ); + +void brw_init_caches( struct brw_context *brw ); +void brw_destroy_caches( struct brw_context *brw ); +void brw_state_cache_bo_delete(struct brw_cache *cache, struct brw_winsys_buffer *bo); + +/*********************************************************************** + * brw_state_batch.c + */ +#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->batch, (s), sizeof(*(s)), IGNORE_CLIPRECTS) +#define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) ) + +GLboolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + GLuint sz ); +void brw_destroy_batch_cache( struct brw_context *brw ); +void brw_clear_batch_cache( struct brw_context *brw ); + +/*********************************************************************** + * brw_wm_surface_state.c + */ + +/*********************************************************************** + * brw_state_debug.c + */ +void brw_update_dirty_counts( unsigned mesa, + unsigned brw, + unsigned cache ); + + + +#endif diff --git a/src/gallium/drivers/i965/brw_state_batch.c b/src/gallium/drivers/i965/brw_state_batch.c new file mode 100644 index 00000000000..7d212e5c247 --- /dev/null +++ b/src/gallium/drivers/i965/brw_state_batch.c @@ -0,0 +1,98 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + + +#include "brw_state.h" +#include "brw_batchbuffer.h" + + + +/* A facility similar to the data caching code above, which aims to + * prevent identical commands being issued repeatedly. + */ +GLboolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + GLuint sz ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + struct header *newheader = (struct header *)data; + + if (brw->flags.always_emit_state) { + brw_batchbuffer_data(brw->batch, data, sz, IGNORE_CLIPRECTS); + return GL_TRUE; + } + + while (item) { + if (item->header->opcode == newheader->opcode) { + if (item->sz == sz && memcmp(item->header, newheader, sz) == 0) + return GL_FALSE; + if (item->sz != sz) { + FREE(item->header); + item->header = MALLOC(sz); + item->sz = sz; + } + goto emit; + } + item = item->next; + } + + assert(!item); + item = CALLOC_STRUCT(brw_cached_batch_item); + item->header = MALLOC(sz); + item->sz = sz; + item->next = brw->cached_batch_items; + brw->cached_batch_items = item; + + emit: + memcpy(item->header, newheader, sz); + brw_batchbuffer_data(brw->batch, data, sz, IGNORE_CLIPRECTS); + return GL_TRUE; +} + +void brw_clear_batch_cache( struct brw_context *brw ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + + while (item) { + struct brw_cached_batch_item *next = item->next; + free((void *)item->header); + free(item); + item = next; + } + + brw->cached_batch_items = NULL; +} + +void brw_destroy_batch_cache( struct brw_context *brw ) +{ + brw_clear_batch_cache(brw); +} diff --git a/src/gallium/drivers/i965/brw_state_cache.c b/src/gallium/drivers/i965/brw_state_cache.c new file mode 100644 index 00000000000..16b643ceb28 --- /dev/null +++ b/src/gallium/drivers/i965/brw_state_cache.c @@ -0,0 +1,617 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +/** @file brw_state_cache.c + * + * This file implements a simple static state cache for 965. The consumers + * can query the hash table of state using a cache_id, opaque key data, + * and list of buffers that will be used in relocations, and receive the + * corresponding state buffer object of state (plus associated auxiliary + * data) in return. + * + * The inner workings are a simple hash table based on a CRC of the key data. + * The cache_id and relocation target buffers associated with the state + * buffer are included as auxiliary key data, but are not part of the hash + * value (this should be fixed, but will likely be fixed instead by making + * consumers use structured keys). + * + * Replacement is not implemented. Instead, when the cache gets too big, at + * a safe point (unlock) we throw out all of the cache data and let it + * regenerate for the next rendering operation. + * + * The reloc structs need to be included as key data, otherwise the + * non-unique values stuffed in the offset in key data through + * brw_cache_data() may result in successful probe for state buffers + * even when the buffer being referenced doesn't match. The result would be + * that the same state cache entry is used twice for different buffers, + * only one of the two buffers referenced gets put into the offset, and the + * incorrect program is run for the other instance. + */ +#include "util/u_memory.h" + +#include "brw_debug.h" +#include "brw_state.h" +#include "brw_batchbuffer.h" + +/* XXX: Fixme - have to include these to get the sizes of the prog_key + * structs: + */ +#include "brw_wm.h" +#include "brw_vs.h" +#include "brw_clip.h" +#include "brw_sf.h" +#include "brw_gs.h" + + +static GLuint +hash_key(const void *key, GLuint key_size, + struct brw_winsys_reloc *relocs, GLuint nr_relocs) +{ + GLuint *ikey = (GLuint *)key; + GLuint hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) { + hash ^= ikey[i]; + hash = (hash << 5) | (hash >> 27); + } + + /* Include the BO pointers as key data as well */ + ikey = (GLuint *)relocs; + key_size = nr_relocs * sizeof(struct brw_winsys_reloc); + for (i = 0; i < key_size/4; i++) { + hash ^= ikey[i]; + hash = (hash << 5) | (hash >> 27); + } + + return hash; +} + + +/** + * Marks a new buffer as being chosen for the given cache id. + */ +static void +update_cache_last(struct brw_cache *cache, enum brw_cache_id cache_id, + struct brw_winsys_buffer *bo) +{ + if (bo == cache->last_bo[cache_id]) + return; /* no change */ + + bo_reference( &cache->last_bo[cache_id], bo ); + + cache->brw->state.dirty.cache |= 1 << cache_id; +} + + +static struct brw_cache_item * +search_cache(struct brw_cache *cache, enum brw_cache_id cache_id, + GLuint hash, const void *key, GLuint key_size, + struct brw_winsys_reloc *relocs, GLuint nr_relocs) +{ + struct brw_cache_item *c; + +#if 0 + int bucketcount = 0; + + for (c = cache->items[hash % cache->size]; c; c = c->next) + bucketcount++; + + debug_printf("bucket %d/%d = %d/%d items\n", hash % cache->size, + cache->size, bucketcount, cache->n_items); +#endif + + for (c = cache->items[hash % cache->size]; c; c = c->next) { + if (c->cache_id == cache_id && + c->hash == hash && + c->key_size == key_size && + memcmp(c->key, key, key_size) == 0 && + c->nr_relocs == nr_relocs && + memcmp(c->relocs, relocs, nr_relocs * sizeof *relocs) == 0) + return c; + } + + return NULL; +} + + +static void +rehash(struct brw_cache *cache) +{ + struct brw_cache_item **items; + struct brw_cache_item *c, *next; + GLuint size, i; + + size = cache->size * 3; + items = (struct brw_cache_item**) CALLOC(size, sizeof(*items)); + + for (i = 0; i < cache->size; i++) + for (c = cache->items[i]; c; c = next) { + next = c->next; + c->next = items[c->hash % size]; + items[c->hash % size] = c; + } + + FREE(cache->items); + cache->items = items; + cache->size = size; +} + + +/** + * Returns the buffer object matching cache_id and key, or NULL. + */ +boolean +brw_search_cache(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *key, + GLuint key_size, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + void *aux_return, + struct brw_winsys_buffer **bo_out) +{ + struct brw_cache_item *item; + GLuint hash = hash_key(key, key_size, relocs, nr_relocs); + + item = search_cache(cache, cache_id, hash, key, key_size, + relocs, nr_relocs); + + if (item) { + if (aux_return) + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + + update_cache_last(cache, cache_id, item->bo); + bo_reference(bo_out, item->bo); + return TRUE; + } + + return FALSE; +} + + +enum pipe_error +brw_upload_cache( struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *key, + GLuint key_size, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + const void *data, + GLuint data_size, + const void *aux, + void *aux_return, + struct brw_winsys_buffer **bo_out) +{ + struct brw_cache_item *item = CALLOC_STRUCT(brw_cache_item); + GLuint hash = hash_key(key, key_size, relocs, nr_relocs); + GLuint relocs_size = nr_relocs * sizeof relocs[0]; + GLuint aux_size = cache->aux_size[cache_id]; + enum pipe_error ret; + void *tmp; + int i; + + /* Create the buffer object to contain the data. For now, use a + * single buffer type to describe all cached state atoms. Later, + * may want to take advantage of hardware distinctions between + * these various entities. + */ + ret = cache->sws->bo_alloc(cache->sws, + cache->buffer_type, + data_size, 1 << 6, + bo_out); + if (ret) + return ret; + + + /* Set up the memory containing the key, aux_data, and relocs */ + tmp = MALLOC(key_size + aux_size + relocs_size); + + memcpy(tmp, key, key_size); + memcpy((char *)tmp + key_size, aux, cache->aux_size[cache_id]); + memcpy((char *)tmp + key_size + aux_size, relocs, relocs_size); + for (i = 0; i < nr_relocs; i++) { + p_atomic_inc(&relocs[i].bo->reference.count); + } + + item->cache_id = cache_id; + item->key = tmp; + item->hash = hash; + item->key_size = key_size; + item->relocs = (struct brw_winsys_reloc *)((char *)tmp + key_size + aux_size); + item->nr_relocs = nr_relocs; + bo_reference( &item->bo, *bo_out ); + item->data_size = data_size; + + if (cache->n_items > cache->size * 1.5) + rehash(cache); + + hash %= cache->size; + item->next = cache->items[hash]; + cache->items[hash] = item; + cache->n_items++; + + if (aux_return) { + assert(cache->aux_size[cache_id]); + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + } + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("upload %s: %d bytes to cache id %d\n", + cache->name[cache_id], + data_size, cache_id); + + /* Copy data to the buffer */ + ret = cache->sws->bo_subdata(item->bo, + cache_id, + 0, data_size, data, + relocs, nr_relocs); + if (ret) + return ret; + + update_cache_last(cache, cache_id, item->bo); + + return PIPE_OK; +} + + +/** + * This doesn't really work with aux data. Use search/upload instead + */ +enum pipe_error +brw_cache_data_sz(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *data, + GLuint data_size, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + struct brw_winsys_buffer **bo_out) +{ + struct brw_cache_item *item; + GLuint hash = hash_key(data, data_size, relocs, nr_relocs); + + item = search_cache(cache, cache_id, hash, data, data_size, + relocs, nr_relocs); + if (item) { + update_cache_last(cache, cache_id, item->bo); + + bo_reference(bo_out, item->bo); + return PIPE_OK; + } + + return brw_upload_cache(cache, cache_id, + data, data_size, + relocs, nr_relocs, + data, data_size, + NULL, NULL, + bo_out); +} + + +/** + * Wrapper around brw_cache_data_sz using the cache_id's canonical key size. + * + * If nr_relocs is nonzero, brw_search_cache()/brw_upload_cache() would be + * better to use, as the potentially changing offsets in the data-used-as-key + * will result in excessive cache misses. + * + * XXX: above is no longer true -- can we remove some code? + */ +enum pipe_error +brw_cache_data(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *data, + struct brw_winsys_reloc *relocs, + GLuint nr_relocs, + struct brw_winsys_buffer **bo_out) +{ + return brw_cache_data_sz(cache, cache_id, data, cache->key_size[cache_id], + relocs, nr_relocs, bo_out); +} + + +static void +brw_init_cache_id(struct brw_cache *cache, + const char *name, + enum brw_cache_id id, + GLuint key_size, + GLuint aux_size) +{ + cache->name[id] = strdup(name); + cache->key_size[id] = key_size; + cache->aux_size[id] = aux_size; +} + + +static void +brw_init_general_state_cache(struct brw_context *brw) +{ + struct brw_cache *cache = &brw->cache; + + cache->brw = brw; + cache->sws = brw->sws; + + cache->buffer_type = BRW_BUFFER_TYPE_GENERAL_STATE; + + cache->size = 7; + cache->n_items = 0; + cache->items = (struct brw_cache_item **) + CALLOC(cache->size, sizeof(struct brw_cache_item)); + + brw_init_cache_id(cache, + "CC_VP", + BRW_CC_VP, + sizeof(struct brw_cc_viewport), + 0); + + brw_init_cache_id(cache, + "CC_UNIT", + BRW_CC_UNIT, + sizeof(struct brw_cc_unit_state), + 0); + + brw_init_cache_id(cache, + "WM_PROG", + BRW_WM_PROG, + sizeof(struct brw_wm_prog_key), + sizeof(struct brw_wm_prog_data)); + + brw_init_cache_id(cache, + "SAMPLER_DEFAULT_COLOR", + BRW_SAMPLER_DEFAULT_COLOR, + sizeof(struct brw_sampler_default_color), + 0); + + brw_init_cache_id(cache, + "SAMPLER", + BRW_SAMPLER, + 0, /* variable key/data size */ + 0); + + brw_init_cache_id(cache, + "WM_UNIT", + BRW_WM_UNIT, + sizeof(struct brw_wm_unit_state), + 0); + + brw_init_cache_id(cache, + "SF_PROG", + BRW_SF_PROG, + sizeof(struct brw_sf_prog_key), + sizeof(struct brw_sf_prog_data)); + + brw_init_cache_id(cache, + "SF_VP", + BRW_SF_VP, + sizeof(struct brw_sf_viewport), + 0); + + brw_init_cache_id(cache, + "SF_UNIT", + BRW_SF_UNIT, + sizeof(struct brw_sf_unit_state), + 0); + + brw_init_cache_id(cache, + "VS_UNIT", + BRW_VS_UNIT, + sizeof(struct brw_vs_unit_state), + 0); + + brw_init_cache_id(cache, + "VS_PROG", + BRW_VS_PROG, + sizeof(struct brw_vs_prog_key), + sizeof(struct brw_vs_prog_data)); + + brw_init_cache_id(cache, + "CLIP_UNIT", + BRW_CLIP_UNIT, + sizeof(struct brw_clip_unit_state), + 0); + + brw_init_cache_id(cache, + "CLIP_PROG", + BRW_CLIP_PROG, + sizeof(struct brw_clip_prog_key), + sizeof(struct brw_clip_prog_data)); + + brw_init_cache_id(cache, + "GS_UNIT", + BRW_GS_UNIT, + sizeof(struct brw_gs_unit_state), + 0); + + brw_init_cache_id(cache, + "GS_PROG", + BRW_GS_PROG, + sizeof(struct brw_gs_prog_key), + sizeof(struct brw_gs_prog_data)); +} + + +static void +brw_init_surface_state_cache(struct brw_context *brw) +{ + struct brw_cache *cache = &brw->surface_cache; + + cache->brw = brw; + cache->sws = brw->sws; + + cache->buffer_type = BRW_BUFFER_TYPE_SURFACE_STATE; + + cache->size = 7; + cache->n_items = 0; + cache->items = (struct brw_cache_item **) + CALLOC(cache->size, sizeof(struct brw_cache_item)); + + brw_init_cache_id(cache, + "SS_SURFACE", + BRW_SS_SURFACE, + sizeof(struct brw_surface_state), + 0); + + brw_init_cache_id(cache, + "SS_SURF_BIND", + BRW_SS_SURF_BIND, + 0, + 0); +} + + +void +brw_init_caches(struct brw_context *brw) +{ + brw_init_general_state_cache(brw); + brw_init_surface_state_cache(brw); +} + + +static void +brw_clear_cache(struct brw_context *brw, struct brw_cache *cache) +{ + struct brw_cache_item *c, *next; + GLuint i; + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("%s\n", __FUNCTION__); + + for (i = 0; i < cache->size; i++) { + for (c = cache->items[i]; c; c = next) { + int j; + + next = c->next; + + for (j = 0; j < c->nr_relocs; j++) + bo_reference(&c->relocs[j].bo, NULL); + + bo_reference(&c->bo, NULL); + FREE((void *)c->key); + FREE(c); + } + cache->items[i] = NULL; + } + + cache->n_items = 0; + + if (brw->curbe.last_buf) { + FREE(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + } + + brw->state.dirty.mesa |= ~0; + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + +/* Clear all entries from the cache that point to the given bo. + * + * This lets us release memory for reuse earlier for known-dead buffers, + * at the cost of walking the entire hash table. + */ +void +brw_state_cache_bo_delete(struct brw_cache *cache, struct brw_winsys_buffer *bo) +{ + struct brw_cache_item **prev; + GLuint i; + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("%s\n", __FUNCTION__); + + for (i = 0; i < cache->size; i++) { + for (prev = &cache->items[i]; *prev;) { + struct brw_cache_item *c = *prev; + + if (cache->sws->bo_references(c->bo, bo)) { + int j; + + *prev = c->next; + + for (j = 0; j < c->nr_relocs; j++) + bo_reference(&c->relocs[j].bo, NULL); + + bo_reference(&c->bo, NULL); + + FREE((void *)c->key); + FREE(c); + cache->n_items--; + } else { + prev = &c->next; + } + } + } +} + +void +brw_state_cache_check_size(struct brw_context *brw) +{ + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items); + + /* un-tuned guess. We've got around 20 state objects for a total of around + * 32k, so 1000 of them is around 1.5MB. + */ + if (brw->cache.n_items > 1000) + brw_clear_cache(brw, &brw->cache); + + if (brw->surface_cache.n_items > 1000) + brw_clear_cache(brw, &brw->surface_cache); +} + + +static void +brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache) +{ + GLuint i; + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("%s\n", __FUNCTION__); + + brw_clear_cache(brw, cache); + for (i = 0; i < BRW_MAX_CACHE; i++) { + bo_reference(&cache->last_bo[i], NULL); + FREE(cache->name[i]); + } + FREE(cache->items); + cache->items = NULL; + cache->size = 0; +} + + +void +brw_destroy_caches(struct brw_context *brw) +{ + brw_destroy_cache(brw, &brw->cache); + brw_destroy_cache(brw, &brw->surface_cache); +} diff --git a/src/gallium/drivers/i965/brw_state_debug.c b/src/gallium/drivers/i965/brw_state_debug.c new file mode 100644 index 00000000000..049c278c93e --- /dev/null +++ b/src/gallium/drivers/i965/brw_state_debug.c @@ -0,0 +1,153 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + + + +#include "brw_context.h" +#include "brw_state.h" + + +struct dirty_bit_map { + uint32_t bit; + char *name; + uint32_t count; +}; + +#define DEFINE_BIT(name) {name, #name, 0} + +static struct dirty_bit_map mesa_bits[] = { + DEFINE_BIT(PIPE_NEW_DEPTH_STENCIL_ALPHA), + DEFINE_BIT(PIPE_NEW_RAST), + DEFINE_BIT(PIPE_NEW_BLEND), + DEFINE_BIT(PIPE_NEW_VIEWPORT), + DEFINE_BIT(PIPE_NEW_SAMPLERS), + DEFINE_BIT(PIPE_NEW_VERTEX_BUFFER), + DEFINE_BIT(PIPE_NEW_VERTEX_ELEMENT), + DEFINE_BIT(PIPE_NEW_FRAGMENT_SHADER), + DEFINE_BIT(PIPE_NEW_VERTEX_SHADER), + DEFINE_BIT(PIPE_NEW_FRAGMENT_CONSTANTS), + DEFINE_BIT(PIPE_NEW_VERTEX_CONSTANTS), + DEFINE_BIT(PIPE_NEW_CLIP), + DEFINE_BIT(PIPE_NEW_INDEX_BUFFER), + DEFINE_BIT(PIPE_NEW_INDEX_RANGE), + DEFINE_BIT(PIPE_NEW_BLEND_COLOR), + DEFINE_BIT(PIPE_NEW_POLYGON_STIPPLE), + DEFINE_BIT(PIPE_NEW_FRAMEBUFFER_DIMENSIONS), + DEFINE_BIT(PIPE_NEW_DEPTH_BUFFER), + DEFINE_BIT(PIPE_NEW_COLOR_BUFFERS), + DEFINE_BIT(PIPE_NEW_QUERY), + DEFINE_BIT(PIPE_NEW_SCISSOR), + DEFINE_BIT(PIPE_NEW_BOUND_TEXTURES), + DEFINE_BIT(PIPE_NEW_NR_CBUFS), + {0, 0, 0} +}; + +static struct dirty_bit_map brw_bits[] = { + DEFINE_BIT(BRW_NEW_URB_FENCE), + DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM), + DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM), + DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS), + DEFINE_BIT(BRW_NEW_CURBE_OFFSETS), + DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE), + DEFINE_BIT(BRW_NEW_PRIMITIVE), + DEFINE_BIT(BRW_NEW_CONTEXT), + DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS), + DEFINE_BIT(BRW_NEW_PSP), + DEFINE_BIT(BRW_NEW_WM_SURFACES), + DEFINE_BIT(BRW_NEW_xxx), + DEFINE_BIT(BRW_NEW_INDICES), + {0, 0, 0} +}; + +static struct dirty_bit_map cache_bits[] = { + DEFINE_BIT(CACHE_NEW_CC_VP), + DEFINE_BIT(CACHE_NEW_CC_UNIT), + DEFINE_BIT(CACHE_NEW_WM_PROG), + DEFINE_BIT(CACHE_NEW_SAMPLER_DEFAULT_COLOR), + DEFINE_BIT(CACHE_NEW_SAMPLER), + DEFINE_BIT(CACHE_NEW_WM_UNIT), + DEFINE_BIT(CACHE_NEW_SF_PROG), + DEFINE_BIT(CACHE_NEW_SF_VP), + DEFINE_BIT(CACHE_NEW_SF_UNIT), + DEFINE_BIT(CACHE_NEW_VS_UNIT), + DEFINE_BIT(CACHE_NEW_VS_PROG), + DEFINE_BIT(CACHE_NEW_GS_UNIT), + DEFINE_BIT(CACHE_NEW_GS_PROG), + DEFINE_BIT(CACHE_NEW_CLIP_VP), + DEFINE_BIT(CACHE_NEW_CLIP_UNIT), + DEFINE_BIT(CACHE_NEW_CLIP_PROG), + DEFINE_BIT(CACHE_NEW_SURFACE), + DEFINE_BIT(CACHE_NEW_SURF_BIND), + {0, 0, 0} +}; + + +static void +brw_update_dirty_count(struct dirty_bit_map *bit_map, int32_t bits) +{ + int i; + + for (i = 0; i < 32; i++) { + if (bit_map[i].bit == 0) + return; + + if (bit_map[i].bit & bits) + bit_map[i].count++; + } +} + +static void +brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits) +{ + int i; + + for (i = 0; i < 32; i++) { + if (bit_map[i].bit == 0) + return; + + debug_printf("0x%08x: %12d (%s)\n", + bit_map[i].bit, bit_map[i].count, bit_map[i].name); + } +} + +void +brw_update_dirty_counts( unsigned mesa, + unsigned brw, + unsigned cache ) +{ + static int dirty_count = 0; + + brw_update_dirty_count(mesa_bits, mesa); + brw_update_dirty_count(brw_bits, brw); + brw_update_dirty_count(cache_bits, cache); + if (dirty_count++ % 1000 == 0) { + brw_print_dirty_count(mesa_bits, mesa); + brw_print_dirty_count(brw_bits, brw); + brw_print_dirty_count(cache_bits, cache); + debug_printf("\n"); + } +} diff --git a/src/gallium/drivers/i965/brw_state_upload.c b/src/gallium/drivers/i965/brw_state_upload.c new file mode 100644 index 00000000000..f8b91eff816 --- /dev/null +++ b/src/gallium/drivers/i965/brw_state_upload.c @@ -0,0 +1,270 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_batchbuffer.h" +#include "brw_debug.h" + +const struct brw_tracked_state *atoms[] = +{ +/* &brw_wm_input_sizes, */ + &brw_vs_prog, + &brw_gs_prog, + &brw_clip_prog, + &brw_sf_prog, + &brw_wm_prog, + + /* Once all the programs are done, we know how large urb entry + * sizes need to be and can decide if we need to change the urb + * layout. + */ + &brw_curbe_offsets, + &brw_recalculate_urb_fence, + + &brw_cc_vp, + &brw_cc_unit, + + &brw_vs_surfaces, /* must do before unit */ + /*&brw_wm_constant_surface,*/ /* must do before wm surfaces/bind bo */ + &brw_wm_surfaces, /* must do before samplers and unit */ + &brw_wm_samplers, + + &brw_wm_unit, + &brw_sf_vp, + &brw_sf_unit, + &brw_vs_unit, /* always required, enabled or not */ + &brw_clip_unit, + &brw_gs_unit, + + /* Command packets: + */ + &brw_invarient_state, + &brw_state_base_address, + + &brw_binding_table_pointers, + &brw_blend_constant_color, + + &brw_depthbuffer, + &brw_polygon_stipple, + &brw_line_stipple, + + &brw_psp_urb_cbs, + + &brw_drawing_rect, + &brw_indices, + &brw_index_buffer, + &brw_vertices, + + &brw_curbe_buffer +}; + + +void brw_init_state( struct brw_context *brw ) +{ + brw_init_caches(brw); +} + + +void brw_destroy_state( struct brw_context *brw ) +{ + brw_destroy_caches(brw); + brw_destroy_batch_cache(brw); +} + +/*********************************************************************** + */ + +static GLboolean check_state( const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + return ((a->mesa & b->mesa) || + (a->brw & b->brw) || + (a->cache & b->cache)); +} + +static void accumulate_state( struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + a->mesa |= b->mesa; + a->brw |= b->brw; + a->cache |= b->cache; +} + + +static void xor_states( struct brw_state_flags *result, + const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + result->mesa = a->mesa ^ b->mesa; + result->brw = a->brw ^ b->brw; + result->cache = a->cache ^ b->cache; +} + +static void +brw_clear_validated_bos(struct brw_context *brw) +{ + int i; + + /* Clear the last round of validated bos */ + for (i = 0; i < brw->state.validated_bo_count; i++) { + bo_reference(&brw->state.validated_bos[i], NULL); + } + brw->state.validated_bo_count = 0; +} + + +/*********************************************************************** + * Emit all state: + */ +enum pipe_error brw_validate_state( struct brw_context *brw ) +{ + struct brw_state_flags *state = &brw->state.dirty; + GLuint i; + int ret; + + brw_clear_validated_bos(brw); + brw_add_validated_bo(brw, brw->batch->buf); + + if (brw->flags.always_emit_state) { + state->mesa |= ~0; + state->brw |= ~0; + state->cache |= ~0; + } + + if (state->mesa == 0 && + state->cache == 0 && + state->brw == 0) + return 0; + + if (brw->state.dirty.brw & BRW_NEW_CONTEXT) + brw_clear_batch_cache(brw); + + /* do prepare stage for all atoms */ + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; + + if (check_state(state, &atom->dirty)) { + if (atom->prepare) { + ret = atom->prepare(brw); + if (ret) + return ret; + } + } + } + + /* Make sure that the textures which are referenced by the current + * brw fragment program are actually present/valid. + * If this fails, we can experience GPU lock-ups. + */ + { + const struct brw_fragment_shader *fp = brw->curr.fragment_shader; + if (fp) { + assert(fp->info.file_max[TGSI_FILE_SAMPLER] < (int)brw->curr.num_samplers); + /*assert(fp->info.texture_max <= brw->curr.num_textures);*/ + } + } + + return 0; +} + + +enum pipe_error brw_upload_state(struct brw_context *brw) +{ + struct brw_state_flags *state = &brw->state.dirty; + int ret; + int i; + + brw_clear_validated_bos(brw); + + if (BRW_DEBUG) { + /* Debug version which enforces various sanity checks on the + * state flags which are generated and checked to help ensure + * state atoms are ordered correctly in the list. + */ + struct brw_state_flags examined, prev; + memset(&examined, 0, sizeof(examined)); + prev = *state; + + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; + struct brw_state_flags generated; + + assert(atom->dirty.mesa || + atom->dirty.brw || + atom->dirty.cache); + + if (check_state(state, &atom->dirty)) { + if (atom->emit) { + ret = atom->emit( brw ); + if (ret) + return ret; + } + } + + accumulate_state(&examined, &atom->dirty); + + /* generated = (prev ^ state) + * if (examined & generated) + * fail; + */ + xor_states(&generated, &prev, state); + assert(!check_state(&examined, &generated)); + prev = *state; + } + } + else { + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; + + if (check_state(state, &atom->dirty)) { + if (atom->emit) { + ret = atom->emit( brw ); + if (ret) + return ret; + } + } + } + } + + if (BRW_DEBUG & DEBUG_STATE) { + brw_update_dirty_counts( state->mesa, + state->brw, + state->cache ); + } + + /* Clear dirty flags: + */ + memset(state, 0, sizeof(*state)); + return 0; +} diff --git a/src/gallium/drivers/i965/brw_structs.h b/src/gallium/drivers/i965/brw_structs.h new file mode 100644 index 00000000000..bf10bc04de7 --- /dev/null +++ b/src/gallium/drivers/i965/brw_structs.h @@ -0,0 +1,1576 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_STRUCTS_H +#define BRW_STRUCTS_H + +#include "brw_types.h" + +/** Number of general purpose registers (VS, WM, etc) */ +#define BRW_MAX_GRF 128 + +/** Number of message register file registers */ +#define BRW_MAX_MRF 16 + + +/* Command packets: + */ +struct header +{ + GLuint length:16; + GLuint opcode:16; +}; + + +union header_union +{ + struct header bits; + GLuint dword; +}; + +struct brw_3d_control +{ + struct + { + GLuint length:8; + GLuint notify_enable:1; + GLuint pad:3; + GLuint wc_flush_enable:1; + GLuint depth_stall_enable:1; + GLuint operation:2; + GLuint opcode:16; + } header; + + struct + { + GLuint pad:2; + GLuint dest_addr_type:1; + GLuint dest_addr:29; + } dest; + + GLuint dword2; + GLuint dword3; +}; + + +struct brw_3d_primitive +{ + struct + { + GLuint length:8; + GLuint pad:2; + GLuint topology:5; + GLuint indexed:1; + GLuint opcode:16; + } header; + + GLuint verts_per_instance; + GLuint start_vert_location; + GLuint instance_count; + GLuint start_instance_location; + GLuint base_vert_location; +}; + +/* These seem to be passed around as function args, so it works out + * better to keep them as #defines: + */ +#define BRW_FLUSH_READ_CACHE 0x1 +#define BRW_FLUSH_STATE_CACHE 0x2 +#define BRW_INHIBIT_FLUSH_RENDER_CACHE 0x4 +#define BRW_FLUSH_SNAPSHOT_COUNTERS 0x8 + +struct brw_mi_flush +{ + GLuint flags:4; + GLuint pad:12; + GLuint opcode:16; +}; + +struct brw_vf_statistics +{ + GLuint statistics_enable:1; + GLuint pad:15; + GLuint opcode:16; +}; + + + +struct brw_binding_table_pointers +{ + struct header header; + GLuint vs; + GLuint gs; + GLuint clp; + GLuint sf; + GLuint wm; +}; + + +struct brw_blend_constant_color +{ + struct header header; + GLfloat blend_constant_color[4]; +}; + + +struct brw_depthbuffer +{ + union header_union header; + + union { + struct { + GLuint pitch:18; + GLuint format:3; + GLuint pad:2; + GLuint software_tiled_rendering_mode:2; + GLuint depth_offset_disable:1; + GLuint tile_walk:1; + GLuint tiled_surface:1; + GLuint pad2:1; + GLuint surface_type:3; + } bits; + GLuint dword; + } dword1; + + GLuint dword2_base_addr; + + union { + struct { + GLuint pad:1; + GLuint mipmap_layout:1; + GLuint lod:4; + GLuint width:13; + GLuint height:13; + } bits; + GLuint dword; + } dword3; + + union { + struct { + GLuint pad:10; + GLuint min_array_element:11; + GLuint depth:11; + } bits; + GLuint dword; + } dword4; +}; + +struct brw_depthbuffer_g4x +{ + union header_union header; + + union { + struct { + GLuint pitch:18; + GLuint format:3; + GLuint pad:2; + GLuint software_tiled_rendering_mode:2; + GLuint depth_offset_disable:1; + GLuint tile_walk:1; + GLuint tiled_surface:1; + GLuint pad2:1; + GLuint surface_type:3; + } bits; + GLuint dword; + } dword1; + + GLuint dword2_base_addr; + + union { + struct { + GLuint pad:1; + GLuint mipmap_layout:1; + GLuint lod:4; + GLuint width:13; + GLuint height:13; + } bits; + GLuint dword; + } dword3; + + union { + struct { + GLuint pad:10; + GLuint min_array_element:11; + GLuint depth:11; + } bits; + GLuint dword; + } dword4; + + union { + struct { + GLuint xoffset:16; + GLuint yoffset:16; + } bits; + GLuint dword; + } dword5; /* NEW in Integrated Graphics Device */ +}; + +struct brw_drawrect +{ + struct header header; + GLuint xmin:16; + GLuint ymin:16; + GLuint xmax:16; + GLuint ymax:16; + GLuint xorg:16; + GLuint yorg:16; +}; + + + + +struct brw_global_depth_offset_clamp +{ + struct header header; + GLfloat depth_offset_clamp; +}; + +struct brw_indexbuffer +{ + union { + struct + { + GLuint length:8; + GLuint index_format:2; + GLuint cut_index_enable:1; + GLuint pad:5; + GLuint opcode:16; + } bits; + GLuint dword; + + } header; + + GLuint buffer_start; + GLuint buffer_end; +}; + +/* NEW in Integrated Graphics Device */ +struct brw_aa_line_parameters +{ + struct header header; + + struct { + GLuint aa_coverage_scope:8; + GLuint pad0:8; + GLuint aa_coverage_bias:8; + GLuint pad1:8; + } bits0; + + struct { + GLuint aa_coverage_endcap_slope:8; + GLuint pad0:8; + GLuint aa_coverage_endcap_bias:8; + GLuint pad1:8; + } bits1; +}; + +struct brw_line_stipple +{ + struct header header; + + struct + { + GLuint pattern:16; + GLuint pad:16; + } bits0; + + struct + { + GLuint repeat_count:9; + GLuint pad:7; + GLuint inverse_repeat_count:16; + } bits1; +}; + + +struct brw_pipelined_state_pointers +{ + struct header header; + + struct { + GLuint pad:5; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE */ + } vs; + + struct + { + GLuint enable:1; + GLuint pad:4; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE */ + } gs; + + struct + { + GLuint enable:1; + GLuint pad:4; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE */ + } clp; + + struct + { + GLuint pad:5; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE */ + } sf; + + struct + { + GLuint pad:5; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE */ + } wm; + + struct + { + GLuint pad:5; + GLuint offset:27; /* Offset from GENERAL_STATE_BASE. KW: check me! */ + } cc; +}; + + +struct brw_polygon_stipple_offset +{ + struct header header; + + struct { + GLuint y_offset:5; + GLuint pad:3; + GLuint x_offset:5; + GLuint pad0:19; + } bits0; +}; + + + +struct brw_polygon_stipple +{ + struct header header; + GLuint stipple[32]; +}; + + + +struct brw_pipeline_select +{ + struct + { + GLuint pipeline_select:1; + GLuint pad:15; + GLuint opcode:16; + } header; +}; + + +struct brw_pipe_control +{ + struct + { + GLuint length:8; + GLuint notify_enable:1; + GLuint texture_cache_flush_enable:1; + GLuint indirect_state_pointers_disable:1; + GLuint instruction_state_cache_flush_enable:1; + GLuint write_cache_flush_enable:1; + GLuint depth_stall_enable:1; + GLuint post_sync_operation:2; + + GLuint opcode:16; + } header; + + struct + { + GLuint pad:2; + GLuint dest_addr_type:1; + GLuint dest_addr:29; + } bits1; + + GLuint data0; + GLuint data1; +}; + + +struct brw_urb_fence +{ + struct + { + GLuint length:8; + GLuint vs_realloc:1; + GLuint gs_realloc:1; + GLuint clp_realloc:1; + GLuint sf_realloc:1; + GLuint vfe_realloc:1; + GLuint cs_realloc:1; + GLuint pad:2; + GLuint opcode:16; + } header; + + struct + { + GLuint vs_fence:10; + GLuint gs_fence:10; + GLuint clp_fence:10; + GLuint pad:2; + } bits0; + + struct + { + GLuint sf_fence:10; + GLuint vf_fence:10; + GLuint cs_fence:11; + GLuint pad:1; + } bits1; +}; + +struct brw_cs_urb_state +{ + struct header header; + + struct + { + GLuint nr_urb_entries:3; + GLuint pad:1; + GLuint urb_entry_size:5; + GLuint pad0:23; + } bits0; +}; + +struct brw_constant_buffer +{ + struct + { + GLuint length:8; + GLuint valid:1; + GLuint pad:7; + GLuint opcode:16; + } header; + + struct + { + GLuint buffer_length:6; + GLuint buffer_address:26; + } bits0; +}; + +struct brw_state_base_address +{ + struct header header; + + struct + { + GLuint modify_enable:1; + GLuint pad:4; + GLuint general_state_address:27; + } bits0; + + struct + { + GLuint modify_enable:1; + GLuint pad:4; + GLuint surface_state_address:27; + } bits1; + + struct + { + GLuint modify_enable:1; + GLuint pad:4; + GLuint indirect_object_state_address:27; + } bits2; + + struct + { + GLuint modify_enable:1; + GLuint pad:11; + GLuint general_state_upper_bound:20; + } bits3; + + struct + { + GLuint modify_enable:1; + GLuint pad:11; + GLuint indirect_object_state_upper_bound:20; + } bits4; +}; + +struct brw_state_prefetch +{ + struct header header; + + struct + { + GLuint prefetch_count:3; + GLuint pad:3; + GLuint prefetch_pointer:26; + } bits0; +}; + +struct brw_system_instruction_pointer +{ + struct header header; + + struct + { + GLuint pad:4; + GLuint system_instruction_pointer:28; + } bits0; +}; + + + + +/* State structs for the various fixed function units: + */ + + +struct thread0 +{ + GLuint pad0:1; + GLuint grf_reg_count:3; + GLuint pad1:2; + GLuint kernel_start_pointer:26; /* Offset from GENERAL_STATE_BASE */ +}; + +struct thread1 +{ + GLuint ext_halt_exception_enable:1; + GLuint sw_exception_enable:1; + GLuint mask_stack_exception_enable:1; + GLuint timeout_exception_enable:1; + GLuint illegal_op_exception_enable:1; + GLuint pad0:3; + GLuint depth_coef_urb_read_offset:6; /* WM only */ + GLuint pad1:2; + GLuint floating_point_mode:1; + GLuint thread_priority:1; + GLuint binding_table_entry_count:8; + GLuint pad3:5; + GLuint single_program_flow:1; +}; + +struct thread2 +{ + GLuint per_thread_scratch_space:4; + GLuint pad0:6; + GLuint scratch_space_base_pointer:22; +}; + + +struct thread3 +{ + GLuint dispatch_grf_start_reg:4; + GLuint urb_entry_read_offset:6; + GLuint pad0:1; + GLuint urb_entry_read_length:6; + GLuint pad1:1; + GLuint const_urb_entry_read_offset:6; + GLuint pad2:1; + GLuint const_urb_entry_read_length:6; + GLuint pad3:1; +}; + + + +struct brw_clip_unit_state +{ + struct thread0 thread0; + struct + { + GLuint pad0:7; + GLuint sw_exception_enable:1; + GLuint pad1:3; + GLuint mask_stack_exception_enable:1; + GLuint pad2:1; + GLuint illegal_op_exception_enable:1; + GLuint pad3:2; + GLuint floating_point_mode:1; + GLuint thread_priority:1; + GLuint binding_table_entry_count:8; + GLuint pad4:5; + GLuint single_program_flow:1; + } thread1; + + struct thread2 thread2; + struct thread3 thread3; + + struct + { + GLuint pad0:9; + GLuint gs_output_stats:1; /* not always */ + GLuint stats_enable:1; + GLuint nr_urb_entries:7; + GLuint pad1:1; + GLuint urb_entry_allocation_size:5; + GLuint pad2:1; + GLuint max_threads:5; /* may be less */ + GLuint pad3:2; + } thread4; + + struct + { + GLuint pad0:13; + GLuint clip_mode:3; + GLuint userclip_enable_flags:8; + GLuint userclip_must_clip:1; + GLuint negative_w_clip_test:1; + GLuint guard_band_enable:1; + GLuint viewport_z_clip_enable:1; + GLuint viewport_xy_clip_enable:1; + GLuint vertex_position_space:1; + GLuint api_mode:1; + GLuint pad2:1; + } clip5; + + struct + { + GLuint pad0:5; + GLuint clipper_viewport_state_ptr:27; + } clip6; + + + GLfloat viewport_xmin; + GLfloat viewport_xmax; + GLfloat viewport_ymin; + GLfloat viewport_ymax; +}; + + + +struct brw_cc_unit_state +{ + struct brw_cc0 + { + GLuint pad0:3; + GLuint bf_stencil_pass_depth_pass_op:3; + GLuint bf_stencil_pass_depth_fail_op:3; + GLuint bf_stencil_fail_op:3; + GLuint bf_stencil_func:3; + GLuint bf_stencil_enable:1; + GLuint pad1:2; + GLuint stencil_write_enable:1; + GLuint stencil_pass_depth_pass_op:3; + GLuint stencil_pass_depth_fail_op:3; + GLuint stencil_fail_op:3; + GLuint stencil_func:3; + GLuint stencil_enable:1; + } cc0; + + + struct brw_cc1 + { + GLuint bf_stencil_ref:8; + GLuint stencil_write_mask:8; + GLuint stencil_test_mask:8; + GLuint stencil_ref:8; + } cc1; + + + struct brw_cc2 + { + GLuint logicop_enable:1; + GLuint pad0:10; + GLuint depth_write_enable:1; + GLuint depth_test_function:3; + GLuint depth_test:1; + GLuint bf_stencil_write_mask:8; + GLuint bf_stencil_test_mask:8; + } cc2; + + + struct brw_cc3 + { + GLuint pad0:8; + GLuint alpha_test_func:3; + GLuint alpha_test:1; + GLuint blend_enable:1; + GLuint ia_blend_enable:1; + GLuint pad1:1; + GLuint alpha_test_format:1; + GLuint pad2:16; + } cc3; + + struct brw_cc4 + { + GLuint pad0:5; + GLuint cc_viewport_state_offset:27; /* Offset from GENERAL_STATE_BASE */ + } cc4; + + struct brw_cc5 + { + GLuint pad0:2; + GLuint ia_dest_blend_factor:5; + GLuint ia_src_blend_factor:5; + GLuint ia_blend_function:3; + GLuint statistics_enable:1; + GLuint logicop_func:4; + GLuint pad1:11; + GLuint dither_enable:1; + } cc5; + + struct brw_cc6 + { + GLuint clamp_post_alpha_blend:1; + GLuint clamp_pre_alpha_blend:1; + GLuint clamp_range:2; + GLuint pad0:11; + GLuint y_dither_offset:2; + GLuint x_dither_offset:2; + GLuint dest_blend_factor:5; + GLuint src_blend_factor:5; + GLuint blend_function:3; + } cc6; + + struct brw_cc7 { + union { + GLfloat f; + GLubyte ub[4]; + } alpha_ref; + } cc7; +}; + + + +struct brw_sf_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + GLuint pad0:10; + GLuint stats_enable:1; + GLuint nr_urb_entries:7; + GLuint pad1:1; + GLuint urb_entry_allocation_size:5; + GLuint pad2:1; + GLuint max_threads:6; + GLuint pad3:1; + } thread4; + + struct + { + GLuint front_winding:1; + GLuint viewport_transform:1; + GLuint pad0:3; + GLuint sf_viewport_state_offset:27; /* Offset from GENERAL_STATE_BASE */ + } sf5; + + struct + { + GLuint pad0:9; + GLuint dest_org_vbias:4; + GLuint dest_org_hbias:4; + GLuint scissor:1; + GLuint disable_2x2_trifilter:1; + GLuint disable_zero_pix_trifilter:1; + GLuint point_rast_rule:2; + GLuint line_endcap_aa_region_width:2; + GLuint line_width:4; + GLuint fast_scissor_disable:1; + GLuint cull_mode:2; + GLuint aa_enable:1; + } sf6; + + struct + { + GLuint point_size:11; + GLuint use_point_size_state:1; + GLuint subpixel_precision:1; + GLuint sprite_point:1; + GLuint pad0:10; + GLuint aa_line_distance_mode:1; + GLuint trifan_pv:2; + GLuint linestrip_pv:2; + GLuint tristrip_pv:2; + GLuint line_last_pixel_enable:1; + } sf7; + +}; + + +struct brw_gs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + GLuint pad0:8; + GLuint rendering_enable:1; /* for IGDNG */ + GLuint pad4:1; + GLuint stats_enable:1; + GLuint nr_urb_entries:7; + GLuint pad1:1; + GLuint urb_entry_allocation_size:5; + GLuint pad2:1; + GLuint max_threads:5; + GLuint pad3:2; + } thread4; + + struct + { + GLuint sampler_count:3; + GLuint pad0:2; + GLuint sampler_state_pointer:27; + } gs5; + + + struct + { + GLuint max_vp_index:4; + GLuint pad0:12; + GLuint svbi_post_inc_value:10; + GLuint pad1:1; + GLuint svbi_post_inc_enable:1; + GLuint svbi_payload:1; + GLuint discard_adjaceny:1; + GLuint reorder_enable:1; + GLuint pad2:1; + } gs6; +}; + + +struct brw_vs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + GLuint pad0:10; + GLuint stats_enable:1; + GLuint nr_urb_entries:7; + GLuint pad1:1; + GLuint urb_entry_allocation_size:5; + GLuint pad2:1; + GLuint max_threads:6; + GLuint pad3:1; + } thread4; + + struct + { + GLuint sampler_count:3; + GLuint pad0:2; + GLuint sampler_state_pointer:27; + } vs5; + + struct + { + GLuint vs_enable:1; + GLuint vert_cache_disable:1; + GLuint pad0:30; + } vs6; +}; + + +struct brw_wm_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct { + GLuint stats_enable:1; + GLuint depth_buffer_clear:1; + GLuint sampler_count:3; + GLuint sampler_state_pointer:27; + } wm4; + + struct + { + GLuint enable_8_pix:1; + GLuint enable_16_pix:1; + GLuint enable_32_pix:1; + GLuint enable_con_32_pix:1; + GLuint enable_con_64_pix:1; + GLuint pad0:5; + GLuint legacy_global_depth_bias:1; + GLuint line_stipple:1; + GLuint depth_offset:1; + GLuint polygon_stipple:1; + GLuint line_aa_region_width:2; + GLuint line_endcap_aa_region_width:2; + GLuint early_depth_test:1; + GLuint thread_dispatch_enable:1; + GLuint program_uses_depth:1; + GLuint program_computes_depth:1; + GLuint program_uses_killpixel:1; + GLuint legacy_line_rast: 1; + GLuint transposed_urb_read_enable:1; + GLuint max_threads:7; + } wm5; + + GLfloat global_depth_offset_constant; + GLfloat global_depth_offset_scale; + + /* for IGDNG only */ + struct { + GLuint pad0:1; + GLuint grf_reg_count_1:3; + GLuint pad1:2; + GLuint kernel_start_pointer_1:26; + } wm8; + + struct { + GLuint pad0:1; + GLuint grf_reg_count_2:3; + GLuint pad1:2; + GLuint kernel_start_pointer_2:26; + } wm9; + + struct { + GLuint pad0:1; + GLuint grf_reg_count_3:3; + GLuint pad1:2; + GLuint kernel_start_pointer_3:26; + } wm10; +}; + +struct brw_sampler_default_color { + GLfloat color[4]; +}; + +struct brw_sampler_state +{ + + struct brw_ss0 + { + GLuint shadow_function:3; + GLuint lod_bias:11; + GLuint min_filter:3; + GLuint mag_filter:3; + GLuint mip_filter:2; + GLuint base_level:5; + GLuint pad:1; + GLuint lod_preclamp:1; + GLuint default_color_mode:1; + GLuint pad0:1; + GLuint disable:1; + } ss0; + + struct brw_ss1 + { + GLuint r_wrap_mode:3; + GLuint t_wrap_mode:3; + GLuint s_wrap_mode:3; + GLuint pad:3; + GLuint max_lod:10; + GLuint min_lod:10; + } ss1; + + + struct brw_ss2 + { + GLuint pad:5; + GLuint default_color_pointer:27; + } ss2; + + struct brw_ss3 + { + GLuint pad:19; + GLuint max_aniso:3; + GLuint chroma_key_mode:1; + GLuint chroma_key_index:2; + GLuint chroma_key_enable:1; + GLuint monochrome_filter_width:3; + GLuint monochrome_filter_height:3; + } ss3; +}; + + +struct brw_clipper_viewport +{ + GLfloat xmin; + GLfloat xmax; + GLfloat ymin; + GLfloat ymax; +}; + +struct brw_cc_viewport +{ + GLfloat min_depth; + GLfloat max_depth; +}; + +struct brw_sf_viewport +{ + struct { + GLfloat m00; + GLfloat m11; + GLfloat m22; + GLfloat m30; + GLfloat m31; + GLfloat m32; + } viewport; + + /* scissor coordinates are inclusive */ + struct { + GLshort xmin; + GLshort ymin; + GLshort xmax; + GLshort ymax; + } scissor; +}; + +/* Documented in the subsystem/shared-functions/sampler chapter... + */ +struct brw_surface_state +{ + struct brw_surf_ss0 { + GLuint cube_pos_z:1; + GLuint cube_neg_z:1; + GLuint cube_pos_y:1; + GLuint cube_neg_y:1; + GLuint cube_pos_x:1; + GLuint cube_neg_x:1; + GLuint pad:4; + GLuint mipmap_layout_mode:1; + GLuint vert_line_stride_ofs:1; + GLuint vert_line_stride:1; + GLuint color_blend:1; + GLuint writedisable_blue:1; + GLuint writedisable_green:1; + GLuint writedisable_red:1; + GLuint writedisable_alpha:1; + GLuint surface_format:9; /**< BRW_SURFACEFORMAT_x */ + GLuint data_return_format:1; + GLuint pad0:1; + GLuint surface_type:3; /**< BRW_SURFACE_1D/2D/3D/CUBE */ + } ss0; + + struct brw_surf_ss1 { + GLuint base_addr; + } ss1; + + struct brw_surf_ss2 { + GLuint pad:2; + GLuint mip_count:4; + GLuint width:13; + GLuint height:13; + } ss2; + + struct brw_surf_ss3 { + GLuint tile_walk:1; + GLuint tiled_surface:1; + GLuint pad:1; + GLuint pitch:18; + GLuint depth:11; + } ss3; + + struct brw_surf_ss4 { + GLuint multisample_position_palette_index:3; + GLuint pad1:1; + GLuint num_multisamples:3; + GLuint pad0:1; + GLuint render_target_view_extent:9; + GLuint min_array_elt:11; + GLuint min_lod:4; + } ss4; + + struct brw_surf_ss5 { + GLuint pad1:16; + GLuint llc_mapping:1; + GLuint mlc_mapping:1; + GLuint gfdt:1; + GLuint gfdt_src:1; + GLuint y_offset:4; + GLuint pad0:1; + GLuint x_offset:7; + } ss5; /* New in G4X */ + +}; + + + +struct brw_vertex_buffer_state +{ + struct { + GLuint pitch:11; + GLuint pad:15; + GLuint access_type:1; + GLuint vb_index:5; + } vb0; + + GLuint start_addr; + GLuint max_index; +#if 1 + GLuint instance_data_step_rate; /* not included for sequential/random vertices? */ +#endif +}; + +#define BRW_VBP_MAX 17 + +struct brw_vb_array_state { + struct header header; + struct brw_vertex_buffer_state vb[BRW_VBP_MAX]; +}; + + +struct brw_vertex_element_state +{ + struct + { + GLuint src_offset:11; + GLuint pad:5; + GLuint src_format:9; + GLuint pad0:1; + GLuint valid:1; + GLuint vertex_buffer_index:5; + } ve0; + + struct + { + GLuint dst_offset:8; + GLuint pad:8; + GLuint vfcomponent3:4; + GLuint vfcomponent2:4; + GLuint vfcomponent1:4; + GLuint vfcomponent0:4; + } ve1; +}; + +#define BRW_VEP_MAX 18 + +struct brw_vertex_element_packet { + struct header header; + struct brw_vertex_element_state ve[BRW_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ +}; + + +struct brw_urb_immediate { + GLuint opcode:4; + GLuint offset:6; + GLuint swizzle_control:2; + GLuint pad:1; + GLuint allocate:1; + GLuint used:1; + GLuint complete:1; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; +}; + +/* Instruction format for the execution units: + */ + +struct brw_instruction +{ + struct + { + GLuint opcode:7; + GLuint pad:1; + GLuint access_mode:1; + GLuint mask_control:1; + GLuint dependency_control:2; + GLuint compression_control:2; + GLuint thread_control:2; + GLuint predicate_control:4; + GLuint predicate_inverse:1; + GLuint execution_size:3; + GLuint destreg__conditionalmod:4; /* destreg - send, conditionalmod - others */ + GLuint pad0:2; + GLuint debug_control:1; + GLuint saturate:1; + } header; + + union { + struct + { + GLuint dest_reg_file:2; + GLuint dest_reg_type:3; + GLuint src0_reg_file:2; + GLuint src0_reg_type:3; + GLuint src1_reg_file:2; + GLuint src1_reg_type:3; + GLuint pad:1; + GLuint dest_subreg_nr:5; + GLuint dest_reg_nr:8; + GLuint dest_horiz_stride:2; + GLuint dest_address_mode:1; + } da1; + + struct + { + GLuint dest_reg_file:2; + GLuint dest_reg_type:3; + GLuint src0_reg_file:2; + GLuint src0_reg_type:3; + GLuint src1_reg_file:2; /* 0x00000c00 */ + GLuint src1_reg_type:3; /* 0x00007000 */ + GLuint pad:1; + GLint dest_indirect_offset:10; /* offset against the deref'd address reg */ + GLuint dest_subreg_nr:3; /* subnr for the address reg a0.x */ + GLuint dest_horiz_stride:2; + GLuint dest_address_mode:1; + } ia1; + + struct + { + GLuint dest_reg_file:2; + GLuint dest_reg_type:3; + GLuint src0_reg_file:2; + GLuint src0_reg_type:3; + GLuint src1_reg_file:2; + GLuint src1_reg_type:3; + GLuint pad:1; + GLuint dest_writemask:4; + GLuint dest_subreg_nr:1; + GLuint dest_reg_nr:8; + GLuint pad1:2; + GLuint dest_address_mode:1; + } da16; + + struct + { + GLuint dest_reg_file:2; + GLuint dest_reg_type:3; + GLuint src0_reg_file:2; + GLuint src0_reg_type:3; + GLuint pad0:6; + GLuint dest_writemask:4; + GLint dest_indirect_offset:6; + GLuint dest_subreg_nr:3; + GLuint pad1:2; + GLuint dest_address_mode:1; + } ia16; + } bits1; + + + union { + struct + { + GLuint src0_subreg_nr:5; + GLuint src0_reg_nr:8; + GLuint src0_abs:1; + GLuint src0_negate:1; + GLuint src0_address_mode:1; + GLuint src0_horiz_stride:2; + GLuint src0_width:3; + GLuint src0_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad:6; + } da1; + + struct + { + GLint src0_indirect_offset:10; + GLuint src0_subreg_nr:3; + GLuint src0_abs:1; + GLuint src0_negate:1; + GLuint src0_address_mode:1; + GLuint src0_horiz_stride:2; + GLuint src0_width:3; + GLuint src0_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad:6; + } ia1; + + struct + { + GLuint src0_swz_x:2; + GLuint src0_swz_y:2; + GLuint src0_subreg_nr:1; + GLuint src0_reg_nr:8; + GLuint src0_abs:1; + GLuint src0_negate:1; + GLuint src0_address_mode:1; + GLuint src0_swz_z:2; + GLuint src0_swz_w:2; + GLuint pad0:1; + GLuint src0_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad1:6; + } da16; + + struct + { + GLuint src0_swz_x:2; + GLuint src0_swz_y:2; + GLint src0_indirect_offset:6; + GLuint src0_subreg_nr:3; + GLuint src0_abs:1; + GLuint src0_negate:1; + GLuint src0_address_mode:1; + GLuint src0_swz_z:2; + GLuint src0_swz_w:2; + GLuint pad0:1; + GLuint src0_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad1:6; + } ia16; + + struct + { + GLuint pad:26; + GLuint end_of_thread:1; + GLuint pad1:1; + GLuint sfid:4; + } send_igdng; /* for IGDNG only */ + + } bits2; + + union + { + struct + { + GLuint src1_subreg_nr:5; + GLuint src1_reg_nr:8; + GLuint src1_abs:1; + GLuint src1_negate:1; + GLuint src1_address_mode:1; + GLuint src1_horiz_stride:2; + GLuint src1_width:3; + GLuint src1_vert_stride:4; + GLuint pad0:7; + } da1; + + struct + { + GLuint src1_swz_x:2; + GLuint src1_swz_y:2; + GLuint src1_subreg_nr:1; + GLuint src1_reg_nr:8; + GLuint src1_abs:1; + GLuint src1_negate:1; + GLuint src1_address_mode:1; + GLuint src1_swz_z:2; + GLuint src1_swz_w:2; + GLuint pad1:1; + GLuint src1_vert_stride:4; + GLuint pad2:7; + } da16; + + struct + { + GLint src1_indirect_offset:10; + GLuint src1_subreg_nr:3; + GLuint src1_abs:1; + GLuint src1_negate:1; + GLuint src1_address_mode:1; + GLuint src1_horiz_stride:2; + GLuint src1_width:3; + GLuint src1_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad1:6; + } ia1; + + struct + { + GLuint src1_swz_x:2; + GLuint src1_swz_y:2; + GLint src1_indirect_offset:6; + GLuint src1_subreg_nr:3; + GLuint src1_abs:1; + GLuint src1_negate:1; + GLuint pad0:1; + GLuint src1_swz_z:2; + GLuint src1_swz_w:2; + GLuint pad1:1; + GLuint src1_vert_stride:4; + GLuint flag_reg_nr:1; + GLuint pad2:6; + } ia16; + + + struct + { + GLint jump_count:16; /* note: signed */ + GLuint pop_count:4; + GLuint pad0:12; + } if_else; + + struct { + GLuint function:4; + GLuint int_type:1; + GLuint precision:1; + GLuint saturate:1; + GLuint data_type:1; + GLuint pad0:8; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } math; + + struct { + GLuint function:4; + GLuint int_type:1; + GLuint precision:1; + GLuint saturate:1; + GLuint data_type:1; + GLuint snapshot:1; + GLuint pad0:10; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } math_igdng; + + struct { + GLuint binding_table_index:8; + GLuint sampler:4; + GLuint return_format:2; + GLuint msg_type:2; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } sampler; + + struct { + GLuint binding_table_index:8; + GLuint sampler:4; + GLuint msg_type:4; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } sampler_g4x; + + struct { + GLuint binding_table_index:8; + GLuint sampler:4; + GLuint msg_type:4; + GLuint simd_mode:2; + GLuint pad0:1; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } sampler_igdng; + + struct brw_urb_immediate urb; + + struct { + GLuint opcode:4; + GLuint offset:6; + GLuint swizzle_control:2; + GLuint pad:1; + GLuint allocate:1; + GLuint used:1; + GLuint complete:1; + GLuint pad0:3; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } urb_igdng; + + struct { + GLuint binding_table_index:8; + GLuint msg_control:4; + GLuint msg_type:2; + GLuint target_cache:2; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } dp_read; + + struct { + GLuint binding_table_index:8; + GLuint msg_control:3; + GLuint msg_type:3; + GLuint target_cache:2; + GLuint pad0:3; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } dp_read_igdng; + + struct { + GLuint binding_table_index:8; + GLuint msg_control:3; + GLuint pixel_scoreboard_clear:1; + GLuint msg_type:3; + GLuint send_commit_msg:1; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } dp_write; + + struct { + GLuint binding_table_index:8; + GLuint msg_control:3; + GLuint pixel_scoreboard_clear:1; + GLuint msg_type:3; + GLuint send_commit_msg:1; + GLuint pad0:3; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } dp_write_igdng; + + struct { + GLuint pad:16; + GLuint response_length:4; + GLuint msg_length:4; + GLuint msg_target:4; + GLuint pad1:3; + GLuint end_of_thread:1; + } generic; + + struct { + GLuint pad:19; + GLuint header_present:1; + GLuint response_length:5; + GLuint msg_length:4; + GLuint pad1:2; + GLuint end_of_thread:1; + } generic_igdng; + + GLint d; + GLuint ud; + float f; + } bits3; +}; + + +#endif diff --git a/src/gallium/drivers/i965/brw_structs_dump.c b/src/gallium/drivers/i965/brw_structs_dump.c new file mode 100644 index 00000000000..cd40fc6d618 --- /dev/null +++ b/src/gallium/drivers/i965/brw_structs_dump.c @@ -0,0 +1,1247 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * @file + * Dump i965 data structures. + * + * Generated automatically from brw_structs.h by brw_structs_dump.py. + */ + +#include "util/u_debug.h" + +#include "brw_types.h" +#include "brw_structs.h" +#include "brw_structs_dump.h" + +void +brw_dump_3d_control(const struct brw_3d_control *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.notify_enable = 0x%x\n", (*ptr).header.notify_enable); + debug_printf("\t\t.header.wc_flush_enable = 0x%x\n", (*ptr).header.wc_flush_enable); + debug_printf("\t\t.header.depth_stall_enable = 0x%x\n", (*ptr).header.depth_stall_enable); + debug_printf("\t\t.header.operation = 0x%x\n", (*ptr).header.operation); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.dest.dest_addr_type = 0x%x\n", (*ptr).dest.dest_addr_type); + debug_printf("\t\t.dest.dest_addr = 0x%x\n", (*ptr).dest.dest_addr); + debug_printf("\t\t.dword2 = 0x%x\n", (*ptr).dword2); + debug_printf("\t\t.dword3 = 0x%x\n", (*ptr).dword3); +} + +void +brw_dump_3d_primitive(const struct brw_3d_primitive *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.topology = 0x%x\n", (*ptr).header.topology); + debug_printf("\t\t.header.indexed = 0x%x\n", (*ptr).header.indexed); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.verts_per_instance = 0x%x\n", (*ptr).verts_per_instance); + debug_printf("\t\t.start_vert_location = 0x%x\n", (*ptr).start_vert_location); + debug_printf("\t\t.instance_count = 0x%x\n", (*ptr).instance_count); + debug_printf("\t\t.start_instance_location = 0x%x\n", (*ptr).start_instance_location); + debug_printf("\t\t.base_vert_location = 0x%x\n", (*ptr).base_vert_location); +} + +void +brw_dump_aa_line_parameters(const struct brw_aa_line_parameters *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.aa_coverage_scope = 0x%x\n", (*ptr).bits0.aa_coverage_scope); + debug_printf("\t\t.bits0.aa_coverage_bias = 0x%x\n", (*ptr).bits0.aa_coverage_bias); + debug_printf("\t\t.bits1.aa_coverage_endcap_slope = 0x%x\n", (*ptr).bits1.aa_coverage_endcap_slope); + debug_printf("\t\t.bits1.aa_coverage_endcap_bias = 0x%x\n", (*ptr).bits1.aa_coverage_endcap_bias); +} + +void +brw_dump_binding_table_pointers(const struct brw_binding_table_pointers *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.vs = 0x%x\n", (*ptr).vs); + debug_printf("\t\t.gs = 0x%x\n", (*ptr).gs); + debug_printf("\t\t.clp = 0x%x\n", (*ptr).clp); + debug_printf("\t\t.sf = 0x%x\n", (*ptr).sf); + debug_printf("\t\t.wm = 0x%x\n", (*ptr).wm); +} + +void +brw_dump_blend_constant_color(const struct brw_blend_constant_color *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.blend_constant_color[0] = %f\n", (*ptr).blend_constant_color[0]); + debug_printf("\t\t.blend_constant_color[1] = %f\n", (*ptr).blend_constant_color[1]); + debug_printf("\t\t.blend_constant_color[2] = %f\n", (*ptr).blend_constant_color[2]); + debug_printf("\t\t.blend_constant_color[3] = %f\n", (*ptr).blend_constant_color[3]); +} + +void +brw_dump_cc0(const struct brw_cc0 *ptr) +{ + debug_printf("\t\t.bf_stencil_pass_depth_pass_op = 0x%x\n", (*ptr).bf_stencil_pass_depth_pass_op); + debug_printf("\t\t.bf_stencil_pass_depth_fail_op = 0x%x\n", (*ptr).bf_stencil_pass_depth_fail_op); + debug_printf("\t\t.bf_stencil_fail_op = 0x%x\n", (*ptr).bf_stencil_fail_op); + debug_printf("\t\t.bf_stencil_func = 0x%x\n", (*ptr).bf_stencil_func); + debug_printf("\t\t.bf_stencil_enable = 0x%x\n", (*ptr).bf_stencil_enable); + debug_printf("\t\t.stencil_write_enable = 0x%x\n", (*ptr).stencil_write_enable); + debug_printf("\t\t.stencil_pass_depth_pass_op = 0x%x\n", (*ptr).stencil_pass_depth_pass_op); + debug_printf("\t\t.stencil_pass_depth_fail_op = 0x%x\n", (*ptr).stencil_pass_depth_fail_op); + debug_printf("\t\t.stencil_fail_op = 0x%x\n", (*ptr).stencil_fail_op); + debug_printf("\t\t.stencil_func = 0x%x\n", (*ptr).stencil_func); + debug_printf("\t\t.stencil_enable = 0x%x\n", (*ptr).stencil_enable); +} + +void +brw_dump_cc1(const struct brw_cc1 *ptr) +{ + debug_printf("\t\t.bf_stencil_ref = 0x%x\n", (*ptr).bf_stencil_ref); + debug_printf("\t\t.stencil_write_mask = 0x%x\n", (*ptr).stencil_write_mask); + debug_printf("\t\t.stencil_test_mask = 0x%x\n", (*ptr).stencil_test_mask); + debug_printf("\t\t.stencil_ref = 0x%x\n", (*ptr).stencil_ref); +} + +void +brw_dump_cc2(const struct brw_cc2 *ptr) +{ + debug_printf("\t\t.logicop_enable = 0x%x\n", (*ptr).logicop_enable); + debug_printf("\t\t.depth_write_enable = 0x%x\n", (*ptr).depth_write_enable); + debug_printf("\t\t.depth_test_function = 0x%x\n", (*ptr).depth_test_function); + debug_printf("\t\t.depth_test = 0x%x\n", (*ptr).depth_test); + debug_printf("\t\t.bf_stencil_write_mask = 0x%x\n", (*ptr).bf_stencil_write_mask); + debug_printf("\t\t.bf_stencil_test_mask = 0x%x\n", (*ptr).bf_stencil_test_mask); +} + +void +brw_dump_cc3(const struct brw_cc3 *ptr) +{ + debug_printf("\t\t.alpha_test_func = 0x%x\n", (*ptr).alpha_test_func); + debug_printf("\t\t.alpha_test = 0x%x\n", (*ptr).alpha_test); + debug_printf("\t\t.blend_enable = 0x%x\n", (*ptr).blend_enable); + debug_printf("\t\t.ia_blend_enable = 0x%x\n", (*ptr).ia_blend_enable); + debug_printf("\t\t.alpha_test_format = 0x%x\n", (*ptr).alpha_test_format); +} + +void +brw_dump_cc4(const struct brw_cc4 *ptr) +{ + debug_printf("\t\t.cc_viewport_state_offset = 0x%x\n", (*ptr).cc_viewport_state_offset); +} + +void +brw_dump_cc5(const struct brw_cc5 *ptr) +{ + debug_printf("\t\t.ia_dest_blend_factor = 0x%x\n", (*ptr).ia_dest_blend_factor); + debug_printf("\t\t.ia_src_blend_factor = 0x%x\n", (*ptr).ia_src_blend_factor); + debug_printf("\t\t.ia_blend_function = 0x%x\n", (*ptr).ia_blend_function); + debug_printf("\t\t.statistics_enable = 0x%x\n", (*ptr).statistics_enable); + debug_printf("\t\t.logicop_func = 0x%x\n", (*ptr).logicop_func); + debug_printf("\t\t.dither_enable = 0x%x\n", (*ptr).dither_enable); +} + +void +brw_dump_cc6(const struct brw_cc6 *ptr) +{ + debug_printf("\t\t.clamp_post_alpha_blend = 0x%x\n", (*ptr).clamp_post_alpha_blend); + debug_printf("\t\t.clamp_pre_alpha_blend = 0x%x\n", (*ptr).clamp_pre_alpha_blend); + debug_printf("\t\t.clamp_range = 0x%x\n", (*ptr).clamp_range); + debug_printf("\t\t.y_dither_offset = 0x%x\n", (*ptr).y_dither_offset); + debug_printf("\t\t.x_dither_offset = 0x%x\n", (*ptr).x_dither_offset); + debug_printf("\t\t.dest_blend_factor = 0x%x\n", (*ptr).dest_blend_factor); + debug_printf("\t\t.src_blend_factor = 0x%x\n", (*ptr).src_blend_factor); + debug_printf("\t\t.blend_function = 0x%x\n", (*ptr).blend_function); +} + +void +brw_dump_cc7(const struct brw_cc7 *ptr) +{ + debug_printf("\t\t.alpha_ref.f = %f\n", (*ptr).alpha_ref.f); + debug_printf("\t\t.alpha_ref.ub[0] = 0x%x\n", (*ptr).alpha_ref.ub[0]); + debug_printf("\t\t.alpha_ref.ub[1] = 0x%x\n", (*ptr).alpha_ref.ub[1]); + debug_printf("\t\t.alpha_ref.ub[2] = 0x%x\n", (*ptr).alpha_ref.ub[2]); + debug_printf("\t\t.alpha_ref.ub[3] = 0x%x\n", (*ptr).alpha_ref.ub[3]); +} + +void +brw_dump_cc_unit_state(const struct brw_cc_unit_state *ptr) +{ + debug_printf("\t\t.cc0.bf_stencil_pass_depth_pass_op = 0x%x\n", (*ptr).cc0.bf_stencil_pass_depth_pass_op); + debug_printf("\t\t.cc0.bf_stencil_pass_depth_fail_op = 0x%x\n", (*ptr).cc0.bf_stencil_pass_depth_fail_op); + debug_printf("\t\t.cc0.bf_stencil_fail_op = 0x%x\n", (*ptr).cc0.bf_stencil_fail_op); + debug_printf("\t\t.cc0.bf_stencil_func = 0x%x\n", (*ptr).cc0.bf_stencil_func); + debug_printf("\t\t.cc0.bf_stencil_enable = 0x%x\n", (*ptr).cc0.bf_stencil_enable); + debug_printf("\t\t.cc0.stencil_write_enable = 0x%x\n", (*ptr).cc0.stencil_write_enable); + debug_printf("\t\t.cc0.stencil_pass_depth_pass_op = 0x%x\n", (*ptr).cc0.stencil_pass_depth_pass_op); + debug_printf("\t\t.cc0.stencil_pass_depth_fail_op = 0x%x\n", (*ptr).cc0.stencil_pass_depth_fail_op); + debug_printf("\t\t.cc0.stencil_fail_op = 0x%x\n", (*ptr).cc0.stencil_fail_op); + debug_printf("\t\t.cc0.stencil_func = 0x%x\n", (*ptr).cc0.stencil_func); + debug_printf("\t\t.cc0.stencil_enable = 0x%x\n", (*ptr).cc0.stencil_enable); + debug_printf("\t\t.cc1.bf_stencil_ref = 0x%x\n", (*ptr).cc1.bf_stencil_ref); + debug_printf("\t\t.cc1.stencil_write_mask = 0x%x\n", (*ptr).cc1.stencil_write_mask); + debug_printf("\t\t.cc1.stencil_test_mask = 0x%x\n", (*ptr).cc1.stencil_test_mask); + debug_printf("\t\t.cc1.stencil_ref = 0x%x\n", (*ptr).cc1.stencil_ref); + debug_printf("\t\t.cc2.logicop_enable = 0x%x\n", (*ptr).cc2.logicop_enable); + debug_printf("\t\t.cc2.depth_write_enable = 0x%x\n", (*ptr).cc2.depth_write_enable); + debug_printf("\t\t.cc2.depth_test_function = 0x%x\n", (*ptr).cc2.depth_test_function); + debug_printf("\t\t.cc2.depth_test = 0x%x\n", (*ptr).cc2.depth_test); + debug_printf("\t\t.cc2.bf_stencil_write_mask = 0x%x\n", (*ptr).cc2.bf_stencil_write_mask); + debug_printf("\t\t.cc2.bf_stencil_test_mask = 0x%x\n", (*ptr).cc2.bf_stencil_test_mask); + debug_printf("\t\t.cc3.alpha_test_func = 0x%x\n", (*ptr).cc3.alpha_test_func); + debug_printf("\t\t.cc3.alpha_test = 0x%x\n", (*ptr).cc3.alpha_test); + debug_printf("\t\t.cc3.blend_enable = 0x%x\n", (*ptr).cc3.blend_enable); + debug_printf("\t\t.cc3.ia_blend_enable = 0x%x\n", (*ptr).cc3.ia_blend_enable); + debug_printf("\t\t.cc3.alpha_test_format = 0x%x\n", (*ptr).cc3.alpha_test_format); + debug_printf("\t\t.cc4.cc_viewport_state_offset = 0x%x\n", (*ptr).cc4.cc_viewport_state_offset); + debug_printf("\t\t.cc5.ia_dest_blend_factor = 0x%x\n", (*ptr).cc5.ia_dest_blend_factor); + debug_printf("\t\t.cc5.ia_src_blend_factor = 0x%x\n", (*ptr).cc5.ia_src_blend_factor); + debug_printf("\t\t.cc5.ia_blend_function = 0x%x\n", (*ptr).cc5.ia_blend_function); + debug_printf("\t\t.cc5.statistics_enable = 0x%x\n", (*ptr).cc5.statistics_enable); + debug_printf("\t\t.cc5.logicop_func = 0x%x\n", (*ptr).cc5.logicop_func); + debug_printf("\t\t.cc5.dither_enable = 0x%x\n", (*ptr).cc5.dither_enable); + debug_printf("\t\t.cc6.clamp_post_alpha_blend = 0x%x\n", (*ptr).cc6.clamp_post_alpha_blend); + debug_printf("\t\t.cc6.clamp_pre_alpha_blend = 0x%x\n", (*ptr).cc6.clamp_pre_alpha_blend); + debug_printf("\t\t.cc6.clamp_range = 0x%x\n", (*ptr).cc6.clamp_range); + debug_printf("\t\t.cc6.y_dither_offset = 0x%x\n", (*ptr).cc6.y_dither_offset); + debug_printf("\t\t.cc6.x_dither_offset = 0x%x\n", (*ptr).cc6.x_dither_offset); + debug_printf("\t\t.cc6.dest_blend_factor = 0x%x\n", (*ptr).cc6.dest_blend_factor); + debug_printf("\t\t.cc6.src_blend_factor = 0x%x\n", (*ptr).cc6.src_blend_factor); + debug_printf("\t\t.cc6.blend_function = 0x%x\n", (*ptr).cc6.blend_function); + debug_printf("\t\t.cc7.alpha_ref.f = %f\n", (*ptr).cc7.alpha_ref.f); + debug_printf("\t\t.cc7.alpha_ref.ub[0] = 0x%x\n", (*ptr).cc7.alpha_ref.ub[0]); + debug_printf("\t\t.cc7.alpha_ref.ub[1] = 0x%x\n", (*ptr).cc7.alpha_ref.ub[1]); + debug_printf("\t\t.cc7.alpha_ref.ub[2] = 0x%x\n", (*ptr).cc7.alpha_ref.ub[2]); + debug_printf("\t\t.cc7.alpha_ref.ub[3] = 0x%x\n", (*ptr).cc7.alpha_ref.ub[3]); +} + +void +brw_dump_cc_viewport(const struct brw_cc_viewport *ptr) +{ + debug_printf("\t\t.min_depth = %f\n", (*ptr).min_depth); + debug_printf("\t\t.max_depth = %f\n", (*ptr).max_depth); +} + +void +brw_dump_clip_unit_state(const struct brw_clip_unit_state *ptr) +{ + debug_printf("\t\t.thread0.grf_reg_count = 0x%x\n", (*ptr).thread0.grf_reg_count); + debug_printf("\t\t.thread0.kernel_start_pointer = 0x%x\n", (*ptr).thread0.kernel_start_pointer); + debug_printf("\t\t.thread1.sw_exception_enable = 0x%x\n", (*ptr).thread1.sw_exception_enable); + debug_printf("\t\t.thread1.mask_stack_exception_enable = 0x%x\n", (*ptr).thread1.mask_stack_exception_enable); + debug_printf("\t\t.thread1.illegal_op_exception_enable = 0x%x\n", (*ptr).thread1.illegal_op_exception_enable); + debug_printf("\t\t.thread1.floating_point_mode = 0x%x\n", (*ptr).thread1.floating_point_mode); + debug_printf("\t\t.thread1.thread_priority = 0x%x\n", (*ptr).thread1.thread_priority); + debug_printf("\t\t.thread1.binding_table_entry_count = 0x%x\n", (*ptr).thread1.binding_table_entry_count); + debug_printf("\t\t.thread1.single_program_flow = 0x%x\n", (*ptr).thread1.single_program_flow); + debug_printf("\t\t.thread2.per_thread_scratch_space = 0x%x\n", (*ptr).thread2.per_thread_scratch_space); + debug_printf("\t\t.thread2.scratch_space_base_pointer = 0x%x\n", (*ptr).thread2.scratch_space_base_pointer); + debug_printf("\t\t.thread3.dispatch_grf_start_reg = 0x%x\n", (*ptr).thread3.dispatch_grf_start_reg); + debug_printf("\t\t.thread3.urb_entry_read_offset = 0x%x\n", (*ptr).thread3.urb_entry_read_offset); + debug_printf("\t\t.thread3.urb_entry_read_length = 0x%x\n", (*ptr).thread3.urb_entry_read_length); + debug_printf("\t\t.thread3.const_urb_entry_read_offset = 0x%x\n", (*ptr).thread3.const_urb_entry_read_offset); + debug_printf("\t\t.thread3.const_urb_entry_read_length = 0x%x\n", (*ptr).thread3.const_urb_entry_read_length); + debug_printf("\t\t.thread4.gs_output_stats = 0x%x\n", (*ptr).thread4.gs_output_stats); + debug_printf("\t\t.thread4.stats_enable = 0x%x\n", (*ptr).thread4.stats_enable); + debug_printf("\t\t.thread4.nr_urb_entries = 0x%x\n", (*ptr).thread4.nr_urb_entries); + debug_printf("\t\t.thread4.urb_entry_allocation_size = 0x%x\n", (*ptr).thread4.urb_entry_allocation_size); + debug_printf("\t\t.thread4.max_threads = 0x%x\n", (*ptr).thread4.max_threads); + debug_printf("\t\t.clip5.clip_mode = 0x%x\n", (*ptr).clip5.clip_mode); + debug_printf("\t\t.clip5.userclip_enable_flags = 0x%x\n", (*ptr).clip5.userclip_enable_flags); + debug_printf("\t\t.clip5.userclip_must_clip = 0x%x\n", (*ptr).clip5.userclip_must_clip); + debug_printf("\t\t.clip5.negative_w_clip_test = 0x%x\n", (*ptr).clip5.negative_w_clip_test); + debug_printf("\t\t.clip5.guard_band_enable = 0x%x\n", (*ptr).clip5.guard_band_enable); + debug_printf("\t\t.clip5.viewport_z_clip_enable = 0x%x\n", (*ptr).clip5.viewport_z_clip_enable); + debug_printf("\t\t.clip5.viewport_xy_clip_enable = 0x%x\n", (*ptr).clip5.viewport_xy_clip_enable); + debug_printf("\t\t.clip5.vertex_position_space = 0x%x\n", (*ptr).clip5.vertex_position_space); + debug_printf("\t\t.clip5.api_mode = 0x%x\n", (*ptr).clip5.api_mode); + debug_printf("\t\t.clip6.clipper_viewport_state_ptr = 0x%x\n", (*ptr).clip6.clipper_viewport_state_ptr); + debug_printf("\t\t.viewport_xmin = %f\n", (*ptr).viewport_xmin); + debug_printf("\t\t.viewport_xmax = %f\n", (*ptr).viewport_xmax); + debug_printf("\t\t.viewport_ymin = %f\n", (*ptr).viewport_ymin); + debug_printf("\t\t.viewport_ymax = %f\n", (*ptr).viewport_ymax); +} + +void +brw_dump_clipper_viewport(const struct brw_clipper_viewport *ptr) +{ + debug_printf("\t\t.xmin = %f\n", (*ptr).xmin); + debug_printf("\t\t.xmax = %f\n", (*ptr).xmax); + debug_printf("\t\t.ymin = %f\n", (*ptr).ymin); + debug_printf("\t\t.ymax = %f\n", (*ptr).ymax); +} + +void +brw_dump_constant_buffer(const struct brw_constant_buffer *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.valid = 0x%x\n", (*ptr).header.valid); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.buffer_length = 0x%x\n", (*ptr).bits0.buffer_length); + debug_printf("\t\t.bits0.buffer_address = 0x%x\n", (*ptr).bits0.buffer_address); +} + +void +brw_dump_cs_urb_state(const struct brw_cs_urb_state *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.nr_urb_entries = 0x%x\n", (*ptr).bits0.nr_urb_entries); + debug_printf("\t\t.bits0.urb_entry_size = 0x%x\n", (*ptr).bits0.urb_entry_size); +} + +void +brw_dump_depthbuffer(const struct brw_depthbuffer *ptr) +{ + debug_printf("\t\t.header.bits.length = 0x%x\n", (*ptr).header.bits.length); + debug_printf("\t\t.header.bits.opcode = 0x%x\n", (*ptr).header.bits.opcode); + debug_printf("\t\t.dword1.bits.pitch = 0x%x\n", (*ptr).dword1.bits.pitch); + debug_printf("\t\t.dword1.bits.format = 0x%x\n", (*ptr).dword1.bits.format); + debug_printf("\t\t.dword1.bits.software_tiled_rendering_mode = 0x%x\n", (*ptr).dword1.bits.software_tiled_rendering_mode); + debug_printf("\t\t.dword1.bits.depth_offset_disable = 0x%x\n", (*ptr).dword1.bits.depth_offset_disable); + debug_printf("\t\t.dword1.bits.tile_walk = 0x%x\n", (*ptr).dword1.bits.tile_walk); + debug_printf("\t\t.dword1.bits.tiled_surface = 0x%x\n", (*ptr).dword1.bits.tiled_surface); + debug_printf("\t\t.dword1.bits.surface_type = 0x%x\n", (*ptr).dword1.bits.surface_type); + debug_printf("\t\t.dword2_base_addr = 0x%x\n", (*ptr).dword2_base_addr); + debug_printf("\t\t.dword3.bits.mipmap_layout = 0x%x\n", (*ptr).dword3.bits.mipmap_layout); + debug_printf("\t\t.dword3.bits.lod = 0x%x\n", (*ptr).dword3.bits.lod); + debug_printf("\t\t.dword3.bits.width = 0x%x\n", (*ptr).dword3.bits.width); + debug_printf("\t\t.dword3.bits.height = 0x%x\n", (*ptr).dword3.bits.height); + debug_printf("\t\t.dword4.bits.min_array_element = 0x%x\n", (*ptr).dword4.bits.min_array_element); + debug_printf("\t\t.dword4.bits.depth = 0x%x\n", (*ptr).dword4.bits.depth); +} + +void +brw_dump_depthbuffer_g4x(const struct brw_depthbuffer_g4x *ptr) +{ + debug_printf("\t\t.header.bits.length = 0x%x\n", (*ptr).header.bits.length); + debug_printf("\t\t.header.bits.opcode = 0x%x\n", (*ptr).header.bits.opcode); + debug_printf("\t\t.dword1.bits.pitch = 0x%x\n", (*ptr).dword1.bits.pitch); + debug_printf("\t\t.dword1.bits.format = 0x%x\n", (*ptr).dword1.bits.format); + debug_printf("\t\t.dword1.bits.software_tiled_rendering_mode = 0x%x\n", (*ptr).dword1.bits.software_tiled_rendering_mode); + debug_printf("\t\t.dword1.bits.depth_offset_disable = 0x%x\n", (*ptr).dword1.bits.depth_offset_disable); + debug_printf("\t\t.dword1.bits.tile_walk = 0x%x\n", (*ptr).dword1.bits.tile_walk); + debug_printf("\t\t.dword1.bits.tiled_surface = 0x%x\n", (*ptr).dword1.bits.tiled_surface); + debug_printf("\t\t.dword1.bits.surface_type = 0x%x\n", (*ptr).dword1.bits.surface_type); + debug_printf("\t\t.dword2_base_addr = 0x%x\n", (*ptr).dword2_base_addr); + debug_printf("\t\t.dword3.bits.mipmap_layout = 0x%x\n", (*ptr).dword3.bits.mipmap_layout); + debug_printf("\t\t.dword3.bits.lod = 0x%x\n", (*ptr).dword3.bits.lod); + debug_printf("\t\t.dword3.bits.width = 0x%x\n", (*ptr).dword3.bits.width); + debug_printf("\t\t.dword3.bits.height = 0x%x\n", (*ptr).dword3.bits.height); + debug_printf("\t\t.dword4.bits.min_array_element = 0x%x\n", (*ptr).dword4.bits.min_array_element); + debug_printf("\t\t.dword4.bits.depth = 0x%x\n", (*ptr).dword4.bits.depth); + debug_printf("\t\t.dword5.bits.xoffset = 0x%x\n", (*ptr).dword5.bits.xoffset); + debug_printf("\t\t.dword5.bits.yoffset = 0x%x\n", (*ptr).dword5.bits.yoffset); +} + +void +brw_dump_drawrect(const struct brw_drawrect *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.xmin = 0x%x\n", (*ptr).xmin); + debug_printf("\t\t.ymin = 0x%x\n", (*ptr).ymin); + debug_printf("\t\t.xmax = 0x%x\n", (*ptr).xmax); + debug_printf("\t\t.ymax = 0x%x\n", (*ptr).ymax); + debug_printf("\t\t.xorg = 0x%x\n", (*ptr).xorg); + debug_printf("\t\t.yorg = 0x%x\n", (*ptr).yorg); +} + +void +brw_dump_global_depth_offset_clamp(const struct brw_global_depth_offset_clamp *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.depth_offset_clamp = %f\n", (*ptr).depth_offset_clamp); +} + +void +brw_dump_gs_unit_state(const struct brw_gs_unit_state *ptr) +{ + debug_printf("\t\t.thread0.grf_reg_count = 0x%x\n", (*ptr).thread0.grf_reg_count); + debug_printf("\t\t.thread0.kernel_start_pointer = 0x%x\n", (*ptr).thread0.kernel_start_pointer); + debug_printf("\t\t.thread1.ext_halt_exception_enable = 0x%x\n", (*ptr).thread1.ext_halt_exception_enable); + debug_printf("\t\t.thread1.sw_exception_enable = 0x%x\n", (*ptr).thread1.sw_exception_enable); + debug_printf("\t\t.thread1.mask_stack_exception_enable = 0x%x\n", (*ptr).thread1.mask_stack_exception_enable); + debug_printf("\t\t.thread1.timeout_exception_enable = 0x%x\n", (*ptr).thread1.timeout_exception_enable); + debug_printf("\t\t.thread1.illegal_op_exception_enable = 0x%x\n", (*ptr).thread1.illegal_op_exception_enable); + debug_printf("\t\t.thread1.depth_coef_urb_read_offset = 0x%x\n", (*ptr).thread1.depth_coef_urb_read_offset); + debug_printf("\t\t.thread1.floating_point_mode = 0x%x\n", (*ptr).thread1.floating_point_mode); + debug_printf("\t\t.thread1.thread_priority = 0x%x\n", (*ptr).thread1.thread_priority); + debug_printf("\t\t.thread1.binding_table_entry_count = 0x%x\n", (*ptr).thread1.binding_table_entry_count); + debug_printf("\t\t.thread1.single_program_flow = 0x%x\n", (*ptr).thread1.single_program_flow); + debug_printf("\t\t.thread2.per_thread_scratch_space = 0x%x\n", (*ptr).thread2.per_thread_scratch_space); + debug_printf("\t\t.thread2.scratch_space_base_pointer = 0x%x\n", (*ptr).thread2.scratch_space_base_pointer); + debug_printf("\t\t.thread3.dispatch_grf_start_reg = 0x%x\n", (*ptr).thread3.dispatch_grf_start_reg); + debug_printf("\t\t.thread3.urb_entry_read_offset = 0x%x\n", (*ptr).thread3.urb_entry_read_offset); + debug_printf("\t\t.thread3.urb_entry_read_length = 0x%x\n", (*ptr).thread3.urb_entry_read_length); + debug_printf("\t\t.thread3.const_urb_entry_read_offset = 0x%x\n", (*ptr).thread3.const_urb_entry_read_offset); + debug_printf("\t\t.thread3.const_urb_entry_read_length = 0x%x\n", (*ptr).thread3.const_urb_entry_read_length); + debug_printf("\t\t.thread4.rendering_enable = 0x%x\n", (*ptr).thread4.rendering_enable); + debug_printf("\t\t.thread4.stats_enable = 0x%x\n", (*ptr).thread4.stats_enable); + debug_printf("\t\t.thread4.nr_urb_entries = 0x%x\n", (*ptr).thread4.nr_urb_entries); + debug_printf("\t\t.thread4.urb_entry_allocation_size = 0x%x\n", (*ptr).thread4.urb_entry_allocation_size); + debug_printf("\t\t.thread4.max_threads = 0x%x\n", (*ptr).thread4.max_threads); + debug_printf("\t\t.gs5.sampler_count = 0x%x\n", (*ptr).gs5.sampler_count); + debug_printf("\t\t.gs5.sampler_state_pointer = 0x%x\n", (*ptr).gs5.sampler_state_pointer); + debug_printf("\t\t.gs6.max_vp_index = 0x%x\n", (*ptr).gs6.max_vp_index); + debug_printf("\t\t.gs6.svbi_post_inc_value = 0x%x\n", (*ptr).gs6.svbi_post_inc_value); + debug_printf("\t\t.gs6.svbi_post_inc_enable = 0x%x\n", (*ptr).gs6.svbi_post_inc_enable); + debug_printf("\t\t.gs6.svbi_payload = 0x%x\n", (*ptr).gs6.svbi_payload); + debug_printf("\t\t.gs6.discard_adjaceny = 0x%x\n", (*ptr).gs6.discard_adjaceny); + debug_printf("\t\t.gs6.reorder_enable = 0x%x\n", (*ptr).gs6.reorder_enable); +} + +void +brw_dump_indexbuffer(const struct brw_indexbuffer *ptr) +{ + debug_printf("\t\t.header.bits.length = 0x%x\n", (*ptr).header.bits.length); + debug_printf("\t\t.header.bits.index_format = 0x%x\n", (*ptr).header.bits.index_format); + debug_printf("\t\t.header.bits.cut_index_enable = 0x%x\n", (*ptr).header.bits.cut_index_enable); + debug_printf("\t\t.header.bits.opcode = 0x%x\n", (*ptr).header.bits.opcode); + debug_printf("\t\t.buffer_start = 0x%x\n", (*ptr).buffer_start); + debug_printf("\t\t.buffer_end = 0x%x\n", (*ptr).buffer_end); +} + +void +brw_dump_line_stipple(const struct brw_line_stipple *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.pattern = 0x%x\n", (*ptr).bits0.pattern); + debug_printf("\t\t.bits1.repeat_count = 0x%x\n", (*ptr).bits1.repeat_count); + debug_printf("\t\t.bits1.inverse_repeat_count = 0x%x\n", (*ptr).bits1.inverse_repeat_count); +} + +void +brw_dump_mi_flush(const struct brw_mi_flush *ptr) +{ + debug_printf("\t\t.flags = 0x%x\n", (*ptr).flags); + debug_printf("\t\t.opcode = 0x%x\n", (*ptr).opcode); +} + +void +brw_dump_pipe_control(const struct brw_pipe_control *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.notify_enable = 0x%x\n", (*ptr).header.notify_enable); + debug_printf("\t\t.header.texture_cache_flush_enable = 0x%x\n", (*ptr).header.texture_cache_flush_enable); + debug_printf("\t\t.header.indirect_state_pointers_disable = 0x%x\n", (*ptr).header.indirect_state_pointers_disable); + debug_printf("\t\t.header.instruction_state_cache_flush_enable = 0x%x\n", (*ptr).header.instruction_state_cache_flush_enable); + debug_printf("\t\t.header.write_cache_flush_enable = 0x%x\n", (*ptr).header.write_cache_flush_enable); + debug_printf("\t\t.header.depth_stall_enable = 0x%x\n", (*ptr).header.depth_stall_enable); + debug_printf("\t\t.header.post_sync_operation = 0x%x\n", (*ptr).header.post_sync_operation); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits1.dest_addr_type = 0x%x\n", (*ptr).bits1.dest_addr_type); + debug_printf("\t\t.bits1.dest_addr = 0x%x\n", (*ptr).bits1.dest_addr); + debug_printf("\t\t.data0 = 0x%x\n", (*ptr).data0); + debug_printf("\t\t.data1 = 0x%x\n", (*ptr).data1); +} + +void +brw_dump_pipeline_select(const struct brw_pipeline_select *ptr) +{ + debug_printf("\t\t.header.pipeline_select = 0x%x\n", (*ptr).header.pipeline_select); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); +} + +void +brw_dump_pipelined_state_pointers(const struct brw_pipelined_state_pointers *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.vs.offset = 0x%x\n", (*ptr).vs.offset); + debug_printf("\t\t.gs.enable = 0x%x\n", (*ptr).gs.enable); + debug_printf("\t\t.gs.offset = 0x%x\n", (*ptr).gs.offset); + debug_printf("\t\t.clp.enable = 0x%x\n", (*ptr).clp.enable); + debug_printf("\t\t.clp.offset = 0x%x\n", (*ptr).clp.offset); + debug_printf("\t\t.sf.offset = 0x%x\n", (*ptr).sf.offset); + debug_printf("\t\t.wm.offset = 0x%x\n", (*ptr).wm.offset); + debug_printf("\t\t.cc.offset = 0x%x\n", (*ptr).cc.offset); +} + +void +brw_dump_polygon_stipple(const struct brw_polygon_stipple *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.stipple[0] = 0x%x\n", (*ptr).stipple[0]); + debug_printf("\t\t.stipple[1] = 0x%x\n", (*ptr).stipple[1]); + debug_printf("\t\t.stipple[2] = 0x%x\n", (*ptr).stipple[2]); + debug_printf("\t\t.stipple[3] = 0x%x\n", (*ptr).stipple[3]); + debug_printf("\t\t.stipple[4] = 0x%x\n", (*ptr).stipple[4]); + debug_printf("\t\t.stipple[5] = 0x%x\n", (*ptr).stipple[5]); + debug_printf("\t\t.stipple[6] = 0x%x\n", (*ptr).stipple[6]); + debug_printf("\t\t.stipple[7] = 0x%x\n", (*ptr).stipple[7]); + debug_printf("\t\t.stipple[8] = 0x%x\n", (*ptr).stipple[8]); + debug_printf("\t\t.stipple[9] = 0x%x\n", (*ptr).stipple[9]); + debug_printf("\t\t.stipple[10] = 0x%x\n", (*ptr).stipple[10]); + debug_printf("\t\t.stipple[11] = 0x%x\n", (*ptr).stipple[11]); + debug_printf("\t\t.stipple[12] = 0x%x\n", (*ptr).stipple[12]); + debug_printf("\t\t.stipple[13] = 0x%x\n", (*ptr).stipple[13]); + debug_printf("\t\t.stipple[14] = 0x%x\n", (*ptr).stipple[14]); + debug_printf("\t\t.stipple[15] = 0x%x\n", (*ptr).stipple[15]); + debug_printf("\t\t.stipple[16] = 0x%x\n", (*ptr).stipple[16]); + debug_printf("\t\t.stipple[17] = 0x%x\n", (*ptr).stipple[17]); + debug_printf("\t\t.stipple[18] = 0x%x\n", (*ptr).stipple[18]); + debug_printf("\t\t.stipple[19] = 0x%x\n", (*ptr).stipple[19]); + debug_printf("\t\t.stipple[20] = 0x%x\n", (*ptr).stipple[20]); + debug_printf("\t\t.stipple[21] = 0x%x\n", (*ptr).stipple[21]); + debug_printf("\t\t.stipple[22] = 0x%x\n", (*ptr).stipple[22]); + debug_printf("\t\t.stipple[23] = 0x%x\n", (*ptr).stipple[23]); + debug_printf("\t\t.stipple[24] = 0x%x\n", (*ptr).stipple[24]); + debug_printf("\t\t.stipple[25] = 0x%x\n", (*ptr).stipple[25]); + debug_printf("\t\t.stipple[26] = 0x%x\n", (*ptr).stipple[26]); + debug_printf("\t\t.stipple[27] = 0x%x\n", (*ptr).stipple[27]); + debug_printf("\t\t.stipple[28] = 0x%x\n", (*ptr).stipple[28]); + debug_printf("\t\t.stipple[29] = 0x%x\n", (*ptr).stipple[29]); + debug_printf("\t\t.stipple[30] = 0x%x\n", (*ptr).stipple[30]); + debug_printf("\t\t.stipple[31] = 0x%x\n", (*ptr).stipple[31]); +} + +void +brw_dump_polygon_stipple_offset(const struct brw_polygon_stipple_offset *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.y_offset = 0x%x\n", (*ptr).bits0.y_offset); + debug_printf("\t\t.bits0.x_offset = 0x%x\n", (*ptr).bits0.x_offset); +} + +void +brw_dump_sampler_default_color(const struct brw_sampler_default_color *ptr) +{ + debug_printf("\t\t.color[0] = %f\n", (*ptr).color[0]); + debug_printf("\t\t.color[1] = %f\n", (*ptr).color[1]); + debug_printf("\t\t.color[2] = %f\n", (*ptr).color[2]); + debug_printf("\t\t.color[3] = %f\n", (*ptr).color[3]); +} + +void +brw_dump_sampler_state(const struct brw_sampler_state *ptr) +{ + debug_printf("\t\t.ss0.shadow_function = 0x%x\n", (*ptr).ss0.shadow_function); + debug_printf("\t\t.ss0.lod_bias = 0x%x\n", (*ptr).ss0.lod_bias); + debug_printf("\t\t.ss0.min_filter = 0x%x\n", (*ptr).ss0.min_filter); + debug_printf("\t\t.ss0.mag_filter = 0x%x\n", (*ptr).ss0.mag_filter); + debug_printf("\t\t.ss0.mip_filter = 0x%x\n", (*ptr).ss0.mip_filter); + debug_printf("\t\t.ss0.base_level = 0x%x\n", (*ptr).ss0.base_level); + debug_printf("\t\t.ss0.lod_preclamp = 0x%x\n", (*ptr).ss0.lod_preclamp); + debug_printf("\t\t.ss0.default_color_mode = 0x%x\n", (*ptr).ss0.default_color_mode); + debug_printf("\t\t.ss0.disable = 0x%x\n", (*ptr).ss0.disable); + debug_printf("\t\t.ss1.r_wrap_mode = 0x%x\n", (*ptr).ss1.r_wrap_mode); + debug_printf("\t\t.ss1.t_wrap_mode = 0x%x\n", (*ptr).ss1.t_wrap_mode); + debug_printf("\t\t.ss1.s_wrap_mode = 0x%x\n", (*ptr).ss1.s_wrap_mode); + debug_printf("\t\t.ss1.max_lod = 0x%x\n", (*ptr).ss1.max_lod); + debug_printf("\t\t.ss1.min_lod = 0x%x\n", (*ptr).ss1.min_lod); + debug_printf("\t\t.ss2.default_color_pointer = 0x%x\n", (*ptr).ss2.default_color_pointer); + debug_printf("\t\t.ss3.max_aniso = 0x%x\n", (*ptr).ss3.max_aniso); + debug_printf("\t\t.ss3.chroma_key_mode = 0x%x\n", (*ptr).ss3.chroma_key_mode); + debug_printf("\t\t.ss3.chroma_key_index = 0x%x\n", (*ptr).ss3.chroma_key_index); + debug_printf("\t\t.ss3.chroma_key_enable = 0x%x\n", (*ptr).ss3.chroma_key_enable); + debug_printf("\t\t.ss3.monochrome_filter_width = 0x%x\n", (*ptr).ss3.monochrome_filter_width); + debug_printf("\t\t.ss3.monochrome_filter_height = 0x%x\n", (*ptr).ss3.monochrome_filter_height); +} + +void +brw_dump_sf_unit_state(const struct brw_sf_unit_state *ptr) +{ + debug_printf("\t\t.thread0.grf_reg_count = 0x%x\n", (*ptr).thread0.grf_reg_count); + debug_printf("\t\t.thread0.kernel_start_pointer = 0x%x\n", (*ptr).thread0.kernel_start_pointer); + debug_printf("\t\t.thread1.ext_halt_exception_enable = 0x%x\n", (*ptr).thread1.ext_halt_exception_enable); + debug_printf("\t\t.thread1.sw_exception_enable = 0x%x\n", (*ptr).thread1.sw_exception_enable); + debug_printf("\t\t.thread1.mask_stack_exception_enable = 0x%x\n", (*ptr).thread1.mask_stack_exception_enable); + debug_printf("\t\t.thread1.timeout_exception_enable = 0x%x\n", (*ptr).thread1.timeout_exception_enable); + debug_printf("\t\t.thread1.illegal_op_exception_enable = 0x%x\n", (*ptr).thread1.illegal_op_exception_enable); + debug_printf("\t\t.thread1.depth_coef_urb_read_offset = 0x%x\n", (*ptr).thread1.depth_coef_urb_read_offset); + debug_printf("\t\t.thread1.floating_point_mode = 0x%x\n", (*ptr).thread1.floating_point_mode); + debug_printf("\t\t.thread1.thread_priority = 0x%x\n", (*ptr).thread1.thread_priority); + debug_printf("\t\t.thread1.binding_table_entry_count = 0x%x\n", (*ptr).thread1.binding_table_entry_count); + debug_printf("\t\t.thread1.single_program_flow = 0x%x\n", (*ptr).thread1.single_program_flow); + debug_printf("\t\t.thread2.per_thread_scratch_space = 0x%x\n", (*ptr).thread2.per_thread_scratch_space); + debug_printf("\t\t.thread2.scratch_space_base_pointer = 0x%x\n", (*ptr).thread2.scratch_space_base_pointer); + debug_printf("\t\t.thread3.dispatch_grf_start_reg = 0x%x\n", (*ptr).thread3.dispatch_grf_start_reg); + debug_printf("\t\t.thread3.urb_entry_read_offset = 0x%x\n", (*ptr).thread3.urb_entry_read_offset); + debug_printf("\t\t.thread3.urb_entry_read_length = 0x%x\n", (*ptr).thread3.urb_entry_read_length); + debug_printf("\t\t.thread3.const_urb_entry_read_offset = 0x%x\n", (*ptr).thread3.const_urb_entry_read_offset); + debug_printf("\t\t.thread3.const_urb_entry_read_length = 0x%x\n", (*ptr).thread3.const_urb_entry_read_length); + debug_printf("\t\t.thread4.stats_enable = 0x%x\n", (*ptr).thread4.stats_enable); + debug_printf("\t\t.thread4.nr_urb_entries = 0x%x\n", (*ptr).thread4.nr_urb_entries); + debug_printf("\t\t.thread4.urb_entry_allocation_size = 0x%x\n", (*ptr).thread4.urb_entry_allocation_size); + debug_printf("\t\t.thread4.max_threads = 0x%x\n", (*ptr).thread4.max_threads); + debug_printf("\t\t.sf5.front_winding = 0x%x\n", (*ptr).sf5.front_winding); + debug_printf("\t\t.sf5.viewport_transform = 0x%x\n", (*ptr).sf5.viewport_transform); + debug_printf("\t\t.sf5.sf_viewport_state_offset = 0x%x\n", (*ptr).sf5.sf_viewport_state_offset); + debug_printf("\t\t.sf6.dest_org_vbias = 0x%x\n", (*ptr).sf6.dest_org_vbias); + debug_printf("\t\t.sf6.dest_org_hbias = 0x%x\n", (*ptr).sf6.dest_org_hbias); + debug_printf("\t\t.sf6.scissor = 0x%x\n", (*ptr).sf6.scissor); + debug_printf("\t\t.sf6.disable_2x2_trifilter = 0x%x\n", (*ptr).sf6.disable_2x2_trifilter); + debug_printf("\t\t.sf6.disable_zero_pix_trifilter = 0x%x\n", (*ptr).sf6.disable_zero_pix_trifilter); + debug_printf("\t\t.sf6.point_rast_rule = 0x%x\n", (*ptr).sf6.point_rast_rule); + debug_printf("\t\t.sf6.line_endcap_aa_region_width = 0x%x\n", (*ptr).sf6.line_endcap_aa_region_width); + debug_printf("\t\t.sf6.line_width = 0x%x\n", (*ptr).sf6.line_width); + debug_printf("\t\t.sf6.fast_scissor_disable = 0x%x\n", (*ptr).sf6.fast_scissor_disable); + debug_printf("\t\t.sf6.cull_mode = 0x%x\n", (*ptr).sf6.cull_mode); + debug_printf("\t\t.sf6.aa_enable = 0x%x\n", (*ptr).sf6.aa_enable); + debug_printf("\t\t.sf7.point_size = 0x%x\n", (*ptr).sf7.point_size); + debug_printf("\t\t.sf7.use_point_size_state = 0x%x\n", (*ptr).sf7.use_point_size_state); + debug_printf("\t\t.sf7.subpixel_precision = 0x%x\n", (*ptr).sf7.subpixel_precision); + debug_printf("\t\t.sf7.sprite_point = 0x%x\n", (*ptr).sf7.sprite_point); + debug_printf("\t\t.sf7.aa_line_distance_mode = 0x%x\n", (*ptr).sf7.aa_line_distance_mode); + debug_printf("\t\t.sf7.trifan_pv = 0x%x\n", (*ptr).sf7.trifan_pv); + debug_printf("\t\t.sf7.linestrip_pv = 0x%x\n", (*ptr).sf7.linestrip_pv); + debug_printf("\t\t.sf7.tristrip_pv = 0x%x\n", (*ptr).sf7.tristrip_pv); + debug_printf("\t\t.sf7.line_last_pixel_enable = 0x%x\n", (*ptr).sf7.line_last_pixel_enable); +} + +void +brw_dump_sf_viewport(const struct brw_sf_viewport *ptr) +{ + debug_printf("\t\t.viewport.m00 = %f\n", (*ptr).viewport.m00); + debug_printf("\t\t.viewport.m11 = %f\n", (*ptr).viewport.m11); + debug_printf("\t\t.viewport.m22 = %f\n", (*ptr).viewport.m22); + debug_printf("\t\t.viewport.m30 = %f\n", (*ptr).viewport.m30); + debug_printf("\t\t.viewport.m31 = %f\n", (*ptr).viewport.m31); + debug_printf("\t\t.viewport.m32 = %f\n", (*ptr).viewport.m32); + debug_printf("\t\t.scissor.xmin = 0x%x\n", (*ptr).scissor.xmin); + debug_printf("\t\t.scissor.ymin = 0x%x\n", (*ptr).scissor.ymin); + debug_printf("\t\t.scissor.xmax = 0x%x\n", (*ptr).scissor.xmax); + debug_printf("\t\t.scissor.ymax = 0x%x\n", (*ptr).scissor.ymax); +} + +void +brw_dump_ss0(const struct brw_ss0 *ptr) +{ + debug_printf("\t\t.shadow_function = 0x%x\n", (*ptr).shadow_function); + debug_printf("\t\t.lod_bias = 0x%x\n", (*ptr).lod_bias); + debug_printf("\t\t.min_filter = 0x%x\n", (*ptr).min_filter); + debug_printf("\t\t.mag_filter = 0x%x\n", (*ptr).mag_filter); + debug_printf("\t\t.mip_filter = 0x%x\n", (*ptr).mip_filter); + debug_printf("\t\t.base_level = 0x%x\n", (*ptr).base_level); + debug_printf("\t\t.lod_preclamp = 0x%x\n", (*ptr).lod_preclamp); + debug_printf("\t\t.default_color_mode = 0x%x\n", (*ptr).default_color_mode); + debug_printf("\t\t.disable = 0x%x\n", (*ptr).disable); +} + +void +brw_dump_ss1(const struct brw_ss1 *ptr) +{ + debug_printf("\t\t.r_wrap_mode = 0x%x\n", (*ptr).r_wrap_mode); + debug_printf("\t\t.t_wrap_mode = 0x%x\n", (*ptr).t_wrap_mode); + debug_printf("\t\t.s_wrap_mode = 0x%x\n", (*ptr).s_wrap_mode); + debug_printf("\t\t.max_lod = 0x%x\n", (*ptr).max_lod); + debug_printf("\t\t.min_lod = 0x%x\n", (*ptr).min_lod); +} + +void +brw_dump_ss2(const struct brw_ss2 *ptr) +{ + debug_printf("\t\t.default_color_pointer = 0x%x\n", (*ptr).default_color_pointer); +} + +void +brw_dump_ss3(const struct brw_ss3 *ptr) +{ + debug_printf("\t\t.max_aniso = 0x%x\n", (*ptr).max_aniso); + debug_printf("\t\t.chroma_key_mode = 0x%x\n", (*ptr).chroma_key_mode); + debug_printf("\t\t.chroma_key_index = 0x%x\n", (*ptr).chroma_key_index); + debug_printf("\t\t.chroma_key_enable = 0x%x\n", (*ptr).chroma_key_enable); + debug_printf("\t\t.monochrome_filter_width = 0x%x\n", (*ptr).monochrome_filter_width); + debug_printf("\t\t.monochrome_filter_height = 0x%x\n", (*ptr).monochrome_filter_height); +} + +void +brw_dump_state_base_address(const struct brw_state_base_address *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.modify_enable = 0x%x\n", (*ptr).bits0.modify_enable); + debug_printf("\t\t.bits0.general_state_address = 0x%x\n", (*ptr).bits0.general_state_address); + debug_printf("\t\t.bits1.modify_enable = 0x%x\n", (*ptr).bits1.modify_enable); + debug_printf("\t\t.bits1.surface_state_address = 0x%x\n", (*ptr).bits1.surface_state_address); + debug_printf("\t\t.bits2.modify_enable = 0x%x\n", (*ptr).bits2.modify_enable); + debug_printf("\t\t.bits2.indirect_object_state_address = 0x%x\n", (*ptr).bits2.indirect_object_state_address); + debug_printf("\t\t.bits3.modify_enable = 0x%x\n", (*ptr).bits3.modify_enable); + debug_printf("\t\t.bits3.general_state_upper_bound = 0x%x\n", (*ptr).bits3.general_state_upper_bound); + debug_printf("\t\t.bits4.modify_enable = 0x%x\n", (*ptr).bits4.modify_enable); + debug_printf("\t\t.bits4.indirect_object_state_upper_bound = 0x%x\n", (*ptr).bits4.indirect_object_state_upper_bound); +} + +void +brw_dump_state_prefetch(const struct brw_state_prefetch *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.prefetch_count = 0x%x\n", (*ptr).bits0.prefetch_count); + debug_printf("\t\t.bits0.prefetch_pointer = 0x%x\n", (*ptr).bits0.prefetch_pointer); +} + +void +brw_dump_surf_ss0(const struct brw_surf_ss0 *ptr) +{ + debug_printf("\t\t.cube_pos_z = 0x%x\n", (*ptr).cube_pos_z); + debug_printf("\t\t.cube_neg_z = 0x%x\n", (*ptr).cube_neg_z); + debug_printf("\t\t.cube_pos_y = 0x%x\n", (*ptr).cube_pos_y); + debug_printf("\t\t.cube_neg_y = 0x%x\n", (*ptr).cube_neg_y); + debug_printf("\t\t.cube_pos_x = 0x%x\n", (*ptr).cube_pos_x); + debug_printf("\t\t.cube_neg_x = 0x%x\n", (*ptr).cube_neg_x); + debug_printf("\t\t.mipmap_layout_mode = 0x%x\n", (*ptr).mipmap_layout_mode); + debug_printf("\t\t.vert_line_stride_ofs = 0x%x\n", (*ptr).vert_line_stride_ofs); + debug_printf("\t\t.vert_line_stride = 0x%x\n", (*ptr).vert_line_stride); + debug_printf("\t\t.color_blend = 0x%x\n", (*ptr).color_blend); + debug_printf("\t\t.writedisable_blue = 0x%x\n", (*ptr).writedisable_blue); + debug_printf("\t\t.writedisable_green = 0x%x\n", (*ptr).writedisable_green); + debug_printf("\t\t.writedisable_red = 0x%x\n", (*ptr).writedisable_red); + debug_printf("\t\t.writedisable_alpha = 0x%x\n", (*ptr).writedisable_alpha); + debug_printf("\t\t.surface_format = 0x%x\n", (*ptr).surface_format); + debug_printf("\t\t.data_return_format = 0x%x\n", (*ptr).data_return_format); + debug_printf("\t\t.surface_type = 0x%x\n", (*ptr).surface_type); +} + +void +brw_dump_surf_ss1(const struct brw_surf_ss1 *ptr) +{ + debug_printf("\t\t.base_addr = 0x%x\n", (*ptr).base_addr); +} + +void +brw_dump_surf_ss2(const struct brw_surf_ss2 *ptr) +{ + debug_printf("\t\t.mip_count = 0x%x\n", (*ptr).mip_count); + debug_printf("\t\t.width = 0x%x\n", (*ptr).width); + debug_printf("\t\t.height = 0x%x\n", (*ptr).height); +} + +void +brw_dump_surf_ss3(const struct brw_surf_ss3 *ptr) +{ + debug_printf("\t\t.tile_walk = 0x%x\n", (*ptr).tile_walk); + debug_printf("\t\t.tiled_surface = 0x%x\n", (*ptr).tiled_surface); + debug_printf("\t\t.pitch = 0x%x\n", (*ptr).pitch); + debug_printf("\t\t.depth = 0x%x\n", (*ptr).depth); +} + +void +brw_dump_surf_ss4(const struct brw_surf_ss4 *ptr) +{ + debug_printf("\t\t.multisample_position_palette_index = 0x%x\n", (*ptr).multisample_position_palette_index); + debug_printf("\t\t.num_multisamples = 0x%x\n", (*ptr).num_multisamples); + debug_printf("\t\t.render_target_view_extent = 0x%x\n", (*ptr).render_target_view_extent); + debug_printf("\t\t.min_array_elt = 0x%x\n", (*ptr).min_array_elt); + debug_printf("\t\t.min_lod = 0x%x\n", (*ptr).min_lod); +} + +void +brw_dump_surf_ss5(const struct brw_surf_ss5 *ptr) +{ + debug_printf("\t\t.llc_mapping = 0x%x\n", (*ptr).llc_mapping); + debug_printf("\t\t.mlc_mapping = 0x%x\n", (*ptr).mlc_mapping); + debug_printf("\t\t.gfdt = 0x%x\n", (*ptr).gfdt); + debug_printf("\t\t.gfdt_src = 0x%x\n", (*ptr).gfdt_src); + debug_printf("\t\t.y_offset = 0x%x\n", (*ptr).y_offset); + debug_printf("\t\t.x_offset = 0x%x\n", (*ptr).x_offset); +} + +void +brw_dump_surface_state(const struct brw_surface_state *ptr) +{ + debug_printf("\t\t.ss0.cube_pos_z = 0x%x\n", (*ptr).ss0.cube_pos_z); + debug_printf("\t\t.ss0.cube_neg_z = 0x%x\n", (*ptr).ss0.cube_neg_z); + debug_printf("\t\t.ss0.cube_pos_y = 0x%x\n", (*ptr).ss0.cube_pos_y); + debug_printf("\t\t.ss0.cube_neg_y = 0x%x\n", (*ptr).ss0.cube_neg_y); + debug_printf("\t\t.ss0.cube_pos_x = 0x%x\n", (*ptr).ss0.cube_pos_x); + debug_printf("\t\t.ss0.cube_neg_x = 0x%x\n", (*ptr).ss0.cube_neg_x); + debug_printf("\t\t.ss0.mipmap_layout_mode = 0x%x\n", (*ptr).ss0.mipmap_layout_mode); + debug_printf("\t\t.ss0.vert_line_stride_ofs = 0x%x\n", (*ptr).ss0.vert_line_stride_ofs); + debug_printf("\t\t.ss0.vert_line_stride = 0x%x\n", (*ptr).ss0.vert_line_stride); + debug_printf("\t\t.ss0.color_blend = 0x%x\n", (*ptr).ss0.color_blend); + debug_printf("\t\t.ss0.writedisable_blue = 0x%x\n", (*ptr).ss0.writedisable_blue); + debug_printf("\t\t.ss0.writedisable_green = 0x%x\n", (*ptr).ss0.writedisable_green); + debug_printf("\t\t.ss0.writedisable_red = 0x%x\n", (*ptr).ss0.writedisable_red); + debug_printf("\t\t.ss0.writedisable_alpha = 0x%x\n", (*ptr).ss0.writedisable_alpha); + debug_printf("\t\t.ss0.surface_format = 0x%x\n", (*ptr).ss0.surface_format); + debug_printf("\t\t.ss0.data_return_format = 0x%x\n", (*ptr).ss0.data_return_format); + debug_printf("\t\t.ss0.surface_type = 0x%x\n", (*ptr).ss0.surface_type); + debug_printf("\t\t.ss1.base_addr = 0x%x\n", (*ptr).ss1.base_addr); + debug_printf("\t\t.ss2.mip_count = 0x%x\n", (*ptr).ss2.mip_count); + debug_printf("\t\t.ss2.width = 0x%x\n", (*ptr).ss2.width); + debug_printf("\t\t.ss2.height = 0x%x\n", (*ptr).ss2.height); + debug_printf("\t\t.ss3.tile_walk = 0x%x\n", (*ptr).ss3.tile_walk); + debug_printf("\t\t.ss3.tiled_surface = 0x%x\n", (*ptr).ss3.tiled_surface); + debug_printf("\t\t.ss3.pitch = 0x%x\n", (*ptr).ss3.pitch); + debug_printf("\t\t.ss3.depth = 0x%x\n", (*ptr).ss3.depth); + debug_printf("\t\t.ss4.multisample_position_palette_index = 0x%x\n", (*ptr).ss4.multisample_position_palette_index); + debug_printf("\t\t.ss4.num_multisamples = 0x%x\n", (*ptr).ss4.num_multisamples); + debug_printf("\t\t.ss4.render_target_view_extent = 0x%x\n", (*ptr).ss4.render_target_view_extent); + debug_printf("\t\t.ss4.min_array_elt = 0x%x\n", (*ptr).ss4.min_array_elt); + debug_printf("\t\t.ss4.min_lod = 0x%x\n", (*ptr).ss4.min_lod); + debug_printf("\t\t.ss5.llc_mapping = 0x%x\n", (*ptr).ss5.llc_mapping); + debug_printf("\t\t.ss5.mlc_mapping = 0x%x\n", (*ptr).ss5.mlc_mapping); + debug_printf("\t\t.ss5.gfdt = 0x%x\n", (*ptr).ss5.gfdt); + debug_printf("\t\t.ss5.gfdt_src = 0x%x\n", (*ptr).ss5.gfdt_src); + debug_printf("\t\t.ss5.y_offset = 0x%x\n", (*ptr).ss5.y_offset); + debug_printf("\t\t.ss5.x_offset = 0x%x\n", (*ptr).ss5.x_offset); +} + +void +brw_dump_system_instruction_pointer(const struct brw_system_instruction_pointer *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.system_instruction_pointer = 0x%x\n", (*ptr).bits0.system_instruction_pointer); +} + +void +brw_dump_urb_fence(const struct brw_urb_fence *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.vs_realloc = 0x%x\n", (*ptr).header.vs_realloc); + debug_printf("\t\t.header.gs_realloc = 0x%x\n", (*ptr).header.gs_realloc); + debug_printf("\t\t.header.clp_realloc = 0x%x\n", (*ptr).header.clp_realloc); + debug_printf("\t\t.header.sf_realloc = 0x%x\n", (*ptr).header.sf_realloc); + debug_printf("\t\t.header.vfe_realloc = 0x%x\n", (*ptr).header.vfe_realloc); + debug_printf("\t\t.header.cs_realloc = 0x%x\n", (*ptr).header.cs_realloc); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.bits0.vs_fence = 0x%x\n", (*ptr).bits0.vs_fence); + debug_printf("\t\t.bits0.gs_fence = 0x%x\n", (*ptr).bits0.gs_fence); + debug_printf("\t\t.bits0.clp_fence = 0x%x\n", (*ptr).bits0.clp_fence); + debug_printf("\t\t.bits1.sf_fence = 0x%x\n", (*ptr).bits1.sf_fence); + debug_printf("\t\t.bits1.vf_fence = 0x%x\n", (*ptr).bits1.vf_fence); + debug_printf("\t\t.bits1.cs_fence = 0x%x\n", (*ptr).bits1.cs_fence); +} + +void +brw_dump_urb_immediate(const struct brw_urb_immediate *ptr) +{ + debug_printf("\t\t.opcode = 0x%x\n", (*ptr).opcode); + debug_printf("\t\t.offset = 0x%x\n", (*ptr).offset); + debug_printf("\t\t.swizzle_control = 0x%x\n", (*ptr).swizzle_control); + debug_printf("\t\t.allocate = 0x%x\n", (*ptr).allocate); + debug_printf("\t\t.used = 0x%x\n", (*ptr).used); + debug_printf("\t\t.complete = 0x%x\n", (*ptr).complete); + debug_printf("\t\t.response_length = 0x%x\n", (*ptr).response_length); + debug_printf("\t\t.msg_length = 0x%x\n", (*ptr).msg_length); + debug_printf("\t\t.msg_target = 0x%x\n", (*ptr).msg_target); + debug_printf("\t\t.end_of_thread = 0x%x\n", (*ptr).end_of_thread); +} + +void +brw_dump_vb_array_state(const struct brw_vb_array_state *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.vb[0].vb0.pitch = 0x%x\n", (*ptr).vb[0].vb0.pitch); + debug_printf("\t\t.vb[0].vb0.access_type = 0x%x\n", (*ptr).vb[0].vb0.access_type); + debug_printf("\t\t.vb[0].vb0.vb_index = 0x%x\n", (*ptr).vb[0].vb0.vb_index); + debug_printf("\t\t.vb[0].start_addr = 0x%x\n", (*ptr).vb[0].start_addr); + debug_printf("\t\t.vb[0].max_index = 0x%x\n", (*ptr).vb[0].max_index); + debug_printf("\t\t.vb[0].instance_data_step_rate = 0x%x\n", (*ptr).vb[0].instance_data_step_rate); + debug_printf("\t\t.vb[1].vb0.pitch = 0x%x\n", (*ptr).vb[1].vb0.pitch); + debug_printf("\t\t.vb[1].vb0.access_type = 0x%x\n", (*ptr).vb[1].vb0.access_type); + debug_printf("\t\t.vb[1].vb0.vb_index = 0x%x\n", (*ptr).vb[1].vb0.vb_index); + debug_printf("\t\t.vb[1].start_addr = 0x%x\n", (*ptr).vb[1].start_addr); + debug_printf("\t\t.vb[1].max_index = 0x%x\n", (*ptr).vb[1].max_index); + debug_printf("\t\t.vb[1].instance_data_step_rate = 0x%x\n", (*ptr).vb[1].instance_data_step_rate); + debug_printf("\t\t.vb[2].vb0.pitch = 0x%x\n", (*ptr).vb[2].vb0.pitch); + debug_printf("\t\t.vb[2].vb0.access_type = 0x%x\n", (*ptr).vb[2].vb0.access_type); + debug_printf("\t\t.vb[2].vb0.vb_index = 0x%x\n", (*ptr).vb[2].vb0.vb_index); + debug_printf("\t\t.vb[2].start_addr = 0x%x\n", (*ptr).vb[2].start_addr); + debug_printf("\t\t.vb[2].max_index = 0x%x\n", (*ptr).vb[2].max_index); + debug_printf("\t\t.vb[2].instance_data_step_rate = 0x%x\n", (*ptr).vb[2].instance_data_step_rate); + debug_printf("\t\t.vb[3].vb0.pitch = 0x%x\n", (*ptr).vb[3].vb0.pitch); + debug_printf("\t\t.vb[3].vb0.access_type = 0x%x\n", (*ptr).vb[3].vb0.access_type); + debug_printf("\t\t.vb[3].vb0.vb_index = 0x%x\n", (*ptr).vb[3].vb0.vb_index); + debug_printf("\t\t.vb[3].start_addr = 0x%x\n", (*ptr).vb[3].start_addr); + debug_printf("\t\t.vb[3].max_index = 0x%x\n", (*ptr).vb[3].max_index); + debug_printf("\t\t.vb[3].instance_data_step_rate = 0x%x\n", (*ptr).vb[3].instance_data_step_rate); + debug_printf("\t\t.vb[4].vb0.pitch = 0x%x\n", (*ptr).vb[4].vb0.pitch); + debug_printf("\t\t.vb[4].vb0.access_type = 0x%x\n", (*ptr).vb[4].vb0.access_type); + debug_printf("\t\t.vb[4].vb0.vb_index = 0x%x\n", (*ptr).vb[4].vb0.vb_index); + debug_printf("\t\t.vb[4].start_addr = 0x%x\n", (*ptr).vb[4].start_addr); + debug_printf("\t\t.vb[4].max_index = 0x%x\n", (*ptr).vb[4].max_index); + debug_printf("\t\t.vb[4].instance_data_step_rate = 0x%x\n", (*ptr).vb[4].instance_data_step_rate); + debug_printf("\t\t.vb[5].vb0.pitch = 0x%x\n", (*ptr).vb[5].vb0.pitch); + debug_printf("\t\t.vb[5].vb0.access_type = 0x%x\n", (*ptr).vb[5].vb0.access_type); + debug_printf("\t\t.vb[5].vb0.vb_index = 0x%x\n", (*ptr).vb[5].vb0.vb_index); + debug_printf("\t\t.vb[5].start_addr = 0x%x\n", (*ptr).vb[5].start_addr); + debug_printf("\t\t.vb[5].max_index = 0x%x\n", (*ptr).vb[5].max_index); + debug_printf("\t\t.vb[5].instance_data_step_rate = 0x%x\n", (*ptr).vb[5].instance_data_step_rate); + debug_printf("\t\t.vb[6].vb0.pitch = 0x%x\n", (*ptr).vb[6].vb0.pitch); + debug_printf("\t\t.vb[6].vb0.access_type = 0x%x\n", (*ptr).vb[6].vb0.access_type); + debug_printf("\t\t.vb[6].vb0.vb_index = 0x%x\n", (*ptr).vb[6].vb0.vb_index); + debug_printf("\t\t.vb[6].start_addr = 0x%x\n", (*ptr).vb[6].start_addr); + debug_printf("\t\t.vb[6].max_index = 0x%x\n", (*ptr).vb[6].max_index); + debug_printf("\t\t.vb[6].instance_data_step_rate = 0x%x\n", (*ptr).vb[6].instance_data_step_rate); + debug_printf("\t\t.vb[7].vb0.pitch = 0x%x\n", (*ptr).vb[7].vb0.pitch); + debug_printf("\t\t.vb[7].vb0.access_type = 0x%x\n", (*ptr).vb[7].vb0.access_type); + debug_printf("\t\t.vb[7].vb0.vb_index = 0x%x\n", (*ptr).vb[7].vb0.vb_index); + debug_printf("\t\t.vb[7].start_addr = 0x%x\n", (*ptr).vb[7].start_addr); + debug_printf("\t\t.vb[7].max_index = 0x%x\n", (*ptr).vb[7].max_index); + debug_printf("\t\t.vb[7].instance_data_step_rate = 0x%x\n", (*ptr).vb[7].instance_data_step_rate); + debug_printf("\t\t.vb[8].vb0.pitch = 0x%x\n", (*ptr).vb[8].vb0.pitch); + debug_printf("\t\t.vb[8].vb0.access_type = 0x%x\n", (*ptr).vb[8].vb0.access_type); + debug_printf("\t\t.vb[8].vb0.vb_index = 0x%x\n", (*ptr).vb[8].vb0.vb_index); + debug_printf("\t\t.vb[8].start_addr = 0x%x\n", (*ptr).vb[8].start_addr); + debug_printf("\t\t.vb[8].max_index = 0x%x\n", (*ptr).vb[8].max_index); + debug_printf("\t\t.vb[8].instance_data_step_rate = 0x%x\n", (*ptr).vb[8].instance_data_step_rate); + debug_printf("\t\t.vb[9].vb0.pitch = 0x%x\n", (*ptr).vb[9].vb0.pitch); + debug_printf("\t\t.vb[9].vb0.access_type = 0x%x\n", (*ptr).vb[9].vb0.access_type); + debug_printf("\t\t.vb[9].vb0.vb_index = 0x%x\n", (*ptr).vb[9].vb0.vb_index); + debug_printf("\t\t.vb[9].start_addr = 0x%x\n", (*ptr).vb[9].start_addr); + debug_printf("\t\t.vb[9].max_index = 0x%x\n", (*ptr).vb[9].max_index); + debug_printf("\t\t.vb[9].instance_data_step_rate = 0x%x\n", (*ptr).vb[9].instance_data_step_rate); + debug_printf("\t\t.vb[10].vb0.pitch = 0x%x\n", (*ptr).vb[10].vb0.pitch); + debug_printf("\t\t.vb[10].vb0.access_type = 0x%x\n", (*ptr).vb[10].vb0.access_type); + debug_printf("\t\t.vb[10].vb0.vb_index = 0x%x\n", (*ptr).vb[10].vb0.vb_index); + debug_printf("\t\t.vb[10].start_addr = 0x%x\n", (*ptr).vb[10].start_addr); + debug_printf("\t\t.vb[10].max_index = 0x%x\n", (*ptr).vb[10].max_index); + debug_printf("\t\t.vb[10].instance_data_step_rate = 0x%x\n", (*ptr).vb[10].instance_data_step_rate); + debug_printf("\t\t.vb[11].vb0.pitch = 0x%x\n", (*ptr).vb[11].vb0.pitch); + debug_printf("\t\t.vb[11].vb0.access_type = 0x%x\n", (*ptr).vb[11].vb0.access_type); + debug_printf("\t\t.vb[11].vb0.vb_index = 0x%x\n", (*ptr).vb[11].vb0.vb_index); + debug_printf("\t\t.vb[11].start_addr = 0x%x\n", (*ptr).vb[11].start_addr); + debug_printf("\t\t.vb[11].max_index = 0x%x\n", (*ptr).vb[11].max_index); + debug_printf("\t\t.vb[11].instance_data_step_rate = 0x%x\n", (*ptr).vb[11].instance_data_step_rate); + debug_printf("\t\t.vb[12].vb0.pitch = 0x%x\n", (*ptr).vb[12].vb0.pitch); + debug_printf("\t\t.vb[12].vb0.access_type = 0x%x\n", (*ptr).vb[12].vb0.access_type); + debug_printf("\t\t.vb[12].vb0.vb_index = 0x%x\n", (*ptr).vb[12].vb0.vb_index); + debug_printf("\t\t.vb[12].start_addr = 0x%x\n", (*ptr).vb[12].start_addr); + debug_printf("\t\t.vb[12].max_index = 0x%x\n", (*ptr).vb[12].max_index); + debug_printf("\t\t.vb[12].instance_data_step_rate = 0x%x\n", (*ptr).vb[12].instance_data_step_rate); + debug_printf("\t\t.vb[13].vb0.pitch = 0x%x\n", (*ptr).vb[13].vb0.pitch); + debug_printf("\t\t.vb[13].vb0.access_type = 0x%x\n", (*ptr).vb[13].vb0.access_type); + debug_printf("\t\t.vb[13].vb0.vb_index = 0x%x\n", (*ptr).vb[13].vb0.vb_index); + debug_printf("\t\t.vb[13].start_addr = 0x%x\n", (*ptr).vb[13].start_addr); + debug_printf("\t\t.vb[13].max_index = 0x%x\n", (*ptr).vb[13].max_index); + debug_printf("\t\t.vb[13].instance_data_step_rate = 0x%x\n", (*ptr).vb[13].instance_data_step_rate); + debug_printf("\t\t.vb[14].vb0.pitch = 0x%x\n", (*ptr).vb[14].vb0.pitch); + debug_printf("\t\t.vb[14].vb0.access_type = 0x%x\n", (*ptr).vb[14].vb0.access_type); + debug_printf("\t\t.vb[14].vb0.vb_index = 0x%x\n", (*ptr).vb[14].vb0.vb_index); + debug_printf("\t\t.vb[14].start_addr = 0x%x\n", (*ptr).vb[14].start_addr); + debug_printf("\t\t.vb[14].max_index = 0x%x\n", (*ptr).vb[14].max_index); + debug_printf("\t\t.vb[14].instance_data_step_rate = 0x%x\n", (*ptr).vb[14].instance_data_step_rate); + debug_printf("\t\t.vb[15].vb0.pitch = 0x%x\n", (*ptr).vb[15].vb0.pitch); + debug_printf("\t\t.vb[15].vb0.access_type = 0x%x\n", (*ptr).vb[15].vb0.access_type); + debug_printf("\t\t.vb[15].vb0.vb_index = 0x%x\n", (*ptr).vb[15].vb0.vb_index); + debug_printf("\t\t.vb[15].start_addr = 0x%x\n", (*ptr).vb[15].start_addr); + debug_printf("\t\t.vb[15].max_index = 0x%x\n", (*ptr).vb[15].max_index); + debug_printf("\t\t.vb[15].instance_data_step_rate = 0x%x\n", (*ptr).vb[15].instance_data_step_rate); + debug_printf("\t\t.vb[16].vb0.pitch = 0x%x\n", (*ptr).vb[16].vb0.pitch); + debug_printf("\t\t.vb[16].vb0.access_type = 0x%x\n", (*ptr).vb[16].vb0.access_type); + debug_printf("\t\t.vb[16].vb0.vb_index = 0x%x\n", (*ptr).vb[16].vb0.vb_index); + debug_printf("\t\t.vb[16].start_addr = 0x%x\n", (*ptr).vb[16].start_addr); + debug_printf("\t\t.vb[16].max_index = 0x%x\n", (*ptr).vb[16].max_index); + debug_printf("\t\t.vb[16].instance_data_step_rate = 0x%x\n", (*ptr).vb[16].instance_data_step_rate); +} + +void +brw_dump_vertex_buffer_state(const struct brw_vertex_buffer_state *ptr) +{ + debug_printf("\t\t.vb0.pitch = 0x%x\n", (*ptr).vb0.pitch); + debug_printf("\t\t.vb0.access_type = 0x%x\n", (*ptr).vb0.access_type); + debug_printf("\t\t.vb0.vb_index = 0x%x\n", (*ptr).vb0.vb_index); + debug_printf("\t\t.start_addr = 0x%x\n", (*ptr).start_addr); + debug_printf("\t\t.max_index = 0x%x\n", (*ptr).max_index); + debug_printf("\t\t.instance_data_step_rate = 0x%x\n", (*ptr).instance_data_step_rate); +} + +void +brw_dump_vertex_element_packet(const struct brw_vertex_element_packet *ptr) +{ + debug_printf("\t\t.header.length = 0x%x\n", (*ptr).header.length); + debug_printf("\t\t.header.opcode = 0x%x\n", (*ptr).header.opcode); + debug_printf("\t\t.ve[0].ve0.src_offset = 0x%x\n", (*ptr).ve[0].ve0.src_offset); + debug_printf("\t\t.ve[0].ve0.src_format = 0x%x\n", (*ptr).ve[0].ve0.src_format); + debug_printf("\t\t.ve[0].ve0.valid = 0x%x\n", (*ptr).ve[0].ve0.valid); + debug_printf("\t\t.ve[0].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[0].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[0].ve1.dst_offset = 0x%x\n", (*ptr).ve[0].ve1.dst_offset); + debug_printf("\t\t.ve[0].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[0].ve1.vfcomponent3); + debug_printf("\t\t.ve[0].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[0].ve1.vfcomponent2); + debug_printf("\t\t.ve[0].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[0].ve1.vfcomponent1); + debug_printf("\t\t.ve[0].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[0].ve1.vfcomponent0); + debug_printf("\t\t.ve[1].ve0.src_offset = 0x%x\n", (*ptr).ve[1].ve0.src_offset); + debug_printf("\t\t.ve[1].ve0.src_format = 0x%x\n", (*ptr).ve[1].ve0.src_format); + debug_printf("\t\t.ve[1].ve0.valid = 0x%x\n", (*ptr).ve[1].ve0.valid); + debug_printf("\t\t.ve[1].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[1].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[1].ve1.dst_offset = 0x%x\n", (*ptr).ve[1].ve1.dst_offset); + debug_printf("\t\t.ve[1].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[1].ve1.vfcomponent3); + debug_printf("\t\t.ve[1].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[1].ve1.vfcomponent2); + debug_printf("\t\t.ve[1].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[1].ve1.vfcomponent1); + debug_printf("\t\t.ve[1].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[1].ve1.vfcomponent0); + debug_printf("\t\t.ve[2].ve0.src_offset = 0x%x\n", (*ptr).ve[2].ve0.src_offset); + debug_printf("\t\t.ve[2].ve0.src_format = 0x%x\n", (*ptr).ve[2].ve0.src_format); + debug_printf("\t\t.ve[2].ve0.valid = 0x%x\n", (*ptr).ve[2].ve0.valid); + debug_printf("\t\t.ve[2].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[2].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[2].ve1.dst_offset = 0x%x\n", (*ptr).ve[2].ve1.dst_offset); + debug_printf("\t\t.ve[2].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[2].ve1.vfcomponent3); + debug_printf("\t\t.ve[2].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[2].ve1.vfcomponent2); + debug_printf("\t\t.ve[2].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[2].ve1.vfcomponent1); + debug_printf("\t\t.ve[2].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[2].ve1.vfcomponent0); + debug_printf("\t\t.ve[3].ve0.src_offset = 0x%x\n", (*ptr).ve[3].ve0.src_offset); + debug_printf("\t\t.ve[3].ve0.src_format = 0x%x\n", (*ptr).ve[3].ve0.src_format); + debug_printf("\t\t.ve[3].ve0.valid = 0x%x\n", (*ptr).ve[3].ve0.valid); + debug_printf("\t\t.ve[3].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[3].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[3].ve1.dst_offset = 0x%x\n", (*ptr).ve[3].ve1.dst_offset); + debug_printf("\t\t.ve[3].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[3].ve1.vfcomponent3); + debug_printf("\t\t.ve[3].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[3].ve1.vfcomponent2); + debug_printf("\t\t.ve[3].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[3].ve1.vfcomponent1); + debug_printf("\t\t.ve[3].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[3].ve1.vfcomponent0); + debug_printf("\t\t.ve[4].ve0.src_offset = 0x%x\n", (*ptr).ve[4].ve0.src_offset); + debug_printf("\t\t.ve[4].ve0.src_format = 0x%x\n", (*ptr).ve[4].ve0.src_format); + debug_printf("\t\t.ve[4].ve0.valid = 0x%x\n", (*ptr).ve[4].ve0.valid); + debug_printf("\t\t.ve[4].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[4].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[4].ve1.dst_offset = 0x%x\n", (*ptr).ve[4].ve1.dst_offset); + debug_printf("\t\t.ve[4].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[4].ve1.vfcomponent3); + debug_printf("\t\t.ve[4].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[4].ve1.vfcomponent2); + debug_printf("\t\t.ve[4].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[4].ve1.vfcomponent1); + debug_printf("\t\t.ve[4].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[4].ve1.vfcomponent0); + debug_printf("\t\t.ve[5].ve0.src_offset = 0x%x\n", (*ptr).ve[5].ve0.src_offset); + debug_printf("\t\t.ve[5].ve0.src_format = 0x%x\n", (*ptr).ve[5].ve0.src_format); + debug_printf("\t\t.ve[5].ve0.valid = 0x%x\n", (*ptr).ve[5].ve0.valid); + debug_printf("\t\t.ve[5].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[5].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[5].ve1.dst_offset = 0x%x\n", (*ptr).ve[5].ve1.dst_offset); + debug_printf("\t\t.ve[5].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[5].ve1.vfcomponent3); + debug_printf("\t\t.ve[5].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[5].ve1.vfcomponent2); + debug_printf("\t\t.ve[5].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[5].ve1.vfcomponent1); + debug_printf("\t\t.ve[5].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[5].ve1.vfcomponent0); + debug_printf("\t\t.ve[6].ve0.src_offset = 0x%x\n", (*ptr).ve[6].ve0.src_offset); + debug_printf("\t\t.ve[6].ve0.src_format = 0x%x\n", (*ptr).ve[6].ve0.src_format); + debug_printf("\t\t.ve[6].ve0.valid = 0x%x\n", (*ptr).ve[6].ve0.valid); + debug_printf("\t\t.ve[6].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[6].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[6].ve1.dst_offset = 0x%x\n", (*ptr).ve[6].ve1.dst_offset); + debug_printf("\t\t.ve[6].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[6].ve1.vfcomponent3); + debug_printf("\t\t.ve[6].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[6].ve1.vfcomponent2); + debug_printf("\t\t.ve[6].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[6].ve1.vfcomponent1); + debug_printf("\t\t.ve[6].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[6].ve1.vfcomponent0); + debug_printf("\t\t.ve[7].ve0.src_offset = 0x%x\n", (*ptr).ve[7].ve0.src_offset); + debug_printf("\t\t.ve[7].ve0.src_format = 0x%x\n", (*ptr).ve[7].ve0.src_format); + debug_printf("\t\t.ve[7].ve0.valid = 0x%x\n", (*ptr).ve[7].ve0.valid); + debug_printf("\t\t.ve[7].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[7].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[7].ve1.dst_offset = 0x%x\n", (*ptr).ve[7].ve1.dst_offset); + debug_printf("\t\t.ve[7].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[7].ve1.vfcomponent3); + debug_printf("\t\t.ve[7].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[7].ve1.vfcomponent2); + debug_printf("\t\t.ve[7].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[7].ve1.vfcomponent1); + debug_printf("\t\t.ve[7].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[7].ve1.vfcomponent0); + debug_printf("\t\t.ve[8].ve0.src_offset = 0x%x\n", (*ptr).ve[8].ve0.src_offset); + debug_printf("\t\t.ve[8].ve0.src_format = 0x%x\n", (*ptr).ve[8].ve0.src_format); + debug_printf("\t\t.ve[8].ve0.valid = 0x%x\n", (*ptr).ve[8].ve0.valid); + debug_printf("\t\t.ve[8].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[8].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[8].ve1.dst_offset = 0x%x\n", (*ptr).ve[8].ve1.dst_offset); + debug_printf("\t\t.ve[8].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[8].ve1.vfcomponent3); + debug_printf("\t\t.ve[8].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[8].ve1.vfcomponent2); + debug_printf("\t\t.ve[8].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[8].ve1.vfcomponent1); + debug_printf("\t\t.ve[8].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[8].ve1.vfcomponent0); + debug_printf("\t\t.ve[9].ve0.src_offset = 0x%x\n", (*ptr).ve[9].ve0.src_offset); + debug_printf("\t\t.ve[9].ve0.src_format = 0x%x\n", (*ptr).ve[9].ve0.src_format); + debug_printf("\t\t.ve[9].ve0.valid = 0x%x\n", (*ptr).ve[9].ve0.valid); + debug_printf("\t\t.ve[9].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[9].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[9].ve1.dst_offset = 0x%x\n", (*ptr).ve[9].ve1.dst_offset); + debug_printf("\t\t.ve[9].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[9].ve1.vfcomponent3); + debug_printf("\t\t.ve[9].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[9].ve1.vfcomponent2); + debug_printf("\t\t.ve[9].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[9].ve1.vfcomponent1); + debug_printf("\t\t.ve[9].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[9].ve1.vfcomponent0); + debug_printf("\t\t.ve[10].ve0.src_offset = 0x%x\n", (*ptr).ve[10].ve0.src_offset); + debug_printf("\t\t.ve[10].ve0.src_format = 0x%x\n", (*ptr).ve[10].ve0.src_format); + debug_printf("\t\t.ve[10].ve0.valid = 0x%x\n", (*ptr).ve[10].ve0.valid); + debug_printf("\t\t.ve[10].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[10].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[10].ve1.dst_offset = 0x%x\n", (*ptr).ve[10].ve1.dst_offset); + debug_printf("\t\t.ve[10].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[10].ve1.vfcomponent3); + debug_printf("\t\t.ve[10].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[10].ve1.vfcomponent2); + debug_printf("\t\t.ve[10].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[10].ve1.vfcomponent1); + debug_printf("\t\t.ve[10].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[10].ve1.vfcomponent0); + debug_printf("\t\t.ve[11].ve0.src_offset = 0x%x\n", (*ptr).ve[11].ve0.src_offset); + debug_printf("\t\t.ve[11].ve0.src_format = 0x%x\n", (*ptr).ve[11].ve0.src_format); + debug_printf("\t\t.ve[11].ve0.valid = 0x%x\n", (*ptr).ve[11].ve0.valid); + debug_printf("\t\t.ve[11].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[11].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[11].ve1.dst_offset = 0x%x\n", (*ptr).ve[11].ve1.dst_offset); + debug_printf("\t\t.ve[11].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[11].ve1.vfcomponent3); + debug_printf("\t\t.ve[11].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[11].ve1.vfcomponent2); + debug_printf("\t\t.ve[11].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[11].ve1.vfcomponent1); + debug_printf("\t\t.ve[11].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[11].ve1.vfcomponent0); + debug_printf("\t\t.ve[12].ve0.src_offset = 0x%x\n", (*ptr).ve[12].ve0.src_offset); + debug_printf("\t\t.ve[12].ve0.src_format = 0x%x\n", (*ptr).ve[12].ve0.src_format); + debug_printf("\t\t.ve[12].ve0.valid = 0x%x\n", (*ptr).ve[12].ve0.valid); + debug_printf("\t\t.ve[12].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[12].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[12].ve1.dst_offset = 0x%x\n", (*ptr).ve[12].ve1.dst_offset); + debug_printf("\t\t.ve[12].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[12].ve1.vfcomponent3); + debug_printf("\t\t.ve[12].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[12].ve1.vfcomponent2); + debug_printf("\t\t.ve[12].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[12].ve1.vfcomponent1); + debug_printf("\t\t.ve[12].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[12].ve1.vfcomponent0); + debug_printf("\t\t.ve[13].ve0.src_offset = 0x%x\n", (*ptr).ve[13].ve0.src_offset); + debug_printf("\t\t.ve[13].ve0.src_format = 0x%x\n", (*ptr).ve[13].ve0.src_format); + debug_printf("\t\t.ve[13].ve0.valid = 0x%x\n", (*ptr).ve[13].ve0.valid); + debug_printf("\t\t.ve[13].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[13].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[13].ve1.dst_offset = 0x%x\n", (*ptr).ve[13].ve1.dst_offset); + debug_printf("\t\t.ve[13].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[13].ve1.vfcomponent3); + debug_printf("\t\t.ve[13].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[13].ve1.vfcomponent2); + debug_printf("\t\t.ve[13].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[13].ve1.vfcomponent1); + debug_printf("\t\t.ve[13].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[13].ve1.vfcomponent0); + debug_printf("\t\t.ve[14].ve0.src_offset = 0x%x\n", (*ptr).ve[14].ve0.src_offset); + debug_printf("\t\t.ve[14].ve0.src_format = 0x%x\n", (*ptr).ve[14].ve0.src_format); + debug_printf("\t\t.ve[14].ve0.valid = 0x%x\n", (*ptr).ve[14].ve0.valid); + debug_printf("\t\t.ve[14].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[14].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[14].ve1.dst_offset = 0x%x\n", (*ptr).ve[14].ve1.dst_offset); + debug_printf("\t\t.ve[14].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[14].ve1.vfcomponent3); + debug_printf("\t\t.ve[14].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[14].ve1.vfcomponent2); + debug_printf("\t\t.ve[14].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[14].ve1.vfcomponent1); + debug_printf("\t\t.ve[14].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[14].ve1.vfcomponent0); + debug_printf("\t\t.ve[15].ve0.src_offset = 0x%x\n", (*ptr).ve[15].ve0.src_offset); + debug_printf("\t\t.ve[15].ve0.src_format = 0x%x\n", (*ptr).ve[15].ve0.src_format); + debug_printf("\t\t.ve[15].ve0.valid = 0x%x\n", (*ptr).ve[15].ve0.valid); + debug_printf("\t\t.ve[15].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[15].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[15].ve1.dst_offset = 0x%x\n", (*ptr).ve[15].ve1.dst_offset); + debug_printf("\t\t.ve[15].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[15].ve1.vfcomponent3); + debug_printf("\t\t.ve[15].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[15].ve1.vfcomponent2); + debug_printf("\t\t.ve[15].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[15].ve1.vfcomponent1); + debug_printf("\t\t.ve[15].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[15].ve1.vfcomponent0); + debug_printf("\t\t.ve[16].ve0.src_offset = 0x%x\n", (*ptr).ve[16].ve0.src_offset); + debug_printf("\t\t.ve[16].ve0.src_format = 0x%x\n", (*ptr).ve[16].ve0.src_format); + debug_printf("\t\t.ve[16].ve0.valid = 0x%x\n", (*ptr).ve[16].ve0.valid); + debug_printf("\t\t.ve[16].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[16].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[16].ve1.dst_offset = 0x%x\n", (*ptr).ve[16].ve1.dst_offset); + debug_printf("\t\t.ve[16].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[16].ve1.vfcomponent3); + debug_printf("\t\t.ve[16].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[16].ve1.vfcomponent2); + debug_printf("\t\t.ve[16].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[16].ve1.vfcomponent1); + debug_printf("\t\t.ve[16].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[16].ve1.vfcomponent0); + debug_printf("\t\t.ve[17].ve0.src_offset = 0x%x\n", (*ptr).ve[17].ve0.src_offset); + debug_printf("\t\t.ve[17].ve0.src_format = 0x%x\n", (*ptr).ve[17].ve0.src_format); + debug_printf("\t\t.ve[17].ve0.valid = 0x%x\n", (*ptr).ve[17].ve0.valid); + debug_printf("\t\t.ve[17].ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve[17].ve0.vertex_buffer_index); + debug_printf("\t\t.ve[17].ve1.dst_offset = 0x%x\n", (*ptr).ve[17].ve1.dst_offset); + debug_printf("\t\t.ve[17].ve1.vfcomponent3 = 0x%x\n", (*ptr).ve[17].ve1.vfcomponent3); + debug_printf("\t\t.ve[17].ve1.vfcomponent2 = 0x%x\n", (*ptr).ve[17].ve1.vfcomponent2); + debug_printf("\t\t.ve[17].ve1.vfcomponent1 = 0x%x\n", (*ptr).ve[17].ve1.vfcomponent1); + debug_printf("\t\t.ve[17].ve1.vfcomponent0 = 0x%x\n", (*ptr).ve[17].ve1.vfcomponent0); +} + +void +brw_dump_vertex_element_state(const struct brw_vertex_element_state *ptr) +{ + debug_printf("\t\t.ve0.src_offset = 0x%x\n", (*ptr).ve0.src_offset); + debug_printf("\t\t.ve0.src_format = 0x%x\n", (*ptr).ve0.src_format); + debug_printf("\t\t.ve0.valid = 0x%x\n", (*ptr).ve0.valid); + debug_printf("\t\t.ve0.vertex_buffer_index = 0x%x\n", (*ptr).ve0.vertex_buffer_index); + debug_printf("\t\t.ve1.dst_offset = 0x%x\n", (*ptr).ve1.dst_offset); + debug_printf("\t\t.ve1.vfcomponent3 = 0x%x\n", (*ptr).ve1.vfcomponent3); + debug_printf("\t\t.ve1.vfcomponent2 = 0x%x\n", (*ptr).ve1.vfcomponent2); + debug_printf("\t\t.ve1.vfcomponent1 = 0x%x\n", (*ptr).ve1.vfcomponent1); + debug_printf("\t\t.ve1.vfcomponent0 = 0x%x\n", (*ptr).ve1.vfcomponent0); +} + +void +brw_dump_vf_statistics(const struct brw_vf_statistics *ptr) +{ + debug_printf("\t\t.statistics_enable = 0x%x\n", (*ptr).statistics_enable); + debug_printf("\t\t.opcode = 0x%x\n", (*ptr).opcode); +} + +void +brw_dump_vs_unit_state(const struct brw_vs_unit_state *ptr) +{ + debug_printf("\t\t.thread0.grf_reg_count = 0x%x\n", (*ptr).thread0.grf_reg_count); + debug_printf("\t\t.thread0.kernel_start_pointer = 0x%x\n", (*ptr).thread0.kernel_start_pointer); + debug_printf("\t\t.thread1.ext_halt_exception_enable = 0x%x\n", (*ptr).thread1.ext_halt_exception_enable); + debug_printf("\t\t.thread1.sw_exception_enable = 0x%x\n", (*ptr).thread1.sw_exception_enable); + debug_printf("\t\t.thread1.mask_stack_exception_enable = 0x%x\n", (*ptr).thread1.mask_stack_exception_enable); + debug_printf("\t\t.thread1.timeout_exception_enable = 0x%x\n", (*ptr).thread1.timeout_exception_enable); + debug_printf("\t\t.thread1.illegal_op_exception_enable = 0x%x\n", (*ptr).thread1.illegal_op_exception_enable); + debug_printf("\t\t.thread1.depth_coef_urb_read_offset = 0x%x\n", (*ptr).thread1.depth_coef_urb_read_offset); + debug_printf("\t\t.thread1.floating_point_mode = 0x%x\n", (*ptr).thread1.floating_point_mode); + debug_printf("\t\t.thread1.thread_priority = 0x%x\n", (*ptr).thread1.thread_priority); + debug_printf("\t\t.thread1.binding_table_entry_count = 0x%x\n", (*ptr).thread1.binding_table_entry_count); + debug_printf("\t\t.thread1.single_program_flow = 0x%x\n", (*ptr).thread1.single_program_flow); + debug_printf("\t\t.thread2.per_thread_scratch_space = 0x%x\n", (*ptr).thread2.per_thread_scratch_space); + debug_printf("\t\t.thread2.scratch_space_base_pointer = 0x%x\n", (*ptr).thread2.scratch_space_base_pointer); + debug_printf("\t\t.thread3.dispatch_grf_start_reg = 0x%x\n", (*ptr).thread3.dispatch_grf_start_reg); + debug_printf("\t\t.thread3.urb_entry_read_offset = 0x%x\n", (*ptr).thread3.urb_entry_read_offset); + debug_printf("\t\t.thread3.urb_entry_read_length = 0x%x\n", (*ptr).thread3.urb_entry_read_length); + debug_printf("\t\t.thread3.const_urb_entry_read_offset = 0x%x\n", (*ptr).thread3.const_urb_entry_read_offset); + debug_printf("\t\t.thread3.const_urb_entry_read_length = 0x%x\n", (*ptr).thread3.const_urb_entry_read_length); + debug_printf("\t\t.thread4.stats_enable = 0x%x\n", (*ptr).thread4.stats_enable); + debug_printf("\t\t.thread4.nr_urb_entries = 0x%x\n", (*ptr).thread4.nr_urb_entries); + debug_printf("\t\t.thread4.urb_entry_allocation_size = 0x%x\n", (*ptr).thread4.urb_entry_allocation_size); + debug_printf("\t\t.thread4.max_threads = 0x%x\n", (*ptr).thread4.max_threads); + debug_printf("\t\t.vs5.sampler_count = 0x%x\n", (*ptr).vs5.sampler_count); + debug_printf("\t\t.vs5.sampler_state_pointer = 0x%x\n", (*ptr).vs5.sampler_state_pointer); + debug_printf("\t\t.vs6.vs_enable = 0x%x\n", (*ptr).vs6.vs_enable); + debug_printf("\t\t.vs6.vert_cache_disable = 0x%x\n", (*ptr).vs6.vert_cache_disable); +} + +void +brw_dump_wm_unit_state(const struct brw_wm_unit_state *ptr) +{ + debug_printf("\t\t.thread0.grf_reg_count = 0x%x\n", (*ptr).thread0.grf_reg_count); + debug_printf("\t\t.thread0.kernel_start_pointer = 0x%x\n", (*ptr).thread0.kernel_start_pointer); + debug_printf("\t\t.thread1.ext_halt_exception_enable = 0x%x\n", (*ptr).thread1.ext_halt_exception_enable); + debug_printf("\t\t.thread1.sw_exception_enable = 0x%x\n", (*ptr).thread1.sw_exception_enable); + debug_printf("\t\t.thread1.mask_stack_exception_enable = 0x%x\n", (*ptr).thread1.mask_stack_exception_enable); + debug_printf("\t\t.thread1.timeout_exception_enable = 0x%x\n", (*ptr).thread1.timeout_exception_enable); + debug_printf("\t\t.thread1.illegal_op_exception_enable = 0x%x\n", (*ptr).thread1.illegal_op_exception_enable); + debug_printf("\t\t.thread1.depth_coef_urb_read_offset = 0x%x\n", (*ptr).thread1.depth_coef_urb_read_offset); + debug_printf("\t\t.thread1.floating_point_mode = 0x%x\n", (*ptr).thread1.floating_point_mode); + debug_printf("\t\t.thread1.thread_priority = 0x%x\n", (*ptr).thread1.thread_priority); + debug_printf("\t\t.thread1.binding_table_entry_count = 0x%x\n", (*ptr).thread1.binding_table_entry_count); + debug_printf("\t\t.thread1.single_program_flow = 0x%x\n", (*ptr).thread1.single_program_flow); + debug_printf("\t\t.thread2.per_thread_scratch_space = 0x%x\n", (*ptr).thread2.per_thread_scratch_space); + debug_printf("\t\t.thread2.scratch_space_base_pointer = 0x%x\n", (*ptr).thread2.scratch_space_base_pointer); + debug_printf("\t\t.thread3.dispatch_grf_start_reg = 0x%x\n", (*ptr).thread3.dispatch_grf_start_reg); + debug_printf("\t\t.thread3.urb_entry_read_offset = 0x%x\n", (*ptr).thread3.urb_entry_read_offset); + debug_printf("\t\t.thread3.urb_entry_read_length = 0x%x\n", (*ptr).thread3.urb_entry_read_length); + debug_printf("\t\t.thread3.const_urb_entry_read_offset = 0x%x\n", (*ptr).thread3.const_urb_entry_read_offset); + debug_printf("\t\t.thread3.const_urb_entry_read_length = 0x%x\n", (*ptr).thread3.const_urb_entry_read_length); + debug_printf("\t\t.wm4.stats_enable = 0x%x\n", (*ptr).wm4.stats_enable); + debug_printf("\t\t.wm4.depth_buffer_clear = 0x%x\n", (*ptr).wm4.depth_buffer_clear); + debug_printf("\t\t.wm4.sampler_count = 0x%x\n", (*ptr).wm4.sampler_count); + debug_printf("\t\t.wm4.sampler_state_pointer = 0x%x\n", (*ptr).wm4.sampler_state_pointer); + debug_printf("\t\t.wm5.enable_8_pix = 0x%x\n", (*ptr).wm5.enable_8_pix); + debug_printf("\t\t.wm5.enable_16_pix = 0x%x\n", (*ptr).wm5.enable_16_pix); + debug_printf("\t\t.wm5.enable_32_pix = 0x%x\n", (*ptr).wm5.enable_32_pix); + debug_printf("\t\t.wm5.enable_con_32_pix = 0x%x\n", (*ptr).wm5.enable_con_32_pix); + debug_printf("\t\t.wm5.enable_con_64_pix = 0x%x\n", (*ptr).wm5.enable_con_64_pix); + debug_printf("\t\t.wm5.legacy_global_depth_bias = 0x%x\n", (*ptr).wm5.legacy_global_depth_bias); + debug_printf("\t\t.wm5.line_stipple = 0x%x\n", (*ptr).wm5.line_stipple); + debug_printf("\t\t.wm5.depth_offset = 0x%x\n", (*ptr).wm5.depth_offset); + debug_printf("\t\t.wm5.polygon_stipple = 0x%x\n", (*ptr).wm5.polygon_stipple); + debug_printf("\t\t.wm5.line_aa_region_width = 0x%x\n", (*ptr).wm5.line_aa_region_width); + debug_printf("\t\t.wm5.line_endcap_aa_region_width = 0x%x\n", (*ptr).wm5.line_endcap_aa_region_width); + debug_printf("\t\t.wm5.early_depth_test = 0x%x\n", (*ptr).wm5.early_depth_test); + debug_printf("\t\t.wm5.thread_dispatch_enable = 0x%x\n", (*ptr).wm5.thread_dispatch_enable); + debug_printf("\t\t.wm5.program_uses_depth = 0x%x\n", (*ptr).wm5.program_uses_depth); + debug_printf("\t\t.wm5.program_computes_depth = 0x%x\n", (*ptr).wm5.program_computes_depth); + debug_printf("\t\t.wm5.program_uses_killpixel = 0x%x\n", (*ptr).wm5.program_uses_killpixel); + debug_printf("\t\t.wm5.legacy_line_rast = 0x%x\n", (*ptr).wm5.legacy_line_rast); + debug_printf("\t\t.wm5.transposed_urb_read_enable = 0x%x\n", (*ptr).wm5.transposed_urb_read_enable); + debug_printf("\t\t.wm5.max_threads = 0x%x\n", (*ptr).wm5.max_threads); + debug_printf("\t\t.global_depth_offset_constant = %f\n", (*ptr).global_depth_offset_constant); + debug_printf("\t\t.global_depth_offset_scale = %f\n", (*ptr).global_depth_offset_scale); + debug_printf("\t\t.wm8.grf_reg_count_1 = 0x%x\n", (*ptr).wm8.grf_reg_count_1); + debug_printf("\t\t.wm8.kernel_start_pointer_1 = 0x%x\n", (*ptr).wm8.kernel_start_pointer_1); + debug_printf("\t\t.wm9.grf_reg_count_2 = 0x%x\n", (*ptr).wm9.grf_reg_count_2); + debug_printf("\t\t.wm9.kernel_start_pointer_2 = 0x%x\n", (*ptr).wm9.kernel_start_pointer_2); + debug_printf("\t\t.wm10.grf_reg_count_3 = 0x%x\n", (*ptr).wm10.grf_reg_count_3); + debug_printf("\t\t.wm10.kernel_start_pointer_3 = 0x%x\n", (*ptr).wm10.kernel_start_pointer_3); +} + diff --git a/src/gallium/drivers/i965/brw_structs_dump.h b/src/gallium/drivers/i965/brw_structs_dump.h new file mode 100644 index 00000000000..7c02dbfe33f --- /dev/null +++ b/src/gallium/drivers/i965/brw_structs_dump.h @@ -0,0 +1,276 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * @file + * Dump i965 data structures. + * + * Generated automatically from brw_structs.h by brw_structs_dump.py. + */ + +#ifndef BRW_STRUCTS_DUMP_H +#define BRW_STRUCTS_DUMP_H + +struct brw_3d_control; +struct brw_3d_primitive; +struct brw_aa_line_parameters; +struct brw_binding_table_pointers; +struct brw_blend_constant_color; +struct brw_cc0; +struct brw_cc1; +struct brw_cc2; +struct brw_cc3; +struct brw_cc4; +struct brw_cc5; +struct brw_cc6; +struct brw_cc7; +struct brw_cc_unit_state; +struct brw_cc_viewport; +struct brw_clip_unit_state; +struct brw_clipper_viewport; +struct brw_constant_buffer; +struct brw_cs_urb_state; +struct brw_depthbuffer; +struct brw_depthbuffer_g4x; +struct brw_drawrect; +struct brw_global_depth_offset_clamp; +struct brw_gs_unit_state; +struct brw_indexbuffer; +struct brw_line_stipple; +struct brw_mi_flush; +struct brw_pipe_control; +struct brw_pipeline_select; +struct brw_pipelined_state_pointers; +struct brw_polygon_stipple; +struct brw_polygon_stipple_offset; +struct brw_sampler_default_color; +struct brw_sampler_state; +struct brw_sf_unit_state; +struct brw_sf_viewport; +struct brw_ss0; +struct brw_ss1; +struct brw_ss2; +struct brw_ss3; +struct brw_state_base_address; +struct brw_state_prefetch; +struct brw_surf_ss0; +struct brw_surf_ss1; +struct brw_surf_ss2; +struct brw_surf_ss3; +struct brw_surf_ss4; +struct brw_surf_ss5; +struct brw_surface_state; +struct brw_system_instruction_pointer; +struct brw_urb_fence; +struct brw_urb_immediate; +struct brw_vb_array_state; +struct brw_vertex_buffer_state; +struct brw_vertex_element_packet; +struct brw_vertex_element_state; +struct brw_vf_statistics; +struct brw_vs_unit_state; +struct brw_wm_unit_state; + +void +brw_dump_3d_control(const struct brw_3d_control *ptr); + +void +brw_dump_3d_primitive(const struct brw_3d_primitive *ptr); + +void +brw_dump_aa_line_parameters(const struct brw_aa_line_parameters *ptr); + +void +brw_dump_binding_table_pointers(const struct brw_binding_table_pointers *ptr); + +void +brw_dump_blend_constant_color(const struct brw_blend_constant_color *ptr); + +void +brw_dump_cc0(const struct brw_cc0 *ptr); + +void +brw_dump_cc1(const struct brw_cc1 *ptr); + +void +brw_dump_cc2(const struct brw_cc2 *ptr); + +void +brw_dump_cc3(const struct brw_cc3 *ptr); + +void +brw_dump_cc4(const struct brw_cc4 *ptr); + +void +brw_dump_cc5(const struct brw_cc5 *ptr); + +void +brw_dump_cc6(const struct brw_cc6 *ptr); + +void +brw_dump_cc7(const struct brw_cc7 *ptr); + +void +brw_dump_cc_unit_state(const struct brw_cc_unit_state *ptr); + +void +brw_dump_cc_viewport(const struct brw_cc_viewport *ptr); + +void +brw_dump_clip_unit_state(const struct brw_clip_unit_state *ptr); + +void +brw_dump_clipper_viewport(const struct brw_clipper_viewport *ptr); + +void +brw_dump_constant_buffer(const struct brw_constant_buffer *ptr); + +void +brw_dump_cs_urb_state(const struct brw_cs_urb_state *ptr); + +void +brw_dump_depthbuffer(const struct brw_depthbuffer *ptr); + +void +brw_dump_depthbuffer_g4x(const struct brw_depthbuffer_g4x *ptr); + +void +brw_dump_drawrect(const struct brw_drawrect *ptr); + +void +brw_dump_global_depth_offset_clamp(const struct brw_global_depth_offset_clamp *ptr); + +void +brw_dump_gs_unit_state(const struct brw_gs_unit_state *ptr); + +void +brw_dump_indexbuffer(const struct brw_indexbuffer *ptr); + +void +brw_dump_line_stipple(const struct brw_line_stipple *ptr); + +void +brw_dump_mi_flush(const struct brw_mi_flush *ptr); + +void +brw_dump_pipe_control(const struct brw_pipe_control *ptr); + +void +brw_dump_pipeline_select(const struct brw_pipeline_select *ptr); + +void +brw_dump_pipelined_state_pointers(const struct brw_pipelined_state_pointers *ptr); + +void +brw_dump_polygon_stipple(const struct brw_polygon_stipple *ptr); + +void +brw_dump_polygon_stipple_offset(const struct brw_polygon_stipple_offset *ptr); + +void +brw_dump_sampler_default_color(const struct brw_sampler_default_color *ptr); + +void +brw_dump_sampler_state(const struct brw_sampler_state *ptr); + +void +brw_dump_sf_unit_state(const struct brw_sf_unit_state *ptr); + +void +brw_dump_sf_viewport(const struct brw_sf_viewport *ptr); + +void +brw_dump_ss0(const struct brw_ss0 *ptr); + +void +brw_dump_ss1(const struct brw_ss1 *ptr); + +void +brw_dump_ss2(const struct brw_ss2 *ptr); + +void +brw_dump_ss3(const struct brw_ss3 *ptr); + +void +brw_dump_state_base_address(const struct brw_state_base_address *ptr); + +void +brw_dump_state_prefetch(const struct brw_state_prefetch *ptr); + +void +brw_dump_surf_ss0(const struct brw_surf_ss0 *ptr); + +void +brw_dump_surf_ss1(const struct brw_surf_ss1 *ptr); + +void +brw_dump_surf_ss2(const struct brw_surf_ss2 *ptr); + +void +brw_dump_surf_ss3(const struct brw_surf_ss3 *ptr); + +void +brw_dump_surf_ss4(const struct brw_surf_ss4 *ptr); + +void +brw_dump_surf_ss5(const struct brw_surf_ss5 *ptr); + +void +brw_dump_surface_state(const struct brw_surface_state *ptr); + +void +brw_dump_system_instruction_pointer(const struct brw_system_instruction_pointer *ptr); + +void +brw_dump_urb_fence(const struct brw_urb_fence *ptr); + +void +brw_dump_urb_immediate(const struct brw_urb_immediate *ptr); + +void +brw_dump_vb_array_state(const struct brw_vb_array_state *ptr); + +void +brw_dump_vertex_buffer_state(const struct brw_vertex_buffer_state *ptr); + +void +brw_dump_vertex_element_packet(const struct brw_vertex_element_packet *ptr); + +void +brw_dump_vertex_element_state(const struct brw_vertex_element_state *ptr); + +void +brw_dump_vf_statistics(const struct brw_vf_statistics *ptr); + +void +brw_dump_vs_unit_state(const struct brw_vs_unit_state *ptr); + +void +brw_dump_wm_unit_state(const struct brw_wm_unit_state *ptr); + + +#endif /* BRW_STRUCTS_DUMP_H */ diff --git a/src/gallium/drivers/i965/brw_structs_dump.py b/src/gallium/drivers/i965/brw_structs_dump.py new file mode 100755 index 00000000000..6dba49ad911 --- /dev/null +++ b/src/gallium/drivers/i965/brw_structs_dump.py @@ -0,0 +1,291 @@ +#!/usr/bin/env python +''' +Generates dumpers for the i965 state strucutures using pygccxml. + +Run as + + PYTHONPATH=/path/to/pygccxml-1.0.0 python brw_structs_dump.py + +Jose Fonseca <[email protected]> +''' + +copyright = ''' +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + ''' + +import os +import sys +import re + +from pygccxml import parser +from pygccxml import declarations + +from pygccxml.declarations import algorithm +from pygccxml.declarations import decl_visitor +from pygccxml.declarations import type_traits +from pygccxml.declarations import type_visitor + + +enums = True + + +def vars_filter(variable): + name = variable.name + return not re.match('^pad\d*', name) and name != 'dword' + + +class decl_dumper_t(decl_visitor.decl_visitor_t): + + def __init__(self, stream, instance = '', decl = None): + decl_visitor.decl_visitor_t.__init__(self) + self.stream = stream + self._instance = instance + self.decl = decl + + def clone(self): + return decl_dumper_t(self.stream, self._instance, self.decl) + + def visit_class(self): + class_ = self.decl + assert self.decl.class_type in ('struct', 'union') + + for variable in class_.variables(recursive = False): + if vars_filter(variable): + dump_type(self.stream, self._instance + '.' + variable.name, variable.type) + + def visit_enumeration(self): + if enums: + self.stream.write(' switch(%s) {\n' % ("(*ptr)" + self._instance,)) + for name, value in self.decl.values: + self.stream.write(' case %s:\n' % (name,)) + self.stream.write(' debug_printf("\\t\\t%s = %s\\n");\n' % (self._instance, name)) + self.stream.write(' break;\n') + self.stream.write(' default:\n') + self.stream.write(' debug_printf("\\t\\t%s = %%i\\n", %s);\n' % (self._instance, "(*ptr)" + self._instance)) + self.stream.write(' break;\n') + self.stream.write(' }\n') + else: + self.stream.write(' debug_printf("\\t\\t%s = %%i\\n", %s);\n' % (self._instance, "(*ptr)" + self._instance)) + + +def dump_decl(stream, instance, decl): + dumper = decl_dumper_t(stream, instance, decl) + algorithm.apply_visitor(dumper, decl) + + +class type_dumper_t(type_visitor.type_visitor_t): + + def __init__(self, stream, instance, type_): + type_visitor.type_visitor_t.__init__(self) + self.stream = stream + self.instance = instance + self.type = type_ + + def clone(self): + return type_dumper_t(self.instance, self.type) + + def visit_bool(self): + self.print_instance('%i') + + def visit_char(self): + #self.print_instance('%i') + self.print_instance('0x%x') + + def visit_unsigned_char(self): + #self.print_instance('%u') + self.print_instance('0x%x') + + def visit_signed_char(self): + #self.print_instance('%i') + self.print_instance('0x%x') + + def visit_wchar(self): + self.print_instance('0x%x') + + def visit_short_int(self): + #self.print_instance('%i') + self.print_instance('0x%x') + + def visit_short_unsigned_int(self): + #self.print_instance('%u') + self.print_instance('0x%x') + + def visit_int(self): + #self.print_instance('%i') + self.print_instance('0x%x') + + def visit_unsigned_int(self): + #self.print_instance('%u') + self.print_instance('0x%x') + + def visit_long_int(self): + #self.print_instance('%li') + self.print_instance('0x%lx') + + def visit_long_unsigned_int(self): + #self.print_instance('%lu') + self.print_instance('%0xlx') + + def visit_long_long_int(self): + #self.print_instance('%lli') + self.print_instance('%0xllx') + + def visit_long_long_unsigned_int(self): + #self.print_instance('%llu') + self.print_instance('0x%llx') + + def visit_float(self): + self.print_instance('%f') + + def visit_double(self): + self.print_instance('%f') + + def visit_array(self): + for i in range(type_traits.array_size(self.type)): + dump_type(self.stream, self.instance + '[%i]' % i, type_traits.base_type(self.type)) + + def visit_pointer(self): + self.print_instance('%p') + + def visit_declarated(self): + #stream.write('decl = %r\n' % self.type.decl_string) + decl = type_traits.remove_declarated(self.type) + dump_decl(self.stream, self.instance, decl) + + def print_instance(self, format): + self.stream.write(' debug_printf("\\t\\t%s = %s\\n", %s);\n' % (self.instance, format, "(*ptr)" + self.instance)) + + + +def dump_type(stream, instance, type_): + type_ = type_traits.remove_alias(type_) + visitor = type_dumper_t(stream, instance, type_) + algorithm.apply_visitor(visitor, type_) + + +def dump_struct_interface(stream, class_, suffix = ';'): + name = class_.name + assert name.startswith('brw_'); + name = name[:4] + 'dump_' + name[4:] + stream.write('void\n') + stream.write('%s(const struct %s *ptr)%s\n' % (name, class_.name, suffix)) + + +def dump_struct_implementation(stream, decls, class_): + dump_struct_interface(stream, class_, suffix = '') + stream.write('{\n') + dump_decl(stream, '', class_) + stream.write('}\n') + stream.write('\n') + + +def dump_header(stream): + stream.write(copyright.strip() + '\n') + stream.write('\n') + stream.write('/**\n') + stream.write(' * @file\n') + stream.write(' * Dump i965 data structures.\n') + stream.write(' *\n') + stream.write(' * Generated automatically from brw_structs.h by brw_structs_dump.py.\n') + stream.write(' */\n') + stream.write('\n') + + +def dump_interfaces(decls, global_ns, names): + stream = open('brw_structs_dump.h', 'wt') + + dump_header(stream) + + stream.write('#ifndef BRW_STRUCTS_DUMP_H\n') + stream.write('#define BRW_STRUCTS_DUMP_H\n') + stream.write('\n') + + for name in names: + stream.write('struct %s;\n' % (name,)) + stream.write('\n') + + for name in names: + (class_,) = global_ns.classes(name = name) + dump_struct_interface(stream, class_) + stream.write('\n') + stream.write('\n') + + stream.write('#endif /* BRW_STRUCTS_DUMP_H */\n') + + +def dump_implementations(decls, global_ns, names): + stream = open('brw_structs_dump.c', 'wt') + + dump_header(stream) + + stream.write('#include "util/u_debug.h"\n') + stream.write('\n') + stream.write('#include "brw_types.h"\n') + stream.write('#include "brw_structs.h"\n') + stream.write('#include "brw_structs_dump.h"\n') + stream.write('\n') + + for name in names: + (class_,) = global_ns.classes(name = name) + dump_struct_implementation(stream, decls, class_) + + +def decl_filter(decl): + '''Filter the declarations we're interested in''' + name = decl.name + return name.startswith('brw_') and name not in ('brw_instruction',) + + +def main(): + + config = parser.config_t( + include_paths = [ + '../../include', + ], + compiler = 'gcc', + ) + + headers = [ + 'brw_types.h', + 'brw_structs.h', + ] + + decls = parser.parse(headers, config, parser.COMPILATION_MODE.ALL_AT_ONCE) + global_ns = declarations.get_global_namespace(decls) + + names = [] + for class_ in global_ns.classes(decl_filter): + names.append(class_.name) + names.sort() + + dump_interfaces(decls, global_ns, names) + dump_implementations(decls, global_ns, names) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/drivers/i965/brw_swtnl.c b/src/gallium/drivers/i965/brw_swtnl.c new file mode 100644 index 00000000000..464013e7c40 --- /dev/null +++ b/src/gallium/drivers/i965/brw_swtnl.c @@ -0,0 +1,95 @@ + +#include "brw_context.h" +#include "brw_pipe_rast.h" + + +static GLboolean need_swtnl( struct brw_context *brw ) +{ + const struct pipe_rasterizer_state *rast = &brw->curr.rast->templ; + + /* If we don't require strict OpenGL conformance, never + * use fallbacks. If we're forcing fallbacks, always + * use fallfacks. + */ + if (brw->flags.no_swtnl) + return FALSE; + + if (brw->flags.force_swtnl) + return TRUE; + + /* Exceeding hw limits on number of VS inputs? + */ + if (brw->curr.num_vertex_elements == 0 || + brw->curr.num_vertex_elements >= BRW_VEP_MAX) { + return TRUE; + } + + /* Position array with zero stride? + * + * XXX: position isn't always at zero... + * XXX: eliminate zero-stride arrays + */ + { + int ve0_vb = brw->curr.vertex_element[0].vertex_buffer_index; + + if (brw->curr.vertex_buffer[ve0_vb].stride == 0) + return TRUE; + } + + /* XXX: short-circuit + */ + return FALSE; + + if (brw->reduced_primitive == PIPE_PRIM_TRIANGLES) { + if (rast->poly_smooth) + return TRUE; + + } + + if (brw->reduced_primitive == PIPE_PRIM_LINES || + (brw->reduced_primitive == PIPE_PRIM_TRIANGLES && + (rast->fill_cw == PIPE_POLYGON_MODE_LINE || + rast->fill_ccw == PIPE_POLYGON_MODE_LINE))) + { + /* BRW hardware will do AA lines, but they are non-conformant it + * seems. TBD whether we keep this fallback: + */ + if (rast->line_smooth) + return TRUE; + + /* XXX: was a fallback in mesa (gs doesn't get enough + * information to know when to reset stipple counter), but there + * must be a way around it. + */ + if (rast->line_stipple_enable && + (brw->reduced_primitive == PIPE_PRIM_TRIANGLES || + brw->primitive == PIPE_PRIM_LINE_LOOP || + brw->primitive == PIPE_PRIM_LINE_STRIP)) + return TRUE; + } + + + if (brw->reduced_primitive == PIPE_PRIM_POINTS || + (brw->reduced_primitive == PIPE_PRIM_TRIANGLES && + (rast->fill_cw == PIPE_POLYGON_MODE_POINT || + rast->fill_ccw == PIPE_POLYGON_MODE_POINT))) + { + if (rast->point_smooth) + return TRUE; + } + + /* BRW hardware doesn't handle CLAMP texturing correctly; + * brw_wm_sampler_state:translate_wrap_mode() treats CLAMP + * as CLAMP_TO_EDGE instead. If we're using CLAMP, and + * we want strict conformance, force the fallback. + * + * XXX: need a workaround for this. + */ + + /* Nothing stopping us from the fast path now */ + return FALSE; +} + + + + diff --git a/src/gallium/drivers/i965/brw_types.h b/src/gallium/drivers/i965/brw_types.h new file mode 100644 index 00000000000..89e08a5c804 --- /dev/null +++ b/src/gallium/drivers/i965/brw_types.h @@ -0,0 +1,21 @@ +#ifndef BRW_TYPES_H +#define BRW_TYPES_H + +#include "pipe/p_compiler.h" + +typedef uint32_t GLuint; +typedef uint8_t GLubyte; +typedef uint16_t GLushort; +typedef int32_t GLint; +typedef int8_t GLbyte; +typedef int16_t GLshort; +typedef float GLfloat; + +/* no GLenum, translate all away */ + +typedef uint8_t GLboolean; + +#define GL_FALSE FALSE +#define GL_TRUE TRUE + +#endif diff --git a/src/gallium/drivers/i965/brw_urb.c b/src/gallium/drivers/i965/brw_urb.c new file mode 100644 index 00000000000..907ec56c6ca --- /dev/null +++ b/src/gallium/drivers/i965/brw_urb.c @@ -0,0 +1,263 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + + +#include "brw_batchbuffer.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_debug.h" + +#define VS 0 +#define GS 1 +#define CLP 2 +#define SF 3 +#define CS 4 + +/** @file brw_urb.c + * + * Manages the division of the URB space between the various fixed-function + * units. + * + * See the Thread Initiation Management section of the GEN4 B-Spec, and + * the individual *_STATE structures for restrictions on numbers of + * entries and threads. + */ + +/* + * Generally, a unit requires a min_nr_entries based on how many entries + * it produces before the downstream unit gets unblocked and can use and + * dereference some of its handles. + * + * The SF unit preallocates a PUE at the start of thread dispatch, and only + * uses that one. So it requires one entry per thread. + * + * For CLIP, the SF unit will hold the previous primitive while the + * next is getting assembled, meaning that linestrips require 3 CLIP VUEs + * (vertices) to ensure continued processing, trifans require 4, and tristrips + * require 5. There can be 1 or 2 threads, and each has the same requirement. + * + * GS has the same requirement as CLIP, but it never handles tristrips, + * so we can lower the minimum to 4 for the POLYGONs (trifans) it produces. + * We only run it single-threaded. + * + * For VS, the number of entries may be 8, 12, 16, or 32 (or 64 on G4X). + * Each thread processes 2 preallocated VUEs (vertices) at a time, and they + * get streamed down as soon as threads processing earlier vertices get + * theirs accepted. + * + * Each unit will take the number of URB entries we give it (based on the + * entry size calculated in brw_vs_emit.c for VUEs, brw_sf_emit.c for PUEs, + * and brw_curbe.c for the CURBEs) and decide its maximum number of + * threads it can support based on that. in brw_*_state.c. + * + * XXX: Are the min_entry_size numbers useful? + * XXX: Verify min_nr_entries, esp for VS. + * XXX: Verify SF min_entry_size. + */ +static const struct urb_limits { + GLuint min_nr_entries; + GLuint preferred_nr_entries; + GLuint min_entry_size; + GLuint max_entry_size; +} limits[CS+1] = { + { 16, 32, 1, 5 }, /* vs */ + { 4, 8, 1, 5 }, /* gs */ + { 5, 10, 1, 5 }, /* clp */ + { 1, 8, 1, 12 }, /* sf */ + { 1, 4, 1, 32 } /* cs */ +}; + + +static GLboolean check_urb_layout( struct brw_context *brw ) +{ + brw->urb.vs_start = 0; + brw->urb.gs_start = brw->urb.nr_vs_entries * brw->urb.vsize; + brw->urb.clip_start = brw->urb.gs_start + brw->urb.nr_gs_entries * brw->urb.vsize; + brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize; + brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize; + + return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= URB_SIZES(brw); +} + +/* Most minimal update, forces re-emit of URB fence packet after GS + * unit turned on/off. + */ +static int recalculate_urb_fence( struct brw_context *brw ) +{ + GLuint csize = brw->curbe.total_size; + GLuint vsize = brw->vs.prog_data->urb_entry_size; + GLuint sfsize = brw->sf.prog_data->urb_entry_size; + + if (csize < limits[CS].min_entry_size) + csize = limits[CS].min_entry_size; + + if (vsize < limits[VS].min_entry_size) + vsize = limits[VS].min_entry_size; + + if (sfsize < limits[SF].min_entry_size) + sfsize = limits[SF].min_entry_size; + + if (brw->urb.vsize < vsize || + brw->urb.sfsize < sfsize || + brw->urb.csize < csize || + (brw->urb.constrained && (brw->urb.vsize > vsize || + brw->urb.sfsize > sfsize || + brw->urb.csize > csize))) { + + + brw->urb.csize = csize; + brw->urb.sfsize = sfsize; + brw->urb.vsize = vsize; + + brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; + brw->urb.nr_gs_entries = limits[GS].preferred_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries; + brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries; + brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries; + + brw->urb.constrained = 0; + + if (BRW_IS_IGDNG(brw)) { + brw->urb.nr_vs_entries = 128; + brw->urb.nr_sf_entries = 48; + if (check_urb_layout(brw)) { + goto done; + } else { + brw->urb.constrained = 1; + brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; + brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries; + } + } else if (BRW_IS_G4X(brw)) { + brw->urb.nr_vs_entries = 64; + if (check_urb_layout(brw)) { + goto done; + } else { + brw->urb.constrained = 1; + brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; + } + } + + if (BRW_DEBUG & DEBUG_MIN_URB) { + brw->urb.nr_vs_entries = limits[VS].min_nr_entries; + brw->urb.nr_gs_entries = limits[GS].min_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].min_nr_entries; + brw->urb.nr_sf_entries = limits[SF].min_nr_entries; + brw->urb.nr_cs_entries = limits[CS].min_nr_entries; + brw->urb.constrained = 1; + } + + if (!check_urb_layout(brw)) { + brw->urb.nr_vs_entries = limits[VS].min_nr_entries; + brw->urb.nr_gs_entries = limits[GS].min_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].min_nr_entries; + brw->urb.nr_sf_entries = limits[SF].min_nr_entries; + brw->urb.nr_cs_entries = limits[CS].min_nr_entries; + + /* Mark us as operating with constrained nr_entries, so that next + * time we recalculate we'll resize the fences in the hope of + * escaping constrained mode and getting back to normal performance. + */ + brw->urb.constrained = 1; + + if (!check_urb_layout(brw)) { + /* This is impossible, given the maximal sizes of urb + * entries and the values for minimum nr of entries + * provided above. + */ + debug_printf("couldn't calculate URB layout!\n"); + exit(1); + } + + if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)) + debug_printf("URB CONSTRAINED\n"); + } + +done: + if (BRW_DEBUG & DEBUG_URB) + debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", + brw->urb.vs_start, + brw->urb.gs_start, + brw->urb.clip_start, + brw->urb.sf_start, + brw->urb.cs_start, + URB_SIZES(brw)); + + brw->state.dirty.brw |= BRW_NEW_URB_FENCE; + } + + return 0; +} + + +const struct brw_tracked_state brw_recalculate_urb_fence = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_CURBE_OFFSETS, + .cache = (CACHE_NEW_VS_PROG | + CACHE_NEW_SF_PROG) + }, + .prepare = recalculate_urb_fence +}; + + + + + +int brw_upload_urb_fence(struct brw_context *brw) +{ + struct brw_urb_fence uf; + memset(&uf, 0, sizeof(uf)); + + uf.header.opcode = CMD_URB_FENCE; + uf.header.length = sizeof(uf)/4-2; + uf.header.vs_realloc = 1; + uf.header.gs_realloc = 1; + uf.header.clp_realloc = 1; + uf.header.sf_realloc = 1; + uf.header.vfe_realloc = 1; + uf.header.cs_realloc = 1; + + /* The ordering below is correct, not the layout in the + * instruction. + * + * There are 256/384 urb reg pairs in total. + */ + uf.bits0.vs_fence = brw->urb.gs_start; + uf.bits0.gs_fence = brw->urb.clip_start; + uf.bits0.clp_fence = brw->urb.sf_start; + uf.bits1.sf_fence = brw->urb.cs_start; + uf.bits1.cs_fence = URB_SIZES(brw); + + BRW_BATCH_STRUCT(brw, &uf); + return 0; +} diff --git a/src/gallium/drivers/i965/brw_util.c b/src/gallium/drivers/i965/brw_util.c new file mode 100644 index 00000000000..458058d668d --- /dev/null +++ b/src/gallium/drivers/i965/brw_util.c @@ -0,0 +1,38 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_util.h" +#include "brw_defines.h" + + + + diff --git a/src/gallium/drivers/i965/brw_util.h b/src/gallium/drivers/i965/brw_util.h new file mode 100644 index 00000000000..b5f9a36e7bc --- /dev/null +++ b/src/gallium/drivers/i965/brw_util.h @@ -0,0 +1,44 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_UTIL_H +#define BRW_UTIL_H + +#include "brw_types.h" + +extern GLuint brw_count_bits( GLuint val ); +extern GLuint brw_translate_blend_factor( unsigned factor ); +extern GLuint brw_translate_blend_equation( unsigned mode ); + + + +#endif diff --git a/src/gallium/drivers/i965/brw_vs.c b/src/gallium/drivers/i965/brw_vs.c new file mode 100644 index 00000000000..14a1c3bcf18 --- /dev/null +++ b/src/gallium/drivers/i965/brw_vs.c @@ -0,0 +1,145 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "tgsi/tgsi_dump.h" + +#include "brw_context.h" +#include "brw_vs.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_pipe_rast.h" + + + +static enum pipe_error do_vs_prog( struct brw_context *brw, + struct brw_vertex_shader *vp, + struct brw_vs_prog_key *key, + struct brw_winsys_buffer **bo_out) +{ + enum pipe_error ret; + GLuint program_size; + const GLuint *program; + struct brw_vs_compile c; + + memset(&c, 0, sizeof(c)); + memcpy(&c.key, key, sizeof(*key)); + + brw_init_compile(brw, &c.func); + c.vp = vp; + + c.prog_data.nr_outputs = vp->info.num_outputs; + c.prog_data.nr_inputs = vp->info.num_inputs; + + /* XXX: we want edgeflag handling to be integrated to the vertex + * shader, but are currently faking the edgeflag output: + */ + if (c.key.copy_edgeflag) { + c.prog_data.output_edgeflag = c.prog_data.nr_outputs; + c.prog_data.nr_outputs++; + } + else { + c.prog_data.output_edgeflag = ~0; + } + + + if (1) + tgsi_dump(c.vp->tokens, 0); + + /* Emit GEN4 code. + */ + brw_vs_emit(&c); + + /* get the program + */ + ret = brw_get_program(&c.func, &program, &program_size); + if (ret) + return ret; + + ret = brw_upload_cache( &brw->cache, BRW_VS_PROG, + &c.key, brw_vs_prog_key_size(&c.key), + NULL, 0, + program, program_size, + &c.prog_data, + &brw->vs.prog_data, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + + +static enum pipe_error brw_upload_vs_prog(struct brw_context *brw) +{ + struct brw_vs_prog_key key; + struct brw_vertex_shader *vp = brw->curr.vertex_shader; + struct brw_fs_signature *sig = &brw->curr.fragment_shader->signature; + enum pipe_error ret; + + memset(&key, 0, sizeof(key)); + + key.program_string_id = vp->id; + key.nr_userclip = brw->curr.ucp.nr; + key.copy_edgeflag = (brw->curr.rast->templ.fill_ccw != PIPE_POLYGON_MODE_FILL || + brw->curr.rast->templ.fill_cw != PIPE_POLYGON_MODE_FILL); + + memcpy(&key.fs_signature, sig, brw_fs_signature_size(sig)); + + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache, BRW_VS_PROG, + &key, brw_vs_prog_key_size(&key), + NULL, 0, + &brw->vs.prog_data, + &brw->vs.prog_bo)) + return PIPE_OK; + + ret = do_vs_prog(brw, vp, &key, &brw->vs.prog_bo); + if (ret) + return ret; + + return PIPE_OK; +} + + +/* See brw_vs.c: + */ +const struct brw_tracked_state brw_vs_prog = { + .dirty = { + .mesa = (PIPE_NEW_CLIP | + PIPE_NEW_RAST | + PIPE_NEW_FRAGMENT_SIGNATURE), + .brw = BRW_NEW_VERTEX_PROGRAM, + .cache = 0 + }, + .prepare = brw_upload_vs_prog +}; diff --git a/src/gallium/drivers/i965/brw_vs.h b/src/gallium/drivers/i965/brw_vs.h new file mode 100644 index 00000000000..3d1598d02b9 --- /dev/null +++ b/src/gallium/drivers/i965/brw_vs.h @@ -0,0 +1,109 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_VS_H +#define BRW_VS_H + + +#include "brw_context.h" +#include "brw_eu.h" + + +struct brw_vs_prog_key { + GLuint program_string_id; + GLuint nr_userclip:4; + GLuint copy_edgeflag:1; + GLuint pad:26; + struct brw_fs_signature fs_signature; +}; + +#define brw_vs_prog_key_size(s) (offsetof(struct brw_vs_prog_key, fs_signature) + \ + brw_fs_signature_size(&(s)->fs_signature)) + + +#define MAX_IF_DEPTH 32 +#define MAX_LOOP_DEPTH 32 + +struct brw_vs_compile { + struct brw_compile func; + struct brw_vs_prog_key key; + struct brw_vs_prog_data prog_data; + struct brw_chipset chipset; + + struct brw_vertex_shader *vp; + + GLuint nr_inputs; + GLuint nr_outputs; + GLuint nr_immediates; + GLfloat immediate[128][4]; + + GLboolean copy_edgeflag; + + GLuint overflow_grf_start; + GLuint overflow_count; + + GLuint first_tmp; + GLuint last_tmp; + + struct brw_reg r0; + struct brw_reg r1; + struct brw_reg regs[TGSI_FILE_COUNT][128]; + struct brw_reg tmp; + struct brw_reg stack; + + struct { + GLboolean used_in_src; + struct brw_reg reg; + } output_regs[128]; + + struct brw_reg userplane[6]; + + /** we may need up to 3 constants per instruction (if use_const_buffer) */ + struct { + GLint index; + struct brw_reg reg; + } current_const[3]; + + struct brw_instruction *if_inst[MAX_IF_DEPTH]; + struct brw_instruction *loop_inst[MAX_LOOP_DEPTH]; + GLuint insn; + GLuint if_depth; + GLuint loop_depth; + GLuint end_offset; + + struct brw_indirect stack_index; +}; + + +void brw_vs_emit( struct brw_vs_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965/brw_vs_emit.c b/src/gallium/drivers/i965/brw_vs_emit.c new file mode 100644 index 00000000000..1d0fff0d9ea --- /dev/null +++ b/src/gallium/drivers/i965/brw_vs_emit.c @@ -0,0 +1,1673 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "pipe/p_shader_tokens.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_info.h" + +#include "brw_context.h" +#include "brw_vs.h" +#include "brw_debug.h" +#include "brw_disasm.h" + +/* Choose one of the 4 vec4's which can be packed into each 16-wide reg. + */ +static INLINE struct brw_reg brw_vec4_grf_repeat( GLuint reg, GLuint slot ) +{ + int nr = reg + slot/2; + int subnr = (slot%2) * 4; + + return stride(brw_vec4_grf(nr, subnr), 0, 4, 1); +} + + +static struct brw_reg get_tmp( struct brw_vs_compile *c ) +{ + struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + +static void release_tmps( struct brw_vs_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +static boolean is_position_output( struct brw_vs_compile *c, + unsigned vs_output ) +{ + struct brw_vertex_shader *vs = c->vp; + + if (vs_output == c->prog_data.output_edgeflag) { + return FALSE; + } + else { + unsigned semantic = vs->info.output_semantic_name[vs_output]; + unsigned index = vs->info.output_semantic_index[vs_output]; + + return (semantic == TGSI_SEMANTIC_POSITION && + index == 0); + } +} + + +static boolean find_output_slot( struct brw_vs_compile *c, + unsigned vs_output, + unsigned *fs_input_slot ) +{ + struct brw_vertex_shader *vs = c->vp; + + if (vs_output == c->prog_data.output_edgeflag) { + *fs_input_slot = c->key.fs_signature.nr_inputs; + return TRUE; + } + else { + unsigned semantic = vs->info.output_semantic_name[vs_output]; + unsigned index = vs->info.output_semantic_index[vs_output]; + unsigned i; + + for (i = 0; i < c->key.fs_signature.nr_inputs; i++) { + if (c->key.fs_signature.input[i].semantic == semantic && + c->key.fs_signature.input[i].semantic_index == index) { + *fs_input_slot = i; + return TRUE; + } + } + } + + return FALSE; +} + + +/** + * Preallocate GRF register before code emit. + * Do things as simply as possible. Allocate and populate all regs + * ahead of time. + */ +static void brw_vs_alloc_regs( struct brw_vs_compile *c ) +{ + GLuint i, reg = 0, subreg = 0, mrf; + int attributes_in_vue; + + /* Determine whether to use a real constant buffer or use a block + * of GRF registers for constants. The later is faster but only + * works if everything fits in the GRF. + * XXX this heuristic/check may need some fine tuning... + */ + if (c->vp->info.file_max[TGSI_FILE_CONSTANT] + 1 + + c->vp->info.file_max[TGSI_FILE_IMMEDIATE] + 1 + + c->vp->info.file_max[TGSI_FILE_TEMPORARY] + 1 + 21 > BRW_MAX_GRF) + c->vp->use_const_buffer = GL_TRUE; + else { + /* XXX: immediates can go elsewhere if necessary: + */ + assert(c->vp->info.file_max[TGSI_FILE_IMMEDIATE] + 1 + + c->vp->info.file_max[TGSI_FILE_TEMPORARY] + 1 + 21 <= BRW_MAX_GRF); + + c->vp->use_const_buffer = GL_FALSE; + } + + /*printf("use_const_buffer = %d\n", c->vp->use_const_buffer);*/ + + /* r0 -- reserved as usual + */ + c->r0 = brw_vec8_grf(reg, 0); + reg++; + + /* User clip planes from curbe: + */ + if (c->key.nr_userclip) { + /* Skip over fixed planes: Or never read them into vs unit? + */ + subreg += 6; + + for (i = 0; i < c->key.nr_userclip; i++, subreg++) { + c->userplane[i] = + stride( brw_vec4_grf(reg+subreg/2, (subreg%2) * 4), 0, 4, 1); + } + + /* Deal with curbe alignment: + */ + subreg = align(subreg, 2); + /*reg += ((6 + c->key.nr_userclip + 3) / 4) * 2;*/ + } + + + /* Immediates: always in the curbe. + * + * XXX: Can try to encode some immediates as brw immediates + * XXX: Make sure ureg sets minimal immediate size and respect it + * here. + */ + for (i = 0; i < c->vp->info.immediate_count; i++, subreg++) { + c->regs[TGSI_FILE_IMMEDIATE][i] = + stride( brw_vec4_grf(reg+subreg/2, (subreg%2) * 4), 0, 4, 1); + } + c->prog_data.nr_params = c->vp->info.immediate_count * 4; + + + /* Vertex constant buffer. + * + * Constants from the buffer can be either cached in the curbe or + * loaded as needed from the actual constant buffer. + */ + if (!c->vp->use_const_buffer) { + GLuint nr_params = c->vp->info.file_max[TGSI_FILE_CONSTANT] + 1; + + for (i = 0; i < nr_params; i++, subreg++) { + c->regs[TGSI_FILE_CONSTANT][i] = + stride( brw_vec4_grf(reg+subreg/2, (subreg%2) * 4), 0, 4, 1); + } + + c->prog_data.nr_params += nr_params * 4; + } + + /* All regs allocated + */ + reg += (subreg + 1) / 2; + c->prog_data.curb_read_length = reg - 1; + + + /* Allocate input regs: + */ + c->nr_inputs = c->vp->info.num_inputs; + for (i = 0; i < c->nr_inputs; i++) { + c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } + + /* If there are no inputs, we'll still be reading one attribute's worth + * because it's required -- see urb_read_length setting. + */ + if (c->nr_inputs == 0) + reg++; + + + + /* Allocate outputs. The non-position outputs go straight into message regs. + */ + c->nr_outputs = c->prog_data.nr_outputs; + + if (c->chipset.is_igdng) + mrf = 8; + else + mrf = 4; + + + if (c->key.fs_signature.nr_inputs > BRW_MAX_MRF) { + c->overflow_grf_start = reg; + c->overflow_count = c->key.fs_signature.nr_inputs - BRW_MAX_MRF; + reg += c->overflow_count; + } + + /* XXX: need to access vertex output semantics here: + */ + for (i = 0; i < c->nr_outputs; i++) { + unsigned slot; + + /* XXX: Put output position in slot zero always. Clipper, etc, + * need access to this reg. + */ + if (is_position_output(c, i)) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); /* copy to mrf 0 */ + reg++; + } + else if (find_output_slot(c, i, &slot)) { + + if (0 /* is_psize_output(c, i) */ ) { + /* c->psize_out.grf = reg; */ + /* c->psize_out.mrf = i; */ + } + + /* The first (16-4) outputs can go straight into the message regs. + */ + if (slot + mrf < BRW_MAX_MRF) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(slot + mrf); + } + else { + int grf = c->overflow_grf_start + slot - BRW_MAX_MRF; + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(grf, 0); + } + } + else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_null_reg(); + } + } + + /* Allocate program temporaries: + */ + + for (i = 0; i < c->vp->info.file_max[TGSI_FILE_TEMPORARY]+1; i++) { + c->regs[TGSI_FILE_TEMPORARY][i] = brw_vec8_grf(reg, 0); + reg++; + } + + /* Address reg(s). Don't try to use the internal address reg until + * deref time. + */ + for (i = 0; i < c->vp->info.file_max[TGSI_FILE_ADDRESS]+1; i++) { + c->regs[TGSI_FILE_ADDRESS][i] = brw_reg(BRW_GENERAL_REGISTER_FILE, + reg, + 0, + BRW_REGISTER_TYPE_D, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XXXX, + BRW_WRITEMASK_X); + reg++; + } + + if (c->vp->use_const_buffer) { + for (i = 0; i < 3; i++) { + c->current_const[i].index = -1; + c->current_const[i].reg = brw_vec8_grf(reg, 0); + reg++; + } + } + +#if 0 + for (i = 0; i < 128; i++) { + if (c->output_regs[i].used_in_src) { + c->output_regs[i].reg = brw_vec8_grf(reg, 0); + reg++; + } + } +#endif + + if (c->vp->has_flow_control) { + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0); + reg += 2; + } + + /* Some opcodes need an internal temporary: + */ + c->first_tmp = reg; + c->last_tmp = reg; /* for allocation purposes */ + + /* Each input reg holds data from two vertices. The + * urb_read_length is the number of registers read from *each* + * vertex urb, so is half the amount: + */ + c->prog_data.urb_read_length = (c->nr_inputs + 1) / 2; + + /* Setting this field to 0 leads to undefined behavior according to the + * the VS_STATE docs. Our VUEs will always have at least one attribute + * sitting in them, even if it's padding. + */ + if (c->prog_data.urb_read_length == 0) + c->prog_data.urb_read_length = 1; + + /* The VS VUEs are shared by VF (outputting our inputs) and VS, so size + * them to fit the biggest thing they need to. + */ + attributes_in_vue = MAX2(c->nr_outputs, c->nr_inputs); + + if (c->chipset.is_igdng) + c->prog_data.urb_entry_size = (attributes_in_vue + 6 + 3) / 4; + else + c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4; + + c->prog_data.total_grf = reg; + + if (BRW_DEBUG & DEBUG_VS) { + debug_printf("%s NumAddrRegs %d\n", __FUNCTION__, + c->vp->info.file_max[TGSI_FILE_ADDRESS]+1); + debug_printf("%s NumTemps %d\n", __FUNCTION__, + c->vp->info.file_max[TGSI_FILE_TEMPORARY]+1); + debug_printf("%s reg = %d\n", __FUNCTION__, reg); + } +} + + +/** + * If an instruction uses a temp reg both as a src and the dest, we + * sometimes need to allocate an intermediate temporary. + */ +static void unalias1( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg )) +{ + if (dst.file == arg0.file && dst.nr == arg0.nr) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0); + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } + else { + func(c, dst, arg0); + } +} + +/** + * \sa unalias2 + * Checkes if 2-operand instruction needs an intermediate temporary. + */ +static void unalias2( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg )) +{ + if ((dst.file == arg0.file && dst.nr == arg0.nr) || + (dst.file == arg1.file && dst.nr == arg1.nr)) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0, arg1); + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } + else { + func(c, dst, arg0, arg1); + } +} + +/** + * \sa unalias2 + * Checkes if 3-operand instruction needs an intermediate temporary. + */ +static void unalias3( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + struct brw_reg arg2, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg, + struct brw_reg )) +{ + if ((dst.file == arg0.file && dst.nr == arg0.nr) || + (dst.file == arg1.file && dst.nr == arg1.nr) || + (dst.file == arg2.file && dst.nr == arg2.nr)) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0, arg1, arg2); + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } + else { + func(c, dst, arg0, arg1, arg2); + } +} + +static void emit_sop( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + GLuint cond) +{ + brw_MOV(p, dst, brw_imm_f(0.0f)); + brw_CMP(p, brw_null_reg(), cond, arg0, arg1); + brw_MOV(p, dst, brw_imm_f(1.0f)); + brw_set_predicate_control_flag_value(p, 0xff); +} + +static void emit_seq( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ); +} + +static void emit_sne( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); +} +static void emit_slt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L); +} + +static void emit_sle( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE); +} + +static void emit_sgt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G); +} + +static void emit_sge( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE); +} + +static void emit_max( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg1, arg0); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + +static void emit_min( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg0, arg1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + + +static void emit_math1( struct brw_vs_compile *c, + GLuint function, + struct brw_reg dst, + struct brw_reg arg0, + GLuint precision) +{ + /* There are various odd behaviours with SEND on the simulator. In + * addition there are documented issues with the fact that the GEN4 + * processor doesn't do dependency control properly on SEND + * results. So, on balance, this kludge to get around failures + * with writemasked math results looks like it might be necessary + * whether that turns out to be a simulator bug or not: + */ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + GLboolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + +static void emit_math2( struct brw_vs_compile *c, + GLuint function, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + GLuint precision) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + GLboolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_message_reg(3), arg1); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + +static void emit_exp_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_X) { + struct brw_reg tmp = get_tmp(c); + struct brw_reg tmp_d = retype(tmp, BRW_REGISTER_TYPE_D); + + /* tmp_d = floor(arg0.x) */ + brw_RNDD(p, tmp_d, brw_swizzle1(arg0, 0)); + + /* result[0] = 2.0 ^ tmp */ + + /* Adjust exponent for floating point: + * exp += 127 + */ + brw_ADD(p, brw_writemask(tmp_d, BRW_WRITEMASK_X), tmp_d, brw_imm_d(127)); + + /* Install exponent and sign. + * Excess drops off the edge: + */ + brw_SHL(p, brw_writemask(retype(dst, BRW_REGISTER_TYPE_D), BRW_WRITEMASK_X), + tmp_d, brw_imm_d(23)); + + release_tmp(c, tmp); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_Y) { + /* result[1] = arg0.x - floor(arg0.x) */ + brw_FRC(p, brw_writemask(dst, BRW_WRITEMASK_Y), brw_swizzle1(arg0, 0)); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_Z) { + /* As with the LOG instruction, we might be better off just + * doing a taylor expansion here, seeing as we have to do all + * the prep work. + * + * If mathbox partial precision is too low, consider also: + * result[3] = result[0] * EXP(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_EXP, + brw_writemask(dst, BRW_WRITEMASK_Z), + brw_swizzle1(arg0, 0), + BRW_MATH_PRECISION_FULL); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_W), brw_imm_f(1)); + } +} + + +static void emit_log_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + struct brw_reg tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + struct brw_reg arg0_ud = retype(arg0, BRW_REGISTER_TYPE_UD); + GLboolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) { + tmp = get_tmp(c); + tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + } + + /* Perform mant = frexpf(fabsf(x), &exp), adjust exp and mnt + * according to spec: + * + * These almost look likey they could be joined up, but not really + * practical: + * + * result[0].f = (x.i & ((1<<31)-1) >> 23) - 127 + * result[1].i = (x.i & ((1<<23)-1) + (127<<23) + */ + if (dst.dw1.bits.writemask & BRW_WRITEMASK_XZ) { + brw_AND(p, + brw_writemask(tmp_ud, BRW_WRITEMASK_X), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1U<<31)-1)); + + brw_SHR(p, + brw_writemask(tmp_ud, BRW_WRITEMASK_X), + tmp_ud, + brw_imm_ud(23)); + + brw_ADD(p, + brw_writemask(tmp, BRW_WRITEMASK_X), + retype(tmp_ud, BRW_REGISTER_TYPE_D), /* does it matter? */ + brw_imm_d(-127)); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_YZ) { + brw_AND(p, + brw_writemask(tmp_ud, BRW_WRITEMASK_Y), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1<<23)-1)); + + brw_OR(p, + brw_writemask(tmp_ud, BRW_WRITEMASK_Y), + tmp_ud, + brw_imm_ud(127<<23)); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_Z) { + /* result[2] = result[0] + LOG2(result[1]); */ + + /* Why bother? The above is just a hint how to do this with a + * taylor series. Maybe we *should* use a taylor series as by + * the time all the above has been done it's almost certainly + * quicker than calling the mathbox, even with low precision. + * + * Options are: + * - result[0] + mathbox.LOG2(result[1]) + * - mathbox.LOG2(arg0.x) + * - result[0] + inline_taylor_approx(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_LOG, + brw_writemask(tmp, BRW_WRITEMASK_Z), + brw_swizzle1(tmp, 1), + BRW_MATH_PRECISION_FULL); + + brw_ADD(p, + brw_writemask(tmp, BRW_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(tmp, 0)); + } + + if (dst.dw1.bits.writemask & BRW_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(tmp, BRW_WRITEMASK_W), brw_imm_f(1)); + } + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + +/* Need to unalias - consider swizzles: r0 = DST r0.xxxx r1 + */ +static void emit_dst_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1) +{ + struct brw_compile *p = &c->func; + + /* There must be a better way to do this: + */ + if (dst.dw1.bits.writemask & BRW_WRITEMASK_X) + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_X), brw_imm_f(1.0)); + if (dst.dw1.bits.writemask & BRW_WRITEMASK_Y) + brw_MUL(p, brw_writemask(dst, BRW_WRITEMASK_Y), arg0, arg1); + if (dst.dw1.bits.writemask & BRW_WRITEMASK_Z) + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_Z), arg0); + if (dst.dw1.bits.writemask & BRW_WRITEMASK_W) + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_W), arg1); +} + + +static void emit_xpd( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg t, + struct brw_reg u) +{ + brw_MUL(p, brw_null_reg(), brw_swizzle(t, 1,2,0,3), brw_swizzle(u,2,0,1,3)); + brw_MAC(p, dst, negate(brw_swizzle(t, 2,0,1,3)), brw_swizzle(u,1,2,0,3)); +} + + +static void emit_lit_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + struct brw_reg tmp = dst; + GLboolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_YZ), brw_imm_f(0)); + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_XW), brw_imm_f(1)); + + /* Need to use BRW_EXECUTE_8 and also do an 8-wide compare in order + * to get all channels active inside the IF. In the clipping code + * we run with NoMask, so it's not an option and we can use + * BRW_EXECUTE_1 for all comparisions. + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,0), brw_imm_f(0)); + if_insn = brw_IF(p, BRW_EXECUTE_8); + { + brw_MOV(p, brw_writemask(dst, BRW_WRITEMASK_Y), brw_swizzle1(arg0,0)); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,1), brw_imm_f(0)); + brw_MOV(p, brw_writemask(tmp, BRW_WRITEMASK_Z), brw_swizzle1(arg0,1)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + emit_math2(c, + BRW_MATH_FUNCTION_POW, + brw_writemask(dst, BRW_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(arg0, 3), + BRW_MATH_PRECISION_PARTIAL); + } + + brw_ENDIF(p, if_insn); + + release_tmp(c, tmp); +} + +static void emit_lrp_noalias(struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + struct brw_reg arg2) +{ + struct brw_compile *p = &c->func; + + brw_ADD(p, dst, negate(arg0), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst, arg2); + brw_MAC(p, dst, arg0, arg1); +} + +/** 3 or 4-component vector normalization */ +static void emit_nrm( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + int num_comps) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + + /* tmp = dot(arg0, arg0) */ + if (num_comps == 3) + brw_DP3(p, tmp, arg0, arg0); + else + brw_DP4(p, tmp, arg0, arg0); + + /* tmp = 1 / sqrt(tmp) */ + emit_math1(c, BRW_MATH_FUNCTION_RSQ, tmp, tmp, BRW_MATH_PRECISION_FULL); + + /* dst = arg0 * tmp */ + brw_MUL(p, dst, arg0, tmp); + + release_tmp(c, tmp); +} + + +static struct brw_reg +get_constant(struct brw_vs_compile *c, + GLuint argIndex, + GLuint index, + GLboolean relAddr) +{ + struct brw_compile *p = &c->func; + struct brw_reg const_reg; + struct brw_reg const2_reg; + + assert(argIndex < 3); + + if (c->current_const[argIndex].index != index || relAddr) { + struct brw_reg addrReg = c->regs[TGSI_FILE_ADDRESS][0]; + + c->current_const[argIndex].index = index; + +#if 0 + printf(" fetch const[%d] for arg %d into reg %d\n", + src.Index, argIndex, c->current_const[argIndex].reg.nr); +#endif + /* need to fetch the constant now */ + brw_dp_READ_4_vs(p, + c->current_const[argIndex].reg,/* writeback dest */ + 0, /* oword */ + relAddr, /* relative indexing? */ + addrReg, /* address register */ + 16 * index, /* byte offset */ + SURF_INDEX_VERT_CONST_BUFFER /* binding table index */ + ); + + if (relAddr) { + /* second read */ + const2_reg = get_tmp(c); + + /* use upper half of address reg for second read */ + addrReg = stride(addrReg, 0, 4, 0); + addrReg.subnr = 16; + + brw_dp_READ_4_vs(p, + const2_reg, /* writeback dest */ + 1, /* oword */ + relAddr, /* relative indexing? */ + addrReg, /* address register */ + 16 * index, /* byte offset */ + SURF_INDEX_VERT_CONST_BUFFER + ); + } + } + + const_reg = c->current_const[argIndex].reg; + + if (relAddr) { + /* merge the two Owords into the constant register */ + /* const_reg[7..4] = const2_reg[7..4] */ + brw_MOV(p, + suboffset(stride(const_reg, 0, 4, 1), 4), + suboffset(stride(const2_reg, 0, 4, 1), 4)); + release_tmp(c, const2_reg); + } + else { + /* replicate lower four floats into upper half (to get XYZWXYZW) */ + const_reg = stride(const_reg, 0, 4, 0); + const_reg.subnr = 0; + } + + return const_reg; +} + + + +/* TODO: relative addressing! + */ +static struct brw_reg get_reg( struct brw_vs_compile *c, + enum tgsi_file_type file, + GLuint index ) +{ + switch (file) { + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + case TGSI_FILE_CONSTANT: + assert(c->regs[file][index].nr != 0); + return c->regs[file][index]; + + case TGSI_FILE_ADDRESS: + assert(index == 0); + return c->regs[file][index]; + + case TGSI_FILE_NULL: /* undef values */ + return brw_null_reg(); + + default: + assert(0); + return brw_null_reg(); + } +} + + +/** + * Indirect addressing: get reg[[arg] + offset]. + */ +static struct brw_reg deref( struct brw_vs_compile *c, + struct brw_reg arg, + GLint offset) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = vec4(get_tmp(c)); + struct brw_reg addr_reg = c->regs[TGSI_FILE_ADDRESS][0]; + struct brw_reg vp_address = retype(vec1(addr_reg), BRW_REGISTER_TYPE_UW); + GLuint byte_offset = arg.nr * 32 + arg.subnr + offset * 16; + struct brw_reg indirect = brw_vec4_indirect(0,0); + + { + brw_push_insn_state(p); + brw_set_access_mode(p, BRW_ALIGN_1); + + /* This is pretty clunky - load the address register twice and + * fetch each 4-dword value in turn. There must be a way to do + * this in a single pass, but I couldn't get it to work. + */ + brw_ADD(p, brw_address_reg(0), vp_address, brw_imm_d(byte_offset)); + brw_MOV(p, tmp, indirect); + + brw_ADD(p, brw_address_reg(0), suboffset(vp_address, 8), brw_imm_d(byte_offset)); + brw_MOV(p, suboffset(tmp, 4), indirect); + + brw_pop_insn_state(p); + } + + /* NOTE: tmp not released */ + return vec8(tmp); +} + + +/** + * Get brw reg corresponding to the instruction's [argIndex] src reg. + * TODO: relative addressing! + */ +static struct brw_reg +get_src_reg( struct brw_vs_compile *c, + GLuint argIndex, + GLuint file, + GLint index, + GLboolean relAddr ) +{ + + switch (file) { + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + if (relAddr) { + return deref(c, c->regs[file][0], index); + } + else { + assert(c->regs[file][index].nr != 0); + return c->regs[file][index]; + } + + case TGSI_FILE_IMMEDIATE: + return c->regs[file][index]; + + case TGSI_FILE_CONSTANT: + if (c->vp->use_const_buffer) { + return get_constant(c, argIndex, index, relAddr); + } + else if (relAddr) { + return deref(c, c->regs[TGSI_FILE_CONSTANT][0], index); + } + else { + assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index]; + } + case TGSI_FILE_ADDRESS: + assert(index == 0); + return c->regs[file][index]; + + case TGSI_FILE_NULL: + /* this is a normal case since we loop over all three src args */ + return brw_null_reg(); + + default: + assert(0); + return brw_null_reg(); + } +} + + +static void emit_arl( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + GLboolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_RNDD(p, tmp, arg0); /* tmp = round(arg0) */ + brw_MUL(p, dst, tmp, brw_imm_d(16)); /* dst = tmp * 16 */ + + if (need_tmp) + release_tmp(c, tmp); +} + + +/** + * Return the brw reg for the given instruction's src argument. + */ +static struct brw_reg get_arg( struct brw_vs_compile *c, + const struct tgsi_full_src_register *src, + GLuint argIndex ) +{ + struct brw_reg reg; + + if (src->Register.File == TGSI_FILE_NULL) + return brw_null_reg(); + + reg = get_src_reg(c, argIndex, + src->Register.File, + src->Register.Index, + src->Register.Indirect); + + /* Convert 3-bit swizzle to 2-bit. + */ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(src->Register.SwizzleX, + src->Register.SwizzleY, + src->Register.SwizzleZ, + src->Register.SwizzleW); + + reg.negate = src->Register.Negate ? 1 : 0; + + /* XXX: abs, absneg + */ + + return reg; +} + + +/** + * Get brw register for the given program dest register. + */ +static struct brw_reg get_dst( struct brw_vs_compile *c, + unsigned file, + unsigned index, + unsigned writemask ) +{ + struct brw_reg reg; + + switch (file) { + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_OUTPUT: + assert(c->regs[file][index].nr != 0); + reg = c->regs[file][index]; + break; + case TGSI_FILE_ADDRESS: + assert(index == 0); + reg = c->regs[file][index]; + break; + case TGSI_FILE_NULL: + /* we may hit this for OPCODE_END, OPCODE_KIL, etc */ + reg = brw_null_reg(); + break; + default: + assert(0); + reg = brw_null_reg(); + } + + reg.dw1.bits.writemask = writemask; + + return reg; +} + + + + +/** + * Post-vertex-program processing. Send the results to the URB. + */ +static void emit_vertex_write( struct brw_vs_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_reg m0 = brw_message_reg(0); + struct brw_reg pos = c->regs[TGSI_FILE_OUTPUT][VERT_RESULT_HPOS]; + struct brw_reg ndc; + int eot; + int i; + GLuint len_vertext_header = 2; + + if (c->key.copy_edgeflag) { + brw_MOV(p, + get_reg(c, TGSI_FILE_OUTPUT, c->prog_data.output_edgeflag), + brw_imm_f(1)); + } + + /* Build ndc coords */ + ndc = get_tmp(c); + /* ndc = 1.0 / pos.w */ + emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); + /* ndc.xyz = pos * ndc */ + brw_MUL(p, brw_writemask(ndc, BRW_WRITEMASK_XYZ), pos, ndc); + + /* Update the header for point size, user clipping flags, and -ve rhw + * workaround. + */ + if (c->prog_data.writes_psiz || + c->key.nr_userclip || + c->chipset.is_965) + { + struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + GLuint i; + + brw_MOV(p, header1, brw_imm_ud(0)); + + brw_set_access_mode(p, BRW_ALIGN_16); + + if (c->prog_data.writes_psiz) { + struct brw_reg psiz = c->regs[TGSI_FILE_OUTPUT][VERT_RESULT_PSIZ]; + brw_MUL(p, brw_writemask(header1, BRW_WRITEMASK_W), brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); + brw_AND(p, brw_writemask(header1, BRW_WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8)); + } + + for (i = 0; i < c->key.nr_userclip; i++) { + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, brw_null_reg(), pos, c->userplane[i]); + brw_OR(p, brw_writemask(header1, BRW_WRITEMASK_W), header1, brw_imm_ud(1<<i)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + + /* i965 clipping workaround: + * 1) Test for -ve rhw + * 2) If set, + * set ndc = (0,0,0,0) + * set ucp[6] = 1 + * + * Later, clipping will detect ucp[6] and ensure the primitive is + * clipped against all fixed planes. + */ + if (c->chipset.is_965) { + brw_CMP(p, + vec8(brw_null_reg()), + BRW_CONDITIONAL_L, + brw_swizzle1(ndc, 3), + brw_imm_f(0)); + + brw_OR(p, brw_writemask(header1, BRW_WRITEMASK_W), header1, brw_imm_ud(1<<6)); + brw_MOV(p, ndc, brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + + brw_set_access_mode(p, BRW_ALIGN_1); /* why? */ + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), header1); + brw_set_access_mode(p, BRW_ALIGN_16); + + release_tmp(c, header1); + } + else { + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0)); + } + + /* Emit the (interleaved) headers for the two vertices - an 8-reg + * of zeros followed by two sets of NDC coordinates: + */ + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, offset(m0, 2), ndc); + + if (c->chipset.is_igdng) { + /* There are 20 DWs (D0-D19) in VUE vertex header on IGDNG */ + brw_MOV(p, offset(m0, 3), pos); /* a portion of vertex header */ + /* m4, m5 contain the distances from vertex to the user clip planeXXX. + * Seems it is useless for us. + * m6 is used for aligning, so that the remainder of vertex element is + * reg-aligned. + */ + brw_MOV(p, offset(m0, 7), pos); /* the remainder of vertex element */ + len_vertext_header = 6; + } else { + brw_MOV(p, offset(m0, 3), pos); + len_vertext_header = 2; + } + + eot = (c->overflow_count == 0); + + brw_urb_WRITE(p, + brw_null_reg(), /* dest */ + 0, /* starting mrf reg nr */ + c->r0, /* src */ + 0, /* allocate */ + 1, /* used */ + MIN2(c->nr_outputs + 1 + len_vertext_header, (BRW_MAX_MRF-1)), /* msg len */ + 0, /* response len */ + eot, /* eot */ + eot, /* writes complete */ + 0, /* urb destination offset */ + BRW_URB_SWIZZLE_INTERLEAVE); + + /* Not all of the vertex outputs/results fit into the MRF. + * Move the overflowed attributes from the GRF to the MRF and + * issue another brw_urb_WRITE(). + */ + for (i = 0; i < c->overflow_count; i += BRW_MAX_MRF) { + unsigned nr = MIN2(c->overflow_count - i, BRW_MAX_MRF); + GLuint j; + + eot = (i + nr >= c->overflow_count); + + /* XXX I'm not 100% sure about which MRF regs to use here. Starting + * at mrf[4] atm... + */ + for (j = 0; j < nr; j++) { + brw_MOV(p, brw_message_reg(4+j), + brw_vec8_grf(c->overflow_grf_start + i + j, 0)); + } + + brw_urb_WRITE(p, + brw_null_reg(), /* dest */ + 4, /* starting mrf reg nr */ + c->r0, /* src */ + 0, /* allocate */ + 1, /* used */ + nr+1, /* msg len */ + 0, /* response len */ + eot, /* eot */ + eot, /* writes complete */ + i-1, /* urb destination offset */ + BRW_URB_SWIZZLE_INTERLEAVE); + } +} + + +/** + * Called after code generation to resolve subroutine calls and the + * END instruction. + * \param end_inst points to brw code for END instruction + * \param last_inst points to last instruction emitted before vertex write + */ +static void +post_vs_emit( struct brw_vs_compile *c, + struct brw_instruction *end_inst, + struct brw_instruction *last_inst ) +{ + GLint offset; + + brw_resolve_cals(&c->func); + + /* patch up the END code to jump past subroutines, etc */ + offset = last_inst - end_inst; + if (offset > 1) { + brw_set_src1(end_inst, brw_imm_d(offset * 16)); + } else { + end_inst->header.opcode = BRW_OPCODE_NOP; + } +} + +static uint32_t +get_predicate(const struct tgsi_full_instruction *inst) +{ + /* XXX: disabling for now + */ +#if 0 + if (inst->dst.CondMask == COND_TR) + return BRW_PREDICATE_NONE; + + /* All of GLSL only produces predicates for COND_NE and one channel per + * vector. Fail badly if someone starts doing something else, as it might + * mean infinite looping or something. + * + * We'd like to support all the condition codes, but our hardware doesn't + * quite match the Mesa IR, which is modeled after the NV extensions. For + * those, the instruction may update the condition codes or not, then any + * later instruction may use one of those condition codes. For gen4, the + * instruction may update the flags register based on one of the condition + * codes output by the instruction, and then further instructions may + * predicate on that. We can probably support this, but it won't + * necessarily be easy. + */ +/* assert(inst->dst.CondMask == COND_NE); */ + + switch (inst->dst.CondSwizzle) { + case SWIZZLE_XXXX: + return BRW_PREDICATE_ALIGN16_REPLICATE_X; + case SWIZZLE_YYYY: + return BRW_PREDICATE_ALIGN16_REPLICATE_Y; + case SWIZZLE_ZZZZ: + return BRW_PREDICATE_ALIGN16_REPLICATE_Z; + case SWIZZLE_WWWW: + return BRW_PREDICATE_ALIGN16_REPLICATE_W; + default: + debug_printf("Unexpected predicate: 0x%08x\n", + inst->dst.CondMask); + return BRW_PREDICATE_NORMAL; + } +#else + return BRW_PREDICATE_NORMAL; +#endif +} + +static void emit_insn(struct brw_vs_compile *c, + const struct tgsi_full_instruction *inst) +{ + unsigned opcode = inst->Instruction.Opcode; + unsigned label = inst->Label.Label; + struct brw_compile *p = &c->func; + struct brw_reg args[3], dst; + GLuint i; + +#if 0 + printf("%d: ", insn); + _mesa_print_instruction(inst); +#endif + + /* Get argument regs. + */ + for (i = 0; i < 3; i++) { + args[i] = get_arg(c, &inst->Src[i], i); + } + + /* Get dest regs. Note that it is possible for a reg to be both + * dst and arg, given the static allocation of registers. So + * care needs to be taken emitting multi-operation instructions. + */ + dst = get_dst(c, + inst->Dst[0].Register.File, + inst->Dst[0].Register.Index, + inst->Dst[0].Register.WriteMask); + + /* XXX: saturate + */ + if (inst->Instruction.Saturate != TGSI_SAT_NONE) { + debug_printf("Unsupported saturate in vertex shader"); + } + + switch (opcode) { + case TGSI_OPCODE_ABS: + brw_MOV(p, dst, brw_abs(args[0])); + break; + case TGSI_OPCODE_ADD: + brw_ADD(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_COS: + emit_math1(c, BRW_MATH_FUNCTION_COS, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_DP3: + brw_DP3(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP4: + brw_DP4(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DPH: + brw_DPH(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_NRM: + emit_nrm(c, dst, args[0], 3); + break; + case TGSI_OPCODE_NRM4: + emit_nrm(c, dst, args[0], 4); + break; + case TGSI_OPCODE_DST: + unalias2(c, dst, args[0], args[1], emit_dst_noalias); + break; + case TGSI_OPCODE_EXP: + unalias1(c, dst, args[0], emit_exp_noalias); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_ARL: + emit_arl(c, dst, args[0]); + break; + case TGSI_OPCODE_FLR: + brw_RNDD(p, dst, args[0]); + break; + case TGSI_OPCODE_FRC: + brw_FRC(p, dst, args[0]); + break; + case TGSI_OPCODE_LOG: + unalias1(c, dst, args[0], emit_log_noalias); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_LIT: + unalias1(c, dst, args[0], emit_lit_noalias); + break; + case TGSI_OPCODE_LRP: + unalias3(c, dst, args[0], args[1], args[2], emit_lrp_noalias); + break; + case TGSI_OPCODE_MAD: + brw_MOV(p, brw_acc_reg(), args[2]); + brw_MAC(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MAX: + emit_max(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MIN: + emit_min(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MOV: + brw_MOV(p, dst, args[0]); + break; + case TGSI_OPCODE_MUL: + brw_MUL(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_POW: + emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, + brw_swizzle(args[0], 0,0,0,0), BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SIN: + emit_math1(c, BRW_MATH_FUNCTION_SIN, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLT: + emit_slt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLE: + emit_sle(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SUB: + brw_ADD(p, dst, args[0], negate(args[1])); + break; + case TGSI_OPCODE_TRUNC: + /* round toward zero */ + brw_RNDZ(p, dst, args[0]); + break; + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_IF: + assert(c->if_depth < MAX_IF_DEPTH); + c->if_inst[c->if_depth] = brw_IF(p, BRW_EXECUTE_8); + /* Note that brw_IF smashes the predicate_control field. */ + c->if_inst[c->if_depth]->header.predicate_control = get_predicate(inst); + c->if_depth++; + break; + case TGSI_OPCODE_ELSE: + c->if_inst[c->if_depth-1] = brw_ELSE(p, c->if_inst[c->if_depth-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(c->if_depth > 0); + brw_ENDIF(p, c->if_inst[--c->if_depth]); + break; + case TGSI_OPCODE_BGNLOOP: + c->loop_inst[c->loop_depth++] = brw_DO(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_BRK: + brw_set_predicate_control(p, get_predicate(inst)); + brw_BREAK(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_CONT: + brw_set_predicate_control(p, get_predicate(inst)); + brw_CONT(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_ENDLOOP: + { + struct brw_instruction *inst0, *inst1; + GLuint br = 1; + + c->loop_depth--; + + if (c->chipset.is_igdng) + br = 2; + + inst0 = inst1 = brw_WHILE(p, c->loop_inst[c->loop_depth]); + /* patch all the BREAK/CONT instructions from last BEGINLOOP */ + while (inst0 > c->loop_inst[c->loop_depth]) { + inst0--; + if (inst0->header.opcode == TGSI_OPCODE_BRK) { + inst0->bits3.if_else.jump_count = br * (inst1 - inst0 + 1); + inst0->bits3.if_else.pop_count = 0; + } + else if (inst0->header.opcode == TGSI_OPCODE_CONT) { + inst0->bits3.if_else.jump_count = br * (inst1 - inst0); + inst0->bits3.if_else.pop_count = 0; + } + } + } + break; + case TGSI_OPCODE_BRA: + brw_set_predicate_control(p, get_predicate(inst)); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_CAL: + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1d(c->stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(4)); + brw_save_call(p, label, p->nr_insn); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_RET: + brw_ADD(p, get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1d(c->stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + break; + case TGSI_OPCODE_END: + c->end_offset = p->nr_insn; + /* this instruction will get patched later to jump past subroutine + * code, etc. + */ + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_BGNSUB: + brw_save_label(p, p->nr_insn, p->nr_insn); + break; + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + default: + debug_printf("Unsupported opcode %i (%s) in vertex shader", + opcode, + tgsi_get_opcode_name(opcode)); + } + + /* Set the predication update on the last instruction of the native + * instruction sequence. + * + * This would be problematic if it was set on a math instruction, + * but that shouldn't be the case with the current GLSL compiler. + */ +#if 0 + /* XXX: disabled + */ + if (inst->CondUpdate) { + struct brw_instruction *hw_insn = &p->store[p->nr_insn - 1]; + + assert(hw_insn->header.destreg__conditionalmod == 0); + hw_insn->header.destreg__conditionalmod = BRW_CONDITIONAL_NZ; + } +#endif + + release_tmps(c); +} + + +/* Emit the vertex program instructions here. + */ +void brw_vs_emit(struct brw_vs_compile *c) +{ + struct brw_compile *p = &c->func; + const struct tgsi_token *tokens = c->vp->tokens; + struct brw_instruction *end_inst, *last_inst; + struct tgsi_parse_context parse; + struct tgsi_full_instruction *inst; + + if (BRW_DEBUG & DEBUG_VS) + tgsi_dump(c->vp->tokens, 0); + + c->stack_index = brw_indirect(0, 0); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_access_mode(p, BRW_ALIGN_16); + + + /* Static register allocation + */ + brw_vs_alloc_regs(c); + + if (c->vp->has_flow_control) { + brw_MOV(p, get_addr_reg(c->stack_index), brw_address(c->stack)); + } + + /* Instructions + */ + tgsi_parse_init( &parse, tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + case TGSI_TOKEN_TYPE_IMMEDIATE: + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + inst = &parse.FullToken.FullInstruction; + emit_insn( c, inst ); + break; + + default: + assert( 0 ); + } + } + tgsi_parse_free( &parse ); + + end_inst = &p->store[c->end_offset]; + last_inst = &p->store[p->nr_insn]; + + /* The END instruction will be patched to jump to this code */ + emit_vertex_write(c); + + post_vs_emit(c, end_inst, last_inst); + + if (BRW_DEBUG & DEBUG_VS) { + debug_printf("vs-native:\n"); + brw_disasm(stderr, p->store, p->nr_insn); + } +} diff --git a/src/gallium/drivers/i965/brw_vs_state.c b/src/gallium/drivers/i965/brw_vs_state.c new file mode 100644 index 00000000000..dadbb622e4d --- /dev/null +++ b/src/gallium/drivers/i965/brw_vs_state.c @@ -0,0 +1,201 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + + +#include "brw_debug.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +struct brw_vs_unit_key { + unsigned int total_grf; + unsigned int urb_entry_read_length; + unsigned int curb_entry_read_length; + + unsigned int curbe_offset; + + unsigned int nr_urb_entries, urb_size; + + unsigned int nr_surfaces; +}; + +static void +vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key) +{ + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_VS_PROG */ + key->total_grf = brw->vs.prog_data->total_grf; + key->urb_entry_read_length = brw->vs.prog_data->urb_read_length; + key->curb_entry_read_length = brw->vs.prog_data->curb_read_length; + + /* BRW_NEW_URB_FENCE */ + key->nr_urb_entries = brw->urb.nr_vs_entries; + key->urb_size = brw->urb.vsize; + + /* BRW_NEW_NR_VS_SURFACES */ + key->nr_surfaces = brw->vs.nr_surfaces; + + /* PIPE_NEW_CLIP */ + if (brw->curr.ucp.nr) { + /* Note that we read in the userclip planes as well, hence + * clip_start: + */ + key->curbe_offset = brw->curbe.clip_start; + } + else { + key->curbe_offset = brw->curbe.vs_start; + } +} + +static enum pipe_error +vs_unit_create_from_key(struct brw_context *brw, + struct brw_vs_unit_key *key, + struct brw_winsys_reloc *reloc, + struct brw_winsys_buffer **bo_out) +{ + enum pipe_error ret; + struct brw_vs_unit_state vs; + int chipset_max_threads; + + memset(&vs, 0, sizeof(vs)); + + vs.thread0.kernel_start_pointer = 0; /* reloc */ + vs.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1; + vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + /* Choosing multiple program flow means that we may get 2-vertex threads, + * which will have the channel mask for dwords 4-7 enabled in the thread, + * and those dwords will be written to the second URB handle when we + * brw_urb_WRITE() results. + */ + vs.thread1.single_program_flow = 0; + + if (BRW_IS_IGDNG(brw)) + vs.thread1.binding_table_entry_count = 0; /* hardware requirement */ + else + vs.thread1.binding_table_entry_count = key->nr_surfaces; + + vs.thread3.urb_entry_read_length = key->urb_entry_read_length; + vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length; + vs.thread3.dispatch_grf_start_reg = 1; + vs.thread3.urb_entry_read_offset = 0; + vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; + + if (BRW_IS_IGDNG(brw)) + vs.thread4.nr_urb_entries = key->nr_urb_entries >> 2; + else + vs.thread4.nr_urb_entries = key->nr_urb_entries; + + vs.thread4.urb_entry_allocation_size = key->urb_size - 1; + + if (BRW_IS_IGDNG(brw)) + chipset_max_threads = 72; + else if (BRW_IS_G4X(brw)) + chipset_max_threads = 32; + else + chipset_max_threads = 16; + + vs.thread4.max_threads = CLAMP(key->nr_urb_entries / 2, + 1, chipset_max_threads) - 1; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + vs.thread4.max_threads = 0; + + /* No samplers for ARB_vp programs: + */ + /* It has to be set to 0 for IGDNG + */ + vs.vs5.sampler_count = 0; + + if (BRW_DEBUG & DEBUG_STATS) + vs.thread4.stats_enable = 1; + + /* Vertex program always enabled: + */ + vs.vs6.vs_enable = 1; + + ret = brw_upload_cache(&brw->cache, BRW_VS_UNIT, + key, sizeof(*key), + reloc, 1, + &vs, sizeof(vs), + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + +static int prepare_vs_unit(struct brw_context *brw) +{ + struct brw_vs_unit_key key; + enum pipe_error ret; + struct brw_winsys_reloc reloc[1]; + unsigned grf_reg_count; + + vs_unit_populate_key(brw, &key); + + grf_reg_count = (align(key.total_grf, 16) / 16 - 1); + + /* Emit VS program relocation */ + make_reloc(&reloc[0], + BRW_USAGE_STATE, + grf_reg_count << 1, + offsetof(struct brw_vs_unit_state, thread0), + brw->vs.prog_bo); + + + if (brw_search_cache(&brw->cache, BRW_VS_UNIT, + &key, sizeof(key), + reloc, 1, + NULL, + &brw->vs.state_bo)) + return PIPE_OK; + + ret = vs_unit_create_from_key(brw, &key, reloc, &brw->vs.state_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_vs_unit = { + .dirty = { + .mesa = (PIPE_NEW_CLIP), + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_NR_VS_SURFACES | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_VS_PROG + }, + .prepare = prepare_vs_unit, +}; diff --git a/src/gallium/drivers/i965/brw_vs_surface_state.c b/src/gallium/drivers/i965/brw_vs_surface_state.c new file mode 100644 index 00000000000..177a5170d23 --- /dev/null +++ b/src/gallium/drivers/i965/brw_vs_surface_state.c @@ -0,0 +1,232 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_winsys.h" + +/* XXX: disabled true constant buffer functionality + */ + + +/* Creates a new VS constant buffer reflecting the current VS program's + * constants, if needed by the VS program. + * + * Otherwise, constants go through the CURBEs using the brw_constant_buffer + * state atom. + */ +#if 0 +static struct brw_winsys_buffer * +brw_vs_update_constant_buffer(struct brw_context *brw) +{ + /* XXX: true constant buffers + */ + struct brw_vertex_program *vp = + (struct brw_vertex_program *) brw->vertex_program; + const struct gl_program_parameter_list *params = vp->program.Base.Parameters; + const int size = params->NumParameters * 4 * sizeof(GLfloat); + drm_intel_bo *const_buffer; + + /* BRW_NEW_VERTEX_PROGRAM */ + if (!vp->use_const_buffer) + return NULL; + + const_buffer = brw->sws->bo_alloc(brw->sws, + BRW_BUFFER_TYPE_SHADER_CONSTANTS, + size, 64); + + /* _NEW_PROGRAM_CONSTANTS */ + brw->sws->bo_subdata(const_buffer, 0, size, params->ParameterValues, + NULL, 0); + + return const_buffer; +} +#endif + +/** + * Update the surface state for a VS constant buffer. + * + * Sets brw->vs.surf_bo[surf] and brw->vp->const_buffer. + */ +#if 0 +static void +brw_update_vs_constant_surface( struct brw_context *brw, + GLuint surf) +{ + struct brw_surface_key key; + struct pipe_buffer *cb = brw->curr.vs_constants; + enum pipe_error ret; + + assert(surf == 0); + + /* If we're in this state update atom, we need to update VS constants, so + * free the old buffer and create a new one for the new contents. + */ + ret = brw_vs_update_constant_buffer(brw, &vp->const_buffer); + if (ret) + return ret; + + /* If there's no constant buffer, then no surface BO is needed to point at + * it. + */ + if (vp->const_buffer == NULL) { + bo_reference(brw->vs.surf_bo[surf], NULL); + return PIPE_OK; + } + + memset(&key, 0, sizeof(key)); + + key.format = PIPE_FORMAT_R32G32B32A32_FLOAT; + key.bo = vp->const_buffer; + key.depthmode = GL_NONE; + key.pitch = params->NumParameters; + key.width = params->NumParameters; + key.height = 1; + key.depth = 1; + key.cpp = 16; + + /* + printf("%s:\n", __FUNCTION__); + printf(" width %d height %d depth %d cpp %d pitch %d\n", + key.width, key.height, key.depth, key.cpp, key.pitch); + */ + + if (brw_search_cache(&brw->surface_cache, + BRW_SS_SURFACE, + &key, sizeof(key), + &key.bo, key.bo ? 1 : 0, + NULL, + &brw->vs.surf_bo[surf])) + return PIPE_OK; + + ret = brw_create_constant_surface(brw, &key + &brw->vs.surf_bo[surf]); + if (ret) + return ret; + + return PIPE_OK; +} +#endif + + +/** + * Constructs the binding table for the VS surface state. + */ +static enum pipe_error +brw_vs_get_binding_table(struct brw_context *brw, + struct brw_winsys_buffer **bo_out) +{ +#if 0 + static GLuint data[BRW_VS_MAX_SURF]; /* always zero */ + struct brw_winsys_reloc reloc[BRW_VS_MAX_SURF]; + int i; + + /* Emit binding table relocations to surface state */ + for (i = 0; i < BRW_VS_MAX_SURF; i++) { + make_reloc(&reloc[i], + BRW_USAGE_STATE, + 0, + i * 4, + brw->vs.surf_bo[i]); + } + + ret = brw_cache_data( &brw->surface_cache, + BRW_SS_SURF_BIND, + NULL, 0, + reloc, nr_reloc, + data, sizeof data, + NULL, NULL, + bo_out); + if (ret) + return ret; + + FREE(data); + return PIPE_OK; +#else + return PIPE_OK; +#endif +} + +/** + * Vertex shader surfaces (constant buffer). + * + * This consumes the state updates for the constant buffer needing + * to be updated, and produces BRW_NEW_NR_VS_SURFACES for the VS unit and + * CACHE_NEW_SURF_BIND for the binding table upload. + */ +static enum pipe_error prepare_vs_surfaces(struct brw_context *brw ) +{ + enum pipe_error ret; + +#if 0 + int i; + int nr_surfaces = 0; + + brw_update_vs_constant_surface(ctx, SURF_INDEX_VERT_CONST_BUFFER); + + for (i = 0; i < BRW_VS_MAX_SURF; i++) { + if (brw->vs.surf_bo[i] != NULL) { + nr_surfaces = i + 1; + } + } + + if (brw->vs.nr_surfaces != nr_surfaces) { + brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES; + brw->vs.nr_surfaces = nr_surfaces; + } +#endif + + /* Note that we don't end up updating the bind_bo if we don't have a + * surface to be pointing at. This should be relatively harmless, as it + * just slightly increases our working set size. + */ + if (brw->vs.nr_surfaces != 0) { + ret = brw_vs_get_binding_table(brw, &brw->vs.bind_bo); + if (ret) + return ret; + } + + return PIPE_OK; +} + +const struct brw_tracked_state brw_vs_surfaces = { + .dirty = { + .mesa = (PIPE_NEW_VERTEX_CONSTANTS | + PIPE_NEW_VERTEX_SHADER), + .brw = 0, + .cache = 0 + }, + .prepare = prepare_vs_surfaces, +}; + + + diff --git a/src/gallium/drivers/i965/brw_winsys.h b/src/gallium/drivers/i965/brw_winsys.h new file mode 100644 index 00000000000..a242e31218a --- /dev/null +++ b/src/gallium/drivers/i965/brw_winsys.h @@ -0,0 +1,309 @@ +/************************************************************************** + * + * Copyright © 2009 Jakob Bornecrantz + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_WINSYS_H +#define BRW_WINSYS_H + +#include "pipe/p_compiler.h" +#include "pipe/p_defines.h" +#include "pipe/p_refcnt.h" + +struct brw_winsys; +struct pipe_fence_handle; + +/* Not sure why the winsys needs this: + */ +#define BRW_BATCH_SIZE (32*1024) + +struct brw_winsys_screen; + +/* Need a tiny bit of information inside the abstract buffer struct: + */ +struct brw_winsys_buffer { + struct pipe_reference reference; + struct brw_winsys_screen *sws; + unsigned size; +}; + + +/* Should be possible to validate usages above against buffer creation + * types, below: + */ +enum brw_buffer_type +{ + BRW_BUFFER_TYPE_TEXTURE, + BRW_BUFFER_TYPE_SCANOUT, /**< a texture used for scanning out from */ + BRW_BUFFER_TYPE_VERTEX, + BRW_BUFFER_TYPE_CURBE, + BRW_BUFFER_TYPE_QUERY, + BRW_BUFFER_TYPE_SHADER_CONSTANTS, + BRW_BUFFER_TYPE_SHADER_SCRATCH, + BRW_BUFFER_TYPE_BATCH, + BRW_BUFFER_TYPE_GENERAL_STATE, + BRW_BUFFER_TYPE_SURFACE_STATE, + BRW_BUFFER_TYPE_PIXEL, /* image uploads, pbo's, etc */ + BRW_BUFFER_TYPE_GENERIC, /* unknown */ + BRW_BUFFER_TYPE_MAX /* Count of possible values */ +}; + + +/* Describe the usage of a particular buffer in a relocation. The DRM + * winsys will translate these back to GEM read/write domain flags. + */ +enum brw_buffer_usage { + BRW_USAGE_STATE, /* INSTRUCTION, 0 */ + BRW_USAGE_QUERY_RESULT, /* INSTRUCTION, INSTRUCTION */ + BRW_USAGE_RENDER_TARGET, /* RENDER, 0 */ + BRW_USAGE_DEPTH_BUFFER, /* RENDER, RENDER */ + BRW_USAGE_BLIT_SOURCE, /* RENDER, 0 */ + BRW_USAGE_BLIT_DEST, /* RENDER, RENDER */ + BRW_USAGE_SAMPLER, /* SAMPLER, 0 */ + BRW_USAGE_VERTEX, /* VERTEX, 0 */ + BRW_USAGE_SCRATCH, /* 0, 0 */ + BRW_USAGE_MAX +}; + +enum brw_buffer_data_type { + BRW_DATA_GS_CC_VP, + BRW_DATA_GS_CC_UNIT, + BRW_DATA_GS_WM_PROG, + BRW_DATA_GS_SAMPLER_DEFAULT_COLOR, + BRW_DATA_GS_SAMPLER, + BRW_DATA_GS_WM_UNIT, + BRW_DATA_GS_SF_PROG, + BRW_DATA_GS_SF_VP, + BRW_DATA_GS_SF_UNIT, + BRW_DATA_GS_VS_UNIT, + BRW_DATA_GS_VS_PROG, + BRW_DATA_GS_GS_UNIT, + BRW_DATA_GS_GS_PROG, + BRW_DATA_GS_CLIP_VP, + BRW_DATA_GS_CLIP_UNIT, + BRW_DATA_GS_CLIP_PROG, + BRW_DATA_SS_SURFACE, + BRW_DATA_SS_SURF_BIND, + BRW_DATA_CONSTANT_BUFFER, + BRW_DATA_BATCH_BUFFER, + BRW_DATA_OTHER, + BRW_DATA_MAX +}; + + +/* Matches the i915_drm definitions: + */ +#define BRW_TILING_NONE 0 +#define BRW_TILING_X 1 +#define BRW_TILING_Y 2 + + +/* Relocations to be applied with subdata in a call to sws->bo_subdata, below. + * + * Effectively this encodes: + * + * (unsigned *)(subdata + offset) = bo->offset + delta + */ +struct brw_winsys_reloc { + enum brw_buffer_usage usage; /* debug only */ + unsigned delta; + unsigned offset; + struct brw_winsys_buffer *bo; +}; + +static INLINE void make_reloc(struct brw_winsys_reloc *reloc, + enum brw_buffer_usage usage, + unsigned delta, + unsigned offset, + struct brw_winsys_buffer *bo) +{ + reloc->usage = usage; + reloc->delta = delta; + reloc->offset = offset; + reloc->bo = bo; /* Note - note taking a reference yet */ +} + + + +struct brw_winsys_screen { + + + /** + * Buffer functions. + */ + + /*@{*/ + /** + * Create a buffer. + */ + enum pipe_error (*bo_alloc)(struct brw_winsys_screen *sws, + enum brw_buffer_type type, + unsigned size, + unsigned alignment, + struct brw_winsys_buffer **bo_out); + + /* Destroy a buffer when our refcount goes to zero: + */ + void (*bo_destroy)(struct brw_winsys_buffer *buffer); + + /* delta -- added to b2->offset, and written into buffer + * offset -- location above value is written to within buffer + */ + enum pipe_error (*bo_emit_reloc)(struct brw_winsys_buffer *buffer, + enum brw_buffer_usage usage, + unsigned delta, + unsigned offset, + struct brw_winsys_buffer *b2); + + enum pipe_error (*bo_exec)(struct brw_winsys_buffer *buffer, + unsigned bytes_used); + + enum pipe_error (*bo_subdata)(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + size_t offset, + size_t size, + const void *data, + const struct brw_winsys_reloc *reloc, + unsigned nr_reloc ); + + boolean (*bo_is_busy)(struct brw_winsys_buffer *buffer); + boolean (*bo_references)(struct brw_winsys_buffer *a, + struct brw_winsys_buffer *b); + + /* XXX: couldn't this be handled by returning true/false on + * bo_emit_reloc? + */ + enum pipe_error (*check_aperture_space)(struct brw_winsys_screen *iws, + struct brw_winsys_buffer **buffers, + unsigned count); + + /** + * Map a buffer. + */ + void *(*bo_map)(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + unsigned offset, + unsigned length, + boolean write, + boolean discard, + boolean flush_explicit); + + void (*bo_flush_range)(struct brw_winsys_buffer *buffer, + unsigned offset, + unsigned length); + + /** + * Unmap a buffer. + */ + void (*bo_unmap)(struct brw_winsys_buffer *buffer); + /*@}*/ + + + /* Wait for buffer to go idle. Similar to map+unmap, but doesn't + * mark buffer contents as dirty. + */ + void (*bo_wait_idle)(struct brw_winsys_buffer *buffer); + + /** + * Destroy the winsys. + */ + void (*destroy)(struct brw_winsys_screen *iws); +}; + +static INLINE void * +bo_map_read(struct brw_winsys_screen *sws, struct brw_winsys_buffer *buf) +{ + return sws->bo_map( buf, + BRW_DATA_OTHER, + 0, buf->size, + FALSE, FALSE, FALSE ); +} + +static INLINE void +bo_reference(struct brw_winsys_buffer **ptr, struct brw_winsys_buffer *buf) +{ + struct brw_winsys_buffer *old_buf = *ptr; + + if (pipe_reference(&(*ptr)->reference, &buf->reference)) + old_buf->sws->bo_destroy(old_buf); + + *ptr = buf; +} + + +/** + * Create brw pipe_screen. + */ +struct pipe_screen *brw_create_screen(struct brw_winsys_screen *iws, unsigned pci_id); + +/** + * Create a brw pipe_context. + */ +struct pipe_context *brw_create_context(struct pipe_screen *screen); + +/** + * Get the brw_winsys buffer backing the texture. + * + * TODO UGLY + */ +struct pipe_texture; +boolean brw_texture_get_winsys_buffer(struct pipe_texture *texture, + struct brw_winsys_buffer **buffer, + unsigned *stride); + +/** + * Wrap a brw_winsys buffer with a texture blanket. + * + * TODO UGLY + */ +struct pipe_texture * +brw_texture_blanket_winsys_buffer(struct pipe_screen *screen, + const struct pipe_texture *template, + unsigned pitch, + unsigned tiling, + struct brw_winsys_buffer *buffer); + + +/************************************************************************* + * Cooperative dumping between winsys and driver. TODO: make this + * driver-only by wrapping calls to winsys->bo_subdata(). + */ + +#ifdef DEBUG +extern int BRW_DUMP; +#else +#define BRW_DUMP 0 +#endif + +#define DUMP_ASM 0x1 +#define DUMP_STATE 0x2 +#define DUMP_BATCH 0x4 + +void brw_dump_data( unsigned pci_id, + enum brw_buffer_data_type data_type, + unsigned offset, + const void *data, + size_t size ); + + +#endif diff --git a/src/gallium/drivers/i965/brw_winsys_debug.c b/src/gallium/drivers/i965/brw_winsys_debug.c new file mode 100644 index 00000000000..f8f6a539bc9 --- /dev/null +++ b/src/gallium/drivers/i965/brw_winsys_debug.c @@ -0,0 +1,87 @@ +#include "brw_winsys.h" +#include "brw_disasm.h" +#include "brw_structs_dump.h" +#include "brw_structs.h" +#include "intel_decode.h" + + +void brw_dump_data( unsigned pci_id, + enum brw_buffer_data_type data_type, + unsigned offset, + const void *data, + size_t size ) +{ + if (BRW_DUMP & DUMP_ASM) { + switch (data_type) { + case BRW_DATA_GS_WM_PROG: + case BRW_DATA_GS_SF_PROG: + case BRW_DATA_GS_VS_PROG: + case BRW_DATA_GS_GS_PROG: + case BRW_DATA_GS_CLIP_PROG: + brw_disasm( stderr, data, size / sizeof(struct brw_instruction) ); + break; + default: + break; + } + } + + if (BRW_DUMP & DUMP_STATE) { + switch (data_type) { + case BRW_DATA_GS_CC_VP: + brw_dump_cc_viewport( data ); + break; + case BRW_DATA_GS_CC_UNIT: + brw_dump_cc_unit_state( data ); + break; + case BRW_DATA_GS_SAMPLER_DEFAULT_COLOR: + brw_dump_sampler_default_color( data ); + break; + case BRW_DATA_GS_SAMPLER: + brw_dump_sampler_state( data ); + break; + case BRW_DATA_GS_WM_UNIT: + brw_dump_wm_unit_state( data ); + break; + case BRW_DATA_GS_SF_VP: + brw_dump_sf_viewport( data ); + break; + case BRW_DATA_GS_SF_UNIT: + brw_dump_sf_unit_state( data ); + break; + case BRW_DATA_GS_VS_UNIT: + brw_dump_vs_unit_state( data ); + break; + case BRW_DATA_GS_GS_UNIT: + brw_dump_gs_unit_state( data ); + break; + case BRW_DATA_GS_CLIP_VP: + brw_dump_clipper_viewport( data ); + break; + case BRW_DATA_GS_CLIP_UNIT: + brw_dump_clip_unit_state( data ); + break; + case BRW_DATA_SS_SURFACE: + brw_dump_surface_state( data ); + break; + case BRW_DATA_SS_SURF_BIND: + break; + case BRW_DATA_OTHER: + break; + case BRW_DATA_CONSTANT_BUFFER: + break; + default: + break; + } + } + + if (BRW_DUMP & DUMP_BATCH) { + switch (data_type) { + case BRW_DATA_BATCH_BUFFER: + intel_decode(data, size / 4, offset, pci_id); + break; + default: + break; + } + } +} + diff --git a/src/gallium/drivers/i965/brw_wm.c b/src/gallium/drivers/i965/brw_wm.c new file mode 100644 index 00000000000..fdf820a9aae --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm.c @@ -0,0 +1,319 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ +#include "tgsi/tgsi_info.h" + +#include "brw_context.h" +#include "brw_screen.h" +#include "brw_util.h" +#include "brw_wm.h" +#include "brw_state.h" +#include "brw_debug.h" +#include "brw_pipe_rast.h" + + +/** Return number of src args for given instruction */ +GLuint brw_wm_nr_args( GLuint opcode ) +{ + switch (opcode) { + case WM_FRONTFACING: + case WM_PIXELXY: + return 0; + case WM_CINTERP: + case WM_WPOSXY: + case WM_DELTAXY: + return 1; + case WM_LINTERP: + case WM_PIXELW: + return 2; + case WM_FB_WRITE: + case WM_PINTERP: + return 3; + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXP: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXD: + /* sampler arg is held as a field in the instruction, not in an + * actual register: + */ + return tgsi_get_opcode_info(opcode)->num_src - 1; + + default: + assert(opcode < MAX_OPCODE); + return tgsi_get_opcode_info(opcode)->num_src; + } +} + + +GLuint brw_wm_is_scalar_result( GLuint opcode ) +{ + switch (opcode) { + case TGSI_OPCODE_COS: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_POW: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SIN: + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + case TGSI_OPCODE_DST: + return 1; + + default: + return 0; + } +} + + +/** + * Do GPU code generation for shaders without flow control. Shaders + * without flow control instructions can more readily be analysed for + * SSA-style optimizations. + */ +static void +brw_wm_linear_shader_emit(struct brw_context *brw, struct brw_wm_compile *c) +{ + /* Augment fragment program. Add instructions for pre- and + * post-fragment-program tasks such as interpolation and fogging. + */ + brw_wm_pass_fp(c); + + /* Translate to intermediate representation. Build register usage + * chains. + */ + brw_wm_pass0(c); + + /* Dead code removal. + */ + brw_wm_pass1(c); + + /* Register allocation. + * Divide by two because we operate on 16 pixels at a time and require + * two GRF entries for each logical shader register. + */ + c->grf_limit = BRW_WM_MAX_GRF / 2; + + brw_wm_pass2(c); + + /* how many general-purpose registers are used */ + c->prog_data.total_grf = c->max_wm_grf; + + /* Scratch space is used for register spilling */ + if (c->last_scratch) { + c->prog_data.total_scratch = c->last_scratch + 0x40; + } + else { + c->prog_data.total_scratch = 0; + } + + /* Emit GEN4 code. + */ + brw_wm_emit(c); +} + + +/** + * All Mesa program -> GPU code generation goes through this function. + * Depending on the instructions used (i.e. flow control instructions) + * we'll use one of two code generators. + */ +static enum pipe_error do_wm_prog( struct brw_context *brw, + struct brw_fragment_shader *fp, + struct brw_wm_prog_key *key, + struct brw_winsys_buffer **bo_out) +{ + enum pipe_error ret; + struct brw_wm_compile *c; + const GLuint *program; + GLuint program_size; + + if (brw->wm.compile_data == NULL) { + brw->wm.compile_data = MALLOC(sizeof(*brw->wm.compile_data)); + if (!brw->wm.compile_data) + return PIPE_ERROR_OUT_OF_MEMORY; + } + + c = brw->wm.compile_data; + memset(c, 0, sizeof *c); + + c->key = *key; + c->fp = fp; + c->env_param = NULL; /*brw->intel.ctx.FragmentProgram.Parameters;*/ + + brw_init_compile(brw, &c->func); + + /* + * Shader which use GLSL features such as flow control are handled + * differently from "simple" shaders. + */ + if (fp->has_flow_control) { + c->dispatch_width = 8; + /* XXX: GLSL support + */ + exit(1); + /* brw_wm_branching_shader_emit(brw, c); */ + } + else { + c->dispatch_width = 16; + brw_wm_linear_shader_emit(brw, c); + } + + if (BRW_DEBUG & DEBUG_WM) + debug_printf("\n"); + + /* get the program + */ + ret = brw_get_program(&c->func, &program, &program_size); + if (ret) + return ret; + + ret = brw_upload_cache( &brw->cache, BRW_WM_PROG, + &c->key, sizeof(c->key), + NULL, 0, + program, program_size, + &c->prog_data, + &brw->wm.prog_data, + bo_out ); + if (ret) + return ret; + + return PIPE_OK; +} + + + +static void brw_wm_populate_key( struct brw_context *brw, + struct brw_wm_prog_key *key ) +{ + unsigned lookup, line_aa; + unsigned i; + + memset(key, 0, sizeof(*key)); + + /* PIPE_NEW_FRAGMENT_SHADER + * PIPE_NEW_DEPTH_STENCIL_ALPHA + */ + lookup = (brw->curr.zstencil->iz_lookup | + brw->curr.fragment_shader->iz_lookup); + + + /* PIPE_NEW_RAST + * BRW_NEW_REDUCED_PRIMITIVE + */ + switch (brw->reduced_primitive) { + case PIPE_PRIM_POINTS: + line_aa = AA_NEVER; + break; + case PIPE_PRIM_LINES: + line_aa = (brw->curr.rast->templ.line_smooth ? + AA_ALWAYS : AA_NEVER); + break; + default: + line_aa = brw->curr.rast->unfilled_aa_line; + break; + } + + brw_wm_lookup_iz(line_aa, + lookup, + brw->curr.fragment_shader->uses_depth, + key); + + /* PIPE_NEW_RAST */ + key->flat_shade = brw->curr.rast->templ.flatshade; + + + /* PIPE_NEW_BOUND_TEXTURES */ + for (i = 0; i < brw->curr.num_textures; i++) { + const struct brw_texture *tex = brw_texture(brw->curr.texture[i]); + + if (tex->base.format == PIPE_FORMAT_YCBCR) + key->yuvtex_mask |= 1 << i; + + if (tex->base.format == PIPE_FORMAT_YCBCR_REV) + key->yuvtex_swap_mask |= 1 << i; + + /* XXX: shadow texture + */ + /* key->shadowtex_mask |= 1<<i; */ + } + + /* CACHE_NEW_VS_PROG */ + key->vp_nr_outputs = brw->vs.prog_data->nr_outputs; + + key->nr_cbufs = brw->curr.fb.nr_cbufs; + + key->nr_inputs = brw->curr.fragment_shader->info.num_inputs; + + /* The unique fragment program ID */ + key->program_string_id = brw->curr.fragment_shader->id; +} + + +static enum pipe_error brw_prepare_wm_prog(struct brw_context *brw) +{ + struct brw_wm_prog_key key; + struct brw_fragment_shader *fs = brw->curr.fragment_shader; + enum pipe_error ret; + + brw_wm_populate_key(brw, &key); + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache, BRW_WM_PROG, + &key, sizeof(key), + NULL, 0, + &brw->wm.prog_data, + &brw->wm.prog_bo)) + return PIPE_OK; + + ret = do_wm_prog(brw, fs, &key, &brw->wm.prog_bo); + if (ret) + return ret; + + return PIPE_OK; +} + + +const struct brw_tracked_state brw_wm_prog = { + .dirty = { + .mesa = (PIPE_NEW_FRAGMENT_SHADER | + PIPE_NEW_DEPTH_STENCIL_ALPHA | + PIPE_NEW_RAST | + PIPE_NEW_NR_CBUFS | + PIPE_NEW_BOUND_TEXTURES), + .brw = (BRW_NEW_WM_INPUT_DIMENSIONS | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = CACHE_NEW_VS_PROG, + }, + .prepare = brw_prepare_wm_prog +}; + diff --git a/src/gallium/drivers/i965/brw_wm.h b/src/gallium/drivers/i965/brw_wm.h new file mode 100644 index 00000000000..f1ca9f63696 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm.h @@ -0,0 +1,344 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#ifndef BRW_WM_H +#define BRW_WM_H + +#include "brw_context.h" +#include "brw_eu.h" + +#define SATURATE (1<<5) + +/* A big lookup table is used to figure out which and how many + * additional regs will inserted before the main payload in the WM + * program execution. These mainly relate to depth and stencil + * processing and the early-depth-test optimization. + */ +#define IZ_PS_KILL_ALPHATEST_BIT 0x1 +#define IZ_PS_COMPUTES_DEPTH_BIT 0x2 +#define IZ_DEPTH_WRITE_ENABLE_BIT 0x4 +#define IZ_DEPTH_TEST_ENABLE_BIT 0x8 +#define IZ_STENCIL_WRITE_ENABLE_BIT 0x10 +#define IZ_STENCIL_TEST_ENABLE_BIT 0x20 +#define IZ_BIT_MAX 0x40 + +#define AA_NEVER 0 +#define AA_SOMETIMES 1 +#define AA_ALWAYS 2 + +struct brw_wm_prog_key { + GLuint source_depth_reg:3; + GLuint aa_dest_stencil_reg:3; + GLuint dest_depth_reg:3; + GLuint nr_depth_regs:3; + GLuint computes_depth:1; + GLuint source_depth_to_render_target:1; + GLuint flat_shade:1; + GLuint runtime_check_aads_emit:1; + + GLuint shadowtex_mask:16; + GLuint yuvtex_mask:16; + GLuint yuvtex_swap_mask:16; /* UV swaped */ + + GLuint vp_nr_outputs:6; + GLuint nr_inputs:6; + GLuint nr_cbufs:3; + GLuint has_flow_control:1; + + GLuint program_string_id; +}; + + +/* A bit of a glossary: + * + * brw_wm_value: A computed value or program input. Values are + * constant, they are created once and are never modified. When a + * fragment program register is written or overwritten, new values are + * created fresh, preserving the rule that values are constant. + * + * brw_wm_ref: A reference to a value. Wherever a value used is by an + * instruction or as a program output, that is tracked with an + * instance of this struct. All references to a value occur after it + * is created. After the last reference, a value is dead and can be + * discarded. + * + * brw_wm_grf: Represents a physical hardware register. May be either + * empty or hold a value. Register allocation is the process of + * assigning values to grf registers. This occurs in pass2 and the + * brw_wm_grf struct is not used before that. + * + * Fragment program registers: These are time-varying constructs that + * are hard to reason about and which we translate away in pass0. A + * single fragment program register element (eg. temp[0].x) will be + * translated to one or more brw_wm_value structs, one for each time + * that temp[0].x is written to during the program. + */ + + + +/* Used in pass2 to track register allocation. + */ +struct brw_wm_grf { + struct brw_wm_value *value; + GLuint nextuse; +}; + +struct brw_wm_value { + struct brw_reg hw_reg; /* emitted to this reg, may not always be there */ + struct brw_wm_ref *lastuse; + struct brw_wm_grf *resident; + GLuint contributes_to_output:1; + GLuint spill_slot:16; /* if non-zero, spill immediately after calculation */ +}; + +struct brw_wm_ref { + struct brw_reg hw_reg; /* nr filled in in pass2, everything else, pass0 */ + struct brw_wm_value *value; + struct brw_wm_ref *prevuse; + GLuint unspill_reg:7; /* unspill to reg */ + GLuint emitted:1; + GLuint insn:24; +}; + +struct brw_wm_instruction { + struct brw_wm_value *dst[4]; + struct brw_wm_ref *src[3][4]; + GLuint opcode:8; + GLuint saturate:1; + GLuint writemask:4; + GLuint sampler:4; + GLuint tex_unit:4; /* texture/sampler unit for texture instructions */ + GLuint target:4; /* TGSI_TEXTURE_x for texture instructions, + * target binding table index for FB_WRITE + */ + GLuint eot:1; /* End of thread indicator for FB_WRITE*/ +}; + + +#define BRW_WM_MAX_INSN 2048 +#define BRW_WM_MAX_GRF 128 /* hardware limit */ +#define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) +#define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) +#define BRW_WM_MAX_PARAM 256 +#define BRW_WM_MAX_CONST 256 +#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS +#define BRW_WM_MAX_SUBROUTINE 16 + + +/* New opcodes to track internal operations required for WM unit. + * These are added early so that the registers used can be tracked, + * freed and reused like those of other instructions. + */ +#define MAX_OPCODE TGSI_OPCODE_LAST +#define WM_PIXELXY (MAX_OPCODE) +#define WM_DELTAXY (MAX_OPCODE + 1) +#define WM_PIXELW (MAX_OPCODE + 2) +#define WM_LINTERP (MAX_OPCODE + 3) +#define WM_PINTERP (MAX_OPCODE + 4) +#define WM_CINTERP (MAX_OPCODE + 5) +#define WM_WPOSXY (MAX_OPCODE + 6) +#define WM_FB_WRITE (MAX_OPCODE + 7) +#define WM_FRONTFACING (MAX_OPCODE + 8) +#define MAX_WM_OPCODE (MAX_OPCODE + 9) + +#define BRW_FILE_PAYLOAD (TGSI_FILE_COUNT) +#define PAYLOAD_DEPTH (PIPE_MAX_SHADER_INPUTS) /* ?? */ + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +struct brw_fp_src { + unsigned file:4; + unsigned index:16; + unsigned swizzle:8; + unsigned indirect:1; + unsigned negate:1; + unsigned abs:1; +}; + +struct brw_fp_dst { + unsigned file:4; + unsigned index:16; + unsigned writemask:4; + unsigned indirect:1; + unsigned saturate:1; +}; + +struct brw_fp_instruction { + struct brw_fp_dst dst; + struct brw_fp_src src[3]; + unsigned opcode:8; + unsigned target:8; /* XXX: special usage for FB_WRITE */ + unsigned tex_unit:4; + unsigned sampler:4; + unsigned pad:8; +}; + + +struct brw_wm_compile { + struct brw_compile func; + struct brw_wm_prog_key key; + struct brw_wm_prog_data prog_data; + + struct brw_fragment_shader *fp; + + GLfloat (*env_param)[4]; + + enum { + START, + PASS2_DONE + } state; + + /* Initial pass - translate fp instructions to fp instructions, + * simplifying and adding instructions for interpolation and + * framebuffer writes. + */ + struct { + GLfloat v[4]; + unsigned nr; + } immediate[BRW_WM_MAX_CONST+3]; + GLuint nr_immediates; + + struct brw_fp_instruction fp_instructions[BRW_WM_MAX_INSN]; + GLuint nr_fp_insns; + GLuint fp_temp; + GLuint fp_interp_emitted; + GLuint fp_fragcolor_emitted; + GLuint fp_first_internal_temp; + + struct brw_fp_src fp_pixel_xy; + struct brw_fp_src fp_delta_xy; + struct brw_fp_src fp_pixel_w; + + + /* Subsequent passes using SSA representation: + */ + struct brw_wm_value vreg[BRW_WM_MAX_VREG]; + GLuint nr_vreg; + + struct brw_wm_value creg[BRW_WM_MAX_PARAM]; + GLuint nr_creg; + + struct { + struct brw_wm_value depth[4]; /* includes r0/r1 */ + struct brw_wm_value input_interp[PIPE_MAX_SHADER_INPUTS]; + } payload; + + + const struct brw_wm_ref *pass0_fp_reg[BRW_FILE_PAYLOAD+1][256][4]; + + struct brw_wm_ref undef_ref; + struct brw_wm_value undef_value; + + struct brw_wm_ref refs[BRW_WM_MAX_REF]; + GLuint nr_refs; + + struct brw_wm_instruction instruction[BRW_WM_MAX_INSN]; + GLuint nr_insns; + + struct brw_wm_grf pass2_grf[BRW_WM_MAX_GRF/2]; + + GLuint grf_limit; + GLuint max_wm_grf; + GLuint last_scratch; + + GLuint cur_inst; /**< index of current instruction */ + + GLboolean out_of_regs; /**< ran out of GRF registers? */ + + /** Mapping from Mesa registers to hardware registers */ + struct { + GLboolean inited; + struct brw_reg reg; + } wm_regs[BRW_FILE_PAYLOAD+1][256][4]; + + GLboolean used_grf[BRW_WM_MAX_GRF]; + GLuint first_free_grf; + struct brw_reg stack; + struct brw_reg emit_mask_reg; + GLuint tmp_regs[BRW_WM_MAX_GRF]; + GLuint tmp_index; + GLuint tmp_max; + GLuint subroutines[BRW_WM_MAX_SUBROUTINE]; + GLuint dispatch_width; + + /** we may need up to 3 constants per instruction (if use_const_buffer) */ + struct { + GLint index; + struct brw_reg reg; + } current_const[3]; + + GLuint error; +}; + + +GLuint brw_wm_nr_args( GLuint opcode ); +GLuint brw_wm_is_scalar_result( GLuint opcode ); + +int brw_wm_pass_fp( struct brw_wm_compile *c ); +void brw_wm_pass0( struct brw_wm_compile *c ); +void brw_wm_pass1( struct brw_wm_compile *c ); +void brw_wm_pass2( struct brw_wm_compile *c ); +void brw_wm_emit( struct brw_wm_compile *c ); + +void brw_wm_print_value( struct brw_wm_compile *c, + struct brw_wm_value *value ); + +void brw_wm_print_ref( struct brw_wm_compile *c, + struct brw_wm_ref *ref ); + +void brw_wm_print_insn( struct brw_wm_compile *c, + struct brw_wm_instruction *inst ); + +void brw_wm_print_program( struct brw_wm_compile *c, + const char *stage ); + +void brw_wm_print_fp_program( struct brw_wm_compile *c, + const char *stage ); + +void brw_wm_lookup_iz( GLuint line_aa, + GLuint lookup, + GLboolean ps_uses_depth, + struct brw_wm_prog_key *key ); + +void brw_wm_branching_shader_emit(struct brw_context *brw, struct brw_wm_compile *c); + +void emit_ddxy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + GLboolean is_ddx, + const struct brw_reg *arg0); + +#endif diff --git a/src/gallium/drivers/i965/brw_wm_constant_buffer.c b/src/gallium/drivers/i965/brw_wm_constant_buffer.c new file mode 100644 index 00000000000..6434c6acf73 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_constant_buffer.c @@ -0,0 +1,165 @@ +/* XXX: Constant buffers disabled + */ + + +/** + * Create the constant buffer surface. Vertex/fragment shader constants will be + * read from this buffer with Data Port Read instructions/messages. + */ +enum pipe_error +brw_create_constant_surface( struct brw_context *brw, + struct brw_surface_key *key, + struct brw_winsys_buffer **bo_out ) +{ + const GLint w = key->width - 1; + struct brw_winsys_buffer *bo; + struct brw_winsys_reloc reloc[1]; + enum pipe_error ret; + + /* Emit relocation to surface contents */ + make_reloc(&reloc[0], + BRW_USAGE_SAMPLER, + 0, + offsetof(struct brw_surface_state, ss1), + key->bo); + + + memset(&surf, 0, sizeof(surf)); + + surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + surf.ss0.surface_type = BRW_SURFACE_BUFFER; + surf.ss0.surface_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; + + surf.ss1.base_addr = 0; /* reloc */ + + surf.ss2.width = w & 0x7f; /* bits 6:0 of size or width */ + surf.ss2.height = (w >> 7) & 0x1fff; /* bits 19:7 of size or width */ + surf.ss3.depth = (w >> 20) & 0x7f; /* bits 26:20 of size or width */ + surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */ + brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */ + + ret = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE, + key, sizeof(*key), + reloc, Elements(reloc), + &surf, sizeof(surf), + NULL, NULL, + &bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + + + +/** + * Update the surface state for a WM constant buffer. + * The constant buffer will be (re)allocated here if needed. + */ +static enum pipe_error +brw_update_wm_constant_surface( struct brw_context *brw, + GLuint surf) +{ + struct brw_surface_key key; + struct brw_fragment_shader *fp = brw->curr.fragment_shader; + struct pipe_buffer *cbuf = brw->curr.fragment_constants; + int pitch = cbuf->size / (4 * sizeof(float)); + enum pipe_error ret; + + /* If we're in this state update atom, we need to update WM constants, so + * free the old buffer and create a new one for the new contents. + */ + ret = brw_wm_update_constant_buffer(brw, &fp->const_buffer); + if (ret) + return ret; + + /* If there's no constant buffer, then no surface BO is needed to point at + * it. + */ + if (cbuf == NULL) { + bo_reference(&brw->wm.surf_bo[surf], NULL); + return PIPE_OK; + } + + memset(&key, 0, sizeof(key)); + + key.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + key.ss0.surface_type = BRW_SURFACE_BUFFER; + key.ss0.surface_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; + + key.bo = brw_buffer(cbuf)->bo; + + key.ss2.width = (pitch-1) & 0x7f; /* bits 6:0 of size or width */ + key.ss2.height = ((pitch-1) >> 7) & 0x1fff; /* bits 19:7 of size or width */ + key.ss3.depth = ((pitch-1) >> 20) & 0x7f; /* bits 26:20 of size or width */ + key.ss3.pitch = (pitch * 4 * sizeof(float)) - 1; /* ignored?? */ + brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */ + + + /* + printf("%s:\n", __FUNCTION__); + printf(" width %d height %d depth %d cpp %d pitch %d\n", + key.width, key.height, key.depth, key.cpp, key.pitch); + */ + + if (brw_search_cache(&brw->surface_cache, + BRW_SS_SURFACE, + &key, sizeof(key), + &key.bo, 1, + NULL, + &brw->wm.surf_bo[surf])) + return PIPE_OK; + + ret = brw_create_constant_surface(brw, &key, &brw->wm.surf_bo[surf]); + if (ret) + return ret; + + brw->state.dirty.brw |= BRW_NEW_WM_SURFACES; + return PIPE_OK; +} + +/** + * Updates surface / buffer for fragment shader constant buffer, if + * one is required. + * + * This consumes the state updates for the constant buffer, and produces + * BRW_NEW_WM_SURFACES to get picked up by brw_prepare_wm_surfaces for + * inclusion in the binding table. + */ +static enum pipe_error prepare_wm_constant_surface(struct brw_context *brw ) +{ + struct brw_fragment_program *fp = + (struct brw_fragment_program *) brw->fragment_program; + GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER; + + ret = brw_wm_update_constant_buffer(brw, + &fp->const_buffer); + if (ret) + return ret; + + /* If there's no constant buffer, then no surface BO is needed to point at + * it. + */ + if (fp->const_buffer == 0) { + if (brw->wm.surf_bo[surf] != NULL) { + bo_reference(&brw->wm.surf_bo[surf], NULL); + brw->state.dirty.brw |= BRW_NEW_WM_SURFACES; + } + return PIPE_OK; + } + + ret = brw_update_wm_constant_surface(ctx, surf); + if (ret) + return ret; + + return PIPE_OK +} + +const struct brw_tracked_state brw_wm_constant_surface = { + .dirty = { + .mesa = (_NEW_PROGRAM_CONSTANTS), + .brw = (BRW_NEW_FRAGMENT_PROGRAM), + .cache = 0 + }, + .prepare = prepare_wm_constant_surface, +}; diff --git a/src/gallium/drivers/i965/brw_wm_debug.c b/src/gallium/drivers/i965/brw_wm_debug.c new file mode 100644 index 00000000000..3d11fa074cc --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_debug.c @@ -0,0 +1,256 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "tgsi/tgsi_info.h" + +#include "brw_context.h" +#include "brw_wm.h" + +static void print_writemask( unsigned writemask ) +{ + if (writemask != BRW_WRITEMASK_XYZW) + debug_printf(".%s%s%s%s", + (writemask & BRW_WRITEMASK_X) ? "x" : "", + (writemask & BRW_WRITEMASK_Y) ? "y" : "", + (writemask & BRW_WRITEMASK_Z) ? "z" : "", + (writemask & BRW_WRITEMASK_W) ? "w" : ""); +} + +static void print_swizzle( unsigned swizzle ) +{ + char *swz = "xyzw"; + if (swizzle != BRW_SWIZZLE_XYZW) + debug_printf(".%c%c%c%c", + swz[BRW_GET_SWZ(swizzle, X)], + swz[BRW_GET_SWZ(swizzle, Y)], + swz[BRW_GET_SWZ(swizzle, Z)], + swz[BRW_GET_SWZ(swizzle, W)]); +} + +static void print_opcode( unsigned opcode ) +{ + switch (opcode) { + case WM_PIXELXY: + debug_printf("PIXELXY"); + break; + case WM_DELTAXY: + debug_printf("DELTAXY"); + break; + case WM_PIXELW: + debug_printf("PIXELW"); + break; + case WM_WPOSXY: + debug_printf("WPOSXY"); + break; + case WM_PINTERP: + debug_printf("PINTERP"); + break; + case WM_LINTERP: + debug_printf("LINTERP"); + break; + case WM_CINTERP: + debug_printf("CINTERP"); + break; + case WM_FB_WRITE: + debug_printf("FB_WRITE"); + break; + case WM_FRONTFACING: + debug_printf("FRONTFACING"); + break; + default: + debug_printf("%s", tgsi_get_opcode_info(opcode)->mnemonic); + break; + } +} + +void brw_wm_print_value( struct brw_wm_compile *c, + struct brw_wm_value *value ) +{ + assert(value); + if (c->state >= PASS2_DONE) + brw_print_reg(value->hw_reg); + else if( value == &c->undef_value ) + debug_printf("undef"); + else if( value - c->vreg >= 0 && + value - c->vreg < BRW_WM_MAX_VREG) + debug_printf("r%d", value - c->vreg); + else if (value - c->creg >= 0 && + value - c->creg < BRW_WM_MAX_PARAM) + debug_printf("c%d", value - c->creg); + else if (value - c->payload.input_interp >= 0 && + value - c->payload.input_interp < PIPE_MAX_SHADER_INPUTS) + debug_printf("i%d", value - c->payload.input_interp); + else if (value - c->payload.depth >= 0 && + value - c->payload.depth < PIPE_MAX_SHADER_INPUTS) + debug_printf("d%d", value - c->payload.depth); + else + debug_printf("?"); +} + +void brw_wm_print_ref( struct brw_wm_compile *c, + struct brw_wm_ref *ref ) +{ + struct brw_reg hw_reg = ref->hw_reg; + + if (ref->unspill_reg) + debug_printf("UNSPILL(%x)/", ref->value->spill_slot); + + if (c->state >= PASS2_DONE) + brw_print_reg(ref->hw_reg); + else { + debug_printf("%s", hw_reg.negate ? "-" : ""); + debug_printf("%s", hw_reg.abs ? "abs/" : ""); + brw_wm_print_value(c, ref->value); + if ((hw_reg.nr&1) || hw_reg.subnr) { + debug_printf("->%d.%d", (hw_reg.nr&1), hw_reg.subnr); + } + } +} + +void brw_wm_print_insn( struct brw_wm_compile *c, + struct brw_wm_instruction *inst ) +{ + GLuint i, arg; + GLuint nr_args = brw_wm_nr_args(inst->opcode); + + debug_printf("["); + for (i = 0; i < 4; i++) { + if (inst->dst[i]) { + brw_wm_print_value(c, inst->dst[i]); + if (inst->dst[i]->spill_slot) + debug_printf("/SPILL(%x)",inst->dst[i]->spill_slot); + } + else + debug_printf("#"); + if (i < 3) + debug_printf(","); + } + debug_printf("]"); + print_writemask(inst->writemask); + + debug_printf(" = "); + print_opcode(inst->opcode); + + if (inst->saturate) + debug_printf("_SAT"); + + for (arg = 0; arg < nr_args; arg++) { + + debug_printf(" ["); + + for (i = 0; i < 4; i++) { + if (inst->src[arg][i]) { + brw_wm_print_ref(c, inst->src[arg][i]); + } + else + debug_printf("%%"); + + if (i < 3) + debug_printf(","); + else + debug_printf("]"); + } + } + debug_printf("\n"); +} + +void brw_wm_print_program( struct brw_wm_compile *c, + const char *stage ) +{ + GLuint insn; + + debug_printf("%s:\n", stage); + for (insn = 0; insn < c->nr_insns; insn++) + brw_wm_print_insn(c, &c->instruction[insn]); + debug_printf("\n"); +} + +static const char *file_strings[TGSI_FILE_COUNT+1] = { + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMPLER", + "ADDR", + "IMM", + "LOOP", + "PAYLOAD" +}; + +static void brw_wm_print_fp_insn( struct brw_wm_compile *c, + struct brw_fp_instruction *inst ) +{ + GLuint i; + GLuint nr_args = brw_wm_nr_args(inst->opcode); + + print_opcode(inst->opcode); + if (inst->dst.saturate) + debug_printf("_SAT"); + debug_printf(" "); + + if (inst->dst.indirect) + debug_printf("["); + + debug_printf("%s[%d]", + file_strings[inst->dst.file], + inst->dst.index ); + print_writemask(inst->dst.writemask); + + if (inst->dst.indirect) + debug_printf("]"); + + debug_printf(nr_args ? ", " : "\n"); + + for (i = 0; i < nr_args; i++) { + debug_printf("%s%s%s[%d]%s", + inst->src[i].negate ? "-" : "", + inst->src[i].abs ? "ABS(" : "", + file_strings[inst->src[i].file], + inst->src[i].index, + inst->src[i].abs ? ")" : ""); + print_swizzle(inst->src[i].swizzle); + debug_printf("%s", i == nr_args - 1 ? "\n" : ", "); + } +} + + +void brw_wm_print_fp_program( struct brw_wm_compile *c, + const char *stage ) +{ + GLuint insn; + + debug_printf("%s:\n", stage); + for (insn = 0; insn < c->nr_fp_insns; insn++) + brw_wm_print_fp_insn(c, &c->fp_instructions[insn]); + debug_printf("\n"); +} + diff --git a/src/gallium/drivers/i965/brw_wm_emit.c b/src/gallium/drivers/i965/brw_wm_emit.c new file mode 100644 index 00000000000..7e57d0306bc --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_emit.c @@ -0,0 +1,1521 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" +#include "tgsi/tgsi_info.h" + +#include "brw_context.h" +#include "brw_wm.h" +#include "brw_debug.h" +#include "brw_disasm.h" + +/* Not quite sure how correct this is - need to understand horiz + * vs. vertical strides a little better. + */ +static INLINE struct brw_reg sechalf( struct brw_reg reg ) +{ + if (reg.vstride) + reg.nr++; + return reg; +} + +/* Payload R0: + * + * R0.0 -- pixel mask, one bit for each of 4 pixels in 4 quads, + * corresponding to each of the 16 execution channels. + * R0.1..8 -- ? + * R1.0 -- triangle vertex 0.X + * R1.1 -- triangle vertex 0.Y + * R1.2 -- quad 0 x,y coords (2 packed uwords) + * R1.3 -- quad 1 x,y coords (2 packed uwords) + * R1.4 -- quad 2 x,y coords (2 packed uwords) + * R1.5 -- quad 3 x,y coords (2 packed uwords) + * R1.6 -- ? + * R1.7 -- ? + * R1.8 -- ? + */ + + +static void emit_pixel_xy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + if (mask & BRW_WRITEMASK_X) { + brw_ADD(p, + vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + } + + if (mask & BRW_WRITEMASK_Y) { + brw_ADD(p, + vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw,5), 2, 4, 0), + brw_imm_v(0x11001100)); + } + + brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); +} + + + +static void emit_delta_xy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & BRW_WRITEMASK_X) { + brw_ADD(p, + dst[0], + retype(arg0[0], BRW_REGISTER_TYPE_UW), + negate(r1)); + } + + if (mask & BRW_WRITEMASK_Y) { + brw_ADD(p, + dst[1], + retype(arg0[1], BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + + } +} + +static void emit_wpos_xy(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) +{ + struct brw_compile *p = &c->func; + + if (mask & BRW_WRITEMASK_X) { + /* X' = X */ + brw_MOV(p, + dst[0], + retype(arg0[0], BRW_REGISTER_TYPE_W)); + } + + /* XXX: is this needed any more, or is this a NOOP? + */ + if (mask & BRW_WRITEMASK_Y) { +#if 0 + /* Y' = height - 1 - Y */ + brw_ADD(p, + dst[1], + negate(retype(arg0[1], BRW_REGISTER_TYPE_W)), + brw_imm_d(c->key.drawable_height - 1)); +#else + brw_MOV(p, + dst[0], + retype(arg0[0], BRW_REGISTER_TYPE_W)); +#endif + } +} + + +static void emit_pixel_w( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas) +{ + /* Don't need this if all you are doing is interpolating color, for + * instance. + */ + if (mask & BRW_WRITEMASK_W) { + struct brw_reg interp3 = brw_vec1_grf(arg0[0].nr+1, 4); + + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + brw_LINE(p, brw_null_reg(), interp3, deltas[0]); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), deltas[1]); + + /* Calc w */ + brw_math_16( p, dst[3], + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} + + + +static void emit_linterp( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas ) +{ + struct brw_reg interp[4]; + GLuint nr = arg0[0].nr; + GLuint i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_LINE(p, brw_null_reg(), interp[i], deltas[0]); + brw_MAC(p, dst[i], suboffset(interp[i],1), deltas[1]); + } + } +} + + +static void emit_pinterp( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas, + const struct brw_reg *w) +{ + struct brw_reg interp[4]; + GLuint nr = arg0[0].nr; + GLuint i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_LINE(p, brw_null_reg(), interp[i], deltas[0]); + brw_MAC(p, dst[i], suboffset(interp[i],1), deltas[1]); + } + } + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MUL(p, dst[i], dst[i], w[3]); + } + } +} + + +static void emit_cinterp( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0 ) +{ + struct brw_reg interp[4]; + GLuint nr = arg0[0].nr; + GLuint i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MOV(p, dst[i], suboffset(interp[i],3)); /* TODO: optimize away like other moves */ + } + } +} + +/* Sets the destination channels to 1.0 or 0.0 according to glFrontFacing. */ +static void emit_frontfacing( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask ) +{ + struct brw_reg r1_6ud = retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_UD); + GLuint i; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MOV(p, dst[i], brw_imm_f(0.0)); + } + } + + /* bit 31 is "primitive is back face", so checking < (1 << 31) gives + * us front face + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, r1_6ud, brw_imm_ud(1 << 31)); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MOV(p, dst[i], brw_imm_f(1.0)); + } + } + brw_set_predicate_control_flag_value(p, 0xff); +} + +/* For OPCODE_DDX and OPCODE_DDY, per channel of output we've got input + * looking like: + * + * arg0: q0.tl q0.tr q0.bl q0.br q1.tl q1.tr q1.bl q1.br + * + * and we're trying to produce: + * + * DDX DDY + * dst: (q0.tr - q0.tl) (q0.tl - q0.bl) + * (q0.tr - q0.tl) (q0.tr - q0.br) + * (q0.br - q0.bl) (q0.tl - q0.bl) + * (q0.br - q0.bl) (q0.tr - q0.br) + * (q1.tr - q1.tl) (q1.tl - q1.bl) + * (q1.tr - q1.tl) (q1.tr - q1.br) + * (q1.br - q1.bl) (q1.tl - q1.bl) + * (q1.br - q1.bl) (q1.tr - q1.br) + * + * and add two more quads if in 16-pixel dispatch mode. + * + * For DDX, it ends up being easy: width = 2, horiz=0 gets us the same result + * for each pair, and vertstride = 2 jumps us 2 elements after processing a + * pair. But for DDY, it's harder, as we want to produce the pairs swizzled + * between each other. We could probably do it like ddx and swizzle the right + * order later, but bail for now and just produce + * ((q0.tl - q0.bl)x4 (q1.tl - q1.bl)x4) + */ +void emit_ddxy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + GLboolean is_ddx, + const struct brw_reg *arg0) +{ + int i; + struct brw_reg src0, src1; + + if (mask & SATURATE) + brw_set_saturate(p, 1); + for (i = 0; i < 4; i++ ) { + if (mask & (1<<i)) { + if (is_ddx) { + src0 = brw_reg(arg0[i].file, arg0[i].nr, 1, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_2, + BRW_WIDTH_2, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, BRW_WRITEMASK_XYZW); + src1 = brw_reg(arg0[i].file, arg0[i].nr, 0, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_2, + BRW_WIDTH_2, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, BRW_WRITEMASK_XYZW); + } else { + src0 = brw_reg(arg0[i].file, arg0[i].nr, 0, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_4, + BRW_WIDTH_4, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, BRW_WRITEMASK_XYZW); + src1 = brw_reg(arg0[i].file, arg0[i].nr, 2, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_4, + BRW_WIDTH_4, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, BRW_WRITEMASK_XYZW); + } + brw_ADD(p, dst[i], src0, negate(src1)); + } + } + if (mask & SATURATE) + brw_set_saturate(p, 0); +} + +static void emit_alu1( struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0 ) +{ + GLuint i; + + if (mask & SATURATE) + brw_set_saturate(p, 1); + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + func(p, dst[i], arg0[i]); + } + } + + if (mask & SATURATE) + brw_set_saturate(p, 0); +} + + +static void emit_alu2( struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + GLuint i; + + if (mask & SATURATE) + brw_set_saturate(p, 1); + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + func(p, dst[i], arg0[i], arg1[i]); + } + } + + if (mask & SATURATE) + brw_set_saturate(p, 0); +} + + +static void emit_mad( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2 ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MUL(p, dst[i], arg0[i], arg1[i]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_ADD(p, dst[i], dst[i], arg2[i]); + brw_set_saturate(p, 0); + } + } +} + +static void emit_trunc( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_RNDZ(p, dst[i], arg0[i]); + } + } +} + +static void emit_lrp( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2 ) +{ + GLuint i; + + /* Uses dst as a temporary: + */ + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + /* Can I use the LINE instruction for this? + */ + brw_ADD(p, dst[i], negate(arg0[i]), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst[i], arg2[i]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MAC(p, dst[i], arg0[i], arg1[i]); + brw_set_saturate(p, 0); + } + } +} + +static void emit_sop( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + GLuint cond, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_MOV(p, dst[i], brw_imm_f(0)); + brw_CMP(p, brw_null_reg(), cond, arg0[i], arg1[i]); + brw_MOV(p, dst[i], brw_imm_f(1.0)); + brw_set_predicate_control_flag_value(p, 0xff); + } + } +} + +static void emit_slt( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_L, arg0, arg1); +} + +static void emit_sle( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_LE, arg0, arg1); +} + +static void emit_sgt( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_G, arg0, arg1); +} + +static void emit_sge( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_GE, arg0, arg1); +} + +static void emit_seq( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_EQ, arg0, arg1); +} + +static void emit_sne( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + emit_sop(p, dst, mask, BRW_CONDITIONAL_NEQ, arg0, arg1); +} + +static void emit_cmp( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2 ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg2[i]); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0[i], brw_imm_f(0)); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg1[i]); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } +} + +static void emit_max( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg0[i]); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0[i], arg1[i]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg1[i]); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } +} + +static void emit_min( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg1[i]); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0[i], arg1[i]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[i], arg0[i]); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } +} + + +static void emit_dp3( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + int dst_chan = ffs(mask & BRW_WRITEMASK_XYZW) - 1; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; /* Do not emit dead code */ + + assert(util_is_power_of_two(mask & BRW_WRITEMASK_XYZW)); + + brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]); + brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MAC(p, dst[dst_chan], arg0[2], arg1[2]); + brw_set_saturate(p, 0); +} + + +static void emit_dp4( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + int dst_chan = ffs(mask & BRW_WRITEMASK_XYZW) - 1; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; /* Do not emit dead code */ + + assert(util_is_power_of_two(mask & BRW_WRITEMASK_XYZW)); + + brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]); + brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]); + brw_MAC(p, brw_null_reg(), arg0[2], arg1[2]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MAC(p, dst[dst_chan], arg0[3], arg1[3]); + brw_set_saturate(p, 0); +} + + +static void emit_dph( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + const int dst_chan = ffs(mask & BRW_WRITEMASK_XYZW) - 1; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; /* Do not emit dead code */ + + assert(util_is_power_of_two(mask & BRW_WRITEMASK_XYZW)); + + brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]); + brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]); + brw_MAC(p, dst[dst_chan], arg0[2], arg1[2]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_ADD(p, dst[dst_chan], dst[dst_chan], arg1[3]); + brw_set_saturate(p, 0); +} + + +static void emit_xpd( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1 ) +{ + GLuint i; + + assert(!(mask & BRW_WRITEMASK_W) == BRW_WRITEMASK_X); + + for (i = 0 ; i < 3; i++) { + if (mask & (1<<i)) { + GLuint i2 = (i+2)%3; + GLuint i1 = (i+1)%3; + + brw_MUL(p, brw_null_reg(), negate(arg0[i2]), arg1[i1]); + + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MAC(p, dst[i], arg0[i1], arg1[i2]); + brw_set_saturate(p, 0); + } + } +} + + +static void emit_math1( struct brw_compile *p, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0 ) +{ + int dst_chan = ffs(mask & BRW_WRITEMASK_XYZW) - 1; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; /* Do not emit dead code */ + + assert(util_is_power_of_two(mask & BRW_WRITEMASK_XYZW)); + + brw_MOV(p, brw_message_reg(2), arg0[0]); + + /* Send two messages to perform all 16 operations: + */ + brw_math_16(p, + dst[dst_chan], + function, + (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_PRECISION_FULL); +} + + +static void emit_math2( struct brw_compile *p, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) +{ + int dst_chan = ffs(mask & BRW_WRITEMASK_XYZW) - 1; + + if (!(mask & BRW_WRITEMASK_XYZW)) + return; /* Do not emit dead code */ + + assert(util_is_power_of_two(mask & BRW_WRITEMASK_XYZW)); + + brw_push_insn_state(p); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, brw_message_reg(2), arg0[0]); + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, brw_message_reg(4), sechalf(arg0[0])); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, brw_message_reg(3), arg1[0]); + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, brw_message_reg(5), sechalf(arg1[0])); + + + /* Send two messages to perform all 16 operations: + */ + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_math(p, + dst[dst_chan], + function, + (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); + + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_math(p, + offset(dst[dst_chan],1), + function, + (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 4, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); + + brw_pop_insn_state(p); +} + + + +static void emit_tex( struct brw_wm_compile *c, + const struct brw_wm_instruction *inst, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *coord, + GLuint sampler) +{ + struct brw_compile *p = &c->func; + GLuint msgLength, responseLength; + GLuint i, nr; + GLuint emit; + GLuint msg_type; + GLboolean shadow = FALSE; + + /* How many input regs are there? + */ + switch (inst->target) { + case TGSI_TEXTURE_1D: + emit = BRW_WRITEMASK_X; + nr = 1; + break; + case TGSI_TEXTURE_SHADOW1D: + emit = BRW_WRITEMASK_XW; + nr = 4; + shadow = TRUE; + break; + case TGSI_TEXTURE_2D: + emit = BRW_WRITEMASK_XY; + nr = 2; + break; + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + emit = BRW_WRITEMASK_XYW; + nr = 4; + shadow = TRUE; + break; + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + emit = BRW_WRITEMASK_XYZ; + nr = 3; + break; + default: + /* unexpected target */ + abort(); + } + + msgLength = 1; + + for (i = 0; i < nr; i++) { + static const GLuint swz[4] = {0,1,2,2}; + if (emit & (1<<i)) + brw_MOV(p, brw_message_reg(msgLength+1), coord[swz[i]]); + else + brw_MOV(p, brw_message_reg(msgLength+1), brw_imm_f(0)); + msgLength += 2; + } + + responseLength = 8; /* always */ + + if (BRW_IS_IGDNG(p->brw)) { + if (shadow) + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG; + } else { + if (shadow) + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE; + } + + brw_SAMPLE(p, + retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), + BTI_TEXTURE(inst->tex_unit), + sampler, /* sampler index */ + inst->writemask, + msg_type, + responseLength, + msgLength, + 0, + 1, + BRW_SAMPLER_SIMD_MODE_SIMD16); +} + + +static void emit_txb( struct brw_wm_compile *c, + const struct brw_wm_instruction *inst, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *coord, + GLuint sampler ) +{ + struct brw_compile *p = &c->func; + GLuint msgLength; + GLuint msg_type; + /* Shadow ignored for txb. + */ + switch (inst->target) { + case TGSI_TEXTURE_1D: + case TGSI_TEXTURE_SHADOW1D: + brw_MOV(p, brw_message_reg(2), coord[0]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + break; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + brw_MOV(p, brw_message_reg(2), coord[0]); + brw_MOV(p, brw_message_reg(4), coord[1]); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + break; + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + brw_MOV(p, brw_message_reg(2), coord[0]); + brw_MOV(p, brw_message_reg(4), coord[1]); + brw_MOV(p, brw_message_reg(6), coord[2]); + break; + default: + /* unexpected target */ + abort(); + } + + brw_MOV(p, brw_message_reg(8), coord[3]); + msgLength = 9; + + if (BRW_IS_IGDNG(p->brw)) + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; + + brw_SAMPLE(p, + retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), + BTI_TEXTURE(inst->tex_unit), + sampler, /* sampler index */ + inst->writemask, + msg_type, + 8, /* responseLength */ + msgLength, + 0, + 1, + BRW_SAMPLER_SIMD_MODE_SIMD16); +} + + +static void emit_lit( struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0 ) +{ + assert((mask & BRW_WRITEMASK_XW) == 0); + + if (mask & BRW_WRITEMASK_Y) { + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[1], arg0[0]); + brw_set_saturate(p, 0); + } + + if (mask & BRW_WRITEMASK_Z) { + emit_math2(p, BRW_MATH_FUNCTION_POW, + &dst[2], + BRW_WRITEMASK_X | (mask & SATURATE), + &arg0[1], + &arg0[3]); + } + + /* Ordinarily you'd use an iff statement to skip or shortcircuit + * some of the POW calculations above, but 16-wide iff statements + * seem to lock c1 hardware, so this is a nasty workaround: + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_LE, arg0[0], brw_imm_f(0)); + { + if (mask & BRW_WRITEMASK_Y) + brw_MOV(p, dst[1], brw_imm_f(0)); + + if (mask & BRW_WRITEMASK_Z) + brw_MOV(p, dst[2], brw_imm_f(0)); + } + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + + +/* Kill pixel - set execution mask to zero for those pixels which + * fail. + */ +static void emit_kil( struct brw_wm_compile *c, + struct brw_reg *arg0) +{ + struct brw_compile *p = &c->func; + struct brw_reg r0uw = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + GLuint i; + + /* XXX - usually won't need 4 compares! + */ + for (i = 0; i < 4; i++) { + brw_push_insn_state(p); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0[i], brw_imm_f(0)); + brw_set_predicate_control_flag_value(p, 0xff); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_AND(p, r0uw, brw_flag_reg(), r0uw); + brw_pop_insn_state(p); + } +} + +/* KILLP kills the pixels that are currently executing, not based on a test + * of the arguments. + */ +static void emit_killp( struct brw_wm_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg r0uw = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); /* IMASK */ + brw_AND(p, r0uw, c->emit_mask_reg, r0uw); + brw_pop_insn_state(p); +} + +static void fire_fb_write( struct brw_wm_compile *c, + GLuint base_reg, + GLuint nr, + GLuint target, + GLuint eot ) +{ + struct brw_compile *p = &c->func; + + /* Pass through control information: + */ +/* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + + /* Send framebuffer write message: */ +/* send (16) null.0<1>:uw m0 r0.0<8;8,1>:uw 0x85a04000:ud { Align1 EOT } */ + brw_fb_WRITE(p, + retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + target, + nr, + 0, + eot); +} + + +static void emit_aa( struct brw_wm_compile *c, + struct brw_reg *arg1, + GLuint reg ) +{ + struct brw_compile *p = &c->func; + GLuint comp = c->key.aa_dest_stencil_reg / 2; + GLuint off = c->key.aa_dest_stencil_reg % 2; + struct brw_reg aa = offset(arg1[comp], off); + + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); /* ?? */ + brw_MOV(p, brw_message_reg(reg), aa); + brw_pop_insn_state(p); +} + + +/* Post-fragment-program processing. Send the results to the + * framebuffer. + * \param arg0 the fragment color + * \param arg1 the pass-through depth value + * \param arg2 the shader-computed depth value + */ +static void emit_fb_write( struct brw_wm_compile *c, + struct brw_reg *arg0, + struct brw_reg *arg1, + struct brw_reg *arg2, + GLuint target, + GLuint eot) +{ + struct brw_compile *p = &c->func; + GLuint nr = 2; + GLuint channel; + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + + /* I don't really understand how this achieves the color interleave + * (ie RGBARGBA) in the result: [Do the saturation here] + */ + { + brw_push_insn_state(p); + + for (channel = 0; channel < 4; channel++) { + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, + brw_message_reg(nr + channel), + arg0[channel]); + + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, + brw_message_reg(nr + channel + 4), + sechalf(arg0[channel])); + } + + /* skip over the regs populated above: + */ + nr += 8; + + brw_pop_insn_state(p); + } + + if (c->key.source_depth_to_render_target) + { + if (c->key.computes_depth) + brw_MOV(p, brw_message_reg(nr), arg2[2]); + else + brw_MOV(p, brw_message_reg(nr), arg1[1]); /* ? */ + + nr += 2; + } + + if (c->key.dest_depth_reg) + { + GLuint comp = c->key.dest_depth_reg / 2; + GLuint off = c->key.dest_depth_reg % 2; + + if (off != 0) { + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + brw_MOV(p, brw_message_reg(nr), offset(arg1[comp],1)); + /* 2nd half? */ + brw_MOV(p, brw_message_reg(nr+1), arg1[comp+1]); + brw_pop_insn_state(p); + } + else { + brw_MOV(p, brw_message_reg(nr), arg1[comp]); + } + nr += 2; + } + + if (!c->key.runtime_check_aads_emit) { + if (c->key.aa_dest_stencil_reg) + emit_aa(c, arg1, 2); + + fire_fb_write(c, 0, nr, target, eot); + } + else { + struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); + struct brw_reg ip = brw_ip_reg(); + struct brw_instruction *jmp; + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, + v1_null_ud, + get_element_ud(brw_vec8_grf(1,0), 6), + brw_imm_ud(1<<26)); + + jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); + { + emit_aa(c, arg1, 2); + fire_fb_write(c, 0, nr, target, eot); + /* note - thread killed in subroutine */ + } + brw_land_fwd_jump(p, jmp); + + /* ELSE: Shuffle up one register to fill in the hole left for AA: + */ + fire_fb_write(c, 1, nr-1, target, eot); + } +} + + +/** + * Move a GPR to scratch memory. + */ +static void emit_spill( struct brw_wm_compile *c, + struct brw_reg reg, + GLuint slot ) +{ + struct brw_compile *p = &c->func; + + /* + mov (16) m2.0<1>:ud r2.0<8;8,1>:ud { Align1 Compr } + */ + brw_MOV(p, brw_message_reg(2), reg); + + /* + mov (1) r0.2<1>:d 0x00000080:d { Align1 NoMask } + send (16) null.0<1>:uw m1 r0.0<8;8,1>:uw 0x053003ff:ud { Align1 } + */ + brw_dp_WRITE_16(p, + retype(vec16(brw_vec8_grf(0, 0)), BRW_REGISTER_TYPE_UW), + slot); +} + + +/** + * Load a GPR from scratch memory. + */ +static void emit_unspill( struct brw_wm_compile *c, + struct brw_reg reg, + GLuint slot ) +{ + struct brw_compile *p = &c->func; + + /* Slot 0 is the undef value. + */ + if (slot == 0) { + brw_MOV(p, reg, brw_imm_f(0)); + return; + } + + /* + mov (1) r0.2<1>:d 0x000000c0:d { Align1 NoMask } + send (16) r110.0<1>:uw m1 r0.0<8;8,1>:uw 0x041243ff:ud { Align1 } + */ + + brw_dp_READ_16(p, + retype(vec16(reg), BRW_REGISTER_TYPE_UW), + slot); +} + + +/** + * Retrieve up to 4 GEN4 register pairs for the given wm reg: + * Args with unspill_reg != 0 will be loaded from scratch memory. + */ +static void get_argument_regs( struct brw_wm_compile *c, + struct brw_wm_ref *arg[], + struct brw_reg *regs ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + if (arg[i]) { + if (arg[i]->unspill_reg) + emit_unspill(c, + brw_vec8_grf(arg[i]->unspill_reg, 0), + arg[i]->value->spill_slot); + + regs[i] = arg[i]->hw_reg; + } + else { + regs[i] = brw_null_reg(); + } + } +} + + +/** + * For values that have a spill_slot!=0, write those regs to scratch memory. + */ +static void spill_values( struct brw_wm_compile *c, + struct brw_wm_value *values, + GLuint nr ) +{ + GLuint i; + + for (i = 0; i < nr; i++) + if (values[i].spill_slot) + emit_spill(c, values[i].hw_reg, values[i].spill_slot); +} + + +/* Emit the fragment program instructions here. + */ +void brw_wm_emit( struct brw_wm_compile *c ) +{ + struct brw_compile *p = &c->func; + GLuint insn; + + brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); + + /* Check if any of the payload regs need to be spilled: + */ + spill_values(c, c->payload.depth, 4); + spill_values(c, c->creg, c->nr_creg); + spill_values(c, c->payload.input_interp, PIPE_MAX_SHADER_INPUTS); + + + for (insn = 0; insn < c->nr_insns; insn++) { + + struct brw_wm_instruction *inst = &c->instruction[insn]; + struct brw_reg args[3][4], dst[4]; + GLuint i, dst_flags; + + /* Get argument regs: + */ + for (i = 0; i < 3; i++) + get_argument_regs(c, inst->src[i], args[i]); + + /* Get dest regs: + */ + for (i = 0; i < 4; i++) + if (inst->dst[i]) + dst[i] = inst->dst[i]->hw_reg; + else + dst[i] = brw_null_reg(); + + /* Flags + */ + dst_flags = inst->writemask; + if (inst->saturate) + dst_flags |= SATURATE; + + switch (inst->opcode) { + /* Generated instructions for calculating triangle interpolants: + */ + case WM_PIXELXY: + emit_pixel_xy(p, dst, dst_flags); + break; + + case WM_DELTAXY: + emit_delta_xy(p, dst, dst_flags, args[0]); + break; + + case WM_WPOSXY: + emit_wpos_xy(c, dst, dst_flags, args[0]); + break; + + case WM_PIXELW: + emit_pixel_w(p, dst, dst_flags, args[0], args[1]); + break; + + case WM_LINTERP: + emit_linterp(p, dst, dst_flags, args[0], args[1]); + break; + + case WM_PINTERP: + emit_pinterp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case WM_CINTERP: + emit_cinterp(p, dst, dst_flags, args[0]); + break; + + case WM_FB_WRITE: + emit_fb_write(c, args[0], args[1], args[2], inst->target, inst->eot); + break; + + case WM_FRONTFACING: + emit_frontfacing(p, dst, dst_flags); + break; + + /* Straightforward arithmetic: + */ + case TGSI_OPCODE_ADD: + emit_alu2(p, brw_ADD, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_FRC: + emit_alu1(p, brw_FRC, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_FLR: + emit_alu1(p, brw_RNDD, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_DDX: + emit_ddxy(p, dst, dst_flags, GL_TRUE, args[0]); + break; + + case TGSI_OPCODE_DDY: + emit_ddxy(p, dst, dst_flags, GL_FALSE, args[0]); + break; + + case TGSI_OPCODE_DP3: + emit_dp3(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_DP4: + emit_dp4(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_DPH: + emit_dph(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_TRUNC: + emit_trunc(p, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_LRP: + emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MAD: + emit_mad(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MOV: + emit_alu1(p, brw_MOV, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_MUL: + emit_alu2(p, brw_MUL, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, dst_flags, args[0], args[1]); + break; + + /* Higher math functions: + */ + case TGSI_OPCODE_RCP: + emit_math1(p, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_RSQ: + emit_math1(p, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_SIN: + emit_math1(p, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_COS: + emit_math1(p, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_EX2: + emit_math1(p, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_LG2: + emit_math1(p, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_SCS: + /* There is an scs math function, but it would need some + * fixup for 16-element execution. + */ + if (dst_flags & BRW_WRITEMASK_X) + emit_math1(p, BRW_MATH_FUNCTION_COS, dst, (dst_flags&SATURATE)|BRW_WRITEMASK_X, args[0]); + if (dst_flags & BRW_WRITEMASK_Y) + emit_math1(p, BRW_MATH_FUNCTION_SIN, dst+1, (dst_flags&SATURATE)|BRW_WRITEMASK_X, args[0]); + break; + + case TGSI_OPCODE_POW: + emit_math2(p, BRW_MATH_FUNCTION_POW, dst, dst_flags, args[0], args[1]); + break; + + /* Comparisons: + */ + case TGSI_OPCODE_CMP: + emit_cmp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MAX: + emit_max(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_MIN: + emit_min(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_SLT: + emit_slt(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_SLE: + emit_sle(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_LIT: + emit_lit(p, dst, dst_flags, args[0]); + break; + + /* Texturing operations: + */ + case TGSI_OPCODE_TEX: + emit_tex(c, inst, dst, dst_flags, args[0], inst->sampler); + break; + + case TGSI_OPCODE_TXB: + emit_txb(c, inst, dst, dst_flags, args[0], inst->sampler); + break; + + case TGSI_OPCODE_KIL: + emit_kil(c, args[0]); + break; + + case TGSI_OPCODE_KILP: + emit_killp(c); + break; + + default: + debug_printf("Unsupported opcode %i (%s) in fragment shader\n", + inst->opcode, + tgsi_get_opcode_info(inst->opcode)->mnemonic); + } + + for (i = 0; i < 4; i++) + if (inst->dst[i] && inst->dst[i]->spill_slot) + emit_spill(c, + inst->dst[i]->hw_reg, + inst->dst[i]->spill_slot); + } + + if (BRW_DEBUG & DEBUG_WM) { + debug_printf("wm-native:\n"); + brw_disasm(stderr, p->store, p->nr_insn); + } +} diff --git a/src/gallium/drivers/i965/brw_wm_fp.c b/src/gallium/drivers/i965/brw_wm_fp.c new file mode 100644 index 00000000000..9c5b527f897 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_fp.c @@ -0,0 +1,1224 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "pipe/p_shader_tokens.h" + +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_info.h" +#include "tgsi/tgsi_util.h" + +#include "brw_wm.h" +#include "brw_util.h" +#include "brw_debug.h" + + +/*********************************************************************** + * Source regs + */ + +static struct brw_fp_src src_reg(GLuint file, GLuint idx) +{ + struct brw_fp_src reg; + reg.file = file; + reg.index = idx; + reg.swizzle = BRW_SWIZZLE_XYZW; + reg.indirect = 0; + reg.negate = 0; + reg.abs = 0; + return reg; +} + +static struct brw_fp_src src_reg_from_dst(struct brw_fp_dst dst) +{ + return src_reg(dst.file, dst.index); +} + +static struct brw_fp_src src_undef( void ) +{ + return src_reg(TGSI_FILE_NULL, 0); +} + +static GLboolean src_is_undef(struct brw_fp_src src) +{ + return src.file == TGSI_FILE_NULL; +} + +static struct brw_fp_src src_swizzle( struct brw_fp_src reg, int x, int y, int z, int w ) +{ + unsigned swz = reg.swizzle; + + reg.swizzle = ( BRW_GET_SWZ(swz, x) << 0 | + BRW_GET_SWZ(swz, y) << 2 | + BRW_GET_SWZ(swz, z) << 4 | + BRW_GET_SWZ(swz, w) << 6 ); + + return reg; +} + +static struct brw_fp_src src_scalar( struct brw_fp_src reg, int x ) +{ + return src_swizzle(reg, x, x, x, x); +} + +static struct brw_fp_src src_abs( struct brw_fp_src src ) +{ + src.negate = 0; + src.abs = 1; + return src; +} + +static struct brw_fp_src src_negate( struct brw_fp_src src ) +{ + src.negate = 1; + src.abs = 0; + return src; +} + + +static int match_or_expand_immediate( const float *v, + unsigned nr, + float *v2, + unsigned *nr2, + unsigned *swizzle ) +{ + unsigned i, j; + + *swizzle = 0; + + for (i = 0; i < nr; i++) { + boolean found = FALSE; + + for (j = 0; j < *nr2 && !found; j++) { + if (v[i] == v2[j]) { + *swizzle |= j << (i * 2); + found = TRUE; + } + } + + if (!found) { + if (*nr2 >= 4) + return FALSE; + + v2[*nr2] = v[i]; + *swizzle |= *nr2 << (i * 2); + (*nr2)++; + } + } + + return TRUE; +} + + + +/* Internally generated immediates: overkill... + */ +static struct brw_fp_src src_imm( struct brw_wm_compile *c, + const GLfloat *v, + unsigned nr) +{ + unsigned i, j; + unsigned swizzle; + + /* Could do a first pass where we examine all existing immediates + * without expanding. + */ + + for (i = 0; i < c->nr_immediates; i++) { + if (match_or_expand_immediate( v, + nr, + c->immediate[i].v, + &c->immediate[i].nr, + &swizzle )) + goto out; + } + + if (c->nr_immediates < Elements(c->immediate)) { + i = c->nr_immediates++; + if (match_or_expand_immediate( v, + nr, + c->immediate[i].v, + &c->immediate[i].nr, + &swizzle )) + goto out; + } + + c->error = 1; + return src_undef(); + +out: + /* Make sure that all referenced elements are from this immediate. + * Has the effect of making size-one immediates into scalars. + */ + for (j = nr; j < 4; j++) + swizzle |= (swizzle & 0x3) << (j * 2); + + return src_swizzle( src_reg( TGSI_FILE_IMMEDIATE, i ), + BRW_GET_SWZ(swizzle, X), + BRW_GET_SWZ(swizzle, Y), + BRW_GET_SWZ(swizzle, Z), + BRW_GET_SWZ(swizzle, W) ); +} + + + +static struct brw_fp_src src_imm1f( struct brw_wm_compile *c, + GLfloat f ) +{ + return src_imm(c, &f, 1); +} + +static struct brw_fp_src src_imm4f( struct brw_wm_compile *c, + GLfloat x, + GLfloat y, + GLfloat z, + GLfloat w) +{ + GLfloat f[4] = {x,y,z,w}; + return src_imm(c, f, 4); +} + + + +/*********************************************************************** + * Dest regs + */ + +static struct brw_fp_dst dst_reg(GLuint file, GLuint idx) +{ + struct brw_fp_dst reg; + reg.file = file; + reg.index = idx; + reg.writemask = BRW_WRITEMASK_XYZW; + reg.indirect = 0; + reg.saturate = 0; + return reg; +} + +static struct brw_fp_dst dst_mask( struct brw_fp_dst reg, int mask ) +{ + reg.writemask &= mask; + return reg; +} + +static struct brw_fp_dst dst_undef( void ) +{ + return dst_reg(TGSI_FILE_NULL, 0); +} + +static boolean dst_is_undef( struct brw_fp_dst dst ) +{ + return dst.file == TGSI_FILE_NULL; +} + +static struct brw_fp_dst dst_saturate( struct brw_fp_dst reg, boolean flag ) +{ + reg.saturate = flag; + return reg; +} + +static struct brw_fp_dst get_temp( struct brw_wm_compile *c ) +{ + int bit = ffs( ~c->fp_temp ); + + if (!bit) { + debug_printf("%s: out of temporaries\n", __FILE__); + } + + c->fp_temp |= 1<<(bit-1); + return dst_reg(TGSI_FILE_TEMPORARY, c->fp_first_internal_temp+(bit-1)); +} + + +static void release_temp( struct brw_wm_compile *c, struct brw_fp_dst temp ) +{ + c->fp_temp &= ~(1 << (temp.index - c->fp_first_internal_temp)); +} + + +/*********************************************************************** + * Instructions + */ + +static struct brw_fp_instruction *get_fp_inst(struct brw_wm_compile *c) +{ + return &c->fp_instructions[c->nr_fp_insns++]; +} + +static struct brw_fp_instruction * emit_tex_op(struct brw_wm_compile *c, + GLuint op, + struct brw_fp_dst dest, + GLuint tex_unit, + GLuint target, + GLuint sampler, + struct brw_fp_src src0, + struct brw_fp_src src1, + struct brw_fp_src src2 ) +{ + struct brw_fp_instruction *inst = get_fp_inst(c); + + if (tex_unit || target) + assert(op == TGSI_OPCODE_TXP || + op == TGSI_OPCODE_TXB || + op == TGSI_OPCODE_TEX || + op == WM_FB_WRITE); + + inst->opcode = op; + inst->dst = dest; + inst->tex_unit = tex_unit; + inst->target = target; + inst->sampler = sampler; + inst->src[0] = src0; + inst->src[1] = src1; + inst->src[2] = src2; + + return inst; +} + + +static INLINE void emit_op3(struct brw_wm_compile *c, + GLuint op, + struct brw_fp_dst dest, + struct brw_fp_src src0, + struct brw_fp_src src1, + struct brw_fp_src src2 ) +{ + emit_tex_op(c, op, dest, 0, 0, 0, src0, src1, src2); +} + + +static INLINE void emit_op2(struct brw_wm_compile *c, + GLuint op, + struct brw_fp_dst dest, + struct brw_fp_src src0, + struct brw_fp_src src1) +{ + emit_tex_op(c, op, dest, 0, 0, 0, src0, src1, src_undef()); +} + +static INLINE void emit_op1(struct brw_wm_compile *c, + GLuint op, + struct brw_fp_dst dest, + struct brw_fp_src src0) +{ + emit_tex_op(c, op, dest, 0, 0, 0, src0, src_undef(), src_undef()); +} + +static INLINE void emit_op0(struct brw_wm_compile *c, + GLuint op, + struct brw_fp_dst dest) +{ + emit_tex_op(c, op, dest, 0, 0, 0, src_undef(), src_undef(), src_undef()); +} + + + +/* Many opcodes produce the same value across all the result channels. + * We'd rather not have to support that splatting in the opcode implementations, + * and brw_wm_pass*.c wants to optimize them out by shuffling references around + * anyway. We can easily get both by emitting the opcode to one channel, and + * then MOVing it to the others, which brw_wm_pass*.c already understands. + */ +static void emit_scalar_insn(struct brw_wm_compile *c, + unsigned opcode, + struct brw_fp_dst dst, + struct brw_fp_src src0, + struct brw_fp_src src1, + struct brw_fp_src src2 ) +{ + unsigned first_chan = ffs(dst.writemask) - 1; + unsigned first_mask = 1 << first_chan; + + if (dst.writemask == 0) + return; + + emit_op3( c, opcode, + dst_mask(dst, first_mask), + src0, src1, src2 ); + + if (dst.writemask != first_mask) { + emit_op1(c, TGSI_OPCODE_MOV, + dst_mask(dst, ~first_mask), + src_scalar(src_reg_from_dst(dst), first_chan)); + } +} + + +/*********************************************************************** + * Special instructions for interpolation and other tasks + */ + +static struct brw_fp_src get_pixel_xy( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->fp_pixel_xy)) { + struct brw_fp_dst pixel_xy = get_temp(c); + struct brw_fp_src payload_r0_depth = src_reg(BRW_FILE_PAYLOAD, PAYLOAD_DEPTH); + + + /* Emit the out calculations, and hold onto the results. Use + * two instructions as a temporary is required. + */ + /* pixel_xy.xy = PIXELXY payload[0]; + */ + emit_op1(c, + WM_PIXELXY, + dst_mask(pixel_xy, BRW_WRITEMASK_XY), + payload_r0_depth); + + c->fp_pixel_xy = src_reg_from_dst(pixel_xy); + } + + return c->fp_pixel_xy; +} + +static struct brw_fp_src get_delta_xy( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->fp_delta_xy)) { + struct brw_fp_dst delta_xy = get_temp(c); + struct brw_fp_src pixel_xy = get_pixel_xy(c); + struct brw_fp_src payload_r0_depth = src_reg(BRW_FILE_PAYLOAD, PAYLOAD_DEPTH); + + /* deltas.xy = DELTAXY pixel_xy, payload[0] + */ + emit_op3(c, + WM_DELTAXY, + dst_mask(delta_xy, BRW_WRITEMASK_XY), + pixel_xy, + payload_r0_depth, + src_undef()); + + c->fp_delta_xy = src_reg_from_dst(delta_xy); + } + + return c->fp_delta_xy; +} + +static struct brw_fp_src get_pixel_w( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->fp_pixel_w)) { + struct brw_fp_dst pixel_w = get_temp(c); + struct brw_fp_src deltas = get_delta_xy(c); + + /* XXX: assuming position is always first -- valid? + */ + struct brw_fp_src interp_wpos = src_reg(BRW_FILE_PAYLOAD, 0); + + /* deltas.xyw = DELTAS2 deltas.xy, payload.interp_wpos.x + */ + emit_op3(c, + WM_PIXELW, + dst_mask(pixel_w, BRW_WRITEMASK_W), + interp_wpos, + deltas, + src_undef()); + + + c->fp_pixel_w = src_reg_from_dst(pixel_w); + } + + return c->fp_pixel_w; +} + + +/*********************************************************************** + * Emit INTERP instructions ahead of first use of each attrib. + */ + +static void emit_interp( struct brw_wm_compile *c, + GLuint idx, + GLuint semantic, + GLuint interp_mode ) +{ + struct brw_fp_dst dst = dst_reg(TGSI_FILE_INPUT, idx); + struct brw_fp_src interp = src_reg(BRW_FILE_PAYLOAD, idx); + struct brw_fp_src deltas = get_delta_xy(c); + + /* Need to use PINTERP on attributes which have been + * multiplied by 1/W in the SF program, and LINTERP on those + * which have not: + */ + switch (semantic) { + case TGSI_SEMANTIC_POSITION: + /* Have to treat wpos.xy specially: + */ + emit_op1(c, + WM_WPOSXY, + dst_mask(dst, BRW_WRITEMASK_XY), + get_pixel_xy(c)); + + /* TGSI_FILE_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw + */ + emit_op2(c, + WM_LINTERP, + dst_mask(dst, BRW_WRITEMASK_ZW), + interp, + deltas); + break; + + case TGSI_SEMANTIC_COLOR: + if (c->key.flat_shade) { + emit_op1(c, + WM_CINTERP, + dst, + interp); + } + else if (interp_mode == TGSI_INTERPOLATE_LINEAR) { + emit_op2(c, + WM_LINTERP, + dst, + interp, + deltas); + } + else { + emit_op3(c, + WM_PINTERP, + dst, + interp, + deltas, + get_pixel_w(c)); + } + + break; + + case TGSI_SEMANTIC_FOG: + /* Interpolate the fog coordinate */ + emit_op3(c, + WM_PINTERP, + dst_mask(dst, BRW_WRITEMASK_X), + interp, + deltas, + get_pixel_w(c)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_YZ), + src_imm1f(c, 0.0)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_W), + src_imm1f(c, 1.0)); + break; + + case TGSI_SEMANTIC_FACE: + /* XXX review/test this case */ + emit_op0(c, + WM_FRONTFACING, + dst_mask(dst, BRW_WRITEMASK_X)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_YZ), + src_imm1f(c, 0.0)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_W), + src_imm1f(c, 1.0)); + break; + + case TGSI_SEMANTIC_PSIZE: + /* XXX review/test this case */ + emit_op3(c, + WM_PINTERP, + dst_mask(dst, BRW_WRITEMASK_XY), + interp, + deltas, + get_pixel_w(c)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_Z), + src_imm1f(c, 0.0f)); + + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_W), + src_imm1f(c, 1.0f)); + break; + + default: + switch (interp_mode) { + case TGSI_INTERPOLATE_CONSTANT: + emit_op1(c, + WM_CINTERP, + dst, + interp); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_op2(c, + WM_LINTERP, + dst, + interp, + deltas); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + emit_op3(c, + WM_PINTERP, + dst, + interp, + deltas, + get_pixel_w(c)); + break; + } + break; + } +} + + +/*********************************************************************** + * Expand various instructions here to simpler forms. + */ +static void precalc_dst( struct brw_wm_compile *c, + struct brw_fp_dst dst, + struct brw_fp_src src0, + struct brw_fp_src src1 ) +{ + if (dst.writemask & BRW_WRITEMASK_Y) { + /* dst.y = mul src0.y, src1.y + */ + emit_op2(c, + TGSI_OPCODE_MUL, + dst_mask(dst, BRW_WRITEMASK_Y), + src0, + src1); + } + + if (dst.writemask & BRW_WRITEMASK_XZ) { + /* dst.z = mov src0.zzzz + */ + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_Z), + src_scalar(src0, Z)); + + /* dst.x = imm1f(1.0) + */ + emit_op1(c, + TGSI_OPCODE_MOV, + dst_saturate(dst_mask(dst, BRW_WRITEMASK_X), 0), + src_imm1f(c, 1.0)); + } + if (dst.writemask & BRW_WRITEMASK_W) { + /* dst.w = mov src1.w + */ + emit_op1(c, + TGSI_OPCODE_MOV, + dst_mask(dst, BRW_WRITEMASK_W), + src1); + } +} + + +static void precalc_lit( struct brw_wm_compile *c, + struct brw_fp_dst dst, + struct brw_fp_src src0 ) +{ + if (dst.writemask & BRW_WRITEMASK_XW) { + /* dst.xw = imm(1.0f) + */ + emit_op1(c, + TGSI_OPCODE_MOV, + dst_saturate(dst_mask(dst, BRW_WRITEMASK_XW), 0), + src_imm1f(c, 1.0f)); + } + + if (dst.writemask & BRW_WRITEMASK_YZ) { + emit_op1(c, + TGSI_OPCODE_LIT, + dst_mask(dst, BRW_WRITEMASK_YZ), + src0); + } +} + + +/** + * Some TEX instructions require extra code, cube map coordinate + * normalization, or coordinate scaling for RECT textures, etc. + * This function emits those extra instructions and the TEX + * instruction itself. + */ +static void precalc_tex( struct brw_wm_compile *c, + struct brw_fp_dst dst, + unsigned target, + unsigned unit, + struct brw_fp_src src0, + struct brw_fp_src sampler ) +{ + struct brw_fp_src coord = src_undef(); + struct brw_fp_dst tmp = dst_undef(); + + assert(unit < BRW_MAX_TEX_UNIT); + + /* Cubemap: find longest component of coord vector and normalize + * it. + */ + if (target == TGSI_TEXTURE_CUBE) { + struct brw_fp_src tmpsrc; + + tmp = get_temp(c); + tmpsrc = src_reg_from_dst(tmp); + + /* tmp = abs(src0) */ + emit_op1(c, + TGSI_OPCODE_MOV, + tmp, + src_abs(src0)); + + /* tmp.X = MAX(tmp.X, tmp.Y) */ + emit_op2(c, TGSI_OPCODE_MAX, + dst_mask(tmp, BRW_WRITEMASK_X), + src_scalar(tmpsrc, X), + src_scalar(tmpsrc, Y)); + + /* tmp.X = MAX(tmp.X, tmp.Z) */ + emit_op2(c, TGSI_OPCODE_MAX, + dst_mask(tmp, BRW_WRITEMASK_X), + tmpsrc, + src_scalar(tmpsrc, Z)); + + /* tmp.X = 1 / tmp.X */ + emit_op1(c, TGSI_OPCODE_RCP, + dst_mask(tmp, BRW_WRITEMASK_X), + tmpsrc); + + /* tmp = src0 * tmp.xxxx */ + emit_op2(c, TGSI_OPCODE_MUL, + tmp, + src0, + src_scalar(tmpsrc, X)); + + coord = tmpsrc; + } + else if (target == TGSI_TEXTURE_RECT || + target == TGSI_TEXTURE_SHADOWRECT) { + /* XXX: need a mechanism for internally generated constants. + */ + coord = src0; + } + else { + coord = src0; + } + + /* Need to emit YUV texture conversions by hand. Probably need to + * do this here - the alternative is in brw_wm_emit.c, but the + * conversion requires allocating a temporary variable which we + * don't have the facility to do that late in the compilation. + */ + if (c->key.yuvtex_mask & (1 << unit)) { + /* convert ycbcr to RGBA */ + GLboolean swap_uv = c->key.yuvtex_swap_mask & (1<<unit); + struct brw_fp_dst tmp = get_temp(c); + struct brw_fp_src tmpsrc = src_reg_from_dst(tmp); + struct brw_fp_src C0 = src_imm4f( c, -.5, -.0625, -.5, 1.164 ); + struct brw_fp_src C1 = src_imm4f( c, 1.596, -0.813, 2.018, -.391 ); + + /* tmp = TEX ... + */ + emit_tex_op(c, + TGSI_OPCODE_TEX, + dst_saturate(tmp, dst.saturate), + unit, + target, + sampler.index, + coord, + src_undef(), + src_undef()); + + /* tmp.xyz = ADD TMP, C0 + */ + emit_op2(c, TGSI_OPCODE_ADD, + dst_mask(tmp, BRW_WRITEMASK_XYZ), + tmpsrc, + C0); + + /* YUV.y = MUL YUV.y, C0.w + */ + emit_op2(c, TGSI_OPCODE_MUL, + dst_mask(tmp, BRW_WRITEMASK_Y), + tmpsrc, + src_scalar(C0, W)); + + /* + * if (UV swaped) + * RGB.xyz = MAD YUV.zzx, C1, YUV.y + * else + * RGB.xyz = MAD YUV.xxz, C1, YUV.y + */ + + emit_op3(c, TGSI_OPCODE_MAD, + dst_mask(dst, BRW_WRITEMASK_XYZ), + ( swap_uv ? + src_swizzle(tmpsrc, Z,Z,X,X) : + src_swizzle(tmpsrc, X,X,Z,Z)), + C1, + src_scalar(tmpsrc, Y)); + + /* RGB.y = MAD YUV.z, C1.w, RGB.y + */ + emit_op3(c, + TGSI_OPCODE_MAD, + dst_mask(dst, BRW_WRITEMASK_Y), + src_scalar(tmpsrc, Z), + src_scalar(C1, W), + src_scalar(src_reg_from_dst(dst), Y)); + + release_temp(c, tmp); + } + else { + /* ordinary RGBA tex instruction */ + emit_tex_op(c, + TGSI_OPCODE_TEX, + dst, + unit, + target, + sampler.index, + coord, + src_undef(), + src_undef()); + } + + /* XXX: add GL_EXT_texture_swizzle support to gallium -- by + * generating shader varients in mesa state tracker. + */ + + /* Release this temp if we ended up allocating it: + */ + if (!dst_is_undef(tmp)) + release_temp(c, tmp); +} + + +/** + * Check if the given TXP instruction really needs the divide-by-W step. + */ +static GLboolean projtex( struct brw_wm_compile *c, + unsigned target, + struct brw_fp_src src ) +{ + /* Only try to detect the simplest cases. Could detect (later) + * cases where we are trying to emit code like RCP {1.0}, MUL x, + * {1.0}, and so on. + * + * More complex cases than this typically only arise from + * user-provided fragment programs anyway: + */ + if (target == TGSI_TEXTURE_CUBE) + return GL_FALSE; /* ut2004 gun rendering !?! */ + + if (src.file == TGSI_FILE_INPUT && + BRW_GET_SWZ(src.swizzle, W) == W && + c->fp->info.input_interpolate[src.index] != TGSI_INTERPOLATE_PERSPECTIVE) + return GL_FALSE; + + return GL_TRUE; +} + + +/** + * Emit code for TXP. + */ +static void precalc_txp( struct brw_wm_compile *c, + struct brw_fp_dst dst, + unsigned target, + unsigned unit, + struct brw_fp_src src0, + struct brw_fp_src sampler ) +{ + if (projtex(c, target, src0)) { + struct brw_fp_dst tmp = get_temp(c); + + /* tmp0.w = RCP inst.arg[0][3] + */ + emit_op1(c, + TGSI_OPCODE_RCP, + dst_mask(tmp, BRW_WRITEMASK_W), + src_scalar(src0, W)); + + /* tmp0.xyz = MUL inst.arg[0], tmp0.wwww + */ + emit_op2(c, + TGSI_OPCODE_MUL, + dst_mask(tmp, BRW_WRITEMASK_XYZ), + src0, + src_scalar(src_reg_from_dst(tmp), W)); + + /* dst = TEX tmp0 + */ + precalc_tex(c, + dst, + target, + unit, + src_reg_from_dst(tmp), + sampler ); + + release_temp(c, tmp); + } + else + { + /* dst = TEX src0 + */ + precalc_tex(c, dst, target, unit, src0, sampler); + } +} + + +/* XXX: note this returns a src_reg. + */ +static struct brw_fp_src +find_output_by_semantic( struct brw_wm_compile *c, + unsigned semantic, + unsigned index ) +{ + const struct tgsi_shader_info *info = &c->fp->info; + unsigned i; + + for (i = 0; i < info->num_outputs; i++) + if (info->output_semantic_name[i] == semantic && + info->output_semantic_index[i] == index) + return src_reg( TGSI_FILE_OUTPUT, i ); + + /* If not found, return some arbitrary immediate value: + * + * XXX: this is a good idea but immediates are up generating extra + * curbe entries atm, as they would have in the original driver. + */ + return src_reg( TGSI_FILE_OUTPUT, 0 ); /* src_imm1f(c, 1.0); */ +} + + +static void emit_fb_write( struct brw_wm_compile *c ) +{ + struct brw_fp_src payload_r0_depth = src_reg(BRW_FILE_PAYLOAD, PAYLOAD_DEPTH); + struct brw_fp_src outdepth = find_output_by_semantic(c, TGSI_SEMANTIC_POSITION, 0); + GLuint i; + + + outdepth = src_scalar(outdepth, Z); + + for (i = 0 ; i < c->key.nr_cbufs; i++) { + struct brw_fp_src outcolor; + + outcolor = find_output_by_semantic(c, TGSI_SEMANTIC_COLOR, i); + + /* Use emit_tex_op so that we can specify the inst->target + * field, which is abused to contain the FB write target and the + * EOT marker + */ + emit_tex_op(c, WM_FB_WRITE, + dst_undef(), + (i == c->key.nr_cbufs - 1), /* EOT */ + i, + 0, /* no sampler */ + outcolor, + payload_r0_depth, + outdepth); + } +} + + +static struct brw_fp_dst translate_dst( struct brw_wm_compile *c, + const struct tgsi_full_dst_register *dst, + unsigned saturate ) +{ + struct brw_fp_dst out; + + out.file = dst->Register.File; + out.index = dst->Register.Index; + out.writemask = dst->Register.WriteMask; + out.indirect = dst->Register.Indirect; + out.saturate = (saturate == TGSI_SAT_ZERO_ONE); + + if (out.indirect) { + assert(dst->Indirect.File == TGSI_FILE_ADDRESS); + assert(dst->Indirect.Index == 0); + } + + return out; +} + + +static struct brw_fp_src translate_src( struct brw_wm_compile *c, + const struct tgsi_full_src_register *src ) +{ + struct brw_fp_src out; + + out.file = src->Register.File; + out.index = src->Register.Index; + out.indirect = src->Register.Indirect; + + out.swizzle = ((src->Register.SwizzleX << 0) | + (src->Register.SwizzleY << 2) | + (src->Register.SwizzleZ << 4) | + (src->Register.SwizzleW << 6)); + + switch (tgsi_util_get_full_src_register_sign_mode( src, 0 )) { + case TGSI_UTIL_SIGN_CLEAR: + out.abs = 1; + out.negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + out.abs = 1; + out.negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + out.abs = 0; + out.negate = 1; + break; + + case TGSI_UTIL_SIGN_KEEP: + default: + out.abs = 0; + out.negate = 0; + break; + } + + if (out.indirect) { + assert(src->Indirect.File == TGSI_FILE_ADDRESS); + assert(src->Indirect.Index == 0); + } + + return out; +} + + + +static void emit_insn( struct brw_wm_compile *c, + const struct tgsi_full_instruction *inst ) +{ + unsigned opcode = inst->Instruction.Opcode; + struct brw_fp_dst dst; + struct brw_fp_src src[3]; + int i; + + dst = translate_dst( c, &inst->Dst[0], + inst->Instruction.Saturate ); + + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) + src[i] = translate_src( c, &inst->Src[i] ); + + switch (opcode) { + case TGSI_OPCODE_ABS: + emit_op1(c, TGSI_OPCODE_MOV, + dst, + src_abs(src[0])); + break; + + case TGSI_OPCODE_SUB: + emit_op2(c, TGSI_OPCODE_ADD, + dst, + src[0], + src_negate(src[1])); + break; + + case TGSI_OPCODE_SCS: + emit_op1(c, TGSI_OPCODE_SCS, + dst_mask(dst, BRW_WRITEMASK_XY), + src[0]); + break; + + case TGSI_OPCODE_DST: + precalc_dst(c, dst, src[0], src[1]); + break; + + case TGSI_OPCODE_LIT: + precalc_lit(c, dst, src[0]); + break; + + case TGSI_OPCODE_TEX: + precalc_tex(c, dst, + inst->Texture.Texture, + src[1].index, /* use sampler unit for tex idx */ + src[0], /* coord */ + src[1]); /* sampler */ + break; + + case TGSI_OPCODE_TXP: + precalc_txp(c, dst, + inst->Texture.Texture, + src[1].index, /* use sampler unit for tex idx */ + src[0], /* coord */ + src[1]); /* sampler */ + break; + + case TGSI_OPCODE_TXB: + /* XXX: TXB not done + */ + precalc_tex(c, dst, + inst->Texture.Texture, + src[1].index, /* use sampler unit for tex idx*/ + src[0], + src[1]); + break; + + case TGSI_OPCODE_XPD: + emit_op2(c, TGSI_OPCODE_XPD, + dst_mask(dst, BRW_WRITEMASK_XYZ), + src[0], + src[1]); + break; + + case TGSI_OPCODE_KIL: + emit_op1(c, TGSI_OPCODE_KIL, + dst_mask(dst_undef(), 0), + src[0]); + break; + + case TGSI_OPCODE_END: + emit_fb_write(c); + break; + default: + if (!c->key.has_flow_control && + brw_wm_is_scalar_result(opcode)) + emit_scalar_insn(c, opcode, dst, src[0], src[1], src[2]); + else + emit_op3(c, opcode, dst, src[0], src[1], src[2]); + break; + } +} + +/** + * Initial pass for fragment program code generation. + * This function is used by both the GLSL and non-GLSL paths. + */ +int brw_wm_pass_fp( struct brw_wm_compile *c ) +{ + struct brw_fragment_shader *fs = c->fp; + struct tgsi_parse_context parse; + struct tgsi_full_instruction *inst; + struct tgsi_full_declaration *decl; + const float *imm; + GLuint size; + GLuint i; + + if (BRW_DEBUG & DEBUG_WM) { + debug_printf("pre-fp:\n"); + tgsi_dump(fs->tokens, 0); + } + + c->fp_pixel_xy = src_undef(); + c->fp_delta_xy = src_undef(); + c->fp_pixel_w = src_undef(); + c->nr_fp_insns = 0; + c->nr_immediates = 0; + + + /* Loop over all instructions doing assorted simplifications and + * transformations. + */ + tgsi_parse_init( &parse, fs->tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* Turn intput declarations into special WM_* instructions. + * + * XXX: For non-branching shaders, consider deferring variable + * initialization as late as possible to minimize register + * usage. This is how the original BRW driver worked. + * + * In a branching shader, must preamble instructions at decl + * time, as instruction order in the shader does not + * correspond to the order instructions are executed in the + * wild. + * + * This is where special instructions such as WM_CINTERP, + * WM_LINTERP, WM_PINTERP and WM_WPOSXY are emitted to + * compute shader inputs from the payload registers and pixel + * position. + */ + decl = &parse.FullToken.FullDeclaration; + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + unsigned attrib; + + first = decl->Range.First; + last = decl->Range.Last; + mask = decl->Declaration.UsageMask; + + for (attrib = first; attrib <= last; attrib++) { + emit_interp(c, + attrib, + decl->Semantic.Name, + decl->Declaration.Interpolate ); + } + } + + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* Unlike VS programs we can probably manage fine encoding + * immediate values directly into the emitted EU + * instructions, as we probably only need to reference one + * float value per instruction. Just save the data for now + * and use directly later. + */ + i = c->nr_immediates++; + imm = &parse.FullToken.FullImmediate.u[i].Float; + size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1; + + if (c->nr_immediates >= BRW_WM_MAX_CONST) + return PIPE_ERROR_OUT_OF_MEMORY; + + for (i = 0; i < size; i++) + c->immediate[c->nr_immediates].v[i] = imm[i]; + + for (; i < 4; i++) + c->immediate[c->nr_immediates].v[i] = 0.0; + + c->immediate[c->nr_immediates].nr = size; + c->nr_immediates++; + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + inst = &parse.FullToken.FullInstruction; + emit_insn(c, inst); + break; + } + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_fp_program( c, "pass_fp" ); + debug_printf("\n"); + } + + return c->error; +} + diff --git a/src/gallium/drivers/i965/brw_wm_glsl.c b/src/gallium/drivers/i965/brw_wm_glsl.c new file mode 100644 index 00000000000..3b3afc39d3c --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_glsl.c @@ -0,0 +1,2032 @@ +#include "util/u_math.h" + + +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_wm.h" + + +static struct brw_reg get_dst_reg(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + GLuint component); + + +static void +reclaim_temps(struct brw_wm_compile *c); + + +/** Mark GRF register as used. */ +static void +prealloc_grf(struct brw_wm_compile *c, int r) +{ + c->used_grf[r] = GL_TRUE; +} + + +/** Mark given GRF register as not in use. */ +static void +release_grf(struct brw_wm_compile *c, int r) +{ + /*assert(c->used_grf[r]);*/ + c->used_grf[r] = GL_FALSE; + c->first_free_grf = MIN2(c->first_free_grf, r); +} + + +/** Return index of a free GRF, mark it as used. */ +static int +alloc_grf(struct brw_wm_compile *c) +{ + GLuint r; + for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) { + if (!c->used_grf[r]) { + c->used_grf[r] = GL_TRUE; + c->first_free_grf = r + 1; /* a guess */ + return r; + } + } + + /* no free temps, try to reclaim some */ + reclaim_temps(c); + c->first_free_grf = 0; + + /* try alloc again */ + for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) { + if (!c->used_grf[r]) { + c->used_grf[r] = GL_TRUE; + c->first_free_grf = r + 1; /* a guess */ + return r; + } + } + + for (r = 0; r < BRW_WM_MAX_GRF; r++) { + assert(c->used_grf[r]); + } + + /* really, no free GRF regs found */ + if (!c->out_of_regs) { + /* print warning once per compilation */ + debug_printf("%s: ran out of registers for fragment program", __FUNCTION__); + c->out_of_regs = GL_TRUE; + } + + return -1; +} + + +/** Return number of GRF registers used */ +static int +num_grf_used(const struct brw_wm_compile *c) +{ + int r; + for (r = BRW_WM_MAX_GRF - 1; r >= 0; r--) + if (c->used_grf[r]) + return r + 1; + return 0; +} + + + +/** + * Record the mapping of a Mesa register to a hardware register. + */ +static void set_reg(struct brw_wm_compile *c, int file, int index, + int component, struct brw_reg reg) +{ + c->wm_regs[file][index][component].reg = reg; + c->wm_regs[file][index][component].inited = GL_TRUE; +} + +static struct brw_reg alloc_tmp(struct brw_wm_compile *c) +{ + struct brw_reg reg; + + /* if we need to allocate another temp, grow the tmp_regs[] array */ + if (c->tmp_index == c->tmp_max) { + int r = alloc_grf(c); + if (r < 0) { + /*printf("Out of temps in %s\n", __FUNCTION__);*/ + r = 50; /* XXX random register! */ + } + c->tmp_regs[ c->tmp_max++ ] = r; + } + + /* form the GRF register */ + reg = brw_vec8_grf(c->tmp_regs[ c->tmp_index++ ], 0); + /*printf("alloc_temp %d\n", reg.nr);*/ + assert(reg.nr < BRW_WM_MAX_GRF); + return reg; + +} + +/** + * Save current temp register info. + * There must be a matching call to release_tmps(). + */ +static int mark_tmps(struct brw_wm_compile *c) +{ + return c->tmp_index; +} + +static struct brw_reg lookup_tmp( struct brw_wm_compile *c, int index ) +{ + return brw_vec8_grf( c->tmp_regs[ index ], 0 ); +} + +static void release_tmps(struct brw_wm_compile *c, int mark) +{ + c->tmp_index = mark; +} + +/** + * Convert Mesa src register to brw register. + * + * Since we're running in SOA mode each Mesa register corresponds to four + * hardware registers. We allocate the hardware registers as needed here. + * + * \param file register file, one of PROGRAM_x + * \param index register number + * \param component src component (X=0, Y=1, Z=2, W=3) + * \param nr not used?!? + * \param neg negate value? + * \param abs take absolute value? + */ +static struct brw_reg +get_reg(struct brw_wm_compile *c, int file, int index, int component, + int nr, GLuint neg, GLuint abs) +{ + struct brw_reg reg; + switch (file) { + case TGSI_FILE_NULL: + return brw_null_reg(); + + case TGSI_FILE_CONSTANT: + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + case BRW_FILE_PAYLOAD: + break; + + default: + debug_printf("%s: Unexpected file type\n", __FUNCTION__); + return brw_null_reg(); + } + + assert(index < 256); + assert(component < 4); + + /* see if we've already allocated a HW register for this Mesa register */ + if (c->wm_regs[file][index][component].inited) { + /* yes, re-use */ + reg = c->wm_regs[file][index][component].reg; + } + else { + /* no, allocate new register */ + int grf = alloc_grf(c); + /*printf("alloc grf %d for reg %d:%d.%d\n", grf, file, index, component);*/ + if (grf < 0) { + /* totally out of temps */ + grf = 51; /* XXX random register! */ + } + + reg = brw_vec8_grf(grf, 0); + /*printf("Alloc new grf %d for %d.%d\n", reg.nr, index, component);*/ + + set_reg(c, file, index, component, reg); + } + + if (neg & (1 << component)) { + reg = negate(reg); + } + if (abs) + reg = brw_abs(reg); + return reg; +} + + + + +/** + * Find first/last instruction that references each temporary register. + */ +GLboolean +_mesa_find_temp_intervals(const struct prog_instruction *instructions, + GLuint numInstructions, + GLint intBegin[MAX_PROGRAM_TEMPS], + GLint intEnd[MAX_PROGRAM_TEMPS]) +{ + struct loop_info + { + GLuint Start, End; /**< Start, end instructions of loop */ + }; + struct loop_info loopStack[MAX_LOOP_NESTING]; + GLuint loopStackDepth = 0; + GLuint i; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++){ + intBegin[i] = intEnd[i] = -1; + } + + /* Scan instructions looking for temporary registers */ + for (i = 0; i < numInstructions; i++) { + const struct prog_instruction *inst = instructions + i; + if (inst->Opcode == OPCODE_BGNLOOP) { + loopStack[loopStackDepth].Start = i; + loopStack[loopStackDepth].End = inst->BranchTarget; + loopStackDepth++; + } + else if (inst->Opcode == OPCODE_ENDLOOP) { + loopStackDepth--; + } + else if (inst->Opcode == OPCODE_CAL) { + return GL_FALSE; + } + else { + const GLuint numSrc = 3; + GLuint j; + for (j = 0; j < numSrc; j++) { + if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) { + const GLuint index = inst->SrcReg[j].Index; + if (inst->SrcReg[j].RelAddr) + return GL_FALSE; + update_interval(intBegin, intEnd, index, i); + if (loopStackDepth > 0) { + /* extend temp register's interval to end of loop */ + GLuint loopEnd = loopStack[loopStackDepth - 1].End; + update_interval(intBegin, intEnd, index, loopEnd); + } + } + } + if (inst->DstReg.File == PROGRAM_TEMPORARY) { + const GLuint index = inst->DstReg.Index; + if (inst->DstReg.RelAddr) + return GL_FALSE; + update_interval(intBegin, intEnd, index, i); + if (loopStackDepth > 0) { + /* extend temp register's interval to end of loop */ + GLuint loopEnd = loopStack[loopStackDepth - 1].End; + update_interval(intBegin, intEnd, index, loopEnd); + } + } + } + } + + return GL_TRUE; +} + + +/** + * This is called if we run out of GRF registers. Examine the live intervals + * of temp regs in the program and free those which won't be used again. + */ +static void +reclaim_temps(struct brw_wm_compile *c) +{ + GLint intBegin[BRW_WM_MAX_TEMPS]; + GLint intEnd[BRW_WM_MAX_TEMPS]; + int index; + + /*printf("Reclaim temps:\n");*/ + + _mesa_find_temp_intervals(c->fp_instructions, c->nr_fp_insns, + intBegin, intEnd); + + for (index = 0; index < BRW_WM_MAX_TEMPS; index++) { + if (intEnd[index] != -1 && intEnd[index] < c->cur_inst) { + /* program temp[i] can be freed */ + int component; + /*printf(" temp[%d] is dead\n", index);*/ + for (component = 0; component < 4; component++) { + if (c->wm_regs[TGSI_FILE_TEMPORARY][index][component].inited) { + int r = c->wm_regs[TGSI_FILE_TEMPORARY][index][component].reg.nr; + release_grf(c, r); + /* + printf(" Reclaim temp %d, reg %d at inst %d\n", + index, r, c->cur_inst); + */ + c->wm_regs[TGSI_FILE_TEMPORARY][index][component].inited = GL_FALSE; + } + } + } + } +} + + + + +/** + * Preallocate registers. This sets up the Mesa to hardware register + * mapping for certain registers, such as constants (uniforms/state vars) + * and shader inputs. + */ +static void prealloc_reg(struct brw_wm_compile *c) +{ + int i, j; + struct brw_reg reg; + int urb_read_length = 0; + GLuint inputs = FRAG_BIT_WPOS | c->fp_interp_emitted; + GLuint reg_index = 0; + + memset(c->used_grf, GL_FALSE, sizeof(c->used_grf)); + c->first_free_grf = 0; + + for (i = 0; i < 4; i++) { + if (i < c->key.nr_depth_regs) + reg = brw_vec8_grf(i * 2, 0); + else + reg = brw_vec8_grf(0, 0); + set_reg(c, TGSI_FILE_PAYLOAD, PAYLOAD_DEPTH, i, reg); + } + reg_index += 2 * c->key.nr_depth_regs; + + /* constants */ + { + const GLuint nr_params = c->fp->program.Base.Parameters->NumParameters; + const GLuint nr_temps = c->fp->program.Base.NumTemporaries; + + /* use a real constant buffer, or just use a section of the GRF? */ + /* XXX this heuristic may need adjustment... */ + if ((nr_params + nr_temps) * 4 + reg_index > 80) + c->fp->use_const_buffer = GL_TRUE; + else + c->fp->use_const_buffer = GL_FALSE; + /*printf("WM use_const_buffer = %d\n", c->fp->use_const_buffer);*/ + + if (c->fp->use_const_buffer) { + /* We'll use a real constant buffer and fetch constants from + * it with a dataport read message. + */ + + /* number of float constants in CURBE */ + c->prog_data.nr_params = 0; + } + else { + const struct gl_program_parameter_list *plist = + c->fp->program.Base.Parameters; + int index = 0; + + /* number of float constants in CURBE */ + c->prog_data.nr_params = 4 * nr_params; + + /* loop over program constants (float[4]) */ + for (i = 0; i < nr_params; i++) { + /* loop over XYZW channels */ + for (j = 0; j < 4; j++, index++) { + reg = brw_vec1_grf(reg_index + index / 8, index % 8); + /* Save pointer to parameter/constant value. + * Constants will be copied in prepare_constant_buffer() + */ + c->prog_data.param[index] = &plist->ParameterValues[i][j]; + set_reg(c, TGSI_FILE_STATE_VAR, i, j, reg); + } + } + /* number of constant regs used (each reg is float[8]) */ + c->nr_creg = 2 * ((4 * nr_params + 15) / 16); + reg_index += c->nr_creg; + } + } + + /* fragment shader inputs */ + for (i = 0; i < VERT_RESULT_MAX; i++) { + int fp_input; + + if (i >= VERT_RESULT_VAR0) + fp_input = i - VERT_RESULT_VAR0 + FRAG_ATTRIB_VAR0; + else if (i <= VERT_RESULT_TEX7) + fp_input = i; + else + fp_input = -1; + + if (fp_input >= 0 && inputs & (1 << fp_input)) { + urb_read_length = reg_index; + reg = brw_vec8_grf(reg_index, 0); + for (j = 0; j < 4; j++) + set_reg(c, TGSI_FILE_PAYLOAD, fp_input, j, reg); + } + if (c->key.nr_vp_outputs > i) { + reg_index += 2; + } + } + + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = urb_read_length; + c->prog_data.curb_read_length = c->nr_creg; + c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0); + reg_index++; + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0); + reg_index += 2; + + /* mark GRF regs [0..reg_index-1] as in-use */ + for (i = 0; i < reg_index; i++) + prealloc_grf(c, i); + + /* Don't use GRF 126, 127. Using them seems to lead to GPU lock-ups */ + prealloc_grf(c, 126); + prealloc_grf(c, 127); + + for (i = 0; i < c->nr_fp_insns; i++) { + const struct brw_fp_instruction *inst = &c->fp_instructions[i]; + struct brw_reg dst[4]; + + switch (inst->Opcode) { + case OPCODE_TEX: + case OPCODE_TXB: + /* Allocate the channels of texture results contiguously, + * since they are written out that way by the sampler unit. + */ + for (j = 0; j < 4; j++) { + dst[j] = get_dst_reg(c, inst, j); + if (j != 0) + assert(dst[j].nr == dst[j - 1].nr + 1); + } + break; + default: + break; + } + } + + /* An instruction may reference up to three constants. + * They'll be found in these registers. + * XXX alloc these on demand! + */ + if (c->fp->use_const_buffer) { + for (i = 0; i < 3; i++) { + c->current_const[i].index = -1; + c->current_const[i].reg = brw_vec8_grf(alloc_grf(c), 0); + } + } +#if 0 + printf("USE CONST BUFFER? %d\n", c->fp->use_const_buffer); + printf("AFTER PRE_ALLOC, reg_index = %d\n", reg_index); +#endif +} + + +/** + * Check if any of the instruction's src registers are constants, uniforms, + * or statevars. If so, fetch any constants that we don't already have in + * the three GRF slots. + */ +static void fetch_constants(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint i; + + /* loop over instruction src regs */ + for (i = 0; i < 3; i++) { + const struct prog_src_register *src = &inst->SrcReg[i]; + if (src->File == TGSI_FILE_IMMEDIATE || + src->File == TGSI_FILE_CONSTANT) { + c->current_const[i].index = src->Index; + +#if 0 + printf(" fetch const[%d] for arg %d into reg %d\n", + src->Index, i, c->current_const[i].reg.nr); +#endif + + /* need to fetch the constant now */ + brw_dp_READ_4(p, + c->current_const[i].reg, /* writeback dest */ + src->RelAddr, /* relative indexing? */ + 16 * src->Index, /* byte offset */ + SURF_INDEX_FRAG_CONST_BUFFER/* binding table index */ + ); + } + } +} + + +/** + * Convert Mesa dst register to brw register. + */ +static struct brw_reg get_dst_reg(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + GLuint component) +{ + const int nr = 1; + return get_reg(c, inst->DstReg.File, inst->DstReg.Index, component, nr, + 0, 0); +} + + +static struct brw_reg +get_src_reg_const(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + GLuint srcRegIndex, GLuint component) +{ + /* We should have already fetched the constant from the constant + * buffer in fetch_constants(). Now we just have to return a + * register description that extracts the needed component and + * smears it across all eight vector components. + */ + const struct prog_src_register *src = &inst->SrcReg[srcRegIndex]; + struct brw_reg const_reg; + + assert(component < 4); + assert(srcRegIndex < 3); + assert(c->current_const[srcRegIndex].index != -1); + const_reg = c->current_const[srcRegIndex].reg; + + /* extract desired float from the const_reg, and smear */ + const_reg = stride(const_reg, 0, 1, 0); + const_reg.subnr = component * 4; + + if (src->Negate) + const_reg = negate(const_reg); + if (src->Abs) + const_reg = brw_abs(const_reg); + +#if 0 + printf(" form const[%d].%d for arg %d, reg %d\n", + c->current_const[srcRegIndex].index, + component, + srcRegIndex, + const_reg.nr); +#endif + + return const_reg; +} + + +/** + * Convert Mesa src register to brw register. + */ +static struct brw_reg get_src_reg(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + GLuint srcRegIndex, GLuint channel) +{ + const struct prog_src_register *src = &inst->SrcReg[srcRegIndex]; + const GLuint nr = 1; + const GLuint component = BRW_GET_SWZ(src->Swizzle, channel); + + /* Extended swizzle terms */ + if (component == SWIZZLE_ZERO) { + return brw_imm_f(0.0F); + } + else if (component == SWIZZLE_ONE) { + return brw_imm_f(1.0F); + } + + if (c->fp->use_const_buffer && + (src->File == TGSI_FILE_STATE_VAR || + src->File == TGSI_FILE_CONSTANT || + src->File == TGSI_FILE_UNIFORM)) { + return get_src_reg_const(c, inst, srcRegIndex, component); + } + else { + /* other type of source register */ + return get_reg(c, src->File, src->Index, component, nr, + src->Negate, src->Abs); + } +} + + +/** + * Same as \sa get_src_reg() but if the register is a immediate, emit + * a brw_reg encoding the immediate. + * Note that a brw instruction only allows one src operand to be a immediate. + * For instructions with more than one operand, only the second can be a + * immediate. This means that we treat some immediates as constants + * (which why TGSI_FILE_IMMEDIATE is checked in fetch_constants()). + * + */ +static struct brw_reg get_src_reg_imm(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + GLuint srcRegIndex, GLuint channel) +{ + const struct prog_src_register *src = &inst->SrcReg[srcRegIndex]; + if (src->File == TGSI_FILE_IMMEDIATE) { + /* an immediate */ + const int component = BRW_GET_SWZ(src->Swizzle, channel); + const GLfloat *param = + c->fp->program.Base.Parameters->ParameterValues[src->Index]; + GLfloat value = param[component]; + if (src->Negate) + value = -value; + if (src->Abs) + value = FABSF(value); +#if 0 + printf(" form immed value %f for chan %d\n", value, channel); +#endif + return brw_imm_f(value); + } + else { + return get_src_reg(c, inst, srcRegIndex, channel); + } +} + + +/** + * Subroutines are minimal support for resusable instruction sequences. + * They are implemented as simply as possible to minimise overhead: there + * is no explicit support for communication between the caller and callee + * other than saving the return address in a temporary register, nor is + * there any automatic local storage. This implies that great care is + * required before attempting reentrancy or any kind of nested + * subroutine invocations. + */ +static void invoke_subroutine( struct brw_wm_compile *c, + enum _subroutine subroutine, + void (*emit)( struct brw_wm_compile * ) ) +{ + struct brw_compile *p = &c->func; + + assert( subroutine < BRW_WM_MAX_SUBROUTINE ); + + if( c->subroutines[ subroutine ] ) { + /* subroutine previously emitted: reuse existing instructions */ + + int mark = mark_tmps( c ); + struct brw_reg return_address = retype( alloc_tmp( c ), + BRW_REGISTER_TYPE_UD ); + int here = p->nr_insn; + + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_ADD( p, return_address, brw_ip_reg(), brw_imm_ud( 2 << 4 ) ); + + brw_ADD( p, brw_ip_reg(), brw_ip_reg(), + brw_imm_d( ( c->subroutines[ subroutine ] - + here - 1 ) << 4 ) ); + brw_pop_insn_state(p); + + release_tmps( c, mark ); + } else { + /* previously unused subroutine: emit, and mark for later reuse */ + + int mark = mark_tmps( c ); + struct brw_reg return_address = retype( alloc_tmp( c ), + BRW_REGISTER_TYPE_UD ); + struct brw_instruction *calc; + int base = p->nr_insn; + + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + calc = brw_ADD( p, return_address, brw_ip_reg(), brw_imm_ud( 0 ) ); + brw_pop_insn_state(p); + + c->subroutines[ subroutine ] = p->nr_insn; + + emit( c ); + + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_MOV( p, brw_ip_reg(), return_address ); + brw_pop_insn_state(p); + + brw_set_src1( calc, brw_imm_ud( ( p->nr_insn - base ) << 4 ) ); + + release_tmps( c, mark ); + } +} + +static void emit_trunc( struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + struct brw_reg src, dst; + dst = get_dst_reg(c, inst, i); + src = get_src_reg(c, inst, 0, i); + brw_RNDZ(p, dst, src); + } + } + brw_set_saturate(p, 0); +} + +static void emit_mov( struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + struct brw_reg src, dst; + dst = get_dst_reg(c, inst, i); + /* XXX some moves from immediate value don't work reliably!!! */ + /*src = get_src_reg_imm(c, inst, 0, i);*/ + src = get_src_reg(c, inst, 0, i); + brw_MOV(p, dst, src); + } + } + brw_set_saturate(p, 0); +} + +static void emit_pixel_xy(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); + + struct brw_reg dst0, dst1; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + + dst0 = get_dst_reg(c, inst, 0); + dst1 = get_dst_reg(c, inst, 1); + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + if (mask & WRITEMASK_X) { + brw_ADD(p, + vec8(retype(dst0, BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + } + + if (mask & WRITEMASK_Y) { + brw_ADD(p, + vec8(retype(dst1, BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 5), 2, 4, 0), + brw_imm_v(0x11001100)); + } +} + +static void emit_delta_xy(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg dst0, dst1, src0, src1; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + + dst0 = get_dst_reg(c, inst, 0); + dst1 = get_dst_reg(c, inst, 1); + src0 = get_src_reg(c, inst, 0, 0); + src1 = get_src_reg(c, inst, 0, 1); + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & WRITEMASK_X) { + brw_ADD(p, + dst0, + retype(src0, BRW_REGISTER_TYPE_UW), + negate(r1)); + } + + if (mask & WRITEMASK_Y) { + brw_ADD(p, + dst1, + retype(src1, BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + + } +} + +static void fire_fb_write( struct brw_wm_compile *c, + GLuint base_reg, + GLuint nr, + GLuint target, + GLuint eot) +{ + struct brw_compile *p = &c->func; + /* Pass through control information: + */ + /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + /* Send framebuffer write message: */ + brw_fb_WRITE(p, + retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + target, + nr, + 0, + eot); +} + +static void emit_fb_write(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + int nr = 2; + int channel; + GLuint target, eot; + struct brw_reg src0; + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + + brw_push_insn_state(p); + for (channel = 0; channel < 4; channel++) { + src0 = get_src_reg(c, inst, 0, channel); + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + brw_MOV(p, brw_message_reg(nr + channel), src0); + } + /* skip over the regs populated above: */ + nr += 8; + brw_pop_insn_state(p); + + if (c->key.source_depth_to_render_target) { + if (c->key.computes_depth) { + src0 = get_src_reg(c, inst, 2, 2); + brw_MOV(p, brw_message_reg(nr), src0); + } + else { + src0 = get_src_reg(c, inst, 1, 1); + brw_MOV(p, brw_message_reg(nr), src0); + } + + nr += 2; + } + + if (c->key.dest_depth_reg) { + const GLuint comp = c->key.dest_depth_reg / 2; + const GLuint off = c->key.dest_depth_reg % 2; + + if (off != 0) { + /* XXX this code needs review/testing */ + struct brw_reg arg1_0 = get_src_reg(c, inst, 1, comp); + struct brw_reg arg1_1 = get_src_reg(c, inst, 1, comp+1); + + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + brw_MOV(p, brw_message_reg(nr), offset(arg1_0, 1)); + /* 2nd half? */ + brw_MOV(p, brw_message_reg(nr+1), arg1_1); + brw_pop_insn_state(p); + } + else + { + struct brw_reg src = get_src_reg(c, inst, 1, 1); + brw_MOV(p, brw_message_reg(nr), src); + } + nr += 2; + } + + target = inst->Aux >> 1; + eot = inst->Aux & 1; + fire_fb_write(c, 0, nr, target, eot); +} + +static void emit_pixel_w( struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + if (mask & WRITEMASK_W) { + struct brw_reg dst, src0, delta0, delta1; + struct brw_reg interp3; + + dst = get_dst_reg(c, inst, 3); + src0 = get_src_reg(c, inst, 0, 0); + delta0 = get_src_reg(c, inst, 1, 0); + delta1 = get_src_reg(c, inst, 1, 1); + + interp3 = brw_vec1_grf(src0.nr+1, 4); + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + brw_LINE(p, brw_null_reg(), interp3, delta0); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), delta1); + + /* Calc w */ + brw_math_16( p, dst, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} + +static void emit_linterp(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst, delta0, delta1; + struct brw_reg src0; + GLuint nr, i; + + src0 = get_src_reg(c, inst, 0, 0); + delta0 = get_src_reg(c, inst, 1, 0); + delta1 = get_src_reg(c, inst, 1, 1); + nr = src0.nr; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + brw_LINE(p, brw_null_reg(), interp[i], delta0); + brw_MAC(p, dst, suboffset(interp[i],1), delta1); + } + } +} + +static void emit_cinterp(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + + struct brw_reg interp[4]; + struct brw_reg dst, src0; + GLuint nr, i; + + src0 = get_src_reg(c, inst, 0, 0); + nr = src0.nr; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + brw_MOV(p, dst, suboffset(interp[i],3)); + } + } +} + +static void emit_pinterp(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + + struct brw_reg interp[4]; + struct brw_reg dst, delta0, delta1; + struct brw_reg src0, w; + GLuint nr, i; + + src0 = get_src_reg(c, inst, 0, 0); + delta0 = get_src_reg(c, inst, 1, 0); + delta1 = get_src_reg(c, inst, 1, 1); + w = get_src_reg(c, inst, 2, 3); + nr = src0.nr; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + brw_LINE(p, brw_null_reg(), interp[i], delta0); + brw_MAC(p, dst, suboffset(interp[i],1), + delta1); + brw_MUL(p, dst, dst, w); + } + } +} + +/* Sets the destination channels to 1.0 or 0.0 according to glFrontFacing. */ +static void emit_frontfacing(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg r1_6ud = retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_UD); + struct brw_reg dst; + GLuint mask = inst->DstReg.WriteMask; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + brw_MOV(p, dst, brw_imm_f(0.0)); + } + } + + /* bit 31 is "primitive is back face", so checking < (1 << 31) gives + * us front face + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, r1_6ud, brw_imm_ud(1 << 31)); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + brw_MOV(p, dst, brw_imm_f(1.0)); + } + } + brw_set_predicate_control_flag_value(p, 0xff); +} + +static void emit_xpd(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + for (i = 0; i < 4; i++) { + GLuint i2 = (i+2)%3; + GLuint i1 = (i+1)%3; + if (mask & (1<<i)) { + struct brw_reg src0, src1, dst; + dst = get_dst_reg(c, inst, i); + src0 = negate(get_src_reg(c, inst, 0, i2)); + src1 = get_src_reg_imm(c, inst, 1, i1); + brw_MUL(p, brw_null_reg(), src0, src1); + src0 = get_src_reg(c, inst, 0, i1); + src1 = get_src_reg_imm(c, inst, 1, i2); + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + brw_MAC(p, dst, src0, src1); + brw_set_saturate(p, 0); + } + } + brw_set_saturate(p, 0); +} + +static void emit_dp3(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_reg src0[3], src1[3], dst; + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + int dst_chan = ffs(mask & WRITEMASK_XYZW) - 1; + + if (!(mask & WRITEMASK_XYZW)) + return; + + assert(is_power_of_two(mask & WRITEMASK_XYZW)); + + for (i = 0; i < 3; i++) { + src0[i] = get_src_reg(c, inst, 0, i); + src1[i] = get_src_reg_imm(c, inst, 1, i); + } + + dst = get_dst_reg(c, inst, dst_chan); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, 0); +} + +static void emit_dp4(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + int dst_chan = ffs(mask & WRITEMASK_XYZW) - 1; + + if (!(mask & WRITEMASK_XYZW)) + return; + + assert(is_power_of_two(mask & WRITEMASK_XYZW)); + + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, inst, 0, i); + src1[i] = get_src_reg_imm(c, inst, 1, i); + } + dst = get_dst_reg(c, inst, dst_chan); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, brw_null_reg(), src0[2], src1[2]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); +} + +static void emit_dph(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + int dst_chan = ffs(mask & WRITEMASK_XYZW) - 1; + + if (!(mask & WRITEMASK_XYZW)) + return; + + assert(is_power_of_two(mask & WRITEMASK_XYZW)); + + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, inst, 0, i); + src1[i] = get_src_reg_imm(c, inst, 1, i); + } + dst = get_dst_reg(c, inst, dst_chan); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_ADD(p, dst, dst, src1[3]); + brw_set_saturate(p, 0); +} + +/** + * Emit a scalar instruction, like RCP, RSQ, LOG, EXP. + * Note that the result of the function is smeared across the dest + * register's X, Y, Z and W channels (subject to writemasking of course). + */ +static void emit_math1(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, GLuint func) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + GLuint mask = inst->DstReg.WriteMask; + int dst_chan = ffs(mask & WRITEMASK_XYZW) - 1; + + if (!(mask & WRITEMASK_XYZW)) + return; + + assert(is_power_of_two(mask & WRITEMASK_XYZW)); + + /* Get first component of source register */ + dst = get_dst_reg(c, inst, dst_chan); + src0 = get_src_reg(c, inst, 0, 0); + + brw_MOV(p, brw_message_reg(2), src0); + brw_math(p, + dst, + func, + (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + +static void emit_rcp(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_INV); +} + +static void emit_rsq(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); +} + +static void emit_sin(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); +} + +static void emit_cos(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_COS); +} + +static void emit_ex2(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); +} + +static void emit_lg2(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); +} + +static void emit_add(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + GLuint mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg(c, inst, 0, i); + src1 = get_src_reg_imm(c, inst, 1, i); + brw_ADD(p, dst, src0, src1); + } + } + brw_set_saturate(p, 0); +} + +static void emit_arl(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, addr_reg; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + addr_reg = brw_uw8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ADDRESS, 0); + src0 = get_src_reg(c, inst, 0, 0); /* channel 0 */ + brw_MOV(p, addr_reg, src0); + brw_set_saturate(p, 0); +} + + +static void emit_mul(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + GLuint mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg(c, inst, 0, i); + src1 = get_src_reg_imm(c, inst, 1, i); + brw_MUL(p, dst, src0, src1); + } + } + brw_set_saturate(p, 0); +} + +static void emit_frc(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + GLuint mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg_imm(c, inst, 0, i); + brw_FRC(p, dst, src0); + } + } + if (inst->SaturateMode != SATURATE_OFF) + brw_set_saturate(p, 0); +} + +static void emit_flr(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + GLuint mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg_imm(c, inst, 0, i); + brw_RNDD(p, dst, src0); + } + } + brw_set_saturate(p, 0); +} + + +static void emit_min_max(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + const GLuint mask = inst->DstReg.WriteMask; + const int mark = mark_tmps(c); + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + struct brw_reg real_dst = get_dst_reg(c, inst, i); + struct brw_reg src0 = get_src_reg(c, inst, 0, i); + struct brw_reg src1 = get_src_reg(c, inst, 1, i); + struct brw_reg dst; + /* if dst==src0 or dst==src1 we need to use a temp reg */ + GLboolean use_temp = brw_same_reg(dst, src0) || + brw_same_reg(dst, src1); + if (use_temp) + dst = alloc_tmp(c); + else + dst = real_dst; + + /* + printf(" Min/max: dst %d src0 %d src1 %d\n", + dst.nr, src0.nr, src1.nr); + */ + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + if (inst->Opcode == OPCODE_MIN) + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); + else + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, src1, src0); + + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + if (use_temp) + brw_MOV(p, real_dst, dst); + } + } + brw_pop_insn_state(p); + release_tmps(c, mark); +} + +static void emit_pow(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst, src0, src1; + GLuint mask = inst->DstReg.WriteMask; + int dst_chan = ffs(mask & WRITEMASK_XYZW) - 1; + + if (!(mask & WRITEMASK_XYZW)) + return; + + assert(is_power_of_two(mask & WRITEMASK_XYZW)); + + dst = get_dst_reg(c, inst, dst_chan); + src0 = get_src_reg_imm(c, inst, 0, 0); + src1 = get_src_reg_imm(c, inst, 1, 0); + + brw_MOV(p, brw_message_reg(2), src0); + brw_MOV(p, brw_message_reg(3), src1); + + brw_math(p, + dst, + BRW_MATH_FUNCTION_POW, + (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + +static void emit_lrp(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + struct brw_reg dst, tmp1, tmp2, src0, src1, src2; + int i; + int mark = mark_tmps(c); + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg(c, inst, 0, i); + + src1 = get_src_reg_imm(c, inst, 1, i); + + if (src1.nr == dst.nr) { + tmp1 = alloc_tmp(c); + brw_MOV(p, tmp1, src1); + } else + tmp1 = src1; + + src2 = get_src_reg(c, inst, 2, i); + if (src2.nr == dst.nr) { + tmp2 = alloc_tmp(c); + brw_MOV(p, tmp2, src2); + } else + tmp2 = src2; + + brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst, tmp2); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0, tmp1); + brw_set_saturate(p, 0); + } + release_tmps(c, mark); + } +} + +/** + * For GLSL shaders, this KIL will be unconditional. + * It may be contained inside an IF/ENDIF structure of course. + */ +static void emit_kil(struct brw_wm_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK + brw_AND(p, depth, c->emit_mask_reg, depth); + brw_pop_insn_state(p); +} + +static void emit_mad(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + struct brw_reg dst, src0, src1, src2; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg(c, inst, 0, i); + src1 = get_src_reg_imm(c, inst, 1, i); + src2 = get_src_reg_imm(c, inst, 2, i); + brw_MUL(p, dst, src0, src1); + + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_ADD(p, dst, dst, src2); + brw_set_saturate(p, 0); + } + } +} + +static void emit_sop(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, GLuint cond) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + struct brw_reg dst, src0, src1; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<<i)) { + dst = get_dst_reg(c, inst, i); + src0 = get_src_reg(c, inst, 0, i); + src1 = get_src_reg_imm(c, inst, 1, i); + brw_push_insn_state(p); + brw_CMP(p, brw_null_reg(), cond, src0, src1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(0.0)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(1.0)); + brw_pop_insn_state(p); + } + } +} + +static void emit_slt(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_L); +} + +static void emit_sle(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_LE); +} + +static void emit_sgt(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_G); +} + +static void emit_sge(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_GE); +} + +static void emit_seq(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_EQ); +} + +static void emit_sne(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_NEQ); +} + +static INLINE struct brw_reg high_words( struct brw_reg reg ) +{ + return stride( suboffset( retype( reg, BRW_REGISTER_TYPE_W ), 1 ), + 0, 8, 2 ); +} + +static INLINE struct brw_reg low_words( struct brw_reg reg ) +{ + return stride( retype( reg, BRW_REGISTER_TYPE_W ), 0, 8, 2 ); +} + +static INLINE struct brw_reg even_bytes( struct brw_reg reg ) +{ + return stride( retype( reg, BRW_REGISTER_TYPE_B ), 0, 16, 2 ); +} + +static INLINE struct brw_reg odd_bytes( struct brw_reg reg ) +{ + return stride( suboffset( retype( reg, BRW_REGISTER_TYPE_B ), 1 ), + 0, 16, 2 ); +} + + + +static void emit_wpos_xy(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + GLuint mask = inst->DstReg.WriteMask; + struct brw_reg src0[2], dst[2]; + + dst[0] = get_dst_reg(c, inst, 0); + dst[1] = get_dst_reg(c, inst, 1); + + src0[0] = get_src_reg(c, inst, 0, 0); + src0[1] = get_src_reg(c, inst, 0, 1); + + /* Calculate the pixel offset from window bottom left into destination + * X and Y channels. + */ + if (mask & WRITEMASK_X) { + /* X' = X */ + brw_MOV(p, + dst[0], + retype(src0[0], BRW_REGISTER_TYPE_W)); + } + + if (mask & WRITEMASK_Y) { + /* Y' = height - 1 - Y */ + brw_ADD(p, + dst[1], + negate(retype(src0[1], BRW_REGISTER_TYPE_W)), + brw_imm_d(c->key.drawable_height - 1)); + } +} + +/* TODO + BIAS on SIMD8 not working yet... + */ +static void emit_txb(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst[4], src[4], payload_reg; + /* Note: tex_unit was already looked up through SamplerTextures[] */ + const GLuint unit = inst->tex_unit; + GLuint i; + GLuint msg_type; + + assert(unit < BRW_MAX_TEX_UNIT); + + payload_reg = get_reg(c, TGSI_FILE_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); + + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, inst, 0, i); + + switch (inst->tex_target) { + case TEXTURE_1D_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); /* s coord */ + brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); /* t coord */ + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); /* r coord */ + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + case TEXTURE_3D_INDEX: + case TEXTURE_CUBE_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), src[2]); + break; + default: + /* invalid target */ + abort(); + } + brw_MOV(p, brw_message_reg(5), src[3]); /* bias */ + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); /* ref (unused?) */ + + if (BRW_IS_IGDNG(p->brw)) { + msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG; + } else { + /* Does it work well on SIMD8? */ + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; + } + + brw_SAMPLE(p, + retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ + 1, /* msg_reg_nr */ + retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ + SURF_INDEX_TEXTURE(unit), + unit, /* sampler */ + inst->DstReg.WriteMask, /* writemask */ + msg_type, /* msg_type */ + 4, /* response_length */ + 4, /* msg_length */ + 0, /* eot */ + 1, + BRW_SAMPLER_SIMD_MODE_SIMD8); +} + + +static void emit_tex(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst[4], src[4], payload_reg; + /* Note: tex_unit was already looked up through SamplerTextures[] */ + const GLuint unit = inst->tex_unit; + GLuint msg_len; + GLuint i, nr; + GLuint emit; + GLboolean shadow = (c->key.shadowtex_mask & (1<<unit)) ? 1 : 0; + GLuint msg_type; + + assert(unit < BRW_MAX_TEX_UNIT); + + payload_reg = get_reg(c, TGSI_FILE_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); + + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, inst, 0, i); + + switch (inst->tex_target) { + case TEXTURE_1D_INDEX: + emit = WRITEMASK_X; + nr = 1; + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + emit = WRITEMASK_XY; + nr = 2; + break; + case TEXTURE_3D_INDEX: + case TEXTURE_CUBE_INDEX: + emit = WRITEMASK_XYZ; + nr = 3; + break; + default: + /* invalid target */ + abort(); + } + msg_len = 1; + + /* move/load S, T, R coords */ + for (i = 0; i < nr; i++) { + static const GLuint swz[4] = {0,1,2,2}; + if (emit & (1<<i)) + brw_MOV(p, brw_message_reg(msg_len+1), src[swz[i]]); + else + brw_MOV(p, brw_message_reg(msg_len+1), brw_imm_f(0)); + msg_len += 1; + } + + if (shadow) { + brw_MOV(p, brw_message_reg(5), brw_imm_f(0)); /* lod / bias */ + brw_MOV(p, brw_message_reg(6), src[2]); /* ref value / R coord */ + } + + if (BRW_IS_IGDNG(p->brw)) { + if (shadow) + msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG; + } else { + /* Does it work for shadow on SIMD8 ? */ + msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE; + } + + brw_SAMPLE(p, + retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ + 1, /* msg_reg_nr */ + retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ + SURF_INDEX_TEXTURE(unit), + unit, /* sampler */ + inst->DstReg.WriteMask, /* writemask */ + msg_type, /* msg_type */ + 4, /* response_length */ + shadow ? 6 : 4, /* msg_length */ + 0, /* eot */ + 1, + BRW_SAMPLER_SIMD_MODE_SIMD8); + + if (shadow) + brw_MOV(p, dst[3], brw_imm_f(1.0)); +} + + +/** + * Resolve subroutine calls after code emit is done. + */ +static void post_wm_emit( struct brw_wm_compile *c ) +{ + brw_resolve_cals(&c->func); +} + +static void +get_argument_regs(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst, + int index, + struct brw_reg *regs, + int mask) +{ + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1 << i)) + regs[i] = get_src_reg(c, inst, index, i); + } +} + +static void brw_wm_emit_branching_shader(struct brw_context *brw, struct brw_wm_compile *c) +{ +#define MAX_IF_DEPTH 32 +#define MAX_LOOP_DEPTH 32 + struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH]; + GLuint i, if_depth = 0, loop_depth = 0; + struct brw_compile *p = &c->func; + struct brw_indirect stack_index = brw_indirect(0, 0); + + c->out_of_regs = GL_FALSE; + + prealloc_reg(c); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); + + for (i = 0; i < c->nr_fp_insns; i++) { + const struct brw_fp_instruction *inst = &c->fp_instructions[i]; + int dst_flags; + struct brw_reg args[3][4], dst[4]; + int j; + + c->cur_inst = i; + +#if 0 + debug_printf("Inst %d: ", i); + _mesa_print_instruction(inst); +#endif + + /* fetch any constants that this instruction needs */ + if (c->fp->use_const_buffer) + fetch_constants(c, inst); + + if (inst->CondUpdate) + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); + + dst_flags = inst->DstReg.WriteMask; + if (inst->SaturateMode == SATURATE_ZERO_ONE) + dst_flags |= SATURATE; + + switch (inst->Opcode) { + case WM_PIXELXY: + emit_pixel_xy(c, inst); + break; + case WM_DELTAXY: + emit_delta_xy(c, inst); + break; + case WM_PIXELW: + emit_pixel_w(c, inst); + break; + case WM_LINTERP: + emit_linterp(c, inst); + break; + case WM_PINTERP: + emit_pinterp(c, inst); + break; + case WM_CINTERP: + emit_cinterp(c, inst); + break; + case WM_WPOSXY: + emit_wpos_xy(c, inst); + break; + case WM_FB_WRITE: + emit_fb_write(c, inst); + break; + case WM_FRONTFACING: + emit_frontfacing(c, inst); + break; + case OPCODE_ADD: + emit_add(c, inst); + break; + case OPCODE_ARL: + emit_arl(c, inst); + break; + case OPCODE_FRC: + emit_frc(c, inst); + break; + case OPCODE_FLR: + emit_flr(c, inst); + break; + case OPCODE_LRP: + emit_lrp(c, inst); + break; + case OPCODE_TRUNC: + emit_trunc(c, inst); + break; + case OPCODE_MOV: + emit_mov(c, inst); + break; + case OPCODE_DP3: + emit_dp3(c, inst); + break; + case OPCODE_DP4: + emit_dp4(c, inst); + break; + case OPCODE_XPD: + emit_xpd(c, inst); + break; + case OPCODE_DPH: + emit_dph(c, inst); + break; + case OPCODE_RCP: + emit_rcp(c, inst); + break; + case OPCODE_RSQ: + emit_rsq(c, inst); + break; + case OPCODE_SIN: + emit_sin(c, inst); + break; + case OPCODE_COS: + emit_cos(c, inst); + break; + case OPCODE_EX2: + emit_ex2(c, inst); + break; + case OPCODE_LG2: + emit_lg2(c, inst); + break; + case OPCODE_MIN: + case OPCODE_MAX: + emit_min_max(c, inst); + break; + case OPCODE_DDX: + case OPCODE_DDY: + for (j = 0; j < 4; j++) { + if (inst->DstReg.WriteMask & (1 << j)) + dst[j] = get_dst_reg(c, inst, j); + else + dst[j] = brw_null_reg(); + } + get_argument_regs(c, inst, 0, args[0], WRITEMASK_XYZW); + emit_ddxy(p, dst, dst_flags, (inst->Opcode == OPCODE_DDX), + args[0]); + break; + case OPCODE_SLT: + emit_slt(c, inst); + break; + case OPCODE_SLE: + emit_sle(c, inst); + break; + case OPCODE_SGT: + emit_sgt(c, inst); + break; + case OPCODE_SGE: + emit_sge(c, inst); + break; + case OPCODE_SEQ: + emit_seq(c, inst); + break; + case OPCODE_SNE: + emit_sne(c, inst); + break; + case OPCODE_MUL: + emit_mul(c, inst); + break; + case OPCODE_POW: + emit_pow(c, inst); + break; + case OPCODE_MAD: + emit_mad(c, inst); + break; + case OPCODE_TEX: + emit_tex(c, inst); + break; + case OPCODE_TXB: + emit_txb(c, inst); + break; + case OPCODE_KIL_NV: + emit_kil(c); + break; + case OPCODE_IF: + assert(if_depth < MAX_IF_DEPTH); + if_inst[if_depth++] = brw_IF(p, BRW_EXECUTE_8); + break; + case OPCODE_ELSE: + if_inst[if_depth-1] = brw_ELSE(p, if_inst[if_depth-1]); + break; + case OPCODE_ENDIF: + assert(if_depth > 0); + brw_ENDIF(p, if_inst[--if_depth]); + break; + case OPCODE_BGNSUB: + brw_save_label(p, inst->Comment, p->nr_insn); + break; + case OPCODE_ENDSUB: + /* no-op */ + break; + case OPCODE_CAL: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1ud(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(4)); + brw_save_call(&c->func, inst->label, p->nr_insn); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_pop_insn_state(p); + break; + + case OPCODE_RET: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1ud(stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_pop_insn_state(p); + + break; + case OPCODE_BGNLOOP: + /* XXX may need to invalidate the current_constant regs */ + loop_inst[loop_depth++] = brw_DO(p, BRW_EXECUTE_8); + break; + case OPCODE_BRK: + brw_BREAK(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case OPCODE_CONT: + brw_CONT(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case OPCODE_ENDLOOP: + { + struct brw_instruction *inst0, *inst1; + GLuint br = 1; + + if (BRW_IS_IGDNG(brw)) + br = 2; + + loop_depth--; + inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]); + /* patch all the BREAK/CONT instructions from last BGNLOOP */ + while (inst0 > loop_inst[loop_depth]) { + inst0--; + if (inst0->header.opcode == BRW_OPCODE_BREAK) { + inst0->bits3.if_else.jump_count = br * (inst1 - inst0 + 1); + inst0->bits3.if_else.pop_count = 0; + } + else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) { + inst0->bits3.if_else.jump_count = br * (inst1 - inst0); + inst0->bits3.if_else.pop_count = 0; + } + } + } + break; + default: + debug_printf("unsupported IR in fragment shader %d\n", + inst->Opcode); + } + + if (inst->CondUpdate) + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + else + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + post_wm_emit(c); + + if (BRW_DEBUG & DEBUG_WM) { + debug_printf("wm-native:\n"); + brw_disasm(stderr, p->store, p->nr_insn); + } +} + +/** + * Do GPU code generation for shaders that use GLSL features such as + * flow control. Other shaders will be compiled with the + */ +void brw_wm_branching_shader_emit(struct brw_context *brw, struct brw_wm_compile *c) +{ + if (BRW_DEBUG & DEBUG_WM) { + debug_printf("%s:\n", __FUNCTION__); + } + + /* initial instruction translation/simplification */ + brw_wm_pass_fp(c); + + /* actual code generation */ + brw_wm_emit_branching_shader(brw, c); + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "brw_wm_branching_shader_emit done"); + } + + c->prog_data.total_grf = num_grf_used(c); + c->prog_data.total_scratch = 0; +} diff --git a/src/gallium/drivers/i965/brw_wm_iz.c b/src/gallium/drivers/i965/brw_wm_iz.c new file mode 100644 index 00000000000..6f1e9fcc3c1 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_iz.c @@ -0,0 +1,156 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_wm.h" + + +#undef P /* prompted depth */ +#undef C /* computed */ +#undef N /* non-promoted? */ + +#define P 0 +#define C 1 +#define N 2 + +const struct { + GLuint mode:2; + GLuint sd_present:1; + GLuint sd_to_rt:1; + GLuint dd_present:1; + GLuint ds_present:1; +} wm_iz_table[IZ_BIT_MAX] = +{ + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 0, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 } +}; + +/** + * \param line_aa AA_NEVER, AA_ALWAYS or AA_SOMETIMES + * \param lookup bitmask of IZ_* flags + */ +void brw_wm_lookup_iz( GLuint line_aa, + GLuint lookup, + GLboolean ps_uses_depth, + struct brw_wm_prog_key *key ) +{ + GLuint reg = 2; + + assert (lookup < IZ_BIT_MAX); + + if (lookup & IZ_PS_COMPUTES_DEPTH_BIT) + key->computes_depth = 1; + + if (wm_iz_table[lookup].sd_present || ps_uses_depth) { + key->source_depth_reg = reg; + reg += 2; + } + + if (wm_iz_table[lookup].sd_to_rt) + key->source_depth_to_render_target = 1; + + if (wm_iz_table[lookup].ds_present || line_aa != AA_NEVER) { + key->aa_dest_stencil_reg = reg; + key->runtime_check_aads_emit = (!wm_iz_table[lookup].ds_present && + line_aa == AA_SOMETIMES); + reg++; + } + + if (wm_iz_table[lookup].dd_present) { + key->dest_depth_reg = reg; + reg+=2; + } + + key->nr_depth_regs = (reg+1)/2; +} + diff --git a/src/gallium/drivers/i965/brw_wm_pass0.c b/src/gallium/drivers/i965/brw_wm_pass0.c new file mode 100644 index 00000000000..0bacad2b0f0 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_pass0.c @@ -0,0 +1,366 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "brw_debug.h" +#include "brw_wm.h" + + + +/*********************************************************************** + */ + +static struct brw_wm_ref *get_ref( struct brw_wm_compile *c ) +{ + assert(c->nr_refs < BRW_WM_MAX_REF); + return &c->refs[c->nr_refs++]; +} + +static struct brw_wm_value *get_value( struct brw_wm_compile *c) +{ + assert(c->nr_refs < BRW_WM_MAX_VREG); + return &c->vreg[c->nr_vreg++]; +} + +/** return pointer to a newly allocated instruction */ +static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) +{ + assert(c->nr_insns < BRW_WM_MAX_INSN); + return &c->instruction[c->nr_insns++]; +} + +/*********************************************************************** + */ + +/** Init the "undef" register */ +static void pass0_init_undef( struct brw_wm_compile *c) +{ + struct brw_wm_ref *ref = &c->undef_ref; + ref->value = &c->undef_value; + ref->hw_reg = brw_vec8_grf(0, 0); + ref->insn = 0; + ref->prevuse = NULL; +} + +/** Set a FP register to a value */ +static void pass0_set_fpreg_value( struct brw_wm_compile *c, + GLuint file, + GLuint idx, + GLuint component, + struct brw_wm_value *value ) +{ + struct brw_wm_ref *ref = get_ref(c); + ref->value = value; + ref->hw_reg = brw_vec8_grf(0, 0); + ref->insn = 0; + ref->prevuse = NULL; + c->pass0_fp_reg[file][idx][component] = ref; +} + +/** Set a FP register to a ref */ +static void pass0_set_fpreg_ref( struct brw_wm_compile *c, + GLuint file, + GLuint idx, + GLuint component, + const struct brw_wm_ref *src_ref ) +{ + c->pass0_fp_reg[file][idx][component] = src_ref; +} + +static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c, + unsigned idx, + unsigned component) +{ + GLuint i = idx * 4 + component; + + if (i >= BRW_WM_MAX_PARAM) { + debug_printf("%s: out of params\n", __FUNCTION__); + c->prog_data.error = 1; + return NULL; + } + else { + struct brw_wm_ref *ref = get_ref(c); + + c->nr_creg = MAX2(c->nr_creg, (i+16)/16); + + /* Push the offsets into hw_reg. These will be added to the + * real register numbers once one is allocated in pass2. + */ + ref->hw_reg = brw_vec1_grf((i&8)?1:0, i%8); + ref->value = &c->creg[i/16]; + ref->insn = 0; + ref->prevuse = NULL; + + return ref; + } +} + + + + +/* Lookup our internal registers + */ +static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, + GLuint file, + GLuint idx, + GLuint component ) +{ + const struct brw_wm_ref *ref = c->pass0_fp_reg[file][idx][component]; + + if (!ref) { + switch (file) { + case TGSI_FILE_INPUT: + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_OUTPUT: + case BRW_FILE_PAYLOAD: + /* should already be done?? */ + break; + + case TGSI_FILE_CONSTANT: + ref = get_param_ref(c, + c->fp->info.immediate_count + idx, + component); + break; + + case TGSI_FILE_IMMEDIATE: + ref = get_param_ref(c, + idx, + component); + break; + + default: + assert(0); + break; + } + + c->pass0_fp_reg[file][idx][component] = ref; + } + + if (!ref) + ref = &c->undef_ref; + + return ref; +} + + + +/*********************************************************************** + * Straight translation to internal instruction format + */ + +static void pass0_set_dst( struct brw_wm_compile *c, + struct brw_wm_instruction *out, + const struct brw_fp_instruction *inst, + GLuint writemask ) +{ + const struct brw_fp_dst dst = inst->dst; + GLuint i; + + for (i = 0; i < 4; i++) { + if (writemask & (1<<i)) { + out->dst[i] = get_value(c); + pass0_set_fpreg_value(c, dst.file, dst.index, i, out->dst[i]); + } + } + + out->writemask = writemask; +} + + +static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c, + struct brw_fp_src src, + GLuint i ) +{ + return pass0_get_reg(c, src.file, src.index, BRW_GET_SWZ(src.swizzle,i)); +} + + +static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, + struct brw_fp_src src, + GLuint i, + struct brw_wm_instruction *insn) +{ + const struct brw_wm_ref *ref = get_fp_src_reg_ref(c, src, i); + struct brw_wm_ref *newref = get_ref(c); + + newref->value = ref->value; + newref->hw_reg = ref->hw_reg; + + if (insn) { + newref->insn = insn - c->instruction; + newref->prevuse = newref->value->lastuse; + newref->value->lastuse = newref; + } + + if (src.negate) + newref->hw_reg.negate ^= 1; + + if (src.abs) { + newref->hw_reg.negate = 0; + newref->hw_reg.abs = 1; + } + + return newref; +} + + +static void +translate_insn(struct brw_wm_compile *c, + const struct brw_fp_instruction *inst) +{ + struct brw_wm_instruction *out = get_instruction(c); + GLuint writemask = inst->dst.writemask; + GLuint nr_args = brw_wm_nr_args(inst->opcode); + GLuint i, j; + + /* Copy some data out of the instruction + */ + out->opcode = inst->opcode; + out->saturate = inst->dst.saturate; + out->tex_unit = inst->tex_unit; + out->target = inst->target; + + /* Nasty hack: + */ + out->eot = (inst->opcode == WM_FB_WRITE && + inst->tex_unit != 0); + + + /* Args: + */ + for (i = 0; i < nr_args; i++) { + for (j = 0; j < 4; j++) { + out->src[i][j] = get_new_ref(c, inst->src[i], j, out); + } + } + + /* Dst: + */ + pass0_set_dst(c, out, inst, writemask); +} + + + +/*********************************************************************** + * Optimize moves and swizzles away: + */ +static void pass0_precalc_mov( struct brw_wm_compile *c, + const struct brw_fp_instruction *inst ) +{ + const struct brw_fp_dst dst = inst->dst; + GLuint writemask = dst.writemask; + struct brw_wm_ref *refs[4]; + GLuint i; + + /* Get the effect of a MOV by manipulating our register table: + * First get all refs, then assign refs. This ensures that "in-place" + * swizzles such as: + * MOV t, t.xxyx + * are handled correctly. Previously, these two steps were done in + * one loop and the above case was incorrectly handled. + */ + for (i = 0; i < 4; i++) { + refs[i] = get_new_ref(c, inst->src[0], i, NULL); + } + for (i = 0; i < 4; i++) { + if (writemask & (1 << i)) { + pass0_set_fpreg_ref( c, dst.file, dst.index, i, refs[i]); + } + } +} + + +/* Initialize payload "registers". + */ +static void pass0_init_payload( struct brw_wm_compile *c ) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + GLuint j = i >= c->key.nr_depth_regs ? 0 : i; + pass0_set_fpreg_value( c, BRW_FILE_PAYLOAD, PAYLOAD_DEPTH, i, + &c->payload.depth[j] ); + } + + for (i = 0; i < c->key.nr_inputs; i++) + pass0_set_fpreg_value( c, BRW_FILE_PAYLOAD, i, 0, + &c->payload.input_interp[i] ); +} + + +/*********************************************************************** + * PASS 0 + * + * Work forwards to give each calculated value a unique number. Where + * an instruction produces duplicate values (eg DP3), all are given + * the same number. + * + * Translate away swizzling and eliminate non-saturating moves. + * + * Translate instructions from our fp_instruction structs to our + * internal brw_wm_instruction representation. + */ +void brw_wm_pass0( struct brw_wm_compile *c ) +{ + GLuint insn; + + c->nr_vreg = 0; + c->nr_insns = 0; + + pass0_init_undef(c); + pass0_init_payload(c); + + for (insn = 0; insn < c->nr_fp_insns; insn++) { + const struct brw_fp_instruction *inst = &c->fp_instructions[insn]; + + /* Optimize away moves, otherwise emit translated instruction: + */ + switch (inst->opcode) { + case TGSI_OPCODE_MOV: + if (!inst->dst.saturate) { + pass0_precalc_mov(c, inst); + } + else { + translate_insn(c, inst); + } + break; + default: + translate_insn(c, inst); + break; + } + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass0"); + } +} diff --git a/src/gallium/drivers/i965/brw_wm_pass1.c b/src/gallium/drivers/i965/brw_wm_pass1.c new file mode 100644 index 00000000000..005747f00ba --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_pass1.c @@ -0,0 +1,292 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_wm.h" +#include "brw_debug.h" + + +static GLuint get_tracked_mask(struct brw_wm_compile *c, + struct brw_wm_instruction *inst) +{ + GLuint i; + for (i = 0; i < 4; i++) { + if (inst->writemask & (1<<i)) { + if (!inst->dst[i]->contributes_to_output) { + inst->writemask &= ~(1<<i); + inst->dst[i] = 0; + } + } + } + + return inst->writemask; +} + +/* Remove a reference from a value's usage chain. + */ +static void unlink_ref(struct brw_wm_ref *ref) +{ + struct brw_wm_value *value = ref->value; + + if (ref == value->lastuse) { + value->lastuse = ref->prevuse; + } + else { + struct brw_wm_ref *i = value->lastuse; + while (i->prevuse != ref) i = i->prevuse; + i->prevuse = ref->prevuse; + } +} + +static void track_arg(struct brw_wm_compile *c, + struct brw_wm_instruction *inst, + GLuint arg, + GLuint readmask) +{ + GLuint i; + + for (i = 0; i < 4; i++) { + struct brw_wm_ref *ref = inst->src[arg][i]; + if (ref) { + if (readmask & (1<<i)) { + ref->value->contributes_to_output = 1; + } + else { + unlink_ref(ref); + inst->src[arg][i] = NULL; + } + } + } +} + +static GLuint get_texcoord_mask( GLuint tex_idx ) +{ + switch (tex_idx) { + case TGSI_TEXTURE_1D: + return BRW_WRITEMASK_X; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + return BRW_WRITEMASK_XY; + case TGSI_TEXTURE_3D: + return BRW_WRITEMASK_XYZ; + case TGSI_TEXTURE_CUBE: + return BRW_WRITEMASK_XYZ; + + case TGSI_TEXTURE_SHADOW1D: + return BRW_WRITEMASK_XZ; + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + return BRW_WRITEMASK_XYZ; + default: + assert(0); + return 0; + } +} + + +/* Step two: Basically this is dead code elimination. + * + * Iterate backwards over instructions, noting which values + * contribute to the final result. Adjust writemasks to only + * calculate these values. + */ +void brw_wm_pass1( struct brw_wm_compile *c ) +{ + GLint insn; + + for (insn = c->nr_insns-1; insn >= 0; insn--) { + struct brw_wm_instruction *inst = &c->instruction[insn]; + GLuint writemask; + GLuint read0, read1, read2; + + if (inst->opcode == TGSI_OPCODE_KIL) { + track_arg(c, inst, 0, BRW_WRITEMASK_XYZW); /* All args contribute to final */ + continue; + } + + if (inst->opcode == WM_FB_WRITE) { + track_arg(c, inst, 0, BRW_WRITEMASK_XYZW); + track_arg(c, inst, 1, BRW_WRITEMASK_XYZW); + if (c->key.source_depth_to_render_target && + c->key.computes_depth) + track_arg(c, inst, 2, BRW_WRITEMASK_Z); + else + track_arg(c, inst, 2, 0); + continue; + } + + /* Lookup all the registers which were written by this + * instruction and get a mask of those that contribute to the output: + */ + writemask = get_tracked_mask(c, inst); + if (!writemask) { + GLuint arg; + for (arg = 0; arg < 3; arg++) + track_arg(c, inst, arg, 0); + continue; + } + + read0 = 0; + read1 = 0; + read2 = 0; + + /* Mark all inputs which contribute to the marked outputs: + */ + switch (inst->opcode) { + case TGSI_OPCODE_ABS: + case TGSI_OPCODE_FLR: + case TGSI_OPCODE_FRC: + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_TRUNC: + read0 = writemask; + break; + + case TGSI_OPCODE_SUB: + case TGSI_OPCODE_SLT: + case TGSI_OPCODE_SLE: + case TGSI_OPCODE_SGE: + case TGSI_OPCODE_SGT: + case TGSI_OPCODE_SEQ: + case TGSI_OPCODE_SNE: + case TGSI_OPCODE_ADD: + case TGSI_OPCODE_MAX: + case TGSI_OPCODE_MIN: + case TGSI_OPCODE_MUL: + read0 = writemask; + read1 = writemask; + break; + + case TGSI_OPCODE_DDX: + case TGSI_OPCODE_DDY: + read0 = writemask; + break; + + case TGSI_OPCODE_MAD: + case TGSI_OPCODE_CMP: + case TGSI_OPCODE_LRP: + read0 = writemask; + read1 = writemask; + read2 = writemask; + break; + + case TGSI_OPCODE_XPD: + if (writemask & BRW_WRITEMASK_X) read0 |= BRW_WRITEMASK_YZ; + if (writemask & BRW_WRITEMASK_Y) read0 |= BRW_WRITEMASK_XZ; + if (writemask & BRW_WRITEMASK_Z) read0 |= BRW_WRITEMASK_XY; + read1 = read0; + break; + + case TGSI_OPCODE_COS: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SIN: + case TGSI_OPCODE_SCS: + case WM_CINTERP: + case WM_PIXELXY: + read0 = BRW_WRITEMASK_X; + break; + + case TGSI_OPCODE_POW: + read0 = BRW_WRITEMASK_X; + read1 = BRW_WRITEMASK_X; + break; + + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXP: + read0 = get_texcoord_mask(inst->target); + break; + + case TGSI_OPCODE_TXB: + read0 = get_texcoord_mask(inst->target) | BRW_WRITEMASK_W; + break; + + case WM_WPOSXY: + read0 = writemask & BRW_WRITEMASK_XY; + break; + + case WM_DELTAXY: + read0 = writemask & BRW_WRITEMASK_XY; + read1 = BRW_WRITEMASK_X; + break; + + case WM_PIXELW: + read0 = BRW_WRITEMASK_X; + read1 = BRW_WRITEMASK_XY; + break; + + case WM_LINTERP: + read0 = BRW_WRITEMASK_X; + read1 = BRW_WRITEMASK_XY; + break; + + case WM_PINTERP: + read0 = BRW_WRITEMASK_X; /* interpolant */ + read1 = BRW_WRITEMASK_XY; /* deltas */ + read2 = BRW_WRITEMASK_W; /* pixel w */ + break; + + case TGSI_OPCODE_DP3: + read0 = BRW_WRITEMASK_XYZ; + read1 = BRW_WRITEMASK_XYZ; + break; + + case TGSI_OPCODE_DPH: + read0 = BRW_WRITEMASK_XYZ; + read1 = BRW_WRITEMASK_XYZW; + break; + + case TGSI_OPCODE_DP4: + read0 = BRW_WRITEMASK_XYZW; + read1 = BRW_WRITEMASK_XYZW; + break; + + case TGSI_OPCODE_LIT: + read0 = BRW_WRITEMASK_XYW; + break; + + case TGSI_OPCODE_DST: + case WM_FRONTFACING: + case TGSI_OPCODE_KILP: + default: + break; + } + + track_arg(c, inst, 0, read0); + track_arg(c, inst, 1, read1); + track_arg(c, inst, 2, read2); + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass1"); + } +} diff --git a/src/gallium/drivers/i965/brw_wm_pass2.c b/src/gallium/drivers/i965/brw_wm_pass2.c new file mode 100644 index 00000000000..19248b45195 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_pass2.c @@ -0,0 +1,334 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + + +#include "brw_debug.h" +#include "brw_wm.h" + + +/* Use these to force spilling so that that functionality can be + * tested with known-good examples rather than having to construct new + * tests. + */ +#define TEST_PAYLOAD_SPILLS 0 +#define TEST_DST_SPILLS 0 + +static void spill_value(struct brw_wm_compile *c, + struct brw_wm_value *value); + +static void prealloc_reg(struct brw_wm_compile *c, + struct brw_wm_value *value, + GLuint reg) +{ + if (value->lastuse) { + /* Set nextuse to zero, it will be corrected by + * update_register_usage(). + */ + c->pass2_grf[reg].value = value; + c->pass2_grf[reg].nextuse = 0; + + value->resident = &c->pass2_grf[reg]; + value->hw_reg = brw_vec8_grf(reg*2, 0); + + if (TEST_PAYLOAD_SPILLS) + spill_value(c, value); + } +} + + +/* Initialize all the register values. Do the initial setup + * calculations for interpolants. + */ +static void init_registers( struct brw_wm_compile *c ) +{ + GLuint reg = 0; + GLuint j; + + for (j = 0; j < c->grf_limit; j++) + c->pass2_grf[j].nextuse = BRW_WM_MAX_INSN; + + /* Pre-allocate incoming payload regs: + */ + for (j = 0; j < c->key.nr_depth_regs; j++) + prealloc_reg(c, &c->payload.depth[j], reg++); + + for (j = 0; j < c->nr_creg; j++) + prealloc_reg(c, &c->creg[j], reg++); + + reg++; /* XXX: skip over position output */ + + /* XXX: currently just hope the VS outputs line up with FS inputs: + */ + for (j = 0; j < c->key.nr_inputs; j++) + prealloc_reg(c, &c->payload.input_interp[j], reg++); + + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = (c->key.nr_inputs + 1) * 2; + c->prog_data.curb_read_length = c->nr_creg * 2; + + /* Note this allocation: + */ + c->max_wm_grf = reg * 2; +} + + +/* Update the nextuse value for each register in our file. + */ +static void update_register_usage(struct brw_wm_compile *c, + GLuint thisinsn) +{ + GLuint i; + + for (i = 1; i < c->grf_limit; i++) { + struct brw_wm_grf *grf = &c->pass2_grf[i]; + + /* Only search those which can change: + */ + if (grf->nextuse < thisinsn) { + const struct brw_wm_ref *ref = grf->value->lastuse; + + /* Has last use of value been passed? + */ + if (ref->insn < thisinsn) { + grf->value->resident = 0; + grf->value = 0; + grf->nextuse = BRW_WM_MAX_INSN; + } + else { + /* Else loop through chain to update: + */ + while (ref->prevuse && ref->prevuse->insn >= thisinsn) + ref = ref->prevuse; + + grf->nextuse = ref->insn; + } + } + } +} + + +static void spill_value(struct brw_wm_compile *c, + struct brw_wm_value *value) +{ + /* Allocate a spill slot. Note that allocations start from 0x40 - + * the first slot is reserved to mean "undef" in brw_wm_emit.c + */ + if (!value->spill_slot) { + c->last_scratch += 0x40; + value->spill_slot = c->last_scratch; + } + + /* The spill will be done in brw_wm_emit.c immediately after the + * value is calculated, so we can just take this reg without any + * further work. + */ + value->resident->value = NULL; + value->resident->nextuse = BRW_WM_MAX_INSN; + value->resident = NULL; +} + + + +/* Search for contiguous region with the most distant nearest + * member. Free regs count as very distant. + * + * TODO: implement spill-to-reg so that we can rearrange discontigous + * free regs and then spill the oldest non-free regs in sequence. + * This would mean inserting instructions in this pass. + */ +static GLuint search_contiguous_regs(struct brw_wm_compile *c, + GLuint nr, + GLuint thisinsn) +{ + struct brw_wm_grf *grf = c->pass2_grf; + GLuint furthest = 0; + GLuint reg = 0; + GLuint i, j; + + /* Start search at 1: r0 is special and can't be used or spilled. + */ + for (i = 1; i < c->grf_limit && furthest < BRW_WM_MAX_INSN; i++) { + GLuint group_nextuse = BRW_WM_MAX_INSN; + + for (j = 0; j < nr; j++) { + if (grf[i+j].nextuse < group_nextuse) + group_nextuse = grf[i+j].nextuse; + } + + if (group_nextuse > furthest) { + furthest = group_nextuse; + reg = i; + } + } + + assert(furthest != thisinsn); + + /* Any non-empty regs will need to be spilled: + */ + for (j = 0; j < nr; j++) + if (grf[reg+j].value) + spill_value(c, grf[reg+j].value); + + return reg; +} + + +static void alloc_contiguous_dest(struct brw_wm_compile *c, + struct brw_wm_value *dst[], + GLuint nr, + GLuint thisinsn) +{ + GLuint reg = search_contiguous_regs(c, nr, thisinsn); + GLuint i; + + for (i = 0; i < nr; i++) { + if (!dst[i]) { + /* Need to grab a dummy value in TEX case. Don't introduce + * it into the tracking scheme. + */ + dst[i] = &c->vreg[c->nr_vreg++]; + } + else { + assert(!dst[i]->resident); + assert(c->pass2_grf[reg+i].nextuse != thisinsn); + + c->pass2_grf[reg+i].value = dst[i]; + c->pass2_grf[reg+i].nextuse = thisinsn; + + dst[i]->resident = &c->pass2_grf[reg+i]; + } + + dst[i]->hw_reg = brw_vec8_grf((reg+i)*2, 0); + } + + if ((reg+nr)*2 > c->max_wm_grf) + c->max_wm_grf = (reg+nr) * 2; +} + + +static void load_args(struct brw_wm_compile *c, + struct brw_wm_instruction *inst) +{ + GLuint thisinsn = inst - c->instruction; + GLuint i,j; + + for (i = 0; i < 3; i++) { + for (j = 0; j < 4; j++) { + struct brw_wm_ref *ref = inst->src[i][j]; + + if (ref) { + if (!ref->value->resident) { + /* Need to bring the value in from scratch space. The code for + * this will be done in brw_wm_emit.c, here we just do the + * register allocation and mark the ref as requiring a fill. + */ + GLuint reg = search_contiguous_regs(c, 1, thisinsn); + + c->pass2_grf[reg].value = ref->value; + c->pass2_grf[reg].nextuse = thisinsn; + + ref->value->resident = &c->pass2_grf[reg]; + + /* Note that a fill is required: + */ + ref->unspill_reg = reg*2; + } + + /* Adjust the hw_reg to point at the value's current location: + */ + assert(ref->value == ref->value->resident->value); + ref->hw_reg.nr += (ref->value->resident - c->pass2_grf) * 2; + } + } + } +} + + + +/* Step 3: Work forwards once again. Perform register allocations, + * taking into account instructions like TEX which require contiguous + * result registers. Where necessary spill registers to scratch space + * and reload later. + */ +void brw_wm_pass2( struct brw_wm_compile *c ) +{ + GLuint insn; + GLuint i; + + init_registers(c); + + for (insn = 0; insn < c->nr_insns; insn++) { + struct brw_wm_instruction *inst = &c->instruction[insn]; + + /* Update registers' nextuse values: + */ + update_register_usage(c, insn); + + /* May need to unspill some args. + */ + load_args(c, inst); + + /* Allocate registers to hold results: + */ + switch (inst->opcode) { + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXP: + alloc_contiguous_dest(c, inst->dst, 4, insn); + break; + + default: + for (i = 0; i < 4; i++) { + if (inst->writemask & (1<<i)) { + assert(inst->dst[i]); + alloc_contiguous_dest(c, &inst->dst[i], 1, insn); + } + } + break; + } + + if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) { + for (i = 0; i < 4; i++) + if (inst->dst[i]) + spill_value(c, inst->dst[i]); + } + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass2"); + } + + c->state = PASS2_DONE; + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass2/done"); + } +} diff --git a/src/gallium/drivers/i965/brw_wm_sampler_state.c b/src/gallium/drivers/i965/brw_wm_sampler_state.c new file mode 100644 index 00000000000..4e99ac703a9 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_sampler_state.c @@ -0,0 +1,228 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_screen.h" + + +/* Samplers aren't strictly wm state from the hardware's perspective, + * but that is the only situation in which we use them in this driver. + */ + + + +static enum pipe_error +upload_default_color( struct brw_context *brw, + const GLfloat *color, + struct brw_winsys_buffer **bo_out ) +{ + struct brw_sampler_default_color sdc; + enum pipe_error ret; + + COPY_4V(sdc.color, color); + + ret = brw_cache_data( &brw->cache, BRW_SAMPLER_DEFAULT_COLOR, &sdc, + NULL, 0, bo_out ); + if (ret) + return ret; + + return PIPE_OK; +} + + +struct wm_sampler_key { + int sampler_count; + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; +}; + + +/** Sets up the cache key for sampler state for all texture units */ +static void +brw_wm_sampler_populate_key(struct brw_context *brw, + struct wm_sampler_key *key) +{ + int i; + + memset(key, 0, sizeof(*key)); + + key->sampler_count = MIN2(brw->curr.num_textures, + brw->curr.num_samplers); + + for (i = 0; i < key->sampler_count; i++) { + const struct brw_texture *tex = brw_texture(brw->curr.texture[i]); + const struct brw_sampler *sampler = brw->curr.sampler[i]; + struct brw_sampler_state *entry = &key->sampler[i]; + + entry->ss0 = sampler->ss0; + entry->ss1 = sampler->ss1; + entry->ss2.default_color_pointer = 0; /* reloc */ + entry->ss3 = sampler->ss3; + + /* Cube-maps on 965 and later must use the same wrap mode for all 3 + * coordinate dimensions. Futher, only CUBE and CLAMP are valid. + */ + if (tex->base.target == PIPE_TEXTURE_CUBE) { + if (FALSE && + (sampler->ss0.min_filter != BRW_MAPFILTER_NEAREST || + sampler->ss0.mag_filter != BRW_MAPFILTER_NEAREST)) { + entry->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE; + entry->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE; + entry->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE; + } else { + entry->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + entry->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + entry->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + } + } else if (tex->base.target == PIPE_TEXTURE_1D) { + /* There's a bug in 1D texture sampling - it actually pays + * attention to the wrap_t value, though it should not. + * Override the wrap_t value here to GL_REPEAT to keep + * any nonexistent border pixels from floating in. + */ + entry->ss1.t_wrap_mode = BRW_TEXCOORDMODE_WRAP; + } + } +} + + +static enum pipe_error +brw_wm_sampler_update_default_colors(struct brw_context *brw) +{ + enum pipe_error ret; + int nr = MIN2(brw->curr.num_textures, + brw->curr.num_samplers); + int i; + + for (i = 0; i < nr; i++) { + const struct brw_texture *tex = brw_texture(brw->curr.texture[i]); + const struct brw_sampler *sampler = brw->curr.sampler[i]; + const float *bc; + + if (pf_is_depth_or_stencil(tex->base.format)) { + float bordercolor[4] = { + sampler->border_color[0], + sampler->border_color[0], + sampler->border_color[0], + sampler->border_color[0] + }; + + bc = bordercolor; + } + else { + bc = sampler->border_color; + } + + /* GL specs that border color for depth textures is taken from the + * R channel, while the hardware uses A. Spam R into all the + * channels for safety. + */ + ret = upload_default_color(brw, + bc, + &brw->wm.sdc_bo[i]); + if (ret) + return ret; + } + + return PIPE_OK; +} + + + +/* All samplers must be uploaded in a single contiguous array. + */ +static int upload_wm_samplers( struct brw_context *brw ) +{ + struct wm_sampler_key key; + struct brw_winsys_reloc reloc[BRW_MAX_TEX_UNIT]; + enum pipe_error ret; + int i; + + brw_wm_sampler_update_default_colors(brw); + brw_wm_sampler_populate_key(brw, &key); + + if (brw->wm.sampler_count != key.sampler_count) { + brw->wm.sampler_count = key.sampler_count; + brw->state.dirty.cache |= CACHE_NEW_SAMPLER; + } + + if (brw->wm.sampler_count == 0) { + bo_reference(&brw->wm.sampler_bo, NULL); + return PIPE_OK; + } + + /* Emit SDC relocations */ + for (i = 0; i < key.sampler_count; i++) { + make_reloc( &reloc[i], + BRW_USAGE_SAMPLER, + 0, + i * sizeof(struct brw_sampler_state) + + offsetof(struct brw_sampler_state, ss2), + brw->wm.sdc_bo[i]); + } + + + if (brw_search_cache(&brw->cache, BRW_SAMPLER, + &key, sizeof(key), + reloc, key.sampler_count, + NULL, + &brw->wm.sampler_bo)) + return PIPE_OK; + + /* If we didnt find it in the cache, compute the state and put it in the + * cache. + */ + ret = brw_upload_cache(&brw->cache, BRW_SAMPLER, + &key, sizeof(key), + reloc, key.sampler_count, + &key.sampler, sizeof(key.sampler), + NULL, NULL, + &brw->wm.sampler_bo); + if (ret) + return ret; + + + return 0; +} + +const struct brw_tracked_state brw_wm_samplers = { + .dirty = { + .mesa = PIPE_NEW_BOUND_TEXTURES | PIPE_NEW_SAMPLERS, + .brw = 0, + .cache = 0 + }, + .prepare = upload_wm_samplers, +}; + + diff --git a/src/gallium/drivers/i965/brw_wm_state.c b/src/gallium/drivers/i965/brw_wm_state.c new file mode 100644 index 00000000000..ee970ac75bc --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_state.c @@ -0,0 +1,339 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "util/u_math.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_wm.h" +#include "brw_debug.h" +#include "brw_pipe_rast.h" + +/*********************************************************************** + * WM unit - fragment programs and rasterization + */ + +struct brw_wm_unit_key { + unsigned int total_grf, total_scratch; + unsigned int urb_entry_read_length; + unsigned int curb_entry_read_length; + unsigned int dispatch_grf_start_reg; + + unsigned int curbe_offset; + unsigned int urb_size; + + unsigned int max_threads; + + unsigned int nr_surfaces, sampler_count; + GLboolean uses_depth, computes_depth, uses_kill, has_flow_control; + GLboolean polygon_stipple, stats_wm, line_stipple, offset_enable; + GLfloat offset_units, offset_factor; +}; + +static void +wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) +{ + const struct brw_fragment_shader *fp = brw->curr.fragment_shader; + + memset(key, 0, sizeof(*key)); + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + key->max_threads = 1; + else { + /* WM maximum threads is number of EUs times number of threads per EU. */ + if (BRW_IS_IGDNG(brw)) + key->max_threads = 12 * 6; + else if (BRW_IS_G4X(brw)) + key->max_threads = 10 * 5; + else + key->max_threads = 8 * 4; + } + + /* CACHE_NEW_WM_PROG */ + key->total_grf = brw->wm.prog_data->total_grf; + key->urb_entry_read_length = brw->wm.prog_data->urb_read_length; + key->curb_entry_read_length = brw->wm.prog_data->curb_read_length; + key->dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; + key->total_scratch = align(brw->wm.prog_data->total_scratch, 1024); + + /* BRW_NEW_URB_FENCE */ + key->urb_size = brw->urb.vsize; + + /* BRW_NEW_CURBE_OFFSETS */ + key->curbe_offset = brw->curbe.wm_start; + + /* BRW_NEW_NR_SURFACEs */ + key->nr_surfaces = brw->wm.nr_surfaces; + + /* CACHE_NEW_SAMPLER */ + key->sampler_count = brw->wm.sampler_count; + + /* PIPE_NEW_RAST */ + key->polygon_stipple = brw->curr.rast->templ.poly_stipple_enable; + + /* PIPE_NEW_FRAGMENT_PROGRAM */ + key->uses_depth = fp->uses_depth; + key->computes_depth = fp->info.writes_z; + + /* PIPE_NEW_DEPTH_BUFFER + * + * Override for NULL depthbuffer case, required by the Pixel Shader Computed + * Depth field. + */ + if (brw->curr.fb.zsbuf == NULL) + key->computes_depth = 0; + + /* PIPE_NEW_DEPTH_STENCIL_ALPHA */ + key->uses_kill = (fp->info.uses_kill || + brw->curr.zstencil->cc3.alpha_test); + + key->has_flow_control = fp->has_flow_control; + + /* temporary sanity check assertion */ + assert(fp->has_flow_control == 0); + + /* PIPE_NEW_QUERY */ + key->stats_wm = (brw->query.stats_wm != 0); + + /* PIPE_NEW_RAST */ + key->line_stipple = brw->curr.rast->templ.line_stipple_enable; + + + key->offset_enable = (brw->curr.rast->templ.offset_cw || + brw->curr.rast->templ.offset_ccw); + + key->offset_units = brw->curr.rast->templ.offset_units; + key->offset_factor = brw->curr.rast->templ.offset_scale; +} + +/** + * Setup wm hardware state. See page 225 of Volume 2 + */ +static enum pipe_error +wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key, + struct brw_winsys_reloc *reloc, + unsigned nr_reloc, + struct brw_winsys_buffer **bo_out) +{ + struct brw_wm_unit_state wm; + enum pipe_error ret; + + memset(&wm, 0, sizeof(wm)); + + wm.thread0.grf_reg_count = align(key->total_grf, 16) / 16 - 1; + wm.thread0.kernel_start_pointer = 0; /* reloc */ + wm.thread1.depth_coef_urb_read_offset = 1; + wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + + if (BRW_IS_IGDNG(brw)) + wm.thread1.binding_table_entry_count = 0; /* hardware requirement */ + else + wm.thread1.binding_table_entry_count = key->nr_surfaces; + + if (key->total_scratch != 0) { + wm.thread2.scratch_space_base_pointer = 0; /* reloc */ + wm.thread2.per_thread_scratch_space = key->total_scratch / 1024 - 1; + } else { + wm.thread2.scratch_space_base_pointer = 0; + wm.thread2.per_thread_scratch_space = 0; + } + + wm.thread3.dispatch_grf_start_reg = key->dispatch_grf_start_reg; + wm.thread3.urb_entry_read_length = key->urb_entry_read_length; + wm.thread3.urb_entry_read_offset = 0; + wm.thread3.const_urb_entry_read_length = key->curb_entry_read_length; + wm.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; + + if (BRW_IS_IGDNG(brw)) + wm.wm4.sampler_count = 0; /* hardware requirement */ + else + wm.wm4.sampler_count = (key->sampler_count + 1) / 4; + + /* reloc */ + wm.wm4.sampler_state_pointer = 0; + + wm.wm5.program_uses_depth = key->uses_depth; + wm.wm5.program_computes_depth = key->computes_depth; + wm.wm5.program_uses_killpixel = key->uses_kill; + + if (key->has_flow_control) + wm.wm5.enable_8_pix = 1; + else + wm.wm5.enable_16_pix = 1; + + wm.wm5.max_threads = key->max_threads - 1; + wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */ + wm.wm5.legacy_line_rast = 0; + wm.wm5.legacy_global_depth_bias = 0; + wm.wm5.early_depth_test = 1; /* never need to disable */ + wm.wm5.line_aa_region_width = 0; + wm.wm5.line_endcap_aa_region_width = 1; + + wm.wm5.polygon_stipple = key->polygon_stipple; + + if (key->offset_enable) { + wm.wm5.depth_offset = 1; + /* Something wierd going on with legacy_global_depth_bias, + * offset_constant, scaling and MRD. This value passes glean + * but gives some odd results elsewere (eg. the + * quad-offset-units test). + */ + wm.global_depth_offset_constant = key->offset_units * 2; + + /* This is the only value that passes glean: + */ + wm.global_depth_offset_scale = key->offset_factor; + } + + wm.wm5.line_stipple = key->line_stipple; + + if ((BRW_DEBUG & DEBUG_STATS) || key->stats_wm) + wm.wm4.stats_enable = 1; + + ret = brw_upload_cache(&brw->cache, BRW_WM_UNIT, + key, sizeof(*key), + reloc, nr_reloc, + &wm, sizeof(wm), + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + + +static enum pipe_error upload_wm_unit( struct brw_context *brw ) +{ + struct brw_wm_unit_key key; + struct brw_winsys_reloc reloc[3]; + unsigned nr_reloc = 0; + enum pipe_error ret; + unsigned grf_reg_count; + unsigned per_thread_scratch_space; + unsigned stats_enable; + unsigned sampler_count; + + wm_unit_populate_key(brw, &key); + + + /* Allocate the necessary scratch space if we haven't already. Don't + * bother reducing the allocation later, since we use scratch so + * rarely. + */ + assert(key.total_scratch <= 12 * 1024); + if (key.total_scratch) { + GLuint total = key.total_scratch * key.max_threads; + + /* Do we need a new buffer: + */ + if (brw->wm.scratch_bo && total > brw->wm.scratch_bo->size) + bo_reference(&brw->wm.scratch_bo, NULL); + + if (brw->wm.scratch_bo == NULL) { + ret = brw->sws->bo_alloc(brw->sws, + BRW_BUFFER_TYPE_SHADER_SCRATCH, + total, + 4096, + &brw->wm.scratch_bo); + if (ret) + return ret; + } + } + + + /* XXX: temporary: + */ + grf_reg_count = (align(key.total_grf, 16) / 16 - 1); + per_thread_scratch_space = key.total_scratch / 1024 - 1; + stats_enable = (BRW_DEBUG & DEBUG_STATS) || key.stats_wm; + sampler_count = BRW_IS_IGDNG(brw) ? 0 :(key.sampler_count + 1) / 4; + + /* Emit WM program relocation */ + make_reloc(&reloc[nr_reloc++], + BRW_USAGE_STATE, + grf_reg_count << 1, + offsetof(struct brw_wm_unit_state, thread0), + brw->wm.prog_bo); + + /* Emit scratch space relocation */ + if (key.total_scratch != 0) { + make_reloc(&reloc[nr_reloc++], + BRW_USAGE_SCRATCH, + per_thread_scratch_space, + offsetof(struct brw_wm_unit_state, thread2), + brw->wm.scratch_bo); + } + + /* Emit sampler state relocation */ + if (key.sampler_count != 0) { + make_reloc(&reloc[nr_reloc++], + BRW_USAGE_STATE, + stats_enable | (sampler_count << 2), + offsetof(struct brw_wm_unit_state, wm4), + brw->wm.sampler_bo); + } + + + if (brw_search_cache(&brw->cache, BRW_WM_UNIT, + &key, sizeof(key), + reloc, nr_reloc, + NULL, + &brw->wm.state_bo)) + return PIPE_OK; + + ret = wm_unit_create_from_key(brw, &key, + reloc, nr_reloc, + &brw->wm.state_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_wm_unit = { + .dirty = { + .mesa = (PIPE_NEW_FRAGMENT_SHADER | + PIPE_NEW_DEPTH_BUFFER | + PIPE_NEW_RAST | + PIPE_NEW_DEPTH_STENCIL_ALPHA | + PIPE_NEW_QUERY), + + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_NR_WM_SURFACES), + + .cache = (CACHE_NEW_WM_PROG | + CACHE_NEW_SAMPLER) + }, + .prepare = upload_wm_unit, +}; + diff --git a/src/gallium/drivers/i965/brw_wm_surface_state.c b/src/gallium/drivers/i965/brw_wm_surface_state.c new file mode 100644 index 00000000000..f92b8198ed3 --- /dev/null +++ b/src/gallium/drivers/i965/brw_wm_surface_state.c @@ -0,0 +1,294 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell <[email protected]> + */ + +#include "pipe/p_format.h" + +#include "brw_batchbuffer.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_screen.h" + + + + +static enum pipe_error +brw_update_texture_surface( struct brw_context *brw, + struct brw_texture *tex, + struct brw_winsys_buffer **bo_out) +{ + struct brw_winsys_reloc reloc[1]; + enum pipe_error ret; + + /* Emit relocation to surface contents */ + make_reloc(&reloc[0], + BRW_USAGE_SAMPLER, + 0, + offsetof(struct brw_surface_state, ss1), + tex->bo); + + if (brw_search_cache(&brw->surface_cache, + BRW_SS_SURFACE, + &tex->ss, sizeof tex->ss, + reloc, Elements(reloc), + NULL, + bo_out)) + return PIPE_OK; + + ret = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE, + &tex->ss, sizeof tex->ss, + reloc, Elements(reloc), + &tex->ss, sizeof tex->ss, + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + + + + + + + + +/** + * Sets up a surface state structure to point at the given region. + * While it is only used for the front/back buffer currently, it should be + * usable for further buffers when doing ARB_draw_buffer support. + */ +static enum pipe_error +brw_update_render_surface(struct brw_context *brw, + struct brw_surface *surface, + struct brw_winsys_buffer **bo_out) +{ + struct brw_surf_ss0 blend_ss0 = brw->curr.blend->ss0; + struct brw_surface_state ss; + struct brw_winsys_reloc reloc[1]; + enum pipe_error ret; + + /* XXX: we will only be rendering to this surface: + */ + make_reloc(&reloc[0], + BRW_USAGE_RENDER_TARGET, + 0, + offsetof(struct brw_surface_state, ss1), + surface->bo); + + /* Surfaces are potentially shared between contexts, so can't + * scribble the in-place ss0 value in the surface. + */ + memcpy(&ss, &surface->ss, sizeof ss); + + ss.ss0.color_blend = blend_ss0.color_blend; + ss.ss0.writedisable_blue = blend_ss0.writedisable_blue; + ss.ss0.writedisable_green = blend_ss0.writedisable_green; + ss.ss0.writedisable_red = blend_ss0.writedisable_red; + ss.ss0.writedisable_alpha = blend_ss0.writedisable_alpha; + + if (brw_search_cache(&brw->surface_cache, + BRW_SS_SURFACE, + &ss, sizeof(ss), + reloc, Elements(reloc), + NULL, + bo_out)) + return PIPE_OK; + + ret = brw_upload_cache(&brw->surface_cache, + BRW_SS_SURFACE, + &ss, sizeof ss, + reloc, Elements(reloc), + &ss, sizeof ss, + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + + +/** + * Constructs the binding table for the WM surface state, which maps unit + * numbers to surface state objects. + */ +static enum pipe_error +brw_wm_get_binding_table(struct brw_context *brw, + struct brw_winsys_buffer **bo_out ) +{ + enum pipe_error ret; + struct brw_winsys_reloc reloc[BRW_WM_MAX_SURF]; + uint32_t data[BRW_WM_MAX_SURF]; + GLuint nr_relocs = 0; + GLuint data_size = brw->wm.nr_surfaces * sizeof data[0]; + int i; + + assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF); + assert(brw->wm.nr_surfaces > 0); + + /* Emit binding table relocations to surface state + */ + for (i = 0; i < brw->wm.nr_surfaces; i++) { + if (brw->wm.surf_bo[i]) { + make_reloc(&reloc[nr_relocs++], + BRW_USAGE_STATE, + 0, + i * sizeof(GLuint), + brw->wm.surf_bo[i]); + } + } + + /* Note there is no key for this search beyond the values in the + * relocation array: + */ + if (brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND, + NULL, 0, + reloc, nr_relocs, + NULL, + bo_out)) + return PIPE_OK; + + /* Upload zero data, will all be overwitten with relocation + * offsets: + */ + for (i = 0; i < brw->wm.nr_surfaces; i++) + data[i] = 0; + + ret = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND, + NULL, 0, + reloc, nr_relocs, + data, data_size, + NULL, NULL, + bo_out); + if (ret) + return ret; + + return PIPE_OK; +} + +static enum pipe_error prepare_wm_surfaces(struct brw_context *brw ) +{ + enum pipe_error ret; + int nr_surfaces = 0; + GLuint i; + + /* PIPE_NEW_COLOR_BUFFERS | PIPE_NEW_BLEND + * + * Update surfaces for drawing buffers. Mixes in colormask and + * blend state. + * + * XXX: no color buffer case + */ + for (i = 0; i < brw->curr.fb.nr_cbufs; i++) { + ret = brw_update_render_surface(brw, + brw_surface(brw->curr.fb.cbufs[i]), + &brw->wm.surf_bo[BTI_COLOR_BUF(i)]); + if (ret) + return ret; + + nr_surfaces = BTI_COLOR_BUF(i) + 1; + } + + + + /* PIPE_NEW_FRAGMENT_CONSTANTS + */ +#if 0 + if (brw->curr.fragment_constants) { + ret = brw_update_fragment_constant_surface( + brw, + brw->curr.fragment_constants, + &brw->wm.surf_bo[BTI_FRAGMENT_CONSTANTS]); + + if (ret) + return ret; + + nr_surfaces = BTI_FRAGMENT_CONSTANTS + 1; + } + else { + bo_reference(&brw->wm.surf_bo[SURF_FRAG_CONSTANTS], NULL); + } +#endif + + + /* PIPE_NEW_TEXTURE + */ + for (i = 0; i < brw->curr.num_textures; i++) { + ret = brw_update_texture_surface(brw, + brw_texture(brw->curr.texture[i]), + &brw->wm.surf_bo[BTI_TEXTURE(i)]); + if (ret) + return ret; + + nr_surfaces = BTI_TEXTURE(i) + 1; + } + + /* Clear any inactive entries: + */ + for (i = brw->curr.fb.nr_cbufs; i < BRW_MAX_DRAW_BUFFERS; i++) + bo_reference(&brw->wm.surf_bo[BTI_COLOR_BUF(i)], NULL); + + if (!brw->curr.fragment_constants) + bo_reference(&brw->wm.surf_bo[BTI_FRAGMENT_CONSTANTS], NULL); + + /* XXX: no pipe_max_textures define?? */ + for (i = brw->curr.num_textures; i < PIPE_MAX_SAMPLERS; i++) + bo_reference(&brw->wm.surf_bo[BTI_TEXTURE(i)], NULL); + + if (brw->wm.nr_surfaces != nr_surfaces) { + brw->wm.nr_surfaces = nr_surfaces; + brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES; + } + + ret = brw_wm_get_binding_table(brw, &brw->wm.bind_bo); + if (ret) + return ret; + + return PIPE_OK; +} + +const struct brw_tracked_state brw_wm_surfaces = { + .dirty = { + .mesa = (PIPE_NEW_COLOR_BUFFERS | + PIPE_NEW_BOUND_TEXTURES | + PIPE_NEW_FRAGMENT_CONSTANTS | + PIPE_NEW_BLEND), + .brw = (BRW_NEW_CONTEXT | + BRW_NEW_WM_SURFACES), + .cache = 0 + }, + .prepare = prepare_wm_surfaces, +}; + + + diff --git a/src/gallium/drivers/i965/intel_decode.c b/src/gallium/drivers/i965/intel_decode.c new file mode 100644 index 00000000000..3166958bad3 --- /dev/null +++ b/src/gallium/drivers/i965/intel_decode.c @@ -0,0 +1,1790 @@ +/* -*- c-basic-offset: 4 -*- */ +/* + * Copyright © 2007 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt <[email protected]> + * + */ + +/** @file intel_decode.c + * This file contains code to print out batchbuffer contents in a + * human-readable format. + * + * The current version only supports i915 packets, and only pretty-prints a + * subset of them. The intention is for it to make just a best attempt to + * decode, but never crash in the process. + */ + +#include <stdio.h> +#include <stdarg.h> +#include <string.h> +#include <inttypes.h> + +#include "intel_decode.h" + +/*#include "intel_chipset.h"*/ +#define IS_965(x) 1 /* XXX */ +#define IS_9XX(x) 1 /* XXX */ + +#define BUFFER_FAIL(_count, _len, _name) do { \ + fprintf(out, "Buffer size too small in %s (%d < %d)\n", \ + (_name), (_count), (_len)); \ + (*failures)++; \ + return count; \ +} while (0) + +static FILE *out; +static uint32_t saved_s2 = 0, saved_s4 = 0; +static char saved_s2_set = 0, saved_s4_set = 0; + +static float +int_as_float(uint32_t intval) +{ + union intfloat { + uint32_t i; + float f; + } uval; + + uval.i = intval; + return uval.f; +} + +static void +instr_out(const uint32_t *data, uint32_t hw_offset, unsigned int index, + char *fmt, ...) +{ + va_list va; + + fprintf(out, "0x%08x: 0x%08x:%s ", hw_offset + index * 4, data[index], + index == 0 ? "" : " "); + va_start(va, fmt); + vfprintf(out, fmt, va); + va_end(va); +} + + +static int +decode_mi(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + unsigned int opcode; + + struct { + uint32_t opcode; + int len_mask; + int min_len; + int max_len; + char *name; + } opcodes_mi[] = { + { 0x08, 0, 1, 1, "MI_ARB_ON_OFF" }, + { 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" }, + { 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" }, + { 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" }, + { 0x04, 0, 1, 1, "MI_FLUSH" }, + { 0x22, 0, 3, 3, "MI_LOAD_REGISTER_IMM" }, + { 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" }, + { 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" }, + { 0x00, 0, 1, 1, "MI_NOOP" }, + { 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" }, + { 0x07, 0, 1, 1, "MI_REPORT_HEAD" }, + { 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT" }, + { 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" }, + { 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" }, + { 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" }, + { 0x02, 0, 1, 1, "MI_USER_INTERRUPT" }, + { 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT" }, + }; + + + for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]); + opcode++) { + if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) { + unsigned int len = 1, i; + + instr_out(data, hw_offset, 0, "%s\n", opcodes_mi[opcode].name); + if (opcodes_mi[opcode].max_len > 1) { + len = (data[0] & opcodes_mi[opcode].len_mask) + 2; + if (len < opcodes_mi[opcode].min_len || + len > opcodes_mi[opcode].max_len) + { + fprintf(out, "Bad length (%d) in %s, [%d, %d]\n", + len, opcodes_mi[opcode].name, + opcodes_mi[opcode].min_len, + opcodes_mi[opcode].max_len); + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_mi[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + + return len; + } + } + + instr_out(data, hw_offset, 0, "MI UNKNOWN\n"); + (*failures)++; + return 1; +} + +static int +decode_2d(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + unsigned int opcode, len; + char *format = NULL; + + struct { + uint32_t opcode; + int min_len; + int max_len; + char *name; + } opcodes_2d[] = { + { 0x40, 5, 5, "COLOR_BLT" }, + { 0x43, 6, 6, "SRC_COPY_BLT" }, + { 0x01, 8, 8, "XY_SETUP_BLT" }, + { 0x11, 9, 9, "XY_SETUP_MONO_PATTERN_SL_BLT" }, + { 0x03, 3, 3, "XY_SETUP_CLIP_BLT" }, + { 0x24, 2, 2, "XY_PIXEL_BLT" }, + { 0x25, 3, 3, "XY_SCANLINES_BLT" }, + { 0x26, 4, 4, "Y_TEXT_BLT" }, + { 0x31, 5, 134, "XY_TEXT_IMMEDIATE_BLT" }, + { 0x50, 6, 6, "XY_COLOR_BLT" }, + { 0x51, 6, 6, "XY_PAT_BLT" }, + { 0x76, 8, 8, "XY_PAT_CHROMA_BLT" }, + { 0x72, 7, 135, "XY_PAT_BLT_IMMEDIATE" }, + { 0x77, 9, 137, "XY_PAT_CHROMA_BLT_IMMEDIATE" }, + { 0x52, 9, 9, "XY_MONO_PAT_BLT" }, + { 0x59, 7, 7, "XY_MONO_PAT_FIXED_BLT" }, + { 0x53, 8, 8, "XY_SRC_COPY_BLT" }, + { 0x54, 8, 8, "XY_MONO_SRC_COPY_BLT" }, + { 0x71, 9, 137, "XY_MONO_SRC_COPY_IMMEDIATE_BLT" }, + { 0x55, 9, 9, "XY_FULL_BLT" }, + { 0x55, 9, 137, "XY_FULL_IMMEDIATE_PATTERN_BLT" }, + { 0x56, 9, 9, "XY_FULL_MONO_SRC_BLT" }, + { 0x75, 10, 138, "XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT" }, + { 0x57, 12, 12, "XY_FULL_MONO_PATTERN_BLT" }, + { 0x58, 12, 12, "XY_FULL_MONO_PATTERN_MONO_SRC_BLT" }, + }; + + switch ((data[0] & 0x1fc00000) >> 22) { + case 0x50: + instr_out(data, hw_offset, 0, + "XY_COLOR_BLT (rgb %sabled, alpha %sabled, dst tile %d)\n", + (data[0] & (1 << 20)) ? "en" : "dis", + (data[0] & (1 << 21)) ? "en" : "dis", + (data[0] >> 11) & 1); + + len = (data[0] & 0x000000ff) + 2; + if (len != 6) + fprintf(out, "Bad count in XY_COLOR_BLT\n"); + if (count < 6) + BUFFER_FAIL(count, len, "XY_COLOR_BLT"); + + switch ((data[1] >> 24) & 0x3) { + case 0: + format="8"; + break; + case 1: + format="565"; + break; + case 2: + format="1555"; + break; + case 3: + format="8888"; + break; + } + + instr_out(data, hw_offset, 1, "format %s, pitch %d, " + "clipping %sabled\n", format, + (short)(data[1] & 0xffff), + data[1] & (1 << 30) ? "en" : "dis"); + instr_out(data, hw_offset, 2, "(%d,%d)\n", + data[2] & 0xffff, data[2] >> 16); + instr_out(data, hw_offset, 3, "(%d,%d)\n", + data[3] & 0xffff, data[3] >> 16); + instr_out(data, hw_offset, 4, "offset 0x%08x\n", data[4]); + instr_out(data, hw_offset, 5, "color\n"); + return len; + case 0x53: + instr_out(data, hw_offset, 0, + "XY_SRC_COPY_BLT (rgb %sabled, alpha %sabled, " + "src tile %d, dst tile %d)\n", + (data[0] & (1 << 20)) ? "en" : "dis", + (data[0] & (1 << 21)) ? "en" : "dis", + (data[0] >> 15) & 1, + (data[0] >> 11) & 1); + + len = (data[0] & 0x000000ff) + 2; + if (len != 8) + fprintf(out, "Bad count in XY_SRC_COPY_BLT\n"); + if (count < 8) + BUFFER_FAIL(count, len, "XY_SRC_COPY_BLT"); + + switch ((data[1] >> 24) & 0x3) { + case 0: + format="8"; + break; + case 1: + format="565"; + break; + case 2: + format="1555"; + break; + case 3: + format="8888"; + break; + } + + instr_out(data, hw_offset, 1, "format %s, dst pitch %d, " + "clipping %sabled\n", format, + (short)(data[1] & 0xffff), + data[1] & (1 << 30) ? "en" : "dis"); + instr_out(data, hw_offset, 2, "dst (%d,%d)\n", + data[2] & 0xffff, data[2] >> 16); + instr_out(data, hw_offset, 3, "dst (%d,%d)\n", + data[3] & 0xffff, data[3] >> 16); + instr_out(data, hw_offset, 4, "dst offset 0x%08x\n", data[4]); + instr_out(data, hw_offset, 5, "src (%d,%d)\n", + data[5] & 0xffff, data[5] >> 16); + instr_out(data, hw_offset, 6, "src pitch %d\n", + (short)(data[6] & 0xffff)); + instr_out(data, hw_offset, 7, "src offset 0x%08x\n", data[7]); + return len; + } + + for (opcode = 0; opcode < sizeof(opcodes_2d) / sizeof(opcodes_2d[0]); + opcode++) { + if ((data[0] & 0x1fc00000) >> 22 == opcodes_2d[opcode].opcode) { + unsigned int i; + + len = 1; + instr_out(data, hw_offset, 0, "%s\n", opcodes_2d[opcode].name); + if (opcodes_2d[opcode].max_len > 1) { + len = (data[0] & 0x000000ff) + 2; + if (len < opcodes_2d[opcode].min_len || + len > opcodes_2d[opcode].max_len) + { + fprintf(out, "Bad count in %s\n", opcodes_2d[opcode].name); + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_2d[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + + return len; + } + } + + instr_out(data, hw_offset, 0, "2D UNKNOWN\n"); + (*failures)++; + return 1; +} + +static int +decode_3d_1c(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + switch ((data[0] & 0x00f80000) >> 19) { + case 0x11: + instr_out(data, hw_offset, 0, "3DSTATE_DEPTH_SUBRECTANGLE_DISALBE\n"); + return 1; + case 0x10: + instr_out(data, hw_offset, 0, "3DSTATE_SCISSOR_ENABLE\n"); + return 1; + case 0x01: + instr_out(data, hw_offset, 0, "3DSTATE_MAP_COORD_SET_I830\n"); + return 1; + case 0x0a: + instr_out(data, hw_offset, 0, "3DSTATE_MAP_CUBE_I830\n"); + return 1; + case 0x05: + instr_out(data, hw_offset, 0, "3DSTATE_MAP_TEX_STREAM_I830\n"); + return 1; + } + + instr_out(data, hw_offset, 0, "3D UNKNOWN\n"); + (*failures)++; + return 1; +} + +/** Sets the string dstname to describe the destination of the PS instruction */ +static void +i915_get_instruction_dst(const uint32_t *data, int i, char *dstname, int do_mask) +{ + uint32_t a0 = data[i]; + int dst_nr = (a0 >> 14) & 0xf; + char dstmask[8]; + char *sat; + + if (do_mask) { + if (((a0 >> 10) & 0xf) == 0xf) { + dstmask[0] = 0; + } else { + int dstmask_index = 0; + + dstmask[dstmask_index++] = '.'; + if (a0 & (1 << 10)) + dstmask[dstmask_index++] = 'x'; + if (a0 & (1 << 11)) + dstmask[dstmask_index++] = 'y'; + if (a0 & (1 << 12)) + dstmask[dstmask_index++] = 'z'; + if (a0 & (1 << 13)) + dstmask[dstmask_index++] = 'w'; + dstmask[dstmask_index++] = 0; + } + + if (a0 & (1 << 22)) + sat = ".sat"; + else + sat = ""; + } else { + dstmask[0] = 0; + sat = ""; + } + + switch ((a0 >> 19) & 0x7) { + case 0: + if (dst_nr > 15) + fprintf(out, "bad destination reg R%d\n", dst_nr); + sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat); + break; + case 4: + if (dst_nr > 0) + fprintf(out, "bad destination reg oC%d\n", dst_nr); + sprintf(dstname, "oC%s%s", dstmask, sat); + break; + case 5: + if (dst_nr > 0) + fprintf(out, "bad destination reg oD%d\n", dst_nr); + sprintf(dstname, "oD%s%s", dstmask, sat); + break; + case 6: + if (dst_nr > 2) + fprintf(out, "bad destination reg U%d\n", dst_nr); + sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat); + break; + default: + sprintf(dstname, "RESERVED"); + break; + } +} + +static char * +i915_get_channel_swizzle(uint32_t select) +{ + switch (select & 0x7) { + case 0: + return (select & 8) ? "-x" : "x"; + case 1: + return (select & 8) ? "-y" : "y"; + case 2: + return (select & 8) ? "-z" : "z"; + case 3: + return (select & 8) ? "-w" : "w"; + case 4: + return (select & 8) ? "-0" : "0"; + case 5: + return (select & 8) ? "-1" : "1"; + default: + return (select & 8) ? "-bad" : "bad"; + } +} + +static void +i915_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name) +{ + switch (src_type) { + case 0: + sprintf(name, "R%d", src_nr); + if (src_nr > 15) + fprintf(out, "bad src reg %s\n", name); + break; + case 1: + if (src_nr < 8) + sprintf(name, "T%d", src_nr); + else if (src_nr == 8) + sprintf(name, "DIFFUSE"); + else if (src_nr == 9) + sprintf(name, "SPECULAR"); + else if (src_nr == 10) + sprintf(name, "FOG"); + else { + fprintf(out, "bad src reg T%d\n", src_nr); + sprintf(name, "RESERVED"); + } + break; + case 2: + sprintf(name, "C%d", src_nr); + if (src_nr > 31) + fprintf(out, "bad src reg %s\n", name); + break; + case 4: + sprintf(name, "oC"); + if (src_nr > 0) + fprintf(out, "bad src reg oC%d\n", src_nr); + break; + case 5: + sprintf(name, "oD"); + if (src_nr > 0) + fprintf(out, "bad src reg oD%d\n", src_nr); + break; + case 6: + sprintf(name, "U%d", src_nr); + if (src_nr > 2) + fprintf(out, "bad src reg %s\n", name); + break; + default: + fprintf(out, "bad src reg type %d\n", src_type); + sprintf(name, "RESERVED"); + break; + } +} + +static void +i915_get_instruction_src0(const uint32_t *data, int i, char *srcname) +{ + uint32_t a0 = data[i]; + uint32_t a1 = data[i + 1]; + int src_nr = (a0 >> 2) & 0x1f; + char *swizzle_x = i915_get_channel_swizzle((a1 >> 28) & 0xf); + char *swizzle_y = i915_get_channel_swizzle((a1 >> 24) & 0xf); + char *swizzle_z = i915_get_channel_swizzle((a1 >> 20) & 0xf); + char *swizzle_w = i915_get_channel_swizzle((a1 >> 16) & 0xf); + char swizzle[100]; + + i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname); + sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); + if (strcmp(swizzle, ".xyzw") != 0) + strcat(srcname, swizzle); +} + +static void +i915_get_instruction_src1(const uint32_t *data, int i, char *srcname) +{ + uint32_t a1 = data[i + 1]; + uint32_t a2 = data[i + 2]; + int src_nr = (a1 >> 8) & 0x1f; + char *swizzle_x = i915_get_channel_swizzle((a1 >> 4) & 0xf); + char *swizzle_y = i915_get_channel_swizzle((a1 >> 0) & 0xf); + char *swizzle_z = i915_get_channel_swizzle((a2 >> 28) & 0xf); + char *swizzle_w = i915_get_channel_swizzle((a2 >> 24) & 0xf); + char swizzle[100]; + + i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname); + sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); + if (strcmp(swizzle, ".xyzw") != 0) + strcat(srcname, swizzle); +} + +static void +i915_get_instruction_src2(const uint32_t *data, int i, char *srcname) +{ + uint32_t a2 = data[i + 2]; + int src_nr = (a2 >> 16) & 0x1f; + char *swizzle_x = i915_get_channel_swizzle((a2 >> 12) & 0xf); + char *swizzle_y = i915_get_channel_swizzle((a2 >> 8) & 0xf); + char *swizzle_z = i915_get_channel_swizzle((a2 >> 4) & 0xf); + char *swizzle_w = i915_get_channel_swizzle((a2 >> 0) & 0xf); + char swizzle[100]; + + i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname); + sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w); + if (strcmp(swizzle, ".xyzw") != 0) + strcat(srcname, swizzle); +} + +static void +i915_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name) +{ + switch (src_type) { + case 0: + sprintf(name, "R%d", src_nr); + if (src_nr > 15) + fprintf(out, "bad src reg %s\n", name); + break; + case 1: + if (src_nr < 8) + sprintf(name, "T%d", src_nr); + else if (src_nr == 8) + sprintf(name, "DIFFUSE"); + else if (src_nr == 9) + sprintf(name, "SPECULAR"); + else if (src_nr == 10) + sprintf(name, "FOG"); + else { + fprintf(out, "bad src reg T%d\n", src_nr); + sprintf(name, "RESERVED"); + } + break; + case 4: + sprintf(name, "oC"); + if (src_nr > 0) + fprintf(out, "bad src reg oC%d\n", src_nr); + break; + case 5: + sprintf(name, "oD"); + if (src_nr > 0) + fprintf(out, "bad src reg oD%d\n", src_nr); + break; + default: + fprintf(out, "bad src reg type %d\n", src_type); + sprintf(name, "RESERVED"); + break; + } +} + +static void +i915_decode_alu1(const uint32_t *data, uint32_t hw_offset, + int i, char *instr_prefix, char *op_name) +{ + char dst[100], src0[100]; + + i915_get_instruction_dst(data, i, dst, 1); + i915_get_instruction_src0(data, i, src0); + + instr_out(data, hw_offset, i++, "%s: %s %s, %s\n", instr_prefix, + op_name, dst, src0); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); +} + +static void +i915_decode_alu2(const uint32_t *data, uint32_t hw_offset, + int i, char *instr_prefix, char *op_name) +{ + char dst[100], src0[100], src1[100]; + + i915_get_instruction_dst(data, i, dst, 1); + i915_get_instruction_src0(data, i, src0); + i915_get_instruction_src1(data, i, src1); + + instr_out(data, hw_offset, i++, "%s: %s %s, %s, %s\n", instr_prefix, + op_name, dst, src0, src1); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); +} + +static void +i915_decode_alu3(const uint32_t *data, uint32_t hw_offset, + int i, char *instr_prefix, char *op_name) +{ + char dst[100], src0[100], src1[100], src2[100]; + + i915_get_instruction_dst(data, i, dst, 1); + i915_get_instruction_src0(data, i, src0); + i915_get_instruction_src1(data, i, src1); + i915_get_instruction_src2(data, i, src2); + + instr_out(data, hw_offset, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix, + op_name, dst, src0, src1, src2); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); +} + +static void +i915_decode_tex(const uint32_t *data, uint32_t hw_offset, int i, char *instr_prefix, + char *tex_name) +{ + uint32_t t0 = data[i]; + uint32_t t1 = data[i + 1]; + char dst_name[100]; + char addr_name[100]; + int sampler_nr; + + i915_get_instruction_dst(data, i, dst_name, 0); + i915_get_instruction_addr((t1 >> 24) & 0x7, + (t1 >> 17) & 0xf, + addr_name); + sampler_nr = t0 & 0xf; + + instr_out(data, hw_offset, i++, "%s: %s %s, S%d, %s\n", instr_prefix, + tex_name, dst_name, sampler_nr, addr_name); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); +} + +static void +i915_decode_dcl(const uint32_t *data, uint32_t hw_offset, int i, char *instr_prefix) +{ + uint32_t d0 = data[i]; + char *sampletype; + int dcl_nr = (d0 >> 14) & 0xf; + char *dcl_x = d0 & (1 << 10) ? "x" : ""; + char *dcl_y = d0 & (1 << 11) ? "y" : ""; + char *dcl_z = d0 & (1 << 12) ? "z" : ""; + char *dcl_w = d0 & (1 << 13) ? "w" : ""; + char dcl_mask[10]; + + switch ((d0 >> 19) & 0x3) { + case 1: + sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w); + if (strcmp(dcl_mask, ".") == 0) + fprintf(out, "bad (empty) dcl mask\n"); + + if (dcl_nr > 10) + fprintf(out, "bad T%d dcl register number\n", dcl_nr); + if (dcl_nr < 8) { + if (strcmp(dcl_mask, ".x") != 0 && + strcmp(dcl_mask, ".xy") != 0 && + strcmp(dcl_mask, ".xz") != 0 && + strcmp(dcl_mask, ".w") != 0 && + strcmp(dcl_mask, ".xyzw") != 0) { + fprintf(out, "bad T%d.%s dcl mask\n", dcl_nr, dcl_mask); + } + instr_out(data, hw_offset, i++, "%s: DCL T%d%s\n", instr_prefix, + dcl_nr, dcl_mask); + } else { + if (strcmp(dcl_mask, ".xz") == 0) + fprintf(out, "errataed bad dcl mask %s\n", dcl_mask); + else if (strcmp(dcl_mask, ".xw") == 0) + fprintf(out, "errataed bad dcl mask %s\n", dcl_mask); + else if (strcmp(dcl_mask, ".xzw") == 0) + fprintf(out, "errataed bad dcl mask %s\n", dcl_mask); + + if (dcl_nr == 8) { + instr_out(data, hw_offset, i++, "%s: DCL DIFFUSE%s\n", instr_prefix, + dcl_mask); + } else if (dcl_nr == 9) { + instr_out(data, hw_offset, i++, "%s: DCL SPECULAR%s\n", instr_prefix, + dcl_mask); + } else if (dcl_nr == 10) { + instr_out(data, hw_offset, i++, "%s: DCL FOG%s\n", instr_prefix, + dcl_mask); + } + } + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + break; + case 3: + switch ((d0 >> 22) & 0x3) { + case 0: + sampletype = "2D"; + break; + case 1: + sampletype = "CUBE"; + break; + case 2: + sampletype = "3D"; + break; + default: + sampletype = "RESERVED"; + break; + } + if (dcl_nr > 15) + fprintf(out, "bad S%d dcl register number\n", dcl_nr); + instr_out(data, hw_offset, i++, "%s: DCL S%d %s\n", instr_prefix, + dcl_nr, sampletype); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + break; + default: + instr_out(data, hw_offset, i++, "%s: DCL RESERVED%d\n", instr_prefix, dcl_nr); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + } +} + +static void +i915_decode_instruction(const uint32_t *data, uint32_t hw_offset, + int i, char *instr_prefix) +{ + switch ((data[i] >> 24) & 0x1f) { + case 0x0: + instr_out(data, hw_offset, i++, "%s: NOP\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + break; + case 0x01: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "ADD"); + break; + case 0x02: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "MOV"); + break; + case 0x03: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "MUL"); + break; + case 0x04: + i915_decode_alu3(data, hw_offset, i, instr_prefix, "MAD"); + break; + case 0x05: + i915_decode_alu3(data, hw_offset, i, instr_prefix, "DP2ADD"); + break; + case 0x06: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "DP3"); + break; + case 0x07: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "DP4"); + break; + case 0x08: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "FRC"); + break; + case 0x09: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "RCP"); + break; + case 0x0a: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "RSQ"); + break; + case 0x0b: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "EXP"); + break; + case 0x0c: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "LOG"); + break; + case 0x0d: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "CMP"); + break; + case 0x0e: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "MIN"); + break; + case 0x0f: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "MAX"); + break; + case 0x10: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "FLR"); + break; + case 0x11: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "MOD"); + break; + case 0x12: + i915_decode_alu1(data, hw_offset, i, instr_prefix, "TRC"); + break; + case 0x13: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "SGE"); + break; + case 0x14: + i915_decode_alu2(data, hw_offset, i, instr_prefix, "SLT"); + break; + case 0x15: + i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLD"); + break; + case 0x16: + i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLDP"); + break; + case 0x17: + i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLDB"); + break; + case 0x19: + i915_decode_dcl(data, hw_offset, i, instr_prefix); + break; + default: + instr_out(data, hw_offset, i++, "%s: unknown\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + instr_out(data, hw_offset, i++, "%s\n", instr_prefix); + break; + } +} + +static int +decode_3d_1d(const uint32_t *data, int count, uint32_t hw_offset, int *failures, int i830) +{ + unsigned int len, i, c, opcode, word, map, sampler, instr; + char *format; + + struct { + uint32_t opcode; + int i830_only; + int min_len; + int max_len; + char *name; + } opcodes_3d_1d[] = { + { 0x8e, 0, 3, 3, "3DSTATE_BUFFER_INFO" }, + { 0x86, 0, 4, 4, "3DSTATE_CHROMA_KEY" }, + { 0x9c, 0, 1, 1, "3DSTATE_CLEAR_PARAMETERS" }, + { 0x88, 0, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" }, + { 0x99, 0, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" }, + { 0x9a, 0, 2, 2, "3DSTATE_DEFAULT_SPECULAR" }, + { 0x98, 0, 2, 2, "3DSTATE_DEFAULT_Z" }, + { 0x97, 0, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" }, + { 0x85, 0, 2, 2, "3DSTATE_DEST_BUFFER_VARIABLES" }, + { 0x80, 0, 5, 5, "3DSTATE_DRAWING_RECTANGLE" }, + { 0x8e, 0, 3, 3, "3DSTATE_BUFFER_INFO" }, + { 0x9d, 0, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" }, + { 0x9e, 0, 4, 4, "3DSTATE_MONO_FILTER" }, + { 0x89, 0, 4, 4, "3DSTATE_FOG_MODE" }, + { 0x8f, 0, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" }, + { 0x81, 0, 3, 3, "3DSTATE_SCISSOR_RECTANGLE" }, + { 0x83, 0, 2, 2, "3DSTATE_SPAN_STIPPLE" }, + { 0x8c, 1, 2, 2, "3DSTATE_MAP_COORD_TRANSFORM_I830" }, + { 0x8b, 1, 2, 2, "3DSTATE_MAP_VERTEX_TRANSFORM_I830" }, + { 0x8d, 1, 3, 3, "3DSTATE_W_STATE_I830" }, + { 0x01, 1, 2, 2, "3DSTATE_COLOR_FACTOR_I830" }, + { 0x02, 1, 2, 2, "3DSTATE_MAP_COORD_SETBIND_I830" }, + }; + + switch ((data[0] & 0x00ff0000) >> 16) { + case 0x07: + /* This instruction is unusual. A 0 length means just 1 DWORD instead of + * 2. The 0 length is specified in one place to be unsupported, but + * stated to be required in another, and 0 length LOAD_INDIRECTs appear + * to cause no harm at least. + */ + instr_out(data, hw_offset, 0, "3DSTATE_LOAD_INDIRECT\n"); + len = (data[0] & 0x000000ff) + 1; + i = 1; + if (data[0] & (0x01 << 8)) { + if (i + 2 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "SIS.0\n"); + instr_out(data, hw_offset, i++, "SIS.1\n"); + } + if (data[0] & (0x02 << 8)) { + if (i + 1 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "DIS.0\n"); + } + if (data[0] & (0x04 << 8)) { + if (i + 2 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "SSB.0\n"); + instr_out(data, hw_offset, i++, "SSB.1\n"); + } + if (data[0] & (0x08 << 8)) { + if (i + 2 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "MSB.0\n"); + instr_out(data, hw_offset, i++, "MSB.1\n"); + } + if (data[0] & (0x10 << 8)) { + if (i + 2 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "PSP.0\n"); + instr_out(data, hw_offset, i++, "PSP.1\n"); + } + if (data[0] & (0x20 << 8)) { + if (i + 2 >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_INDIRECT"); + instr_out(data, hw_offset, i++, "PSC.0\n"); + instr_out(data, hw_offset, i++, "PSC.1\n"); + } + if (len != i) { + fprintf(out, "Bad count in 3DSTATE_LOAD_INDIRECT\n"); + (*failures)++; + return len; + } + return len; + case 0x04: + instr_out(data, hw_offset, 0, "3DSTATE_LOAD_STATE_IMMEDIATE_1\n"); + len = (data[0] & 0x0000000f) + 2; + i = 1; + for (word = 0; word <= 7; word++) { + if (data[0] & (1 << (4 + word))) { + if (i >= count) + BUFFER_FAIL(count, len, "3DSTATE_LOAD_STATE_IMMEDIATE_1"); + + /* save vertex state for decode */ + if (word == 2) { + saved_s2_set = 1; + saved_s2 = data[i]; + } + if (word == 4) { + saved_s4_set = 1; + saved_s4 = data[i]; + } + + instr_out(data, hw_offset, i++, "S%d\n", word); + } + } + if (len != i) { + fprintf(out, "Bad count in 3DSTATE_LOAD_INDIRECT\n"); + (*failures)++; + } + return len; + case 0x00: + instr_out(data, hw_offset, 0, "3DSTATE_MAP_STATE\n"); + len = (data[0] & 0x0000003f) + 2; + instr_out(data, hw_offset, 1, "mask\n"); + + i = 2; + for (map = 0; map <= 15; map++) { + if (data[1] & (1 << map)) { + if (i + 3 >= count) + BUFFER_FAIL(count, len, "3DSTATE_MAP_STATE"); + instr_out(data, hw_offset, i++, "map %d MS2\n", map); + instr_out(data, hw_offset, i++, "map %d MS3\n", map); + instr_out(data, hw_offset, i++, "map %d MS4\n", map); + } + } + if (len != i) { + fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n"); + (*failures)++; + return len; + } + return len; + case 0x06: + instr_out(data, hw_offset, 0, "3DSTATE_PIXEL_SHADER_CONSTANTS\n"); + len = (data[0] & 0x000000ff) + 2; + + i = 2; + for (c = 0; c <= 31; c++) { + if (data[1] & (1 << c)) { + if (i + 4 >= count) + BUFFER_FAIL(count, len, "3DSTATE_PIXEL_SHADER_CONSTANTS"); + instr_out(data, hw_offset, i, "C%d.X = %f\n", + c, int_as_float(data[i])); + i++; + instr_out(data, hw_offset, i, "C%d.Y = %f\n", + c, int_as_float(data[i])); + i++; + instr_out(data, hw_offset, i, "C%d.Z = %f\n", + c, int_as_float(data[i])); + i++; + instr_out(data, hw_offset, i, "C%d.W = %f\n", + c, int_as_float(data[i])); + i++; + } + } + if (len != i) { + fprintf(out, "Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n"); + (*failures)++; + } + return len; + case 0x05: + instr_out(data, hw_offset, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n"); + len = (data[0] & 0x000000ff) + 2; + if ((len - 1) % 3 != 0 || len > 370) { + fprintf(out, "Bad count in 3DSTATE_PIXEL_SHADER_PROGRAM\n"); + (*failures)++; + } + i = 1; + for (instr = 0; instr < (len - 1) / 3; instr++) { + char instr_prefix[10]; + + if (i + 3 >= count) + BUFFER_FAIL(count, len, "3DSTATE_PIXEL_SHADER_PROGRAM"); + sprintf(instr_prefix, "PS%03d", instr); + i915_decode_instruction(data, hw_offset, i, instr_prefix); + i += 3; + } + return len; + case 0x01: + if (i830) + break; + instr_out(data, hw_offset, 0, "3DSTATE_SAMPLER_STATE\n"); + instr_out(data, hw_offset, 1, "mask\n"); + len = (data[0] & 0x0000003f) + 2; + i = 2; + for (sampler = 0; sampler <= 15; sampler++) { + if (data[1] & (1 << sampler)) { + if (i + 3 >= count) + BUFFER_FAIL(count, len, "3DSTATE_SAMPLER_STATE"); + instr_out(data, hw_offset, i++, "sampler %d SS2\n", + sampler); + instr_out(data, hw_offset, i++, "sampler %d SS3\n", + sampler); + instr_out(data, hw_offset, i++, "sampler %d SS4\n", + sampler); + } + } + if (len != i) { + fprintf(out, "Bad count in 3DSTATE_SAMPLER_STATE\n"); + (*failures)++; + } + return len; + case 0x85: + len = (data[0] & 0x0000000f) + 2; + + if (len != 2) + fprintf(out, "Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n"); + if (count < 2) + BUFFER_FAIL(count, len, "3DSTATE_DEST_BUFFER_VARIABLES"); + + instr_out(data, hw_offset, 0, + "3DSTATE_DEST_BUFFER_VARIABLES\n"); + + switch ((data[1] >> 8) & 0xf) { + case 0x0: format = "g8"; break; + case 0x1: format = "x1r5g5b5"; break; + case 0x2: format = "r5g6b5"; break; + case 0x3: format = "a8r8g8b8"; break; + case 0x4: format = "ycrcb_swapy"; break; + case 0x5: format = "ycrcb_normal"; break; + case 0x6: format = "ycrcb_swapuv"; break; + case 0x7: format = "ycrcb_swapuvy"; break; + case 0x8: format = "a4r4g4b4"; break; + case 0x9: format = "a1r5g5b5"; break; + case 0xa: format = "a2r10g10b10"; break; + default: format = "BAD"; break; + } + instr_out(data, hw_offset, 1, "%s format, early Z %sabled\n", + format, + (data[1] & (1 << 31)) ? "en" : "dis"); + return len; + } + + for (opcode = 0; opcode < sizeof(opcodes_3d_1d) / sizeof(opcodes_3d_1d[0]); + opcode++) + { + if (opcodes_3d_1d[opcode].i830_only && !i830) + continue; + + if (((data[0] & 0x00ff0000) >> 16) == opcodes_3d_1d[opcode].opcode) { + len = 1; + + instr_out(data, hw_offset, 0, "%s\n", opcodes_3d_1d[opcode].name); + if (opcodes_3d_1d[opcode].max_len > 1) { + len = (data[0] & 0x0000ffff) + 2; + if (len < opcodes_3d_1d[opcode].min_len || + len > opcodes_3d_1d[opcode].max_len) + { + fprintf(out, "Bad count in %s\n", + opcodes_3d_1d[opcode].name); + (*failures)++; + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_3d_1d[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + + return len; + } + } + + instr_out(data, hw_offset, 0, "3D UNKNOWN\n"); + (*failures)++; + return 1; +} + +static int +decode_3d_primitive(const uint32_t *data, int count, uint32_t hw_offset, + int *failures) +{ + char immediate = (data[0] & (1 << 23)) == 0; + unsigned int len, i; + char *primtype; + + switch ((data[0] >> 18) & 0xf) { + case 0x0: primtype = "TRILIST"; break; + case 0x1: primtype = "TRISTRIP"; break; + case 0x2: primtype = "TRISTRIP_REVERSE"; break; + case 0x3: primtype = "TRIFAN"; break; + case 0x4: primtype = "POLYGON"; break; + case 0x5: primtype = "LINELIST"; break; + case 0x6: primtype = "LINESTRIP"; break; + case 0x7: primtype = "RECTLIST"; break; + case 0x8: primtype = "POINTLIST"; break; + case 0x9: primtype = "DIB"; break; + case 0xa: primtype = "CLEAR_RECT"; break; + default: primtype = "unknown"; break; + } + + /* XXX: 3DPRIM_DIB not supported */ + if (immediate) { + len = (data[0] & 0x0003ffff) + 2; + instr_out(data, hw_offset, 0, "3DPRIMITIVE inline %s\n", primtype); + if (count < len) + BUFFER_FAIL(count, len, "3DPRIMITIVE inline"); + if (!saved_s2_set || !saved_s4_set) { + fprintf(out, "unknown vertex format\n"); + for (i = 1; i < len; i++) { + instr_out(data, hw_offset, i, + " vertex data (%f float)\n", + int_as_float(data[i])); + } + } else { + unsigned int vertex = 0; + for (i = 1; i < len;) { + unsigned int tc; + +#define VERTEX_OUT(fmt, ...) do { \ + if (i < len) \ + instr_out(data, hw_offset, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \ + else \ + fprintf(out, " missing data in V%d\n", vertex); \ + i++; \ +} while (0) + + VERTEX_OUT("X = %f", int_as_float(data[i])); + VERTEX_OUT("Y = %f", int_as_float(data[i])); + switch (saved_s4 >> 6 & 0x7) { + case 0x1: + VERTEX_OUT("Z = %f", int_as_float(data[i])); + break; + case 0x2: + VERTEX_OUT("Z = %f", int_as_float(data[i])); + VERTEX_OUT("W = %f", int_as_float(data[i])); + break; + case 0x3: + break; + case 0x4: + VERTEX_OUT("W = %f", int_as_float(data[i])); + break; + default: + fprintf(out, "bad S4 position mask\n"); + } + + if (saved_s4 & (1 << 10)) { + VERTEX_OUT("color = (A=0x%02x, R=0x%02x, G=0x%02x, " + "B=0x%02x)", + data[i] >> 24, + (data[i] >> 16) & 0xff, + (data[i] >> 8) & 0xff, + data[i] & 0xff); + } + if (saved_s4 & (1 << 11)) { + VERTEX_OUT("spec = (A=0x%02x, R=0x%02x, G=0x%02x, " + "B=0x%02x)", + data[i] >> 24, + (data[i] >> 16) & 0xff, + (data[i] >> 8) & 0xff, + data[i] & 0xff); + } + if (saved_s4 & (1 << 12)) + VERTEX_OUT("width = 0x%08x)", data[i]); + + for (tc = 0; tc <= 7; tc++) { + switch ((saved_s2 >> (tc * 4)) & 0xf) { + case 0x0: + VERTEX_OUT("T%d.X = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.Y = %f", tc, int_as_float(data[i])); + break; + case 0x1: + VERTEX_OUT("T%d.X = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.Y = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.Z = %f", tc, int_as_float(data[i])); + break; + case 0x2: + VERTEX_OUT("T%d.X = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.Y = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.Z = %f", tc, int_as_float(data[i])); + VERTEX_OUT("T%d.W = %f", tc, int_as_float(data[i])); + break; + case 0x3: + VERTEX_OUT("T%d.X = %f", tc, int_as_float(data[i])); + break; + case 0x4: + VERTEX_OUT("T%d.XY = 0x%08x half-float", tc, data[i]); + break; + case 0x5: + VERTEX_OUT("T%d.XY = 0x%08x half-float", tc, data[i]); + VERTEX_OUT("T%d.ZW = 0x%08x half-float", tc, data[i]); + break; + case 0xf: + break; + default: + fprintf(out, "bad S2.T%d format\n", tc); + } + } + vertex++; + } + } + } else { + /* indirect vertices */ + len = data[0] & 0x0000ffff; /* index count */ + if (data[0] & (1 << 17)) { + /* random vertex access */ + if (count < (len + 1) / 2 + 1) { + BUFFER_FAIL(count, (len + 1) / 2 + 1, + "3DPRIMITIVE random indirect"); + } + instr_out(data, hw_offset, 0, + "3DPRIMITIVE random indirect %s (%d)\n", primtype, len); + if (len == 0) { + /* vertex indices continue until 0xffff is found */ + for (i = 1; i < count; i++) { + if ((data[i] & 0xffff) == 0xffff) { + instr_out(data, hw_offset, i, + " indices: (terminator)\n"); + return i; + } else if ((data[i] >> 16) == 0xffff) { + instr_out(data, hw_offset, i, + " indices: 0x%04x, " + "(terminator)\n", + data[i] & 0xffff); + return i; + } else { + instr_out(data, hw_offset, i, + " indices: 0x%04x, 0x%04x\n", + data[i] & 0xffff, data[i] >> 16); + } + } + fprintf(out, + "3DPRIMITIVE: no terminator found in index buffer\n"); + (*failures)++; + return count; + } else { + /* fixed size vertex index buffer */ + for (i = 0; i < len; i += 2) { + if (i * 2 == len - 1) { + instr_out(data, hw_offset, i, + " indices: 0x%04x\n", + data[i] & 0xffff); + } else { + instr_out(data, hw_offset, i, + " indices: 0x%04x, 0x%04x\n", + data[i] & 0xffff, data[i] >> 16); + } + } + } + return (len + 1) / 2 + 1; + } else { + /* sequential vertex access */ + if (count < 2) + BUFFER_FAIL(count, 2, "3DPRIMITIVE seq indirect"); + instr_out(data, hw_offset, 0, + "3DPRIMITIVE sequential indirect %s, %d starting from " + "%d\n", primtype, len, data[1] & 0xffff); + instr_out(data, hw_offset, 1, " start\n"); + return 2; + } + } + + return len; +} + +static int +decode_3d(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + unsigned int opcode; + + struct { + uint32_t opcode; + int min_len; + int max_len; + char *name; + } opcodes_3d[] = { + { 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" }, + { 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" }, + { 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" }, + { 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" }, + { 0x15, 1, 1, "3DSTATE_FOG_COLOR" }, + { 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" }, + { 0x0d, 1, 1, "3DSTATE_MODES_4" }, + { 0x0c, 1, 1, "3DSTATE_MODES_5" }, + { 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" }, + }; + + switch ((data[0] & 0x1f000000) >> 24) { + case 0x1f: + return decode_3d_primitive(data, count, hw_offset, failures); + case 0x1d: + return decode_3d_1d(data, count, hw_offset, failures, 0); + case 0x1c: + return decode_3d_1c(data, count, hw_offset, failures); + } + + for (opcode = 0; opcode < sizeof(opcodes_3d) / sizeof(opcodes_3d[0]); + opcode++) { + if ((data[0] & 0x1f000000) >> 24 == opcodes_3d[opcode].opcode) { + unsigned int len = 1, i; + + instr_out(data, hw_offset, 0, "%s\n", opcodes_3d[opcode].name); + if (opcodes_3d[opcode].max_len > 1) { + len = (data[0] & 0xff) + 2; + if (len < opcodes_3d[opcode].min_len || + len > opcodes_3d[opcode].max_len) + { + fprintf(out, "Bad count in %s\n", opcodes_3d[opcode].name); + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_3d[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + return len; + } + } + + instr_out(data, hw_offset, 0, "3D UNKNOWN\n"); + (*failures)++; + return 1; +} + +static const char * +get_965_surfacetype(unsigned int surfacetype) +{ + switch (surfacetype) { + case 0: return "1D"; + case 1: return "2D"; + case 2: return "3D"; + case 3: return "CUBE"; + case 4: return "BUFFER"; + case 7: return "NULL"; + default: return "unknown"; + } +} + +static const char * +get_965_depthformat(unsigned int depthformat) +{ + switch (depthformat) { + case 0: return "s8_z24float"; + case 1: return "z32float"; + case 2: return "z24s8"; + case 5: return "z16"; + default: return "unknown"; + } +} + +static const char * +get_965_element_component(uint32_t data, int component) +{ + uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7; + + switch (component_control) { + case 0: + return "nostore"; + case 1: + switch (component) { + case 0: return "X"; + case 1: return "Y"; + case 2: return "Z"; + case 3: return "W"; + default: return "fail"; + } + case 2: + return "0.0"; + case 3: + return "1.0"; + case 4: + return "0x1"; + case 5: + return "VID"; + default: + return "fail"; + } +} + +static const char * +get_965_prim_type(uint32_t data) +{ + uint32_t primtype = (data >> 10) & 0x1f; + + switch (primtype) { + case 0x01: return "point list"; + case 0x02: return "line list"; + case 0x03: return "line strip"; + case 0x04: return "tri list"; + case 0x05: return "tri strip"; + case 0x06: return "tri fan"; + case 0x07: return "quad list"; + case 0x08: return "quad strip"; + case 0x09: return "line list adj"; + case 0x0a: return "line strip adj"; + case 0x0b: return "tri list adj"; + case 0x0c: return "tri strip adj"; + case 0x0d: return "tri strip reverse"; + case 0x0e: return "polygon"; + case 0x0f: return "rect list"; + case 0x10: return "line loop"; + case 0x11: return "point list bf"; + case 0x12: return "line strip cont"; + case 0x13: return "line strip bf"; + case 0x14: return "line strip cont bf"; + case 0x15: return "tri fan no stipple"; + default: return "fail"; + } +} + +static int +decode_3d_965(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + unsigned int opcode, len; + int i; + + struct { + uint32_t opcode; + int min_len; + int max_len; + char *name; + } opcodes_3d[] = { + { 0x6000, 3, 3, "URB_FENCE" }, + { 0x6001, 2, 2, "CS_URB_STATE" }, + { 0x6002, 2, 2, "CONSTANT_BUFFER" }, + { 0x6101, 6, 6, "STATE_BASE_ADDRESS" }, + { 0x6102, 2, 2 , "STATE_SIP" }, + { 0x6104, 1, 1, "3DSTATE_PIPELINE_SELECT" }, + { 0x680b, 1, 1, "3DSTATE_VF_STATISTICS" }, + { 0x6904, 1, 1, "3DSTATE_PIPELINE_SELECT" }, + { 0x7800, 7, 7, "3DSTATE_PIPELINED_POINTERS" }, + { 0x7801, 6, 6, "3DSTATE_BINDING_TABLE_POINTERS" }, + { 0x780b, 1, 1, "3DSTATE_VF_STATISTICS" }, + { 0x7808, 5, 257, "3DSTATE_VERTEX_BUFFERS" }, + { 0x7809, 3, 256, "3DSTATE_VERTEX_ELEMENTS" }, + { 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" }, + { 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" }, + { 0x7901, 5, 5, "3DSTATE_CONSTANT_COLOR" }, + { 0x7905, 5, 7, "3DSTATE_DEPTH_BUFFER" }, + { 0x7906, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" }, + { 0x7907, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" }, + { 0x7908, 3, 3, "3DSTATE_LINE_STIPPLE" }, + { 0x7909, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" }, + { 0x790a, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" }, + { 0x7b00, 6, 6, "3DPRIMITIVE" }, + }; + + len = (data[0] & 0x0000ffff) + 2; + + switch ((data[0] & 0xffff0000) >> 16) { + case 0x6101: + if (len != 6) + fprintf(out, "Bad count in STATE_BASE_ADDRESS\n"); + if (count < 6) + BUFFER_FAIL(count, len, "STATE_BASE_ADDRESS"); + + instr_out(data, hw_offset, 0, + "STATE_BASE_ADDRESS\n"); + + if (data[1] & 1) { + instr_out(data, hw_offset, 1, "General state at 0x%08x\n", + data[1] & ~1); + } else + instr_out(data, hw_offset, 1, "General state not updated\n"); + + if (data[2] & 1) { + instr_out(data, hw_offset, 2, "Surface state at 0x%08x\n", + data[2] & ~1); + } else + instr_out(data, hw_offset, 2, "Surface state not updated\n"); + + if (data[3] & 1) { + instr_out(data, hw_offset, 3, "Indirect state at 0x%08x\n", + data[3] & ~1); + } else + instr_out(data, hw_offset, 3, "Indirect state not updated\n"); + + if (data[4] & 1) { + instr_out(data, hw_offset, 4, "General state upper bound 0x%08x\n", + data[4] & ~1); + } else + instr_out(data, hw_offset, 4, "General state not updated\n"); + + if (data[5] & 1) { + instr_out(data, hw_offset, 5, "Indirect state upper bound 0x%08x\n", + data[5] & ~1); + } else + instr_out(data, hw_offset, 5, "Indirect state not updated\n"); + + return len; + case 0x7800: + if (len != 7) + fprintf(out, "Bad count in 3DSTATE_PIPELINED_POINTERS\n"); + if (count < 7) + BUFFER_FAIL(count, len, "3DSTATE_PIPELINED_POINTERS"); + + instr_out(data, hw_offset, 0, + "3DSTATE_PIPELINED_POINTERS\n"); + instr_out(data, hw_offset, 1, "VS state\n"); + instr_out(data, hw_offset, 2, "GS state\n"); + instr_out(data, hw_offset, 3, "Clip state\n"); + instr_out(data, hw_offset, 4, "SF state\n"); + instr_out(data, hw_offset, 5, "WM state\n"); + instr_out(data, hw_offset, 6, "CC state\n"); + return len; + case 0x7801: + if (len != 6) + fprintf(out, "Bad count in 3DSTATE_BINDING_TABLE_POINTERS\n"); + if (count < 6) + BUFFER_FAIL(count, len, "3DSTATE_BINDING_TABLE_POINTERS"); + + instr_out(data, hw_offset, 0, + "3DSTATE_BINDING_TABLE_POINTERS\n"); + instr_out(data, hw_offset, 1, "VS binding table\n"); + instr_out(data, hw_offset, 2, "GS binding table\n"); + instr_out(data, hw_offset, 3, "Clip binding table\n"); + instr_out(data, hw_offset, 4, "SF binding table\n"); + instr_out(data, hw_offset, 5, "WM binding table\n"); + + return len; + + case 0x7808: + len = (data[0] & 0xff) + 2; + if ((len - 1) % 4 != 0) + fprintf(out, "Bad count in 3DSTATE_VERTEX_BUFFERS\n"); + if (count < len) + BUFFER_FAIL(count, len, "3DSTATE_VERTEX_BUFFERS"); + instr_out(data, hw_offset, 0, "3DSTATE_VERTEX_BUFFERS\n"); + + for (i = 1; i < len;) { + instr_out(data, hw_offset, i, "buffer %d: %s, pitch %db\n", + data[i] >> 27, + data[i] & (1 << 26) ? "random" : "sequential", + data[i] & 0x07ff); + i++; + instr_out(data, hw_offset, i++, "buffer address\n"); + instr_out(data, hw_offset, i++, "max index\n"); + instr_out(data, hw_offset, i++, "mbz\n"); + } + return len; + + case 0x7809: + len = (data[0] & 0xff) + 2; + if ((len + 1) % 2 != 0) + fprintf(out, "Bad count in 3DSTATE_VERTEX_ELEMENTS\n"); + if (count < len) + BUFFER_FAIL(count, len, "3DSTATE_VERTEX_ELEMENTS"); + instr_out(data, hw_offset, 0, "3DSTATE_VERTEX_ELEMENTS\n"); + + for (i = 1; i < len;) { + instr_out(data, hw_offset, i, "buffer %d: %svalid, type 0x%04x, " + "src offset 0x%04x bytes\n", + data[i] >> 27, + data[i] & (1 << 26) ? "" : "in", + (data[i] >> 16) & 0x1ff, + data[i] & 0x07ff); + i++; + instr_out(data, hw_offset, i, "(%s, %s, %s, %s), " + "dst offset 0x%02x bytes\n", + get_965_element_component(data[i], 0), + get_965_element_component(data[i], 1), + get_965_element_component(data[i], 2), + get_965_element_component(data[i], 3), + (data[i] & 0xff) * 4); + i++; + } + return len; + + case 0x780a: + len = (data[0] & 0xff) + 2; + if (len != 3) + fprintf(out, "Bad count in 3DSTATE_INDEX_BUFFER\n"); + if (count < len) + BUFFER_FAIL(count, len, "3DSTATE_INDEX_BUFFER"); + instr_out(data, hw_offset, 0, "3DSTATE_INDEX_BUFFER\n"); + instr_out(data, hw_offset, 1, "beginning buffer address\n"); + instr_out(data, hw_offset, 2, "ending buffer address\n"); + return len; + + case 0x7900: + if (len != 4) + fprintf(out, "Bad count in 3DSTATE_DRAWING_RECTANGLE\n"); + if (count < 4) + BUFFER_FAIL(count, len, "3DSTATE_DRAWING_RECTANGLE"); + + instr_out(data, hw_offset, 0, + "3DSTATE_DRAWING_RECTANGLE\n"); + instr_out(data, hw_offset, 1, "top left: %d,%d\n", + data[1] & 0xffff, + (data[1] >> 16) & 0xffff); + instr_out(data, hw_offset, 2, "bottom right: %d,%d\n", + data[2] & 0xffff, + (data[2] >> 16) & 0xffff); + instr_out(data, hw_offset, 3, "origin: %d,%d\n", + (int)data[3] & 0xffff, + ((int)data[3] >> 16) & 0xffff); + + return len; + + case 0x7905: + if (len != 5 && len != 6) + fprintf(out, "Bad count in 3DSTATE_DEPTH_BUFFER\n"); + if (count < len) + BUFFER_FAIL(count, len, "3DSTATE_DEPTH_BUFFER"); + + instr_out(data, hw_offset, 0, + "3DSTATE_DEPTH_BUFFER\n"); + instr_out(data, hw_offset, 1, "%s, %s, pitch = %d bytes, %stiled\n", + get_965_surfacetype(data[1] >> 29), + get_965_depthformat((data[1] >> 18) & 0x7), + (data[1] & 0x0001ffff) + 1, + data[1] & (1 << 27) ? "" : "not "); + instr_out(data, hw_offset, 2, "depth offset\n"); + instr_out(data, hw_offset, 3, "%dx%d\n", + ((data[3] & 0x0007ffc0) >> 6) + 1, + ((data[3] & 0xfff80000) >> 19) + 1); + instr_out(data, hw_offset, 4, "volume depth\n"); + if (len == 6) + instr_out(data, hw_offset, 5, "\n"); + + return len; + + case 0x7b00: + len = (data[0] & 0xff) + 2; + if (len != 6) + fprintf(out, "Bad count in 3DPRIMITIVE\n"); + if (count < len) + BUFFER_FAIL(count, len, "3DPRIMITIVE"); + + instr_out(data, hw_offset, 0, + "3DPRIMITIVE: %s %s\n", + get_965_prim_type(data[0]), + (data[0] & (1 << 15)) ? "random" : "sequential"); + instr_out(data, hw_offset, 1, "vertex count\n"); + instr_out(data, hw_offset, 2, "start vertex\n"); + instr_out(data, hw_offset, 3, "instance count\n"); + instr_out(data, hw_offset, 4, "start instance\n"); + instr_out(data, hw_offset, 5, "index bias\n"); + return len; + } + + for (opcode = 0; opcode < sizeof(opcodes_3d) / sizeof(opcodes_3d[0]); + opcode++) { + if ((data[0] & 0xffff0000) >> 16 == opcodes_3d[opcode].opcode) { + unsigned int i; + len = 1; + + instr_out(data, hw_offset, 0, "%s\n", opcodes_3d[opcode].name); + if (opcodes_3d[opcode].max_len > 1) { + len = (data[0] & 0xff) + 2; + if (len < opcodes_3d[opcode].min_len || + len > opcodes_3d[opcode].max_len) + { + fprintf(out, "Bad count in %s\n", opcodes_3d[opcode].name); + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_3d[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + return len; + } + } + + instr_out(data, hw_offset, 0, "3D UNKNOWN\n"); + (*failures)++; + return 1; +} + +static int +decode_3d_i830(const uint32_t *data, int count, uint32_t hw_offset, int *failures) +{ + unsigned int opcode; + + struct { + uint32_t opcode; + int min_len; + int max_len; + char *name; + } opcodes_3d[] = { + { 0x02, 1, 1, "3DSTATE_MODES_3" }, + { 0x03, 1, 1, "3DSTATE_ENABLES_1"}, + { 0x04, 1, 1, "3DSTATE_ENABLES_2"}, + { 0x05, 1, 1, "3DSTATE_VFT0"}, + { 0x06, 1, 1, "3DSTATE_AA"}, + { 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" }, + { 0x08, 1, 1, "3DSTATE_MODES_1" }, + { 0x09, 1, 1, "3DSTATE_STENCIL_TEST" }, + { 0x0a, 1, 1, "3DSTATE_VFT1"}, + { 0x0b, 1, 1, "3DSTATE_INDPT_ALPHA_BLEND" }, + { 0x0c, 1, 1, "3DSTATE_MODES_5" }, + { 0x0d, 1, 1, "3DSTATE_MAP_BLEND_OP" }, + { 0x0e, 1, 1, "3DSTATE_MAP_BLEND_ARG" }, + { 0x0f, 1, 1, "3DSTATE_MODES_2" }, + { 0x15, 1, 1, "3DSTATE_FOG_COLOR" }, + { 0x16, 1, 1, "3DSTATE_MODES_4" }, + }; + + switch ((data[0] & 0x1f000000) >> 24) { + case 0x1f: + return decode_3d_primitive(data, count, hw_offset, failures); + case 0x1d: + return decode_3d_1d(data, count, hw_offset, failures, 1); + case 0x1c: + return decode_3d_1c(data, count, hw_offset, failures); + } + + for (opcode = 0; opcode < sizeof(opcodes_3d) / sizeof(opcodes_3d[0]); + opcode++) { + if ((data[0] & 0x1f000000) >> 24 == opcodes_3d[opcode].opcode) { + unsigned int len = 1, i; + + instr_out(data, hw_offset, 0, "%s\n", opcodes_3d[opcode].name); + if (opcodes_3d[opcode].max_len > 1) { + len = (data[0] & 0xff) + 2; + if (len < opcodes_3d[opcode].min_len || + len > opcodes_3d[opcode].max_len) + { + fprintf(out, "Bad count in %s\n", opcodes_3d[opcode].name); + } + } + + for (i = 1; i < len; i++) { + if (i >= count) + BUFFER_FAIL(count, len, opcodes_3d[opcode].name); + instr_out(data, hw_offset, i, "dword %d\n", i); + } + return len; + } + } + + instr_out(data, hw_offset, 0, "3D UNKNOWN\n"); + (*failures)++; + return 1; +} + +/** + * Decodes an i830-i915 batch buffer, writing the output to stdout. + * + * \param data batch buffer contents + * \param count number of DWORDs to decode in the batch buffer + * \param hw_offset hardware address for the buffer + */ +int +intel_decode(const uint32_t *data, int count, uint32_t hw_offset, uint32_t devid) +{ + int index = 0; + int failures = 0; + + out = stderr; + + while (index < count) { + switch ((data[index] & 0xe0000000) >> 29) { + case 0x0: + index += decode_mi(data + index, count - index, + hw_offset + index * 4, &failures); + break; + case 0x2: + index += decode_2d(data + index, count - index, + hw_offset + index * 4, &failures); + break; + case 0x3: + if (IS_965(devid)) { + index += decode_3d_965(data + index, count - index, + hw_offset + index * 4, &failures); + } else if (IS_9XX(devid)) { + index += decode_3d(data + index, count - index, + hw_offset + index * 4, &failures); + } else { + index += decode_3d_i830(data + index, count - index, + hw_offset + index * 4, &failures); + } + break; + default: + instr_out(data, hw_offset, index, "UNKNOWN\n"); + failures++; + index++; + break; + } + fflush(out); + } + + return failures; +} + +void intel_decode_context_reset(void) +{ + saved_s2_set = 0; + saved_s4_set = 1; +} + diff --git a/src/gallium/drivers/i965/intel_decode.h b/src/gallium/drivers/i965/intel_decode.h new file mode 100644 index 00000000000..7683097b869 --- /dev/null +++ b/src/gallium/drivers/i965/intel_decode.h @@ -0,0 +1,29 @@ +/* + * Copyright © 2007 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt <[email protected]> + * + */ + +int intel_decode(const uint32_t *data, int count, uint32_t hw_offset, uint32_t devid); +void intel_decode_context_reset(void); diff --git a/src/gallium/drivers/i965/intel_structs.h b/src/gallium/drivers/i965/intel_structs.h new file mode 100644 index 00000000000..522e3bd92c2 --- /dev/null +++ b/src/gallium/drivers/i965/intel_structs.h @@ -0,0 +1,132 @@ +#ifndef INTEL_STRUCTS_H +#define INTEL_STRUCTS_H + +struct br0 { + GLuint length:8; + GLuint pad0:3; + GLuint dst_tiled:1; + GLuint pad1:8; + GLuint write_rgb:1; + GLuint write_alpha:1; + GLuint opcode:7; + GLuint client:3; +}; + + +struct br13 { + GLint dest_pitch:16; + GLuint rop:8; + GLuint color_depth:2; + GLuint pad1:3; + GLuint mono_source_transparency:1; + GLuint clipping_enable:1; + GLuint pad0:1; +}; + + + +/* This is an attempt to move some of the 2D interaction in this + * driver to using structs for packets rather than a bunch of #defines + * and dwords. + */ +struct xy_color_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint color; +}; + +struct xy_src_copy_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + + struct { + GLuint src_x1:16; + GLuint src_y1:16; + } dw5; + + struct { + GLint src_pitch:16; + GLuint pad:16; + } dw6; + + GLuint src_base_addr; +}; + +struct xy_setup_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint clip_x1:16; + GLuint clip_y1:16; + } dw2; + + struct { + GLuint clip_x2:16; + GLuint clip_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint background_color; + GLuint foreground_color; + GLuint pattern_base_addr; +}; + + +struct xy_text_immediate_blit { + struct { + GLuint length:8; + GLuint pad2:3; + GLuint dst_tiled:1; + GLuint pad1:4; + GLuint byte_packed:1; + GLuint pad0:5; + GLuint opcode:7; + GLuint client:3; + } dw0; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw1; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw2; + + /* Src bitmap data follows as inline dwords. + */ +}; + + +#define CLIENT_2D 0x2 +#define OPCODE_XY_SETUP_BLT 0x1 +#define OPCODE_XY_COLOR_BLT 0x50 +#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31 + +#endif diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c index 9c59677a741..eea6b5d6a5c 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_arit.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c @@ -629,7 +629,8 @@ lp_build_abs(struct lp_build_context *bld, if(type.floating) { /* Mask out the sign bit */ LLVMTypeRef int_vec_type = lp_build_int_vec_type(type); - LLVMValueRef mask = lp_build_int_const_scalar(type, ((unsigned long long)1 << type.width) - 1); + unsigned long absMask = ~(1 << (type.width - 1)); + LLVMValueRef mask = lp_build_int_const_scalar(type, ((unsigned long long) absMask)); a = LLVMBuildBitCast(bld->builder, a, int_vec_type, ""); a = LLVMBuildAnd(bld->builder, a, mask, ""); a = LLVMBuildBitCast(bld->builder, a, vec_type, ""); @@ -1083,7 +1084,7 @@ lp_build_log(struct lp_build_context *bld, LLVMValueRef x) { /* log(2) */ - LLVMValueRef log2 = lp_build_const_scalar(bld->type, 1.4426950408889634); + LLVMValueRef log2 = lp_build_const_scalar(bld->type, 0.69314718055994529); return lp_build_mul(bld, log2, lp_build_exp2(bld, x)); } @@ -1095,7 +1096,7 @@ lp_build_log(struct lp_build_context *bld, /** * Generate polynomial. - * Ex: x^2 * coeffs[0] + x * coeffs[1] + coeffs[2]. + * Ex: coeffs[0] + x * coeffs[1] + x^2 * coeffs[2]. */ static LLVMValueRef lp_build_polynomial(struct lp_build_context *bld, @@ -1285,13 +1286,13 @@ lp_build_log2_approx(struct lp_build_context *bld, /* mant = (float) mantissa(x) */ mant = LLVMBuildAnd(bld->builder, i, mantmask, ""); mant = LLVMBuildOr(bld->builder, mant, one, ""); - mant = LLVMBuildSIToFP(bld->builder, mant, vec_type, ""); + mant = LLVMBuildBitCast(bld->builder, mant, vec_type, ""); logmant = lp_build_polynomial(bld, mant, lp_build_log2_polynomial, Elements(lp_build_log2_polynomial)); /* This effectively increases the polynomial degree by one, but ensures that log2(1) == 0*/ - logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildMul(bld->builder, mant, bld->one, ""), ""); + logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildSub(bld->builder, mant, bld->one, ""), ""); res = LLVMBuildAdd(bld->builder, logmant, logexp, ""); } diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c index 3eb0e0c57cb..a67c70ff25a 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c @@ -763,7 +763,7 @@ emit_instruction( FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) { src0 = emit_fetch( bld, inst, 0, chan_index ); tmp0 = lp_build_floor(&bld->base, src0); - tmp0 = lp_build_sub(&bld->base, tmp0, src0); + tmp0 = lp_build_sub(&bld->base, src0, tmp0); dst0[chan_index] = tmp0; } break; diff --git a/src/gallium/drivers/llvmpipe/lp_debug.h b/src/gallium/drivers/llvmpipe/lp_debug.h new file mode 100644 index 00000000000..74b27574942 --- /dev/null +++ b/src/gallium/drivers/llvmpipe/lp_debug.h @@ -0,0 +1,71 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef LP_DEBUG_H +#define LP_DEBUG_H + +#include "pipe/p_compiler.h" +#include "util/u_debug.h" + +extern void +st_print_current(void); + + +#define DEBUG_PIPE 0x1 +#define DEBUG_TGSI 0x2 +#define DEBUG_TEX 0x4 +#define DEBUG_ASM 0x8 +#define DEBUG_SETUP 0x10 +#define DEBUG_RAST 0x20 +#define DEBUG_QUERY 0x40 +#define DEBUG_SCREEN 0x80 +#define DEBUG_JIT 0x100 + +#ifdef DEBUG +extern int LP_DEBUG; +#else +#define LP_DEBUG 0 +#endif + +void st_debug_init( void ); + +static INLINE void +LP_DBG( unsigned flag, const char *fmt, ... ) +{ + if (LP_DEBUG & flag) + { + va_list args; + + va_start( args, fmt ); + debug_vprintf( fmt, args ); + va_end( args ); + } +} + + +#endif /* LP_DEBUG_H */ diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c index 19fe2850fd1..9b47415f003 100644 --- a/src/gallium/drivers/llvmpipe/lp_screen.c +++ b/src/gallium/drivers/llvmpipe/lp_screen.c @@ -36,6 +36,24 @@ #include "lp_winsys.h" #include "lp_jit.h" #include "lp_screen.h" +#include "lp_debug.h" + +#ifdef DEBUG +int LP_DEBUG = 0; + +static const struct debug_named_value lp_debug_flags[] = { + { "pipe", DEBUG_PIPE }, + { "tgsi", DEBUG_TGSI }, + { "tex", DEBUG_TEX }, + { "asm", DEBUG_ASM }, + { "setup", DEBUG_SETUP }, + { "rast", DEBUG_RAST }, + { "query", DEBUG_QUERY }, + { "screen", DEBUG_SCREEN }, + { "jit", DEBUG_JIT }, + {NULL, 0} +}; +#endif static const char * @@ -259,6 +277,10 @@ llvmpipe_create_screen(struct llvmpipe_winsys *winsys) { struct llvmpipe_screen *screen = CALLOC_STRUCT(llvmpipe_screen); +#ifdef DEBUG + LP_DEBUG = debug_get_flags_option("LP_DEBUG", lp_debug_flags, 0 ); +#endif + if (!screen) return NULL; diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c index b4aabd4d7cc..b18f17c0cd3 100644 --- a/src/gallium/drivers/llvmpipe/lp_setup.c +++ b/src/gallium/drivers/llvmpipe/lp_setup.c @@ -41,6 +41,7 @@ #include "draw/draw_vertex.h" #include "pipe/p_shader_tokens.h" #include "pipe/p_thread.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "lp_bld_debug.h" @@ -166,7 +167,7 @@ shade_quads(struct llvmpipe_context *llvmpipe, assert((y % 2) == 0); depth = llvmpipe->zsbuf_map + y*llvmpipe->zsbuf_transfer->stride + - 2*x*pf_get_blocksize(llvmpipe->zsbuf_transfer->texture->format); + 2*x*util_format_get_blocksize(llvmpipe->zsbuf_transfer->texture->format); } else depth = NULL; diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c index ee0f69b2af9..22683ff8b42 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_fs.c +++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c @@ -87,6 +87,7 @@ #include "lp_state.h" #include "lp_quad.h" #include "lp_tex_sample.h" +#include "lp_debug.h" static const unsigned char quad_offset_x[4] = {0, 1, 0, 1}; @@ -408,59 +409,58 @@ generate_fragment(struct llvmpipe_context *lp, unsigned i; unsigned chan; -#ifdef DEBUG - tgsi_dump(shader->base.tokens, 0); - if(key->depth.enabled) { - debug_printf("depth.format = %s\n", pf_name(key->zsbuf_format)); - debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE)); - debug_printf("depth.writemask = %u\n", key->depth.writemask); - } - if(key->alpha.enabled) { - debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE)); - debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value); - } - if(key->blend.logicop_enable) { - debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func); - } - else if(key->blend.blend_enable) { - debug_printf("blend.rgb_func = %s\n", debug_dump_blend_func (key->blend.rgb_func, TRUE)); - debug_printf("rgb_src_factor = %s\n", debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE)); - debug_printf("rgb_dst_factor = %s\n", debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE)); - debug_printf("alpha_func = %s\n", debug_dump_blend_func (key->blend.alpha_func, TRUE)); - debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE)); - debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE)); - } - debug_printf("blend.colormask = 0x%x\n", key->blend.colormask); - for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) { - if(key->sampler[i].format) { - debug_printf("sampler[%u] = \n", i); - debug_printf(" .format = %s\n", - pf_name(key->sampler[i].format)); - debug_printf(" .target = %s\n", - debug_dump_tex_target(key->sampler[i].target, TRUE)); - debug_printf(" .pot = %u %u %u\n", - key->sampler[i].pot_width, - key->sampler[i].pot_height, - key->sampler[i].pot_depth); - debug_printf(" .wrap = %s %s %s\n", - debug_dump_tex_wrap(key->sampler[i].wrap_s, TRUE), - debug_dump_tex_wrap(key->sampler[i].wrap_t, TRUE), - debug_dump_tex_wrap(key->sampler[i].wrap_r, TRUE)); - debug_printf(" .min_img_filter = %s\n", - debug_dump_tex_filter(key->sampler[i].min_img_filter, TRUE)); - debug_printf(" .min_mip_filter = %s\n", - debug_dump_tex_mipfilter(key->sampler[i].min_mip_filter, TRUE)); - debug_printf(" .mag_img_filter = %s\n", - debug_dump_tex_filter(key->sampler[i].mag_img_filter, TRUE)); - if(key->sampler[i].compare_mode) - debug_printf(" .compare_mode = %s\n", debug_dump_func(key->sampler[i].compare_func, TRUE)); - debug_printf(" .normalized_coords = %u\n", key->sampler[i].normalized_coords); - debug_printf(" .prefilter = %u\n", key->sampler[i].prefilter); + if (LP_DEBUG & DEBUG_JIT) { + tgsi_dump(shader->base.tokens, 0); + if(key->depth.enabled) { + debug_printf("depth.format = %s\n", pf_name(key->zsbuf_format)); + debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE)); + debug_printf("depth.writemask = %u\n", key->depth.writemask); + } + if(key->alpha.enabled) { + debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE)); + debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value); + } + if(key->blend.logicop_enable) { + debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func); + } + else if(key->blend.blend_enable) { + debug_printf("blend.rgb_func = %s\n", debug_dump_blend_func (key->blend.rgb_func, TRUE)); + debug_printf("rgb_src_factor = %s\n", debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE)); + debug_printf("rgb_dst_factor = %s\n", debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE)); + debug_printf("alpha_func = %s\n", debug_dump_blend_func (key->blend.alpha_func, TRUE)); + debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE)); + debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE)); + } + debug_printf("blend.colormask = 0x%x\n", key->blend.colormask); + for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) { + if(key->sampler[i].format) { + debug_printf("sampler[%u] = \n", i); + debug_printf(" .format = %s\n", + pf_name(key->sampler[i].format)); + debug_printf(" .target = %s\n", + debug_dump_tex_target(key->sampler[i].target, TRUE)); + debug_printf(" .pot = %u %u %u\n", + key->sampler[i].pot_width, + key->sampler[i].pot_height, + key->sampler[i].pot_depth); + debug_printf(" .wrap = %s %s %s\n", + debug_dump_tex_wrap(key->sampler[i].wrap_s, TRUE), + debug_dump_tex_wrap(key->sampler[i].wrap_t, TRUE), + debug_dump_tex_wrap(key->sampler[i].wrap_r, TRUE)); + debug_printf(" .min_img_filter = %s\n", + debug_dump_tex_filter(key->sampler[i].min_img_filter, TRUE)); + debug_printf(" .min_mip_filter = %s\n", + debug_dump_tex_mipfilter(key->sampler[i].min_mip_filter, TRUE)); + debug_printf(" .mag_img_filter = %s\n", + debug_dump_tex_filter(key->sampler[i].mag_img_filter, TRUE)); + if(key->sampler[i].compare_mode) + debug_printf(" .compare_mode = %s\n", debug_dump_func(key->sampler[i].compare_func, TRUE)); + debug_printf(" .normalized_coords = %u\n", key->sampler[i].normalized_coords); + debug_printf(" .prefilter = %u\n", key->sampler[i].prefilter); + } } } -#endif - variant = CALLOC_STRUCT(lp_fragment_shader_variant); if(!variant) return NULL; @@ -599,8 +599,8 @@ generate_fragment(struct llvmpipe_context *lp, } lp_build_conv_mask(builder, fs_type, blend_type, - fs_mask, num_fs, - &blend_mask, 1); + fs_mask, num_fs, + &blend_mask, 1); /* * Blending. @@ -631,16 +631,15 @@ generate_fragment(struct llvmpipe_context *lp, LLVMRunFunctionPassManager(screen->pass, variant->function); -#ifdef DEBUG - LLVMDumpValue(variant->function); - debug_printf("\n"); -#endif + if (LP_DEBUG & DEBUG_JIT) { + LLVMDumpValue(variant->function); + debug_printf("\n"); + } variant->jit_function = (lp_jit_frag_func)LLVMGetPointerToGlobal(screen->engine, variant->function); -#ifdef DEBUG - lp_disassemble(variant->jit_function); -#endif + if (LP_DEBUG & DEBUG_ASM) + lp_disassemble(variant->jit_function); variant->next = shader->variants; shader->variants = variant; diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c index c06ce8b75c1..ba970cac985 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_surface.c +++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c @@ -35,6 +35,8 @@ #include "draw/draw_context.h" +#include "util/u_format.h" + /** * XXX this might get moved someday @@ -88,8 +90,9 @@ llvmpipe_set_framebuffer_state(struct pipe_context *pipe, if (lp->framebuffer.zsbuf) { int depth_bits; double mrd; - depth_bits = pf_get_component_bits(lp->framebuffer.zsbuf->format, - PIPE_FORMAT_COMP_Z); + depth_bits = util_format_get_component_bits(lp->framebuffer.zsbuf->format, + UTIL_FORMAT_COLORSPACE_ZS, + 0); if (depth_bits > 16) { mrd = 0.0000001; } diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index f099f903bd7..2c135029ea2 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -34,6 +34,8 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" + +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -67,10 +69,10 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, /* Allocate storage for whole quads. This is particularly important * for depth surfaces, which are currently stored in a swizzled format. */ - nblocksx = pf_get_nblocksx(pt->format, align(width, 2)); - nblocksy = pf_get_nblocksy(pt->format, align(height, 2)); + nblocksx = util_format_get_nblocksx(pt->format, align(width, 2)); + nblocksy = util_format_get_nblocksy(pt->format, align(height, 2)); - lpt->stride[level] = align(nblocksx * pf_get_blocksize(pt->format), 16); + lpt->stride[level] = align(nblocksx * util_format_get_blocksize(pt->format), 16); lpt->level_offset[level] = buffer_size; @@ -249,11 +251,11 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen, */ if (pt->target == PIPE_TEXTURE_CUBE) { unsigned tex_height = ps->height; - ps->offset += face * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; + ps->offset += face * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { unsigned tex_height = ps->height; - ps->offset += zslice * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; + ps->offset += zslice * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else { assert(face == 0); @@ -312,11 +314,11 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen, */ if (texture->target == PIPE_TEXTURE_CUBE) { unsigned tex_height = u_minify(texture->height0, level); - lpt->offset += face * pf_get_nblocksy(texture->format, tex_height) * pt->stride; + lpt->offset += face * util_format_get_nblocksy(texture->format, tex_height) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { unsigned tex_height = u_minify(texture->height0, level); - lpt->offset += zslice * pf_get_nblocksy(texture->format, tex_height) * pt->stride; + lpt->offset += zslice * util_format_get_nblocksy(texture->format, tex_height) * pt->stride; } else { assert(face == 0); @@ -377,8 +379,8 @@ llvmpipe_transfer_map( struct pipe_screen *_screen, } xfer_map = map + llvmpipe_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); /*printf("map = %p xfer map = %p\n", map, xfer_map);*/ return xfer_map; } diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h index 62990f9b6a6..9aee9e49566 100644 --- a/src/gallium/drivers/nouveau/nouveau_stateobj.h +++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h @@ -112,20 +112,30 @@ so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so) { struct nouveau_pushbuf *pb = chan->pushbuf; unsigned nr, i; + int ret = 0; nr = so->cur - so->push; - if (pb->remaining < nr) - nouveau_pushbuf_flush(chan, nr); + /* This will flush if we need space. + * We don't actually need the marker. + */ + if ((ret = nouveau_pushbuf_marker_emit(chan, nr, so->cur_reloc))) { + debug_printf("so_emit failed marker emit with error %d\n", ret); + return; + } pb->remaining -= nr; memcpy(pb->cur, so->push, nr * 4); for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset, + if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset, r->bo, r->data, 0, r->flags, - r->vor, r->tor); + r->vor, r->tor))) { + debug_printf("so_emit failed reloc with error %d\n", ret); + goto out; + } } +out: pb->cur += nr; } @@ -134,26 +144,45 @@ so_emit_reloc_markers(struct nouveau_channel *chan, struct nouveau_stateobj *so) { struct nouveau_pushbuf *pb = chan->pushbuf; unsigned i; + int ret = 0; if (!so) return; i = so->cur_reloc << 1; - if (pb->remaining < i) - nouveau_pushbuf_flush(chan, i); + /* This will flush if we need space. + * We don't actually need the marker. + */ + if ((ret = nouveau_pushbuf_marker_emit(chan, i, i))) { + debug_printf("so_emit_reloc_markers failed marker emit with" \ + "error %d\n", ret); + return; + } pb->remaining -= i; for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->packet, 0, + if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, + r->packet, 0, (r->flags & (NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RDWR)) | - NOUVEAU_BO_DUMMY, 0, 0); - nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->data, 0, + NOUVEAU_BO_DUMMY, 0, 0))) { + debug_printf("so_emit_reloc_markers failed reloc" \ + "with error %d\n", ret); + pb->remaining += ((so->cur_reloc - i) << 1); + return; + } + if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, + r->data, 0, r->flags | NOUVEAU_BO_DUMMY, - r->vor, r->tor); + r->vor, r->tor))) { + debug_printf("so_emit_reloc_markers failed reloc" \ + "with error %d\n", ret); + pb->remaining += ((so->cur_reloc - i) << 1) - 1; + return; + } } } diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c index 932893eef59..12df7fd1997 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -1,5 +1,6 @@ #include "pipe/p_context.h" #include "pipe/p_format.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -133,6 +134,9 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, assert(sub_w == w || util_is_pot(sub_w)); assert(sub_h == h || util_is_pot(sub_h)); + MARK_RING (chan, 8 + ((w+sub_w)/sub_w)*((h+sub_h)/sub_h)*17, 2 + + ((w+sub_w)/sub_w)*((h+sub_h)/sub_h)*2); + BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_DMA_IMAGE, 1); OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); @@ -155,10 +159,10 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, sub_w = MIN2(sub_w, w - x); /* Must be 64-byte aligned */ - assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format)) & 63)); + assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format)) & 63)); BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1); - OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format), + OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format), NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION, 9); @@ -177,7 +181,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, OUT_RING (chan, src_pitch | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE); - OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * pf_get_blocksize(src->texture->format), + OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * util_format_get_blocksize(src->texture->format), NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RING (chan, 0); } @@ -198,11 +202,11 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, unsigned src_pitch = ((struct nv04_surface *)src)->pitch; unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; unsigned dst_offset = dst->offset + dy * dst_pitch + - dx * pf_get_blocksize(dst->texture->format); + dx * util_format_get_blocksize(dst->texture->format); unsigned src_offset = src->offset + sy * src_pitch + - sx * pf_get_blocksize(src->texture->format); + sx * util_format_get_blocksize(src->texture->format); - WAIT_RING (chan, 3 + ((h / 2047) + 1) * 9); + MARK_RING (chan, 3 + ((h / 2047) + 1) * 9, 2 + ((h / 2047) + 1) * 2); BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2); OUT_RELOCo(chan, src_bo, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); @@ -219,7 +223,7 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR); OUT_RING (chan, src_pitch); OUT_RING (chan, dst_pitch); - OUT_RING (chan, w * pf_get_blocksize(src->texture->format)); + OUT_RING (chan, w * util_format_get_blocksize(src->texture->format)); OUT_RING (chan, count); OUT_RING (chan, 0x0101); OUT_RING (chan, 0); @@ -250,7 +254,7 @@ nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst, if (format < 0) return 1; - WAIT_RING (chan, 12); + MARK_RING (chan, 12, 4); BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE, 2); OUT_RELOCo(chan, src_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); @@ -315,7 +319,7 @@ nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst, gdirect_format = nv04_rect_format(dst->format); assert(gdirect_format >= 0); - WAIT_RING (chan, 16); + MARK_RING (chan, 16, 4); BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE, 2); OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c index d66d6c6346c..8446073ae80 100644 --- a/src/gallium/drivers/nv04/nv04_transfer.c +++ b/src/gallium/drivers/nv04/nv04_transfer.c @@ -1,6 +1,7 @@ #include <pipe/p_state.h> #include <pipe/p_defines.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <nouveau/nouveau_winsys.h> @@ -151,7 +152,7 @@ nv04_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c index 6a52b6af362..908482ad854 100644 --- a/src/gallium/drivers/nv10/nv10_miptree.c +++ b/src/gallium/drivers/nv10/nv10_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv10_context.h" @@ -23,9 +24,9 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt) for (l = 0; l <= pt->last_level; l++) { if (swizzled) - nv10mt->level[l].pitch = pf_get_stride(pt->format, width); + nv10mt->level[l].pitch = util_format_get_stride(pt->format, width); else - nv10mt->level[l].pitch = pf_get_stride(pt->format, pt->width0); + nv10mt->level[l].pitch = util_format_get_stride(pt->format, pt->width0); nv10mt->level[l].pitch = (nv10mt->level[l].pitch + 63) & ~63; nv10mt->level[l].image_offset = diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c index 06bb5134173..c664973e904 100644 --- a/src/gallium/drivers/nv10/nv10_transfer.c +++ b/src/gallium/drivers/nv10/nv10_transfer.c @@ -1,6 +1,7 @@ #include <pipe/p_state.h> #include <pipe/p_defines.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <nouveau/nouveau_winsys.h> @@ -151,7 +152,7 @@ nv10_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c index e2e01bd849b..d1291a92e0a 100644 --- a/src/gallium/drivers/nv20/nv20_miptree.c +++ b/src/gallium/drivers/nv20/nv20_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv20_context.h" @@ -27,9 +28,9 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv20mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + nv20mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - nv20mt->level[l].pitch = pf_get_stride(pt->format, width); + nv20mt->level[l].pitch = util_format_get_stride(pt->format, width); nv20mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c index 26a73c5143a..69b79c809f4 100644 --- a/src/gallium/drivers/nv20/nv20_transfer.c +++ b/src/gallium/drivers/nv20/nv20_transfer.c @@ -1,6 +1,7 @@ #include <pipe/p_state.h> #include <pipe/p_defines.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <nouveau/nouveau_winsys.h> @@ -151,7 +152,7 @@ nv20_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv30/nv30_context.c b/src/gallium/drivers/nv30/nv30_context.c index d8300fd69f6..46a821a48b1 100644 --- a/src/gallium/drivers/nv30/nv30_context.c +++ b/src/gallium/drivers/nv30/nv30_context.c @@ -58,6 +58,9 @@ nv30_create(struct pipe_screen *pscreen, unsigned pctx_id) nv30->pipe.is_texture_referenced = nouveau_is_texture_referenced; nv30->pipe.is_buffer_referenced = nouveau_is_buffer_referenced; + screen->base.channel->user_private = nv30; + screen->base.channel->flush_notify = nv30_state_flush_notify; + nv30_init_query_functions(nv30); nv30_init_surface_functions(nv30); nv30_init_state_functions(nv30); diff --git a/src/gallium/drivers/nv30/nv30_context.h b/src/gallium/drivers/nv30/nv30_context.h index 026cc82e0af..864ddaeb598 100644 --- a/src/gallium/drivers/nv30/nv30_context.h +++ b/src/gallium/drivers/nv30/nv30_context.h @@ -183,6 +183,7 @@ extern void nv30_fragtex_bind(struct nv30_context *); /* nv30_state.c and friends */ extern boolean nv30_state_validate(struct nv30_context *nv30); extern void nv30_state_emit(struct nv30_context *nv30); +extern void nv30_state_flush_notify(struct nouveau_channel *chan); extern struct nv30_state_entry nv30_state_rasterizer; extern struct nv30_state_entry nv30_state_scissor; extern struct nv30_state_entry nv30_state_stipple; diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c index 920fe64c32f..ce95d9700f6 100644 --- a/src/gallium/drivers/nv30/nv30_miptree.c +++ b/src/gallium/drivers/nv30/nv30_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv30_context.h" @@ -29,9 +30,9 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv30mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + nv30mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - nv30mt->level[l].pitch = pf_get_stride(pt->format, width); + nv30mt->level[l].pitch = util_format_get_stride(pt->format, width); nv30mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c index 621b8846c8e..ac52d946f02 100644 --- a/src/gallium/drivers/nv30/nv30_state_emit.c +++ b/src/gallium/drivers/nv30/nv30_state_emit.c @@ -41,7 +41,7 @@ nv30_state_emit(struct nv30_context *nv30) struct nouveau_channel *chan = nv30->screen->base.channel; struct nv30_state *state = &nv30->state; struct nv30_screen *screen = nv30->screen; - unsigned i, samplers; + unsigned i; uint64_t states; if (nv30->pctx_id != screen->cur_pctx) { @@ -63,6 +63,14 @@ nv30_state_emit(struct nv30_context *nv30) } state->dirty = 0; +} + +void +nv30_state_flush_notify(struct nouveau_channel *chan) +{ + struct nv30_context *nv30 = chan->user_private; + struct nv30_state *state = &nv30->state; + unsigned i, samplers; so_emit_reloc_markers(chan, state->hw[NV30_STATE_FB]); for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c index e29bfbd3efd..2255a02caed 100644 --- a/src/gallium/drivers/nv30/nv30_transfer.c +++ b/src/gallium/drivers/nv30/nv30_transfer.c @@ -1,6 +1,7 @@ #include <pipe/p_state.h> #include <pipe/p_defines.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <nouveau/nouveau_winsys.h> @@ -151,7 +152,7 @@ nv30_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv40/nv40_context.c b/src/gallium/drivers/nv40/nv40_context.c index 7f008274a4e..eb9cce4c786 100644 --- a/src/gallium/drivers/nv40/nv40_context.c +++ b/src/gallium/drivers/nv40/nv40_context.c @@ -58,6 +58,9 @@ nv40_create(struct pipe_screen *pscreen, unsigned pctx_id) nv40->pipe.is_texture_referenced = nouveau_is_texture_referenced; nv40->pipe.is_buffer_referenced = nouveau_is_buffer_referenced; + screen->base.channel->user_private = nv40; + screen->base.channel->flush_notify = nv40_state_flush_notify; + nv40_init_query_functions(nv40); nv40_init_surface_functions(nv40); nv40_init_state_functions(nv40); diff --git a/src/gallium/drivers/nv40/nv40_context.h b/src/gallium/drivers/nv40/nv40_context.h index 06172e88175..83fcf1785d9 100644 --- a/src/gallium/drivers/nv40/nv40_context.h +++ b/src/gallium/drivers/nv40/nv40_context.h @@ -203,6 +203,7 @@ extern void nv40_fragtex_bind(struct nv40_context *); extern boolean nv40_state_validate(struct nv40_context *nv40); extern boolean nv40_state_validate_swtnl(struct nv40_context *nv40); extern void nv40_state_emit(struct nv40_context *nv40); +extern void nv40_state_flush_notify(struct nouveau_channel *chan); extern struct nv40_state_entry nv40_state_rasterizer; extern struct nv40_state_entry nv40_state_scissor; extern struct nv40_state_entry nv40_state_stipple; diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c index 89ddf373e9e..b974e68a077 100644 --- a/src/gallium/drivers/nv40/nv40_miptree.c +++ b/src/gallium/drivers/nv40/nv40_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv40_context.h" @@ -31,9 +32,9 @@ nv40_miptree_layout(struct nv40_miptree *mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - mt->level[l].pitch = pf_get_stride(pt->format, width); + mt->level[l].pitch = util_format_get_stride(pt->format, width); mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c index 980ed217ecc..789ed16126a 100644 --- a/src/gallium/drivers/nv40/nv40_state_emit.c +++ b/src/gallium/drivers/nv40/nv40_state_emit.c @@ -57,7 +57,7 @@ nv40_state_emit(struct nv40_context *nv40) struct nouveau_channel *chan = nv40->screen->base.channel; struct nv40_state *state = &nv40->state; struct nv40_screen *screen = nv40->screen; - unsigned i, samplers; + unsigned i; uint64_t states; if (nv40->pctx_id != screen->cur_pctx) { @@ -87,6 +87,14 @@ nv40_state_emit(struct nv40_context *nv40) } state->dirty = 0; +} + +void +nv40_state_flush_notify(struct nouveau_channel *chan) +{ + struct nv40_context *nv40 = chan->user_private; + struct nv40_state *state = &nv40->state; + unsigned i, samplers; so_emit_reloc_markers(chan, state->hw[NV40_STATE_FB]); for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) { diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c index ed5be1cf879..b084a38b482 100644 --- a/src/gallium/drivers/nv40/nv40_transfer.c +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -1,6 +1,7 @@ #include <pipe/p_state.h> #include <pipe/p_defines.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <nouveau/nouveau_winsys.h> @@ -151,7 +152,7 @@ nv40_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 79135f2f362..5578a5838fb 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -126,7 +126,7 @@ struct nv50_state { unsigned viewport_bypass; struct nouveau_stateobj *tsc_upload; struct nouveau_stateobj *tic_upload; - unsigned miptree_nr; + unsigned miptree_nr[PIPE_SHADER_TYPES]; struct nouveau_stateobj *vertprog; struct nouveau_stateobj *fragprog; struct nouveau_stateobj *programs; @@ -162,10 +162,10 @@ struct nv50_context { unsigned vtxbuf_nr; struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS]; unsigned vtxelt_nr; - struct nv50_sampler_stateobj *sampler[PIPE_MAX_SAMPLERS]; - unsigned sampler_nr; - struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS]; - unsigned miptree_nr; + struct nv50_sampler_stateobj *sampler[PIPE_SHADER_TYPES][PIPE_MAX_SAMPLERS]; + unsigned sampler_nr[PIPE_SHADER_TYPES]; + struct nv50_miptree *miptree[PIPE_SHADER_TYPES][PIPE_MAX_SAMPLERS]; + unsigned miptree_nr[PIPE_SHADER_TYPES]; uint16_t vbo_fifo; }; @@ -218,7 +218,7 @@ extern void nv50_state_flush_notify(struct nouveau_channel *chan); extern void nv50_so_init_sifc(struct nv50_context *nv50, struct nouveau_stateobj *so, struct nouveau_bo *bo, unsigned reloc, - unsigned size); + unsigned offset, unsigned size); /* nv50_tex.c */ extern void nv50_tex_validate(struct nv50_context *); diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 40ee6659993..3f1edf0a139 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -23,6 +23,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "nv50_context.h" @@ -55,6 +56,20 @@ get_tile_mode(unsigned ny, unsigned d) return tile_mode | 0x10; } +static INLINE unsigned +get_zslice_offset(unsigned tile_mode, unsigned z, unsigned pitch, unsigned nb_h) +{ + unsigned tile_h = get_tile_height(tile_mode); + unsigned tile_d = get_tile_depth(tile_mode); + + /* pitch_2d == to next slice within this volume-tile */ + /* pitch_3d == size (in bytes) of a volume-tile */ + unsigned pitch_2d = tile_h * 64; + unsigned pitch_3d = tile_d * align(nb_h, tile_h) * pitch; + + return (z % tile_d) * pitch_2d + (z / tile_d) * pitch_3d; +} + static struct pipe_texture * nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) { @@ -91,10 +106,10 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) for (l = 0; l <= pt->last_level; l++) { struct nv50_miptree_level *lvl = &mt->level[l]; - unsigned nblocksy = pf_get_nblocksy(pt->format, height); + unsigned nblocksy = util_format_get_nblocksy(pt->format, height); lvl->image_offset = CALLOC(mt->image_nr, sizeof(int)); - lvl->pitch = align(pf_get_stride(pt->format, width), 64); + lvl->pitch = align(util_format_get_stride(pt->format, width), 64); lvl->tile_mode = get_tile_mode(nblocksy, depth); width = u_minify(width, 1); @@ -116,7 +131,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) unsigned tile_d = get_tile_depth(lvl->tile_mode); size = lvl->pitch; - size *= align(pf_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h); + size *= align(util_format_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h); size *= align(u_minify(pt->depth0, l), tile_d); lvl->image_offset[i] = mt->total_size; @@ -130,6 +145,8 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) mt->level[0].tile_mode, tile_flags, &mt->base.bo); if (ret) { + for (l = 0; l < pt->last_level; ++l) + FREE(mt->level[l].image_offset); FREE(mt); return NULL; } @@ -169,6 +186,10 @@ static void nv50_miptree_destroy(struct pipe_texture *pt) { struct nv50_miptree *mt = nv50_miptree(pt); + unsigned l; + + for (l = 0; l < pt->last_level; ++l) + FREE(mt->level[l].image_offset); nouveau_bo_ref(NULL, &mt->base.bo); FREE(mt); @@ -182,15 +203,10 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, struct nv50_miptree *mt = nv50_miptree(pt); struct nv50_miptree_level *lvl = &mt->level[level]; struct pipe_surface *ps; - int img; + unsigned img = 0; if (pt->target == PIPE_TEXTURE_CUBE) img = face; - else - if (pt->target == PIPE_TEXTURE_3D) - img = zslice; - else - img = 0; ps = CALLOC_STRUCT(pipe_surface); if (!ps) @@ -206,6 +222,12 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, ps->zslice = zslice; ps->offset = lvl->image_offset[img]; + if (pt->target == PIPE_TEXTURE_3D) { + unsigned nb_h = util_format_get_nblocksy(pt->format, ps->height); + ps->offset += get_zslice_offset(lvl->tile_mode, zslice, + lvl->pitch, nb_h); + } + return ps; } diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index f0fe7e61684..679c28ce4b1 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -98,9 +98,17 @@ struct nv50_reg { #define NV50_MOD_ABS 2 #define NV50_MOD_SAT 4 -/* arbitrary limits */ -#define MAX_IF_DEPTH 4 -#define MAX_LOOP_DEPTH 4 +/* STACK: Conditionals and loops have to use the (per warp) stack. + * Stack entries consist of an entry type (divergent path, join at), + * a mask indicating the active threads of the warp, and an address. + * MPs can store 12 stack entries internally, if we need more (and + * we probably do), we have to create a stack buffer in VRAM. + */ +/* impose low limits for now */ +#define NV50_MAX_COND_NESTING 4 +#define NV50_MAX_LOOP_NESTING 3 + +#define JOIN_ON(e) e; pc->p->exec_tail->inst[1] |= 2 struct nv50_pc { struct nv50_program *p; @@ -119,10 +127,11 @@ struct nv50_pc { struct nv50_reg *param; int param_nr; struct nv50_reg *immd; - float *immd_buf; + uint32_t *immd_buf; int immd_nr; struct nv50_reg **addr; int addr_nr; + uint8_t addr_alloc; /* set bit indicates used for TGSI_FILE_ADDRESS */ struct nv50_reg *temp_temp[16]; unsigned temp_temp_nr; @@ -131,17 +140,19 @@ struct nv50_pc { struct nv50_reg *r_brdc; struct nv50_reg *r_dst[4]; + struct nv50_reg reg_instances[16]; + unsigned reg_instance_nr; + unsigned interp_mode[32]; /* perspective interpolation registers */ struct nv50_reg *iv_p; struct nv50_reg *iv_c; - struct nv50_program_exec *if_cond; - struct nv50_program_exec *if_insn[MAX_IF_DEPTH]; - struct nv50_program_exec *br_join[MAX_IF_DEPTH]; - struct nv50_program_exec *br_loop[MAX_LOOP_DEPTH]; /* for BRK branch */ + struct nv50_program_exec *if_insn[NV50_MAX_COND_NESTING]; + struct nv50_program_exec *if_join[NV50_MAX_COND_NESTING]; + struct nv50_program_exec *loop_brka[NV50_MAX_LOOP_NESTING]; int if_lvl, loop_lvl; - unsigned loop_pos[MAX_LOOP_DEPTH]; + unsigned loop_pos[NV50_MAX_LOOP_NESTING]; /* current instruction and total number of insns */ unsigned insn_cur; @@ -150,6 +161,20 @@ struct nv50_pc { boolean allow32; }; +static INLINE struct nv50_reg * +reg_instance(struct nv50_pc *pc, struct nv50_reg *reg) +{ + struct nv50_reg *ri; + + assert(pc->reg_instance_nr < 16); + ri = &pc->reg_instances[pc->reg_instance_nr++]; + if (reg) { + *ri = *reg; + reg->mod = 0; + } + return ri; +} + static INLINE void ctor_reg(struct nv50_reg *reg, unsigned type, int index, int hw) { @@ -176,8 +201,7 @@ terminate_mbb(struct nv50_pc *pc) /* remove records of temporary address register values */ for (i = 0; i < NV50_SU_MAX_ADDR; ++i) - if (pc->r_addr[i].index < 0) - pc->r_addr[i].rhw = -1; + pc->r_addr[i].rhw = -1; } static void @@ -342,25 +366,34 @@ static void kill_temp_temp(struct nv50_pc *pc) { int i; - + for (i = 0; i < pc->temp_temp_nr; i++) free_temp(pc, pc->temp_temp[i]); pc->temp_temp_nr = 0; } static int -ctor_immd(struct nv50_pc *pc, float x, float y, float z, float w) +ctor_immd_4u32(struct nv50_pc *pc, + uint32_t x, uint32_t y, uint32_t z, uint32_t w) { - pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * 4 * sizeof(float)), - (pc->immd_nr + 1) * 4 * sizeof(float)); + unsigned size = pc->immd_nr * 4 * sizeof(uint32_t); + + pc->immd_buf = REALLOC(pc->immd_buf, size, size + 4 * sizeof(uint32_t)); + pc->immd_buf[(pc->immd_nr * 4) + 0] = x; pc->immd_buf[(pc->immd_nr * 4) + 1] = y; pc->immd_buf[(pc->immd_nr * 4) + 2] = z; pc->immd_buf[(pc->immd_nr * 4) + 3] = w; - + return pc->immd_nr++; } +static INLINE int +ctor_immd_4f32(struct nv50_pc *pc, float x, float y, float z, float w) +{ + return ctor_immd_4u32(pc, fui(x), fui(y), fui(z), fui(w)); +} + static struct nv50_reg * alloc_immd(struct nv50_pc *pc, float f) { @@ -368,11 +401,11 @@ alloc_immd(struct nv50_pc *pc, float f) unsigned hw; for (hw = 0; hw < pc->immd_nr * 4; hw++) - if (pc->immd_buf[hw] == f) + if (pc->immd_buf[hw] == fui(f)) break; if (hw == pc->immd_nr * 4) - hw = ctor_immd(pc, f, -f, 0.5 * f, 0) * 4; + hw = ctor_immd_4f32(pc, f, -f, 0.5 * f, 0) * 4; ctor_reg(r, P_IMMD, -1, hw); return r; @@ -464,22 +497,24 @@ set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e) static INLINE void set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e) { - unsigned val; - float f = pc->immd_buf[imm->hw]; + union { + float f; + uint32_t ui; + } u; + u.ui = pc->immd_buf[imm->hw]; - if (imm->mod & NV50_MOD_ABS) - f = fabsf(f); - val = fui((imm->mod & NV50_MOD_NEG) ? -f : f); + u.f = (imm->mod & NV50_MOD_ABS) ? fabsf(u.f) : u.f; + u.f = (imm->mod & NV50_MOD_NEG) ? -u.f : u.f; set_long(pc, e); - /*XXX: can't be predicated - bits overlap.. catch cases where both - * are required and avoid them. */ + /* XXX: can't be predicated - bits overlap; cases where both + * are required should be avoided by using pc->allow32 */ set_pred(pc, 0, 0, e); set_pred_wr(pc, 0, 0, e); e->inst[1] |= 0x00000002 | 0x00000001; - e->inst[0] |= (val & 0x3f) << 16; - e->inst[1] |= (val >> 6) << 2; + e->inst[0] |= (u.ui & 0x3f) << 16; + e->inst[1] |= (u.ui >> 6) << 2; } static INLINE void @@ -511,21 +546,24 @@ emit_add_addr_imm(struct nv50_pc *pc, struct nv50_reg *dst, static struct nv50_reg * alloc_addr(struct nv50_pc *pc, struct nv50_reg *ref) { - int i; struct nv50_reg *a_tgsi = NULL, *a = NULL; + int i; + uint8_t avail = ~pc->addr_alloc; if (!ref) { - /* allocate for TGSI address reg */ - for (i = 0; i < NV50_SU_MAX_ADDR; ++i) { - if (pc->r_addr[i].index >= 0) - continue; - if (pc->r_addr[i].rhw >= 0 && - pc->r_addr[i].acc == pc->insn_cur) - continue; + /* allocate for TGSI_FILE_ADDRESS */ + while (avail) { + i = ffs(avail) - 1; - pc->r_addr[i].rhw = -1; - pc->r_addr[i].index = i; - return &pc->r_addr[i]; + if (pc->r_addr[i].rhw < 0 || + pc->r_addr[i].acc != pc->insn_cur) { + pc->addr_alloc |= (1 << i); + + pc->r_addr[i].rhw = -1; + pc->r_addr[i].index = i; + return &pc->r_addr[i]; + } + avail &= ~(1 << i); } assert(0); return NULL; @@ -533,15 +571,16 @@ alloc_addr(struct nv50_pc *pc, struct nv50_reg *ref) /* Allocate and set an address reg so we can access 'ref'. * - * If and r_addr has index < 0, it is not reserved for TGSI, - * and index will be the negative of the TGSI addr index the - * value in rhw is relative to, or -256 if rhw is an offset - * from 0. If rhw < 0, the reg has not been initialized. + * If and r_addr->index will be -1 or the hw index the value + * value in rhw is relative to. If rhw < 0, the reg has not + * been initialized or is in use for TGSI_FILE_ADDRESS. */ - for (i = NV50_SU_MAX_ADDR - 1; i >= 0; --i) { - if (pc->r_addr[i].index >= 0) /* occupied for TGSI */ - continue; - if (pc->r_addr[i].rhw < 0) { /* unused */ + while (avail) { /* only consider regs that are not TGSI */ + i = ffs(avail) - 1; + avail &= ~(1 << i); + + if ((!a || a->rhw >= 0) && pc->r_addr[i].rhw < 0) { + /* prefer an usused reg with low hw index */ a = &pc->r_addr[i]; continue; } @@ -551,8 +590,8 @@ alloc_addr(struct nv50_pc *pc, struct nv50_reg *ref) if (ref->hw - pc->r_addr[i].rhw >= 128) continue; - if ((ref->acc >= 0 && pc->r_addr[i].index == -256) || - (ref->acc < 0 && -pc->r_addr[i].index == ref->index)) { + if ((ref->acc >= 0 && pc->r_addr[i].index < 0) || + (ref->acc < 0 && pc->r_addr[i].index == ref->index)) { pc->r_addr[i].acc = pc->insn_cur; return &pc->r_addr[i]; } @@ -566,7 +605,7 @@ alloc_addr(struct nv50_pc *pc, struct nv50_reg *ref) a->rhw = ref->hw & ~0x7f; a->acc = pc->insn_cur; - a->index = a_tgsi ? -ref->index : -256; + a->index = a_tgsi ? ref->index : -1; return a; } @@ -644,7 +683,7 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) if (src->type == P_IMMD || src->type == P_CONST) { set_long(pc, e); set_data(pc, src, 0x7f, 9, e); - e->inst[1] |= 0x20000000; /* src0 const? */ + e->inst[1] |= 0x20000000; /* mov from c[] */ } else { if (src->type == P_ATTR) { set_long(pc, e); @@ -659,9 +698,9 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) if (is_long(e) && !is_immd(e)) { e->inst[1] |= 0x04000000; /* 32-bit */ - e->inst[1] |= 0x0000c000; /* "subsubop" 0x3 */ + e->inst[1] |= 0x0000c000; /* 32-bit c[] load / lane mask 0:1 */ if (!(e->inst[1] & 0x20000000)) - e->inst[1] |= 0x00030000; /* "subsubop" 0xf */ + e->inst[1] |= 0x00030000; /* lane mask 2:3 */ } else e->inst[0] |= 0x00008000; @@ -676,6 +715,17 @@ emit_mov_immdval(struct nv50_pc *pc, struct nv50_reg *dst, float f) FREE(imm); } +static void +emit_nop(struct nv50_pc *pc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0xf0000000; + set_long(pc, e); + e->inst[1] = 0xe0000000; + emit(pc, e); +} + static boolean check_swap_src_0_1(struct nv50_pc *pc, struct nv50_reg **s0, struct nv50_reg **s1) @@ -795,6 +845,33 @@ set_src_2(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e) } static void +emit_mov_from_pred(struct nv50_pc *pc, struct nv50_reg *dst, int pred) +{ + struct nv50_program_exec *e = exec(pc); + + assert(dst->type == P_TEMP); + e->inst[1] = 0x20000000 | (pred << 12); + set_long(pc, e); + set_dst(pc, dst, e); + + emit(pc, e); +} + +static void +emit_mov_to_pred(struct nv50_pc *pc, int pred, struct nv50_reg *src) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0x000001fc; + e->inst[1] = 0xa0000008; + set_long(pc, e); + set_pred_wr(pc, 1, pred, e); + set_src_0_restricted(pc, src, e); + + emit(pc, e); +} + +static void emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1) { @@ -898,7 +975,6 @@ static INLINE void emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1) { - assert(src0 != src1); src1->mod ^= NV50_MOD_NEG; emit_add(pc, dst, src0, src1); src1->mod ^= NV50_MOD_NEG; @@ -967,7 +1043,6 @@ static INLINE void emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0, struct nv50_reg *src1, struct nv50_reg *src2) { - assert(src2 != src0 && src2 != src1); src2->mod ^= NV50_MOD_NEG; emit_mad(pc, dst, src0, src1, src2); src2->mod ^= NV50_MOD_NEG; @@ -1120,7 +1195,6 @@ emit_set(struct nv50_pc *pc, unsigned ccode, struct nv50_reg *dst, int wp, set_src_1(pc, src1, e); emit(pc, e); - pc->if_cond = pc->p->exec_tail; /* record for OPCODE_IF */ /* cvt.f32.u32/s32 (?) if we didn't only write the predicate */ if (rdst) @@ -1242,24 +1316,128 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src) { struct nv50_program_exec *e; const int r_pred = 1; - unsigned cvn = CVT_F32_F32; - - if (src->mod & NV50_MOD_NEG) - cvn |= CVT_NEG; - /* write predicate reg */ - emit_cvt(pc, NULL, src, r_pred, CVTOP_RN, cvn); - /* conditional discard */ e = exec(pc); - e->inst[0] = 0x00000002; + e->inst[0] = 0x00000002; /* discard */ + set_long(pc, e); /* sets cond code to ALWAYS */ + + if (src) { + unsigned cvn = CVT_F32_F32; + + set_pred(pc, 0x1 /* cc = LT */, r_pred, e); + + if (src->mod & NV50_MOD_NEG) + cvn |= CVT_NEG; + /* write predicate reg */ + emit_cvt(pc, NULL, src, r_pred, CVTOP_RN, cvn); + } + + emit(pc, e); +} + +static struct nv50_program_exec * +emit_breakaddr(struct nv50_pc *pc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0x40000002; + set_long(pc, e); + + emit(pc, e); + return e; +} + +static void +emit_break(struct nv50_pc *pc, int pred, unsigned cc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0x50000002; + set_long(pc, e); + if (pred >= 0) + set_pred(pc, cc, pred, e); + + emit(pc, e); +} + +static struct nv50_program_exec * +emit_joinat(struct nv50_pc *pc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0xa0000002; + set_long(pc, e); + + emit(pc, e); + return e; +} + +static struct nv50_program_exec * +emit_branch(struct nv50_pc *pc, int pred, unsigned cc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0x10000002; set_long(pc, e); - set_pred(pc, 0x1 /* LT */, r_pred, e); + if (pred >= 0) + set_pred(pc, cc, pred, e); emit(pc, e); + return pc->p->exec_tail; +} + +static void +emit_ret(struct nv50_pc *pc, int pred, unsigned cc) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0x30000002; + set_long(pc, e); + if (pred >= 0) + set_pred(pc, cc, pred, e); + + emit(pc, e); +} + +#define QOP_ADD 0 +#define QOP_SUBR 1 +#define QOP_SUB 2 +#define QOP_MOV_SRC1 3 + +/* For a quad of threads / top left, top right, bottom left, bottom right + * pixels, do a different operation, and take src0 from a specific thread. + */ +static void +emit_quadop(struct nv50_pc *pc, struct nv50_reg *dst, int wp, int lane_src0, + struct nv50_reg *src0, struct nv50_reg *src1, ubyte qop) +{ + struct nv50_program_exec *e = exec(pc); + + e->inst[0] = 0xc0000000; + e->inst[1] = 0x80000000; + set_long(pc, e); + e->inst[0] |= lane_src0 << 16; + set_src_0(pc, src0, e); + set_src_2(pc, src1, e); + + if (wp >= 0) + set_pred_wr(pc, 1, wp, e); + + if (dst) + set_dst(pc, dst, e); + else { + e->inst[0] |= 0x000001fc; + e->inst[1] |= 0x00000008; + } + + e->inst[0] |= (qop & 3) << 20; + e->inst[1] |= (qop >> 2) << 22; + + emit(pc, e); } static void load_cube_tex_coords(struct nv50_pc *pc, struct nv50_reg *t[4], - struct nv50_reg **src, boolean proj) + struct nv50_reg **src, unsigned arg, boolean proj) { int mod[3] = { src[0]->mod, src[1]->mod, src[2]->mod }; @@ -1276,6 +1454,10 @@ load_cube_tex_coords(struct nv50_pc *pc, struct nv50_reg *t[4], if (proj && 0 /* looks more correct without this */) emit_mul(pc, t[2], t[2], src[3]); + else + if (arg == 4) /* there is no textureProj(samplerCubeShadow) */ + emit_mov(pc, t[3], src[3]); + emit_flop(pc, 0, t[2], t[2]); emit_mul(pc, t[0], src[0], t[2]); @@ -1284,89 +1466,214 @@ load_cube_tex_coords(struct nv50_pc *pc, struct nv50_reg *t[4], } static void -emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, - struct nv50_reg **src, unsigned unit, unsigned type, boolean proj) +load_proj_tex_coords(struct nv50_pc *pc, struct nv50_reg *t[4], + struct nv50_reg **src, unsigned dim, unsigned arg) { - struct nv50_reg *t[4]; - struct nv50_program_exec *e; + unsigned c, mode; + + if (src[0]->type == P_TEMP && src[0]->rhw != -1) { + mode = pc->interp_mode[src[0]->index] | INTERP_PERSPECTIVE; - unsigned c, mode, dim; + t[3]->rhw = src[3]->rhw; + emit_interp(pc, t[3], NULL, (mode & INTERP_CENTROID)); + emit_flop(pc, 0, t[3], t[3]); + for (c = 0; c < dim; ++c) { + t[c]->rhw = src[c]->rhw; + emit_interp(pc, t[c], t[3], mode); + } + if (arg != dim) { /* depth reference value */ + t[dim]->rhw = src[2]->rhw; + emit_interp(pc, t[dim], t[3], mode); + } + } else { + /* XXX: for some reason the blob sometimes uses MAD + * (mad f32 $rX $rY $rZ neg $r63) + */ + emit_flop(pc, 0, t[3], src[3]); + for (c = 0; c < dim; ++c) + emit_mul(pc, t[c], src[c], t[3]); + if (arg != dim) /* depth reference value */ + emit_mul(pc, t[dim], src[2], t[3]); + } +} + +static INLINE void +get_tex_dim(unsigned type, unsigned *dim, unsigned *arg) +{ switch (type) { case TGSI_TEXTURE_1D: - dim = 1; + *arg = *dim = 1; + break; + case TGSI_TEXTURE_SHADOW1D: + *dim = 1; + *arg = 2; break; case TGSI_TEXTURE_UNKNOWN: case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_SHADOW1D: /* XXX: x, z */ case TGSI_TEXTURE_RECT: - dim = 2; + *arg = *dim = 2; + break; + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + *dim = 2; + *arg = 3; break; case TGSI_TEXTURE_3D: case TGSI_TEXTURE_CUBE: - case TGSI_TEXTURE_SHADOW2D: - case TGSI_TEXTURE_SHADOWRECT: /* XXX */ - dim = 3; + *dim = *arg = 3; break; default: assert(0); break; } +} - /* some cards need t[0]'s hw index to be a multiple of 4 */ - alloc_temp4(pc, t, 0); +/* We shouldn't execute TEXLOD if any of the pixels in a quad have + * different LOD values, so branch off groups of equal LOD. + */ +static void +emit_texlod_sequence(struct nv50_pc *pc, struct nv50_reg *tlod, + struct nv50_reg *src, struct nv50_program_exec *tex) +{ + struct nv50_program_exec *join_at; + unsigned i, target = pc->p->exec_size + 7 * 2; - if (type == TGSI_TEXTURE_CUBE) { - load_cube_tex_coords(pc, t, src, proj); - } else - if (proj) { - if (src[0]->type == P_TEMP && src[0]->rhw != -1) { - mode = pc->interp_mode[src[0]->index]; - - t[3]->rhw = src[3]->rhw; - emit_interp(pc, t[3], NULL, (mode & INTERP_CENTROID)); - emit_flop(pc, 0, t[3], t[3]); - - for (c = 0; c < dim; c++) { - t[c]->rhw = src[c]->rhw; - emit_interp(pc, t[c], t[3], - (mode | INTERP_PERSPECTIVE)); - } - } else { - emit_flop(pc, 0, t[3], src[3]); - for (c = 0; c < dim; c++) - emit_mul(pc, t[c], src[c], t[3]); + /* Subtract lod of each pixel from lod of top left pixel, jump + * texlod insn if result is 0, then repeat for 2 other pixels. + */ + join_at = emit_joinat(pc); + emit_quadop(pc, NULL, 0, 0, tlod, tlod, 0x55); + emit_branch(pc, 0, 2)->param.index = target; - /* XXX: for some reason the blob sometimes uses MAD: - * emit_mad(pc, t[c], src[0][c], t[3], t[3]) - * pc->p->exec_tail->inst[1] |= 0x080fc000; - */ + for (i = 1; i < 4; ++i) { + emit_quadop(pc, NULL, 0, i, tlod, tlod, 0x55); + emit_branch(pc, 0, 2)->param.index = target; + } + + emit_mov(pc, tlod, src); /* target */ + emit(pc, tex); /* texlod */ + + join_at->param.index = target + 2 * 2; + JOIN_ON(emit_nop(pc)); /* join _after_ tex */ +} + +static void +emit_texbias_sequence(struct nv50_pc *pc, struct nv50_reg *t[4], unsigned arg, + struct nv50_program_exec *tex) +{ + struct nv50_program_exec *e; + struct nv50_reg imm_1248, *t123[4][4], *r_bits = alloc_temp(pc, NULL); + int r_pred = 0; + unsigned n, c, i, cc[4] = { 0x0a, 0x13, 0x11, 0x10 }; + + pc->allow32 = FALSE; + ctor_reg(&imm_1248, P_IMMD, -1, ctor_immd_4u32(pc, 1, 2, 4, 8) * 4); + + /* Subtract bias value of thread i from bias values of each thread, + * store result in r_pred, and set bit i in r_bits if result was 0. + */ + assert(arg < 4); + for (i = 0; i < 4; ++i, ++imm_1248.hw) { + emit_quadop(pc, NULL, r_pred, i, t[arg], t[arg], 0x55); + emit_mov(pc, r_bits, &imm_1248); + set_pred(pc, 2, r_pred, pc->p->exec_tail); + } + emit_mov_to_pred(pc, r_pred, r_bits); + + /* The lanes of a quad are now grouped by the bit in r_pred they have + * set. Put the input values for TEX into a new register set for each + * group and execute TEX only for a specific group. + * We cannot use the same register set for each group because we need + * the derivatives, which are implicitly calculated, to be correct. + */ + for (i = 1; i < 4; ++i) { + alloc_temp4(pc, t123[i], 0); + + for (c = 0; c <= arg; ++c) + emit_mov(pc, t123[i][c], t[c]); + + *(e = exec(pc)) = *(tex); + e->inst[0] &= ~0x01fc; + set_dst(pc, t123[i][0], e); + set_pred(pc, cc[i], r_pred, e); + emit(pc, e); + } + /* finally TEX on the original regs (where we kept the input) */ + set_pred(pc, cc[0], r_pred, tex); + emit(pc, tex); + + /* put the 3 * n other results into regs for lane 0 */ + n = popcnt4(((e->inst[0] >> 25) & 0x3) | ((e->inst[1] >> 12) & 0xc)); + for (i = 1; i < 4; ++i) { + for (c = 0; c < n; ++c) { + emit_mov(pc, t[c], t123[i][c]); + set_pred(pc, cc[i], r_pred, pc->p->exec_tail); } - } else { - for (c = 0; c < dim; c++) - emit_mov(pc, t[c], src[c]); + free_temp4(pc, t123[i]); } + emit_nop(pc); + free_temp(pc, r_bits); +} + +static void +emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, + struct nv50_reg **src, unsigned unit, unsigned type, + boolean proj, int bias_lod) +{ + struct nv50_reg *t[4]; + struct nv50_program_exec *e; + unsigned c, dim, arg; + + /* t[i] must be within a single 128 bit super-reg */ + alloc_temp4(pc, t, 0); + e = exec(pc); + e->inst[0] = 0xf0000000; set_long(pc, e); - e->inst[0] |= 0xf0000000; - e->inst[1] |= 0x00000004; set_dst(pc, t[0], e); - e->inst[0] |= (unit << 9); - if (dim == 2) - e->inst[0] |= 0x00400000; - else - if (dim == 3) { - e->inst[0] |= 0x00800000; - if (type == TGSI_TEXTURE_CUBE) - e->inst[0] |= 0x08000000; + /* TIC and TSC binding indices (TSC is ignored as TSC_LINKED = TRUE): */ + e->inst[0] |= (unit << 9) /* | (unit << 17) */; + + /* live flag (don't set if TEX results affect input to another TEX): */ + /* e->inst[0] |= 0x00000004; */ + + get_tex_dim(type, &dim, &arg); + + if (type == TGSI_TEXTURE_CUBE) { + e->inst[0] |= 0x08000000; + load_cube_tex_coords(pc, t, src, arg, proj); + } else + if (proj) + load_proj_tex_coords(pc, t, src, dim, arg); + else { + for (c = 0; c < dim; c++) + emit_mov(pc, t[c], src[c]); + if (arg != dim) /* depth reference value (always src.z here) */ + emit_mov(pc, t[dim], src[2]); } e->inst[0] |= (mask & 0x3) << 25; e->inst[1] |= (mask & 0xc) << 12; - emit(pc, e); + if (!bias_lod) { + e->inst[0] |= (arg - 1) << 22; + emit(pc, e); + } else + if (bias_lod < 0) { + e->inst[0] |= arg << 22; + e->inst[1] |= 0x20000000; /* texbias */ + emit_mov(pc, t[arg], src[3]); + emit_texbias_sequence(pc, t, arg, e); + } else { + e->inst[0] |= arg << 22; + e->inst[1] |= 0x40000000; /* texlod */ + emit_mov(pc, t[arg], src[3]); + emit_texlod_sequence(pc, t[arg], src[3], e); + } + #if 1 c = 0; if (mask & 1) emit_mov(pc, dst[0], t[c++]); @@ -1389,46 +1696,14 @@ emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, } static void -emit_branch(struct nv50_pc *pc, int pred, unsigned cc, - struct nv50_program_exec **join) -{ - struct nv50_program_exec *e = exec(pc); - - if (join) { - set_long(pc, e); - e->inst[0] |= 0xa0000002; - emit(pc, e); - *join = e; - e = exec(pc); - } - - set_long(pc, e); - e->inst[0] |= 0x10000002; - if (pred >= 0) - set_pred(pc, cc, pred, e); - emit(pc, e); -} - -static void -emit_nop(struct nv50_pc *pc) -{ - struct nv50_program_exec *e = exec(pc); - - e->inst[0] = 0xf0000000; - set_long(pc, e); - e->inst[1] = 0xe0000000; - emit(pc, e); -} - -static void emit_ddx(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) { struct nv50_program_exec *e = exec(pc); assert(src->type == P_TEMP); - e->inst[0] = 0xc0140000; - e->inst[1] = 0x89800000; + e->inst[0] = (src->mod & NV50_MOD_NEG) ? 0xc0240000 : 0xc0140000; + e->inst[1] = (src->mod & NV50_MOD_NEG) ? 0x86400000 : 0x89800000; set_long(pc, e); set_dst(pc, dst, e); set_src_0(pc, src, e); @@ -1440,25 +1715,16 @@ emit_ddx(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) static void emit_ddy(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src) { - struct nv50_reg *r = src; struct nv50_program_exec *e = exec(pc); assert(src->type == P_TEMP); - if (!(src->mod & NV50_MOD_NEG)) { /* ! double negation */ - r = alloc_temp(pc, NULL); - emit_neg(pc, r, src); - } - - e->inst[0] = 0xc0150000; - e->inst[1] = 0x8a400000; + e->inst[0] = (src->mod & NV50_MOD_NEG) ? 0xc0250000 : 0xc0150000; + e->inst[1] = (src->mod & NV50_MOD_NEG) ? 0x85800000 : 0x8a400000; set_long(pc, e); set_dst(pc, dst, e); - set_src_0(pc, r, e); - set_src_2(pc, r, e); - - if (r != src) - free_temp(pc, r); + set_src_0(pc, src, e); + set_src_2(pc, src, e); emit(pc, e); } @@ -1515,9 +1781,8 @@ convert_to_long(struct nv50_pc *pc, struct nv50_program_exec *e) static boolean negate_supported(const struct tgsi_full_instruction *insn, int i) { - int s; - switch (insn->Instruction.Opcode) { + case TGSI_OPCODE_DDX: case TGSI_OPCODE_DDY: case TGSI_OPCODE_DP3: case TGSI_OPCODE_DP4: @@ -1526,29 +1791,14 @@ negate_supported(const struct tgsi_full_instruction *insn, int i) case TGSI_OPCODE_ADD: case TGSI_OPCODE_SUB: case TGSI_OPCODE_MAD: - break; + return TRUE; case TGSI_OPCODE_POW: if (i == 1) - break; + return TRUE; return FALSE; default: return FALSE; } - - /* Watch out for possible multiple uses of an nv50_reg, we - * can't use nv50_reg::neg in these cases. - */ - for (s = 0; s < insn->Instruction.NumSrcRegs; ++s) { - if (s == i) - continue; - if ((insn->Src[s].Register.Index == - insn->Src[i].Register.Index) && - (insn->Src[s].Register.File == - insn->Src[i].Register.File)) - return FALSE; - } - - return TRUE; } /* Return a read mask for source registers deduced from opcode & write mask. */ @@ -1576,9 +1826,13 @@ nv50_tgsi_src_mask(const struct tgsi_full_instruction *insn, int c) case TGSI_OPCODE_RSQ: case TGSI_OPCODE_SCS: return 0x1; + case TGSI_OPCODE_IF: + return 0x1; case TGSI_OPCODE_LIT: return 0xb; case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXL: case TGSI_OPCODE_TXP: { const struct tgsi_instruction_texture *tex; @@ -1587,13 +1841,17 @@ nv50_tgsi_src_mask(const struct tgsi_full_instruction *insn, int c) tex = &insn->Texture; mask = 0x7; - if (insn->Instruction.Opcode == TGSI_OPCODE_TXP) - mask |= 0x8; + if (insn->Instruction.Opcode != TGSI_OPCODE_TEX && + insn->Instruction.Opcode != TGSI_OPCODE_TXD) + mask |= 0x8; /* bias, lod or proj */ switch (tex->Texture) { case TGSI_TEXTURE_1D: mask &= 0x9; break; + case TGSI_TEXTURE_SHADOW1D: + mask &= 0x5; + break; case TGSI_TEXTURE_2D: mask &= 0xb; break; @@ -1676,7 +1934,7 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src, /* Indicate indirection by setting r->acc < 0 and * use the index field to select the address reg. */ - r = MALLOC_STRUCT(nv50_reg); + r = reg_instance(pc, NULL); swz = tgsi_util_get_src_register_swizzle( &src->Indirect, 0); ctor_reg(r, P_CONST, @@ -1730,6 +1988,8 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src, break; } + if (r && r->acc >= 0 && r != temp) + return reg_instance(pc, r); return r; } @@ -1785,6 +2045,8 @@ nv50_tgsi_dst_revdep(unsigned op, int s, int c) case TGSI_OPCODE_LIT: case TGSI_OPCODE_SCS: case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXL: case TGSI_OPCODE_TXP: /* these take care of dangerous swizzles themselves */ return 0x0; @@ -1826,28 +2088,43 @@ nv50_kill_branch(struct nv50_pc *pc) */ if (has_pred(pc->if_insn[lvl], 0xf)) return FALSE; - assert(pc->if_insn[lvl] && pc->br_join[lvl]); + assert(pc->if_insn[lvl] && pc->if_join[lvl]); - /* We'll use the exec allocated for JOIN_AT (as we can't easily - * update prev's next); if exec_tail is BRK, update the pointer. + /* We'll use the exec allocated for JOIN_AT (we can't easily + * access nv50_program_exec's prev). */ - if (pc->loop_lvl && pc->br_loop[pc->loop_lvl - 1] == pc->p->exec_tail) - pc->br_loop[pc->loop_lvl - 1] = pc->br_join[lvl]; - pc->p->exec_size -= 4; /* remove JOIN_AT and BRA */ - *pc->br_join[lvl] = *pc->p->exec_tail; + *pc->if_join[lvl] = *pc->p->exec_tail; FREE(pc->if_insn[lvl]); FREE(pc->p->exec_tail); - pc->p->exec_tail = pc->br_join[lvl]; + pc->p->exec_tail = pc->if_join[lvl]; pc->p->exec_tail->next = NULL; set_pred(pc, 0xd, 0, pc->p->exec_tail); return TRUE; } +static void +nv50_fp_move_results(struct nv50_pc *pc) +{ + struct nv50_reg reg; + unsigned i; + + ctor_reg(®, P_TEMP, -1, -1); + + for (i = 0; i < pc->result_nr * 4; ++i) { + if (pc->result[i].rhw < 0 || pc->result[i].hw < 0) + continue; + if (pc->result[i].rhw != pc->result[i].hw) { + reg.hw = pc->result[i].rhw; + emit_mov(pc, ®, &pc->result[i]); + } + } +} + static boolean nv50_program_tx_insn(struct nv50_pc *pc, const struct tgsi_full_instruction *inst) @@ -1934,13 +2211,13 @@ nv50_program_tx_insn(struct nv50_pc *pc, emit_arl(pc, dst[0], temp, 4); break; case TGSI_OPCODE_BGNLOOP: + pc->loop_brka[pc->loop_lvl] = emit_breakaddr(pc); pc->loop_pos[pc->loop_lvl++] = pc->p->exec_size; terminate_mbb(pc); break; case TGSI_OPCODE_BRK: - emit_branch(pc, -1, 0, NULL); assert(pc->loop_lvl > 0); - pc->br_loop[pc->loop_lvl - 1] = pc->p->exec_tail; + emit_break(pc, -1, 0); break; case TGSI_OPCODE_CEIL: for (c = 0; c < 4; c++) { @@ -2016,7 +2293,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, emit_mov_immdval(pc, dst[0], 1.0f); break; case TGSI_OPCODE_ELSE: - emit_branch(pc, -1, 0, NULL); + emit_branch(pc, -1, 0); pc->if_insn[--pc->if_lvl]->param.index = pc->p->exec_size; pc->if_insn[pc->if_lvl++] = pc->p->exec_tail; terminate_mbb(pc); @@ -2028,21 +2305,20 @@ nv50_program_tx_insn(struct nv50_pc *pc, if (nv50_kill_branch(pc) == TRUE) break; - if (pc->br_join[pc->if_lvl]) { - pc->br_join[pc->if_lvl]->param.index = pc->p->exec_size; - pc->br_join[pc->if_lvl] = NULL; + if (pc->if_join[pc->if_lvl]) { + pc->if_join[pc->if_lvl]->param.index = pc->p->exec_size; + pc->if_join[pc->if_lvl] = NULL; } terminate_mbb(pc); /* emit a NOP as join point, we could set it on the next * one, but would have to make sure it is long and !immd */ - emit_nop(pc); - pc->p->exec_tail->inst[1] |= 2; + JOIN_ON(emit_nop(pc)); break; case TGSI_OPCODE_ENDLOOP: - emit_branch(pc, -1, 0, NULL); - pc->p->exec_tail->param.index = pc->loop_pos[--pc->loop_lvl]; - pc->br_loop[pc->loop_lvl]->param.index = pc->p->exec_size; + emit_branch(pc, -1, 0)->param.index = + pc->loop_pos[--pc->loop_lvl]; + pc->loop_brka[pc->loop_lvl]->param.index = pc->p->exec_size; terminate_mbb(pc); break; case TGSI_OPCODE_EX2: @@ -2066,21 +2342,23 @@ nv50_program_tx_insn(struct nv50_pc *pc, } break; case TGSI_OPCODE_IF: - /* emitting a join_at may not be necessary */ - assert(pc->if_lvl < MAX_IF_DEPTH); - /* set_pred_wr(pc, 1, 0, pc->if_cond); */ + assert(pc->if_lvl < NV50_MAX_COND_NESTING); emit_cvt(pc, NULL, src[0][0], 0, CVTOP_ABS | CVTOP_RN, CVT_F32_F32); - emit_branch(pc, 0, 2, &pc->br_join[pc->if_lvl]); - pc->if_insn[pc->if_lvl++] = pc->p->exec_tail; + pc->if_join[pc->if_lvl] = emit_joinat(pc); + pc->if_insn[pc->if_lvl++] = emit_branch(pc, 0, 2);; terminate_mbb(pc); break; case TGSI_OPCODE_KIL: + assert(src[0][0] && src[0][1] && src[0][2] && src[0][3]); emit_kil(pc, src[0][0]); emit_kil(pc, src[0][1]); emit_kil(pc, src[0][2]); emit_kil(pc, src[0][3]); break; + case TGSI_OPCODE_KILP: + emit_kil(pc, NULL); + break; case TGSI_OPCODE_LIT: emit_lit(pc, &dst[0], mask, &src[0][0]); break; @@ -2137,6 +2415,11 @@ nv50_program_tx_insn(struct nv50_pc *pc, case TGSI_OPCODE_RCP: emit_flop(pc, 0, brdc, src[0][0]); break; + case TGSI_OPCODE_RET: + if (pc->p->type == PIPE_SHADER_FRAGMENT) + nv50_fp_move_results(pc); + emit_ret(pc, -1, 0); + break; case TGSI_OPCODE_RSQ: emit_flop(pc, 2, brdc, src[0][0]); break; @@ -2187,11 +2470,19 @@ nv50_program_tx_insn(struct nv50_pc *pc, break; case TGSI_OPCODE_TEX: emit_tex(pc, dst, mask, src[0], unit, - inst->Texture.Texture, FALSE); + inst->Texture.Texture, FALSE, 0); + break; + case TGSI_OPCODE_TXB: + emit_tex(pc, dst, mask, src[0], unit, + inst->Texture.Texture, FALSE, -1); + break; + case TGSI_OPCODE_TXL: + emit_tex(pc, dst, mask, src[0], unit, + inst->Texture.Texture, FALSE, 1); break; case TGSI_OPCODE_TXP: emit_tex(pc, dst, mask, src[0], unit, - inst->Texture.Texture, TRUE); + inst->Texture.Texture, TRUE, 0); break; case TGSI_OPCODE_TRUNC: for (c = 0; c < 4; c++) { @@ -2245,20 +2536,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, } } - for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - for (c = 0; c < 4; c++) { - if (!src[i][c]) - continue; - src[i][c]->mod = 0; - if (src[i][c]->index == -1 && src[i][c]->type == P_IMMD) - FREE(src[i][c]); - else - if (src[i][c]->acc < 0 && src[i][c]->type == P_CONST) - FREE(src[i][c]); /* indirect constant */ - } - } - kill_temp_temp(pc); + pc->reg_instance_nr = 0; + return TRUE; } @@ -2541,10 +2821,10 @@ nv50_program_tx_prep(struct nv50_pc *pc) const struct tgsi_full_immediate *imm = &tp.FullToken.FullImmediate; - ctor_immd(pc, imm->u[0].Float, - imm->u[1].Float, - imm->u[2].Float, - imm->u[3].Float); + ctor_immd_4f32(pc, imm->u[0].Float, + imm->u[1].Float, + imm->u[2].Float, + imm->u[3].Float); } break; case TGSI_TOKEN_TYPE_DECLARATION: @@ -2898,24 +3178,6 @@ ctor_nv50_pc(struct nv50_pc *pc, struct nv50_program *p) } static void -nv50_fp_move_results(struct nv50_pc *pc) -{ - struct nv50_reg reg; - unsigned i; - - ctor_reg(®, P_TEMP, -1, -1); - - for (i = 0; i < pc->result_nr * 4; ++i) { - if (pc->result[i].rhw < 0 || pc->result[i].hw < 0) - continue; - if (pc->result[i].rhw != pc->result[i].hw) { - reg.hw = pc->result[i].rhw; - emit_mov(pc, ®, &pc->result[i]); - } - } -} - -static void nv50_program_fixup_insns(struct nv50_pc *pc) { struct nv50_program_exec *e, **bra_list; @@ -3024,7 +3286,7 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p) } static void -nv50_program_upload_data(struct nv50_context *nv50, float *map, +nv50_program_upload_data(struct nv50_context *nv50, uint32_t *map, unsigned start, unsigned count, unsigned cbuf) { struct nouveau_channel *chan = nv50->screen->base.channel; @@ -3072,8 +3334,8 @@ nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p) if (p->param_nr) { unsigned cb; - float *map = pipe_buffer_map(pscreen, nv50->constbuf[p->type], - PIPE_BUFFER_USAGE_CPU_READ); + uint32_t *map = pipe_buffer_map(pscreen, nv50->constbuf[p->type], + PIPE_BUFFER_USAGE_CPU_READ); if (p->type == PIPE_SHADER_VERTEX) cb = NV50_CB_PVP; diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h index 255c7c737ef..4a90c372ce3 100644 --- a/src/gallium/drivers/nv50/nv50_program.h +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -37,7 +37,7 @@ struct nv50_program { struct nouveau_bo *bo; - float *immd; + uint32_t *immd; unsigned immd_nr; unsigned param_nr; diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 5305c93d59a..268c9823f7d 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -93,7 +93,7 @@ nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq) struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_query *q = nv50_query(pq); - WAIT_RING (chan, 5); + MARK_RING (chan, 5, 2); /* flush on lack of space or relocs */ BEGIN_RING(chan, tesla, NV50TCL_QUERY_ADDRESS_HIGH, 4); OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index e1b2f11239a..d443ca3ad06 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -76,6 +76,7 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen, case PIPE_FORMAT_DXT3_RGBA: case PIPE_FORMAT_DXT5_RGBA: case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_Z32_FLOAT: case PIPE_FORMAT_R16G16B16A16_SNORM: case PIPE_FORMAT_R16G16B16A16_UNORM: @@ -97,6 +98,10 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param) switch (param) { case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: return 32; + case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: + return 32; + case PIPE_CAP_MAX_COMBINED_SAMPLERS: + return 64; case PIPE_CAP_NPOT_TEXTURES: return 1; case PIPE_CAP_TWO_SIDED_STENCIL: @@ -122,8 +127,6 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param) case PIPE_CAP_TEXTURE_MIRROR_CLAMP: case PIPE_CAP_TEXTURE_MIRROR_REPEAT: return 1; - case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: - return 0; case PIPE_CAP_TGSI_CONT_SUPPORTED: return 0; case PIPE_CAP_BLEND_EQUATION_SEPARATE: @@ -315,6 +318,9 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_method(so, screen->tesla, 0x1400, 1); so_data (so, 0xf); + /* max TIC (bits 4:8) & TSC (ignored) bindings, per program type */ + so_method(so, screen->tesla, 0x13b4, 1); + so_data (so, 0x54); so_method(so, screen->tesla, 0x13bc, 1); so_data (so, 0x54); /* origin is top left (set to 1 for bottom left) */ @@ -387,7 +393,8 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x00000131 | (NV50_CB_PFP << 12)); - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 64*8*4, &screen->tic); + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, PIPE_SHADER_TYPES*32*32, + &screen->tic); if (ret) { nv50_screen_destroy(pscreen); return NULL; @@ -398,9 +405,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); - so_data (so, 0x000007ff); + so_data (so, PIPE_SHADER_TYPES * 32 - 1); - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 64*8*4, &screen->tsc); + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, PIPE_SHADER_TYPES*32*32, + &screen->tsc); if (ret) { nv50_screen_destroy(pscreen); return NULL; @@ -411,7 +419,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); - so_data (so, 0x00000000); + so_data (so, 0x00000000); /* ignored if TSC_LINKED (0x1234) = 1 */ /* Vertex array limits - max them out */ @@ -425,6 +433,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_data (so, fui(0.0)); so_data (so, fui(1.0)); + /* no dynamic combination of TIC & TSC entries => only BIND_TIC used */ so_method(so, screen->tesla, 0x1234, 1); so_data (so, 1); diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 07318f23947..88aef52d08c 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -196,8 +196,9 @@ nv50_sampler_state_create(struct pipe_context *pipe, } if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - tsc[0] |= (1 << 8); - tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7); + /* XXX: must be deactivated for non-shadow textures */ + tsc[0] |= (1 << 9); + tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7) << 10; } limit = CLAMP(cso->lod_bias, -16.0, 15.0); @@ -215,41 +216,66 @@ nv50_sampler_state_create(struct pipe_context *pipe, return (void *)sso; } -static void -nv50_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler) +static INLINE void +nv50_sampler_state_bind(struct pipe_context *pipe, unsigned type, + unsigned nr, void **sampler) { struct nv50_context *nv50 = nv50_context(pipe); - int i; - nv50->sampler_nr = nr; - for (i = 0; i < nv50->sampler_nr; i++) - nv50->sampler[i] = sampler[i]; + memcpy(nv50->sampler[type], sampler, nr * sizeof(void *)); + nv50->sampler_nr[type] = nr; nv50->dirty |= NV50_NEW_SAMPLER; } static void +nv50_vp_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **s) +{ + nv50_sampler_state_bind(pipe, PIPE_SHADER_VERTEX, nr, s); +} + +static void +nv50_fp_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **s) +{ + nv50_sampler_state_bind(pipe, PIPE_SHADER_FRAGMENT, nr, s); +} + +static void nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso) { FREE(hwcso); } -static void -nv50_set_sampler_texture(struct pipe_context *pipe, unsigned nr, - struct pipe_texture **pt) +static INLINE void +nv50_set_sampler_texture(struct pipe_context *pipe, unsigned type, + unsigned nr, struct pipe_texture **pt) { struct nv50_context *nv50 = nv50_context(pipe); - int i; + unsigned i; for (i = 0; i < nr; i++) - pipe_texture_reference((void *)&nv50->miptree[i], pt[i]); - for (i = nr; i < nv50->miptree_nr; i++) - pipe_texture_reference((void *)&nv50->miptree[i], NULL); + pipe_texture_reference((void *)&nv50->miptree[type][i], pt[i]); + for (i = nr; i < nv50->miptree_nr[type]; i++) + pipe_texture_reference((void *)&nv50->miptree[type][i], NULL); - nv50->miptree_nr = nr; + nv50->miptree_nr[type] = nr; nv50->dirty |= NV50_NEW_TEXTURE; } +static void +nv50_set_vp_sampler_textures(struct pipe_context *pipe, + unsigned nr, struct pipe_texture **pt) +{ + nv50_set_sampler_texture(pipe, PIPE_SHADER_VERTEX, nr, pt); +} + +static void +nv50_set_fp_sampler_textures(struct pipe_context *pipe, + unsigned nr, struct pipe_texture **pt) +{ + nv50_set_sampler_texture(pipe, PIPE_SHADER_FRAGMENT, nr, pt); +} + static void * nv50_rasterizer_state_create(struct pipe_context *pipe, const struct pipe_rasterizer_state *cso) @@ -648,9 +674,11 @@ nv50_init_state_functions(struct nv50_context *nv50) nv50->pipe.delete_blend_state = nv50_blend_state_delete; nv50->pipe.create_sampler_state = nv50_sampler_state_create; - nv50->pipe.bind_fragment_sampler_states = nv50_sampler_state_bind; nv50->pipe.delete_sampler_state = nv50_sampler_state_delete; - nv50->pipe.set_fragment_sampler_textures = nv50_set_sampler_texture; + nv50->pipe.bind_fragment_sampler_states = nv50_fp_sampler_state_bind; + nv50->pipe.bind_vertex_sampler_states = nv50_vp_sampler_state_bind; + nv50->pipe.set_fragment_sampler_textures = nv50_set_fp_sampler_textures; + nv50->pipe.set_vertex_sampler_textures = nv50_set_vp_sampler_textures; nv50->pipe.create_rasterizer_state = nv50_rasterizer_state_create; nv50->pipe.bind_rasterizer_state = nv50_rasterizer_state_bind; diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index c871acaab8d..871e8097b65 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -156,6 +156,30 @@ nv50_state_validate_fb(struct nv50_context *nv50) } static void +nv50_validate_samplers(struct nv50_context *nv50, struct nouveau_stateobj *so, + unsigned p) +{ + struct nouveau_grobj *eng2d = nv50->screen->eng2d; + unsigned i, j, dw = nv50->sampler_nr[p] * 8; + + if (!dw) + return; + nv50_so_init_sifc(nv50, so, nv50->screen->tsc, NOUVEAU_BO_VRAM, + p * (32 * 8 * 4), dw * 4); + + so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), dw); + + for (i = 0; i < nv50->sampler_nr[p]; ++i) { + if (nv50->sampler[p][i]) + so_datap(so, nv50->sampler[p][i]->tsc, 8); + else { + for (j = 0; j < 8; ++j) /* you get punished */ + so_data(so, 0); /* ... for leaving holes */ + } + } +} + +static void nv50_state_emit(struct nv50_context *nv50) { struct nv50_screen *screen = nv50->screen; @@ -246,7 +270,6 @@ boolean nv50_state_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; - struct nouveau_grobj *eng2d = nv50->screen->eng2d; struct nouveau_stateobj *so; unsigned i; @@ -369,22 +392,16 @@ scissor_uptodate: viewport_uptodate: if (nv50->dirty & NV50_NEW_SAMPLER) { - unsigned i; + unsigned nr = 0; - so = so_new(nv50->sampler_nr * 9 + 23 + 4, 2); + for (i = 0; i < PIPE_SHADER_TYPES; ++i) + nr += nv50->sampler_nr[i]; - nv50_so_init_sifc(nv50, so, nv50->screen->tsc, NOUVEAU_BO_VRAM, - nv50->sampler_nr * 8 * 4); + so = so_new(nr * 8 + 24 * PIPE_SHADER_TYPES + 2, 4); - for (i = 0; i < nv50->sampler_nr; i++) { - if (!nv50->sampler[i]) - continue; - so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), 8); - so_datap (so, nv50->sampler[i]->tsc, 8); - } + nv50_validate_samplers(nv50, so, PIPE_SHADER_VERTEX); + nv50_validate_samplers(nv50, so, PIPE_SHADER_FRAGMENT); - so_method(so, tesla, 0x1440, 1); /* sync SIFC */ - so_data (so, 0); so_method(so, tesla, 0x1334, 1); /* flush TSC */ so_data (so, 0); @@ -407,10 +424,13 @@ viewport_uptodate: void nv50_so_init_sifc(struct nv50_context *nv50, struct nouveau_stateobj *so, - struct nouveau_bo *bo, unsigned reloc, unsigned size) + struct nouveau_bo *bo, unsigned reloc, + unsigned offset, unsigned size) { struct nouveau_grobj *eng2d = nv50->screen->eng2d; + reloc |= NOUVEAU_BO_WR; + so_method(so, eng2d, NV50_2D_DST_FORMAT, 2); so_data (so, NV50_2D_DST_FORMAT_R8_UNORM); so_data (so, 1); @@ -418,8 +438,8 @@ void nv50_so_init_sifc(struct nv50_context *nv50, so_data (so, 262144); so_data (so, 65536); so_data (so, 1); - so_reloc (so, bo, 0, reloc | NOUVEAU_BO_WR | NOUVEAU_BO_HIGH, 0, 0); - so_reloc (so, bo, 0, reloc | NOUVEAU_BO_WR | NOUVEAU_BO_LOW, 0, 0); + so_reloc (so, bo, offset, reloc | NOUVEAU_BO_HIGH, 0, 0); + so_reloc (so, bo, offset, reloc | NOUVEAU_BO_LOW, 0, 0); so_method(so, eng2d, NV50_2D_SIFC_UNK0800, 2); so_data (so, 0); so_data (so, NV50_2D_SIFC_FORMAT_R8_UNORM); diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 6bf6f773b0c..79655fc08d5 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -62,6 +62,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) return 1; if (!bo->tile_flags) { + MARK_RING (chan, 9, 2); /* flush on lack of space or relocs */ BEGIN_RING(chan, eng2d, mthd, 2); OUT_RING (chan, format); OUT_RING (chan, 1); @@ -72,6 +73,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) OUT_RELOCh(chan, bo, ps->offset, flags); OUT_RELOCl(chan, bo, ps->offset, flags); } else { + MARK_RING (chan, 11, 2); /* flush on lack of space or relocs */ BEGIN_RING(chan, eng2d, mthd, 5); OUT_RING (chan, format); OUT_RING (chan, 0); diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 417d3679422..c4ca096d6ac 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -25,6 +25,8 @@ #include "nouveau/nouveau_stateobj.h" +#include "util/u_format.h" + #define _MIXED(pf, t0, t1, t2, t3, cr, cg, cb, ca, f) \ { \ PIPE_FORMAT_##pf, \ @@ -68,6 +70,7 @@ static const struct nv50_texture_format nv50_tex_format_list[] = _(DXT5_RGBA, UNORM, C0, C1, C2, C3, DXT5), _MIXED(Z24S8_UNORM, UINT, UNORM, UINT, UINT, C1, C1, C1, ONE, 24_8), + _MIXED(S8Z24_UNORM, UNORM, UINT, UINT, UINT, C0, C0, C0, ONE, 8_24), _(R16G16B16A16_SNORM, UNORM, C0, C1, C2, C3, 16_16_16_16), _(R16G16B16A16_UNORM, SNORM, C0, C1, C2, C3, 16_16_16_16), @@ -85,10 +88,11 @@ static const struct nv50_texture_format nv50_tex_format_list[] = static int nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, - struct nv50_miptree *mt, int unit) + struct nv50_miptree *mt, int unit, unsigned p) { unsigned i; uint32_t mode; + const struct util_format_description *desc; for (i = 0; i < NV50_TEX_FORMAT_LIST_SIZE; i++) if (nv50_tex_format_list[i].pf == mt->base.base.format) @@ -96,7 +100,7 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, if (i == NV50_TEX_FORMAT_LIST_SIZE) return 1; - if (nv50->sampler[unit]->normalized) + if (nv50->sampler[p][unit]->normalized) mode = 0x50001000 | (1 << 31); else { mode = 0x50001000 | (7 << 14); @@ -106,7 +110,10 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, mode |= ((mt->base.bo->tile_mode & 0x0f) << 22) | ((mt->base.bo->tile_mode & 0xf0) << 21); - if (pf_type(mt->base.base.format) == PIPE_FORMAT_TYPE_SRGB) + desc = util_format_description(mt->base.base.format); + assert(desc); + + if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) mode |= 0x0400; switch (mt->base.base.target) { @@ -140,48 +147,78 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, return 0; } -void -nv50_tex_validate(struct nv50_context *nv50) +#ifndef NV50TCL_BIND_TIC +#define NV50TCL_BIND_TIC(n) (0x1448 + 8 * n) +#endif + +static boolean +nv50_validate_textures(struct nv50_context *nv50, struct nouveau_stateobj *so, + unsigned p) { + static const unsigned p_remap[PIPE_SHADER_TYPES] = { 0, 2 }; + struct nouveau_grobj *eng2d = nv50->screen->eng2d; struct nouveau_grobj *tesla = nv50->screen->tesla; - struct nouveau_stateobj *so; - unsigned i, unit, push; - - push = MAX2(nv50->miptree_nr, nv50->state.miptree_nr) * 2 + 23 + 6; - so = so_new(nv50->miptree_nr * 9 + push, nv50->miptree_nr * 2 + 2); + unsigned unit, j, p_hw = p_remap[p]; nv50_so_init_sifc(nv50, so, nv50->screen->tic, NOUVEAU_BO_VRAM, - nv50->miptree_nr * 8 * 4); - - for (i = 0, unit = 0; unit < nv50->miptree_nr; ++unit) { - struct nv50_miptree *mt = nv50->miptree[unit]; + p * (32 * 8 * 4), nv50->miptree_nr[p] * 8 * 4); - if (!mt) - continue; + for (unit = 0; unit < nv50->miptree_nr[p]; ++unit) { + struct nv50_miptree *mt = nv50->miptree[p][unit]; so_method(so, eng2d, NV50_2D_SIFC_DATA | (2 << 29), 8); - if (nv50_tex_construct(nv50, so, mt, unit)) { - NOUVEAU_ERR("failed tex validate\n"); - so_ref(NULL, &so); - return; + if (mt) { + if (nv50_tex_construct(nv50, so, mt, unit, p)) + return FALSE; + /* Set TEX insn $t src binding $unit in program type p + * to TIC, TSC entry (32 * p + unit), mark valid (1). + */ + so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1); + so_data (so, ((32 * p + unit) << 9) | (unit << 1) | 1); + } else { + for (j = 0; j < 8; ++j) + so_data(so, 0); + so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1); + so_data (so, (unit << 1) | 0); } + } + + for (; unit < nv50->state.miptree_nr[p]; unit++) { + /* Make other bindings invalid. */ + so_method(so, tesla, NV50TCL_BIND_TIC(p_hw), 1); + so_data (so, (unit << 1) | 0); + } - so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1); - so_data (so, (i++ << NV50TCL_SET_SAMPLER_TEX_TIC_SHIFT) | - (unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) | - NV50TCL_SET_SAMPLER_TEX_VALID); + nv50->state.miptree_nr[p] = nv50->miptree_nr[p]; + return TRUE; +} + +void +nv50_tex_validate(struct nv50_context *nv50) +{ + struct nouveau_stateobj *so; + struct nouveau_grobj *tesla = nv50->screen->tesla; + unsigned p, push, nrlc; + + for (nrlc = 0, push = 0, p = 0; p < PIPE_SHADER_TYPES; ++p) { + push += MAX2(nv50->miptree_nr[p], nv50->state.miptree_nr[p]); + nrlc += nv50->miptree_nr[p]; } + push = push * 11 + 23 * PIPE_SHADER_TYPES + 4; + nrlc = nrlc * 2 + 2 * PIPE_SHADER_TYPES; + + so = so_new(push, nrlc); - for (; unit < nv50->state.miptree_nr; unit++) { - so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1); - so_data (so, - (unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) | 0); + if (nv50_validate_textures(nv50, so, PIPE_SHADER_VERTEX) == FALSE || + nv50_validate_textures(nv50, so, PIPE_SHADER_FRAGMENT) == FALSE) { + so_ref(NULL, &so); + + NOUVEAU_ERR("failed tex validate\n"); + return; } /* not sure if the following really do what I think: */ - so_method(so, tesla, 0x1440, 1); /* sync SIFC */ - so_data (so, 0); so_method(so, tesla, 0x1330, 1); /* flush TIC */ so_data (so, 0); so_method(so, tesla, 0x1338, 1); /* flush texture caches */ @@ -189,6 +226,4 @@ nv50_tex_validate(struct nv50_context *nv50) so_ref(so, &nv50->state.tic_upload); so_ref(NULL, &so); - nv50->state.miptree_nr = nv50->miptree_nr; } - diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h index d531e611327..b870302019a 100644 --- a/src/gallium/drivers/nv50/nv50_texture.h +++ b/src/gallium/drivers/nv50/nv50_texture.h @@ -82,6 +82,7 @@ #define NV50TIC_0_0_FMT_RGTC1 0x00000027 #define NV50TIC_0_0_FMT_RGTC2 0x00000028 #define NV50TIC_0_0_FMT_24_8 0x00000029 +#define NV50TIC_0_0_FMT_8_24 0x0000002a #define NV50TIC_0_0_FMT_32_DEPTH 0x0000002f #define NV50TIC_0_0_FMT_32_8 0x00000030 diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index 4705f96f572..4d9afa6fedc 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv50_context.h" @@ -16,6 +17,7 @@ struct nv50_transfer { int level_depth; int level_x; int level_y; + int level_z; unsigned nblocksx; unsigned nblocksy; }; @@ -24,10 +26,10 @@ static void nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, unsigned src_offset, int src_pitch, unsigned src_tile_mode, - int sx, int sy, int sw, int sh, int sd, + int sx, int sy, int sz, int sw, int sh, int sd, struct nouveau_bo *dst_bo, unsigned dst_offset, int dst_pitch, unsigned dst_tile_mode, - int dx, int dy, int dw, int dh, int dd, + int dx, int dy, int dz, int dw, int dh, int dd, int cpp, int width, int height, unsigned src_reloc, unsigned dst_reloc) { @@ -56,7 +58,7 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, OUT_RING (chan, sw * cpp); OUT_RING (chan, sh); OUT_RING (chan, sd); - OUT_RING (chan, 0); + OUT_RING (chan, sz); /* copying only 1 zslice per call */ } if (!dst_bo->tile_flags) { @@ -75,13 +77,13 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, OUT_RING (chan, dw * cpp); OUT_RING (chan, dh); OUT_RING (chan, dd); - OUT_RING (chan, 0); + OUT_RING (chan, dz); /* copying only 1 zslice per call */ } while (height) { int line_count = height > 2047 ? 2047 : height; - WAIT_RING (chan, 15); + MARK_RING (chan, 15, 4); /* flush on lack of space or relocs */ BEGIN_RING(chan, m2mf, NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH, 2); OUT_RELOCh(chan, src_bo, src_offset, src_reloc); @@ -118,20 +120,6 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, } } -static INLINE unsigned -get_zslice_offset(unsigned tile_mode, unsigned z, unsigned pitch, unsigned ny) -{ - unsigned tile_h = get_tile_height(tile_mode); - unsigned tile_d = get_tile_depth(tile_mode); - - /* pitch_2d == to next slice within this volume-tile */ - /* pitch_3d == to next slice in next 2D array of blocks */ - unsigned pitch_2d = tile_h * 64; - unsigned pitch_3d = tile_d * align(ny, tile_h) * pitch; - - return (z % tile_d) * pitch_2d + (z / tile_d) * pitch_3d; -} - static struct pipe_transfer * nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice, @@ -153,11 +141,11 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&tx->base.texture, pt); - tx->nblocksx = pf_get_nblocksx(pt->format, u_minify(pt->width0, level)); - tx->nblocksy = pf_get_nblocksy(pt->format, u_minify(pt->height0, level)); + tx->nblocksx = util_format_get_nblocksx(pt->format, u_minify(pt->width0, level)); + tx->nblocksy = util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)); tx->base.width = w; tx->base.height = h; - tx->base.stride = tx->nblocksx * pf_get_blocksize(pt->format); + tx->base.stride = tx->nblocksx * util_format_get_blocksize(pt->format); tx->base.usage = usage; tx->level_pitch = lvl->pitch; @@ -166,8 +154,9 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->level_depth = u_minify(mt->base.base.depth0, level); tx->level_offset = lvl->image_offset[image]; tx->level_tiling = lvl->tile_mode; - tx->level_x = pf_get_nblocksx(pt->format, x); - tx->level_y = pf_get_nblocksy(pt->format, y); + tx->level_z = zslice; + tx->level_x = util_format_get_nblocksx(pt->format, x); + tx->level_y = util_format_get_nblocksy(pt->format, y); ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, tx->nblocksy * tx->base.stride, &tx->bo); if (ret) { @@ -175,25 +164,20 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; } - if (pt->target == PIPE_TEXTURE_3D) - tx->level_offset += get_zslice_offset(lvl->tile_mode, zslice, - lvl->pitch, - tx->nblocksy); - if (usage & PIPE_TRANSFER_READ) { - nx = pf_get_nblocksx(pt->format, tx->base.width); - ny = pf_get_nblocksy(pt->format, tx->base.height); + nx = util_format_get_nblocksx(pt->format, tx->base.width); + ny = util_format_get_nblocksy(pt->format, tx->base.height); nv50_transfer_rect_m2mf(pscreen, mt->base.bo, tx->level_offset, tx->level_pitch, tx->level_tiling, - x, y, + x, y, zslice, tx->nblocksx, tx->nblocksy, tx->level_depth, tx->bo, 0, tx->base.stride, tx->bo->tile_mode, - 0, 0, + 0, 0, 0, tx->nblocksx, tx->nblocksy, 1, - pf_get_blocksize(pt->format), nx, ny, + util_format_get_blocksize(pt->format), nx, ny, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, NOUVEAU_BO_GART); } @@ -208,22 +192,22 @@ nv50_transfer_del(struct pipe_transfer *ptx) struct nv50_miptree *mt = nv50_miptree(ptx->texture); struct pipe_texture *pt = ptx->texture; - unsigned nx = pf_get_nblocksx(pt->format, tx->base.width); - unsigned ny = pf_get_nblocksy(pt->format, tx->base.height); + unsigned nx = util_format_get_nblocksx(pt->format, tx->base.width); + unsigned ny = util_format_get_nblocksy(pt->format, tx->base.height); if (ptx->usage & PIPE_TRANSFER_WRITE) { struct pipe_screen *pscreen = pt->screen; nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride, tx->bo->tile_mode, - 0, 0, + 0, 0, 0, tx->nblocksx, tx->nblocksy, 1, mt->base.bo, tx->level_offset, tx->level_pitch, tx->level_tiling, - tx->level_x, tx->level_y, + tx->level_x, tx->level_y, tx->level_z, tx->nblocksx, tx->nblocksy, tx->level_depth, - pf_get_blocksize(pt->format), nx, ny, + util_format_get_blocksize(pt->format), nx, ny, NOUVEAU_BO_GART, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART); } @@ -282,7 +266,7 @@ nv50_upload_sifc(struct nv50_context *nv50, reloc |= NOUVEAU_BO_WR; - WAIT_RING (chan, 32); + MARK_RING (chan, 32, 2); /* flush on lack of space or relocs */ if (bo->tile_flags) { BEGIN_RING(chan, eng2d, NV50_2D_DST_FORMAT, 5); diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index db54380241f..f7fa0659e8c 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -24,6 +24,8 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" + #include "nv50_context.h" static boolean @@ -62,18 +64,25 @@ nv50_prim(unsigned mode) } static INLINE uint32_t -nv50_vbo_type_to_hw(unsigned type) +nv50_vbo_type_to_hw(enum pipe_format format) { - switch (type) { - case PIPE_FORMAT_TYPE_FLOAT: + const struct util_format_description *desc; + + desc = util_format_description(format); + assert(desc); + + switch (desc->channel[0].type) { + case UTIL_FORMAT_TYPE_FLOAT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT; - case PIPE_FORMAT_TYPE_UNORM: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; - case PIPE_FORMAT_TYPE_SNORM: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; - case PIPE_FORMAT_TYPE_USCALED: + case UTIL_FORMAT_TYPE_UNSIGNED: + if (desc->channel[0].normalized) { + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; + } return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; - case PIPE_FORMAT_TYPE_SSCALED: + case UTIL_FORMAT_TYPE_SIGNED: + if (desc->channel[0].normalized) { + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; + } return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; /* case PIPE_FORMAT_TYPE_UINT: @@ -120,9 +129,15 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) { uint32_t hw_type, hw_size; enum pipe_format pf = ve->src_format; - unsigned size = pf_size_x(pf) << pf_exp2(pf); + const struct util_format_description *desc; + unsigned size; - hw_type = nv50_vbo_type_to_hw(pf_type(pf)); + desc = util_format_description(pf); + assert(desc); + + size = util_format_get_component_bits(pf, UTIL_FORMAT_COLORSPACE_RGB, 0); + + hw_type = nv50_vbo_type_to_hw(pf); hw_size = nv50_vbo_size_to_hw(size, ve->nr_components); if (!hw_type || !hw_size) { @@ -131,7 +146,7 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) return 0x24e80000; } - if (pf_swizzle_x(pf) == 2) /* BGRA */ + if (desc->swizzle[0] == UTIL_FORMAT_SWIZZLE_Z) /* BGRA */ hw_size |= (1 << 31); /* no real swizzle bits :-( */ return (hw_type | hw_size); @@ -319,9 +334,13 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, float *v; int ret; enum pipe_format pf = ve->src_format; + const struct util_format_description *desc; - if ((pf_type(pf) != PIPE_FORMAT_TYPE_FLOAT) || - (pf_size_x(pf) << pf_exp2(pf)) != 32) + desc = util_format_description(pf); + assert(desc); + + if ((desc->channel[0].type != UTIL_FORMAT_TYPE_FLOAT) || + util_format_get_component_bits(pf, UTIL_FORMAT_COLORSPACE_RGB, 0) != 32) return FALSE; ret = nouveau_bo_map(bo, NOUVEAU_BO_RD); @@ -609,7 +628,8 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, for (i = 0; i < nv50->vtxelt_nr; ++i) { struct pipe_vertex_element *ve; struct pipe_vertex_buffer *vb; - unsigned n, type, size; + unsigned n, size; + const struct util_format_description *desc; ve = &nv50->vtxelt[i]; vb = &nv50->vtxbuf[ve->vertex_buffer_index]; @@ -621,8 +641,10 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, emit->map[n] = nouveau_bo(vb->buffer)->map + (start * vb->stride + ve->src_offset); - type = pf_type(ve->src_format); - size = pf_size_x(ve->src_format) << pf_exp2(ve->src_format); + desc = util_format_description(ve->src_format); + assert(desc); + + size = util_format_get_component_bits(ve->src_format, UTIL_FORMAT_COLORSPACE_RGB, 0); assert(ve->nr_components > 0 && ve->nr_components <= 4); diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile index 9c9fc6f64b3..afddcb161fa 100644 --- a/src/gallium/drivers/r300/Makefile +++ b/src/gallium/drivers/r300/Makefile @@ -4,8 +4,8 @@ include $(TOP)/configs/current LIBNAME = r300 C_SOURCES = \ + r300_blit.c \ r300_chipset.c \ - r300_clear.c \ r300_context.c \ r300_debug.c \ r300_emit.c \ @@ -17,7 +17,6 @@ C_SOURCES = \ r300_state.c \ r300_state_derived.c \ r300_state_invariant.c \ - r300_vbo.c \ r300_vs.c \ r300_texture.c \ r300_tgsi_to_rc.c diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript index 97989040d2e..0d2de17be93 100644 --- a/src/gallium/drivers/r300/SConscript +++ b/src/gallium/drivers/r300/SConscript @@ -9,8 +9,8 @@ env.Append(CPPPATH = ['#/src/mesa/drivers/dri/r300/compiler', '#/include', '#/sr r300 = env.ConvenienceLibrary( target = 'r300', source = [ + 'r300_blit.c', 'r300_chipset.c', - 'r300_clear.c', 'r300_context.c', 'r300_debug.c', 'r300_emit.c', diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c new file mode 100644 index 00000000000..ffe066d5369 --- /dev/null +++ b/src/gallium/drivers/r300/r300_blit.c @@ -0,0 +1,130 @@ +/* + * Copyright 2009 Marek Olšák <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_blit.h" +#include "r300_context.h" + +#include "util/u_rect.h" + +static void r300_blitter_save_states(struct r300_context* r300) +{ + util_blitter_save_blend(r300->blitter, r300->blend_state); + util_blitter_save_depth_stencil_alpha(r300->blitter, r300->dsa_state); + util_blitter_save_rasterizer(r300->blitter, r300->rs_state); + util_blitter_save_fragment_shader(r300->blitter, r300->fs); + util_blitter_save_vertex_shader(r300->blitter, r300->vs); +} + +/* Clear currently bound buffers. */ +void r300_clear(struct pipe_context* pipe, + unsigned buffers, + const float* rgba, + double depth, + unsigned stencil) +{ + /* XXX Implement fastfill. + * + * If fastfill is enabled, a few facts should be considered: + * + * 1) Zbuffer must be micro-tiled and whole microtiles must be + * written. + * + * 2) ZB_DEPTHCLEARVALUE is used to clear a zbuffer and Z Mask must be + * equal to 0. + * + * 3) RB3D_COLOR_CLEAR_VALUE is used to clear a colorbuffer and + * RB3D_COLOR_CHANNEL_MASK must be equal to 0. + * + * 4) ZB_CB_CLEAR can be used to make the ZB units help in clearing + * the colorbuffer. The color clear value is supplied through both + * RB3D_COLOR_CLEAR_VALUE and ZB_DEPTHCLEARVALUE, and the colorbuffer + * must be set in ZB_DEPTHOFFSET and ZB_DEPTHPITCH in addition to + * RB3D_COLOROFFSET and RB3D_COLORPITCH. It's obvious that the zbuffer + * will not be cleared and multiple render targets cannot be cleared + * this way either. + * + * 5) For 16-bit integer buffering, compression causes a hung with one or + * two samples and should not be used. + * + * 6) Fastfill must not be used if reading of compressed Z data is disabled + * and writing of compressed Z data is enabled (RD/WR_COMP_ENABLE), + * i.e. it cannot be used to compress the zbuffer. + * (what the hell does that mean and how does it fit in clearing + * the buffers?) + * + * - Marek + */ + + struct r300_context* r300 = r300_context(pipe); + + r300_blitter_save_states(r300); + + util_blitter_clear(r300->blitter, + r300->framebuffer_state.width, + r300->framebuffer_state.height, + r300->framebuffer_state.nr_cbufs, + buffers, rgba, depth, stencil); +} + +/* Copy a block of pixels from one surface to another. */ +void r300_surface_copy(struct pipe_context* pipe, + struct pipe_surface* dst, + unsigned dstx, unsigned dsty, + struct pipe_surface* src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct r300_context* r300 = r300_context(pipe); + + /* Yeah we have to save all those states to ensure this blitter operation + * is really transparent. The states will be restored by the blitter once + * copying is done. */ + r300_blitter_save_states(r300); + util_blitter_save_framebuffer(r300->blitter, &r300->framebuffer_state); + + util_blitter_save_fragment_sampler_states( + r300->blitter, r300->sampler_count, (void**)r300->sampler_states); + + util_blitter_save_fragment_sampler_textures( + r300->blitter, r300->texture_count, + (struct pipe_texture**)r300->textures); + + /* Do a copy */ + util_blitter_copy(r300->blitter, + dst, dstx, dsty, src, srcx, srcy, width, height, TRUE); +} + +/* Fill a region of a surface with a constant value. */ +void r300_surface_fill(struct pipe_context* pipe, + struct pipe_surface* dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value) +{ + struct r300_context* r300 = r300_context(pipe); + + r300_blitter_save_states(r300); + util_blitter_save_framebuffer(r300->blitter, &r300->framebuffer_state); + + util_blitter_fill(r300->blitter, + dst, dstx, dsty, width, height, value); +} diff --git a/src/gallium/drivers/r300/r300_clear.h b/src/gallium/drivers/r300/r300_blit.h index b8fcdf273c7..029e4f98e7d 100644 --- a/src/gallium/drivers/r300/r300_clear.h +++ b/src/gallium/drivers/r300/r300_blit.h @@ -1,5 +1,5 @@ /* - * Copyright 2008 Corbin Simpson <[email protected]> + * Copyright 2008 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,10 +20,11 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#ifndef R300_CLEAR_H -#define R300_CLEAR_H +#ifndef R300_BLIT_H +#define R300_BLIT_H struct pipe_context; +struct pipe_surface; void r300_clear(struct pipe_context* pipe, unsigned buffers, @@ -31,4 +32,17 @@ void r300_clear(struct pipe_context* pipe, double depth, unsigned stencil); -#endif /* R300_CLEAR_H */ +void r300_surface_copy(struct pipe_context* pipe, + struct pipe_surface* dst, + unsigned dstx, unsigned dsty, + struct pipe_surface* src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); + +void r300_surface_fill(struct pipe_context* pipe, + struct pipe_surface* dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value); + +#endif /* R300_BLIT_H */ diff --git a/src/gallium/drivers/r300/r300_clear.c b/src/gallium/drivers/r300/r300_clear.c deleted file mode 100644 index 02d6d504fc0..00000000000 --- a/src/gallium/drivers/r300/r300_clear.c +++ /dev/null @@ -1,38 +0,0 @@ -/* - * Copyright 2008 Corbin Simpson <[email protected]> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#include "r300_clear.h" -#include "r300_context.h" - -#include "util/u_clear.h" - -/* Clears currently bound buffers. */ -void r300_clear(struct pipe_context* pipe, - unsigned buffers, - const float* rgba, - double depth, - unsigned stencil) -{ - /* XXX we can and should do one clear if both color and zs are set */ - util_clear(pipe, &r300_context(pipe)->framebuffer_state, - buffers, rgba, depth, stencil); -} diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 5b337f03ac4..d5c2d63d393 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -28,7 +28,7 @@ #include "util/u_memory.h" #include "util/u_simple_list.h" -#include "r300_clear.h" +#include "r300_blit.h" #include "r300_context.h" #include "r300_flush.h" #include "r300_query.h" @@ -52,6 +52,8 @@ static void r300_destroy_context(struct pipe_context* context) struct r300_context* r300 = r300_context(context); struct r300_query* query, * temp; + util_blitter_destroy(r300->blitter); + util_hash_table_foreach(r300->shader_hash_table, r300_clear_hash_table, NULL); util_hash_table_destroy(r300->shader_hash_table); @@ -124,6 +126,8 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.destroy = r300_destroy_context; r300->context.clear = r300_clear; + r300->context.surface_copy = r300_surface_copy; + r300->context.surface_fill = r300_surface_fill; if (r300screen->caps->has_tcl) { r300->context.draw_arrays = r300_draw_arrays; @@ -175,5 +179,8 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw++; + + r300->blitter = util_blitter_create(&r300->context); + return &r300->context; } diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 0be190392a5..232530b7dc4 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -25,6 +25,8 @@ #include "draw/draw_vertex.h" +#include "util/u_blitter.h" + #include "pipe/p_context.h" #include "pipe/p_inlines.h" @@ -89,6 +91,8 @@ struct r300_rs_block { }; struct r300_sampler_state { + struct pipe_sampler_state state; + uint32_t filter0; /* R300_TX_FILTER0: 0x4400 */ uint32_t filter1; /* R300_TX_FILTER1: 0x4440 */ uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */ @@ -248,6 +252,8 @@ struct r300_context { struct radeon_winsys* winsys; /* Draw module. Used mostly for SW TCL. */ struct draw_context* draw; + /* Accelerated blit support. */ + struct blitter_context* blitter; /* Vertex buffer for rendering. */ struct pipe_buffer* vbo; diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h index 9fcf3ab538c..d142fee0502 100644 --- a/src/gallium/drivers/r300/r300_cs.h +++ b/src/gallium/drivers/r300/r300_cs.h @@ -115,6 +115,15 @@ cs_count -= 3; \ } while (0) +#define OUT_CS_RELOC_NO_OFFSET(bo, rd, wd, flags) do { \ + DBG(cs_context_copy, DBG_CS, "r300: writing relocation for buffer %p, " \ + "domains (%d, %d, %d)\n", \ + bo, rd, wd, flags); \ + assert(bo); \ + cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \ + cs_count -= 2; \ +} while (0) + #define END_CS do { \ if (VERY_VERBOSE_CS) { \ DBG(cs_context_copy, DBG_CS, "r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index dbf316a9b57..199ce3a945d 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -1,5 +1,6 @@ /* * Copyright 2008 Corbin Simpson <[email protected]> + * Copyright 2009 Marek Olšák <[email protected]> * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -22,6 +23,7 @@ /* r300_emit: Functions for emitting state. */ +#include "util/u_format.h" #include "util/u_math.h" #include "r300_context.h" @@ -40,9 +42,16 @@ void r300_emit_blend_state(struct r300_context* r300, CS_LOCALS(r300); BEGIN_CS(8); OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3); - OUT_CS(blend->blend_control); - OUT_CS(blend->alpha_blend_control); - OUT_CS(blend->color_channel_mask); + if (r300->framebuffer_state.nr_cbufs) { + OUT_CS(blend->blend_control); + OUT_CS(blend->alpha_blend_control); + OUT_CS(blend->color_channel_mask); + } else { + OUT_CS(0); + OUT_CS(0); + OUT_CS(0); + /* XXX also disable fastfill here once it's supported */ + } OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop); OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither); END_CS; @@ -149,6 +158,13 @@ static const float * get_shader_constant( vec[1] = 1.0 / tex->height0; break; + /* Texture compare-fail value. */ + /* XXX Since Gallium doesn't support GL_ARB_shadow_ambient, + * this is always (0,0,0,0). */ + case RC_STATE_SHADOW_AMBIENT: + vec[3] = 0; + break; + default: debug_printf("r300: Implementation error: " "Unknown RC_CONSTANT type %d\n", constant->u.State[0]); @@ -276,7 +292,7 @@ void r500_emit_fragment_program_code(struct r300_context* r300, BEGIN_CS(13 + ((code->inst_end + 1) * 6)); - OUT_CS_REG(R500_US_CONFIG, 0); + OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx); OUT_CS_REG(R500_US_CODE_RANGE, R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end)); @@ -330,7 +346,13 @@ void r300_emit_fb_state(struct r300_context* r300, int i; CS_LOCALS(r300); - BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4); + /* Shouldn't fail unless there is a bug in the state tracker. */ + assert(fb->nr_cbufs <= 4); + + BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) + + (fb->zsbuf ? 10 : 0) + 6); + + /* Flush and free renderbuffer caches. */ OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS | R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D); @@ -338,6 +360,10 @@ void r300_emit_fb_state(struct r300_context* r300, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); + /* Set the number of colorbuffers. */ + OUT_CS_REG(R300_RB3D_CCTL, R300_RB3D_CCTL_NUM_MULTIWRITES(fb->nr_cbufs)); + + /* Set up colorbuffers. */ for (i = 0; i < fb->nr_cbufs; i++) { surf = fb->cbufs[i]; tex = (struct r300_texture*)surf->texture; @@ -355,6 +381,12 @@ void r300_emit_fb_state(struct r300_context* r300, r300_translate_out_fmt(surf->format)); } + /* Disable unused colorbuffers. */ + for (; i < 4; i++) { + OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), R300_US_OUT_FMT_UNUSED); + } + + /* Set up a zbuffer. */ if (fb->zsbuf) { surf = fb->zsbuf; tex = (struct r300_texture*)surf->texture; @@ -623,50 +655,68 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } -/* XXX I can't read this and that's not good */ -void r300_emit_aos(struct r300_context* r300, unsigned offset) +static boolean r300_validate_aos(struct r300_context *r300) { struct pipe_vertex_buffer *vbuf = r300->vertex_buffer; struct pipe_vertex_element *velem = r300->vertex_element; - CS_LOCALS(r300); int i; - unsigned aos_count = r300->vertex_element_count; + /* Check if formats and strides are aligned to the size of DWORD. */ + for (i = 0; i < r300->vertex_element_count; i++) { + if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 || + util_format_get_blocksize(velem[i].src_format) % 4 != 0) { + return FALSE; + } + } + return TRUE; +} + +void r300_emit_aos(struct r300_context* r300, unsigned offset) +{ + struct pipe_vertex_buffer *vb1, *vb2, *vbuf = r300->vertex_buffer; + struct pipe_vertex_element *velem = r300->vertex_element; + int i; + unsigned size1, size2, aos_count = r300->vertex_element_count; unsigned packet_size = (aos_count * 3 + 1) / 2; + CS_LOCALS(r300); + + /* XXX Move this checking to a more approriate place. */ + if (!r300_validate_aos(r300)) { + /* XXX We should fallback using Draw. */ + assert(0); + } + BEGIN_CS(2 + packet_size + aos_count * 2); OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size); OUT_CS(aos_count); + for (i = 0; i < aos_count - 1; i += 2) { - int buf_num1 = velem[i].vertex_buffer_index; - int buf_num2 = velem[i+1].vertex_buffer_index; - assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); - assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_blocksize(velem[i+1].src_format) % 4 == 0); - OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | - (pf_get_blocksize(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); - OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset + - offset * vbuf[buf_num1].stride); - OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + - offset * vbuf[buf_num2].stride); + vb1 = &vbuf[velem[i].vertex_buffer_index]; + vb2 = &vbuf[velem[i+1].vertex_buffer_index]; + size1 = util_format_get_blocksize(velem[i].src_format); + size2 = util_format_get_blocksize(velem[i+1].src_format); + + OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) | + R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride)); + OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride); + OUT_CS(vb2->buffer_offset + velem[i+1].src_offset + offset * vb2->stride); } + if (aos_count & 1) { - int buf_num = velem[i].vertex_buffer_index; - assert(vbuf[buf_num].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); - OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); - OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset + - offset * vbuf[buf_num].stride); + vb1 = &vbuf[velem[i].vertex_buffer_index]; + size1 = util_format_get_blocksize(velem[i].src_format); + + OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride)); + OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride); } - /* XXX bare CS reloc */ for (i = 0; i < aos_count; i++) { - cs_winsys->write_cs_reloc(cs_winsys, - vbuf[velem[i].vertex_buffer_index].buffer, - RADEON_GEM_DOMAIN_GTT, - 0, - 0); - cs_count -= 2; + OUT_CS_RELOC_NO_OFFSET(vbuf[velem[i].vertex_buffer_index].buffer, + RADEON_GEM_DOMAIN_GTT, 0, 0); } END_CS; } + #if 0 void r300_emit_draw_packet(struct r300_context* r300) { @@ -841,10 +891,21 @@ void r300_emit_viewport_state(struct r300_context* r300, void r300_emit_texture_count(struct r300_context* r300) { + uint32_t tx_enable = 0; + int i; CS_LOCALS(r300); + /* Notice that texture_count and sampler_count are just sizes + * of the respective arrays. We still have to check for the individual + * elements. */ + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { + if (r300->textures[i]) { + tx_enable |= 1 << i; + } + } + BEGIN_CS(2); - OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); + OUT_CS_REG(R300_TX_ENABLE, tx_enable); END_CS; } @@ -976,18 +1037,20 @@ validate: if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { if (r300screen->caps->is_r500) { - r500_emit_fragment_program_code(r300, &r300->fs->code); + r500_emit_fragment_program_code(r300, &r300->fs->shader->code); } else { - r300_emit_fragment_program_code(r300, &r300->fs->code); + r300_emit_fragment_program_code(r300, &r300->fs->shader->code); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; } if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER_CONSTANTS) { if (r300screen->caps->is_r500) { - r500_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + r500_emit_fs_constant_buffer(r300, + &r300->fs->shader->code.constants); } else { - r300_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + r300_emit_fs_constant_buffer(r300, + &r300->fs->shader->code.constants); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER_CONSTANTS; } diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c index 79b01bb4dc2..4e1b61ca409 100644 --- a/src/gallium/drivers/r300/r300_fs.c +++ b/src/gallium/drivers/r300/r300_fs.c @@ -22,6 +22,9 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "util/u_math.h" +#include "util/u_memory.h" + #include "tgsi/tgsi_dump.h" #include "r300_context.h" @@ -33,8 +36,8 @@ #include "radeon_compiler.h" /* Convert info about FS input semantics to r300_shader_semantics. */ -static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, - struct r300_shader_semantics* fs_inputs) +void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, + struct r300_shader_semantics* fs_inputs) { int i; unsigned index; @@ -66,7 +69,6 @@ static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, } } - static void find_output_registers(struct r300_fragment_program_compiler * compiler, struct r300_fragment_shader * fs) { @@ -95,7 +97,7 @@ static void allocate_hardware_inputs( void * mydata) { struct r300_shader_semantics* inputs = - &((struct r300_fragment_shader*)c->UserData)->inputs; + (struct r300_shader_semantics*)c->UserData; int i, reg = 0; /* Allocate input registers. */ @@ -114,31 +116,45 @@ static void allocate_hardware_inputs( } } -void r300_translate_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs) +static void get_compare_state( + struct r300_context* r300, + struct r300_fragment_program_external_state* state, + unsigned shadow_samplers) +{ + memset(state, 0, sizeof(*state)); + + for (int i = 0; i < r300->sampler_count; i++) { + struct r300_sampler_state* s = r300->sampler_states[i]; + + if (s && s->state.compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + /* XXX Gallium doesn't provide us with any information regarding + * this mode, so we are screwed. I'm setting 0 = LUMINANCE. */ + state->unit[i].depth_texture_mode = 0; + + /* Fortunately, no need to translate this. */ + state->unit[i].texture_compare_func = s->state.compare_func; + } + } +} + +static void r300_translate_fragment_shader( + struct r300_context* r300, + struct r300_fragment_shader_code* shader) { + struct r300_fragment_shader* fs = r300->fs; struct r300_fragment_program_compiler compiler; struct tgsi_to_rc ttr; - /* Initialize. */ - r300_shader_read_fs_inputs(&fs->info, &fs->inputs); - /* Setup the compiler. */ memset(&compiler, 0, sizeof(compiler)); rc_init(&compiler.Base); compiler.Base.Debug = DBG_ON(r300, DBG_FP); - compiler.code = &fs->code; + compiler.code = &shader->code; + compiler.state = shader->compare_state; compiler.is_r500 = r300_screen(r300->context.screen)->caps->is_r500; compiler.AllocateHwInputs = &allocate_hardware_inputs; - compiler.UserData = fs; - - /* XXX: Program compilation depends on texture compare modes, - * which are sampler state. Therefore, programs need to be recompiled - * depending on this state as in the classic Mesa driver. - * - * This is not yet handled correctly. - */ + compiler.UserData = &fs->inputs; find_output_registers(&compiler, fs); @@ -153,6 +169,8 @@ void r300_translate_fragment_shader(struct r300_context* r300, r300_tgsi_to_rc(&ttr, fs->state.tokens); + fs->shadow_samplers = compiler.Base.Program.ShadowSamplers; + /* Invoke the compiler */ r3xx_compile_fragment_program(&compiler); if (compiler.Base.Error) { @@ -164,5 +182,51 @@ void r300_translate_fragment_shader(struct r300_context* r300, /* And, finally... */ rc_destroy(&compiler.Base); - fs->translated = TRUE; +} + +boolean r300_pick_fragment_shader(struct r300_context* r300) +{ + struct r300_fragment_shader* fs = r300->fs; + struct r300_fragment_program_external_state state; + struct r300_fragment_shader_code* ptr; + + if (!fs->first) { + /* Build the fragment shader for the first time. */ + fs->first = fs->shader = CALLOC_STRUCT(r300_fragment_shader_code); + + /* BTW shadow samplers will be known after the first translation, + * therefore we set ~0, which means it should look at all sampler + * states. This choice doesn't have any impact on the correctness. */ + get_compare_state(r300, &fs->shader->compare_state, ~0); + r300_translate_fragment_shader(r300, fs->shader); + return TRUE; + + } else if (fs->shadow_samplers) { + get_compare_state(r300, &state, fs->shadow_samplers); + + /* Check if the currently-bound shader has been compiled + * with the texture-compare state we need. */ + if (memcmp(&fs->shader->compare_state, &state, sizeof(state)) != 0) { + /* Search for the right shader. */ + ptr = fs->first; + while (ptr) { + if (memcmp(&ptr->compare_state, &state, sizeof(state)) == 0) { + fs->shader = ptr; + return TRUE; + } + ptr = ptr->next; + } + + /* Not found, gotta compile a new one. */ + ptr = CALLOC_STRUCT(r300_fragment_shader_code); + ptr->next = fs->first; + fs->first = fs->shader = ptr; + + ptr->compare_state = state; + r300_translate_fragment_shader(r300, ptr); + return TRUE; + } + } + + return FALSE; } diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h index 630e2d0c8a5..40ce874353c 100644 --- a/src/gallium/drivers/r300/r300_fs.h +++ b/src/gallium/drivers/r300/r300_fs.h @@ -30,6 +30,13 @@ #include "radeon_code.h" #include "r300_shader_semantics.h" +struct r300_fragment_shader_code { + struct r300_fragment_program_external_state compare_state; + struct rX00_fragment_program_code code; + + struct r300_fragment_shader_code* next; +}; + struct r300_fragment_shader { /* Parent class */ struct pipe_shader_state state; @@ -37,21 +44,28 @@ struct r300_fragment_shader { struct tgsi_shader_info info; struct r300_shader_semantics inputs; - /* Has this shader been translated yet? */ - boolean translated; + /* Bits 0-15: TRUE if it's a shadow sampler, FALSE otherwise. */ + unsigned shadow_samplers; - /* Compiled code */ - struct rX00_fragment_program_code code; + /* Currently-bound fragment shader. */ + struct r300_fragment_shader_code* shader; + + /* List of the same shaders compiled with different texture-compare + * states. */ + struct r300_fragment_shader_code* first; }; +void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, + struct r300_shader_semantics* fs_inputs); -void r300_translate_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs); +/* Return TRUE if the shader was switched and should be re-emitted. */ +boolean r300_pick_fragment_shader(struct r300_context* r300); -static inline boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs) +static INLINE boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs) { if (!fs) - return FALSE; - return (fs->code.writes_depth) ? TRUE : FALSE; + return FALSE; + return (fs->shader->code.writes_depth) ? TRUE : FALSE; } + #endif /* R300_FS_H */ diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 85b1ea568a3..d8d08fbe264 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -2145,6 +2145,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. /* Unpipelined. */ #define R300_RB3D_CCTL 0x4e00 +# define R300_RB3D_CCTL_NUM_MULTIWRITES(x) (MAX2(((x)-1), 0) << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_1_BUFFER (0 << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_2_BUFFERS (1 << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_3_BUFFERS (2 << 5) @@ -3293,6 +3294,11 @@ enum { */ #define R300_PACKET3_3D_LOAD_VBPNTR 0x00002F00 +# define R300_VBPNTR_SIZE0(x) ((x) >> 2) +# define R300_VBPNTR_STRIDE0(x) (((x) >> 2) << 8) +# define R300_VBPNTR_SIZE1(x) (((x) >> 2) << 16) +# define R300_VBPNTR_STRIDE1(x) (((x) >> 2) << 24) + #define R300_PACKET3_INDX_BUFFER 0x00003300 # define R300_INDX_BUFFER_DST_SHIFT 0 # define R300_INDX_BUFFER_SKIP_SHIFT 16 diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 35b335df6a1..2d70ec2ac94 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -37,7 +37,6 @@ #include "r300_reg.h" #include "r300_render.h" #include "r300_state_derived.h" -#include "r300_vbo.h" /* r300_render: Vertex and index buffer primitive emission. */ #define R300_MAX_VBO_SIZE (1024 * 1024) @@ -76,14 +75,61 @@ static boolean r300_nothing_to_draw(struct r300_context *r300) r300->scissor_state->scissor.empty_area; } +static uint32_t r300_provoking_vertex_fixes(struct r300_context *r300, + unsigned mode) +{ + uint32_t color_control = r300->rs_state->color_control; + + /* By default (see r300_state.c:r300_create_rs_state) color_control is + * initialized to provoking the first vertex. + * + * Triangle fans must be reduced to the second vertex, not the first, in + * Gallium flatshade-first mode, as per the GL spec. + * (http://www.opengl.org/registry/specs/ARB/provoking_vertex.txt) + * + * Quads never provoke correctly in flatshade-first mode. The first + * vertex is never considered as provoking, so only the second, third, + * and fourth vertices can be selected, and both "third" and "last" modes + * select the fourth vertex. This is probably due to D3D lacking quads. + * + * Similarly, polygons reduce to the first, not the last, vertex, when in + * "last" mode, and all other modes start from the second vertex. + * + * ~ C. + */ + + if (r300->rs_state->rs.flatshade_first) { + switch (mode) { + case PIPE_PRIM_TRIANGLE_FAN: + color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_SECOND; + break; + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + case PIPE_PRIM_POLYGON: + color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST; + break; + default: + color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_FIRST; + break; + } + } else { + color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST; + } + + return color_control; +} + static void r300_emit_draw_arrays(struct r300_context *r300, unsigned mode, unsigned count) { CS_LOCALS(r300); - BEGIN_CS(4); - OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); + BEGIN_CS(8); + OUT_CS_REG(R300_GA_COLOR_CONTROL, + r300_provoking_vertex_fixes(r300, mode)); + OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count - 1); OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0); OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | r300_translate_primitive(mode)); @@ -108,7 +154,10 @@ static void r300_emit_draw_elements(struct r300_context *r300, assert((start * indexSize) % 4 == 0); assert(offset_dwords == 0); - BEGIN_CS(10); + BEGIN_CS(14); + OUT_CS_REG(R300_GA_COLOR_CONTROL, + r300_provoking_vertex_fixes(r300, mode)); + OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, minIndex); OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex); OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0); if (indexSize == 4) { @@ -179,7 +228,6 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, return FALSE; } - if (count > 65535) { return FALSE; } @@ -194,7 +242,14 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, return FALSE; } - setup_index_buffer(r300, indexBuffer, indexSize); + if (!r300->winsys->add_buffer(r300->winsys, indexBuffer, + RADEON_GEM_DOMAIN_GTT, 0)) { + return FALSE; + } + + if (!r300->winsys->validate(r300->winsys)) { + return FALSE; + } r300_emit_dirty_state(r300); diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index c0d9797020d..2a8667d4835 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -21,6 +21,7 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_simple_screen.h" @@ -141,6 +142,10 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) return 0; case PIPE_CAP_BLEND_EQUATION_SEPARATE: return 1; + case PIPE_CAP_SM3: + return 1; + case PIPE_CAP_MAX_COMBINED_SAMPLERS: + return 8; default: debug_printf("r300: Implementation error: Bad param %d\n", param); @@ -220,12 +225,18 @@ static boolean check_tex_format(enum pipe_format format, uint32_t usage, /* Z buffer or texture */ case PIPE_FORMAT_Z16_UNORM: + retval = usage & + (PIPE_TEXTURE_USAGE_DEPTH_STENCIL | + PIPE_TEXTURE_USAGE_SAMPLER); + break; + + /* 24bit Z buffer can only be used as a texture on R500. */ case PIPE_FORMAT_Z24X8_UNORM: /* Z buffer with stencil or texture */ case PIPE_FORMAT_Z24S8_UNORM: retval = usage & (PIPE_TEXTURE_USAGE_DEPTH_STENCIL | - PIPE_TEXTURE_USAGE_SAMPLER); + (is_r500 ? PIPE_TEXTURE_USAGE_SAMPLER : 0)); break; /* Definitely unsupported formats. */ @@ -351,8 +362,8 @@ static void* r300_transfer_map(struct pipe_screen* screen, } return map + r300_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); } static void r300_transfer_unmap(struct pipe_screen* screen, diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index a0ebdf30241..49072462ec3 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -332,6 +332,7 @@ static void r300->dirty_state |= R300_NEW_SCISSOR; } r300->dirty_state |= R300_NEW_FRAMEBUFFERS; + r300->dirty_state |= R300_NEW_BLEND; } /* Create fragment shader state. */ @@ -347,6 +348,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe, fs->state.tokens = tgsi_dup_tokens(shader->tokens); tgsi_scan_shader(shader->tokens, &fs->info); + r300_shader_read_fs_inputs(&fs->info, &fs->inputs); return (void*)fs; } @@ -360,11 +362,10 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) if (fs == NULL) { r300->fs = NULL; return; - } else if (!fs->translated) { - r300_translate_fragment_shader(r300, fs); } r300->fs = fs; + r300_pick_fragment_shader(r300); r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | R300_NEW_FRAGMENT_SHADER_CONSTANTS; } @@ -373,7 +374,14 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) static void r300_delete_fs_state(struct pipe_context* pipe, void* shader) { struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader; - rc_constants_destroy(&fs->code.constants); + struct r300_fragment_shader_code *tmp, *ptr = fs->first; + + while (ptr) { + tmp = ptr; + ptr = ptr->next; + rc_constants_destroy(&tmp->code.constants); + FREE(tmp); + } FREE((void*)fs->state.tokens); FREE(shader); } @@ -412,8 +420,6 @@ static void* r300_create_rs_state(struct pipe_context* pipe, if (state->bypass_vs_clip_and_viewport || !r300_screen(pipe->screen)->caps->has_tcl) { rs->vap_control_status |= R300_VAP_TCL_BYPASS; - } else { - rs->rs.bypass_vs_clip_and_viewport = TRUE; } rs->point_size = pack_float_16_6x(state->point_size) | @@ -504,10 +510,6 @@ static void* r300_create_rs_state(struct pipe_context* pipe, rs->color_control = R300_SHADE_MODEL_SMOOTH; } - if (!state->flatshade_first) { - rs->color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST; - } - return (void*)rs; } @@ -545,6 +547,8 @@ static void* int lod_bias; union util_color uc; + sampler->state = *state; + sampler->filter0 |= (r300_translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) | (r300_translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) | @@ -595,6 +599,14 @@ static void r300_bind_sampler_states(struct pipe_context* pipe, } r300->sampler_count = count; + + /* Pick a fragment shader based on the texture compare state. */ + if (r300->fs && (r300->dirty_state & R300_ANY_NEW_SAMPLERS)) { + if (r300_pick_fragment_shader(r300)) { + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | + R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } + } } static void r300_lacks_vertex_textures(struct pipe_context* pipe, @@ -621,8 +633,6 @@ static void r300_set_sampler_textures(struct pipe_context* pipe, return; } - r300->context.flush(&r300->context, 0, NULL); - for (i = 0; i < count; i++) { if (r300->textures[i] != (struct r300_texture*)texture[i]) { pipe_texture_reference((struct pipe_texture**)&r300->textures[i], diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 6af49888b9e..29bc701a86e 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -134,6 +134,16 @@ static void r300_vertex_psc(struct r300_context* r300) uint16_t type, swizzle; enum pipe_format format; unsigned i; + int identity[16] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}; + int* stream_tab; + + /* If TCL is bypassed, map vertex streams to equivalent VS output + * locations. */ + if (r300->rs_state->enable_vte) { + stream_tab = identity; + } else { + stream_tab = r300->vs->stream_loc_notcl; + } /* Vertex shaders have no semantics on their inputs, * so PSC should just route stuff based on the vertex elements, @@ -147,10 +157,10 @@ static void r300_vertex_psc(struct r300_context* r300) format = r300->vertex_element[i].src_format; type = r300_translate_vertex_data_type(format) | - (i << R300_DST_VEC_LOC_SHIFT); + (stream_tab[i] << R300_DST_VEC_LOC_SHIFT); swizzle = r300_translate_vertex_data_swizzle(format); - if (i % 2) { + if (i & 1) { vformat->vap_prog_stream_cntl[i >> 1] |= type << 16; vformat->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16; } else { @@ -159,7 +169,6 @@ static void r300_vertex_psc(struct r300_context* r300) } } - assert(i <= 15); /* Set the last vector in the PSC. */ @@ -178,7 +187,7 @@ static void r300_swtcl_vertex_psc(struct r300_context* r300) uint16_t type, swizzle; enum pipe_format format; unsigned i, attrib_count; - int* vs_output_tab = r300->vs->output_stream_loc_swtcl; + int* vs_output_tab = r300->vs->stream_loc_notcl; /* For each Draw attribute, route it to the fragment shader according * to the vs_output_tab. */ diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index e6c1cb54dac..dbe42edd910 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -28,6 +28,8 @@ #include "pipe/p_format.h" +#include "util/u_format.h" + #include "r300_reg.h" /* Some maths. These should probably find their way to u_math, if needed. */ @@ -443,20 +445,22 @@ static INLINE uint32_t r300_translate_gb_pipes(int pipe_count) static INLINE unsigned pf_component_count(enum pipe_format format) { unsigned count = 0; - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { - return count; + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + count++; } - - if (pf_size_x(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1)) { + count++; + } + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2)) { count++; } - if (pf_size_y(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3)) { count++; } - if (pf_size_z(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0)) { count++; } - if (pf_size_w(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)) { count++; } @@ -467,19 +471,23 @@ static INLINE unsigned pf_component_count(enum pipe_format format) { static INLINE uint16_t r300_translate_vertex_data_type(enum pipe_format format) { uint32_t result = 0; + const struct util_format_description *desc; unsigned components = pf_component_count(format); - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { + desc = util_format_description(format); + + if (desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) { debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); assert(0); } - switch (pf_type(format)) { + switch (desc->channel[0].type) { /* Half-floats, floats, doubles */ - case PIPE_FORMAT_TYPE_FLOAT: - switch (pf_size_x(format)) { - case 4: + case UTIL_FORMAT_TYPE_FLOAT: + switch (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + case 32: result = R300_DATA_TYPE_FLOAT_1 + (components - 1); break; default: @@ -488,19 +496,15 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } break; - /* Normalized unsigned ints */ - case PIPE_FORMAT_TYPE_UNORM: - /* Normalized signed ints */ - case PIPE_FORMAT_TYPE_SNORM: - /* Non-normalized unsigned ints */ - case PIPE_FORMAT_TYPE_USCALED: - /* Non-normalized signed ints */ - case PIPE_FORMAT_TYPE_SSCALED: - switch (pf_size_x(format)) { - case 1: + /* Unsigned ints */ + case UTIL_FORMAT_TYPE_UNSIGNED: + /* Signed ints */ + case UTIL_FORMAT_TYPE_SIGNED: + switch (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + case 8: result = R300_DATA_TYPE_BYTE; break; - case 2: + case 16: if (components > 2) { result = R300_DATA_TYPE_SHORT_4; } else { @@ -510,8 +514,8 @@ r300_translate_vertex_data_type(enum pipe_format format) { default: debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); - debug_printf("r300: pf_size_x(format) == %d\n", - pf_size_x(format)); + debug_printf("r300: util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) == %d\n", + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)); assert(0); } break; @@ -521,12 +525,11 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } - if (pf_type(format) == PIPE_FORMAT_TYPE_SSCALED) { + if (desc->channel[0].type == UTIL_FORMAT_TYPE_SIGNED) { result |= R300_SIGNED; - } else if (pf_type(format) == PIPE_FORMAT_TYPE_UNORM) { + } + if (desc->channel[0].normalized) { result |= R300_NORMALIZE; - } else if (pf_type(format) == PIPE_FORMAT_TYPE_SNORM) { - result |= (R300_SIGNED | R300_NORMALIZE); } return result; @@ -534,17 +537,21 @@ r300_translate_vertex_data_type(enum pipe_format format) { static INLINE uint16_t r300_translate_vertex_data_swizzle(enum pipe_format format) { + const struct util_format_description *desc = util_format_description(format); + + assert(format); - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { + if (desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) { debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); return 0; } - return ((pf_swizzle_x(format) << R300_SWIZZLE_SELECT_X_SHIFT) | - (pf_swizzle_y(format) << R300_SWIZZLE_SELECT_Y_SHIFT) | - (pf_swizzle_z(format) << R300_SWIZZLE_SELECT_Z_SHIFT) | - (pf_swizzle_w(format) << R300_SWIZZLE_SELECT_W_SHIFT) | + return ((desc->swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) | + (desc->swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) | + (desc->swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) | + (desc->swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT) | (0xf << R300_WRITE_ENA_SHIFT)); } diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 46d1cb39b54..bcd4c030f9c 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -43,7 +43,7 @@ void r300_emit_invariant_state(struct r300_context* r300) struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); - BEGIN_CS(24 + (caps->has_tcl ? 2: 0)); + BEGIN_CS(20 + (caps->has_tcl ? 2: 0)); /*** Graphics Backend (GB) ***/ /* Various GB enables */ @@ -70,9 +70,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_US_W_FMT, 0x0); /*** VAP ***/ - /* Max and min vertex index clamp. */ - OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0x0); - OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, 0xffffff); /* Sign/normalize control */ OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, R300_SGN_NORM_NO_ZERO); /* TCL-only stuff */ @@ -84,15 +81,11 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(56 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); - /* Flush PVS. */ - OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); + BEGIN_CS(44 + (caps->has_tcl ? 7 : 0) + (caps->is_r500 ? 4 : 0)); - OUT_CS_REG(R300_SE_VTE_CNTL, R300_VPORT_X_SCALE_ENA | - R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | - R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | - R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT); if (caps->has_tcl) { + /*Flushing PVS is required before the VAP_GB registers can be changed*/ + OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0); OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4); OUT_CS_32F(1.0); OUT_CS_32F(1.0); @@ -123,19 +116,15 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000); OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C); OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525); - OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); if (caps->is_r500) { OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF); } - OUT_CS_REG(R300_ZB_FORMAT, 0x00000002); - OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003); OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000); OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000); OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000); OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000); - OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F); /* XXX */ OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa); diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 63fc6a235ac..9a96206a4dc 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -22,6 +22,7 @@ #include "pipe/p_screen.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -105,7 +106,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level) return 0; } - return align(pf_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32); + return align(util_format_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32); } static void r300_setup_miptree(struct r300_texture* tex) @@ -115,7 +116,7 @@ static void r300_setup_miptree(struct r300_texture* tex) int i; for (i = 0; i <= base->last_level; i++) { - unsigned nblocksy = pf_get_nblocksy(base->format, u_minify(base->height0, i)); + unsigned nblocksy = util_format_get_nblocksy(base->format, u_minify(base->height0, i)); stride = r300_texture_get_stride(tex, i); layer_size = stride * nblocksy; @@ -128,7 +129,7 @@ static void r300_setup_miptree(struct r300_texture* tex) tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; tex->layer_size[i] = layer_size; - tex->pitch[i] = stride / pf_get_blocksize(base->format); + tex->pitch[i] = stride / util_format_get_blocksize(base->format); debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", @@ -244,7 +245,7 @@ static struct pipe_texture* tex->tex.screen = screen; tex->stride_override = *stride; - tex->pitch[0] = *stride / pf_get_blocksize(base->format); + tex->pitch[0] = *stride / util_format_get_blocksize(base->format); r300_setup_flags(tex); r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500); diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c index 9fb2de24032..096cdb20bbe 100644 --- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c +++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c @@ -212,7 +212,8 @@ static void transform_srcreg( dst->Negate = src->Register.Negate ? RC_MASK_XYZW : 0; } -static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src) +static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src, + uint32_t *shadowSamplers) { switch(src.Texture) { case TGSI_TEXTURE_1D: @@ -233,14 +234,17 @@ static void transform_texture(struct rc_instruction * dst, struct tgsi_instructi case TGSI_TEXTURE_SHADOW1D: dst->U.I.TexSrcTarget = RC_TEXTURE_1D; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; case TGSI_TEXTURE_SHADOW2D: dst->U.I.TexSrcTarget = RC_TEXTURE_2D; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; case TGSI_TEXTURE_SHADOWRECT: dst->U.I.TexSrcTarget = RC_TEXTURE_RECT; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; } } @@ -269,7 +273,8 @@ static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_inst /* Texturing. */ if (src->Instruction.Texture) - transform_texture(dst, src->Texture); + transform_texture(dst, src->Texture, + &ttr->compiler->Program.ShadowSamplers); } static void handle_immediate(struct tgsi_to_rc * ttr, struct tgsi_full_immediate * imm) diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c deleted file mode 100644 index d8610dadfae..00000000000 --- a/src/gallium/drivers/r300/r300_vbo.c +++ /dev/null @@ -1,87 +0,0 @@ -/* - * Copyright 2009 Maciej Cencora <[email protected]> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* r300_vbo: Various helpers for emitting vertex buffers. Needs cleanup, - * refactoring, etc. */ - -#include "r300_vbo.h" - -#include "pipe/p_format.h" - -#include "r300_cs.h" -#include "r300_context.h" -#include "r300_state_inlines.h" -#include "r300_reg.h" - -#include "radeon_winsys.h" - -static INLINE int get_buffer_offset(struct r300_context *r300, - unsigned int buf_nr, - unsigned int elem_offset) -{ - return r300->vertex_buffer[buf_nr].buffer_offset + elem_offset; -} -#if 0 -/* XXX not called at all */ -static void setup_vertex_buffers(struct r300_context *r300) -{ - struct pipe_vertex_element *vert_elem; - int i; - - for (i = 0; i < r300->aos_count; i++) - { - vert_elem = &r300->vertex_element[i]; - /* XXX use translate module to convert the data */ - if (!format_is_supported(vert_elem->src_format, - vert_elem->nr_components)) { - assert(0); - /* - struct pipe_buffer *buf; - const unsigned int max_index = r300->vertex_buffers[vert_elem->vertex_buffer_index].max_index; - buf = pipe_buffer_create(r300->context.screen, 4, usage, vert_elem->nr_components * max_index * sizeof(float)); - */ - } - - if (get_buffer_offset(r300, - vert_elem->vertex_buffer_index, - vert_elem->src_offset) % 4) { - /* XXX need to align buffer */ - assert(0); - } - } -} -#endif -/* XXX these shouldn't be asserts since we can work around bad indexbufs */ -void setup_index_buffer(struct r300_context *r300, - struct pipe_buffer* indexBuffer, - unsigned indexSize) -{ - if (!r300->winsys->add_buffer(r300->winsys, indexBuffer, - RADEON_GEM_DOMAIN_GTT, 0)) { - assert(0); - } - - if (!r300->winsys->validate(r300->winsys)) { - assert(0); - } -} diff --git a/src/gallium/drivers/r300/r300_vbo.h b/src/gallium/drivers/r300/r300_vbo.h deleted file mode 100644 index 7afa75899cf..00000000000 --- a/src/gallium/drivers/r300/r300_vbo.h +++ /dev/null @@ -1,36 +0,0 @@ -/* - * Copyright 2009 Maciej Cencora <[email protected]> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef R300_VBO_H -#define R300_VBO_H - -struct r300_context; -struct pipe_buffer; - -void setup_vertex_attributes(struct r300_context *r300); - -void setup_index_buffer(struct r300_context *r300, - struct pipe_buffer* indexBuffer, - unsigned indexSize); - -#endif diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c index 6ab5995244b..c4ed0d712f9 100644 --- a/src/gallium/drivers/r300/r300_vs.c +++ b/src/gallium/drivers/r300/r300_vs.c @@ -148,35 +148,33 @@ static void r300_shader_vap_output_fmt( assert(gen_count <= 8); } -/* Set VS output stream locations for SWTCL. */ -static void r300_stream_locations_swtcl( +/* Sets up stream mapping to equivalent VS outputs if TCL is bypassed + * or isn't present. */ +static void r300_stream_locations_notcl( struct r300_shader_semantics* vs_outputs, - int* output_stream_loc) + int* stream_loc) { int i, tabi = 0, gen_count; - /* XXX Check whether the numbers (0, 1, 2+i, etc.) are correct. - * These should go to VAP_PROG_STREAM_CNTL/DST_VEC_LOC. */ - /* Position. */ - output_stream_loc[tabi++] = 0; + stream_loc[tabi++] = 0; /* Point size. */ if (vs_outputs->psize != ATTR_UNUSED) { - output_stream_loc[tabi++] = 1; + stream_loc[tabi++] = 1; } /* Colors. */ for (i = 0; i < ATTR_COLOR_COUNT; i++) { if (vs_outputs->color[i] != ATTR_UNUSED) { - output_stream_loc[tabi++] = 2 + i; + stream_loc[tabi++] = 2 + i; } } /* Back-face colors. */ for (i = 0; i < ATTR_COLOR_COUNT; i++) { if (vs_outputs->bcolor[i] != ATTR_UNUSED) { - output_stream_loc[tabi++] = 4 + i; + stream_loc[tabi++] = 4 + i; } } @@ -185,7 +183,7 @@ static void r300_stream_locations_swtcl( for (i = 0; i < ATTR_GENERIC_COUNT; i++) { if (vs_outputs->bcolor[i] != ATTR_UNUSED) { assert(tabi < 16); - output_stream_loc[tabi++] = 6 + gen_count; + stream_loc[tabi++] = 6 + gen_count; gen_count++; } } @@ -193,7 +191,7 @@ static void r300_stream_locations_swtcl( /* Fog coordinates. */ if (vs_outputs->fog != ATTR_UNUSED) { assert(tabi < 16); - output_stream_loc[tabi++] = 6 + gen_count; + stream_loc[tabi++] = 6 + gen_count; gen_count++; } @@ -201,7 +199,7 @@ static void r300_stream_locations_swtcl( assert(gen_count <= 8); for (; tabi < 16;) { - output_stream_loc[tabi++] = -1; + stream_loc[tabi++] = -1; } } @@ -259,10 +257,7 @@ void r300_translate_vertex_shader(struct r300_context* r300, /* Initialize. */ r300_shader_read_vs_outputs(&vs->info, &vs->outputs); r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt); - - if (!r300_screen(r300->context.screen)->caps->has_tcl) { - r300_stream_locations_swtcl(&vs->outputs, vs->output_stream_loc_swtcl); - } + r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl); /* Setup the compiler */ rc_init(&compiler.Base); @@ -288,7 +283,7 @@ void r300_translate_vertex_shader(struct r300_context* r300, /* Invoke the compiler */ r3xx_compile_vertex_program(&compiler); if (compiler.Base.Error) { - /* XXX Fail gracefully */ + /* XXX We should fallback using Draw. */ fprintf(stderr, "r300 VP: Compiler error\n"); abort(); } diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h index 283dd5a9e83..67e9db5366f 100644 --- a/src/gallium/drivers/r300/r300_vs.h +++ b/src/gallium/drivers/r300/r300_vs.h @@ -38,9 +38,11 @@ struct r300_vertex_shader { struct tgsi_shader_info info; struct r300_shader_semantics outputs; - int output_stream_loc_swtcl[16]; uint hwfmt[4]; + /* Stream locations for SWTCL or if TCL is bypassed. */ + int stream_loc_notcl[16]; + /* Has this shader been translated yet? */ boolean translated; diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index bc0e2011300..a518248bb18 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -35,6 +35,8 @@ #include "draw/draw_context.h" +#include "util/u_format.h" + /** * XXX this might get moved someday @@ -80,8 +82,9 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (sp->framebuffer.zsbuf) { int depth_bits; double mrd; - depth_bits = pf_get_component_bits(sp->framebuffer.zsbuf->format, - PIPE_FORMAT_COMP_Z); + depth_bits = util_format_get_component_bits(sp->framebuffer.zsbuf->format, + UTIL_FORMAT_COLORSPACE_ZS, + 0); if (depth_bits > 16) { mrd = 0.0000001; } diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 4f946ccfcf8..a9436a33942 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -32,6 +32,8 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" + +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -63,11 +65,11 @@ softpipe_texture_layout(struct pipe_screen *screen, pt->depth0 = depth; for (level = 0; level <= pt->last_level; level++) { - spt->stride[level] = pf_get_stride(pt->format, width); + spt->stride[level] = util_format_get_stride(pt->format, width); spt->level_offset[level] = buffer_size; - buffer_size += (pf_get_nblocksy(pt->format, height) * + buffer_size += (util_format_get_nblocksy(pt->format, height) * ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * spt->stride[level]); @@ -237,11 +239,11 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->zslice = zslice; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset += face * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) * + ps->offset += face * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) * spt->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset += zslice * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) * + ps->offset += zslice * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) * spt->stride[level]; } else { @@ -297,7 +299,7 @@ softpipe_get_tex_transfer(struct pipe_screen *screen, spt = CALLOC_STRUCT(softpipe_transfer); if (spt) { struct pipe_transfer *pt = &spt->base; - int nblocksy = pf_get_nblocksy(texture->format, u_minify(texture->height0, level)); + int nblocksy = util_format_get_nblocksy(texture->format, u_minify(texture->height0, level)); pipe_texture_reference(&pt->texture, texture); pt->x = x; pt->y = y; @@ -374,8 +376,8 @@ softpipe_transfer_map( struct pipe_screen *screen, } xfer_map = map + softpipe_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); /*printf("map = %p xfer map = %p\n", map, xfer_map);*/ return xfer_map; } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 04f61d16c44..112a6fe0cf3 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -33,6 +33,7 @@ */ #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_tile.h" #include "sp_tile_cache.h" @@ -238,7 +239,7 @@ clear_tile(struct softpipe_cached_tile *tile, { uint i, j; - switch (pf_get_blocksize(format)) { + switch (util_format_get_blocksize(format)) { case 1: memset(tile->data.any, clear_value, TILE_SIZE * TILE_SIZE); break; diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h index a851fa67057..0885d9ca741 100644 --- a/src/gallium/drivers/svga/svga_context.h +++ b/src/gallium/drivers/svga/svga_context.h @@ -367,7 +367,7 @@ struct svga_context #define SVGA_NEW_FRAME_BUFFER 0x800 #define SVGA_NEW_STIPPLE 0x1000 #define SVGA_NEW_SCISSOR 0x2000 -#define SVGA_NEW_BLEND_COLOR 0x5000 +#define SVGA_NEW_BLEND_COLOR 0x4000 #define SVGA_NEW_CLIP 0x8000 #define SVGA_NEW_VIEWPORT 0x10000 #define SVGA_NEW_PRESCALE 0x20000 diff --git a/src/gallium/drivers/svga/svga_screen_cache.c b/src/gallium/drivers/svga/svga_screen_cache.c index 8a06383f61e..eff36e0bccb 100644 --- a/src/gallium/drivers/svga/svga_screen_cache.c +++ b/src/gallium/drivers/svga/svga_screen_cache.c @@ -277,6 +277,15 @@ svga_screen_surface_create(struct svga_screen *svgascreen, while(size < key->size.width) size <<= 1; key->size.width = size; + /* Since we're reusing buffers we're effectively transforming all + * of them into dynamic buffers. + * + * It would be nice to not cache long lived static buffers. But there + * is no way to detect the long lived from short lived ones yet. A + * good heuristic would be buffer size. + */ + key->flags &= ~SVGA3D_SURFACE_HINT_STATIC; + key->flags |= SVGA3D_SURFACE_HINT_DYNAMIC; } handle = svga_screen_cache_lookup(svgascreen, key); diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index 3e2cb1a16d2..2224c2d3945 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -29,6 +29,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_thread.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -158,8 +159,8 @@ svga_transfer_dma_band(struct svga_transfer *st, st->base.x + st->base.width, y + h, st->base.zslice + 1, - pf_get_blocksize(texture->base.format)*8/ - (pf_get_blockwidth(texture->base.format)*pf_get_blockheight(texture->base.format))); + util_format_get_blocksize(texture->base.format)*8/ + (util_format_get_blockwidth(texture->base.format)*util_format_get_blockheight(texture->base.format))); box.x = st->base.x; box.y = y; @@ -209,7 +210,7 @@ svga_transfer_dma(struct svga_transfer *st, } else { unsigned y, h, srcy; - unsigned blockheight = pf_get_blockheight(st->base.texture->format); + unsigned blockheight = util_format_get_blockheight(st->base.texture->format); h = st->hw_nblocksy * blockheight; srcy = 0; for(y = 0; y < st->base.height; y += h) { @@ -319,7 +320,7 @@ svga_texture_create(struct pipe_screen *screen, */ #if 0 if((templat->tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) && - !pf_is_compressed(templat->format)) + !util_format_is_compressed(templat->format)) tex->key.flags |= SVGA3D_SURFACE_HINT_RENDERTARGET; #endif @@ -525,7 +526,7 @@ svga_texture_view_surface(struct pipe_context *pipe, { struct svga_screen *ss = svga_screen(tex->base.screen); struct svga_winsys_surface *handle; - int i, j; + uint32_t i, j; unsigned z_offset = 0; SVGA_DBG(DEBUG_PERF, @@ -657,13 +658,11 @@ svga_get_tex_surface(struct pipe_screen *screen, s->real_level = 0; s->real_zslice = 0; } else { - struct svga_winsys_screen *sws = svga_winsys_screen(screen); - SVGA_DBG(DEBUG_VIEWS, "svga: Surface view: no %p, level %u, face %u, z %u, %p\n", pt, level, face, zslice, s); memset(&s->key, 0, sizeof s->key); - sws->surface_reference(sws, &s->handle, tex->handle); + s->handle = tex->handle; s->real_face = face; s->real_level = level; s->real_zslice = zslice; @@ -677,11 +676,14 @@ static void svga_tex_surface_destroy(struct pipe_surface *surf) { struct svga_surface *s = svga_surface(surf); + struct svga_texture *t = svga_texture(surf->texture); struct svga_screen *ss = svga_screen(surf->texture->screen); - SVGA_DBG(DEBUG_DMA, "unref sid %p (tex surface)\n", s->handle); - assert(s->key.cachable == 0); - svga_screen_surface_destroy(ss, &s->key, &s->handle); + if(s->handle != t->handle) { + SVGA_DBG(DEBUG_DMA, "unref sid %p (tex surface)\n", s->handle); + svga_screen_surface_destroy(ss, &s->key, &s->handle); + } + pipe_texture_reference(&surf->texture, NULL); FREE(surf); } @@ -770,8 +772,8 @@ svga_get_tex_transfer(struct pipe_screen *screen, struct svga_screen *ss = svga_screen(screen); struct svga_winsys_screen *sws = ss->sws; struct svga_transfer *st; - unsigned nblocksx = pf_get_nblocksx(texture->format, w); - unsigned nblocksy = pf_get_nblocksy(texture->format, h); + unsigned nblocksx = util_format_get_nblocksx(texture->format, w); + unsigned nblocksy = util_format_get_nblocksy(texture->format, h); /* We can't map texture storage directly */ if (usage & PIPE_TRANSFER_MAP_DIRECTLY) @@ -785,7 +787,7 @@ svga_get_tex_transfer(struct pipe_screen *screen, st->base.y = y; st->base.width = w; st->base.height = h; - st->base.stride = nblocksx*pf_get_blocksize(texture->format); + st->base.stride = nblocksx*util_format_get_blocksize(texture->format); st->base.usage = usage; st->base.face = face; st->base.level = level; @@ -909,7 +911,6 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, unsigned min_lod, unsigned max_lod) { struct svga_screen *ss = svga_screen(pt->screen); - struct svga_winsys_screen *sws = ss->sws; struct svga_texture *tex = svga_texture(pt); struct svga_sampler_view *sv = NULL; SVGA3dSurfaceFormat format = svga_translate_format(pt->format); @@ -931,7 +932,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, if (min_lod == 0 && max_lod >= pt->last_level) view = FALSE; - if (pf_is_compressed(pt->format) && view) { + if (util_format_is_compressed(pt->format) && view) { format = svga_translate_format_render(pt->format); } @@ -960,7 +961,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, sv = CALLOC_STRUCT(svga_sampler_view); pipe_reference_init(&sv->reference, 1); - sv->texture = tex; + pipe_texture_reference(&sv->texture, pt); sv->min_lod = min_lod; sv->max_lod = max_lod; @@ -975,7 +976,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, pt->depth0, pt->last_level); sv->key.cachable = 0; - sws->surface_reference(sws, &sv->handle, tex->handle); + sv->handle = tex->handle; return sv; } @@ -998,7 +999,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, if (!sv->handle) { assert(0); sv->key.cachable = 0; - sws->surface_reference(sws, &sv->handle, tex->handle); + sv->handle = tex->handle; return sv; } @@ -1012,14 +1013,14 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt, void svga_validate_sampler_view(struct svga_context *svga, struct svga_sampler_view *v) { - struct svga_texture *tex = v->texture; + struct svga_texture *tex = svga_texture(v->texture); unsigned numFaces; unsigned age = 0; int i, k; assert(svga); - if (v->handle == v->texture->handle) + if (v->handle == tex->handle) return; age = tex->age; @@ -1047,11 +1048,14 @@ svga_validate_sampler_view(struct svga_context *svga, struct svga_sampler_view * void svga_destroy_sampler_view_priv(struct svga_sampler_view *v) { - struct svga_screen *ss = svga_screen(v->texture->base.screen); - - SVGA_DBG(DEBUG_DMA, "unref sid %p (sampler view)\n", v->handle); - svga_screen_surface_destroy(ss, &v->key, &v->handle); + struct svga_texture *tex = svga_texture(v->texture); + if(v->handle != tex->handle) { + struct svga_screen *ss = svga_screen(v->texture->screen); + SVGA_DBG(DEBUG_DMA, "unref sid %p (sampler view)\n", v->handle); + svga_screen_surface_destroy(ss, &v->key, &v->handle); + } + pipe_texture_reference(&v->texture, NULL); FREE(v); } @@ -1067,7 +1071,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, svga_translate_format(texture->format), stex->handle); - *stride = pf_get_stride(texture->format, texture->width0); + *stride = util_format_get_stride(texture->format, texture->width0); return *buffer != NULL; } diff --git a/src/gallium/drivers/svga/svga_screen_texture.h b/src/gallium/drivers/svga/svga_screen_texture.h index 727f2c51d28..89ae24219fd 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.h +++ b/src/gallium/drivers/svga/svga_screen_texture.h @@ -61,7 +61,7 @@ struct svga_sampler_view { struct pipe_reference reference; - struct svga_texture *texture; + struct pipe_texture *texture; int min_lod; int max_lod; @@ -94,6 +94,13 @@ struct svga_texture * operation. */ struct svga_host_surface_cache_key key; + + /** + * Handle for the host side surface. + * + * This handle is owned by this texture. Views should hold on to a reference + * to this texture and never destroy this handle directly. + */ struct svga_winsys_surface *handle; }; diff --git a/src/gallium/drivers/svga/svga_state_constants.c b/src/gallium/drivers/svga/svga_state_constants.c index 209ed282450..6b0e511cec1 100644 --- a/src/gallium/drivers/svga/svga_state_constants.c +++ b/src/gallium/drivers/svga/svga_state_constants.c @@ -231,7 +231,8 @@ static int emit_vs_consts( struct svga_context *svga, struct svga_tracked_state svga_hw_vs_parameters = { "hw vs params", - (SVGA_NEW_VS_CONST_BUFFER | + (SVGA_NEW_PRESCALE | + SVGA_NEW_VS_CONST_BUFFER | SVGA_NEW_ZERO_STRIDE | SVGA_NEW_VS_RESULT), emit_vs_consts diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c index f1b0daf9f6b..44b7ceb4fa4 100644 --- a/src/gallium/drivers/svga/svga_state_vs.c +++ b/src/gallium/drivers/svga/svga_state_vs.c @@ -25,6 +25,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_defines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "translate/translate.h" @@ -210,7 +211,7 @@ static int update_zero_stride( struct svga_context *svga, mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, vbuffer->buffer, vel->src_offset, - pf_get_blocksize(vel->src_format), + util_format_get_blocksize(vel->src_format), PIPE_BUFFER_USAGE_CPU_READ); translate->set_buffer(translate, vel->vertex_buffer_index, mapped_buffer, diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c index af1d7f3224e..c31b1d86986 100644 --- a/src/gallium/drivers/trace/tr_rbug.c +++ b/src/gallium/drivers/trace/tr_rbug.c @@ -26,6 +26,7 @@ **************************************************************************/ +#include "util/u_format.h" #include "util/u_string.h" #include "util/u_memory.h" #include "util/u_simple_list.h" @@ -203,9 +204,9 @@ trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, &t->width0, 1, &t->height0, 1, &t->depth0, 1, - pf_get_blockwidth(t->format), - pf_get_blockheight(t->format), - pf_get_blocksize(t->format), + util_format_get_blockwidth(t->format), + util_format_get_blockheight(t->format), + util_format_get_blocksize(t->format), t->last_level, t->nr_samples, t->tex_usage, @@ -254,11 +255,11 @@ trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header, rbug_send_texture_read_reply(tr_rbug->con, serial, t->texture->format, - pf_get_blockwidth(t->texture->format), - pf_get_blockheight(t->texture->format), - pf_get_blocksize(t->texture->format), + util_format_get_blockwidth(t->texture->format), + util_format_get_blockheight(t->texture->format), + util_format_get_blocksize(t->texture->format), (uint8_t*)map, - t->stride * pf_get_nblocksy(t->texture->format, t->height), + t->stride * util_format_get_nblocksy(t->texture->format, t->height), t->stride, NULL); diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index f69f7da000d..ac20a47af1e 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -25,6 +25,7 @@ * **************************************************************************/ +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_simple_list.h" @@ -425,7 +426,7 @@ trace_screen_transfer_unmap(struct pipe_screen *_screen, struct pipe_transfer *transfer = tr_trans->transfer; if(tr_trans->map) { - size_t size = pf_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride; + size_t size = util_format_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride; trace_dump_call_begin("pipe_screen", "transfer_write"); diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 69a0970d5f8..fe1390d765f 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -140,7 +140,8 @@ enum pipe_texture_target { PIPE_TEXTURE_1D = 0, PIPE_TEXTURE_2D = 1, PIPE_TEXTURE_3D = 2, - PIPE_TEXTURE_CUBE = 3 + PIPE_TEXTURE_CUBE = 3, + PIPE_MAX_TEXTURE_TYPES }; #define PIPE_TEX_FACE_POS_X 0 diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index e6bba777d34..6bfff1cc59c 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -40,353 +40,133 @@ extern "C" { #endif /** - * The pipe_format enum is a 32-bit wide bitfield that encodes all the - * information needed to uniquely describe a pixel format. - */ - -/** - * Possible format layouts are encoded in the first 2 bits. - * The interpretation of the remaining 30 bits depends on a particular - * format layout. - */ -#define PIPE_FORMAT_LAYOUT_RGBAZS 0 -#define PIPE_FORMAT_LAYOUT_YCBCR 1 -#define PIPE_FORMAT_LAYOUT_DXT 2 /**< XXX temporary? */ -#define PIPE_FORMAT_LAYOUT_MIXED 3 - -static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ -{ - return f & 0x3; -} - -/** - * RGBAZS Format Layout. - */ - -/** - * Format component selectors for RGBAZS & MIXED layout. - */ -#define PIPE_FORMAT_COMP_R 0 -#define PIPE_FORMAT_COMP_G 1 -#define PIPE_FORMAT_COMP_B 2 -#define PIPE_FORMAT_COMP_A 3 -#define PIPE_FORMAT_COMP_0 4 -#define PIPE_FORMAT_COMP_1 5 -#define PIPE_FORMAT_COMP_Z 6 -#define PIPE_FORMAT_COMP_S 7 - -/** - * Format types for RGBAZS layout. - */ -#define PIPE_FORMAT_TYPE_UNKNOWN 0 -#define PIPE_FORMAT_TYPE_FLOAT 1 /**< 16/32/64-bit/channel formats */ -#define PIPE_FORMAT_TYPE_UNORM 2 /**< uints, normalized to [0,1] */ -#define PIPE_FORMAT_TYPE_SNORM 3 /**< ints, normalized to [-1,1] */ -#define PIPE_FORMAT_TYPE_USCALED 4 /**< uints, not normalized */ -#define PIPE_FORMAT_TYPE_SSCALED 5 /**< ints, not normalized */ -#define PIPE_FORMAT_TYPE_SRGB 6 /**< sRGB colorspace */ -#define PIPE_FORMAT_TYPE_FIXED 7 /**< 16.16 fixed point */ - - -/** - * Because the destination vector is assumed to be RGBA FLOAT, we - * need to know how to swizzle and expand components from the source - * vector. - * Let's take U_A1_R5_G5_B5 as an example. X swizzle is A, X size - * is 1 bit and type is UNORM. So we take the most significant bit - * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it - * in the last component of the destination RGBA component. - * Next, Y swizzle is R, Y size is 5 and type is UNORM. We normalize - * those 5 bits into [0.0; 1.0] range and put it into second - * component of the destination vector. Rinse and repeat for - * components Z and W. - * If any of size fields is zero, it means the source format contains - * less than four components. - * If any swizzle is 0 or 1, the corresponding destination component - * should be filled with 0.0 and 1.0, respectively. - */ -typedef uint pipe_format_rgbazs_t; - -static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) -{ - return (f >> shift) & mask; -} - -#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_xyzw(f,i) pf_get(f, 2+((i)*3), 0x7) -#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ -#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ -#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ -#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ -#define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) -#define pf_exp2(f) pf_get(f, 26, 0x7) /**< Scale size by 2 ^ exp2 */ -#define pf_type(f) pf_get(f, 29, 0x7) /**< PIPE_FORMAT_TYPE_ */ - -/** - * Helper macro to encode the above structure into a 32-bit value. - */ -#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP2, TYPE ) (\ - (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ - ((SWZ) << 2) |\ - ((SIZEX) << 14) |\ - ((SIZEY) << 17) |\ - ((SIZEZ) << 20) |\ - ((SIZEW) << 23) |\ - ((EXP2) << 26) |\ - ((TYPE) << 29) ) - -/** - * Helper macro to encode the swizzle part of the structure above. - */ -#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 1-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 2-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_2( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 8-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 3, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 64-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 6, TYPE ) - -typedef uint pipe_format_mixed_t; - -/* NOTE: Use pf_swizzle_* and pf_size_* macros for swizzles and sizes. - */ - -#define pf_mixed_sign_x(f) pf_get( f, 26, 0x1 ) /*< Sign of X */ -#define pf_mixed_sign_y(f) pf_get( f, 27, 0x1 ) /*< Sign of Y */ -#define pf_mixed_sign_z(f) pf_get( f, 28, 0x1 ) /*< Sign of Z */ -#define pf_mixed_sign_w(f) pf_get( f, 29, 0x1 ) /*< Sign of W */ -#define pf_mixed_sign_xyzw(f, i) pf_get( f, 26 + (i), 0x1 ) -#define pf_mixed_normalized(f) pf_get( f, 30, 0x1 ) /*< Type is NORM (1) or SCALED (0) */ -#define pf_mixed_scale8(f) pf_get( f, 31, 0x1 ) /*< Scale size by either one (0) or eight (1) */ - -/** - * Helper macro to encode the above structure into a 32-bit value. - */ -#define _PIPE_FORMAT_MIXED( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, SIGNX, SIGNY, SIGNZ, SIGNW, NORMALIZED, SCALE8 ) (\ - (PIPE_FORMAT_LAYOUT_MIXED << 0) |\ - ((SWZ) << 2) |\ - ((SIZEX) << 14) |\ - ((SIZEY) << 17) |\ - ((SIZEZ) << 20) |\ - ((SIZEW) << 23) |\ - ((SIGNX) << 26) |\ - ((SIGNY) << 27) |\ - ((SIGNZ) << 28) |\ - ((SIGNW) << 29) |\ - ((NORMALIZED) << 30) |\ - ((SCALE8) << 31) ) - -/** - * Shorthand macro for common format swizzles. - */ -#define _PIPE_FORMAT_R001 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RG01 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RGB1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) -#define _PIPE_FORMAT_ABGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_1RGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) -#define _PIPE_FORMAT_1BGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_BGR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) -#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_0Z00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) - -/** - * YCBCR Format Layout. - */ - -/** - * This only contains a flag that indicates whether the format is reversed or - * not. - */ -typedef uint pipe_format_ycbcr_t; - -/** - * Helper macro to encode the above structure into a 32-bit value. - */ -#define _PIPE_FORMAT_YCBCR( REV ) (\ - (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ - ((REV) << 2) ) - -static INLINE uint pf_rev(pipe_format_ycbcr_t f) -{ - return (f >> 2) & 0x1; -} - - -/** - * Compresssed format layouts (this will probably change) - */ -#define _PIPE_FORMAT_DXT( LEVEL, RSIZE, GSIZE, BSIZE, ASIZE, TYPE ) \ - ((PIPE_FORMAT_LAYOUT_DXT << 0) | \ - ((LEVEL) << 2) | \ - ((RSIZE) << 5) | \ - ((GSIZE) << 8) | \ - ((BSIZE) << 11) | \ - ((ASIZE) << 14) | \ - ((TYPE) << 29)) - - - -/** * Texture/surface image formats (preliminary) */ /* KW: Added lots of surface formats to support vertex element layout - * definitions, and eventually render-to-vertex-buffer. Could - * consider making float/int/uint/scaled/normalized a separate - * parameter, but on the other hand there are special cases like - * z24s8, compressed textures, ycbcr, etc that won't fit that model. + * definitions, and eventually render-to-vertex-buffer. */ enum pipe_format { - PIPE_FORMAT_NONE = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ), - PIPE_FORMAT_A8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_X8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_1RGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_B8G8R8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_B8G8R8X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A2B10G10R10_UNORM = _PIPE_FORMAT_RGBAZS_2 ( _PIPE_FORMAT_ABGR, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ - PIPE_FORMAT_A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ - PIPE_FORMAT_I8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ - PIPE_FORMAT_A8L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ - PIPE_FORMAT_L16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ushort luminance */ - PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), - PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), - PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_S8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z24S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_X8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_0Z00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z24X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */ - PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64B64A64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32B32A32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32B32A32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32B32A32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32B32A32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32B32A32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16B16A16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16B16A16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16B16A16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16B16A16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8A8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8X8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8X8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_B6G5R5_SNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_A8B8G8R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_X8B8G8R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8X8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FIXED ), - PIPE_FORMAT_R32G32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FIXED ), - PIPE_FORMAT_R32G32B32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FIXED ), - PIPE_FORMAT_R32G32B32A32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FIXED ), + PIPE_FORMAT_NONE = 0, + PIPE_FORMAT_A8R8G8B8_UNORM = 1, + PIPE_FORMAT_X8R8G8B8_UNORM = 2, + PIPE_FORMAT_B8G8R8A8_UNORM = 3, + PIPE_FORMAT_B8G8R8X8_UNORM = 4, + PIPE_FORMAT_A1R5G5B5_UNORM = 5, + PIPE_FORMAT_A4R4G4B4_UNORM = 6, + PIPE_FORMAT_R5G6B5_UNORM = 7, + PIPE_FORMAT_A2B10G10R10_UNORM = 8, + PIPE_FORMAT_L8_UNORM = 9, /**< ubyte luminance */ + PIPE_FORMAT_A8_UNORM = 10, /**< ubyte alpha */ + PIPE_FORMAT_I8_UNORM = 11, /**< ubyte intensity */ + PIPE_FORMAT_A8L8_UNORM = 12, /**< ubyte alpha, luminance */ + PIPE_FORMAT_L16_UNORM = 13, /**< ushort luminance */ + PIPE_FORMAT_YCBCR = 14, + PIPE_FORMAT_YCBCR_REV = 15, + PIPE_FORMAT_Z16_UNORM = 16, + PIPE_FORMAT_Z32_UNORM = 17, + PIPE_FORMAT_Z32_FLOAT = 18, + PIPE_FORMAT_S8Z24_UNORM = 19, + PIPE_FORMAT_Z24S8_UNORM = 20, + PIPE_FORMAT_X8Z24_UNORM = 21, + PIPE_FORMAT_Z24X8_UNORM = 22, + PIPE_FORMAT_S8_UNORM = 23, /**< ubyte stencil */ + PIPE_FORMAT_R64_FLOAT = 24, + PIPE_FORMAT_R64G64_FLOAT = 25, + PIPE_FORMAT_R64G64B64_FLOAT = 26, + PIPE_FORMAT_R64G64B64A64_FLOAT = 27, + PIPE_FORMAT_R32_FLOAT = 28, + PIPE_FORMAT_R32G32_FLOAT = 29, + PIPE_FORMAT_R32G32B32_FLOAT = 30, + PIPE_FORMAT_R32G32B32A32_FLOAT = 31, + PIPE_FORMAT_R32_UNORM = 32, + PIPE_FORMAT_R32G32_UNORM = 33, + PIPE_FORMAT_R32G32B32_UNORM = 34, + PIPE_FORMAT_R32G32B32A32_UNORM = 35, + PIPE_FORMAT_R32_USCALED = 36, + PIPE_FORMAT_R32G32_USCALED = 37, + PIPE_FORMAT_R32G32B32_USCALED = 38, + PIPE_FORMAT_R32G32B32A32_USCALED = 39, + PIPE_FORMAT_R32_SNORM = 40, + PIPE_FORMAT_R32G32_SNORM = 41, + PIPE_FORMAT_R32G32B32_SNORM = 42, + PIPE_FORMAT_R32G32B32A32_SNORM = 43, + PIPE_FORMAT_R32_SSCALED = 44, + PIPE_FORMAT_R32G32_SSCALED = 45, + PIPE_FORMAT_R32G32B32_SSCALED = 46, + PIPE_FORMAT_R32G32B32A32_SSCALED = 47, + PIPE_FORMAT_R16_UNORM = 48, + PIPE_FORMAT_R16G16_UNORM = 49, + PIPE_FORMAT_R16G16B16_UNORM = 50, + PIPE_FORMAT_R16G16B16A16_UNORM = 51, + PIPE_FORMAT_R16_USCALED = 52, + PIPE_FORMAT_R16G16_USCALED = 53, + PIPE_FORMAT_R16G16B16_USCALED = 54, + PIPE_FORMAT_R16G16B16A16_USCALED = 55, + PIPE_FORMAT_R16_SNORM = 56, + PIPE_FORMAT_R16G16_SNORM = 57, + PIPE_FORMAT_R16G16B16_SNORM = 58, + PIPE_FORMAT_R16G16B16A16_SNORM = 59, + PIPE_FORMAT_R16_SSCALED = 60, + PIPE_FORMAT_R16G16_SSCALED = 61, + PIPE_FORMAT_R16G16B16_SSCALED = 62, + PIPE_FORMAT_R16G16B16A16_SSCALED = 63, + PIPE_FORMAT_R8_UNORM = 64, + PIPE_FORMAT_R8G8_UNORM = 65, + PIPE_FORMAT_R8G8B8_UNORM = 66, + PIPE_FORMAT_R8G8B8A8_UNORM = 67, + PIPE_FORMAT_R8G8B8X8_UNORM = 68, + PIPE_FORMAT_R8_USCALED = 69, + PIPE_FORMAT_R8G8_USCALED = 70, + PIPE_FORMAT_R8G8B8_USCALED = 71, + PIPE_FORMAT_R8G8B8A8_USCALED = 72, + PIPE_FORMAT_R8G8B8X8_USCALED = 73, + PIPE_FORMAT_R8_SNORM = 74, + PIPE_FORMAT_R8G8_SNORM = 75, + PIPE_FORMAT_R8G8B8_SNORM = 76, + PIPE_FORMAT_R8G8B8A8_SNORM = 77, + PIPE_FORMAT_R8G8B8X8_SNORM = 78, + PIPE_FORMAT_B6G5R5_SNORM = 79, + PIPE_FORMAT_A8B8G8R8_SNORM = 80, + PIPE_FORMAT_X8B8G8R8_SNORM = 81, + PIPE_FORMAT_R8_SSCALED = 82, + PIPE_FORMAT_R8G8_SSCALED = 83, + PIPE_FORMAT_R8G8B8_SSCALED = 84, + PIPE_FORMAT_R8G8B8A8_SSCALED = 85, + PIPE_FORMAT_R8G8B8X8_SSCALED = 86, + PIPE_FORMAT_R32_FIXED = 87, + PIPE_FORMAT_R32G32_FIXED = 88, + PIPE_FORMAT_R32G32B32_FIXED = 89, + PIPE_FORMAT_R32G32B32A32_FIXED = 90, /* sRGB formats */ - PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_A8L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_R8G8B8X8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_A8R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_X8R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_1RGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_B8G8R8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_B8G8R8X8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_L8_SRGB = 91, + PIPE_FORMAT_A8L8_SRGB = 92, + PIPE_FORMAT_R8G8B8_SRGB = 93, + PIPE_FORMAT_R8G8B8A8_SRGB = 94, + PIPE_FORMAT_R8G8B8X8_SRGB = 95, + PIPE_FORMAT_A8R8G8B8_SRGB = 96, + PIPE_FORMAT_X8R8G8B8_SRGB = 97, + PIPE_FORMAT_B8G8R8A8_SRGB = 98, + PIPE_FORMAT_B8G8R8X8_SRGB = 99, /* mixed formats */ - PIPE_FORMAT_X8UB8UG8SR8S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_1BGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ), - PIPE_FORMAT_B6UG5SR5S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, 0, 1, 1, 0, 1, 0 ), + PIPE_FORMAT_X8UB8UG8SR8S_NORM = 100, + PIPE_FORMAT_B6UG5SR5S_NORM = 101, /* compressed formats */ - PIPE_FORMAT_DXT1_RGB = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_DXT1_RGBA = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_DXT3_RGBA = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_DXT5_RGBA = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_DXT1_RGB = 102, + PIPE_FORMAT_DXT1_RGBA = 103, + PIPE_FORMAT_DXT3_RGBA = 104, + PIPE_FORMAT_DXT5_RGBA = 105, /* sRGB, compressed */ - PIPE_FORMAT_DXT1_SRGB = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_DXT1_SRGBA = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_DXT3_SRGBA = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_DXT5_SRGBA = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB ) + PIPE_FORMAT_DXT1_SRGB = 106, + PIPE_FORMAT_DXT1_SRGBA = 107, + PIPE_FORMAT_DXT3_SRGBA = 108, + PIPE_FORMAT_DXT5_SRGBA = 109, + + PIPE_FORMAT_COUNT }; /** @@ -394,205 +174,6 @@ enum pipe_format { */ extern const char *pf_name( enum pipe_format format ); -/** - * Return bits for a particular component. - * \param comp component index, starting at 0 - */ -static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) -{ - uint size; - - if (pf_swizzle_x(format) == comp) { - size = pf_size_x(format); - } - else if (pf_swizzle_y(format) == comp) { - size = pf_size_y(format); - } - else if (pf_swizzle_z(format) == comp) { - size = pf_size_z(format); - } - else if (pf_swizzle_w(format) == comp) { - size = pf_size_w(format); - } - else { - size = 0; - } - if (pf_layout( format ) == PIPE_FORMAT_LAYOUT_RGBAZS) - return size << pf_exp2( format ); - return size << (pf_mixed_scale8( format ) * 3); -} - - -/** - * Return total bits needed for the pixel format per block. - */ -static INLINE uint pf_get_blocksizebits( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - case PIPE_FORMAT_LAYOUT_MIXED: - return - pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); - case PIPE_FORMAT_LAYOUT_YCBCR: - assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV ); - return 32; - case PIPE_FORMAT_LAYOUT_DXT: - switch(format) { - case PIPE_FORMAT_DXT1_RGBA: - case PIPE_FORMAT_DXT1_RGB: - case PIPE_FORMAT_DXT1_SRGBA: - case PIPE_FORMAT_DXT1_SRGB: - return 64; - case PIPE_FORMAT_DXT3_RGBA: - case PIPE_FORMAT_DXT5_RGBA: - case PIPE_FORMAT_DXT3_SRGBA: - case PIPE_FORMAT_DXT5_SRGBA: - return 128; - default: - assert( 0 ); - return 0; - } - - default: - assert( 0 ); - return 0; - } -} - -/** - * Return bytes per element for the given format. - */ -static INLINE uint pf_get_blocksize( enum pipe_format format ) -{ - assert(pf_get_blocksizebits(format) % 8 == 0); - return pf_get_blocksizebits(format) / 8; -} - -static INLINE uint pf_get_blockwidth( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_YCBCR: - return 2; - case PIPE_FORMAT_LAYOUT_DXT: - return 4; - default: - return 1; - } -} - -static INLINE uint pf_get_blockheight( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_DXT: - return 4; - default: - return 1; - } -} - -static INLINE unsigned -pf_get_nblocksx(enum pipe_format format, unsigned x) -{ - unsigned blockwidth = pf_get_blockwidth(format); - return (x + blockwidth - 1) / blockwidth; -} - -static INLINE unsigned -pf_get_nblocksy(enum pipe_format format, unsigned y) -{ - unsigned blockheight = pf_get_blockheight(format); - return (y + blockheight - 1) / blockheight; -} - -static INLINE unsigned -pf_get_nblocks(enum pipe_format format, unsigned width, unsigned height) -{ - return pf_get_nblocksx(format, width) * pf_get_nblocksy(format, height); -} - -static INLINE size_t -pf_get_stride(enum pipe_format format, unsigned width) -{ - return pf_get_nblocksx(format, width) * pf_get_blocksize(format); -} - -static INLINE size_t -pf_get_2d_size(enum pipe_format format, size_t stride, unsigned height) -{ - return pf_get_nblocksy(format, height) * stride; -} - -static INLINE boolean -pf_is_depth_or_stencil( enum pipe_format format ) -{ - return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; -} - -static INLINE boolean -pf_is_depth_and_stencil( enum pipe_format format ) -{ - return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) != 0 && - pf_get_component_bits( format, PIPE_FORMAT_COMP_S ) != 0); -} - -/** DEPRECATED: For backwards compatibility */ -static INLINE boolean -pf_is_depth_stencil( enum pipe_format format ) -{ - return pf_is_depth_or_stencil( format ); -} - -static INLINE boolean -pf_is_compressed( enum pipe_format format ) -{ - return pf_layout(format) == PIPE_FORMAT_LAYOUT_DXT ? TRUE : FALSE; -} - -static INLINE boolean -pf_is_ycbcr( enum pipe_format format ) -{ - return pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ? TRUE : FALSE; -} - -static INLINE boolean -pf_has_alpha( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - case PIPE_FORMAT_LAYOUT_MIXED: - /* FIXME: pf_get_component_bits( PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_COMP_A ) should not return 0 right? */ - if(format == PIPE_FORMAT_A8_UNORM || - format == PIPE_FORMAT_A8L8_UNORM || - format == PIPE_FORMAT_A8L8_SRGB) - return TRUE; - return pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) ? TRUE : FALSE; - case PIPE_FORMAT_LAYOUT_YCBCR: - return FALSE; - case PIPE_FORMAT_LAYOUT_DXT: - switch (format) { - case PIPE_FORMAT_DXT1_RGBA: - case PIPE_FORMAT_DXT3_RGBA: - case PIPE_FORMAT_DXT5_RGBA: - case PIPE_FORMAT_DXT1_SRGBA: - case PIPE_FORMAT_DXT3_SRGBA: - case PIPE_FORMAT_DXT5_SRGBA: - return TRUE; - default: - return FALSE; - } - default: - assert( 0 ); - return FALSE; - } -} enum pipe_video_chroma_format { diff --git a/src/gallium/include/pipe/p_refcnt.h b/src/gallium/include/pipe/p_refcnt.h index e252f559006..c1c7415e023 100644 --- a/src/gallium/include/pipe/p_refcnt.h +++ b/src/gallium/include/pipe/p_refcnt.h @@ -62,6 +62,7 @@ pipe_is_referenced(struct pipe_reference *reference) * Update reference counting. * The old thing pointed to, if any, will be unreferenced. * Both 'ptr' and 'reference' may be NULL. + * \return TRUE if the object's refcount hits zero and should be destroyed. */ static INLINE boolean pipe_reference(struct pipe_reference *ptr, struct pipe_reference *reference) diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index c051d4dae84..5da85bbbc24 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -55,6 +55,7 @@ struct tgsi_processor #define TGSI_TOKEN_TYPE_DECLARATION 0 #define TGSI_TOKEN_TYPE_IMMEDIATE 1 #define TGSI_TOKEN_TYPE_INSTRUCTION 2 +#define TGSI_TOKEN_TYPE_PROPERTY 3 struct tgsi_token { @@ -64,16 +65,17 @@ struct tgsi_token }; enum tgsi_file_type { - TGSI_FILE_NULL =0, - TGSI_FILE_CONSTANT =1, - TGSI_FILE_INPUT =2, - TGSI_FILE_OUTPUT =3, - TGSI_FILE_TEMPORARY =4, - TGSI_FILE_SAMPLER =5, - TGSI_FILE_ADDRESS =6, - TGSI_FILE_IMMEDIATE =7, - TGSI_FILE_LOOP =8, - TGSI_FILE_PREDICATE =9, + TGSI_FILE_NULL =0, + TGSI_FILE_CONSTANT =1, + TGSI_FILE_INPUT =2, + TGSI_FILE_OUTPUT =3, + TGSI_FILE_TEMPORARY =4, + TGSI_FILE_SAMPLER =5, + TGSI_FILE_ADDRESS =6, + TGSI_FILE_IMMEDIATE =7, + TGSI_FILE_LOOP =8, + TGSI_FILE_PREDICATE =9, + TGSI_FILE_SYSTEM_VALUE =10, TGSI_FILE_COUNT /**< how many TGSI_FILE_ types */ }; @@ -152,6 +154,22 @@ union tgsi_immediate_data float Float; }; +#define TGSI_PROPERTY_GS_INPUT_PRIM 0 +#define TGSI_PROPERTY_GS_OUTPUT_PRIM 1 +#define TGSI_PROPERTY_GS_MAX_VERTICES 2 +#define TGSI_PROPERTY_COUNT 3 + +struct tgsi_property { + unsigned Type : 4; /**< TGSI_TOKEN_TYPE_PROPERTY */ + unsigned NrTokens : 8; /**< UINT */ + unsigned PropertyName : 8; /**< one of TGSI_PROPERTY */ + unsigned Padding : 12; +}; + +struct tgsi_property_data { + unsigned Data; +}; + /* TGSI opcodes. * * For more information on semantics of opcodes and diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c index 099cf1e0645..4b12243ddff 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.c +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -44,6 +44,7 @@ #include "state_tracker/st_context.h" #include "state_tracker/st_cb_fbo.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_rect.h" diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c index 737bdfdd346..0b60b5be059 100644 --- a/src/gallium/state_trackers/egl/egl_surface.c +++ b/src/gallium/state_trackers/egl/egl_surface.c @@ -12,6 +12,7 @@ #include "state_tracker/drm_api.h" +#include "util/u_format.h" #include "util/u_rect.h" /* diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c index 97ca2afc543..96377414211 100644 --- a/src/gallium/state_trackers/python/st_sample.c +++ b/src/gallium/state_trackers/python/st_sample.c @@ -30,6 +30,7 @@ #include "pipe/p_format.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_tile.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -461,7 +462,7 @@ st_sample_dxt_pixel_block(enum pipe_format format, for(ch = 0; ch < 4; ++ch) rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f; - memcpy(raw, data[i].raw, pf_get_blocksize(format)); + memcpy(raw, data[i].raw, util_format_get_blocksize(format)); } @@ -473,7 +474,7 @@ st_sample_generic_pixel_block(enum pipe_format format, { unsigned i; unsigned x, y, ch; - int blocksize = pf_get_blocksize(format); + int blocksize = util_format_get_blocksize(format); for(i = 0; i < blocksize; ++i) raw[i] = (uint8_t)st_random(); @@ -548,11 +549,11 @@ st_sample_surface(struct st_surface *surface, float *rgba) if (raw) { enum pipe_format format = texture->format; uint x, y; - int nblocksx = pf_get_nblocksx(format, width); - int nblocksy = pf_get_nblocksy(format, height); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); - int blocksize = pf_get_blocksize(format); + int nblocksx = util_format_get_nblocksx(format, width); + int nblocksy = util_format_get_nblocksy(format, height); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); for (y = 0; y < nblocksy; ++y) { diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 43c61af1ffd..a3294e877a6 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -40,6 +40,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "softpipe/sp_winsys.h" @@ -168,8 +169,8 @@ st_softpipe_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index 202ccfc3509..b022d073fda 100755 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -47,7 +47,7 @@ for name, value in globals().items(): formats[value] = name def is_depth_stencil_format(format): - # FIXME: make and use binding to pf_is_depth_stencil + # FIXME: make and use binding to util_format_is_depth_or_stencil return format in ( PIPE_FORMAT_Z32_UNORM, PIPE_FORMAT_Z24S8_UNORM, diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c index eb135c1ff4f..2f984fb7b9a 100644 --- a/src/gallium/state_trackers/vega/api_filters.c +++ b/src/gallium/state_trackers/vega/api_filters.c @@ -38,6 +38,7 @@ #include "pipe/p_screen.h" #include "pipe/p_shader_tokens.h" +#include "util/u_format.h" #include "util/u_memory.h" diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c index 172311851e9..278ba6d46eb 100644 --- a/src/gallium/state_trackers/vega/image.c +++ b/src/gallium/state_trackers/vega/image.c @@ -39,6 +39,7 @@ #include "pipe/p_screen.h" #include "pipe/p_inlines.h" #include "util/u_blit.h" +#include "util/u_format.h" #include "util/u_tile.h" #include "util/u_memory.h" #include "util/u_math.h" diff --git a/src/gallium/state_trackers/vega/mask.c b/src/gallium/state_trackers/vega/mask.c index 868c28239a3..42300bb6d57 100644 --- a/src/gallium/state_trackers/vega/mask.c +++ b/src/gallium/state_trackers/vega/mask.c @@ -36,6 +36,7 @@ #include "pipe/p_context.h" #include "pipe/p_screen.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" struct vg_mask_layer { diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c index 785c9829432..cc73771d358 100644 --- a/src/gallium/state_trackers/vega/paint.c +++ b/src/gallium/state_trackers/vega/paint.c @@ -34,6 +34,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_math.h" diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c index c85dae0282e..64e3a7c5453 100644 --- a/src/gallium/state_trackers/vega/renderer.c +++ b/src/gallium/state_trackers/vega/renderer.c @@ -35,6 +35,7 @@ #include "pipe/p_shader_tokens.h" #include "util/u_draw_quad.h" +#include "util/u_format.h" #include "util/u_simple_shaders.h" #include "util/u_memory.h" #include "util/u_rect.h" @@ -56,7 +57,7 @@ static void setup_shaders(struct renderer *ctx) { struct pipe_context *pipe = ctx->pipe; /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D); } static struct pipe_buffer * diff --git a/src/gallium/state_trackers/vega/vg_tracker.c b/src/gallium/state_trackers/vega/vg_tracker.c index 5a286b14491..e5039132758 100644 --- a/src/gallium/state_trackers/vega/vg_tracker.c +++ b/src/gallium/state_trackers/vega/vg_tracker.c @@ -31,6 +31,7 @@ #include "pipe/p_context.h" #include "pipe/p_inlines.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_math.h" #include "util/u_rect.h" @@ -56,7 +57,7 @@ create_texture(struct pipe_context *pipe, enum pipe_format format, templ.depth0 = 1; templ.last_level = 0; - if (pf_get_component_bits(format, PIPE_FORMAT_COMP_S)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)) { templ.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; } else { templ.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | diff --git a/src/gallium/state_trackers/wgl/stw_framebuffer.c b/src/gallium/state_trackers/wgl/stw_framebuffer.c index 6d095019815..129a6298a77 100644 --- a/src/gallium/state_trackers/wgl/stw_framebuffer.c +++ b/src/gallium/state_trackers/wgl/stw_framebuffer.c @@ -30,6 +30,7 @@ #include "main/context.h" #include "pipe/p_format.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "state_tracker/st_context.h" #include "state_tracker/st_public.h" @@ -267,15 +268,13 @@ stw_framebuffer_allocate( enum pipe_format colorFormat, depthFormat, stencilFormat; colorFormat = pfi->color_format; - - assert(pf_layout( pfi->depth_stencil_format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); - - if(pf_get_component_bits( pfi->depth_stencil_format, PIPE_FORMAT_COMP_Z )) + + if(util_format_get_component_bits(pfi->depth_stencil_format, UTIL_FORMAT_COLORSPACE_ZS, 0)) depthFormat = pfi->depth_stencil_format; else depthFormat = PIPE_FORMAT_NONE; - if(pf_get_component_bits( pfi->depth_stencil_format, PIPE_FORMAT_COMP_S )) + if(util_format_get_component_bits(pfi->depth_stencil_format, UTIL_FORMAT_COLORSPACE_ZS, 1)) stencilFormat = pfi->depth_stencil_format; else stencilFormat = PIPE_FORMAT_NONE; diff --git a/src/gallium/state_trackers/wgl/stw_pixelformat.c b/src/gallium/state_trackers/wgl/stw_pixelformat.c index 7abe5d9f7fa..54cc3614129 100644 --- a/src/gallium/state_trackers/wgl/stw_pixelformat.c +++ b/src/gallium/state_trackers/wgl/stw_pixelformat.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "util/u_debug.h" #include "stw_icd.h" @@ -132,14 +133,12 @@ stw_pixelformat_add( if(stw_dev->pixelformat_extended_count >= STW_MAX_PIXELFORMATS) return; - assert(pf_layout( color->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); - assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_R ) == color->bits.red ); - assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_G ) == color->bits.green ); - assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_B ) == color->bits.blue ); - assert(pf_get_component_bits( color->format, PIPE_FORMAT_COMP_A ) == color->bits.alpha ); - assert(pf_layout( depth->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); - assert(pf_get_component_bits( depth->format, PIPE_FORMAT_COMP_Z ) == depth->bits.depth ); - assert(pf_get_component_bits( depth->format, PIPE_FORMAT_COMP_S ) == depth->bits.stencil ); + assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 0) == color->bits.red); + assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 1) == color->bits.green); + assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 2) == color->bits.blue); + assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 3) == color->bits.alpha); + assert(util_format_get_component_bits(depth->format, UTIL_FORMAT_COLORSPACE_ZS, 0) == depth->bits.depth); + assert(util_format_get_component_bits(depth->format, UTIL_FORMAT_COLORSPACE_ZS, 1) == depth->bits.stencil); pfi = &stw_dev->pixelformats[stw_dev->pixelformat_extended_count]; diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c index 03bc6e37c9c..c776faa53f8 100644 --- a/src/gallium/state_trackers/xorg/xorg_crtc.c +++ b/src/gallium/state_trackers/xorg/xorg_crtc.c @@ -50,6 +50,7 @@ #endif #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_rect.h" #ifdef HAVE_LIBKMS @@ -256,7 +257,7 @@ crtc_load_cursor_argb_kms(xf86CrtcPtr crtc, CARD32 * image) return; err_bo_destroy: - kms_bo_destroy(crtcp->cursor_bo); + kms_bo_destroy(&crtcp->cursor_bo); } #endif @@ -304,10 +305,8 @@ xorg_crtc_cursor_destroy(xf86CrtcPtr crtc) pipe_texture_reference(&crtcp->cursor_tex, NULL); #ifdef HAVE_LIBKMS if (crtcp->cursor_bo) - kms_bo_destroy(crtcp->cursor_bo); + kms_bo_destroy(&crtcp->cursor_bo); #endif - - xfree(crtcp); } /* @@ -319,11 +318,12 @@ crtc_destroy(xf86CrtcPtr crtc) { struct crtc_private *crtcp = crtc->driver_private; - if (crtcp->cursor_tex) - pipe_texture_reference(&crtcp->cursor_tex, NULL); + xorg_crtc_cursor_destroy(crtc); drmModeFreeCrtc(crtcp->drm_crtc); + xfree(crtcp); + crtc->driver_private = NULL; } static const xf86CrtcFuncsRec crtc_funcs = { diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index bdd6a8b2051..fd82f4fa1d1 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -40,6 +40,7 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_rect.h" /* Make all the #if cases in the code esier to read */ @@ -92,7 +93,7 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form case 9: #endif if (exa_priv->depth_stencil_tex && - !pf_is_depth_stencil(exa_priv->depth_stencil_tex->format)) + !util_format_is_depth_or_stencil(exa_priv->depth_stencil_tex->format)) exa_priv->depth_stencil_tex = NULL; /* Fall through */ case DRI2BufferDepth: diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c index da86295c316..8a24aa10a3c 100644 --- a/src/gallium/state_trackers/xorg/xorg_driver.c +++ b/src/gallium/state_trackers/xorg/xorg_driver.c @@ -260,8 +260,7 @@ drv_close_resource_management(ScrnInfoPtr pScrn) #ifdef HAVE_LIBKMS if (ms->kms) - kms_destroy(ms->kms); - ms->kms = NULL; + kms_destroy(&ms->kms); #endif return TRUE; @@ -481,7 +480,7 @@ static void drv_block_handler(int i, pointer blockData, pointer pTimeout, if (num_cliprects) { drmModeClip *clip = alloca(num_cliprects * sizeof(drmModeClip)); BoxPtr rect = REGION_RECTS(dirty); - int i; + int i, ret; /* XXX no need for copy? */ for (i = 0; i < num_cliprects; i++, rect++) { @@ -492,7 +491,11 @@ static void drv_block_handler(int i, pointer blockData, pointer pTimeout, } /* TODO query connector property to see if this is needed */ - drmModeDirtyFB(ms->fd, ms->fb_id, clip, num_cliprects); + ret = drmModeDirtyFB(ms->fd, ms->fb_id, clip, num_cliprects); + if (ret) { + debug_printf("%s: failed to send dirty (%i, %s)\n", + __func__, ret, strerror(-ret)); + } DamageEmpty(ms->damage); } @@ -688,6 +691,9 @@ drv_leave_vt(int scrnIndex, int flags) xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn); int o; + if (ms->winsys_leave_vt) + ms->winsys_leave_vt(pScrn); + for (o = 0; o < config->num_crtc; o++) { xf86CrtcPtr crtc = config->crtc[o]; @@ -750,6 +756,9 @@ drv_enter_vt(int scrnIndex, int flags) if (!xf86SetDesiredModes(pScrn)) return FALSE; + if (ms->winsys_enter_vt) + ms->winsys_enter_vt(pScrn); + return TRUE; } @@ -832,6 +841,7 @@ drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn) modesettingPtr ms = modesettingPTR(pScrn); unsigned handle, stride; struct pipe_texture *tex; + int ret; ms->noEvict = TRUE; @@ -845,16 +855,21 @@ drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn) tex, &stride, &handle)) - return FALSE; + goto err_destroy; - drmModeAddFB(ms->fd, - pScrn->virtualX, - pScrn->virtualY, - pScrn->depth, - pScrn->bitsPerPixel, - stride, - handle, - &ms->fb_id); + ret = drmModeAddFB(ms->fd, + pScrn->virtualX, + pScrn->virtualY, + pScrn->depth, + pScrn->bitsPerPixel, + stride, + handle, + &ms->fb_id); + if (ret) { + debug_printf("%s: failed to create framebuffer (%i, %s)", + __func__, ret, strerror(-ret)); + goto err_destroy; + } pScrn->frameX0 = 0; pScrn->frameY0 = 0; @@ -864,6 +879,10 @@ drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn) pipe_texture_reference(&tex, NULL); return TRUE; + +err_destroy: + pipe_texture_reference(&tex, NULL); + return FALSE; } static Bool @@ -893,13 +912,14 @@ static Bool drv_destroy_front_buffer_kms(ScrnInfoPtr pScrn) { modesettingPtr ms = modesettingPTR(pScrn); + ScreenPtr pScreen = pScrn->pScreen; + PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen); if (!ms->root_bo) return TRUE; kms_bo_unmap(ms->root_bo); - kms_bo_destroy(ms->root_bo); - ms->root_bo = NULL; + kms_bo_destroy(&ms->root_bo); return TRUE; } @@ -910,6 +930,7 @@ drv_create_front_buffer_kms(ScrnInfoPtr pScrn) unsigned handle, stride; struct kms_bo *bo; unsigned attr[8]; + int ret; attr[0] = KMS_BO_TYPE; attr[1] = KMS_BO_TYPE_SCANOUT; @@ -928,14 +949,19 @@ drv_create_front_buffer_kms(ScrnInfoPtr pScrn) if (kms_bo_get_prop(bo, KMS_HANDLE, &handle)) goto err_destroy; - drmModeAddFB(ms->fd, - pScrn->virtualX, - pScrn->virtualY, - pScrn->depth, - pScrn->bitsPerPixel, - stride, - handle, - &ms->fb_id); + ret = drmModeAddFB(ms->fd, + pScrn->virtualX, + pScrn->virtualY, + pScrn->depth, + pScrn->bitsPerPixel, + stride, + handle, + &ms->fb_id); + if (ret) { + debug_printf("%s: failed to create framebuffer (%i, %s)", + __func__, ret, strerror(-ret)); + goto err_destroy; + } pScrn->frameX0 = 0; pScrn->frameY0 = 0; @@ -945,7 +971,7 @@ drv_create_front_buffer_kms(ScrnInfoPtr pScrn) return TRUE; err_destroy: - kms_bo_destroy(bo); + kms_bo_destroy(&bo); return FALSE; } @@ -962,7 +988,7 @@ drv_bind_front_buffer_kms(ScrnInfoPtr pScrn) return FALSE; if (kms_bo_map(ms->root_bo, &ptr)) - return FALSE; + goto err_destroy; pScreen->ModifyPixmapHeader(rootPixmap, pScreen->width, @@ -972,6 +998,10 @@ drv_bind_front_buffer_kms(ScrnInfoPtr pScrn) stride, ptr); return TRUE; + +err_destroy: + kms_bo_destroy(&ms->root_bo); + return FALSE; } #endif /* HAVE_LIBKMS */ diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c index 55b87aefa15..d5c005ebadd 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.c +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -43,6 +43,7 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_rect.h" #include "util/u_math.h" #include "util/u_debug.h" @@ -343,6 +344,9 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg) #if DEBUG_PRINT debug_printf("ExaPrepareSolid(0x%x)\n", fg); #endif + if (!exa->accel) + return FALSE; + if (!exa->pipe) XORG_FALLBACK("accle not enabled"); @@ -361,7 +365,7 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg) XORG_FALLBACK("format %s", pf_name(priv->tex->format)); } - return exa->accel && xorg_solid_bind_state(exa, priv, fg); + return xorg_solid_bind_state(exa, priv, fg); } static void @@ -417,6 +421,10 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir, #if DEBUG_PRINT debug_printf("ExaPrepareCopy\n"); #endif + + if (!exa->accel) + return FALSE; + if (!exa->pipe) XORG_FALLBACK("accle not enabled"); @@ -490,7 +498,7 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir, } - return exa->accel; + return TRUE; } static void @@ -598,15 +606,19 @@ ExaCheckComposite(int op, ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); struct exa_context *exa = ms->exa; - boolean accelerated = xorg_composite_accelerated(op, - pSrcPicture, - pMaskPicture, - pDstPicture); + #if DEBUG_PRINT debug_printf("ExaCheckComposite(%d, %p, %p, %p) = %d\n", op, pSrcPicture, pMaskPicture, pDstPicture, accelerated); #endif - return exa->accel && accelerated; + + if (!exa->accel) + return FALSE; + + return xorg_composite_accelerated(op, + pSrcPicture, + pMaskPicture, + pDstPicture); } @@ -620,6 +632,9 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture, struct exa_context *exa = ms->exa; struct exa_pixmap_priv *priv; + if (!exa->accel) + return FALSE; + #if DEBUG_PRINT debug_printf("ExaPrepareComposite(%d, src=0x%p, mask=0x%p, dst=0x%p)\n", op, pSrcPicture, pMaskPicture, pDstPicture); @@ -678,8 +693,7 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture, render_format_name(pMaskPicture->format)); } - return exa->accel && - xorg_composite_bind_state(exa, op, pSrcPicture, pMaskPicture, + return xorg_composite_bind_state(exa, op, pSrcPicture, pMaskPicture, pDstPicture, pSrc ? exaGetPixmapDriverPrivate(pSrc) : NULL, pMask ? exaGetPixmapDriverPrivate(pMask) : NULL, diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.c b/src/gallium/state_trackers/xorg/xorg_renderer.c index 248d658ac2c..ba15f8a7845 100644 --- a/src/gallium/state_trackers/xorg/xorg_renderer.c +++ b/src/gallium/state_trackers/xorg/xorg_renderer.c @@ -5,6 +5,7 @@ #include "cso_cache/cso_context.h" #include "util/u_draw_quad.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_rect.h" diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h index d5fc18448ef..c0cfbe60616 100644 --- a/src/gallium/state_trackers/xorg/xorg_tracker.h +++ b/src/gallium/state_trackers/xorg/xorg_tracker.h @@ -117,6 +117,8 @@ typedef struct _modesettingRec /* winsys hocks */ Bool (*winsys_screen_init)(ScrnInfoPtr pScr); Bool (*winsys_screen_close)(ScrnInfoPtr pScr); + Bool (*winsys_enter_vt)(ScrnInfoPtr pScr); + Bool (*winsys_leave_vt)(ScrnInfoPtr pScr); void *winsys_priv; #ifdef DRM_MODE_FEATURE_DIRTYFB diff --git a/src/gallium/state_trackers/xorg/xorg_xv.c b/src/gallium/state_trackers/xorg/xorg_xv.c index c965e4b9b54..6b5a41a3727 100644 --- a/src/gallium/state_trackers/xorg/xorg_xv.c +++ b/src/gallium/state_trackers/xorg/xorg_xv.c @@ -13,6 +13,8 @@ #include "pipe/p_screen.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" + /*XXX get these from pipe's texture limits */ #define IMAGE_MAX_WIDTH 2048 #define IMAGE_MAX_HEIGHT 2048 @@ -212,17 +214,67 @@ check_yuv_textures(struct xorg_xv_port_priv *priv, int width, int height) return Success; } +static int +query_image_attributes(ScrnInfoPtr pScrn, + int id, + unsigned short *w, unsigned short *h, + int *pitches, int *offsets) +{ + int size, tmp; + + if (*w > IMAGE_MAX_WIDTH) + *w = IMAGE_MAX_WIDTH; + if (*h > IMAGE_MAX_HEIGHT) + *h = IMAGE_MAX_HEIGHT; + + *w = (*w + 1) & ~1; + if (offsets) + offsets[0] = 0; + + switch (id) { + case FOURCC_YV12: + *h = (*h + 1) & ~1; + size = (*w + 3) & ~3; + if (pitches) { + pitches[0] = size; + } + size *= *h; + if (offsets) { + offsets[1] = size; + } + tmp = ((*w >> 1) + 3) & ~3; + if (pitches) { + pitches[1] = pitches[2] = tmp; + } + tmp *= (*h >> 1); + size += tmp; + if (offsets) { + offsets[2] = size; + } + size += tmp; + break; + case FOURCC_UYVY: + case FOURCC_YUY2: + default: + size = *w << 1; + if (pitches) + pitches[0] = size; + size *= *h; + break; + } + + return size; +} + static void copy_packed_data(ScrnInfoPtr pScrn, struct xorg_xv_port_priv *port, int id, unsigned char *buf, - int srcPitch, int left, int top, - int w, int h) + unsigned short w, unsigned short h) { - unsigned char *src; int i, j; struct pipe_texture **dst = port->yuv[port->current_set]; struct pipe_transfer *ytrans, *utrans, *vtrans; @@ -232,8 +284,6 @@ copy_packed_data(ScrnInfoPtr pScrn, int yidx, uidx, vidx; int y_array_size = w * h; - src = buf + (top * srcPitch) + (left << 1); - ytrans = screen->get_tex_transfer(screen, dst[0], 0, 0, 0, PIPE_TRANSFER_WRITE, @@ -255,15 +305,22 @@ copy_packed_data(ScrnInfoPtr pScrn, switch (id) { case FOURCC_YV12: { - for (i = 0; i < w; ++i) { - for (j = 0; j < h; ++j) { - /*XXX use src? */ - y1 = buf[j*w + i]; - u = buf[(j/2) * (w/2) + i/2 + y_array_size]; - v = buf[(j/2) * (w/2) + i/2 + y_array_size + y_array_size/4]; - ymap[yidx++] = y1; - umap[uidx++] = u; - vmap[vidx++] = v; + int pitches[3], offsets[3]; + unsigned char *y, *u, *v; + query_image_attributes(pScrn, FOURCC_YV12, + &w, &h, pitches, offsets); + + y = buf + offsets[0]; + v = buf + offsets[1]; + u = buf + offsets[2]; + for (i = 0; i < h; ++i) { + for (j = 0; j < w; ++j) { + int yoffset = (w*i+j); + int ii = (i|1), jj = (j|1); + int vuoffset = (w/2)*(ii/2) + (jj/2); + ymap[yidx++] = y[yoffset]; + umap[uidx++] = u[vuoffset]; + vmap[vidx++] = v[vuoffset]; } } } @@ -510,7 +567,6 @@ put_image(ScrnInfoPtr pScrn, ScreenPtr pScreen = screenInfo.screens[pScrn->scrnIndex]; PixmapPtr pPixmap; INT32 x1, x2, y1, y2; - int srcPitch; BoxRec dstBox; int ret; @@ -529,21 +585,12 @@ put_image(ScrnInfoPtr pScrn, width, height)) return Success; - switch (id) { - case FOURCC_UYVY: - case FOURCC_YUY2: - case FOURCC_YV12: - default: - srcPitch = width << 1; - break; - } - ret = check_yuv_textures(pPriv, width, height); if (ret) return ret; - copy_packed_data(pScrn, pPriv, id, buf, srcPitch, + copy_packed_data(pScrn, pPriv, id, buf, src_x, src_y, width, height); if (pDraw->type == DRAWABLE_WINDOW) { @@ -561,38 +608,6 @@ put_image(ScrnInfoPtr pScrn, return Success; } -static int -query_image_attributes(ScrnInfoPtr pScrn, - int id, - unsigned short *w, unsigned short *h, - int *pitches, int *offsets) -{ - int size; - - if (*w > IMAGE_MAX_WIDTH) - *w = IMAGE_MAX_WIDTH; - if (*h > IMAGE_MAX_HEIGHT) - *h = IMAGE_MAX_HEIGHT; - - *w = (*w + 1) & ~1; - if (offsets) - offsets[0] = 0; - - switch (id) { - case FOURCC_UYVY: - case FOURCC_YUY2: - case FOURCC_YV12: - default: - size = *w << 1; - if (pitches) - pitches[0] = size; - size *= *h; - break; - } - - return size; -} - static struct xorg_xv_port_priv * port_priv_create(struct xorg_renderer *r) { diff --git a/src/gallium/winsys/drm/SConscript b/src/gallium/winsys/drm/SConscript index 9f7b383d2d3..66b73a8bf93 100644 --- a/src/gallium/winsys/drm/SConscript +++ b/src/gallium/winsys/drm/SConscript @@ -58,6 +58,11 @@ if env['dri']: 'intel/SConscript', ]) + if 'i965' in env['winsys']: + SConscript([ + 'i965/SConscript', + ]) + if 'radeon' in env['winsys']: SConscript([ 'radeon/SConscript', diff --git a/src/gallium/winsys/drm/i965/Makefile b/src/gallium/winsys/drm/i965/Makefile new file mode 100644 index 00000000000..d8feef6824a --- /dev/null +++ b/src/gallium/winsys/drm/i965/Makefile @@ -0,0 +1,12 @@ +# src/gallium/winsys/drm/intel/Makefile +TOP = ../../../../.. +include $(TOP)/configs/current + +SUBDIRS = gem $(GALLIUM_STATE_TRACKERS_DIRS) + +default install clean: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) $@) || exit 1; \ + fi \ + done diff --git a/src/gallium/winsys/drm/i965/SConscript b/src/gallium/winsys/drm/i965/SConscript new file mode 100644 index 00000000000..50d7b75ed68 --- /dev/null +++ b/src/gallium/winsys/drm/i965/SConscript @@ -0,0 +1,7 @@ +Import('*') + +SConscript(['gem/SConscript',]) + +if 'mesa' in env['statetrackers']: + + SConscript(['dri/SConscript']) diff --git a/src/gallium/winsys/drm/i965/dri/Makefile b/src/gallium/winsys/drm/i965/dri/Makefile new file mode 100644 index 00000000000..f7e81eed87b --- /dev/null +++ b/src/gallium/winsys/drm/i965/dri/Makefile @@ -0,0 +1,26 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = i965_dri.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/dri/libdridrm.a \ + $(TOP)/src/gallium/winsys/drm/i965/gem/libi965drm.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/identity/libidentity.a \ + $(TOP)/src/gallium/drivers/i965/libi965.a + + +DRIVER_SOURCES = + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +include ../../Makefile.template + +DRI_LIB_DEPS += -ldrm_intel + +symlinks: $(TOP)/$(LIB_DIR)/gallium + @rm -f $(TOP)/$(LIB_DIR)/gallium/i965_dri.so diff --git a/src/gallium/winsys/drm/i965/dri/SConscript b/src/gallium/winsys/drm/i965/dri/SConscript new file mode 100644 index 00000000000..233ef464be5 --- /dev/null +++ b/src/gallium/winsys/drm/i965/dri/SConscript @@ -0,0 +1,19 @@ +Import('*') + +env = drienv.Clone() + +env.ParseConfig('pkg-config --cflags --libs libdrm_intel') + +drivers = [ + st_dri, + i965drm, + i965, + trace, +] + +env.LoadableModule( + target ='i965_dri.so', + source = COMMON_GALLIUM_SOURCES, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], + SHLIBPREFIX = '', +) diff --git a/src/gallium/winsys/drm/i965/egl/Makefile b/src/gallium/winsys/drm/i965/egl/Makefile new file mode 100644 index 00000000000..a1b32eb2a79 --- /dev/null +++ b/src/gallium/winsys/drm/i965/egl/Makefile @@ -0,0 +1,29 @@ +TOP = ../../../../../.. +GALLIUMDIR = ../../../.. +include $(TOP)/configs/current + +LIBNAME = EGL_i965.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/state_trackers/egl/libegldrm.a \ + $(GALLIUMDIR)/winsys/drm/i965/gem/libi965drm.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ + $(TOP)/src/gallium/drivers/i965/libi965.a + +DRIVER_SOURCES = + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +DRIVER_EXTRAS = -ldrm_intel + +ASM_SOURCES = + +DRIVER_DEFINES = -I../gem $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../../Makefile.template + +symlinks: diff --git a/src/gallium/winsys/drm/i965/gem/Makefile b/src/gallium/winsys/drm/i965/gem/Makefile new file mode 100644 index 00000000000..6a7497b6be3 --- /dev/null +++ b/src/gallium/winsys/drm/i965/gem/Makefile @@ -0,0 +1,14 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = i965drm + +C_SOURCES = \ + i965_drm_buffer.c \ + i965_drm_api.c + +LIBRARY_INCLUDES = $(shell pkg-config libdrm --cflags-only-I) + +LIBRARY_DEFINES = $(shell pkg-config libdrm --cflags-only-other) + +include ../../../../Makefile.template diff --git a/src/gallium/winsys/drm/i965/gem/SConscript b/src/gallium/winsys/drm/i965/gem/SConscript new file mode 100644 index 00000000000..6256ec6eaf0 --- /dev/null +++ b/src/gallium/winsys/drm/i965/gem/SConscript @@ -0,0 +1,15 @@ +Import('*') + +env = drienv.Clone() + +i965drm_sources = [ + 'i965_drm_api.c', + 'i965_drm_buffer.c', +] + +i965drm = env.ConvenienceLibrary( + target ='i965drm', + source = i965drm_sources, +) + +Export('i965drm') diff --git a/src/gallium/winsys/drm/i965/gem/i965_drm_api.c b/src/gallium/winsys/drm/i965/gem/i965_drm_api.c new file mode 100644 index 00000000000..fc9678d2b62 --- /dev/null +++ b/src/gallium/winsys/drm/i965/gem/i965_drm_api.c @@ -0,0 +1,243 @@ + +#include <stdio.h> +#include "state_tracker/drm_api.h" + +#include "i965_drm_winsys.h" +#include "util/u_memory.h" + +#include "i965/brw_context.h" /* XXX: shouldn't be doing this */ +#include "i965/brw_screen.h" /* XXX: shouldn't be doing this */ + +#include "trace/tr_drm.h" + +/* + * Helper functions + */ + + +static void +i965_libdrm_get_device_id(unsigned int *device_id) +{ + char path[512]; + FILE *file; + void *shutup_gcc; + + /* + * FIXME: Fix this up to use a drm ioctl or whatever. + */ + + snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device"); + file = fopen(path, "r"); + if (!file) { + return; + } + + shutup_gcc = fgets(path, sizeof(path), file); + sscanf(path, "%x", device_id); + fclose(file); +} + +static struct i965_libdrm_buffer * +i965_libdrm_buffer_from_handle(struct i965_libdrm_winsys *idws, + const char* name, unsigned handle) +{ + struct i965_libdrm_buffer *buf = CALLOC_STRUCT(i965_libdrm_buffer); + uint32_t swizzle = 0; + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + if (!buf) + return NULL; + pipe_reference_init(&buf->base.reference, 1); + buf->bo = drm_intel_bo_gem_create_from_name(idws->gem, name, handle); + buf->base.size = buf->bo->size; + buf->base.sws = &idws->base; + buf->flinked = TRUE; + buf->flink = handle; + + + if (!buf->bo) + goto err; + + drm_intel_bo_get_tiling(buf->bo, &buf->tiling, &swizzle); + if (buf->tiling != 0) + buf->map_gtt = TRUE; + + return buf; + +err: + FREE(buf); + return NULL; +} + + +/* + * Exported functions + */ + + +static struct pipe_texture * +i965_libdrm_texture_from_shared_handle(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *template, + const char* name, + unsigned pitch, + unsigned handle) +{ + /* XXX: this is silly -- there should be a way to get directly from + * the "drm_api" struct to ourselves, without peering into + * unrelated code: + */ + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(brw_screen(screen)->sws); + struct i965_libdrm_buffer *buffer; + + if (BRW_DUMP) + debug_printf("%s %s pitch %d handle 0x%x\n", __FUNCTION__, + name, pitch, handle); + + buffer = i965_libdrm_buffer_from_handle(idws, name, handle); + if (!buffer) + return NULL; + + return brw_texture_blanket_winsys_buffer(screen, template, pitch, + buffer->tiling, + &buffer->base); +} + + +static boolean +i965_libdrm_shared_handle_from_texture(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned *pitch, + unsigned *handle) +{ + struct i965_libdrm_buffer *buf = NULL; + struct brw_winsys_buffer *buffer = NULL; + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + if (!brw_texture_get_winsys_buffer(texture, &buffer, pitch)) + return FALSE; + + buf = i965_libdrm_buffer(buffer); + if (!buf->flinked) { + if (drm_intel_bo_flink(buf->bo, &buf->flink)) + return FALSE; + buf->flinked = TRUE; + } + + *handle = buf->flink; + + if (BRW_DUMP) + debug_printf(" -> pitch %d handle 0x%x\n", *pitch, *handle); + + return TRUE; +} + +static boolean +i965_libdrm_local_handle_from_texture(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned *pitch, + unsigned *handle) +{ + struct brw_winsys_buffer *buffer = NULL; + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + if (!brw_texture_get_winsys_buffer(texture, &buffer, pitch)) + return FALSE; + + *handle = i965_libdrm_buffer(buffer)->bo->handle; + + if (BRW_DUMP) + debug_printf(" -> pitch %d handle 0x%x\n", *pitch, *handle); + + return TRUE; +} + +static void +i965_libdrm_winsys_destroy(struct brw_winsys_screen *iws) +{ + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(iws); + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + drm_intel_bufmgr_destroy(idws->gem); + + FREE(idws); +} + +static struct pipe_screen * +i965_libdrm_create_screen(struct drm_api *api, int drmFD, + struct drm_create_screen_arg *arg) +{ + struct i965_libdrm_winsys *idws; + unsigned int deviceID; + + debug_printf("%s\n", __FUNCTION__); + + if (arg != NULL) { + switch(arg->mode) { + case DRM_CREATE_NORMAL: + break; + default: + return NULL; + } + } + + idws = CALLOC_STRUCT(i965_libdrm_winsys); + if (!idws) + return NULL; + + i965_libdrm_get_device_id(&deviceID); + + i965_libdrm_winsys_init_buffer_functions(idws); + + idws->fd = drmFD; + idws->id = deviceID; + + idws->base.destroy = i965_libdrm_winsys_destroy; + + idws->gem = drm_intel_bufmgr_gem_init(idws->fd, BRW_BATCH_SIZE); + drm_intel_bufmgr_gem_enable_reuse(idws->gem); + + idws->send_cmd = !debug_get_bool_option("BRW_NO_HW", FALSE); + + return brw_create_screen(&idws->base, deviceID); +} + +static struct pipe_context * +i965_libdrm_create_context(struct drm_api *api, struct pipe_screen *screen) +{ + return brw_create_context(screen); +} + +static void +destroy(struct drm_api *api) +{ + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + +} + +struct drm_api i965_libdrm_api = +{ + .create_context = i965_libdrm_create_context, + .create_screen = i965_libdrm_create_screen, + .texture_from_shared_handle = i965_libdrm_texture_from_shared_handle, + .shared_handle_from_texture = i965_libdrm_shared_handle_from_texture, + .local_handle_from_texture = i965_libdrm_local_handle_from_texture, + .destroy = destroy, +}; + +struct drm_api * +drm_api_create() +{ + return trace_drm_create(&i965_libdrm_api); +} diff --git a/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c b/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c new file mode 100644 index 00000000000..a4a72b372dd --- /dev/null +++ b/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c @@ -0,0 +1,427 @@ + +#include "i965_drm_winsys.h" +#include "util/u_memory.h" + +#include "i915_drm.h" +#include "intel_bufmgr.h" + + + +const char *names[BRW_BUFFER_TYPE_MAX] = { + "TEXTURE", + "SCANOUT", + "VERTEX", + "CURBE", + "QUERY", + "SHADER_CONSTANTS", + "WM_SCRATCH", + "BATCH", + "GENERAL_STATE", + "SURFACE_STATE", + "PIXEL", + "GENERIC", +}; + +const char *usages[BRW_USAGE_MAX] = { + "STATE", + "QUERY_RESULT", + "RENDER_TARGET", + "DEPTH_BUFFER", + "BLIT_SOURCE", + "BLIT_DEST", + "SAMPLER", + "VERTEX", + "SCRATCH" +}; + + +const char *data_types[BRW_DATA_MAX] = +{ + "GS: CC_VP", + "GS: CC_UNIT", + "GS: WM_PROG", + "GS: SAMPLER_DEFAULT_COLOR", + "GS: SAMPLER", + "GS: WM_UNIT", + "GS: SF_PROG", + "GS: SF_VP", + "GS: SF_UNIT", + "GS: VS_UNIT", + "GS: VS_PROG", + "GS: GS_UNIT", + "GS: GS_PROG", + "GS: CLIP_VP", + "GS: CLIP_UNIT", + "GS: CLIP_PROG", + "SS: SURFACE", + "SS: SURF_BIND", + "CONSTANT DATA", + "BATCH DATA", + "(untyped)" +}; + +static enum pipe_error +i965_libdrm_bo_alloc(struct brw_winsys_screen *sws, + enum brw_buffer_type type, + unsigned size, + unsigned alignment, + struct brw_winsys_buffer **bo_out) +{ + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(sws); + struct i965_libdrm_buffer *buf; + + if (BRW_DUMP) + debug_printf("%s type %s sz %d align %d\n", + __FUNCTION__, names[type], size, alignment ); + + buf = CALLOC_STRUCT(i965_libdrm_buffer); + if (!buf) + return PIPE_ERROR_OUT_OF_MEMORY; + + switch (type) { + case BRW_BUFFER_TYPE_TEXTURE: +/* case BRW_BUFFER_TYPE_SCANOUT:*/ + case BRW_BUFFER_TYPE_VERTEX: + case BRW_BUFFER_TYPE_CURBE: + case BRW_BUFFER_TYPE_QUERY: + case BRW_BUFFER_TYPE_SHADER_CONSTANTS: + case BRW_BUFFER_TYPE_SHADER_SCRATCH: + case BRW_BUFFER_TYPE_BATCH: + case BRW_BUFFER_TYPE_GENERAL_STATE: + case BRW_BUFFER_TYPE_SURFACE_STATE: + case BRW_BUFFER_TYPE_PIXEL: + case BRW_BUFFER_TYPE_GENERIC: + break; + case BRW_BUFFER_TYPE_SCANOUT: + buf->map_gtt = TRUE; + break; + default: + assert(0); + break; + } + + buf->bo = drm_intel_bo_alloc(idws->gem, + names[type], + size, + alignment); + + if (!buf->bo) + goto err; + + pipe_reference_init(&buf->base.reference, 1); + buf->base.size = size; + buf->base.sws = sws; + + *bo_out = &buf->base; + return PIPE_OK; + +err: + assert(0); + FREE(buf); + return PIPE_ERROR_OUT_OF_MEMORY; +} + +static void +i965_libdrm_bo_destroy(struct brw_winsys_buffer *buffer) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + drm_intel_bo_unreference(buf->bo); + FREE(buffer); +} + +static enum pipe_error +i965_libdrm_bo_emit_reloc(struct brw_winsys_buffer *buffer, + enum brw_buffer_usage usage, + unsigned delta, + unsigned offset, + struct brw_winsys_buffer *buffer2) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + struct i965_libdrm_buffer *buf2 = i965_libdrm_buffer(buffer2); + int read, write; + int ret; + + if (BRW_DUMP) + debug_printf("%s buf %p offset %x delta %x buf2 %p/%s/%s\n", + __FUNCTION__, (void *)buffer, + offset, delta, + (void *)buffer2, names[buf2->data_type], usages[usage]); + + switch (usage) { + case BRW_USAGE_STATE: + read = I915_GEM_DOMAIN_INSTRUCTION; + write = 0; + break; + case BRW_USAGE_QUERY_RESULT: + read = I915_GEM_DOMAIN_INSTRUCTION; + write = I915_GEM_DOMAIN_INSTRUCTION; + break; + case BRW_USAGE_RENDER_TARGET: + read = I915_GEM_DOMAIN_RENDER; + write = 0; + break; + case BRW_USAGE_DEPTH_BUFFER: + read = I915_GEM_DOMAIN_RENDER; + write = I915_GEM_DOMAIN_RENDER; + break; + case BRW_USAGE_BLIT_SOURCE: + read = 0; + write = I915_GEM_DOMAIN_RENDER; + break; + case BRW_USAGE_BLIT_DEST: + read = I915_GEM_DOMAIN_RENDER; + write = I915_GEM_DOMAIN_RENDER; + break; + case BRW_USAGE_SAMPLER: + read = I915_GEM_DOMAIN_SAMPLER; + write = 0; + break; + case BRW_USAGE_VERTEX: + read = I915_GEM_DOMAIN_VERTEX; + write = 0; + break; + case BRW_USAGE_SCRATCH: + read = 0; + write = 0; + break; + default: + assert(0); + return -1; + } + + /* Needed?? + ((uint32_t *)buf->bo->virtual)[offset/4] = (delta + + buf2->bo->offset); + */ + + ret = dri_bo_emit_reloc( buf->bo, read, write, delta, offset, buf2->bo ); + if (ret) + return -1; + + return 0; +} + +static enum pipe_error +i965_libdrm_bo_exec(struct brw_winsys_buffer *buffer, + unsigned bytes_used) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(buffer->sws); + int ret; + + if (BRW_DUMP) + debug_printf("execute buffer %p, bytes %d\n", (void *)buffer, bytes_used); + + if (idws->send_cmd) { + ret = dri_bo_exec(buf->bo, bytes_used, NULL, 0, 0); + if (ret) + return PIPE_ERROR; + } + + return PIPE_OK; +} + +static enum pipe_error +i965_libdrm_bo_subdata(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + size_t offset, + size_t size, + const void *data, + const struct brw_winsys_reloc *reloc, + unsigned nr_reloc) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(buffer->sws); + int ret, i; + + (void)data_type; + + if (BRW_DUMP) + debug_printf("%s buf %p off %d sz %d %s relocs: %d\n", + __FUNCTION__, + (void *)buffer, offset, size, + data_types[data_type], + nr_reloc); + + if (BRW_DUMP) + brw_dump_data( idws->id, + data_type, + buf->bo->offset + offset, + data, size ); + + /* XXX: use bo_map_gtt/memcpy/unmap_gtt under some circumstances??? + */ + ret = drm_intel_bo_subdata(buf->bo, offset, size, (void*)data); + if (ret) + return PIPE_ERROR; + + for (i = 0; i < nr_reloc; i++) { + i965_libdrm_bo_emit_reloc(buffer, reloc[i].usage, reloc[i].delta, + reloc[i].offset, reloc[i].bo); + } + + return PIPE_OK; +} + +static boolean +i965_libdrm_bo_is_busy(struct brw_winsys_buffer *buffer) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + boolean ret; + + if (BRW_DUMP) + debug_printf("%s %p\n", __FUNCTION__, (void *)buffer); + + ret = drm_intel_bo_busy(buf->bo); + + if (BRW_DUMP) + debug_printf(" --> %d\n", ret); + + return ret; +} + +static boolean +i965_libdrm_bo_references(struct brw_winsys_buffer *a, + struct brw_winsys_buffer *b) +{ + struct i965_libdrm_buffer *bufa = i965_libdrm_buffer(a); + struct i965_libdrm_buffer *bufb = i965_libdrm_buffer(b); + boolean ret; + + if (BRW_DUMP) + debug_printf("%s %p %p\n", __FUNCTION__, (void *)a, (void *)b); + + ret = drm_intel_bo_references(bufa->bo, bufb->bo); + + if (BRW_DUMP) + debug_printf(" --> %d\n", ret); + + return ret; +} + +/* XXX: couldn't this be handled by returning true/false on + * bo_emit_reloc? + */ +static enum pipe_error +i965_libdrm_check_aperture_space(struct brw_winsys_screen *iws, + struct brw_winsys_buffer **buffers, + unsigned count) +{ + static drm_intel_bo *bos[128]; + int i; + int ret; + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + if (count > Elements(bos)) { + assert(0); + return FALSE; + } + + for (i = 0; i < count; i++) + bos[i] = i965_libdrm_buffer(buffers[i])->bo; + + /* XXX: converting from ??? to pipe_error: + */ + ret = dri_bufmgr_check_aperture_space(bos, count); + + if (BRW_DUMP) + debug_printf(" --> %d (ok == %d)\n", ret, PIPE_OK); + + return ret; +} + +static void * +i965_libdrm_bo_map(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + unsigned offset, + unsigned length, + boolean write, + boolean discard, + boolean flush_explicit) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + int ret; + + + if (BRW_DUMP) + debug_printf("%s %p %s %s\n", __FUNCTION__, (void *)buffer, + write ? "read/write" : "read", + write ? data_types[data_type] : ""); + + if (!buf->map_count) { + if (buf->map_gtt) { + ret = drm_intel_gem_bo_map_gtt(buf->bo); + if (ret) + return NULL; + } + else { + ret = drm_intel_bo_map(buf->bo, write); + if (ret) + return NULL; + } + } + + buf->data_type = data_type; + buf->map_count++; + return buf->bo->virtual; +} + +static void +i965_libdrm_bo_flush_range(struct brw_winsys_buffer *buffer, + unsigned offset, + unsigned length) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + struct i965_libdrm_winsys *idws = i965_libdrm_winsys(buffer->sws); + + if (BRW_DUMP) + debug_printf("%s %s offset %d len %d\n", __FUNCTION__, + data_types[buf->data_type], + offset, length); + + if (BRW_DUMP) + brw_dump_data( idws->id, + buf->data_type, + buf->bo->offset + offset, + buf->bo->virtual + offset, + length ); +} + +static void +i965_libdrm_bo_unmap(struct brw_winsys_buffer *buffer) +{ + struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer); + + if (BRW_DUMP) + debug_printf("%s\n", __FUNCTION__); + + if (--buf->map_count > 0) + return; + + if (buf->map_gtt) + drm_intel_gem_bo_unmap_gtt(buf->bo); + else + drm_intel_bo_unmap(buf->bo); +} + +void +i965_libdrm_winsys_init_buffer_functions(struct i965_libdrm_winsys *idws) +{ + idws->base.bo_alloc = i965_libdrm_bo_alloc; + idws->base.bo_destroy = i965_libdrm_bo_destroy; + idws->base.bo_emit_reloc = i965_libdrm_bo_emit_reloc; + idws->base.bo_exec = i965_libdrm_bo_exec; + idws->base.bo_subdata = i965_libdrm_bo_subdata; + idws->base.bo_is_busy = i965_libdrm_bo_is_busy; + idws->base.bo_references = i965_libdrm_bo_references; + idws->base.check_aperture_space = i965_libdrm_check_aperture_space; + idws->base.bo_map = i965_libdrm_bo_map; + idws->base.bo_flush_range = i965_libdrm_bo_flush_range; + idws->base.bo_unmap = i965_libdrm_bo_unmap; +} diff --git a/src/gallium/winsys/drm/i965/gem/i965_drm_winsys.h b/src/gallium/winsys/drm/i965/gem/i965_drm_winsys.h new file mode 100644 index 00000000000..c6a7d4a8c51 --- /dev/null +++ b/src/gallium/winsys/drm/i965/gem/i965_drm_winsys.h @@ -0,0 +1,64 @@ + +#ifndef INTEL_DRM_WINSYS_H +#define INTEL_DRM_WINSYS_H + +#include "i965/brw_winsys.h" + +#include "drm.h" +#include "intel_bufmgr.h" + + + +/* + * Winsys + */ + + +struct i965_libdrm_winsys +{ + struct brw_winsys_screen base; + drm_intel_bufmgr *gem; + + boolean send_cmd; + + int fd; /**< Drm file discriptor */ + + unsigned id; +}; + +static INLINE struct i965_libdrm_winsys * +i965_libdrm_winsys(struct brw_winsys_screen *iws) +{ + return (struct i965_libdrm_winsys *)iws; +} + +struct i965_libdrm_winsys *i965_libdrm_winsys_create(int fd, unsigned pci_id); + +void i965_libdrm_winsys_init_buffer_functions(struct i965_libdrm_winsys *idws); + + +/* Buffer. + */ +struct i965_libdrm_buffer { + struct brw_winsys_buffer base; + + drm_intel_bo *bo; + + void *ptr; + unsigned map_count; + unsigned data_type; /* valid while mapped */ + unsigned tiling; + + boolean map_gtt; + boolean flinked; + unsigned flink; +}; + +static INLINE struct i965_libdrm_buffer * +i965_libdrm_buffer(struct brw_winsys_buffer *buffer) +{ + return (struct i965_libdrm_buffer *)buffer; +} + + +#endif diff --git a/src/gallium/winsys/drm/i965/xlib/Makefile b/src/gallium/winsys/drm/i965/xlib/Makefile new file mode 100644 index 00000000000..0efa0ca6f9a --- /dev/null +++ b/src/gallium/winsys/drm/i965/xlib/Makefile @@ -0,0 +1,97 @@ +# src/gallium/winsys/xlib/Makefile + +# This makefile produces a "stand-alone" libGL.so which is based on +# Xlib (no DRI HW acceleration) + + +TOP = ../../../../../.. +include $(TOP)/configs/current + + +GL_MAJOR = 1 +GL_MINOR = 5 +GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) + + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/drivers/i965 \ + -I$(TOP)/src/gallium/drivers/i965/include \ + -I$(TOP)/src/gallium/state_trackers/glx/xlib \ + -I$(TOP)/src/gallium/auxiliary \ + -I/usr/include/drm + +XLIB_WINSYS_SOURCES = \ + xlib_i965.c \ + + + +XLIB_WINSYS_OBJECTS = $(XLIB_WINSYS_SOURCES:.c=.o) + + + +LIBS = \ + $(TOP)/src/gallium/drivers/i965/libi965.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ + $(TOP)/src/gallium/state_trackers/glx/xlib/libxlib.a \ + $(TOP)/src/mesa/libglapi.a \ + $(TOP)/src/mesa/libmesagallium.a \ + $(GALLIUM_AUXILIARIES) + +# $(TOP)/src/gallium/drivers/i965/lib/libi9xx.a \ + +.SUFFIXES : .cpp + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(DEFINES) $(CFLAGS) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDE_DIRS) $(DEFINES) $(CXXFLAGS) $< -o $@ + + + +default: $(TOP)/$(LIB_DIR)/gallium $(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME) + +$(TOP)/$(LIB_DIR)/gallium: + @ mkdir -p $(TOP)/$(LIB_DIR)/gallium + +# Make the libGL.so library +$(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME): $(XLIB_WINSYS_OBJECTS) $(LIBS) Makefile + $(TOP)/bin/mklib -o $(GL_LIB) \ + -linker "$(CC)" \ + -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ + -install $(TOP)/$(LIB_DIR)/gallium \ + $(MKLIB_OPTIONS) $(XLIB_WINSYS_OBJECTS) \ + -Wl,--start-group $(LIBS) -Wl,--end-group $(GL_LIB_DEPS) + + +depend: $(XLIB_WINSYS_SOURCES) + @ echo "running $(MKDEP)" + @ rm -f depend # workaround oops on gutsy?!? + @ touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(XLIB_WINSYS_SOURCES) \ + > /dev/null 2>/dev/null + + +install: default + $(INSTALL) -d $(INSTALL_DIR)/include/GL + $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL + @if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \ + $(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ + fi + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h + +clean: + -rm -f *.o + + +include depend diff --git a/src/gallium/winsys/drm/i965/xlib/xlib_i965.c b/src/gallium/winsys/drm/i965/xlib/xlib_i965.c new file mode 100644 index 00000000000..d2b9a1ab311 --- /dev/null +++ b/src/gallium/winsys/drm/i965/xlib/xlib_i965.c @@ -0,0 +1,522 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "util/u_memory.h" +#include "util/u_math.h" +#include "pipe/p_error.h" +#include "pipe/p_context.h" + +#include "xm_winsys.h" + +#include "i965/brw_winsys.h" +#include "i965/brw_screen.h" +#include "i965/brw_reg.h" +#include "i965/brw_structs_dump.h" + +#define MAX_VRAM (128*1024*1024) + + + +extern int brw_disasm (FILE *file, + const struct brw_instruction *inst, + unsigned count ); + +extern int intel_decode(const uint32_t *data, + int count, + uint32_t hw_offset, + uint32_t devid); + +struct xlib_brw_buffer +{ + struct brw_winsys_buffer base; + char *virtual; + unsigned offset; + unsigned type; + int map_count; + boolean modified; +}; + + +/** + * Subclass of brw_winsys_screen for Xlib winsys + */ +struct xlib_brw_winsys +{ + struct brw_winsys_screen base; + struct brw_chipset chipset; + + unsigned size; + unsigned used; +}; + +static struct xlib_brw_winsys * +xlib_brw_winsys( struct brw_winsys_screen *screen ) +{ + return (struct xlib_brw_winsys *)screen; +} + + +static struct xlib_brw_buffer * +xlib_brw_buffer( struct brw_winsys_buffer *buffer ) +{ + return (struct xlib_brw_buffer *)buffer; +} + + + +const char *names[BRW_BUFFER_TYPE_MAX] = { + "TEXTURE", + "SCANOUT", + "VERTEX", + "CURBE", + "QUERY", + "SHADER_CONSTANTS", + "WM_SCRATCH", + "BATCH", + "GENERAL_STATE", + "SURFACE_STATE", + "PIXEL", + "GENERIC", +}; + +const char *usages[BRW_USAGE_MAX] = { + "STATE", + "QUERY_RESULT", + "RENDER_TARGET", + "DEPTH_BUFFER", + "BLIT_SOURCE", + "BLIT_DEST", + "SAMPLER", + "VERTEX", + "SCRATCH" +}; + + +const char *data_types[BRW_DATA_MAX] = +{ + "GS: CC_VP", + "GS: CC_UNIT", + "GS: WM_PROG", + "GS: SAMPLER_DEFAULT_COLOR", + "GS: SAMPLER", + "GS: WM_UNIT", + "GS: SF_PROG", + "GS: SF_VP", + "GS: SF_UNIT", + "GS: VS_UNIT", + "GS: VS_PROG", + "GS: GS_UNIT", + "GS: GS_PROG", + "GS: CLIP_VP", + "GS: CLIP_UNIT", + "GS: CLIP_PROG", + "SS: SURFACE", + "SS: SURF_BIND", + "CONSTANT DATA", + "BATCH DATA", + "(untyped)" +}; + + +static enum pipe_error +xlib_brw_bo_alloc( struct brw_winsys_screen *sws, + enum brw_buffer_type type, + unsigned size, + unsigned alignment, + struct brw_winsys_buffer **bo_out ) +{ + struct xlib_brw_winsys *xbw = xlib_brw_winsys(sws); + struct xlib_brw_buffer *buf; + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s type %s sz %d align %d\n", + __FUNCTION__, names[type], size, alignment ); + + buf = CALLOC_STRUCT(xlib_brw_buffer); + if (!buf) + return PIPE_ERROR_OUT_OF_MEMORY; + + pipe_reference_init(&buf->base.reference, 1); + + buf->offset = align(xbw->used, alignment); + buf->type = type; + buf->virtual = MALLOC(size); + buf->base.size = size; + buf->base.sws = sws; + + xbw->used = align(xbw->used, alignment) + size; + if (xbw->used > MAX_VRAM) + goto err; + + /* XXX: possibly rentrant call to bo_destroy: + */ + bo_reference(bo_out, &buf->base); + return PIPE_OK; + +err: + assert(0); + FREE(buf->virtual); + FREE(buf); + return PIPE_ERROR_OUT_OF_MEMORY; +} + +static void +xlib_brw_bo_destroy( struct brw_winsys_buffer *buffer ) +{ + struct xlib_brw_buffer *buf = xlib_brw_buffer(buffer); + + FREE(buf); +} + +static int +xlib_brw_bo_emit_reloc( struct brw_winsys_buffer *buffer, + enum brw_buffer_usage usage, + unsigned delta, + unsigned offset, + struct brw_winsys_buffer *buffer2) +{ + struct xlib_brw_buffer *buf = xlib_brw_buffer(buffer); + struct xlib_brw_buffer *buf2 = xlib_brw_buffer(buffer2); + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s buf %p offset %x val %x + %x buf2 %p/%s/%s\n", + __FUNCTION__, (void *)buffer, offset, + buf2->offset, delta, + (void *)buffer2, names[buf2->type], usages[usage]); + + *(uint32_t *)(buf->virtual + offset) = buf2->offset + delta; + + return 0; +} + +static int +xlib_brw_bo_exec( struct brw_winsys_buffer *buffer, + unsigned bytes_used ) +{ + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("execute buffer %p, bytes %d\n", (void *)buffer, bytes_used); + + return 0; +} + + + + +static int +xlib_brw_bo_subdata(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + size_t offset, + size_t size, + const void *data, + const struct brw_winsys_reloc *reloc, + unsigned nr_relocs) +{ + struct xlib_brw_buffer *buf = xlib_brw_buffer(buffer); + struct xlib_brw_winsys *xbw = xlib_brw_winsys(buffer->sws); + unsigned i; + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s buf %p off %d sz %d %s relocs: %d\n", + __FUNCTION__, + (void *)buffer, offset, size, + data_types[data_type], + nr_relocs); + + assert(buf->base.size >= offset + size); + memcpy(buf->virtual + offset, data, size); + + /* Apply the relocations: + */ + for (i = 0; i < nr_relocs; i++) { + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("\treloc[%d] usage %s off %d value %x+%x\n", + i, usages[reloc[i].usage], reloc[i].offset, + xlib_brw_buffer(reloc[i].bo)->offset, reloc[i].delta); + + *(unsigned *)(buf->virtual + offset + reloc[i].offset) = + xlib_brw_buffer(reloc[i].bo)->offset + reloc[i].delta; + } + + if (BRW_DUMP) + brw_dump_data( xbw->chipset.pci_id, + data_type, + buf->offset + offset, + buf->virtual + offset, size ); + + + return 0; +} + + +static boolean +xlib_brw_bo_is_busy(struct brw_winsys_buffer *buffer) +{ + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s %p\n", __FUNCTION__, (void *)buffer); + return TRUE; +} + +static boolean +xlib_brw_bo_references(struct brw_winsys_buffer *a, + struct brw_winsys_buffer *b) +{ + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s %p %p\n", __FUNCTION__, (void *)a, (void *)b); + return TRUE; +} + +static enum pipe_error +xlib_brw_check_aperture_space( struct brw_winsys_screen *iws, + struct brw_winsys_buffer **buffers, + unsigned count ) +{ + unsigned tot_size = 0; + unsigned i; + + for (i = 0; i < count; i++) + tot_size += buffers[i]->size; + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s %d bufs, tot_size: %d kb\n", + __FUNCTION__, count, + (tot_size + 1023) / 1024); + + return PIPE_OK; +} + +static void * +xlib_brw_bo_map(struct brw_winsys_buffer *buffer, + enum brw_buffer_data_type data_type, + unsigned offset, + unsigned length, + boolean write, + boolean discard, + boolean explicit) +{ + struct xlib_brw_buffer *buf = xlib_brw_buffer(buffer); + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s %p %s %s\n", __FUNCTION__, (void *)buffer, + write ? "read/write" : "read", + write ? data_types[data_type] : ""); + + if (write) + buf->modified = 1; + + buf->map_count++; + return buf->virtual; +} + + +static void +xlib_brw_bo_flush_range( struct brw_winsys_buffer *buffer, + unsigned offset, + unsigned length ) +{ +} + + +static void +xlib_brw_bo_unmap(struct brw_winsys_buffer *buffer) +{ + struct xlib_brw_buffer *buf = xlib_brw_buffer(buffer); + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s %p\n", __FUNCTION__, (void *)buffer); + + --buf->map_count; + assert(buf->map_count >= 0); + + if (buf->map_count == 0 && + buf->modified) { + + buf->modified = 0; + + /* Consider dumping new buffer contents here, using the + * flush-range info to minimize verbosity. + */ + } +} + + +static void +xlib_brw_bo_wait_idle( struct brw_winsys_buffer *buffer ) +{ +} + + +static void +xlib_brw_winsys_destroy( struct brw_winsys_screen *sws ) +{ + struct xlib_brw_winsys *xbw = xlib_brw_winsys(sws); + + FREE(xbw); +} + +static struct brw_winsys_screen * +xlib_create_brw_winsys_screen( void ) +{ + struct xlib_brw_winsys *ws; + + ws = CALLOC_STRUCT(xlib_brw_winsys); + if (!ws) + return NULL; + + ws->used = 0; + + ws->base.destroy = xlib_brw_winsys_destroy; + ws->base.bo_alloc = xlib_brw_bo_alloc; + ws->base.bo_destroy = xlib_brw_bo_destroy; + ws->base.bo_emit_reloc = xlib_brw_bo_emit_reloc; + ws->base.bo_exec = xlib_brw_bo_exec; + ws->base.bo_subdata = xlib_brw_bo_subdata; + ws->base.bo_is_busy = xlib_brw_bo_is_busy; + ws->base.bo_references = xlib_brw_bo_references; + ws->base.check_aperture_space = xlib_brw_check_aperture_space; + ws->base.bo_map = xlib_brw_bo_map; + ws->base.bo_flush_range = xlib_brw_bo_flush_range; + ws->base.bo_unmap = xlib_brw_bo_unmap; + ws->base.bo_wait_idle = xlib_brw_bo_wait_idle; + + return &ws->base; +} + + +/*********************************************************************** + * Implementation of Xlib co-state-tracker's winsys interface + */ + +static void +xlib_i965_display_surface(struct xmesa_buffer *xm_buffer, + struct pipe_surface *surf) +{ + struct brw_surface *surface = brw_surface(surf); + struct xlib_brw_buffer *bo = xlib_brw_buffer(surface->bo); + + if (BRW_DEBUG & DEBUG_WINSYS) + debug_printf("%s offset %x+%x sz %dx%d\n", __FUNCTION__, + bo->offset, + surface->draw_offset, + surf->width, + surf->height); +} + +static void +xlib_i965_flush_frontbuffer(struct pipe_screen *screen, + struct pipe_surface *surf, + void *context_private) +{ + xlib_i965_display_surface(NULL, surf); +} + + +static struct pipe_screen * +xlib_create_i965_screen( void ) +{ + struct brw_winsys_screen *winsys; + struct pipe_screen *screen; + + winsys = xlib_create_brw_winsys_screen(); + if (winsys == NULL) + return NULL; + + screen = brw_create_screen(winsys, PCI_CHIP_GM45_GM); + if (screen == NULL) + goto fail; + + xlib_brw_winsys(winsys)->chipset = brw_screen(screen)->chipset; + + screen->flush_frontbuffer = xlib_i965_flush_frontbuffer; + return screen; + +fail: + if (winsys) + winsys->destroy( winsys ); + + return NULL; +} + + +static struct pipe_context * +xlib_create_i965_context( struct pipe_screen *screen, + void *context_private ) +{ + struct pipe_context *pipe; + + pipe = brw_create_context(screen); + if (pipe == NULL) + goto fail; + + pipe->priv = context_private; + return pipe; + +fail: + /* Free stuff here */ + return NULL; +} + + + + +struct xm_driver xlib_i965_driver = +{ + .create_pipe_screen = xlib_create_i965_screen, + .create_pipe_context = xlib_create_i965_context, + .display_surface = xlib_i965_display_surface +}; + + +/* Register this driver at library load: + */ +static void _init( void ) __attribute__((constructor)); +static void _init( void ) +{ + xmesa_set_driver( &xlib_i965_driver ); +} + + + +/*********************************************************************** + * + * Butt-ugly hack to convince the linker not to throw away public GL + * symbols (they are all referenced from getprocaddress, I guess). + */ +extern void (*linker_foo(const unsigned char *procName))(); +extern void (*glXGetProcAddress(const unsigned char *procName))(); + +extern void (*linker_foo(const unsigned char *procName))() +{ + return glXGetProcAddress(procName); +} diff --git a/src/gallium/winsys/drm/i965/xorg/Makefile b/src/gallium/winsys/drm/i965/xorg/Makefile new file mode 100644 index 00000000000..d91d0006efd --- /dev/null +++ b/src/gallium/winsys/drm/i965/xorg/Makefile @@ -0,0 +1,57 @@ +TARGET = modesetting_drv.so +CFILES = $(wildcard ./*.c) +OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) +TOP = ../../../../../.. + +include $(TOP)/configs/current + +INCLUDES = \ + $(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \ + -I../gem \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main + +LIBS = \ + $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a \ + $(TOP)/src/gallium/winsys/drm/i965/gem/libi965drm.a \ + $(TOP)/src/gallium/drivers/i965/libi965.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(GALLIUM_AUXILIARIES) + +DRIVER_DEFINES = \ + -DHAVE_CONFIG_H + + +############################################# + + + +all default: $(TARGET) + +$(TARGET): $(OBJECTS) Makefile $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a $(LIBS) + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel + +clean: + rm -rf $(OBJECTS) $(TARGET) + +install: + $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + $(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + + +############################################## + + +.c.o: + $(CC) -c $(CFLAGS) $(INCLUDES) $(DRIVER_DEFINES) $< -o $@ + + +############################################## + +.PHONY = all clean install diff --git a/src/gallium/winsys/drm/i965/xorg/intel_xorg.c b/src/gallium/winsys/drm/i965/xorg/intel_xorg.c new file mode 100644 index 00000000000..ac691cb76b3 --- /dev/null +++ b/src/gallium/winsys/drm/i965/xorg/intel_xorg.c @@ -0,0 +1,147 @@ +/* + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * + * Author: Alan Hourihane <[email protected]> + * Author: Jakob Bornecrantz <[email protected]> + * + */ + +#include "../../../../state_trackers/xorg/xorg_winsys.h" + +static void intel_xorg_identify(int flags); +static Bool intel_xorg_pci_probe(DriverPtr driver, + int entity_num, + struct pci_device *device, + intptr_t match_data); + +static const struct pci_id_match intel_xorg_device_match[] = { + {0x8086, PCI_MATCH_ANY, PCI_MATCH_ANY, PCI_MATCH_ANY, 0, 0, 0}, + {0, 0, 0}, +}; + +static SymTabRec intel_xorg_chipsets[] = { + {PCI_MATCH_ANY, "Intel Graphics Device"}, + {-1, NULL} +}; + +static PciChipsets intel_xorg_pci_devices[] = { + {PCI_MATCH_ANY, PCI_MATCH_ANY, NULL}, + {-1, -1, NULL} +}; + +static XF86ModuleVersionInfo intel_xorg_version = { + "modesetting", + MODULEVENDORSTRING, + MODINFOSTRING1, + MODINFOSTRING2, + XORG_VERSION_CURRENT, + 0, 1, 0, /* major, minor, patch */ + ABI_CLASS_VIDEODRV, + ABI_VIDEODRV_VERSION, + MOD_CLASS_VIDEODRV, + {0, 0, 0, 0} +}; + +/* + * Xorg driver exported structures + */ + +_X_EXPORT DriverRec modesetting = { + 1, + "modesetting", + intel_xorg_identify, + NULL, + xorg_tracker_available_options, + NULL, + 0, + NULL, + intel_xorg_device_match, + intel_xorg_pci_probe +}; + +static MODULESETUPPROTO(intel_xorg_setup); + +_X_EXPORT XF86ModuleData modesettingModuleData = { + &intel_xorg_version, + intel_xorg_setup, + NULL +}; + +/* + * Xorg driver functions + */ + +static pointer +intel_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin) +{ + static Bool setupDone = 0; + + /* This module should be loaded only once, but check to be sure. + */ + if (!setupDone) { + setupDone = 1; + xf86AddDriver(&modesetting, module, HaveDriverFuncs); + + /* + * The return value must be non-NULL on success even though there + * is no TearDownProc. + */ + return (pointer) 1; + } else { + if (errmaj) + *errmaj = LDR_ONCEONLY; + return NULL; + } +} + +static void +intel_xorg_identify(int flags) +{ + xf86PrintChipsets("modesetting", "Driver for Modesetting Kernel Drivers", + intel_xorg_chipsets); +} + +static Bool +intel_xorg_pci_probe(DriverPtr driver, + int entity_num, struct pci_device *device, intptr_t match_data) +{ + ScrnInfoPtr scrn = NULL; + EntityInfoPtr entity; + + scrn = xf86ConfigPciEntity(scrn, 0, entity_num, intel_xorg_pci_devices, + NULL, NULL, NULL, NULL, NULL); + if (scrn != NULL) { + scrn->driverVersion = 1; + scrn->driverName = "i965"; + scrn->name = "modesetting"; + scrn->Probe = NULL; + + entity = xf86GetEntityInfo(entity_num); + + /* Use all the functions from the xorg tracker */ + xorg_tracker_set_functions(scrn); + } + return scrn != NULL; +} diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile index c0ecd9680e2..26aae4122eb 100644 --- a/src/gallium/winsys/drm/intel/dri/Makefile +++ b/src/gallium/winsys/drm/intel/dri/Makefile @@ -24,4 +24,3 @@ DRI_LIB_DEPS += -ldrm_intel symlinks: $(TOP)/$(LIB_DIR)/gallium @rm -f $(TOP)/$(LIB_DIR)/gallium/i965_dri.so - ln -s i915_dri.so $(TOP)/$(LIB_DIR)/gallium/i965_dri.so diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c index 9ed570ff6e4..5ed2a10af1c 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c @@ -32,6 +32,7 @@ intel_drm_get_device_id(unsigned int *device_id) } shutup_gcc = fgets(path, sizeof(path), file); + (void) shutup_gcc; sscanf(path, "%x", device_id); fclose(file); } diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c index b6248a3bcf7..e8b58742ab7 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c @@ -39,7 +39,7 @@ intel_drm_fence_reference(struct intel_winsys *iws, struct intel_drm_fence *old = (struct intel_drm_fence *)*ptr; struct intel_drm_fence *f = (struct intel_drm_fence *)fence; - if (pipe_reference(&(*ptr)->reference, &f->reference)) { + if (pipe_reference(&((struct intel_drm_fence *)(*ptr))->reference, &f->reference)) { if (old->bo) drm_intel_bo_unreference(old->bo); FREE(old); diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index 8d95826c9a2..7106a06492d 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -1,5 +1,6 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "nouveau_drm_api.h" @@ -246,7 +247,7 @@ nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen, return false; *handle = mt->bo->handle; - *stride = pf_get_stride(mt->base.format, mt->base.width0); + *stride = util_format_get_stride(mt->base.format, mt->base.width0); return true; } diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 2a8daed051d..d2367b245a2 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -35,6 +35,7 @@ #include "radeon_bo_gem.h" #include "softpipe/sp_texture.h" #include "r300_context.h" +#include "util/u_format.h" #include "util/u_math.h" #include <X11/Xutil.h> @@ -120,8 +121,8 @@ static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws, const unsigned alignment = 64; unsigned nblocksy, size; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); size = *stride * nblocksy; return radeon_buffer_create(ws, 64, usage, size); @@ -146,7 +147,7 @@ static void *radeon_buffer_map(struct pipe_winsys *ws, int write = 0; if (radeon_bo_is_referenced_by_cs(radeon_buffer->bo, priv->cs)) { - priv->cs->space_flush_fn(priv->cs->space_flush_data); + priv->flush_cb(priv->flush_data); } if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h index bfe2221d1ed..d7f17564a9f 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h @@ -68,6 +68,10 @@ struct radeon_winsys_priv { /* Current CS. */ struct radeon_cs* cs; + + /* Flush CB */ + void (*flush_cb)(void *); + void *flush_data; }; struct radeon_winsys* radeon_pipe_winsys(int fb); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index ba0596c30dc..0875ee41cbf 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -26,6 +26,8 @@ static void radeon_set_flush_cb(struct radeon_winsys *winsys, void (*flush_cb)(void *), void *data) { + winsys->priv->flush_cb = flush_cb; + winsys->priv->flush_data = data; radeon_cs_space_set_flush(winsys->priv->cs, flush_cb, data); } diff --git a/src/gallium/winsys/drm/vmware/core/vmw_screen_ioctl.c b/src/gallium/winsys/drm/vmware/core/vmw_screen_ioctl.c index 51e455f9254..ccd0b418a16 100644 --- a/src/gallium/winsys/drm/vmware/core/vmw_screen_ioctl.c +++ b/src/gallium/winsys/drm/vmware/core/vmw_screen_ioctl.c @@ -468,6 +468,15 @@ vmw_ioctl_init(struct vmw_winsys_screen *vws) VMW_FUNC; memset(&gp_arg, 0, sizeof(gp_arg)); + gp_arg.param = DRM_VMW_PARAM_3D; + ret = drmCommandWriteRead(vws->ioctl.drm_fd, DRM_VMW_GET_PARAM, + &gp_arg, sizeof(gp_arg)); + if (ret || gp_arg.value == 0) { + debug_printf("No 3D enabled (%i, %s)\n", ret, strerror(-ret)); + goto out_err1; + } + + memset(&gp_arg, 0, sizeof(gp_arg)); gp_arg.param = DRM_VMW_PARAM_FIFO_OFFSET; ret = drmCommandWriteRead(vws->ioctl.drm_fd, DRM_VMW_GET_PARAM, &gp_arg, sizeof(gp_arg)); diff --git a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h index 56070a1ba10..2be7e1249b6 100644 --- a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h +++ b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h @@ -25,41 +25,50 @@ * **************************************************************************/ -#ifndef _VMWGFX_DRM_H_ -#define _VMWGFX_DRM_H_ +#ifndef __VMWGFX_DRM_H__ +#define __VMWGFX_DRM_H__ #define DRM_VMW_MAX_SURFACE_FACES 6 #define DRM_VMW_MAX_MIP_LEVELS 24 #define DRM_VMW_EXT_NAME_LEN 128 -#define DRM_VMW_GET_PARAM 1 -#define DRM_VMW_EXTENSION 2 -#define DRM_VMW_CREATE_CONTEXT 3 -#define DRM_VMW_UNREF_CONTEXT 4 -#define DRM_VMW_CREATE_SURFACE 5 -#define DRM_VMW_UNREF_SURFACE 6 -#define DRM_VMW_REF_SURFACE 7 -#define DRM_VMW_EXECBUF 8 -#define DRM_VMW_ALLOC_DMABUF 9 -#define DRM_VMW_UNREF_DMABUF 10 -#define DRM_VMW_FIFO_DEBUG 11 -#define DRM_VMW_FENCE_WAIT 12 -#define DRM_VMW_OVERLAY 13 -#define DRM_VMW_CURSOR_BYPASS 14 +#define DRM_VMW_GET_PARAM 0 +#define DRM_VMW_ALLOC_DMABUF 1 +#define DRM_VMW_UNREF_DMABUF 2 +#define DRM_VMW_CURSOR_BYPASS 3 +/* guarded by DRM_VMW_PARAM_NUM_STREAMS != 0*/ +#define DRM_VMW_CONTROL_STREAM 4 +#define DRM_VMW_CLAIM_STREAM 5 +#define DRM_VMW_UNREF_STREAM 6 +/* guarded by DRM_VMW_PARAM_3D == 1 */ +#define DRM_VMW_CREATE_CONTEXT 7 +#define DRM_VMW_UNREF_CONTEXT 8 +#define DRM_VMW_CREATE_SURFACE 9 +#define DRM_VMW_UNREF_SURFACE 10 +#define DRM_VMW_REF_SURFACE 11 +#define DRM_VMW_EXECBUF 12 +#define DRM_VMW_FIFO_DEBUG 13 +#define DRM_VMW_FENCE_WAIT 14 + /*************************************************************************/ /** * DRM_VMW_GET_PARAM - get device information. * - * Currently we support only one parameter: - * * DRM_VMW_PARAM_FIFO_OFFSET: * Offset to use to map the first page of the FIFO read-only. * The fifo is mapped using the mmap() system call on the drm device. + * + * DRM_VMW_PARAM_OVERLAY_IOCTL: + * Does the driver support the overlay ioctl. */ -#define DRM_VMW_PARAM_FIFO_OFFSET 0 +#define DRM_VMW_PARAM_NUM_STREAMS 0 +#define DRM_VMW_PARAM_NUM_FREE_STREAMS 1 +#define DRM_VMW_PARAM_3D 2 +#define DRM_VMW_PARAM_FIFO_OFFSET 3 + /** * struct drm_vmw_getparam_arg @@ -442,7 +451,7 @@ struct drm_vmw_fence_wait_arg { /*************************************************************************/ /** - * DRM_VMW_OVERLAY - Control overlays. + * DRM_VMW_CONTROL_STREAM - Control overlays, aka streams. * * This IOCTL controls the overlay units of the svga device. * The SVGA overlay units does not work like regular hardware units in @@ -467,7 +476,7 @@ struct drm_vmw_rect { }; /** - * struct drm_vmw_overlay_arg + * struct drm_vmw_control_stream_arg * * @stream_id: Stearm to control * @enabled: If false all following arguments are ignored. @@ -481,10 +490,10 @@ struct drm_vmw_rect { * @src: Source rect, must be within the defined area above. * @dst: Destination rect, x and y may be negative. * - * Argument to the DRM_VMW_OVERLAY Ioctl. + * Argument to the DRM_VMW_CONTROL_STREAM Ioctl. */ -struct drm_vmw_overlay_arg { +struct drm_vmw_control_stream_arg { uint32_t stream_id; uint32_t enabled; @@ -535,4 +544,31 @@ struct drm_vmw_cursor_bypass_arg { int32_t yhot; }; +/*************************************************************************/ +/** + * DRM_VMW_CLAIM_STREAM - Claim a single stream. + */ + +/** + * struct drm_vmw_context_arg + * + * @stream_id: Device unique context ID. + * + * Output argument to the DRM_VMW_CREATE_CONTEXT Ioctl. + * Input argument to the DRM_VMW_UNREF_CONTEXT Ioctl. + */ + +struct drm_vmw_stream_arg { + uint32_t stream_id; + uint32_t pad64; +}; + +/*************************************************************************/ +/** + * DRM_VMW_UNREF_STREAM - Unclaim a stream. + * + * Return a single stream that was claimed by this process. Also makes + * sure that the stream has been stopped. + */ + #endif diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h b/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h index db6b89b8bcd..3efe851a4be 100644 --- a/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h +++ b/src/gallium/winsys/drm/vmware/xorg/vmw_driver.h @@ -66,6 +66,8 @@ Bool vmw_video_init(ScrnInfoPtr pScrn, struct vmw_driver *vmw); Bool vmw_video_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw); +void vmw_video_stop_all(ScrnInfoPtr pScrn, struct vmw_driver *vmw); + /*********************************************************************** * vmw_ioctl.c @@ -86,5 +88,14 @@ void vmw_ioctl_buffer_unmap(struct vmw_driver *vmw, void vmw_ioctl_buffer_destroy(struct vmw_driver *vmw, struct vmw_dma_buffer *buf); +int vmw_ioctl_supports_streams(struct vmw_driver *vmw); + +int vmw_ioctl_num_streams(struct vmw_driver *vmw, + uint32_t *ntot, uint32_t *nfree); + +int vmw_ioctl_unref_stream(struct vmw_driver *vmw, uint32_t stream_id); + +int vmw_ioctl_claim_stream(struct vmw_driver *vmw, uint32_t *out); + #endif diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c b/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c index c84368bab7f..ab2b5fadc49 100644 --- a/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c +++ b/src/gallium/winsys/drm/vmware/xorg/vmw_ioctl.c @@ -56,6 +56,89 @@ struct vmw_dma_buffer uint32_t size; }; +static int +vmw_ioctl_get_param(struct vmw_driver *vmw, uint32_t param, uint64_t *out) +{ + struct drm_vmw_getparam_arg gp_arg; + int ret; + + memset(&gp_arg, 0, sizeof(gp_arg)); + gp_arg.param = param; + ret = drmCommandWriteRead(vmw->fd, DRM_VMW_GET_PARAM, + &gp_arg, sizeof(gp_arg)); + + if (ret == 0) { + *out = gp_arg.value; + } + + return ret; +} + +int +vmw_ioctl_supports_streams(struct vmw_driver *vmw) +{ + uint64_t value; + int ret; + + ret = vmw_ioctl_get_param(vmw, DRM_VMW_PARAM_NUM_STREAMS, &value); + if (ret) + return ret; + + return value ? 0 : -ENOSYS; +} + +int +vmw_ioctl_num_streams(struct vmw_driver *vmw, + uint32_t *ntot, uint32_t *nfree) +{ + uint64_t v1, v2; + int ret; + + ret = vmw_ioctl_get_param(vmw, DRM_VMW_PARAM_NUM_STREAMS, &v1); + if (ret) + return ret; + + ret = vmw_ioctl_get_param(vmw, DRM_VMW_PARAM_NUM_FREE_STREAMS, &v2); + if (ret) + return ret; + + *ntot = (uint32_t)v1; + *nfree = (uint32_t)v2; + + return 0; +} + +int +vmw_ioctl_claim_stream(struct vmw_driver *vmw, uint32_t *out) +{ + struct drm_vmw_stream_arg s_arg; + int ret; + + ret = drmCommandRead(vmw->fd, DRM_VMW_CLAIM_STREAM, + &s_arg, sizeof(s_arg)); + + if (ret) + return -1; + + *out = s_arg.stream_id; + return 0; +} + +int +vmw_ioctl_unref_stream(struct vmw_driver *vmw, uint32_t stream_id) +{ + struct drm_vmw_stream_arg s_arg; + int ret; + + memset(&s_arg, 0, sizeof(s_arg)); + s_arg.stream_id = stream_id; + + ret = drmCommandRead(vmw->fd, DRM_VMW_CLAIM_STREAM, + &s_arg, sizeof(s_arg)); + + return 0; +} + int vmw_ioctl_cursor_bypass(struct vmw_driver *vmw, int xhot, int yhot) { diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c b/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c index 18cb509189a..7c9757cce95 100644 --- a/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c +++ b/src/gallium/winsys/drm/vmware/xorg/vmw_screen.c @@ -124,6 +124,26 @@ vmw_screen_close(ScrnInfoPtr pScrn) return TRUE; } +static Bool +vmw_screen_enter_vt(ScrnInfoPtr pScrn) +{ + debug_printf("%s: enter\n", __func__); + + return TRUE; +} + +static Bool +vmw_screen_leave_vt(ScrnInfoPtr pScrn) +{ + struct vmw_driver *vmw = vmw_driver(pScrn); + + debug_printf("%s: enter\n", __func__); + + vmw_video_stop_all(pScrn, vmw); + + return TRUE; +} + /* * Functions for setting up hooks into the xorg state tracker */ @@ -142,6 +162,8 @@ vmw_screen_pre_init(ScrnInfoPtr pScrn, int flags) ms = modesettingPTR(pScrn); ms->winsys_screen_init = vmw_screen_init; ms->winsys_screen_close = vmw_screen_close; + ms->winsys_enter_vt = vmw_screen_enter_vt; + ms->winsys_leave_vt = vmw_screen_leave_vt; return TRUE; } diff --git a/src/gallium/winsys/drm/vmware/xorg/vmw_video.c b/src/gallium/winsys/drm/vmware/xorg/vmw_video.c index d62c3b7296f..b065b96346a 100644 --- a/src/gallium/winsys/drm/vmware/xorg/vmw_video.c +++ b/src/gallium/winsys/drm/vmware/xorg/vmw_video.c @@ -273,9 +273,20 @@ vmw_video_init(ScrnInfoPtr pScrn, struct vmw_driver *vmw) XF86VideoAdaptorPtr *overlayAdaptors, *newAdaptors = NULL; XF86VideoAdaptorPtr newAdaptor = NULL; int numAdaptors; + unsigned int ntot, nfree; debug_printf("%s: enter\n", __func__); + if (vmw_ioctl_num_streams(vmw, &ntot, &nfree) != 0) { + debug_printf("No stream ioctl support\n"); + return FALSE; + } + + if (nfree == 0) { + debug_printf("No free streams\n"); + return FALSE; + } + numAdaptors = xf86XVListGenericAdaptors(pScrn, &overlayAdaptors); newAdaptor = vmw_video_init_adaptor(pScrn, vmw); @@ -346,7 +357,9 @@ vmw_video_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw) return TRUE; for (i = 0; i < VMWARE_VID_NUM_PORTS; ++i) { - vmw_video_port_cleanup(pScrn, &video->port[i]); + /* make sure the port is stoped as well */ + vmw_xv_stop_video(pScrn, &video->port[i], TRUE); + vmw_ioctl_unref_stream(vmw, video->port[i].streamId); } /* XXX: I'm sure this function is missing code for turning off Xv */ @@ -361,6 +374,38 @@ vmw_video_close(ScrnInfoPtr pScrn, struct vmw_driver *vmw) /* *----------------------------------------------------------------------------- * + * vmw_video_stop_all -- + * + * Stop all video streams from playing. + * + * Results: + * None. + * + * Side effects: + * All buffers are freed. + * + *----------------------------------------------------------------------------- + */ + +void vmw_video_stop_all(ScrnInfoPtr pScrn, struct vmw_driver *vmw) +{ + struct vmw_video_private *video = vmw->video_priv; + int i; + + debug_printf("%s: enter\n", __func__); + + if (!video) + return; + + for (i = 0; i < VMWARE_VID_NUM_PORTS; ++i) { + vmw_xv_stop_video(pScrn, &video->port[i], TRUE); + } +} + + +/* + *----------------------------------------------------------------------------- + * * vmw_video_init_adaptor -- * * Initializes a XF86VideoAdaptor structure with the capabilities and @@ -410,7 +455,7 @@ vmw_video_init_adaptor(ScrnInfoPtr pScrn, struct vmw_driver *vmw) adaptor->pPortPrivates = video->port_ptr; for (i = 0; i < VMWARE_VID_NUM_PORTS; ++i) { - video->port[i].streamId = i; + vmw_ioctl_claim_stream(vmw, &video->port[i].streamId); video->port[i].play = vmw_video_port_init; video->port[i].flags = SVGA_VIDEO_FLAG_COLORKEY; video->port[i].colorKey = VMWARE_VIDEO_COLORKEY; @@ -539,7 +584,7 @@ vmw_video_port_play(ScrnInfoPtr pScrn, struct vmw_video_port *port, short height, RegionPtr clipBoxes) { struct vmw_driver *vmw = vmw_driver(pScrn); - struct drm_vmw_overlay_arg arg; + struct drm_vmw_control_stream_arg arg; unsigned short w, h; int size; int ret; @@ -598,7 +643,7 @@ vmw_video_port_play(ScrnInfoPtr pScrn, struct vmw_video_port *port, } } - ret = drmCommandWrite(vmw->fd, DRM_VMW_OVERLAY, &arg, sizeof(arg)); + ret = drmCommandWrite(vmw->fd, DRM_VMW_CONTROL_STREAM, &arg, sizeof(arg)); if (ret) { vmw_video_port_cleanup(pScrn, port); return XvBadAlloc; @@ -808,7 +853,7 @@ vmw_xv_stop_video(ScrnInfoPtr pScrn, pointer data, Bool cleanup) { struct vmw_driver *vmw = vmw_driver(pScrn); struct vmw_video_port *port = data; - struct drm_vmw_overlay_arg arg; + struct drm_vmw_control_stream_arg arg; int ret; debug_printf("%s: cleanup is %s\n", __func__, cleanup ? "TRUE" : "FALSE"); @@ -824,7 +869,7 @@ vmw_xv_stop_video(ScrnInfoPtr pScrn, pointer data, Bool cleanup) arg.stream_id = port->streamId; arg.enabled = FALSE; - ret = drmCommandWrite(vmw->fd, DRM_VMW_OVERLAY, &arg, sizeof(arg)); + ret = drmCommandWrite(vmw->fd, DRM_VMW_CONTROL_STREAM, &arg, sizeof(arg)); assert(ret == 0); vmw_video_port_cleanup(pScrn, port); diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index d5644c161fa..6ee3ede38cb 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -38,6 +38,7 @@ #include "pipe/internal/p_winsys_screen.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -162,8 +163,8 @@ surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c index b8c8502d7bb..f15bcd37b50 100644 --- a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c +++ b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c @@ -30,6 +30,7 @@ #include <pipe/internal/p_winsys_screen.h> #include <pipe/p_state.h> #include <pipe/p_inlines.h> +#include <util/u_format.h> #include <util/u_memory.h> #include <util/u_math.h> #include <softpipe/sp_winsys.h> @@ -140,8 +141,8 @@ static struct pipe_buffer* xsp_surface_buffer_create const unsigned int ALIGNMENT = 1; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), ALIGNMENT); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), ALIGNMENT); return pws->buffer_create(pws, ALIGNMENT, usage, *stride * nblocksy); diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript index 8f556daf04a..74f6b2fd475 100644 --- a/src/gallium/winsys/gdi/SConscript +++ b/src/gallium/winsys/gdi/SConscript @@ -45,5 +45,5 @@ if env['platform'] == 'windows': env.SharedLibrary( target ='opengl32', source = sources, - LIBS = wgl + glapi + mesa + drivers + auxiliaries + env['LIBS'], + LIBS = wgl + glapi + mesa + drivers + auxiliaries + glsl + env['LIBS'], ) diff --git a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c index 81c46c0a96f..7d076be3a31 100644 --- a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c @@ -39,6 +39,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "llvmpipe/lp_winsys.h" @@ -136,8 +137,8 @@ gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys, gdt->width = width; gdt->height = height; - bpp = pf_get_blocksizebits(format); - cpp = pf_get_blocksize(format); + bpp = util_format_get_blocksizebits(format); + cpp = util_format_get_blocksize(format); gdt->stride = align(width * cpp, alignment); gdt->size = gdt->stride * height; diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c index 173fa5b6fe0..2ad794c3f0f 100644 --- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c @@ -42,6 +42,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "softpipe/sp_winsys.h" @@ -162,8 +163,8 @@ gdi_softpipe_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, @@ -270,10 +271,10 @@ gdi_softpipe_present(struct pipe_screen *screen, memset(&bmi, 0, sizeof(BITMAPINFO)); bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER); - bmi.bmiHeader.biWidth = texture->stride[surface->level] / pf_get_blocksize(surface->format); + bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_blocksize(surface->format); bmi.bmiHeader.biHeight= -(long)surface->height; bmi.bmiHeader.biPlanes = 1; - bmi.bmiHeader.biBitCount = pf_get_blocksizebits(surface->format); + bmi.bmiHeader.biBitCount = util_format_get_blocksizebits(surface->format); bmi.bmiHeader.biCompression = BI_RGB; bmi.bmiHeader.biSizeImage = 0; bmi.bmiHeader.biXPelsPerMeter = 0; diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index 3dc38a78e45..a0293fe9b4b 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -84,11 +84,11 @@ depend: $(XLIB_WINSYS_SOURCES) install: default - $(INSTALL) -d $(INSTALL_DIR)/include/GL - $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) - $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/include/GL + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(DESTDIR)$(INSTALL_DIR)/include/GL @if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \ - $(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ + $(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR); \ fi diff --git a/src/gallium/winsys/xlib/xlib_cell.c b/src/gallium/winsys/xlib/xlib_cell.c index b078de23b76..47ae0519a4b 100644 --- a/src/gallium/winsys/xlib/xlib_cell.c +++ b/src/gallium/winsys/xlib/xlib_cell.c @@ -45,6 +45,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -288,8 +289,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/winsys/xlib/xlib_llvmpipe.c b/src/gallium/winsys/xlib/xlib_llvmpipe.c index 41f3e248e84..2a434b5fd21 100644 --- a/src/gallium/winsys/xlib/xlib_llvmpipe.c +++ b/src/gallium/winsys/xlib/xlib_llvmpipe.c @@ -44,6 +44,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "llvmpipe/lp_winsys.h" @@ -261,10 +262,10 @@ xm_llvmpipe_display(struct xmesa_buffer *xm_buffer, { if (xm_dt->tempImage == NULL) { - assert(pf_get_blockwidth(xm_dt->format) == 1); - assert(pf_get_blockheight(xm_dt->format) == 1); + assert(util_format_get_blockwidth(xm_dt->format) == 1); + assert(util_format_get_blockheight(xm_dt->format) == 1); alloc_shm_ximage(xm_dt, xm_buffer, - xm_dt->stride / pf_get_blocksize(xm_dt->format), + xm_dt->stride / util_format_get_blocksize(xm_dt->format), xm_dt->height); } @@ -330,8 +331,8 @@ xm_displaytarget_create(struct llvmpipe_winsys *winsys, xm_dt->width = width; xm_dt->height = height; - nblocksy = pf_get_nblocksy(format, height); - xm_dt->stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + xm_dt->stride = align(util_format_get_stride(format, width), alignment); size = xm_dt->stride * nblocksy; #ifdef USE_XSHM diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c index 69a5dcc2b78..f7c0099584e 100644 --- a/src/gallium/winsys/xlib/xlib_softpipe.c +++ b/src/gallium/winsys/xlib/xlib_softpipe.c @@ -42,6 +42,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "softpipe/sp_winsys.h" @@ -254,10 +255,10 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b, { if (xm_buf->tempImage == NULL) { - assert(pf_get_blockwidth(surf->texture->format) == 1); - assert(pf_get_blockheight(surf->texture->format) == 1); + assert(util_format_get_blockwidth(surf->texture->format) == 1); + assert(util_format_get_blockheight(surf->texture->format) == 1); alloc_shm_ximage(xm_buf, b, spt->stride[surf->level] / - pf_get_blocksize(surf->texture->format), surf->height); + util_format_get_blocksize(surf->texture->format), surf->height); } ximage = xm_buf->tempImage; @@ -362,8 +363,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy, size; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); size = *stride * nblocksy; #ifdef USE_XSHM diff --git a/src/glsl/Makefile b/src/glsl/Makefile new file mode 100644 index 00000000000..ca7f2d2ac7d --- /dev/null +++ b/src/glsl/Makefile @@ -0,0 +1,15 @@ +# src/glsl/Makefile + +TOP = ../.. + +include $(TOP)/configs/current + +SUBDIRS = pp cl apps + +default install clean: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) $@) || exit 1; \ + fi \ + done + diff --git a/src/glsl/Makefile.template b/src/glsl/Makefile.template new file mode 100644 index 00000000000..974987a0a04 --- /dev/null +++ b/src/glsl/Makefile.template @@ -0,0 +1,50 @@ +# src/glsl/Makefile.template + +# Template makefile for glsl libraries. +# +# Usage: +# The minimum that the including makefile needs to define +# is TOP, LIBNAME and one of of the *_SOURCES. +# +# Optional defines: +# LIBRARY_INCLUDES are appended to the list of includes directories. +# LIBRARY_DEFINES is not used for makedepend, but for compilation. + + +### Basic defines ### + +OBJECTS = $(C_SOURCES:.c=.o) + +INCLUDES = \ + -I. \ + $(LIBRARY_INCLUDES) + + +##### TARGETS ##### + +default: depend lib$(LIBNAME).a + +lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/src/glsl/Makefile.template + $(MKLIB) -o $(LIBNAME) -static $(OBJECTS) + +depend: $(C_SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(C_SOURCES) 2> /dev/null + +# Remove .o and backup files +clean: + rm -f $(OBJECTS) lib$(LIBNAME).a depend depend.bak + +# Dummy target +install: + @echo -n "" + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(LIBRARY_DEFINES) $< -o $@ + +-include depend + diff --git a/src/glsl/SConscript b/src/glsl/SConscript new file mode 100644 index 00000000000..6f1f81b199a --- /dev/null +++ b/src/glsl/SConscript @@ -0,0 +1,68 @@ +import common + +Import('*') + +env = env.Clone() + +sources = [ + 'pp/sl_pp_context.c', + 'pp/sl_pp_define.c', + 'pp/sl_pp_dict.c', + 'pp/sl_pp_error.c', + 'pp/sl_pp_expression.c', + 'pp/sl_pp_extension.c', + 'pp/sl_pp_if.c', + 'pp/sl_pp_line.c', + 'pp/sl_pp_macro.c', + 'pp/sl_pp_pragma.c', + 'pp/sl_pp_process.c', + 'pp/sl_pp_purify.c', + 'pp/sl_pp_token.c', + 'pp/sl_pp_version.c', + 'cl/sl_cl_parse.c', +] + +glsl = env.StaticLibrary( + target = 'glsl', + source = sources, +) + +Export('glsl') + +env = env.Clone() + +if env['platform'] == 'windows': + env.PrependUnique(LIBS = [ + 'user32', + ]) + +env.Prepend(LIBS = [glsl]) + +env.Program( + target = 'purify', + source = ['apps/purify.c'], +) + +env.Program( + target = 'tokenise', + source = ['apps/tokenise.c'], +) + +env.Program( + target = 'version', + source = ['apps/version.c'], +) + +env.Program( + target = 'process', + source = ['apps/process.c'], +) + +glsl_compile = env.Program( + target = 'compile', + source = ['apps/compile.c'], +) + +if env['platform'] == common.default_platform: + # Only export the GLSL compiler when building for the host platform + Export('glsl_compile') diff --git a/src/glsl/apps/Makefile b/src/glsl/apps/Makefile new file mode 100644 index 00000000000..39a0df7fea5 --- /dev/null +++ b/src/glsl/apps/Makefile @@ -0,0 +1,43 @@ +# src/glsl/apps/Makefile + +TOP = ../../.. + +include $(TOP)/configs/current + +LIBS = \ + $(TOP)/src/glsl/pp/libglslpp.a \ + $(TOP)/src/glsl/cl/libglslcl.a + +SOURCES = \ + compile.c \ + process.c \ + purify.c \ + tokenise.c \ + version.c + +APPS = $(SOURCES:%.c=%) + +INCLUDES = -I. + + +##### RULES ##### + +.SUFFIXES: +.SUFFIXES: .c + +.c: + $(APP_CC) $(INCLUDES) $(CFLAGS) $(LDFLAGS) $< $(LIBS) -o $@ + +.c.o: + $(APP_CC) -c $(INCLUDES) $(CFLAGS) $(DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: $(APPS) + +install: + +clean: + -rm -f $(APPS) + -rm -f *.o diff --git a/src/glsl/apps/compile.c b/src/glsl/apps/compile.c new file mode 100644 index 00000000000..c9a830b9f3c --- /dev/null +++ b/src/glsl/apps/compile.c @@ -0,0 +1,191 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <assert.h> +#include "../pp/sl_pp_public.h" +#include "../cl/sl_cl_parse.h" + + +static void +usage(void) +{ + printf("Usage:\n"); + printf(" compile fragment|vertex <source> <output>\n"); +} + +int +main(int argc, + char *argv[]) +{ + FILE *in; + long size; + char *inbuf; + struct sl_pp_purify_options options; + char errmsg[100] = ""; + struct sl_pp_context *context; + unsigned int version; + FILE *out; + unsigned char *outbytes; + unsigned int cboutbytes; + unsigned int shader_type; + + if (argc != 4) { + usage(); + return 1; + } + + if (!strcmp(argv[1], "fragment")) { + shader_type = 1; + } else if (!strcmp(argv[1], "vertex")) { + shader_type = 2; + } else { + usage(); + return 1; + } + + in = fopen(argv[2], "rb"); + if (!in) { + printf("Could not open `%s' for read.\n", argv[2]); + usage(); + return 1; + } + + fseek(in, 0, SEEK_END); + size = ftell(in); + fseek(in, 0, SEEK_SET); + + out = fopen(argv[3], "w"); + if (!out) { + fclose(in); + printf("Could not open `%s' for write.\n", argv[3]); + usage(); + return 1; + } + + inbuf = malloc(size + 1); + if (!inbuf) { + fprintf(out, "$OOMERROR\n"); + + fclose(out); + fclose(in); + printf("Out of memory.\n"); + return 0; + } + + if (fread(inbuf, 1, size, in) != size) { + fprintf(out, "$READERROR\n"); + + free(inbuf); + fclose(out); + fclose(in); + printf("Could not read from `%s'.\n", argv[2]); + return 0; + } + inbuf[size] = '\0'; + + fclose(in); + + memset(&options, 0, sizeof(options)); + + context = sl_pp_context_create(inbuf, &options); + if (!context) { + fprintf(out, "$CONTEXERROR\n"); + + free(inbuf); + fclose(out); + printf("Could not create parse context.\n"); + return 0; + } + + if (sl_pp_version(context, &version)) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + printf("Error: %s\n", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 0; + } + + if (sl_pp_context_add_extension(context, "ARB_draw_buffers", "GL_ARB_draw_buffers") || + sl_pp_context_add_extension(context, "ARB_texture_rectangle", "GL_ARB_texture_rectangle")) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + printf("Error: %s\n", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 0; + } + + if (sl_cl_compile(context, shader_type, 1, &outbytes, &cboutbytes, errmsg, sizeof(errmsg)) == 0) { + unsigned int i; + unsigned int line = 0; + + fprintf(out, "\n/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */"); + fprintf(out, "\n/* %s */", argv[2]); + fprintf(out, "\n\n"); + + for (i = 0; i < cboutbytes; i++) { + unsigned int a; + + if (outbytes[i] < 10) { + a = 1; + } else if (outbytes[i] < 100) { + a = 2; + } else { + a = 3; + } + if (i < cboutbytes - 1) { + a++; + } + if (line + a >= 100) { + fprintf (out, "\n"); + line = 0; + } + line += a; + fprintf (out, "%u", outbytes[i]); + if (i < cboutbytes - 1) { + fprintf (out, ","); + } + } + fprintf (out, "\n"); + free(outbytes); + } else { + fprintf(out, "$SYNTAXERROR: `%s'\n", errmsg); + + printf("Error: %s\n", errmsg); + } + + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 0; +} diff --git a/src/glsl/apps/process.c b/src/glsl/apps/process.c new file mode 100644 index 00000000000..569890210f2 --- /dev/null +++ b/src/glsl/apps/process.c @@ -0,0 +1,381 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <assert.h> +#include "../pp/sl_pp_public.h" + + +int +main(int argc, + char *argv[]) +{ + FILE *in; + long size; + char *inbuf; + struct sl_pp_purify_options options; + struct sl_pp_context *context; + unsigned int version; + struct sl_pp_token_info *outtokens; + FILE *out; + unsigned int i; + + if (argc != 3) { + return 1; + } + + in = fopen(argv[1], "rb"); + if (!in) { + return 1; + } + + fseek(in, 0, SEEK_END); + size = ftell(in); + fseek(in, 0, SEEK_SET); + + out = fopen(argv[2], "wb"); + if (!out) { + fclose(in); + return 1; + } + + inbuf = malloc(size + 1); + if (!inbuf) { + fprintf(out, "$OOMERROR\n"); + + fclose(out); + fclose(in); + return 1; + } + + if (fread(inbuf, 1, size, in) != size) { + fprintf(out, "$READERROR\n"); + + free(inbuf); + fclose(out); + fclose(in); + return 1; + } + inbuf[size] = '\0'; + + fclose(in); + + memset(&options, 0, sizeof(options)); + + context = sl_pp_context_create(inbuf, &options); + if (!context) { + fprintf(out, "$CONTEXERROR\n"); + + free(inbuf); + fclose(out); + return 1; + } + + if (sl_pp_version(context, &version)) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return -1; + } + + if (sl_pp_context_add_extension(context, "ARB_draw_buffers", "GL_ARB_draw_buffers") || + sl_pp_context_add_extension(context, "ARB_texture_rectangle", "GL_ARB_texture_rectangle")) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + printf("Error: %s\n", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 0; + } + + if (sl_pp_context_add_predefined(context, "__GLSL_PP_PREDEFINED_MACRO_TEST", "1")) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + printf("Error: %s\n", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 0; + } + + if (sl_pp_process(context, &outtokens)) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return -1; + } + + free(inbuf); + + for (i = 0; outtokens[i].token != SL_PP_EOF; i++) { + switch (outtokens[i].token) { + case SL_PP_NEWLINE: + fprintf(out, "\n"); + break; + + case SL_PP_COMMA: + fprintf(out, ", "); + break; + + case SL_PP_SEMICOLON: + fprintf(out, "; "); + break; + + case SL_PP_LBRACE: + fprintf(out, "{ "); + break; + + case SL_PP_RBRACE: + fprintf(out, "} "); + break; + + case SL_PP_LPAREN: + fprintf(out, "( "); + break; + + case SL_PP_RPAREN: + fprintf(out, ") "); + break; + + case SL_PP_LBRACKET: + fprintf(out, "[ "); + break; + + case SL_PP_RBRACKET: + fprintf(out, "] "); + break; + + case SL_PP_DOT: + fprintf(out, ". "); + break; + + case SL_PP_INCREMENT: + fprintf(out, "++ "); + break; + + case SL_PP_ADDASSIGN: + fprintf(out, "+= "); + break; + + case SL_PP_PLUS: + fprintf(out, "+ "); + break; + + case SL_PP_DECREMENT: + fprintf(out, "-- "); + break; + + case SL_PP_SUBASSIGN: + fprintf(out, "-= "); + break; + + case SL_PP_MINUS: + fprintf(out, "- "); + break; + + case SL_PP_BITNOT: + fprintf(out, "~ "); + break; + + case SL_PP_NOTEQUAL: + fprintf(out, "!= "); + break; + + case SL_PP_NOT: + fprintf(out, "! "); + break; + + case SL_PP_MULASSIGN: + fprintf(out, "*= "); + break; + + case SL_PP_STAR: + fprintf(out, "* "); + break; + + case SL_PP_DIVASSIGN: + fprintf(out, "/= "); + break; + + case SL_PP_SLASH: + fprintf(out, "/ "); + break; + + case SL_PP_MODASSIGN: + fprintf(out, "%%= "); + break; + + case SL_PP_MODULO: + fprintf(out, "%% "); + break; + + case SL_PP_LSHIFTASSIGN: + fprintf(out, "<<= "); + break; + + case SL_PP_LSHIFT: + fprintf(out, "<< "); + break; + + case SL_PP_LESSEQUAL: + fprintf(out, "<= "); + break; + + case SL_PP_LESS: + fprintf(out, "< "); + break; + + case SL_PP_RSHIFTASSIGN: + fprintf(out, ">>= "); + break; + + case SL_PP_RSHIFT: + fprintf(out, ">> "); + break; + + case SL_PP_GREATEREQUAL: + fprintf(out, ">= "); + break; + + case SL_PP_GREATER: + fprintf(out, "> "); + break; + + case SL_PP_EQUAL: + fprintf(out, "== "); + break; + + case SL_PP_ASSIGN: + fprintf(out, "= "); + break; + + case SL_PP_AND: + fprintf(out, "&& "); + break; + + case SL_PP_BITANDASSIGN: + fprintf(out, "&= "); + break; + + case SL_PP_BITAND: + fprintf(out, "& "); + break; + + case SL_PP_XOR: + fprintf(out, "^^ "); + break; + + case SL_PP_BITXORASSIGN: + fprintf(out, "^= "); + break; + + case SL_PP_BITXOR: + fprintf(out, "^ "); + break; + + case SL_PP_OR: + fprintf(out, "|| "); + break; + + case SL_PP_BITORASSIGN: + fprintf(out, "|= "); + break; + + case SL_PP_BITOR: + fprintf(out, "| "); + break; + + case SL_PP_QUESTION: + fprintf(out, "? "); + break; + + case SL_PP_COLON: + fprintf(out, ": "); + break; + + case SL_PP_IDENTIFIER: + fprintf(out, "%s ", sl_pp_context_cstr(context, outtokens[i].data.identifier)); + break; + + case SL_PP_UINT: + fprintf(out, "%s ", sl_pp_context_cstr(context, outtokens[i].data._uint)); + break; + + case SL_PP_FLOAT: + fprintf(out, "%s ", sl_pp_context_cstr(context, outtokens[i].data._float)); + break; + + case SL_PP_OTHER: + fprintf(out, "%c", outtokens[i].data.other); + break; + + case SL_PP_PRAGMA_OPTIMIZE: + fprintf(out, "#pragma optimize(%s)", outtokens[i].data.pragma ? "on" : "off"); + break; + + case SL_PP_PRAGMA_DEBUG: + fprintf(out, "#pragma debug(%s)", outtokens[i].data.pragma ? "on" : "off"); + break; + + case SL_PP_EXTENSION_REQUIRE: + fprintf(out, "#extension %s : require", sl_pp_context_cstr(context, outtokens[i].data.extension)); + break; + + case SL_PP_EXTENSION_ENABLE: + fprintf(out, "#extension %s : enable", sl_pp_context_cstr(context, outtokens[i].data.extension)); + break; + + case SL_PP_EXTENSION_WARN: + fprintf(out, "#extension %s : warn", sl_pp_context_cstr(context, outtokens[i].data.extension)); + break; + + case SL_PP_EXTENSION_DISABLE: + fprintf(out, "#extension %s : disable", sl_pp_context_cstr(context, outtokens[i].data.extension)); + break; + + case SL_PP_LINE: + fprintf(out, "#line %u %u", outtokens[i].data.line.lineno, outtokens[i].data.line.fileno); + break; + + default: + assert(0); + } + } + + sl_pp_context_destroy(context); + free(outtokens); + fclose(out); + + return 0; +} diff --git a/src/glsl/apps/purify.c b/src/glsl/apps/purify.c new file mode 100644 index 00000000000..8c01f4fc6a3 --- /dev/null +++ b/src/glsl/apps/purify.c @@ -0,0 +1,105 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "../pp/sl_pp_public.h" + + +int +main(int argc, + char *argv[]) +{ + FILE *in; + long size; + char *inbuf; + struct sl_pp_purify_options options; + char *outbuf; + char errmsg[100] = ""; + unsigned int errline = 0; + FILE *out; + + if (argc != 3) { + return 1; + } + + in = fopen(argv[1], "rb"); + if (!in) { + return 1; + } + + fseek(in, 0, SEEK_END); + size = ftell(in); + fseek(in, 0, SEEK_SET); + + out = fopen(argv[2], "wb"); + if (!out) { + fclose(in); + return 1; + } + + inbuf = malloc(size + 1); + if (!inbuf) { + fprintf(out, "$OOMERROR\n"); + + fclose(out); + fclose(in); + return 1; + } + + if (fread(inbuf, 1, size, in) != size) { + fprintf(out, "$READERROR\n"); + + free(inbuf); + fclose(out); + fclose(in); + return 1; + } + inbuf[size] = '\0'; + + fclose(in); + + memset(&options, 0, sizeof(options)); + + if (sl_pp_purify(inbuf, &options, &outbuf, errmsg, sizeof(errmsg), &errline)) { + fprintf(out, "$PURIFYERROR %u: %s\n", errline, errmsg); + + free(inbuf); + fclose(out); + return 1; + } + + free(inbuf); + + fwrite(outbuf, 1, strlen(outbuf), out); + + free(outbuf); + fclose(out); + + return 0; +} diff --git a/src/glsl/apps/tokenise.c b/src/glsl/apps/tokenise.c new file mode 100644 index 00000000000..9ff73157e9c --- /dev/null +++ b/src/glsl/apps/tokenise.c @@ -0,0 +1,333 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <assert.h> +#include "../pp/sl_pp_public.h" + + +int +main(int argc, + char *argv[]) +{ + FILE *in; + long size; + char *inbuf; + struct sl_pp_purify_options options; + struct sl_pp_context *context; + struct sl_pp_token_info *tokens; + FILE *out; + unsigned int i; + + if (argc != 3) { + return 1; + } + + in = fopen(argv[1], "rb"); + if (!in) { + return 1; + } + + fseek(in, 0, SEEK_END); + size = ftell(in); + fseek(in, 0, SEEK_SET); + + out = fopen(argv[2], "wb"); + if (!out) { + fclose(in); + return 1; + } + + inbuf = malloc(size + 1); + if (!inbuf) { + fprintf(out, "$OOMERROR\n"); + + fclose(out); + fclose(in); + return 1; + } + + if (fread(inbuf, 1, size, in) != size) { + fprintf(out, "$READERROR\n"); + + free(inbuf); + fclose(out); + fclose(in); + return 1; + } + inbuf[size] = '\0'; + + fclose(in); + + memset(&options, 0, sizeof(options)); + + context = sl_pp_context_create(inbuf, &options); + if (!context) { + fprintf(out, "$CONTEXERROR\n"); + + free(inbuf); + fclose(out); + return 1; + } + + if (sl_pp_tokenise(context, &tokens)) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return 1; + } + + free(inbuf); + + for (i = 0; tokens[i].token != SL_PP_EOF; i++) { + switch (tokens[i].token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_NEWLINE: + fprintf(out, "\n"); + break; + + case SL_PP_HASH: + fprintf(out, "# "); + break; + + case SL_PP_COMMA: + fprintf(out, ", "); + break; + + case SL_PP_SEMICOLON: + fprintf(out, "; "); + break; + + case SL_PP_LBRACE: + fprintf(out, "{ "); + break; + + case SL_PP_RBRACE: + fprintf(out, "} "); + break; + + case SL_PP_LPAREN: + fprintf(out, "( "); + break; + + case SL_PP_RPAREN: + fprintf(out, ") "); + break; + + case SL_PP_LBRACKET: + fprintf(out, "[ "); + break; + + case SL_PP_RBRACKET: + fprintf(out, "] "); + break; + + case SL_PP_DOT: + fprintf(out, ". "); + break; + + case SL_PP_INCREMENT: + fprintf(out, "++ "); + break; + + case SL_PP_ADDASSIGN: + fprintf(out, "+= "); + break; + + case SL_PP_PLUS: + fprintf(out, "+ "); + break; + + case SL_PP_DECREMENT: + fprintf(out, "-- "); + break; + + case SL_PP_SUBASSIGN: + fprintf(out, "-= "); + break; + + case SL_PP_MINUS: + fprintf(out, "- "); + break; + + case SL_PP_BITNOT: + fprintf(out, "~ "); + break; + + case SL_PP_NOTEQUAL: + fprintf(out, "!= "); + break; + + case SL_PP_NOT: + fprintf(out, "! "); + break; + + case SL_PP_MULASSIGN: + fprintf(out, "*= "); + break; + + case SL_PP_STAR: + fprintf(out, "* "); + break; + + case SL_PP_DIVASSIGN: + fprintf(out, "/= "); + break; + + case SL_PP_SLASH: + fprintf(out, "/ "); + break; + + case SL_PP_MODASSIGN: + fprintf(out, "%%= "); + break; + + case SL_PP_MODULO: + fprintf(out, "%% "); + break; + + case SL_PP_LSHIFTASSIGN: + fprintf(out, "<<= "); + break; + + case SL_PP_LSHIFT: + fprintf(out, "<< "); + break; + + case SL_PP_LESSEQUAL: + fprintf(out, "<= "); + break; + + case SL_PP_LESS: + fprintf(out, "< "); + break; + + case SL_PP_RSHIFTASSIGN: + fprintf(out, ">>= "); + break; + + case SL_PP_RSHIFT: + fprintf(out, ">> "); + break; + + case SL_PP_GREATEREQUAL: + fprintf(out, ">= "); + break; + + case SL_PP_GREATER: + fprintf(out, "> "); + break; + + case SL_PP_EQUAL: + fprintf(out, "== "); + break; + + case SL_PP_ASSIGN: + fprintf(out, "= "); + break; + + case SL_PP_AND: + fprintf(out, "&& "); + break; + + case SL_PP_BITANDASSIGN: + fprintf(out, "&= "); + break; + + case SL_PP_BITAND: + fprintf(out, "& "); + break; + + case SL_PP_XOR: + fprintf(out, "^^ "); + break; + + case SL_PP_BITXORASSIGN: + fprintf(out, "^= "); + break; + + case SL_PP_BITXOR: + fprintf(out, "^ "); + break; + + case SL_PP_OR: + fprintf(out, "|| "); + break; + + case SL_PP_BITORASSIGN: + fprintf(out, "|= "); + break; + + case SL_PP_BITOR: + fprintf(out, "| "); + break; + + case SL_PP_QUESTION: + fprintf(out, "? "); + break; + + case SL_PP_COLON: + fprintf(out, ": "); + break; + + case SL_PP_IDENTIFIER: + fprintf(out, "%s ", sl_pp_context_cstr(context, tokens[i].data.identifier)); + break; + + case SL_PP_UINT: + fprintf(out, "(%s) ", sl_pp_context_cstr(context, tokens[i].data._uint)); + break; + + case SL_PP_FLOAT: + fprintf(out, "(%s) ", sl_pp_context_cstr(context, tokens[i].data._float)); + break; + + case SL_PP_OTHER: + if (tokens[i].data.other == '\'') { + fprintf(out, "'\\'' "); + } else { + fprintf(out, "'%c' ", tokens[i].data.other); + } + break; + + default: + assert(0); + } + } + + sl_pp_context_destroy(context); + free(tokens); + fclose(out); + + return 0; +} diff --git a/src/glsl/apps/version.c b/src/glsl/apps/version.c new file mode 100644 index 00000000000..40a4a069c30 --- /dev/null +++ b/src/glsl/apps/version.c @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <assert.h> +#include "../pp/sl_pp_public.h" + + +int +main(int argc, + char *argv[]) +{ + FILE *in; + long size; + char *inbuf; + struct sl_pp_purify_options options; + struct sl_pp_context *context; + unsigned int version; + FILE *out; + + if (argc != 3) { + return 1; + } + + in = fopen(argv[1], "rb"); + if (!in) { + return 1; + } + + fseek(in, 0, SEEK_END); + size = ftell(in); + fseek(in, 0, SEEK_SET); + + out = fopen(argv[2], "wb"); + if (!out) { + fclose(in); + return 1; + } + + inbuf = malloc(size + 1); + if (!inbuf) { + fprintf(out, "$OOMERROR\n"); + + fclose(out); + fclose(in); + return 1; + } + + if (fread(inbuf, 1, size, in) != size) { + fprintf(out, "$READERROR\n"); + + free(inbuf); + fclose(out); + fclose(in); + return 1; + } + inbuf[size] = '\0'; + + fclose(in); + + memset(&options, 0, sizeof(options)); + + context = sl_pp_context_create(inbuf, &options); + if (!context) { + fprintf(out, "$CONTEXERROR\n"); + + free(inbuf); + fclose(out); + return 1; + } + + if (sl_pp_version(context, &version)) { + fprintf(out, "$ERROR: `%s'\n", sl_pp_context_error_message(context)); + + sl_pp_context_destroy(context); + free(inbuf); + fclose(out); + return -1; + } + + sl_pp_context_destroy(context); + free(inbuf); + + fprintf(out, "%u\n", version); + + fclose(out); + + return 0; +} diff --git a/src/glsl/cl/Makefile b/src/glsl/cl/Makefile new file mode 100644 index 00000000000..04a52df8c33 --- /dev/null +++ b/src/glsl/cl/Makefile @@ -0,0 +1,13 @@ +#src/glsl/cl/Makefile + +TOP = ../../.. + +include $(TOP)/configs/current + +LIBNAME = glslcl + +C_SOURCES = \ + sl_cl_parse.c + +include ../Makefile.template + diff --git a/src/glsl/cl/sl_cl_parse.c b/src/glsl/cl/sl_cl_parse.c new file mode 100644 index 00000000000..5dddf434e15 --- /dev/null +++ b/src/glsl/cl/sl_cl_parse.c @@ -0,0 +1,2823 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <assert.h> +#include <stdlib.h> +#include <string.h> +#include "../pp/sl_pp_public.h" +#include "sl_cl_parse.h" + + +/* revision number - increment after each change affecting emitted output */ +#define REVISION 5 + +/* external declaration (or precision or invariant stmt) */ +#define EXTERNAL_NULL 0 +#define EXTERNAL_FUNCTION_DEFINITION 1 +#define EXTERNAL_DECLARATION 2 +#define DEFAULT_PRECISION 3 +#define INVARIANT_STMT 4 + +/* precision */ +#define PRECISION_DEFAULT 0 +#define PRECISION_LOW 1 +#define PRECISION_MEDIUM 2 +#define PRECISION_HIGH 3 + +/* declaration */ +#define DECLARATION_FUNCTION_PROTOTYPE 1 +#define DECLARATION_INIT_DECLARATOR_LIST 2 + +/* function type */ +#define FUNCTION_ORDINARY 0 +#define FUNCTION_CONSTRUCTOR 1 +#define FUNCTION_OPERATOR 2 + +/* function call type */ +#define FUNCTION_CALL_NONARRAY 0 +#define FUNCTION_CALL_ARRAY 1 + +/* operator type */ +#define OPERATOR_ADDASSIGN 1 +#define OPERATOR_SUBASSIGN 2 +#define OPERATOR_MULASSIGN 3 +#define OPERATOR_DIVASSIGN 4 +/*#define OPERATOR_MODASSIGN 5*/ +/*#define OPERATOR_LSHASSIGN 6*/ +/*#define OPERATOR_RSHASSIGN 7*/ +/*#define OPERATOR_ORASSIGN 8*/ +/*#define OPERATOR_XORASSIGN 9*/ +/*#define OPERATOR_ANDASSIGN 10*/ +#define OPERATOR_LOGICALXOR 11 +/*#define OPERATOR_BITOR 12*/ +/*#define OPERATOR_BITXOR 13*/ +/*#define OPERATOR_BITAND 14*/ +#define OPERATOR_LESS 15 +#define OPERATOR_GREATER 16 +#define OPERATOR_LESSEQUAL 17 +#define OPERATOR_GREATEREQUAL 18 +/*#define OPERATOR_LSHIFT 19*/ +/*#define OPERATOR_RSHIFT 20*/ +#define OPERATOR_MULTIPLY 21 +#define OPERATOR_DIVIDE 22 +/*#define OPERATOR_MODULUS 23*/ +#define OPERATOR_INCREMENT 24 +#define OPERATOR_DECREMENT 25 +#define OPERATOR_PLUS 26 +#define OPERATOR_MINUS 27 +/*#define OPERATOR_COMPLEMENT 28*/ +#define OPERATOR_NOT 29 + +/* init declarator list */ +#define DECLARATOR_NONE 0 +#define DECLARATOR_NEXT 1 + +/* variable declaration */ +#define VARIABLE_NONE 0 +#define VARIABLE_IDENTIFIER 1 +#define VARIABLE_INITIALIZER 2 +#define VARIABLE_ARRAY_EXPLICIT 3 +#define VARIABLE_ARRAY_UNKNOWN 4 + +/* type qualifier */ +#define TYPE_QUALIFIER_NONE 0 +#define TYPE_QUALIFIER_CONST 1 +#define TYPE_QUALIFIER_ATTRIBUTE 2 +#define TYPE_QUALIFIER_VARYING 3 +#define TYPE_QUALIFIER_UNIFORM 4 +#define TYPE_QUALIFIER_FIXEDOUTPUT 5 +#define TYPE_QUALIFIER_FIXEDINPUT 6 + +/* invariant qualifier */ +#define TYPE_VARIANT 90 +#define TYPE_INVARIANT 91 + +/* centroid qualifier */ +#define TYPE_CENTER 95 +#define TYPE_CENTROID 96 + +/* type specifier */ +#define TYPE_SPECIFIER_VOID 0 +#define TYPE_SPECIFIER_BOOL 1 +#define TYPE_SPECIFIER_BVEC2 2 +#define TYPE_SPECIFIER_BVEC3 3 +#define TYPE_SPECIFIER_BVEC4 4 +#define TYPE_SPECIFIER_INT 5 +#define TYPE_SPECIFIER_IVEC2 6 +#define TYPE_SPECIFIER_IVEC3 7 +#define TYPE_SPECIFIER_IVEC4 8 +#define TYPE_SPECIFIER_FLOAT 9 +#define TYPE_SPECIFIER_VEC2 10 +#define TYPE_SPECIFIER_VEC3 11 +#define TYPE_SPECIFIER_VEC4 12 +#define TYPE_SPECIFIER_MAT2 13 +#define TYPE_SPECIFIER_MAT3 14 +#define TYPE_SPECIFIER_MAT4 15 +#define TYPE_SPECIFIER_SAMPLER1D 16 +#define TYPE_SPECIFIER_SAMPLER2D 17 +#define TYPE_SPECIFIER_SAMPLER3D 18 +#define TYPE_SPECIFIER_SAMPLERCUBE 19 +#define TYPE_SPECIFIER_SAMPLER1DSHADOW 20 +#define TYPE_SPECIFIER_SAMPLER2DSHADOW 21 +#define TYPE_SPECIFIER_SAMPLER2DRECT 22 +#define TYPE_SPECIFIER_SAMPLER2DRECTSHADOW 23 +#define TYPE_SPECIFIER_STRUCT 24 +#define TYPE_SPECIFIER_TYPENAME 25 + +/* OpenGL 2.1 */ +#define TYPE_SPECIFIER_MAT23 26 +#define TYPE_SPECIFIER_MAT32 27 +#define TYPE_SPECIFIER_MAT24 28 +#define TYPE_SPECIFIER_MAT42 29 +#define TYPE_SPECIFIER_MAT34 30 +#define TYPE_SPECIFIER_MAT43 31 + +/* type specifier array */ +#define TYPE_SPECIFIER_NONARRAY 0 +#define TYPE_SPECIFIER_ARRAY 1 + +/* structure field */ +#define FIELD_NONE 0 +#define FIELD_NEXT 1 +#define FIELD_ARRAY 2 + +/* operation */ +#define OP_END 0 +#define OP_BLOCK_BEGIN_NO_NEW_SCOPE 1 +#define OP_BLOCK_BEGIN_NEW_SCOPE 2 +#define OP_DECLARE 3 +#define OP_ASM 4 +#define OP_BREAK 5 +#define OP_CONTINUE 6 +#define OP_DISCARD 7 +#define OP_RETURN 8 +#define OP_EXPRESSION 9 +#define OP_IF 10 +#define OP_WHILE 11 +#define OP_DO 12 +#define OP_FOR 13 +#define OP_PUSH_VOID 14 +#define OP_PUSH_BOOL 15 +#define OP_PUSH_INT 16 +#define OP_PUSH_FLOAT 17 +#define OP_PUSH_IDENTIFIER 18 +#define OP_SEQUENCE 19 +#define OP_ASSIGN 20 +#define OP_ADDASSIGN 21 +#define OP_SUBASSIGN 22 +#define OP_MULASSIGN 23 +#define OP_DIVASSIGN 24 +/*#define OP_MODASSIGN 25*/ +/*#define OP_LSHASSIGN 26*/ +/*#define OP_RSHASSIGN 27*/ +/*#define OP_ORASSIGN 28*/ +/*#define OP_XORASSIGN 29*/ +/*#define OP_ANDASSIGN 30*/ +#define OP_SELECT 31 +#define OP_LOGICALOR 32 +#define OP_LOGICALXOR 33 +#define OP_LOGICALAND 34 +/*#define OP_BITOR 35*/ +/*#define OP_BITXOR 36*/ +/*#define OP_BITAND 37*/ +#define OP_EQUAL 38 +#define OP_NOTEQUAL 39 +#define OP_LESS 40 +#define OP_GREATER 41 +#define OP_LESSEQUAL 42 +#define OP_GREATEREQUAL 43 +/*#define OP_LSHIFT 44*/ +/*#define OP_RSHIFT 45*/ +#define OP_ADD 46 +#define OP_SUBTRACT 47 +#define OP_MULTIPLY 48 +#define OP_DIVIDE 49 +/*#define OP_MODULUS 50*/ +#define OP_PREINCREMENT 51 +#define OP_PREDECREMENT 52 +#define OP_PLUS 53 +#define OP_MINUS 54 +/*#define OP_COMPLEMENT 55*/ +#define OP_NOT 56 +#define OP_SUBSCRIPT 57 +#define OP_CALL 58 +#define OP_FIELD 59 +#define OP_POSTINCREMENT 60 +#define OP_POSTDECREMENT 61 +#define OP_PRECISION 62 +#define OP_METHOD 63 + +/* parameter qualifier */ +#define PARAM_QUALIFIER_IN 0 +#define PARAM_QUALIFIER_OUT 1 +#define PARAM_QUALIFIER_INOUT 2 + +/* function parameter */ +#define PARAMETER_NONE 0 +#define PARAMETER_NEXT 1 + +/* function parameter array presence */ +#define PARAMETER_ARRAY_NOT_PRESENT 0 +#define PARAMETER_ARRAY_PRESENT 1 + + +struct parse_dict { + int _void; + int _float; + int _int; + int _bool; + int vec2; + int vec3; + int vec4; + int bvec2; + int bvec3; + int bvec4; + int ivec2; + int ivec3; + int ivec4; + int mat2; + int mat3; + int mat4; + int mat2x3; + int mat3x2; + int mat2x4; + int mat4x2; + int mat3x4; + int mat4x3; + int sampler1D; + int sampler2D; + int sampler3D; + int samplerCube; + int sampler1DShadow; + int sampler2DShadow; + int sampler2DRect; + int sampler2DRectShadow; + + int invariant; + + int centroid; + + int precision; + int lowp; + int mediump; + int highp; + + int _const; + int attribute; + int varying; + int uniform; + int __fixed_output; + int __fixed_input; + + int in; + int out; + int inout; + + int _struct; + + int __constructor; + int __operator; + int ___asm; + + int _if; + int _else; + int _for; + int _while; + int _do; + + int _continue; + int _break; + int _return; + int discard; + + int _false; + int _true; +}; + + +struct parse_context { + struct sl_pp_context *context; + + struct parse_dict dict; + + struct sl_pp_token_info *tokens; + unsigned int tokens_read; + unsigned int tokens_cap; + + unsigned char *out_buf; + unsigned int out_cap; + + unsigned int shader_type; + unsigned int parsing_builtin; + + char error[256]; + int process_error; +}; + + +struct parse_state { + unsigned int in; + unsigned int out; +}; + + +static __inline unsigned int +_emit(struct parse_context *ctx, + unsigned int *out, + unsigned char b) +{ + if (*out == ctx->out_cap) { + ctx->out_cap += 4096; + ctx->out_buf = (unsigned char *)realloc(ctx->out_buf, ctx->out_cap * sizeof(unsigned char)); + } + ctx->out_buf[*out] = b; + return (*out)++; +} + + +static void +_update(struct parse_context *ctx, + unsigned int out, + unsigned char b) +{ + ctx->out_buf[out] = b; +} + + +static void +_error(struct parse_context *ctx, + const char *msg) +{ + if (ctx->error[0] == '\0') { + strcpy(ctx->error, msg); + } +} + + +static const struct sl_pp_token_info * +_fetch_token(struct parse_context *ctx, + unsigned int pos) +{ + if (ctx->process_error) { + return NULL; + } + + while (pos >= ctx->tokens_read) { + if (ctx->tokens_read == ctx->tokens_cap) { + ctx->tokens_cap += 1024; + ctx->tokens = realloc(ctx->tokens, + ctx->tokens_cap * sizeof(struct sl_pp_token_info)); + if (!ctx->tokens) { + _error(ctx, "out of memory"); + ctx->process_error = 1; + return NULL; + } + } + if (sl_pp_process_get(ctx->context, &ctx->tokens[ctx->tokens_read])) { + _error(ctx, sl_pp_context_error_message(ctx->context)); + ctx->process_error = 1; + return NULL; + } + switch (ctx->tokens[ctx->tokens_read].token) { + case SL_PP_COMMA: + case SL_PP_SEMICOLON: + case SL_PP_LBRACE: + case SL_PP_RBRACE: + case SL_PP_LPAREN: + case SL_PP_RPAREN: + case SL_PP_LBRACKET: + case SL_PP_RBRACKET: + case SL_PP_DOT: + case SL_PP_INCREMENT: + case SL_PP_ADDASSIGN: + case SL_PP_PLUS: + case SL_PP_DECREMENT: + case SL_PP_SUBASSIGN: + case SL_PP_MINUS: + case SL_PP_BITNOT: + case SL_PP_NOTEQUAL: + case SL_PP_NOT: + case SL_PP_MULASSIGN: + case SL_PP_STAR: + case SL_PP_DIVASSIGN: + case SL_PP_SLASH: + case SL_PP_MODASSIGN: + case SL_PP_MODULO: + case SL_PP_LSHIFTASSIGN: + case SL_PP_LSHIFT: + case SL_PP_LESSEQUAL: + case SL_PP_LESS: + case SL_PP_RSHIFTASSIGN: + case SL_PP_RSHIFT: + case SL_PP_GREATEREQUAL: + case SL_PP_GREATER: + case SL_PP_EQUAL: + case SL_PP_ASSIGN: + case SL_PP_AND: + case SL_PP_BITANDASSIGN: + case SL_PP_BITAND: + case SL_PP_XOR: + case SL_PP_BITXORASSIGN: + case SL_PP_BITXOR: + case SL_PP_OR: + case SL_PP_BITORASSIGN: + case SL_PP_BITOR: + case SL_PP_QUESTION: + case SL_PP_COLON: + case SL_PP_IDENTIFIER: + case SL_PP_UINT: + case SL_PP_FLOAT: + case SL_PP_EOF: + ctx->tokens_read++; + break; + } + } + return &ctx->tokens[pos]; +} + + +static int +_parse_token(struct parse_context *ctx, + enum sl_pp_token token, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + + if (input && input->token == token) { + ps->in++; + return 0; + } + return -1; +} + + +static int +_parse_id(struct parse_context *ctx, + int id, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + + if (input && input->token == SL_PP_IDENTIFIER && input->data.identifier == id) { + ps->in++; + return 0; + } + return -1; +} + + +static int +_parse_identifier(struct parse_context *ctx, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + + if (input && input->token == SL_PP_IDENTIFIER) { + const char *cstr = sl_pp_context_cstr(ctx->context, input->data.identifier); + + do { + _emit(ctx, &ps->out, *cstr); + } while (*cstr++); + ps->in++; + return 0; + } + return -1; +} + + +static int +_parse_float(struct parse_context *ctx, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + + if (input && input->token == SL_PP_FLOAT) { + const char *cstr = sl_pp_context_cstr(ctx->context, input->data._float); + + _emit(ctx, &ps->out, 1); + do { + _emit(ctx, &ps->out, *cstr); + } while (*cstr++); + ps->in++; + return 0; + } + return -1; +} + + +static int +_parse_uint(struct parse_context *ctx, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + + if (input && input->token == SL_PP_UINT) { + const char *cstr = sl_pp_context_cstr(ctx->context, input->data._uint); + + _emit(ctx, &ps->out, 1); + do { + _emit(ctx, &ps->out, *cstr); + } while (*cstr++); + ps->in++; + return 0; + } + return -1; +} + + +/**************************************/ + + +static int +_parse_unary_expression(struct parse_context *ctx, + struct parse_state *ps); + +static int +_parse_conditional_expression(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_constant_expression(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_primary_expression(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_statement(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_type_specifier(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_declaration(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_statement_list(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_assignment_expression(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_precision(struct parse_context *ctx, + struct parse_state *ps); + + +static int +_parse_overriden_operator(struct parse_context *ctx, + struct parse_state *ps) +{ + unsigned int op; + + if (_parse_token(ctx, SL_PP_INCREMENT, ps) == 0) { + op = OPERATOR_INCREMENT; + } else if (_parse_token(ctx, SL_PP_ADDASSIGN, ps) == 0) { + op = OPERATOR_ADDASSIGN; + } else if (_parse_token(ctx, SL_PP_PLUS, ps) == 0) { + op = OPERATOR_PLUS; + } else if (_parse_token(ctx, SL_PP_DECREMENT, ps) == 0) { + op = OPERATOR_DECREMENT; + } else if (_parse_token(ctx, SL_PP_SUBASSIGN, ps) == 0) { + op = OPERATOR_SUBASSIGN; + } else if (_parse_token(ctx, SL_PP_MINUS, ps) == 0) { + op = OPERATOR_MINUS; + } else if (_parse_token(ctx, SL_PP_NOT, ps) == 0) { + op = OPERATOR_NOT; + } else if (_parse_token(ctx, SL_PP_MULASSIGN, ps) == 0) { + op = OPERATOR_MULASSIGN; + } else if (_parse_token(ctx, SL_PP_STAR, ps) == 0) { + op = OPERATOR_MULTIPLY; + } else if (_parse_token(ctx, SL_PP_DIVASSIGN, ps) == 0) { + op = OPERATOR_DIVASSIGN; + } else if (_parse_token(ctx, SL_PP_SLASH, ps) == 0) { + op = OPERATOR_DIVIDE; + } else if (_parse_token(ctx, SL_PP_LESSEQUAL, ps) == 0) { + op = OPERATOR_LESSEQUAL; + } else if (_parse_token(ctx, SL_PP_LESS, ps) == 0) { + op = OPERATOR_LESS; + } else if (_parse_token(ctx, SL_PP_GREATEREQUAL, ps) == 0) { + op = OPERATOR_GREATEREQUAL; + } else if (_parse_token(ctx, SL_PP_GREATER, ps) == 0) { + op = OPERATOR_GREATER; + } else if (_parse_token(ctx, SL_PP_XOR, ps) == 0) { + op = OPERATOR_LOGICALXOR; + } else { + return -1; + } + + _emit(ctx, &ps->out, op); + return 0; +} + + +static int +_parse_function_decl_identifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, 0); + + if (ctx->parsing_builtin && _parse_id(ctx, ctx->dict.__constructor, &p) == 0) { + _update(ctx, e, FUNCTION_CONSTRUCTOR); + *ps = p; + return 0; + } + + if (ctx->parsing_builtin && _parse_id(ctx, ctx->dict.__operator, &p) == 0) { + _update(ctx, e, FUNCTION_OPERATOR); + if (_parse_overriden_operator(ctx, &p) == 0) { + *ps = p; + return 0; + } + return -1; + } + + if (_parse_identifier(ctx, &p) == 0) { + _update(ctx, e, FUNCTION_ORDINARY); + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_invariant_qualifier(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_id(ctx, ctx->dict.invariant, ps)) { + return -1; + } + _emit(ctx, &ps->out, TYPE_INVARIANT); + return 0; +} + + +static int +_parse_centroid_qualifier(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_id(ctx, ctx->dict.centroid, ps)) { + return -1; + } + _emit(ctx, &ps->out, TYPE_CENTROID); + return 0; +} + + +static int +_parse_type_qualifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + const struct sl_pp_token_info *input = _fetch_token(ctx, p.in); + unsigned int e = _emit(ctx, &p.out, 0); + int id; + + if (!input || input->token != SL_PP_IDENTIFIER) { + return -1; + } + id = input->data.identifier; + + if (id == ctx->dict._const) { + _update(ctx, e, TYPE_QUALIFIER_CONST); + } else if (ctx->shader_type == 2 && id == ctx->dict.attribute) { + _update(ctx, e, TYPE_QUALIFIER_ATTRIBUTE); + } else if (id == ctx->dict.varying) { + _update(ctx, e, TYPE_QUALIFIER_VARYING); + } else if (id == ctx->dict.uniform) { + _update(ctx, e, TYPE_QUALIFIER_UNIFORM); + } else if (ctx->parsing_builtin && id == ctx->dict.__fixed_output) { + _update(ctx, e, TYPE_QUALIFIER_FIXEDOUTPUT); + } else if (ctx->parsing_builtin && id == ctx->dict.__fixed_input) { + _update(ctx, e, TYPE_QUALIFIER_FIXEDINPUT); + } else { + return -1; + } + _parse_token(ctx, SL_PP_IDENTIFIER, &p); + *ps = p; + return 0; +} + + +static int +_parse_struct_declarator(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e; + + if (_parse_identifier(ctx, &p)) { + return -1; + } + e = _emit(ctx, &p.out, FIELD_NONE); + *ps = p; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p)) { + return 0; + } + if (_parse_constant_expression(ctx, &p)) { + _error(ctx, "expected constant integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + _update(ctx, e, FIELD_ARRAY); + *ps = p; + return 0; +} + + +static int +_parse_struct_declarator_list(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_struct_declarator(ctx, &p)) { + return -1; + } + + for (;;) { + *ps = p; + _emit(ctx, &p.out, FIELD_NEXT); + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + return 0; + } + if (_parse_struct_declarator(ctx, &p)) { + return 0; + } + } +} + + +static int +_parse_struct_declaration(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_type_specifier(ctx, &p)) { + return -1; + } + if (_parse_struct_declarator_list(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + _emit(ctx, &p.out, FIELD_NONE); + *ps = p; + return 0; +} + + +static int +_parse_struct_declaration_list(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_struct_declaration(ctx, &p)) { + return -1; + } + + for (;;) { + *ps = p; + _emit(ctx, &p.out, FIELD_NEXT); + if (_parse_struct_declaration(ctx, &p)) { + return 0; + } + } +} + + +static int +_parse_struct_specifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_id(ctx, ctx->dict._struct, &p)) { + return -1; + } + if (_parse_identifier(ctx, &p)) { + _emit(ctx, &p.out, '\0'); + } + if (_parse_token(ctx, SL_PP_LBRACE, &p)) { + _error(ctx, "expected `{'"); + return -1; + } + if (_parse_struct_declaration_list(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACE, &p)) { + return -1; + } + _emit(ctx, &p.out, FIELD_NONE); + *ps = p; + return 0; +} + + +static int +_parse_type_specifier_nonarray(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, 0); + const struct sl_pp_token_info *input; + int id; + + if (_parse_struct_specifier(ctx, &p) == 0) { + _update(ctx, e, TYPE_SPECIFIER_STRUCT); + *ps = p; + return 0; + } + + input = _fetch_token(ctx, p.in); + if (!input || input->token != SL_PP_IDENTIFIER) { + return -1; + } + id = input->data.identifier; + + if (id == ctx->dict._void) { + _update(ctx, e, TYPE_SPECIFIER_VOID); + } else if (id == ctx->dict._float) { + _update(ctx, e, TYPE_SPECIFIER_FLOAT); + } else if (id == ctx->dict._int) { + _update(ctx, e, TYPE_SPECIFIER_INT); + } else if (id == ctx->dict._bool) { + _update(ctx, e, TYPE_SPECIFIER_BOOL); + } else if (id == ctx->dict.vec2) { + _update(ctx, e, TYPE_SPECIFIER_VEC2); + } else if (id == ctx->dict.vec3) { + _update(ctx, e, TYPE_SPECIFIER_VEC3); + } else if (id == ctx->dict.vec4) { + _update(ctx, e, TYPE_SPECIFIER_VEC4); + } else if (id == ctx->dict.bvec2) { + _update(ctx, e, TYPE_SPECIFIER_BVEC2); + } else if (id == ctx->dict.bvec3) { + _update(ctx, e, TYPE_SPECIFIER_BVEC3); + } else if (id == ctx->dict.bvec4) { + _update(ctx, e, TYPE_SPECIFIER_BVEC4); + } else if (id == ctx->dict.ivec2) { + _update(ctx, e, TYPE_SPECIFIER_IVEC2); + } else if (id == ctx->dict.ivec3) { + _update(ctx, e, TYPE_SPECIFIER_IVEC3); + } else if (id == ctx->dict.ivec4) { + _update(ctx, e, TYPE_SPECIFIER_IVEC4); + } else if (id == ctx->dict.mat2) { + _update(ctx, e, TYPE_SPECIFIER_MAT2); + } else if (id == ctx->dict.mat3) { + _update(ctx, e, TYPE_SPECIFIER_MAT3); + } else if (id == ctx->dict.mat4) { + _update(ctx, e, TYPE_SPECIFIER_MAT4); + } else if (id == ctx->dict.mat2x3) { + _update(ctx, e, TYPE_SPECIFIER_MAT23); + } else if (id == ctx->dict.mat3x2) { + _update(ctx, e, TYPE_SPECIFIER_MAT32); + } else if (id == ctx->dict.mat2x4) { + _update(ctx, e, TYPE_SPECIFIER_MAT24); + } else if (id == ctx->dict.mat4x2) { + _update(ctx, e, TYPE_SPECIFIER_MAT42); + } else if (id == ctx->dict.mat3x4) { + _update(ctx, e, TYPE_SPECIFIER_MAT34); + } else if (id == ctx->dict.mat4x3) { + _update(ctx, e, TYPE_SPECIFIER_MAT43); + } else if (id == ctx->dict.sampler1D) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER1D); + } else if (id == ctx->dict.sampler2D) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER2D); + } else if (id == ctx->dict.sampler3D) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER3D); + } else if (id == ctx->dict.samplerCube) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLERCUBE); + } else if (id == ctx->dict.sampler1DShadow) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER1DSHADOW); + } else if (id == ctx->dict.sampler2DShadow) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER2DSHADOW); + } else if (id == ctx->dict.sampler2DRect) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER2DRECT); + } else if (id == ctx->dict.sampler2DRectShadow) { + _update(ctx, e, TYPE_SPECIFIER_SAMPLER2DRECTSHADOW); + } else if (_parse_identifier(ctx, &p) == 0) { + _update(ctx, e, TYPE_SPECIFIER_TYPENAME); + *ps = p; + return 0; + } else { + return -1; + } + + _parse_token(ctx, SL_PP_IDENTIFIER, &p); + *ps = p; + return 0; +} + + +static int +_parse_type_specifier_array(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p)) { + return -1; + } + if (_parse_constant_expression(ctx, &p)) { + _error(ctx, "expected constant integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_type_specifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e; + + if (_parse_type_specifier_nonarray(ctx, &p)) { + return -1; + } + + e = _emit(ctx, &p.out, TYPE_SPECIFIER_ARRAY); + if (_parse_type_specifier_array(ctx, &p)) { + _update(ctx, e, TYPE_SPECIFIER_NONARRAY); + } + *ps = p; + return 0; +} + + +static int +_parse_fully_specified_type(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_invariant_qualifier(ctx, &p)) { + _emit(ctx, &p.out, TYPE_VARIANT); + } + if (_parse_centroid_qualifier(ctx, &p)) { + _emit(ctx, &p.out, TYPE_CENTER); + } + if (_parse_type_qualifier(ctx, &p)) { + _emit(ctx, &p.out, TYPE_QUALIFIER_NONE); + } + if (_parse_precision(ctx, &p)) { + _emit(ctx, &p.out, PRECISION_DEFAULT); + } + if (_parse_type_specifier(ctx, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_function_header(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_fully_specified_type(ctx, &p)) { + return -1; + } + if (_parse_function_decl_identifier(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_parameter_qualifier(struct parse_context *ctx, + struct parse_state *ps) +{ + unsigned int e = _emit(ctx, &ps->out, PARAM_QUALIFIER_IN); + + if (_parse_id(ctx, ctx->dict.out, ps) == 0) { + _update(ctx, e, PARAM_QUALIFIER_OUT); + } else if (_parse_id(ctx, ctx->dict.inout, ps) == 0) { + _update(ctx, e, PARAM_QUALIFIER_INOUT); + } else { + _parse_id(ctx, ctx->dict.in, ps); + } + return 0; +} + + +static int +_parse_function_identifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + unsigned int e; + + if (_parse_identifier(ctx, ps)) { + return -1; + } + e = _emit(ctx, &ps->out, FUNCTION_CALL_NONARRAY); + + p = *ps; + if (_parse_token(ctx, SL_PP_LBRACKET, &p)) { + return 0; + } + if (_parse_constant_expression(ctx, &p)) { + _error(ctx, "expected constant integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + _update(ctx, e, FUNCTION_CALL_ARRAY); + *ps = p; + return 0; +} + + +static int +_parse_function_call_header(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_identifier(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_assign_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int op; + + if (_parse_unary_expression(ctx, &p)) { + return -1; + } + + if (_parse_token(ctx, SL_PP_ASSIGN, &p) == 0) { + op = OP_ASSIGN; + } else if (_parse_token(ctx, SL_PP_MULASSIGN, &p) == 0) { + op = OP_MULASSIGN; + } else if (_parse_token(ctx, SL_PP_DIVASSIGN, &p) == 0) { + op = OP_DIVASSIGN; + } else if (_parse_token(ctx, SL_PP_ADDASSIGN, &p) == 0) { + op = OP_ADDASSIGN; + } else if (_parse_token(ctx, SL_PP_SUBASSIGN, &p) == 0) { + op = OP_SUBASSIGN; + } else { + return -1; + } + + if (_parse_assignment_expression(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, op); + + *ps = p; + return 0; +} + + +static int +_parse_assignment_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_assign_expression(ctx, ps) == 0) { + return 0; + } + + if (_parse_conditional_expression(ctx, ps) == 0) { + return 0; + } + + return -1; +} + + +static int +_parse_function_call_header_with_parameters(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_call_header(ctx, &p)) { + return -1; + } + if (_parse_assignment_expression(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + return 0; + } + if (_parse_assignment_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_END); + } +} + + +static int +_parse_function_call_header_no_parameters(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_function_call_header(ctx, ps)) { + return -1; + } + _parse_id(ctx, ctx->dict._void, ps); + return 0; +} + + +static int +_parse_function_call_generic(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_call_header_with_parameters(ctx, &p) == 0) { + if (_parse_token(ctx, SL_PP_RPAREN, &p) == 0) { + *ps = p; + return 0; + } + _error(ctx, "expected `)'"); + return -1; + } + + p = *ps; + if (_parse_function_call_header_no_parameters(ctx, &p) == 0) { + if (_parse_token(ctx, SL_PP_RPAREN, &p) == 0) { + *ps = p; + return 0; + } + _error(ctx, "expected `)'"); + return -1; + } + + return -1; +} + + +static int +_parse_method_call(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_METHOD); + if (_parse_identifier(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_DOT, &p)) { + return -1; + } + if (_parse_function_call_generic(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_regular_function_call(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_CALL); + if (_parse_function_call_generic(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_function_call(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_regular_function_call(ctx, ps) == 0) { + return 0; + } + + if (_parse_method_call(ctx, ps) == 0) { + return 0; + } + + return -1; +} + + +static int +_parse_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_assignment_expression(ctx, &p)) { + return -1; + } + + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + return 0; + } + if (_parse_assignment_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_SEQUENCE); + } +} + + +static int +_parse_postfix_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + + if (_parse_function_call(ctx, ps)) { + if (_parse_primary_expression(ctx, ps)) { + return -1; + } + } + + for (p = *ps;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_INCREMENT, &p) == 0) { + _emit(ctx, &p.out, OP_POSTINCREMENT); + } else if (_parse_token(ctx, SL_PP_DECREMENT, &p) == 0) { + _emit(ctx, &p.out, OP_POSTDECREMENT); + } else if (_parse_token(ctx, SL_PP_LBRACKET, &p) == 0) { + if (_parse_expression(ctx, &p)) { + _error(ctx, "expected an integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + _emit(ctx, &p.out, OP_SUBSCRIPT); + } else if (_parse_token(ctx, SL_PP_DOT, &p) == 0) { + _emit(ctx, &p.out, OP_FIELD); + if (_parse_identifier(ctx, &p)) { + return 0; + } + } else { + return 0; + } + } +} + + +static int +_parse_unary_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + unsigned int op; + + if (_parse_postfix_expression(ctx, ps) == 0) { + return 0; + } + + p = *ps; + if (_parse_token(ctx, SL_PP_INCREMENT, &p) == 0) { + op = OP_PREINCREMENT; + } else if (_parse_token(ctx, SL_PP_DECREMENT, &p) == 0) { + op = OP_PREDECREMENT; + } else if (_parse_token(ctx, SL_PP_PLUS, &p) == 0) { + op = OP_PLUS; + } else if (_parse_token(ctx, SL_PP_MINUS, &p) == 0) { + op = OP_MINUS; + } else if (_parse_token(ctx, SL_PP_NOT, &p) == 0) { + op = OP_NOT; + } else { + return -1; + } + + if (_parse_unary_expression(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, op); + *ps = p; + return 0; +} + + +static int +_parse_multiplicative_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_unary_expression(ctx, &p)) { + return -1; + } + for (;;) { + unsigned int op; + + *ps = p; + if (_parse_token(ctx, SL_PP_STAR, &p) == 0) { + op = OP_MULTIPLY; + } else if (_parse_token(ctx, SL_PP_SLASH, &p) == 0) { + op = OP_DIVIDE; + } else { + return 0; + } + if (_parse_unary_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, op); + } +} + + +static int +_parse_additive_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_multiplicative_expression(ctx, &p)) { + return -1; + } + for (;;) { + unsigned int op; + + *ps = p; + if (_parse_token(ctx, SL_PP_PLUS, &p) == 0) { + op = OP_ADD; + } else if (_parse_token(ctx, SL_PP_MINUS, &p) == 0) { + op = OP_SUBTRACT; + } else { + return 0; + } + if (_parse_multiplicative_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, op); + } +} + + +static int +_parse_relational_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_additive_expression(ctx, &p)) { + return -1; + } + for (;;) { + unsigned int op; + + *ps = p; + if (_parse_token(ctx, SL_PP_LESS, &p) == 0) { + op = OP_LESS; + } else if (_parse_token(ctx, SL_PP_GREATER, &p) == 0) { + op = OP_GREATER; + } else if (_parse_token(ctx, SL_PP_LESSEQUAL, &p) == 0) { + op = OP_LESSEQUAL; + } else if (_parse_token(ctx, SL_PP_GREATEREQUAL, &p) == 0) { + op = OP_GREATEREQUAL; + } else { + return 0; + } + if (_parse_additive_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, op); + } +} + + +static int +_parse_equality_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_relational_expression(ctx, &p)) { + return -1; + } + for (;;) { + unsigned int op; + + *ps = p; + if (_parse_token(ctx, SL_PP_EQUAL, &p) == 0) { + op = OP_EQUAL; + } else if (_parse_token(ctx, SL_PP_NOTEQUAL, &p) == 0) { + op = OP_NOTEQUAL; + } else { + return 0; + } + if (_parse_relational_expression(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, op); + } +} + + +static int +_parse_logical_and_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_equality_expression(ctx, &p)) { + return -1; + } + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_AND, &p)) { + return 0; + } + if (_parse_equality_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_LOGICALAND); + } +} + + +static int +_parse_logical_xor_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_logical_and_expression(ctx, &p)) { + return -1; + } + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_XOR, &p)) { + return 0; + } + if (_parse_logical_and_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_LOGICALXOR); + } +} + + +static int +_parse_logical_or_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_logical_xor_expression(ctx, &p)) { + return -1; + } + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_OR, &p)) { + return 0; + } + if (_parse_logical_xor_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_LOGICALOR); + } +} + + +static int +_parse_conditional_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_logical_or_expression(ctx, &p)) { + return -1; + } + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_QUESTION, &p)) { + return 0; + } + if (_parse_expression(ctx, &p)) { + return 0; + } + if (_parse_token(ctx, SL_PP_COLON, &p)) { + return 0; + } + if (_parse_conditional_expression(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_SELECT); + } +} + + +static int +_parse_constant_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_conditional_expression(ctx, ps)) { + return -1; + } + _emit(ctx, &ps->out, OP_END); + return 0; +} + + +static int +_parse_parameter_declarator_array(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p)) { + return -1; + } + if (_parse_constant_expression(ctx, &p)) { + _error(ctx, "expected constant integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_parameter_declarator(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e; + + if (_parse_type_specifier(ctx, &p)) { + return -1; + } + if (_parse_identifier(ctx, &p)) { + return -1; + } + e = _emit(ctx, &p.out, PARAMETER_ARRAY_PRESENT); + if (_parse_parameter_declarator_array(ctx, &p)) { + _update(ctx, e, PARAMETER_ARRAY_NOT_PRESENT); + } + *ps = p; + return 0; +} + + +static int +_parse_parameter_type_specifier_array(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p)) { + return -1; + } + if (_parse_constant_expression(ctx, &p)) { + _error(ctx, "expected constant integral expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p)) { + _error(ctx, "expected `]'"); + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_parameter_type_specifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e; + + if (_parse_type_specifier(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, '\0'); + + e = _emit(ctx, &p.out, PARAMETER_ARRAY_PRESENT); + if (_parse_parameter_type_specifier_array(ctx, &p)) { + _update(ctx, e, PARAMETER_ARRAY_NOT_PRESENT); + } + *ps = p; + return 0; +} + + +static int +_parse_parameter_declaration(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, PARAMETER_NEXT); + + (void) e; + + if (_parse_type_qualifier(ctx, &p)) { + _emit(ctx, &p.out, TYPE_QUALIFIER_NONE); + } + _parse_parameter_qualifier(ctx, &p); + if (_parse_precision(ctx, &p)) { + _emit(ctx, &p.out, PRECISION_DEFAULT); + } + if (_parse_parameter_declarator(ctx, &p) == 0) { + *ps = p; + return 0; + } + if (_parse_parameter_type_specifier(ctx, &p) == 0) { + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_function_header_with_parameters(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_header(ctx, &p)) { + return -1; + } + if (_parse_parameter_declaration(ctx, &p)) { + return -1; + } + + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + return 0; + } + if (_parse_parameter_declaration(ctx, &p)) { + return 0; + } + } +} + + +static int +_parse_function_declarator(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_function_header_with_parameters(ctx, ps) == 0) { + return 0; + } + + if (_parse_function_header(ctx, ps) == 0) { + return 0; + } + + return -1; +} + + +static int +_parse_function_prototype(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_header(ctx, &p) == 0) { + if (_parse_id(ctx, ctx->dict._void, &p) == 0) { + if (_parse_token(ctx, SL_PP_RPAREN, &p) == 0) { + _emit(ctx, &p.out, PARAMETER_NONE); + *ps = p; + return 0; + } + _error(ctx, "expected `)'"); + return -1; + } + } + + p = *ps; + if (_parse_function_declarator(ctx, &p) == 0) { + if (_parse_token(ctx, SL_PP_RPAREN, &p) == 0) { + _emit(ctx, &p.out, PARAMETER_NONE); + *ps = p; + return 0; + } + _error(ctx, "expected `)'"); + return -1; + } + + return -1; +} + + +static int +_parse_precision(struct parse_context *ctx, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + int id; + unsigned int precision; + + if (!input || input->token != SL_PP_IDENTIFIER) { + return -1; + } + id = input->data.identifier; + + if (id == ctx->dict.lowp) { + precision = PRECISION_LOW; + } else if (id == ctx->dict.mediump) { + precision = PRECISION_MEDIUM; + } else if (id == ctx->dict.highp) { + precision = PRECISION_HIGH; + } else { + return -1; + } + + _parse_token(ctx, SL_PP_IDENTIFIER, ps); + _emit(ctx, &ps->out, precision); + return 0; +} + + +static int +_parse_prectype(struct parse_context *ctx, + struct parse_state *ps) +{ + const struct sl_pp_token_info *input = _fetch_token(ctx, ps->in); + int id; + unsigned int type; + + if (!input || input->token != SL_PP_IDENTIFIER) { + return -1; + } + id = input->data.identifier; + + if (id == ctx->dict._int) { + type = TYPE_SPECIFIER_INT; + } else if (id == ctx->dict._float) { + type = TYPE_SPECIFIER_FLOAT; + } else if (id == ctx->dict.sampler1D) { + type = TYPE_SPECIFIER_SAMPLER1D; + } else if (id == ctx->dict.sampler2D) { + type = TYPE_SPECIFIER_SAMPLER2D; + } else if (id == ctx->dict.sampler3D) { + type = TYPE_SPECIFIER_SAMPLER3D; + } else if (id == ctx->dict.samplerCube) { + type = TYPE_SPECIFIER_SAMPLERCUBE; + } else if (id == ctx->dict.sampler1DShadow) { + type = TYPE_SPECIFIER_SAMPLER1DSHADOW; + } else if (id == ctx->dict.sampler2DShadow) { + type = TYPE_SPECIFIER_SAMPLER2DSHADOW; + } else if (id == ctx->dict.sampler2DRect) { + type = TYPE_SPECIFIER_SAMPLER2DRECT; + } else if (id == ctx->dict.sampler2DRectShadow) { + type = TYPE_SPECIFIER_SAMPLER2DRECTSHADOW; + } else { + return -1; + } + + _parse_token(ctx, SL_PP_IDENTIFIER, ps); + _emit(ctx, &ps->out, type); + return 0; +} + + +static int +_parse_precision_stmt(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_id(ctx, ctx->dict.precision, &p)) { + return -1; + } + if (_parse_precision(ctx, &p)) { + return -1; + } + if (_parse_prectype(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_floatconstant(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_PUSH_FLOAT); + if (_parse_float(ctx, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_intconstant(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_PUSH_INT); + if (_parse_uint(ctx, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_boolconstant(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_id(ctx, ctx->dict._false, ps) == 0) { + _emit(ctx, &ps->out, OP_PUSH_BOOL); + _emit(ctx, &ps->out, 2); /* radix */ + _emit(ctx, &ps->out, '0'); + _emit(ctx, &ps->out, '\0'); + return 0; + } + + if (_parse_id(ctx, ctx->dict._true, ps) == 0) { + _emit(ctx, &ps->out, OP_PUSH_BOOL); + _emit(ctx, &ps->out, 2); /* radix */ + _emit(ctx, &ps->out, '1'); + _emit(ctx, &ps->out, '\0'); + return 0; + } + + return -1; +} + + +static int +_parse_variable_identifier(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_PUSH_IDENTIFIER); + if (_parse_identifier(ctx, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_primary_expression(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + + if (_parse_floatconstant(ctx, ps) == 0) { + return 0; + } + if (_parse_boolconstant(ctx, ps) == 0) { + return 0; + } + if (_parse_intconstant(ctx, ps) == 0) { + return 0; + } + if (_parse_variable_identifier(ctx, ps) == 0) { + return 0; + } + + p = *ps; + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + return -1; + } + if (_parse_expression(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_RPAREN, &p)) { + return -1; + } + + *ps = p; + return 0; +} + + +static int +_parse_asm_argument(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_variable_identifier(ctx, ps) == 0) { + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_DOT, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_FIELD); + if (_parse_identifier(ctx, &p)) { + return 0; + } + *ps = p; + return 0; + } + + if (_parse_floatconstant(ctx, ps) == 0) { + return 0; + } + + return -1; +} + + +static int +_parse_asm_arguments(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_asm_argument(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + + for (;;) { + *ps = p; + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + return 0; + } + if (_parse_asm_argument(ctx, &p)) { + return 0; + } + _emit(ctx, &p.out, OP_END); + } +} + + +static int +_parse_asm_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_id(ctx, ctx->dict.___asm, &p)) { + return -1; + } + if (_parse_identifier(ctx, &p)) { + return -1; + } + if (_parse_asm_arguments(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_selection_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_IF); + if (_parse_id(ctx, ctx->dict._if, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + _error(ctx, "expected `('"); + return -1; + } + if (_parse_expression(ctx, &p)) { + _error(ctx, "expected an expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RPAREN, &p)) { + _error(ctx, "expected `)'"); + return -1; + } + _emit(ctx, &p.out, OP_END); + if (_parse_statement(ctx, &p)) { + return -1; + } + + *ps = p; + if (_parse_id(ctx, ctx->dict._else, &p) == 0) { + if (_parse_statement(ctx, &p) == 0) { + *ps = p; + return 0; + } + } + + _emit(ctx, &ps->out, OP_EXPRESSION); + _emit(ctx, &ps->out, OP_PUSH_VOID); + _emit(ctx, &ps->out, OP_END); + return 0; +} + + +static int +_parse_expression_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_expression(ctx, &p)) { + _emit(ctx, &p.out, OP_PUSH_VOID); + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_for_init_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, OP_EXPRESSION); + + if (_parse_expression_statement(ctx, &p) == 0) { + *ps = p; + return 0; + } + + if (_parse_declaration(ctx, &p) == 0) { + _update(ctx, e, OP_DECLARE); + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_initializer(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_assignment_expression(ctx, ps) == 0) { + _emit(ctx, &ps->out, OP_END); + return 0; + } + return -1; +} + + +static int +_parse_condition_initializer(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + _emit(ctx, &p.out, OP_DECLARE); + _emit(ctx, &p.out, DECLARATION_INIT_DECLARATOR_LIST); + if (_parse_fully_specified_type(ctx, &p)) { + return -1; + } + _emit(ctx, &p.out, VARIABLE_IDENTIFIER); + if (_parse_identifier(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_ASSIGN, &p)) { + _error(ctx, "expected `='"); + return -1; + } + _emit(ctx, &p.out, VARIABLE_INITIALIZER); + if (_parse_initializer(ctx, &p)) { + _error(ctx, "expected an initialiser"); + return -1; + } + _emit(ctx, &p.out, DECLARATOR_NONE); + *ps = p; + return 0; +} + + +static int +_parse_condition(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + + if (_parse_condition_initializer(ctx, ps) == 0) { + return 0; + } + + p = *ps; + _emit(ctx, &p.out, OP_EXPRESSION); + if (_parse_expression(ctx, &p) == 0) { + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_for_rest_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_condition(ctx, &p)) { + _emit(ctx, &p.out, OP_EXPRESSION); + _emit(ctx, &p.out, OP_PUSH_BOOL); + _emit(ctx, &p.out, 2); + _emit(ctx, &p.out, '1'); + _emit(ctx, &p.out, '\0'); + _emit(ctx, &p.out, OP_END); + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + if (_parse_expression(ctx, &p)) { + _emit(ctx, &p.out, OP_PUSH_VOID); + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_iteration_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_id(ctx, ctx->dict._while, &p) == 0) { + _emit(ctx, &p.out, OP_WHILE); + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + _error(ctx, "expected `('"); + return -1; + } + if (_parse_condition(ctx, &p)) { + _error(ctx, "expected an expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RPAREN, &p)) { + _error(ctx, "expected `)'"); + return -1; + } + if (_parse_statement(ctx, &p)) { + return -1; + } + *ps = p; + return 0; + } + + if (_parse_id(ctx, ctx->dict._do, &p) == 0) { + _emit(ctx, &p.out, OP_DO); + if (_parse_statement(ctx, &p)) { + return -1; + } + if (_parse_id(ctx, ctx->dict._while, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + _error(ctx, "expected `('"); + return -1; + } + if (_parse_expression(ctx, &p)) { + _error(ctx, "expected an expression"); + return -1; + } + if (_parse_token(ctx, SL_PP_RPAREN, &p)) { + _error(ctx, "expected `)'"); + return -1; + } + _emit(ctx, &p.out, OP_END); + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + _error(ctx, "expected `;'"); + return -1; + } + *ps = p; + return 0; + } + + if (_parse_id(ctx, ctx->dict._for, &p) == 0) { + _emit(ctx, &p.out, OP_FOR); + if (_parse_token(ctx, SL_PP_LPAREN, &p)) { + _error(ctx, "expected `('"); + return -1; + } + if (_parse_for_init_statement(ctx, &p)) { + return -1; + } + if (_parse_for_rest_statement(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_RPAREN, &p)) { + _error(ctx, "expected `)'"); + return -1; + } + if (_parse_statement(ctx, &p)) { + return -1; + } + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_jump_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, 0); + + if (_parse_id(ctx, ctx->dict._continue, &p) == 0) { + _update(ctx, e, OP_CONTINUE); + } else if (_parse_id(ctx, ctx->dict._break, &p) == 0) { + _update(ctx, e, OP_BREAK); + } else if (_parse_id(ctx, ctx->dict._return, &p) == 0) { + _update(ctx, e, OP_RETURN); + if (_parse_expression(ctx, &p)) { + _emit(ctx, &p.out, OP_PUSH_VOID); + } + _emit(ctx, &p.out, OP_END); + } else if (ctx->shader_type == 1 && _parse_id(ctx, ctx->dict.discard, &p) == 0) { + _update(ctx, e, OP_DISCARD); + } else { + return -1; + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_simple_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p; + unsigned int e; + + if (_parse_selection_statement(ctx, ps) == 0) { + return 0; + } + + if (_parse_iteration_statement(ctx, ps) == 0) { + return 0; + } + + if (_parse_jump_statement(ctx, ps) == 0) { + return 0; + } + + p = *ps; + e = _emit(ctx, &p.out, OP_EXPRESSION); + if (_parse_expression_statement(ctx, &p) == 0) { + *ps = p; + return 0; + } + + if (_parse_precision_stmt(ctx, &p) == 0) { + _update(ctx, e, OP_PRECISION); + *ps = p; + return 0; + } + + if (ctx->parsing_builtin && _parse_asm_statement(ctx, &p) == 0) { + _update(ctx, e, OP_ASM); + *ps = p; + return 0; + } + + if (_parse_declaration(ctx, &p) == 0) { + _update(ctx, e, OP_DECLARE); + *ps = p; + return 0; + } + + return -1; +} + + +static int +_parse_compound_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACE, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_BLOCK_BEGIN_NEW_SCOPE); + _parse_statement_list(ctx, &p); + if (_parse_token(ctx, SL_PP_RBRACE, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_statement(struct parse_context *ctx, + struct parse_state *ps) +{ + if (_parse_compound_statement(ctx, ps) == 0) { + return 0; + } + + if (_parse_simple_statement(ctx, ps) == 0) { + return 0; + } + + return -1; +} + + +static int +_parse_statement_list(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_statement(ctx, &p)) { + return -1; + } + + for (;;) { + *ps = p; + if (_parse_statement(ctx, &p)) { + return 0; + } + } +} + + +static int +_parse_compound_statement_no_new_scope(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACE, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_BLOCK_BEGIN_NO_NEW_SCOPE); + _parse_statement_list(ctx, &p); + if (_parse_token(ctx, SL_PP_RBRACE, &p)) { + return -1; + } + _emit(ctx, &p.out, OP_END); + *ps = p; + return 0; +} + + +static int +_parse_function_definition(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_function_prototype(ctx, &p)) { + return -1; + } + if (_parse_compound_statement_no_new_scope(ctx, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_invariant_stmt(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_id(ctx, ctx->dict.invariant, &p)) { + return -1; + } + if (_parse_identifier(ctx, &p)) { + return -1; + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_single_declaration(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e; + + if (_parse_fully_specified_type(ctx, &p)) { + return -1; + } + + e = _emit(ctx, &p.out, VARIABLE_IDENTIFIER); + if (_parse_identifier(ctx, &p)) { + _update(ctx, e, VARIABLE_NONE); + *ps = p; + return 0; + } + + e = _emit(ctx, &p.out, VARIABLE_NONE); + *ps = p; + + if (_parse_token(ctx, SL_PP_ASSIGN, &p) == 0) { + _update(ctx, e, VARIABLE_INITIALIZER); + if (_parse_initializer(ctx, &p) == 0) { + *ps = p; + return 0; + } + _error(ctx, "expected an initialiser"); + return -1; + } + p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p) == 0) { + if (_parse_constant_expression(ctx, &p)) { + _update(ctx, e, VARIABLE_ARRAY_UNKNOWN); + } else { + _update(ctx, e, VARIABLE_ARRAY_EXPLICIT); + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p) == 0) { + *ps = p; + return 0; + } + _error(ctx, "expected `]'"); + return -1; + } + return 0; +} + + +static int +_parse_init_declarator_list(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + + if (_parse_single_declaration(ctx, &p)) { + return -1; + } + + for (;;) { + unsigned int e; + + *ps = p; + if (_parse_token(ctx, SL_PP_COMMA, &p)) { + break; + } + _emit(ctx, &p.out, DECLARATOR_NEXT); + _emit(ctx, &p.out, VARIABLE_IDENTIFIER); + if (_parse_identifier(ctx, &p)) { + break; + } + + e = _emit(ctx, &p.out, VARIABLE_NONE); + *ps = p; + + if (_parse_token(ctx, SL_PP_ASSIGN, &p) == 0) { + if (_parse_initializer(ctx, &p) == 0) { + _update(ctx, e, VARIABLE_INITIALIZER); + *ps = p; + continue; + } + _error(ctx, "expected an initialiser"); + break; + } + p = *ps; + + if (_parse_token(ctx, SL_PP_LBRACKET, &p) == 0) { + unsigned int arr; + + if (_parse_constant_expression(ctx, &p)) { + arr = VARIABLE_ARRAY_UNKNOWN; + } else { + arr = VARIABLE_ARRAY_EXPLICIT; + } + if (_parse_token(ctx, SL_PP_RBRACKET, &p) == 0) { + _update(ctx, e, arr); + *ps = p; + continue; + } + _error(ctx, "expected `]'"); + break; + } + p = *ps; + } + + _emit(ctx, &ps->out, DECLARATOR_NONE); + return 0; +} + + +static int +_parse_declaration(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, DECLARATION_FUNCTION_PROTOTYPE); + + if (_parse_function_prototype(ctx, &p)) { + if (_parse_init_declarator_list(ctx, &p)) { + return -1; + } + _update(ctx, e, DECLARATION_INIT_DECLARATOR_LIST); + } + if (_parse_token(ctx, SL_PP_SEMICOLON, &p)) { + _error(ctx, "expected `;'"); + return -1; + } + *ps = p; + return 0; +} + + +static int +_parse_external_declaration(struct parse_context *ctx, + struct parse_state *ps) +{ + struct parse_state p = *ps; + unsigned int e = _emit(ctx, &p.out, 0); + + if (_parse_precision_stmt(ctx, &p) == 0) { + _update(ctx, e, DEFAULT_PRECISION); + *ps = p; + return 0; + } + + if (_parse_function_definition(ctx, &p) == 0) { + _update(ctx, e, EXTERNAL_FUNCTION_DEFINITION); + *ps = p; + return 0; + } + + if (_parse_invariant_stmt(ctx, &p) == 0) { + _update(ctx, e, INVARIANT_STMT); + *ps = p; + return 0; + } + + if (_parse_declaration(ctx, &p) == 0) { + _update(ctx, e, EXTERNAL_DECLARATION); + *ps = p; + return 0; + } + + _error(ctx, "expected an identifier"); + return -1; +} + + +static int +_parse_translation_unit(struct parse_context *ctx, + struct parse_state *ps) +{ + _emit(ctx, &ps->out, REVISION); + if (_parse_external_declaration(ctx, ps)) { + return -1; + } + while (_parse_external_declaration(ctx, ps) == 0) { + } + _emit(ctx, &ps->out, EXTERNAL_NULL); + if (_parse_token(ctx, SL_PP_EOF, ps)) { + return -1; + } + return 0; +} + + +#define ADD_NAME_STR(CTX, NAME, STR)\ + do {\ + (CTX).dict.NAME = sl_pp_context_add_unique_str((CTX).context, (STR));\ + if ((CTX).dict.NAME == -1) {\ + return -1;\ + }\ + } while (0) + +#define ADD_NAME(CTX, NAME) ADD_NAME_STR(CTX, NAME, #NAME) + + +int +sl_cl_compile(struct sl_pp_context *context, + unsigned int shader_type, + unsigned int parsing_builtin, + unsigned char **output, + unsigned int *cboutput, + char *error, + unsigned int cberror) +{ + struct parse_context ctx; + struct parse_state ps; + + ctx.context = context; + + ADD_NAME_STR(ctx, _void, "void"); + ADD_NAME_STR(ctx, _float, "float"); + ADD_NAME_STR(ctx, _int, "int"); + ADD_NAME_STR(ctx, _bool, "bool"); + ADD_NAME(ctx, vec2); + ADD_NAME(ctx, vec3); + ADD_NAME(ctx, vec4); + ADD_NAME(ctx, bvec2); + ADD_NAME(ctx, bvec3); + ADD_NAME(ctx, bvec4); + ADD_NAME(ctx, ivec2); + ADD_NAME(ctx, ivec3); + ADD_NAME(ctx, ivec4); + ADD_NAME(ctx, mat2); + ADD_NAME(ctx, mat3); + ADD_NAME(ctx, mat4); + ADD_NAME(ctx, mat2x3); + ADD_NAME(ctx, mat3x2); + ADD_NAME(ctx, mat2x4); + ADD_NAME(ctx, mat4x2); + ADD_NAME(ctx, mat3x4); + ADD_NAME(ctx, mat4x3); + ADD_NAME(ctx, sampler1D); + ADD_NAME(ctx, sampler2D); + ADD_NAME(ctx, sampler3D); + ADD_NAME(ctx, samplerCube); + ADD_NAME(ctx, sampler1DShadow); + ADD_NAME(ctx, sampler2DShadow); + ADD_NAME(ctx, sampler2DRect); + ADD_NAME(ctx, sampler2DRectShadow); + + ADD_NAME(ctx, invariant); + + ADD_NAME(ctx, centroid); + + ADD_NAME(ctx, precision); + ADD_NAME(ctx, lowp); + ADD_NAME(ctx, mediump); + ADD_NAME(ctx, highp); + + ADD_NAME_STR(ctx, _const, "const"); + ADD_NAME(ctx, attribute); + ADD_NAME(ctx, varying); + ADD_NAME(ctx, uniform); + ADD_NAME(ctx, __fixed_output); + ADD_NAME(ctx, __fixed_input); + + ADD_NAME(ctx, in); + ADD_NAME(ctx, out); + ADD_NAME(ctx, inout); + + ADD_NAME_STR(ctx, _struct, "struct"); + + ADD_NAME(ctx, __constructor); + ADD_NAME(ctx, __operator); + ADD_NAME_STR(ctx, ___asm, "__asm"); + + ADD_NAME_STR(ctx, _if, "if"); + ADD_NAME_STR(ctx, _else, "else"); + ADD_NAME_STR(ctx, _for, "for"); + ADD_NAME_STR(ctx, _while, "while"); + ADD_NAME_STR(ctx, _do, "do"); + + ADD_NAME_STR(ctx, _continue, "continue"); + ADD_NAME_STR(ctx, _break, "break"); + ADD_NAME_STR(ctx, _return, "return"); + ADD_NAME(ctx, discard); + + ADD_NAME_STR(ctx, _false, "false"); + ADD_NAME_STR(ctx, _true, "true"); + + ctx.out_buf = NULL; + ctx.out_cap = 0; + + ctx.shader_type = shader_type; + ctx.parsing_builtin = 1; + + ctx.error[0] = '\0'; + ctx.process_error = 0; + + ctx.tokens_cap = 1024; + ctx.tokens_read = 0; + ctx.tokens = malloc(ctx.tokens_cap * sizeof(struct sl_pp_token_info)); + if (!ctx.tokens) { + strncpy(error, "out of memory", cberror); + return -1; + } + + ps.in = 0; + ps.out = 0; + + if (_parse_translation_unit(&ctx, &ps)) { + strncpy(error, ctx.error, cberror); + free(ctx.tokens); + return -1; + } + + *output = ctx.out_buf; + *cboutput = ps.out; + free(ctx.tokens); + return 0; +} diff --git a/src/glsl/cl/sl_cl_parse.h b/src/glsl/cl/sl_cl_parse.h new file mode 100644 index 00000000000..dd5791d5901 --- /dev/null +++ b/src/glsl/cl/sl_cl_parse.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_CL_PARSE_H +#define SL_CL_PARSE_H + +int +sl_cl_compile(struct sl_pp_context *context, + unsigned int shader_type, + unsigned int parsing_builtin, + unsigned char **output, + unsigned int *cboutput, + char *error, + unsigned int cberror); + +#endif /* SL_CL_PARSE_H */ diff --git a/src/glsl/pp/Makefile b/src/glsl/pp/Makefile new file mode 100644 index 00000000000..819079f6258 --- /dev/null +++ b/src/glsl/pp/Makefile @@ -0,0 +1,26 @@ +#src/glsl/pp/Makefile + +TOP = ../../.. + +include $(TOP)/configs/current + +LIBNAME = glslpp + +C_SOURCES = \ + sl_pp_context.c \ + sl_pp_define.c \ + sl_pp_dict.c \ + sl_pp_error.c \ + sl_pp_expression.c \ + sl_pp_extension.c \ + sl_pp_if.c \ + sl_pp_line.c \ + sl_pp_macro.c \ + sl_pp_pragma.c \ + sl_pp_process.c \ + sl_pp_purify.c \ + sl_pp_token.c \ + sl_pp_version.c + +include ../Makefile.template + diff --git a/src/glsl/pp/sl_pp_context.c b/src/glsl/pp/sl_pp_context.c new file mode 100644 index 00000000000..74a9bdddfdc --- /dev/null +++ b/src/glsl/pp/sl_pp_context.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_public.h" +#include "sl_pp_context.h" + + +struct sl_pp_context * +sl_pp_context_create(const char *input, + const struct sl_pp_purify_options *options) +{ + struct sl_pp_context *context; + + context = calloc(1, sizeof(struct sl_pp_context)); + if (!context) { + return NULL; + } + + if (sl_pp_dict_init(context)) { + sl_pp_context_destroy(context); + return NULL; + } + + context->getc_buf_capacity = 64; + context->getc_buf = malloc(context->getc_buf_capacity * sizeof(char)); + if (!context->getc_buf) { + sl_pp_context_destroy(context); + return NULL; + } + + if (sl_pp_token_buffer_init(&context->tokens, context)) { + sl_pp_context_destroy(context); + return NULL; + } + + context->macro_tail = &context->macro; + context->if_ptr = SL_PP_MAX_IF_NESTING; + context->if_value = 1; + memset(context->error_msg, 0, sizeof(context->error_msg)); + context->error_line = 1; + context->line = 1; + context->file = 0; + + sl_pp_purify_state_init(&context->pure, input, options); + + memset(&context->process_state, 0, sizeof(context->process_state)); + + return context; +} + +void +sl_pp_context_destroy(struct sl_pp_context *context) +{ + if (context) { + free(context->cstr_pool); + sl_pp_macro_free(context->macro); + free(context->getc_buf); + sl_pp_token_buffer_destroy(&context->tokens); + free(context->process_state.out); + free(context); + } +} + +const char * +sl_pp_context_error_message(const struct sl_pp_context *context) +{ + return context->error_msg; +} + +void +sl_pp_context_error_position(const struct sl_pp_context *context, + unsigned int *file, + unsigned int *line) +{ + if (file) { + *file = 0; + } + if (line) { + *line = context->error_line; + } +} + +int +sl_pp_context_add_predefined(struct sl_pp_context *context, + const char *name, + const char *value) +{ + struct sl_pp_predefined pre; + + if (context->num_predefined == SL_PP_MAX_PREDEFINED) { + return -1; + } + + pre.name = sl_pp_context_add_unique_str(context, name); + if (pre.name == -1) { + return -1; + } + + pre.value = sl_pp_context_add_unique_str(context, value); + if (pre.value == -1) { + return -1; + } + + context->predefined[context->num_predefined++] = pre; + return 0; +} + +int +sl_pp_context_add_unique_str(struct sl_pp_context *context, + const char *str) +{ + unsigned int size; + unsigned int offset = 0; + + size = strlen(str) + 1; + + /* Find out if this is a unique string. */ + while (offset < context->cstr_pool_len) { + const char *str2; + unsigned int size2; + + str2 = &context->cstr_pool[offset]; + size2 = strlen(str2) + 1; + if (size == size2 && !memcmp(str, str2, size - 1)) { + return offset; + } + + offset += size2; + } + + if (context->cstr_pool_len + size > context->cstr_pool_max) { + context->cstr_pool_max = (context->cstr_pool_len + size + 0xffff) & ~0xffff; + context->cstr_pool = realloc(context->cstr_pool, context->cstr_pool_max); + } + + if (!context->cstr_pool) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + offset = context->cstr_pool_len; + memcpy(&context->cstr_pool[offset], str, size); + context->cstr_pool_len += size; + + return offset; +} + +const char * +sl_pp_context_cstr(const struct sl_pp_context *context, + int offset) +{ + if (offset == -1) { + return NULL; + } + return &context->cstr_pool[offset]; +} diff --git a/src/glsl/pp/sl_pp_context.h b/src/glsl/pp/sl_pp_context.h new file mode 100644 index 00000000000..3eada380cd1 --- /dev/null +++ b/src/glsl/pp/sl_pp_context.h @@ -0,0 +1,92 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_CONTEXT_H +#define SL_PP_CONTEXT_H + +#include "sl_pp_dict.h" +#include "sl_pp_macro.h" +#include "sl_pp_process.h" +#include "sl_pp_purify.h" +#include "sl_pp_token_util.h" + + +#define SL_PP_MAX_IF_NESTING 64 + +#define SL_PP_MAX_ERROR_MSG 1024 + +#define SL_PP_MAX_EXTENSIONS 16 + +#define SL_PP_MAX_PREDEFINED 16 + +struct sl_pp_extension { + int name; /*< VENDOR_extension_name */ + int name_string; /*< GL_VENDOR_extension_name */ +}; + +struct sl_pp_predefined { + int name; + int value; +}; + +struct sl_pp_context { + char *cstr_pool; + unsigned int cstr_pool_max; + unsigned int cstr_pool_len; + struct sl_pp_dict dict; + + struct sl_pp_macro *macro; + struct sl_pp_macro **macro_tail; + + struct sl_pp_extension extensions[SL_PP_MAX_EXTENSIONS]; + unsigned int num_extensions; + + struct sl_pp_predefined predefined[SL_PP_MAX_PREDEFINED]; + unsigned int num_predefined; + + unsigned int if_stack[SL_PP_MAX_IF_NESTING]; + unsigned int if_ptr; + unsigned int if_value; + + char error_msg[SL_PP_MAX_ERROR_MSG]; + unsigned int error_line; + + unsigned int line; + unsigned int file; + + struct sl_pp_purify_state pure; + + char *getc_buf; + unsigned int getc_buf_size; + unsigned int getc_buf_capacity; + + struct sl_pp_token_buffer tokens; + + struct sl_pp_process_state process_state; +}; + +#endif /* SL_PP_CONTEXT_H */ diff --git a/src/glsl/pp/sl_pp_define.c b/src/glsl/pp/sl_pp_define.c new file mode 100644 index 00000000000..808a6a0d4f1 --- /dev/null +++ b/src/glsl/pp/sl_pp_define.c @@ -0,0 +1,238 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_process.h" +#include "sl_pp_public.h" + + +static void +skip_whitespace(const struct sl_pp_token_info *input, + unsigned int *first, + unsigned int last) +{ + while (*first < last && input[*first].token == SL_PP_WHITESPACE) { + (*first)++; + } +} + + +static int +_parse_formal_args(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int *first, + unsigned int last, + struct sl_pp_macro *macro) +{ + struct sl_pp_macro_formal_arg **arg; + + macro->num_args = 0; + + skip_whitespace(input, first, last); + if (*first < last) { + if (input[*first].token == SL_PP_RPAREN) { + (*first)++; + return 0; + } + } else { + strcpy(context->error_msg, "expected either macro formal argument or `)'"); + return -1; + } + + arg = ¯o->arg; + + for (;;) { + if (*first < last && input[*first].token != SL_PP_IDENTIFIER) { + strcpy(context->error_msg, "expected macro formal argument"); + return -1; + } + + *arg = malloc(sizeof(struct sl_pp_macro_formal_arg)); + if (!*arg) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + (**arg).name = input[*first].data.identifier; + (*first)++; + + (**arg).next = NULL; + arg = &(**arg).next; + + macro->num_args++; + + skip_whitespace(input, first, last); + if (*first < last) { + if (input[*first].token == SL_PP_COMMA) { + (*first)++; + skip_whitespace(input, first, last); + } else if (input[*first].token == SL_PP_RPAREN) { + (*first)++; + return 0; + } else { + strcpy(context->error_msg, "expected either `,' or `)'"); + return -1; + } + } else { + strcpy(context->error_msg, "expected either `,' or `)'"); + return -1; + } + } + + /* Should not gete here. */ +} + + +int +sl_pp_process_define(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + int macro_name = -1; + struct sl_pp_macro *macro; + unsigned int i; + unsigned int body_len; + unsigned int j; + + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + macro_name = input[first].data.identifier; + first++; + } + if (macro_name == -1) { + strcpy(context->error_msg, "expected macro name"); + return -1; + } + + /* Check for reserved macro names */ + { + const char *name = sl_pp_context_cstr(context, macro_name); + + if (strstr(name, "__")) { + strcpy(context->error_msg, "macro names containing `__' are reserved"); + return 1; + } + if (name[0] == 'G' && name[1] == 'L' && name[2] == '_') { + strcpy(context->error_msg, "macro names prefixed with `GL_' are reserved"); + return 1; + } + } + + for (macro = context->macro; macro; macro = macro->next) { + if (macro->name == macro_name) { + break; + } + } + + if (!macro) { + macro = sl_pp_macro_new(); + if (!macro) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + *context->macro_tail = macro; + context->macro_tail = ¯o->next; + } else { + sl_pp_macro_reset(macro); + } + + macro->name = macro_name; + + /* + * If there is no whitespace between macro name and left paren, a macro + * formal argument list follows. This is the only place where the presence + * of a whitespace matters and it's the only reason why we are dealing + * with whitespace at this level. + */ + if (first < last && input[first].token == SL_PP_LPAREN) { + first++; + if (_parse_formal_args(context, input, &first, last, macro)) { + return -1; + } + } + + /* Calculate body size, trim out whitespace, make room for EOF. */ + body_len = 1; + for (i = first; i < last; i++) { + if (input[i].token != SL_PP_WHITESPACE) { + body_len++; + } + } + + macro->body = malloc(sizeof(struct sl_pp_token_info) * body_len); + if (!macro->body) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + for (j = 0, i = first; i < last; i++) { + if (input[i].token != SL_PP_WHITESPACE) { + macro->body[j++] = input[i]; + } + } + macro->body[j++].token = SL_PP_EOF; + + return 0; +} + + +int +sl_pp_process_undef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + int macro_name = -1; + struct sl_pp_macro **pmacro; + struct sl_pp_macro *macro; + + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + macro_name = input[first].data.identifier; + } + if (macro_name == -1) { + return 0; + } + + for (pmacro = &context->macro; *pmacro; pmacro = &(**pmacro).next) { + if ((**pmacro).name == macro_name) { + break; + } + } + if (!*pmacro) { + return 0; + } + + macro = *pmacro; + *pmacro = macro->next; + macro->next = NULL; + sl_pp_macro_free(macro); + + return 0; +} diff --git a/src/glsl/pp/sl_pp_dict.c b/src/glsl/pp/sl_pp_dict.c new file mode 100644 index 00000000000..062139e6ac0 --- /dev/null +++ b/src/glsl/pp/sl_pp_dict.c @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "sl_pp_public.h" +#include "sl_pp_context.h" +#include "sl_pp_dict.h" + + +#define ADD_NAME_STR(CTX, NAME, STR)\ + do {\ + (CTX)->dict.NAME = sl_pp_context_add_unique_str((CTX), (STR));\ + if ((CTX)->dict.NAME == -1) {\ + return -1;\ + }\ + } while (0) + +#define ADD_NAME(CTX, NAME) ADD_NAME_STR(CTX, NAME, #NAME) + + +int +sl_pp_dict_init(struct sl_pp_context *context) +{ + ADD_NAME(context, all); + + ADD_NAME(context, require); + ADD_NAME(context, enable); + ADD_NAME(context, warn); + ADD_NAME(context, disable); + + ADD_NAME(context, defined); + + ADD_NAME_STR(context, ___LINE__, "__LINE__"); + ADD_NAME_STR(context, ___FILE__, "__FILE__"); + ADD_NAME_STR(context, ___VERSION__, "__VERSION__"); + + ADD_NAME(context, optimize); + ADD_NAME(context, debug); + + ADD_NAME(context, off); + ADD_NAME(context, on); + + ADD_NAME(context, define); + ADD_NAME(context, elif); + ADD_NAME_STR(context, _else, "else"); + ADD_NAME(context, endif); + ADD_NAME(context, error); + ADD_NAME(context, extension); + ADD_NAME_STR(context, _if, "if"); + ADD_NAME(context, ifdef); + ADD_NAME(context, ifndef); + ADD_NAME(context, line); + ADD_NAME(context, pragma); + ADD_NAME(context, undef); + + ADD_NAME(context, version); + + ADD_NAME_STR(context, _0, "0"); + ADD_NAME_STR(context, _1, "1"); + + return 0; +} diff --git a/src/glsl/pp/sl_pp_dict.h b/src/glsl/pp/sl_pp_dict.h new file mode 100644 index 00000000000..875217bd309 --- /dev/null +++ b/src/glsl/pp/sl_pp_dict.h @@ -0,0 +1,77 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_DICT_H +#define SL_PP_DICT_H + + +struct sl_pp_context; + +struct sl_pp_dict { + int all; + + int require; + int enable; + int warn; + int disable; + + int defined; + + int ___LINE__; + int ___FILE__; + int ___VERSION__; + + int optimize; + int debug; + + int off; + int on; + + int define; + int elif; + int _else; + int endif; + int error; + int extension; + int _if; + int ifdef; + int ifndef; + int line; + int pragma; + int undef; + + int version; + + int _0; + int _1; +}; + + +int +sl_pp_dict_init(struct sl_pp_context *context); + +#endif /* SL_PP_DICT_H */ diff --git a/src/glsl/pp/sl_pp_error.c b/src/glsl/pp/sl_pp_error.c new file mode 100644 index 00000000000..b628e37ce83 --- /dev/null +++ b/src/glsl/pp/sl_pp_error.c @@ -0,0 +1,270 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_process.h" +#include "sl_pp_public.h" + + +void +sl_pp_process_error(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + unsigned int out_len = 0; + unsigned int i; + + for (i = first; i < last; i++) { + const char *s = NULL; + char buf[2]; + + switch (input[i].token) { + case SL_PP_WHITESPACE: + s = " "; + break; + + case SL_PP_NEWLINE: + s = "\n"; + break; + + case SL_PP_HASH: + s = "#"; + break; + + case SL_PP_COMMA: + s = ","; + break; + + case SL_PP_SEMICOLON: + s = ";"; + break; + + case SL_PP_LBRACE: + s = "{"; + break; + + case SL_PP_RBRACE: + s = "}"; + break; + + case SL_PP_LPAREN: + s = "("; + break; + + case SL_PP_RPAREN: + s = ")"; + break; + + case SL_PP_LBRACKET: + s = "["; + break; + + case SL_PP_RBRACKET: + s = "]"; + break; + + case SL_PP_DOT: + s = "."; + break; + + case SL_PP_INCREMENT: + s = "++"; + break; + + case SL_PP_ADDASSIGN: + s = "+="; + break; + + case SL_PP_PLUS: + s = "+"; + break; + + case SL_PP_DECREMENT: + s = "--"; + break; + + case SL_PP_SUBASSIGN: + s = "-="; + break; + + case SL_PP_MINUS: + s = "-"; + break; + + case SL_PP_BITNOT: + s = "~"; + break; + + case SL_PP_NOTEQUAL: + s = "!="; + break; + + case SL_PP_NOT: + s = "!"; + break; + + case SL_PP_MULASSIGN: + s = "*="; + break; + + case SL_PP_STAR: + s = "*"; + break; + + case SL_PP_DIVASSIGN: + s = "/="; + break; + + case SL_PP_SLASH: + s = "/"; + break; + + case SL_PP_MODASSIGN: + s = "%="; + break; + + case SL_PP_MODULO: + s = "%"; + break; + + case SL_PP_LSHIFTASSIGN: + s = "<<="; + break; + + case SL_PP_LSHIFT: + s = "<<"; + break; + + case SL_PP_LESSEQUAL: + s = "<="; + break; + + case SL_PP_LESS: + s = "<"; + break; + + case SL_PP_RSHIFTASSIGN: + s = ">>="; + break; + + case SL_PP_RSHIFT: + s = ">>"; + break; + + case SL_PP_GREATEREQUAL: + s = ">="; + break; + + case SL_PP_GREATER: + s = ">"; + break; + + case SL_PP_EQUAL: + s = "=="; + break; + + case SL_PP_ASSIGN: + s = "="; + break; + + case SL_PP_AND: + s = "&&"; + break; + + case SL_PP_BITANDASSIGN: + s = "&="; + break; + + case SL_PP_BITAND: + s = "&"; + break; + + case SL_PP_XOR: + s = "^^"; + break; + + case SL_PP_BITXORASSIGN: + s = "^="; + break; + + case SL_PP_BITXOR: + s = "^"; + break; + + case SL_PP_OR: + s = "||"; + break; + + case SL_PP_BITORASSIGN: + s = "|="; + break; + + case SL_PP_BITOR: + s = "|"; + break; + + case SL_PP_QUESTION: + s = "?"; + break; + + case SL_PP_COLON: + s = ":"; + break; + + case SL_PP_IDENTIFIER: + s = sl_pp_context_cstr(context, input[i].data.identifier); + break; + + case SL_PP_UINT: + s = sl_pp_context_cstr(context, input[i].data._uint); + break; + + case SL_PP_FLOAT: + s = sl_pp_context_cstr(context, input[i].data._float); + break; + + case SL_PP_OTHER: + buf[0] = input[i].data.other; + buf[1] = '\0'; + s = buf; + break; + + default: + strcpy(context->error_msg, "internal error"); + return; + } + + while (*s != '\0' && out_len < sizeof(context->error_msg) - 1) { + context->error_msg[out_len++] = *s++; + } + } + + context->error_msg[out_len] = '\0'; +} diff --git a/src/glsl/pp/sl_pp_expression.c b/src/glsl/pp/sl_pp_expression.c new file mode 100644 index 00000000000..ec904787dd7 --- /dev/null +++ b/src/glsl/pp/sl_pp_expression.c @@ -0,0 +1,411 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_expression.h" +#include "sl_pp_public.h" + + +struct parse_context { + struct sl_pp_context *context; + const struct sl_pp_token_info *input; +}; + +static int +_parse_or(struct parse_context *ctx, + int *result); + +static int +_parse_primary(struct parse_context *ctx, + int *result) +{ + if (ctx->input->token == SL_PP_UINT) { + *result = atoi(sl_pp_context_cstr(ctx->context, ctx->input->data._uint)); + ctx->input++; + } else { + if (ctx->input->token != SL_PP_LPAREN) { + strcpy(ctx->context->error_msg, "expected `('"); + return -1; + } + ctx->input++; + if (_parse_or(ctx, result)) { + return -1; + } + if (ctx->input->token != SL_PP_RPAREN) { + strcpy(ctx->context->error_msg, "expected `)'"); + return -1; + } + ctx->input++; + } + return 0; +} + +static int +_parse_unary(struct parse_context *ctx, + int *result) +{ + if (!_parse_primary(ctx, result)) { + return 0; + } + + switch (ctx->input->token) { + case SL_PP_PLUS: + ctx->input++; + if (_parse_unary(ctx, result)) { + return -1; + } + *result = +*result; + break; + + case SL_PP_MINUS: + ctx->input++; + if (_parse_unary(ctx, result)) { + return -1; + } + *result = -*result; + break; + + case SL_PP_NOT: + ctx->input++; + if (_parse_unary(ctx, result)) { + return -1; + } + *result = !*result; + break; + + case SL_PP_BITNOT: + ctx->input++; + if (_parse_unary(ctx, result)) { + return -1; + } + *result = ~*result; + break; + + default: + return -1; + } + + return 0; +} + +static int +_parse_multiplicative(struct parse_context *ctx, + int *result) +{ + if (_parse_unary(ctx, result)) { + return -1; + } + for (;;) { + int right; + + switch (ctx->input->token) { + case SL_PP_STAR: + ctx->input++; + if (_parse_unary(ctx, &right)) { + return -1; + } + *result = *result * right; + break; + + case SL_PP_SLASH: + ctx->input++; + if (_parse_unary(ctx, &right)) { + return -1; + } + *result = *result / right; + break; + + case SL_PP_MODULO: + ctx->input++; + if (_parse_unary(ctx, &right)) { + return -1; + } + *result = *result % right; + break; + + default: + return 0; + } + } +} + +static int +_parse_additive(struct parse_context *ctx, + int *result) +{ + if (_parse_multiplicative(ctx, result)) { + return -1; + } + for (;;) { + int right; + + switch (ctx->input->token) { + case SL_PP_PLUS: + ctx->input++; + if (_parse_multiplicative(ctx, &right)) { + return -1; + } + *result = *result + right; + break; + + case SL_PP_MINUS: + ctx->input++; + if (_parse_multiplicative(ctx, &right)) { + return -1; + } + *result = *result - right; + break; + + default: + return 0; + } + } +} + +static int +_parse_shift(struct parse_context *ctx, + int *result) +{ + if (_parse_additive(ctx, result)) { + return -1; + } + for (;;) { + int right; + + switch (ctx->input->token) { + case SL_PP_LSHIFT: + ctx->input++; + if (_parse_additive(ctx, &right)) { + return -1; + } + *result = *result << right; + break; + + case SL_PP_RSHIFT: + ctx->input++; + if (_parse_additive(ctx, &right)) { + return -1; + } + *result = *result >> right; + break; + + default: + return 0; + } + } +} + +static int +_parse_relational(struct parse_context *ctx, + int *result) +{ + if (_parse_shift(ctx, result)) { + return -1; + } + for (;;) { + int right; + + switch (ctx->input->token) { + case SL_PP_LESSEQUAL: + ctx->input++; + if (_parse_shift(ctx, &right)) { + return -1; + } + *result = *result <= right; + break; + + case SL_PP_GREATEREQUAL: + ctx->input++; + if (_parse_shift(ctx, &right)) { + return -1; + } + *result = *result >= right; + break; + + case SL_PP_LESS: + ctx->input++; + if (_parse_shift(ctx, &right)) { + return -1; + } + *result = *result < right; + break; + + case SL_PP_GREATER: + ctx->input++; + if (_parse_shift(ctx, &right)) { + return -1; + } + *result = *result > right; + break; + + default: + return 0; + } + } +} + +static int +_parse_equality(struct parse_context *ctx, + int *result) +{ + if (_parse_relational(ctx, result)) { + return -1; + } + for (;;) { + int right; + + switch (ctx->input->token) { + case SL_PP_EQUAL: + ctx->input++; + if (_parse_relational(ctx, &right)) { + return -1; + } + *result = *result == right; + break; + + case SL_PP_NOTEQUAL: + ctx->input++; + if (_parse_relational(ctx, &right)) { + return -1; + } + *result = *result != right; + break; + + default: + return 0; + } + } +} + +static int +_parse_bitand(struct parse_context *ctx, + int *result) +{ + if (_parse_equality(ctx, result)) { + return -1; + } + while (ctx->input->token == SL_PP_BITAND) { + int right; + + ctx->input++; + if (_parse_equality(ctx, &right)) { + return -1; + } + *result = *result & right; + } + return 0; +} + +static int +_parse_xor(struct parse_context *ctx, + int *result) +{ + if (_parse_bitand(ctx, result)) { + return -1; + } + while (ctx->input->token == SL_PP_XOR) { + int right; + + ctx->input++; + if (_parse_bitand(ctx, &right)) { + return -1; + } + *result = *result ^ right; + } + return 0; +} + +static int +_parse_bitor(struct parse_context *ctx, + int *result) +{ + if (_parse_xor(ctx, result)) { + return -1; + } + while (ctx->input->token == SL_PP_BITOR) { + int right; + + ctx->input++; + if (_parse_xor(ctx, &right)) { + return -1; + } + *result = *result | right; + } + return 0; +} + +static int +_parse_and(struct parse_context *ctx, + int *result) +{ + if (_parse_bitor(ctx, result)) { + return -1; + } + while (ctx->input->token == SL_PP_AND) { + int right; + + ctx->input++; + if (_parse_bitor(ctx, &right)) { + return -1; + } + *result = *result && right; + } + return 0; +} + +static int +_parse_or(struct parse_context *ctx, + int *result) +{ + if (_parse_and(ctx, result)) { + return -1; + } + while (ctx->input->token == SL_PP_OR) { + int right; + + ctx->input++; + if (_parse_and(ctx, &right)) { + return -1; + } + *result = *result || right; + } + return 0; +} + +int +sl_pp_execute_expression(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + int *result) +{ + struct parse_context ctx; + + ctx.context = context; + ctx.input = input; + + return _parse_or(&ctx, result); +} diff --git a/src/mesa/state_tracker/st_cb_get.h b/src/glsl/pp/sl_pp_expression.h index 8e9f3e93060..377d5b4cbd9 100644 --- a/src/mesa/state_tracker/st_cb_get.h +++ b/src/glsl/pp/sl_pp_expression.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2009 VMware, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,13 +25,16 @@ * **************************************************************************/ +#ifndef SL_PP_EXPRESSION_H +#define SL_PP_EXPRESSION_H -#ifndef ST_CB_GET_H -#define ST_CB_GET_H +#include "sl_pp_context.h" +#include "sl_pp_token.h" -extern void -st_init_get_functions(struct dd_function_table *functions); +int +sl_pp_execute_expression(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + int *result); - -#endif +#endif /* SL_PP_EXPRESSION_H */ diff --git a/src/glsl/pp/sl_pp_extension.c b/src/glsl/pp/sl_pp_extension.c new file mode 100644 index 00000000000..8af5731e840 --- /dev/null +++ b/src/glsl/pp/sl_pp_extension.c @@ -0,0 +1,171 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_process.h" +#include "sl_pp_public.h" + + +int +sl_pp_context_add_extension(struct sl_pp_context *context, + const char *name, + const char *name_string) +{ + struct sl_pp_extension ext; + + if (context->num_extensions == SL_PP_MAX_EXTENSIONS) { + return -1; + } + + ext.name = sl_pp_context_add_unique_str(context, name); + if (ext.name == -1) { + return -1; + } + + ext.name_string = sl_pp_context_add_unique_str(context, name_string); + if (ext.name_string == -1) { + return -1; + } + + context->extensions[context->num_extensions++] = ext; + return 0; +} + +int +sl_pp_process_extension(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last, + struct sl_pp_process_state *state) +{ + int extension_name = -1; + int behavior = -1; + struct sl_pp_token_info out; + + /* Grab the extension name. */ + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + extension_name = input[first].data.identifier; + first++; + } + if (extension_name == -1) { + strcpy(context->error_msg, "expected identifier after `#extension'"); + return -1; + } + + /* Make sure the extension is supported. */ + if (extension_name == context->dict.all) { + out.data.extension = extension_name; + } else { + unsigned int i; + + out.data.extension = -1; + for (i = 0; i < context->num_extensions; i++) { + if (extension_name == context->extensions[i].name_string) { + out.data.extension = extension_name; + break; + } + } + } + + /* Grab the colon separating the extension name and behavior. */ + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + if (first < last && input[first].token == SL_PP_COLON) { + first++; + } else { + strcpy(context->error_msg, "expected `:' after extension name"); + return -1; + } + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + + /* Grab the behavior name. */ + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + behavior = input[first].data.identifier; + first++; + } + if (behavior == -1) { + strcpy(context->error_msg, "expected identifier after `:'"); + return -1; + } + + if (behavior == context->dict.require) { + if (out.data.extension == -1) { + strcpy(context->error_msg, "the required extension is not supported"); + return -1; + } + if (out.data.extension == context->dict.all) { + strcpy(context->error_msg, "invalid behavior for `all' extension: `require'"); + return -1; + } + out.token = SL_PP_EXTENSION_REQUIRE; + } else if (behavior == context->dict.enable) { + if (out.data.extension == -1) { + /* Warning: the extension cannot be enabled. */ + return 0; + } + if (out.data.extension == context->dict.all) { + strcpy(context->error_msg, "invalid behavior for `all' extension: `enable'"); + return -1; + } + out.token = SL_PP_EXTENSION_ENABLE; + } else if (behavior == context->dict.warn) { + if (out.data.extension == -1) { + /* Warning: the extension is not supported. */ + return 0; + } + out.token = SL_PP_EXTENSION_WARN; + } else if (behavior == context->dict.disable) { + if (out.data.extension == -1) { + /* Warning: the extension is not supported. */ + return 0; + } + out.token = SL_PP_EXTENSION_DISABLE; + } else { + strcpy(context->error_msg, "unrecognised behavior name"); + return -1; + } + + /* Grab the end of line. */ + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + if (first < last) { + strcpy(context->error_msg, "expected end of line after behavior name"); + return -1; + } + + if (sl_pp_process_out(state, &out)) { + return -1; + } + + return 0; +} diff --git a/src/glsl/pp/sl_pp_if.c b/src/glsl/pp/sl_pp_if.c new file mode 100644 index 00000000000..f12f0f142c6 --- /dev/null +++ b/src/glsl/pp/sl_pp_if.c @@ -0,0 +1,344 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_expression.h" +#include "sl_pp_process.h" + + +static int +_parse_defined(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer, + struct sl_pp_process_state *state) +{ + struct sl_pp_token_info input; + int parens = 0; + int macro_name; + struct sl_pp_macro *macro; + int defined = 0; + struct sl_pp_token_info result; + + if (sl_pp_token_buffer_skip_white(buffer, &input)) { + return -1; + } + + if (input.token == SL_PP_LPAREN) { + if (sl_pp_token_buffer_skip_white(buffer, &input)) { + return -1; + } + parens = 1; + } + + if (input.token != SL_PP_IDENTIFIER) { + strcpy(context->error_msg, "expected an identifier"); + return -1; + } + + macro_name = input.data.identifier; + for (macro = context->macro; macro; macro = macro->next) { + if (macro->name == macro_name) { + defined = 1; + break; + } + } + + if (parens) { + if (sl_pp_token_buffer_skip_white(buffer, &input)) { + return -1; + } + if (input.token != SL_PP_RPAREN) { + strcpy(context->error_msg, "expected `)'"); + return -1; + } + } + + result.token = SL_PP_UINT; + result.data._uint = (defined ? context->dict._1 : context->dict._0); + + if (sl_pp_process_out(state, &result)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + return 0; +} + +static unsigned int +_evaluate_if_stack(struct sl_pp_context *context) +{ + unsigned int i; + + for (i = context->if_ptr; i < SL_PP_MAX_IF_NESTING; i++) { + if (!(context->if_stack[i] & 1)) { + return 0; + } + } + return 1; +} + +static int +_parse_if(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer) +{ + struct sl_pp_process_state state; + int found_end = 0; + struct sl_pp_token_info eof; + int result; + + if (!context->if_ptr) { + strcpy(context->error_msg, "`#if' nesting too deep"); + return -1; + } + + memset(&state, 0, sizeof(state)); + while (!found_end) { + struct sl_pp_token_info input; + + sl_pp_token_buffer_get(buffer, &input); + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_IDENTIFIER: + if (input.data.identifier == context->dict.defined) { + if (_parse_defined(context, buffer, &state)) { + free(state.out); + return -1; + } + } else { + sl_pp_token_buffer_unget(buffer, &input); + if (sl_pp_macro_expand(context, buffer, NULL, &state, sl_pp_macro_expand_unknown_to_0)) { + free(state.out); + return -1; + } + } + break; + + case SL_PP_NEWLINE: + case SL_PP_EOF: + found_end = 1; + break; + + default: + if (sl_pp_process_out(&state, &input)) { + strcpy(context->error_msg, "out of memory"); + free(state.out); + return -1; + } + } + } + + eof.token = SL_PP_EOF; + if (sl_pp_process_out(&state, &eof)) { + strcpy(context->error_msg, "out of memory"); + free(state.out); + return -1; + } + + if (sl_pp_execute_expression(context, state.out, &result)) { + free(state.out); + return -1; + } + + free(state.out); + + context->if_ptr--; + context->if_stack[context->if_ptr] = result ? 1 : 0; + context->if_value = _evaluate_if_stack(context); + + return 0; +} + +static int +_parse_else(struct sl_pp_context *context) +{ + if (context->if_ptr == SL_PP_MAX_IF_NESTING) { + strcpy(context->error_msg, "no matching `#if'"); + return -1; + } + + /* Bit b1 indicates we already went through #else. */ + if (context->if_stack[context->if_ptr] & 2) { + strcpy(context->error_msg, "no matching `#if'"); + return -1; + } + + /* Invert current condition value and mark that we are in the #else block. */ + context->if_stack[context->if_ptr] = (1 - (context->if_stack[context->if_ptr] & 1)) | 2; + context->if_value = _evaluate_if_stack(context); + + return 0; +} + +int +sl_pp_process_if(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer) +{ + return _parse_if(context, buffer); +} + +int +sl_pp_process_ifdef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + unsigned int i; + + if (!context->if_ptr) { + strcpy(context->error_msg, "`#if' nesting too deep"); + return -1; + } + + for (i = first; i < last; i++) { + switch (input[i].token) { + case SL_PP_IDENTIFIER: + { + struct sl_pp_macro *macro; + int macro_name = input[i].data.identifier; + int defined = 0; + + for (macro = context->macro; macro; macro = macro->next) { + if (macro->name == macro_name) { + defined = 1; + break; + } + } + + context->if_ptr--; + context->if_stack[context->if_ptr] = defined ? 1 : 0; + context->if_value = _evaluate_if_stack(context); + } + return 0; + + case SL_PP_WHITESPACE: + break; + + default: + strcpy(context->error_msg, "expected an identifier"); + return -1; + } + } + + strcpy(context->error_msg, "expected an identifier"); + return -1; +} + +int +sl_pp_process_ifndef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + unsigned int i; + + if (!context->if_ptr) { + strcpy(context->error_msg, "`#if' nesting too deep"); + return -1; + } + + for (i = first; i < last; i++) { + switch (input[i].token) { + case SL_PP_IDENTIFIER: + { + struct sl_pp_macro *macro; + int macro_name = input[i].data.identifier; + int defined = 0; + + for (macro = context->macro; macro; macro = macro->next) { + if (macro->name == macro_name) { + defined = 1; + break; + } + } + + context->if_ptr--; + context->if_stack[context->if_ptr] = defined ? 0 : 1; + context->if_value = _evaluate_if_stack(context); + } + return 0; + + case SL_PP_WHITESPACE: + break; + + default: + strcpy(context->error_msg, "expected an identifier"); + return -1; + } + } + + strcpy(context->error_msg, "expected an identifier"); + return -1; +} + +int +sl_pp_process_elif(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer) +{ + if (_parse_else(context)) { + return -1; + } + + if (context->if_stack[context->if_ptr] & 1) { + context->if_ptr++; + if (_parse_if(context, buffer)) { + return -1; + } + } + + /* We are still in the #if block. */ + context->if_stack[context->if_ptr] = context->if_stack[context->if_ptr] & ~2; + + return 0; +} + +int +sl_pp_process_else(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + return _parse_else(context); +} + +int +sl_pp_process_endif(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last) +{ + if (context->if_ptr == SL_PP_MAX_IF_NESTING) { + strcpy(context->error_msg, "no matching `#if'"); + return -1; + } + + context->if_ptr++; + context->if_value = _evaluate_if_stack(context); + + return 0; +} diff --git a/src/glsl/pp/sl_pp_line.c b/src/glsl/pp/sl_pp_line.c new file mode 100644 index 00000000000..6f7e9eb562c --- /dev/null +++ b/src/glsl/pp/sl_pp_line.c @@ -0,0 +1,127 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_public.h" +#include "sl_pp_process.h" + + +int +sl_pp_process_line(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer, + struct sl_pp_process_state *pstate) +{ + struct sl_pp_process_state state; + int found_end = 0; + int line_number = -1; + int file_number = -1; + unsigned int line; + unsigned int file; + + memset(&state, 0, sizeof(state)); + while (!found_end) { + struct sl_pp_token_info input; + + sl_pp_token_buffer_get(buffer, &input); + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_IDENTIFIER: + sl_pp_token_buffer_unget(buffer, &input); + if (sl_pp_macro_expand(context, buffer, NULL, &state, sl_pp_macro_expand_normal)) { + free(state.out); + return -1; + } + break; + + case SL_PP_NEWLINE: + case SL_PP_EOF: + found_end = 1; + break; + + default: + if (sl_pp_process_out(&state, &input)) { + strcpy(context->error_msg, "out of memory"); + free(state.out); + return -1; + } + } + } + + if (state.out_len > 0 && state.out[0].token == SL_PP_UINT) { + line_number = state.out[0].data._uint; + } else { + strcpy(context->error_msg, "expected a number after `#line'"); + free(state.out); + return -1; + } + + if (state.out_len > 1) { + if (state.out[1].token == SL_PP_UINT) { + file_number = state.out[1].data._uint; + } else { + strcpy(context->error_msg, "expected a number after line number"); + free(state.out); + return -1; + } + + if (state.out_len > 2) { + strcpy(context->error_msg, "expected an end of line after file number"); + free(state.out); + return -1; + } + } + + free(state.out); + + line = atoi(sl_pp_context_cstr(context, line_number)); + if (file_number != -1) { + file = atoi(sl_pp_context_cstr(context, file_number)); + } else { + file = context->file; + } + + if (context->line != line || context->file != file) { + struct sl_pp_token_info ti; + + ti.token = SL_PP_LINE; + ti.data.line.lineno = line; + ti.data.line.fileno = file; + if (sl_pp_process_out(pstate, &ti)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + context->line = line; + context->file = file; + } + + return 0; +} diff --git a/src/glsl/pp/sl_pp_macro.c b/src/glsl/pp/sl_pp_macro.c new file mode 100644 index 00000000000..9f520b8fc53 --- /dev/null +++ b/src/glsl/pp/sl_pp_macro.c @@ -0,0 +1,414 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <stdio.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_public.h" +#include "sl_pp_macro.h" +#include "sl_pp_process.h" + + +static void +_macro_init(struct sl_pp_macro *macro) +{ + macro->name = -1; + macro->num_args = -1; + macro->arg = NULL; + macro->body = NULL; +} + +struct sl_pp_macro * +sl_pp_macro_new(void) +{ + struct sl_pp_macro *macro; + + macro = calloc(1, sizeof(struct sl_pp_macro)); + if (macro) { + _macro_init(macro); + } + return macro; +} + +static void +_macro_destroy(struct sl_pp_macro *macro) +{ + struct sl_pp_macro_formal_arg *arg = macro->arg; + + while (arg) { + struct sl_pp_macro_formal_arg *next_arg = arg->next; + + free(arg); + arg = next_arg; + } + + free(macro->body); +} + +void +sl_pp_macro_free(struct sl_pp_macro *macro) +{ + while (macro) { + struct sl_pp_macro *next_macro = macro->next; + + _macro_destroy(macro); + free(macro); + macro = next_macro; + } +} + +void +sl_pp_macro_reset(struct sl_pp_macro *macro) +{ + _macro_destroy(macro); + _macro_init(macro); +} + +static int +_out_number(struct sl_pp_context *context, + struct sl_pp_process_state *state, + unsigned int number) +{ + char buf[32]; + struct sl_pp_token_info ti; + + sprintf(buf, "%u", number); + + ti.token = SL_PP_UINT; + ti.data._uint = sl_pp_context_add_unique_str(context, buf); + if (sl_pp_process_out(state, &ti)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + return 0; +} + +int +sl_pp_macro_expand(struct sl_pp_context *context, + struct sl_pp_token_buffer *tokens, + struct sl_pp_macro *local, + struct sl_pp_process_state *state, + enum sl_pp_macro_expand_behaviour behaviour) +{ + int mute = (behaviour == sl_pp_macro_expand_mute); + struct sl_pp_token_info input; + int macro_name; + struct sl_pp_macro *macro = NULL; + struct sl_pp_macro *actual_arg = NULL; + unsigned int j; + + if (sl_pp_token_buffer_get(tokens, &input)) { + return -1; + } + + if (input.token != SL_PP_IDENTIFIER) { + strcpy(context->error_msg, "expected an identifier"); + return -1; + } + + macro_name = input.data.identifier; + + /* First look for predefined macros. + */ + + if (macro_name == context->dict.___LINE__) { + if (!mute && _out_number(context, state, context->line)) { + return -1; + } + return 0; + } + if (macro_name == context->dict.___FILE__) { + if (!mute && _out_number(context, state, context->file)) { + return -1; + } + return 0; + } + if (macro_name == context->dict.___VERSION__) { + if (!mute && _out_number(context, state, 110)) { + return -1; + } + return 0; + } + + for (j = 0; j < context->num_predefined; j++) { + if (macro_name == context->predefined[j].name) { + if (!mute) { + struct sl_pp_token_info ti; + + ti.token = SL_PP_UINT; + ti.data._uint = context->predefined[j].value; + if (sl_pp_process_out(state, &ti)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + } + return 0; + } + } + + /* Replace extension names with 1. + */ + for (j = 0; j < context->num_extensions; j++) { + if (macro_name == context->extensions[j].name) { + if (!mute && _out_number(context, state, 1)) { + return -1; + } + return 0; + } + } + + if (local) { + for (macro = local; macro; macro = macro->next) { + if (macro->name == macro_name) { + break; + } + } + } + + if (!macro) { + for (macro = context->macro; macro; macro = macro->next) { + if (macro->name == macro_name) { + break; + } + } + } + + if (!macro) { + if (behaviour == sl_pp_macro_expand_unknown_to_0) { + if (_out_number(context, state, 0)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + } else if (!mute) { + if (sl_pp_process_out(state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + } + return 0; + } + + if (macro->num_args >= 0) { + if (sl_pp_token_buffer_skip_white(tokens, &input)) { + return -1; + } + if (input.token != SL_PP_LPAREN) { + strcpy(context->error_msg, "expected `('"); + return -1; + } + if (sl_pp_token_buffer_skip_white(tokens, &input)) { + return -1; + } + sl_pp_token_buffer_unget(tokens, &input); + } + + if (macro->num_args > 0) { + struct sl_pp_macro_formal_arg *formal_arg = macro->arg; + struct sl_pp_macro **pmacro = &actual_arg; + + for (j = 0; j < (unsigned int)macro->num_args; j++) { + struct sl_pp_process_state arg_state; + int done = 0; + unsigned int paren_nesting = 0; + struct sl_pp_token_info eof; + + memset(&arg_state, 0, sizeof(arg_state)); + + while (!done) { + if (sl_pp_token_buffer_get(tokens, &input)) { + goto fail_arg; + } + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_COMMA: + if (!paren_nesting) { + if (j < (unsigned int)macro->num_args - 1) { + done = 1; + } else { + strcpy(context->error_msg, "too many actual macro arguments"); + goto fail_arg; + } + } else { + if (sl_pp_process_out(&arg_state, &input)) { + strcpy(context->error_msg, "out of memory"); + goto fail_arg; + } + } + break; + + case SL_PP_LPAREN: + paren_nesting++; + if (sl_pp_process_out(&arg_state, &input)) { + goto oom_arg; + } + break; + + case SL_PP_RPAREN: + if (!paren_nesting) { + if (j == (unsigned int)macro->num_args - 1) { + done = 1; + } else { + strcpy(context->error_msg, "too few actual macro arguments"); + goto fail_arg; + } + } else { + paren_nesting--; + if (sl_pp_process_out(&arg_state, &input)) { + goto oom_arg; + } + } + break; + + case SL_PP_IDENTIFIER: + sl_pp_token_buffer_unget(tokens, &input); + if (sl_pp_macro_expand(context, tokens, local, &arg_state, sl_pp_macro_expand_normal)) { + goto fail_arg; + } + break; + + case SL_PP_EOF: + strcpy(context->error_msg, "too few actual macro arguments"); + goto fail_arg; + + default: + if (sl_pp_process_out(&arg_state, &input)) { + goto oom_arg; + } + } + } + + eof.token = SL_PP_EOF; + if (sl_pp_process_out(&arg_state, &eof)) { + goto oom_arg; + } + + *pmacro = sl_pp_macro_new(); + if (!*pmacro) { + goto oom_arg; + } + + (**pmacro).name = formal_arg->name; + (**pmacro).body = arg_state.out; + + formal_arg = formal_arg->next; + pmacro = &(**pmacro).next; + + continue; + +oom_arg: + strcpy(context->error_msg, "out of memory"); +fail_arg: + free(arg_state.out); + goto fail; + } + } + + /* Right paren for non-empty argument list has already been eaten. */ + if (macro->num_args == 0) { + if (sl_pp_token_buffer_skip_white(tokens, &input)) { + goto fail; + } + if (input.token != SL_PP_RPAREN) { + strcpy(context->error_msg, "expected `)'"); + goto fail; + } + } + + /* XXX: This is all wrong, we should be ungetting all tokens + * back to the main token buffer. + */ + { + struct sl_pp_token_buffer buffer; + + /* Seek to the end. + */ + for (j = 0; macro->body[j].token != SL_PP_EOF; j++) { + } + j++; + + /* Create a context-less token buffer since we are not going to underrun + * its internal buffer. + */ + if (sl_pp_token_buffer_init(&buffer, NULL)) { + strcpy(context->error_msg, "out of memory"); + goto fail; + } + + /* Unget the tokens in reverse order so later they will be fetched correctly. + */ + for (; j > 0; j--) { + sl_pp_token_buffer_unget(&buffer, ¯o->body[j - 1]); + } + + /* Expand. + */ + for (;;) { + struct sl_pp_token_info input; + + sl_pp_token_buffer_get(&buffer, &input); + switch (input.token) { + case SL_PP_NEWLINE: + if (sl_pp_process_out(state, &input)) { + strcpy(context->error_msg, "out of memory"); + sl_pp_token_buffer_destroy(&buffer); + goto fail; + } + break; + + case SL_PP_IDENTIFIER: + sl_pp_token_buffer_unget(&buffer, &input); + if (sl_pp_macro_expand(context, &buffer, actual_arg, state, behaviour)) { + sl_pp_token_buffer_destroy(&buffer); + goto fail; + } + break; + + case SL_PP_EOF: + sl_pp_token_buffer_destroy(&buffer); + sl_pp_macro_free(actual_arg); + return 0; + + default: + if (!mute) { + if (sl_pp_process_out(state, &input)) { + strcpy(context->error_msg, "out of memory"); + sl_pp_token_buffer_destroy(&buffer); + goto fail; + } + } + } + } + } + +fail: + sl_pp_macro_free(actual_arg); + return -1; +} diff --git a/src/glsl/pp/sl_pp_macro.h b/src/glsl/pp/sl_pp_macro.h new file mode 100644 index 00000000000..1d210681091 --- /dev/null +++ b/src/glsl/pp/sl_pp_macro.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_MACRO_H +#define SL_PP_MACRO_H + +#include "sl_pp_token.h" + + +struct sl_pp_context; +struct sl_pp_process_state; +struct sl_pp_token_buffer; + +struct sl_pp_macro_formal_arg { + int name; + struct sl_pp_macro_formal_arg *next; +}; + +struct sl_pp_macro { + int name; + int num_args; /* -1 means no args, 0 means `()' */ + struct sl_pp_macro_formal_arg *arg; + struct sl_pp_token_info *body; + struct sl_pp_macro *next; +}; + +struct sl_pp_macro * +sl_pp_macro_new(void); + +void +sl_pp_macro_free(struct sl_pp_macro *macro); + +void +sl_pp_macro_reset(struct sl_pp_macro *macro); + +enum sl_pp_macro_expand_behaviour { + sl_pp_macro_expand_normal, + sl_pp_macro_expand_mute, + sl_pp_macro_expand_unknown_to_0 +}; + +int +sl_pp_macro_expand(struct sl_pp_context *context, + struct sl_pp_token_buffer *tokens, + struct sl_pp_macro *local, + struct sl_pp_process_state *state, + enum sl_pp_macro_expand_behaviour behaviour); + +#endif /* SL_PP_MACRO_H */ diff --git a/src/glsl/pp/sl_pp_pragma.c b/src/glsl/pp/sl_pp_pragma.c new file mode 100644 index 00000000000..caf4c63f657 --- /dev/null +++ b/src/glsl/pp/sl_pp_pragma.c @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_process.h" + + +int +sl_pp_process_pragma(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last, + struct sl_pp_process_state *state) +{ + int pragma_name = -1; + struct sl_pp_token_info out; + int arg_name = -1; + + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + pragma_name = input[first].data.identifier; + first++; + } + if (pragma_name == -1) { + return 0; + } + + if (pragma_name == context->dict.optimize) { + out.token = SL_PP_PRAGMA_OPTIMIZE; + } else if (pragma_name == context->dict.debug) { + out.token = SL_PP_PRAGMA_DEBUG; + } else { + return 0; + } + + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + + if (first < last && input[first].token == SL_PP_LPAREN) { + first++; + } else { + return 0; + } + + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + + if (first < last && input[first].token == SL_PP_IDENTIFIER) { + arg_name = input[first].data.identifier; + first++; + } + if (arg_name == -1) { + return 0; + } + + if (arg_name == context->dict.off) { + out.data.pragma = 0; + } else if (arg_name == context->dict.on) { + out.data.pragma = 1; + } else { + return 0; + } + + while (first < last && input[first].token == SL_PP_WHITESPACE) { + first++; + } + + if (first < last && input[first].token == SL_PP_RPAREN) { + first++; + } else { + return 0; + } + + /* Ignore the tokens that follow. */ + + if (sl_pp_process_out(state, &out)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + return 0; +} diff --git a/src/glsl/pp/sl_pp_process.c b/src/glsl/pp/sl_pp_process.c new file mode 100644 index 00000000000..563ea948e70 --- /dev/null +++ b/src/glsl/pp/sl_pp_process.c @@ -0,0 +1,324 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_context.h" +#include "sl_pp_process.h" +#include "sl_pp_public.h" + + +int +sl_pp_process_out(struct sl_pp_process_state *state, + const struct sl_pp_token_info *token) +{ + if (state->out_len >= state->out_max) { + unsigned int new_max = state->out_max; + + if (new_max < 0x100) { + new_max = 0x100; + } else if (new_max < 0x10000) { + new_max *= 2; + } else { + new_max += 0x10000; + } + + state->out = realloc(state->out, new_max * sizeof(struct sl_pp_token_info)); + if (!state->out) { + return -1; + } + state->out_max = new_max; + } + + state->out[state->out_len++] = *token; + return 0; +} + +int +sl_pp_process_get(struct sl_pp_context *context, + struct sl_pp_token_info *output) +{ + if (!context->process_state.out) { + if (context->line > 1) { + struct sl_pp_token_info ti; + + ti.token = SL_PP_LINE; + ti.data.line.lineno = context->line - 1; + ti.data.line.fileno = context->file; + if (sl_pp_process_out(&context->process_state, &ti)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + + ti.token = SL_PP_NEWLINE; + if (sl_pp_process_out(&context->process_state, &ti)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + } + } + + for (;;) { + struct sl_pp_token_info input; + int found_eof = 0; + + if (context->process_state.out_len) { + *output = context->process_state.out[0]; + + if (context->process_state.out_len > 1) { + unsigned int i; + + for (i = 1; i < context->process_state.out_len; i++) { + context->process_state.out[i - 1] = context->process_state.out[i]; + } + } + context->process_state.out_len--; + + return 0; + } + + if (sl_pp_token_buffer_skip_white(&context->tokens, &input)) { + return -1; + } + if (input.token == SL_PP_HASH) { + if (sl_pp_token_buffer_skip_white(&context->tokens, &input)) { + return -1; + } + switch (input.token) { + case SL_PP_IDENTIFIER: + { + int name; + int found_eol = 0; + struct sl_pp_token_info endof; + struct sl_pp_token_peek peek; + int result = 0; + + /* Directive name. */ + name = input.data.identifier; + + if (sl_pp_token_buffer_skip_white(&context->tokens, &input)) { + return -1; + } + sl_pp_token_buffer_unget(&context->tokens, &input); + + if (sl_pp_token_peek_init(&peek, &context->tokens)) { + return -1; + } + + while (!found_eol) { + if (sl_pp_token_peek_get(&peek, &input)) { + sl_pp_token_peek_destroy(&peek); + return -1; + } + switch (input.token) { + case SL_PP_NEWLINE: + /* Preserve newline just for the sake of line numbering. */ + endof = input; + found_eol = 1; + break; + + case SL_PP_EOF: + endof = input; + found_eof = 1; + found_eol = 1; + break; + } + } + + if (name == context->dict._if) { + struct sl_pp_token_buffer buffer; + + result = sl_pp_token_peek_to_buffer(&peek, &buffer); + if (result == 0) { + result = sl_pp_process_if(context, &buffer); + sl_pp_token_buffer_destroy(&buffer); + } + } else if (name == context->dict.ifdef) { + result = sl_pp_process_ifdef(context, peek.tokens, 0, peek.size - 1); + } else if (name == context->dict.ifndef) { + result = sl_pp_process_ifndef(context, peek.tokens, 0, peek.size - 1); + } else if (name == context->dict.elif) { + struct sl_pp_token_buffer buffer; + + result = sl_pp_token_peek_to_buffer(&peek, &buffer); + if (result == 0) { + result = sl_pp_process_elif(context, &buffer); + sl_pp_token_buffer_destroy(&buffer); + } + } else if (name == context->dict._else) { + result = sl_pp_process_else(context, peek.tokens, 0, peek.size - 1); + } else if (name == context->dict.endif) { + result = sl_pp_process_endif(context, peek.tokens, 0, peek.size - 1); + } else if (context->if_value) { + if (name == context->dict.define) { + result = sl_pp_process_define(context, peek.tokens, 0, peek.size - 1); + } else if (name == context->dict.error) { + sl_pp_process_error(context, peek.tokens, 0, peek.size - 1); + result = -1; + } else if (name == context->dict.extension) { + result = sl_pp_process_extension(context, peek.tokens, 0, peek.size - 1, &context->process_state); + } else if (name == context->dict.line) { + struct sl_pp_token_buffer buffer; + + result = sl_pp_token_peek_to_buffer(&peek, &buffer); + if (result == 0) { + result = sl_pp_process_line(context, &buffer, &context->process_state); + sl_pp_token_buffer_destroy(&buffer); + } + } else if (name == context->dict.pragma) { + result = sl_pp_process_pragma(context, peek.tokens, 0, peek.size - 1, &context->process_state); + } else if (name == context->dict.undef) { + result = sl_pp_process_undef(context, peek.tokens, 0, peek.size - 1); + } else { + strcpy(context->error_msg, "unrecognised directive name"); + result = -1; + } + } + + sl_pp_token_peek_commit(&peek); + sl_pp_token_peek_destroy(&peek); + + if (result) { + return result; + } + + if (sl_pp_process_out(&context->process_state, &endof)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + context->line++; + } + break; + + case SL_PP_NEWLINE: + /* Empty directive. */ + if (sl_pp_process_out(&context->process_state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + context->line++; + break; + + case SL_PP_EOF: + /* Empty directive. */ + if (sl_pp_process_out(&context->process_state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + found_eof = 1; + break; + + default: + strcpy(context->error_msg, "expected a directive name"); + return -1; + } + } else { + int found_eol = 0; + + sl_pp_token_buffer_unget(&context->tokens, &input); + + while (!found_eol) { + if (sl_pp_token_buffer_get(&context->tokens, &input)) { + return -1; + } + + switch (input.token) { + case SL_PP_WHITESPACE: + /* Drop whitespace all together at this point. */ + break; + + case SL_PP_NEWLINE: + /* Preserve newline just for the sake of line numbering. */ + if (sl_pp_process_out(&context->process_state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + context->line++; + found_eol = 1; + break; + + case SL_PP_EOF: + if (sl_pp_process_out(&context->process_state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + found_eof = 1; + found_eol = 1; + break; + + case SL_PP_IDENTIFIER: + sl_pp_token_buffer_unget(&context->tokens, &input); + if (sl_pp_macro_expand(context, &context->tokens, NULL, &context->process_state, + context->if_value ? sl_pp_macro_expand_normal : sl_pp_macro_expand_mute)) { + return -1; + } + break; + + default: + if (context->if_value) { + if (sl_pp_process_out(&context->process_state, &input)) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + } + } + } + } + + if (found_eof) { + if (context->if_ptr != SL_PP_MAX_IF_NESTING) { + strcpy(context->error_msg, "expected `#endif' directive"); + return -1; + } + } + } +} + +int +sl_pp_process(struct sl_pp_context *context, + struct sl_pp_token_info **output) +{ + struct sl_pp_process_state state; + + memset(&state, 0, sizeof(state)); + for (;;) { + struct sl_pp_token_info input; + + if (sl_pp_process_get(context, &input)) { + free(state.out); + return -1; + } + if (sl_pp_process_out(&state, &input)) { + free(state.out); + return -1; + } + if (input.token == SL_PP_EOF) { + *output = state.out; + return 0; + } + } +} diff --git a/src/glsl/pp/sl_pp_process.h b/src/glsl/pp/sl_pp_process.h new file mode 100644 index 00000000000..fe6ff0d4648 --- /dev/null +++ b/src/glsl/pp/sl_pp_process.h @@ -0,0 +1,116 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_PROCESS_H +#define SL_PP_PROCESS_H + +#include "sl_pp_macro.h" +#include "sl_pp_token.h" + + +struct sl_pp_context; + +struct sl_pp_process_state { + struct sl_pp_token_info *out; + unsigned int out_len; + unsigned int out_max; +}; + +int +sl_pp_process_define(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_undef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_if(struct sl_pp_context *context, + struct sl_pp_token_buffer *input); + +int +sl_pp_process_ifdef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_ifndef(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_elif(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer); + +int +sl_pp_process_else(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_endif(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +void +sl_pp_process_error(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last); + +int +sl_pp_process_pragma(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last, + struct sl_pp_process_state *state); + +int +sl_pp_process_extension(struct sl_pp_context *context, + const struct sl_pp_token_info *input, + unsigned int first, + unsigned int last, + struct sl_pp_process_state *state); + +int +sl_pp_process_line(struct sl_pp_context *context, + struct sl_pp_token_buffer *buffer, + struct sl_pp_process_state *state); + +int +sl_pp_process_out(struct sl_pp_process_state *state, + const struct sl_pp_token_info *token); + +#endif /* SL_PP_PROCESS_H */ diff --git a/src/glsl/pp/sl_pp_public.h b/src/glsl/pp/sl_pp_public.h new file mode 100644 index 00000000000..12528d6f8d1 --- /dev/null +++ b/src/glsl/pp/sl_pp_public.h @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_PUBLIC_H +#define SL_PP_PUBLIC_H + + +struct sl_pp_context; + + +#include "sl_pp_purify.h" +#include "sl_pp_token.h" + + +struct sl_pp_context * +sl_pp_context_create(const char *input, + const struct sl_pp_purify_options *options); + +void +sl_pp_context_destroy(struct sl_pp_context *context); + +const char * +sl_pp_context_error_message(const struct sl_pp_context *context); + +void +sl_pp_context_error_position(const struct sl_pp_context *context, + unsigned int *file, + unsigned int *line); + +int +sl_pp_context_add_extension(struct sl_pp_context *context, + const char *name, + const char *name_string); + +int +sl_pp_context_add_predefined(struct sl_pp_context *context, + const char *name, + const char *value); + +int +sl_pp_context_add_unique_str(struct sl_pp_context *context, + const char *str); + +const char * +sl_pp_context_cstr(const struct sl_pp_context *context, + int offset); + +int +sl_pp_version(struct sl_pp_context *context, + unsigned int *version); + +int +sl_pp_process_get(struct sl_pp_context *context, + struct sl_pp_token_info *output); + +int +sl_pp_process(struct sl_pp_context *context, + struct sl_pp_token_info **output); + +#endif /* SL_PP_PUBLIC_H */ diff --git a/src/glsl/pp/sl_pp_purify.c b/src/glsl/pp/sl_pp_purify.c new file mode 100644 index 00000000000..b50f8192517 --- /dev/null +++ b/src/glsl/pp/sl_pp_purify.c @@ -0,0 +1,302 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <stdarg.h> +#include <stdio.h> +#include "sl_pp_purify.h" + + +/* + * Preprocessor purifier performs the following tasks. + * - Convert all variants of newlines into a Unix newline. + * - Merge continued lines into a single long line. + * - Remove line comments and replace block comments with whitespace. + */ + + +static unsigned int +_purify_newline(const char *input, + char *out, + unsigned int *current_line) +{ + if (input[0] == '\n') { + *out = '\n'; + (*current_line)++; + if (input[1] == '\r') { + /* + * The GLSL spec is not explicit about whether this + * combination is a valid newline or not. + * Let's assume it is acceptable. + */ + return 2; + } + return 1; + } + if (input[0] == '\r') { + *out = '\n'; + (*current_line)++; + if (input[1] == '\n') { + return 2; + } + return 1; + } + *out = input[0]; + return 1; +} + + +static unsigned int +_purify_backslash(const char *input, + char *out, + unsigned int *current_line) +{ + unsigned int eaten = 0; + + for (;;) { + if (input[0] == '\\') { + char next; + unsigned int next_eaten; + unsigned int next_line = *current_line; + + eaten++; + input++; + + next_eaten = _purify_newline(input, &next, &next_line); + if (next == '\n') { + /* + * If this is really a line continuation sequence, eat + * it and do not exit the loop. + */ + eaten += next_eaten; + input += next_eaten; + *current_line = next_line; + } else { + /* + * It is an error to put anything between a backslash + * and a newline and still expect it to behave like a line + * continuation sequence. + * Even if it is an innocent whitespace. + */ + *out = '\\'; + break; + } + } else { + eaten += _purify_newline(input, out, current_line); + break; + } + } + return eaten; +} + + +static void +_report_error(char *buf, + unsigned int cbbuf, + const char *msg, + ...) +{ + va_list args; + + va_start(args, msg); + vsnprintf(buf, cbbuf, msg, args); + va_end(args); +} + + +void +sl_pp_purify_state_init(struct sl_pp_purify_state *state, + const char *input, + const struct sl_pp_purify_options *options) +{ + state->options = *options; + state->input = input; + state->current_line = 1; + state->inside_c_comment = 0; +} + + +unsigned int +_purify_comment(struct sl_pp_purify_state *state, + char *output, + unsigned int *current_line, + char *errormsg, + unsigned int cberrormsg) +{ + for (;;) { + unsigned int eaten; + char next; + + eaten = _purify_backslash(state->input, &next, current_line); + state->input += eaten; + while (next == '*') { + eaten = _purify_backslash(state->input, &next, current_line); + state->input += eaten; + if (next == '/') { + *output = ' '; + state->inside_c_comment = 0; + return 1; + } + } + if (next == '\n') { + *output = '\n'; + state->inside_c_comment = 1; + return 1; + } + if (next == '\0') { + _report_error(errormsg, cberrormsg, "expected `*/' but end of translation unit found"); + return 0; + } + } +} + + +unsigned int +sl_pp_purify_getc(struct sl_pp_purify_state *state, + char *output, + unsigned int *current_line, + char *errormsg, + unsigned int cberrormsg) +{ + unsigned int eaten; + + if (state->inside_c_comment) { + return _purify_comment(state, output, current_line, errormsg, cberrormsg); + } + + eaten = _purify_backslash(state->input, output, current_line); + state->input += eaten; + if (*output == '/') { + char next; + unsigned int next_line = *current_line; + + eaten = _purify_backslash(state->input, &next, &next_line); + if (next == '/') { + state->input += eaten; + *current_line = next_line; + + /* Replace a line comment with either a newline or nil. */ + for (;;) { + eaten = _purify_backslash(state->input, &next, current_line); + state->input += eaten; + if (next == '\n' || next == '\0') { + *output = next; + return eaten; + } + } + } else if (next == '*') { + state->input += eaten; + *current_line = next_line; + + return _purify_comment(state, output, current_line, errormsg, cberrormsg); + } + } + return eaten; +} + + +struct out_buf { + char *out; + unsigned int len; + unsigned int capacity; + unsigned int current_line; + char *errormsg; + unsigned int cberrormsg; +}; + + +static int +_out_buf_putc(struct out_buf *obuf, + char c) +{ + if (obuf->len >= obuf->capacity) { + unsigned int new_max = obuf->capacity; + + if (new_max < 0x100) { + new_max = 0x100; + } else if (new_max < 0x10000) { + new_max *= 2; + } else { + new_max += 0x10000; + } + + obuf->out = realloc(obuf->out, new_max); + if (!obuf->out) { + _report_error(obuf->errormsg, obuf->cberrormsg, "out of memory"); + return -1; + } + obuf->capacity = new_max; + } + + obuf->out[obuf->len++] = c; + + return 0; +} + + +int +sl_pp_purify(const char *input, + const struct sl_pp_purify_options *options, + char **output, + char *errormsg, + unsigned int cberrormsg, + unsigned int *errorline) +{ + struct out_buf obuf; + struct sl_pp_purify_state state; + + obuf.out = NULL; + obuf.len = 0; + obuf.capacity = 0; + obuf.current_line = 1; + obuf.errormsg = errormsg; + obuf.cberrormsg = cberrormsg; + + sl_pp_purify_state_init(&state, input, options); + + for (;;) { + unsigned int eaten; + char c; + + eaten = sl_pp_purify_getc(&state, &c, &obuf.current_line, errormsg, cberrormsg); + if (!eaten) { + *errorline = obuf.current_line; + return -1; + } + if (_out_buf_putc(&obuf, c)) { + *errorline = obuf.current_line; + return -1; + } + + if (c == '\0') { + break; + } + } + + *output = obuf.out; + return 0; +} diff --git a/src/glsl/pp/sl_pp_purify.h b/src/glsl/pp/sl_pp_purify.h new file mode 100644 index 00000000000..c0f55cbfd89 --- /dev/null +++ b/src/glsl/pp/sl_pp_purify.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_PURIFY_H +#define SL_PP_PURIFY_H + +struct sl_pp_purify_options { + unsigned int preserve_columns:1; + unsigned int tab_width:4; +}; + +int +sl_pp_purify(const char *input, + const struct sl_pp_purify_options *options, + char **output, + char *errormsg, + unsigned int cberrormsg, + unsigned int *errorline); + +struct sl_pp_purify_state { + struct sl_pp_purify_options options; + const char *input; + unsigned int current_line; + unsigned int inside_c_comment:1; +}; + +void +sl_pp_purify_state_init(struct sl_pp_purify_state *state, + const char *input, + const struct sl_pp_purify_options *options); + +unsigned int +sl_pp_purify_getc(struct sl_pp_purify_state *state, + char *output, + unsigned int *current_line, + char *errormsg, + unsigned int cberrormsg); + +#endif /* SL_PP_PURIFY_H */ diff --git a/src/glsl/pp/sl_pp_token.c b/src/glsl/pp/sl_pp_token.c new file mode 100644 index 00000000000..a7089787005 --- /dev/null +++ b/src/glsl/pp/sl_pp_token.c @@ -0,0 +1,854 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <assert.h> +#include <stdlib.h> +#include <string.h> +#include "sl_pp_public.h" +#include "sl_pp_context.h" +#include "sl_pp_token.h" + + +#define PURE_ERROR 256 + +static int +_pure_getc(struct sl_pp_context *context) +{ + char c; + + if (context->getc_buf_size) { + return context->getc_buf[--context->getc_buf_size]; + } + + if (sl_pp_purify_getc(&context->pure, &c, &context->error_line, context->error_msg, sizeof(context->error_msg)) == 0) { + return PURE_ERROR; + } + return c; +} + + +static void +_pure_ungetc(struct sl_pp_context *context, + int c) +{ + assert(c != PURE_ERROR); + + if (context->getc_buf_size == context->getc_buf_capacity) { + context->getc_buf_capacity += 64; + context->getc_buf = realloc(context->getc_buf, context->getc_buf_capacity * sizeof(char)); + assert(context->getc_buf); + } + + context->getc_buf[context->getc_buf_size++] = (char)c; +} + + +struct lookahead_state { + char buf[256]; + unsigned int pos; + struct sl_pp_context *context; +}; + + +static void +_lookahead_init(struct lookahead_state *lookahead, + struct sl_pp_context *context) +{ + lookahead->pos = 0; + lookahead->context = context; +} + + +static unsigned int +_lookahead_tell(const struct lookahead_state *lookahead) +{ + return lookahead->pos; +} + + +static const void * +_lookahead_buf(const struct lookahead_state *lookahead) +{ + return lookahead->buf; +} + + +static void +_lookahead_revert(struct lookahead_state *lookahead, + unsigned int pos) +{ + assert(pos <= lookahead->pos); + + while (lookahead->pos > pos) { + _pure_ungetc(lookahead->context, lookahead->buf[--lookahead->pos]); + } +} + + +static int +_lookahead_getc(struct lookahead_state *lookahead) +{ + int c; + + assert(lookahead->pos < sizeof(lookahead->buf) / sizeof(lookahead->buf[0])); + + c = _pure_getc(lookahead->context); + if (c != PURE_ERROR) { + lookahead->buf[lookahead->pos++] = (char)c; + } + return c; +} + + +static int +_is_identifier_char(char c) +{ + return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || (c >= '0' && c <= '9') || c == '_'; +} + + +static int +_tokenise_identifier(struct sl_pp_context *context, + struct sl_pp_token_info *out) +{ + int c; + char identifier[256]; /* XXX: Remove this artifical limit. */ + unsigned int i = 0; + + out->token = SL_PP_IDENTIFIER; + out->data.identifier = -1; + + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + identifier[i++] = (char)c; + for (;;) { + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + + if (_is_identifier_char((char)c)) { + if (i >= sizeof(identifier) / sizeof(char) - 1) { + strcpy(context->error_msg, "out of memory"); + _pure_ungetc(context, c); + while (i) { + _pure_ungetc(context, identifier[--i]); + } + return -1; + } + identifier[i++] = (char)c; + } else { + _pure_ungetc(context, c); + break; + } + } + identifier[i] = '\0'; + + out->data.identifier = sl_pp_context_add_unique_str(context, identifier); + if (out->data.identifier == -1) { + while (i) { + _pure_ungetc(context, identifier[--i]); + } + return -1; + } + + return 0; +} + + +/* + * Return the number of consecutive decimal digits in the input stream. + */ +static unsigned int +_parse_float_digits(struct lookahead_state *lookahead) +{ + unsigned int eaten; + + for (eaten = 0;; eaten++) { + unsigned int pos = _lookahead_tell(lookahead); + char c = _lookahead_getc(lookahead); + + if (c < '0' || c > '9') { + _lookahead_revert(lookahead, pos); + break; + } + } + return eaten; +} + + +/* + * Try to match one of the following patterns for the fractional part + * of a floating point number. + * + * digits . [digits] + * . digits + * + * Return 0 if the pattern could not be matched, otherwise the number + * of eaten characters from the input stream. + */ +static unsigned int +_parse_float_frac(struct lookahead_state *lookahead) +{ + unsigned int pos; + int c; + unsigned int eaten; + + pos = _lookahead_tell(lookahead); + c = _lookahead_getc(lookahead); + if (c == '.') { + eaten = _parse_float_digits(lookahead); + if (eaten) { + return eaten + 1; + } + _lookahead_revert(lookahead, pos); + return 0; + } + + _lookahead_revert(lookahead, pos); + eaten = _parse_float_digits(lookahead); + if (eaten) { + c = _lookahead_getc(lookahead); + if (c == '.') { + return eaten + 1 + _parse_float_digits(lookahead); + } + } + + _lookahead_revert(lookahead, pos); + return 0; +} + + +/* + * Try to match the following pattern for the exponential part + * of a floating point number. + * + * (e|E) [(+|-)] digits + * + * Return 0 if the pattern could not be matched, otherwise the number + * of eaten characters from the input stream. + */ +static unsigned int +_parse_float_exp(struct lookahead_state *lookahead) +{ + unsigned int pos, pos2; + int c; + unsigned int eaten, digits; + + pos = _lookahead_tell(lookahead); + c = _lookahead_getc(lookahead); + if (c != 'e' && c != 'E') { + _lookahead_revert(lookahead, pos); + return 0; + } + + pos2 = _lookahead_tell(lookahead); + c = _lookahead_getc(lookahead); + if (c == '-' || c == '+') { + eaten = 2; + } else { + _lookahead_revert(lookahead, pos2); + eaten = 1; + } + + digits = _parse_float_digits(lookahead); + if (!digits) { + _lookahead_revert(lookahead, pos); + return 0; + } + + return eaten + digits; +} + + +/* + * Try to match one of the following patterns for a floating point number. + * + * fract [exp] [(f|F)] + * digits exp [(f|F)] + * + * Return 0 if the pattern could not be matched, otherwise the number + * of eaten characters from the input stream. + */ +static unsigned int +_parse_float(struct lookahead_state *lookahead) +{ + unsigned int eaten; + + eaten = _parse_float_frac(lookahead); + if (eaten) { + unsigned int pos; + int c; + + eaten += _parse_float_exp(lookahead); + + pos = _lookahead_tell(lookahead); + c = _lookahead_getc(lookahead); + if (c == 'f' || c == 'F') { + eaten++; + } else { + _lookahead_revert(lookahead, pos); + } + + return eaten; + } + + eaten = _parse_float_digits(lookahead); + if (eaten) { + unsigned int exponent; + + exponent = _parse_float_exp(lookahead); + if (exponent) { + unsigned int pos; + int c; + + eaten += exponent; + + pos = _lookahead_tell(lookahead); + c = _lookahead_getc(lookahead); + if (c == 'f' || c == 'F') { + eaten++; + } else { + _lookahead_revert(lookahead, pos); + } + + return eaten; + } + } + + _lookahead_revert(lookahead, 0); + return 0; +} + + +static unsigned int +_parse_hex(struct lookahead_state *lookahead) +{ + int c; + unsigned int n; + + c = _lookahead_getc(lookahead); + if (c != '0') { + _lookahead_revert(lookahead, 0); + return 0; + } + + c = _lookahead_getc(lookahead); + if (c != 'x' && c != 'X') { + _lookahead_revert(lookahead, 0); + return 0; + } + + for (n = 2;;) { + unsigned int pos = _lookahead_tell(lookahead); + + c = _lookahead_getc(lookahead); + if ((c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F')) { + n++; + } else { + _lookahead_revert(lookahead, pos); + break; + } + } + + if (n > 2) { + return n; + } + + _lookahead_revert(lookahead, 0); + return 0; +} + + +static unsigned int +_parse_oct(struct lookahead_state *lookahead) +{ + int c; + unsigned int n; + + c = _lookahead_getc(lookahead); + if (c != '0') { + _lookahead_revert(lookahead, 0); + return 0; + } + + for (n = 1;;) { + unsigned int pos = _lookahead_tell(lookahead); + + c = _lookahead_getc(lookahead); + if ((c >= '0' && c <= '7')) { + n++; + } else { + _lookahead_revert(lookahead, pos); + break; + } + } + + return n; +} + + +static unsigned int +_parse_dec(struct lookahead_state *lookahead) +{ + unsigned int n = 0; + + for (;;) { + unsigned int pos = _lookahead_tell(lookahead); + int c = _lookahead_getc(lookahead); + + if ((c >= '0' && c <= '9')) { + n++; + } else { + _lookahead_revert(lookahead, pos); + break; + } + } + + return n; +} + + +static int +_tokenise_number(struct sl_pp_context *context, + struct sl_pp_token_info *out) +{ + struct lookahead_state lookahead; + unsigned int eaten; + unsigned int is_float = 0; + unsigned int pos; + int c; + char number[256]; /* XXX: Remove this artifical limit. */ + + _lookahead_init(&lookahead, context); + + eaten = _parse_float(&lookahead); + if (!eaten) { + eaten = _parse_hex(&lookahead); + if (!eaten) { + eaten = _parse_oct(&lookahead); + if (!eaten) { + eaten = _parse_dec(&lookahead); + } + } + } else { + is_float = 1; + } + + if (!eaten) { + strcpy(context->error_msg, "expected a number"); + return -1; + } + + pos = _lookahead_tell(&lookahead); + c = _lookahead_getc(&lookahead); + _lookahead_revert(&lookahead, pos); + + if (_is_identifier_char(c)) { + strcpy(context->error_msg, "expected a number"); + _lookahead_revert(&lookahead, 0); + return -1; + } + + if (eaten > sizeof(number) - 1) { + strcpy(context->error_msg, "out of memory"); + _lookahead_revert(&lookahead, 0); + return -1; + } + + assert(_lookahead_tell(&lookahead) == eaten); + + memcpy(number, _lookahead_buf(&lookahead), eaten); + number[eaten] = '\0'; + + if (is_float) { + out->token = SL_PP_FLOAT; + out->data._float = sl_pp_context_add_unique_str(context, number); + if (out->data._float == -1) { + _lookahead_revert(&lookahead, 0); + return -1; + } + } else { + out->token = SL_PP_UINT; + out->data._uint = sl_pp_context_add_unique_str(context, number); + if (out->data._uint == -1) { + _lookahead_revert(&lookahead, 0); + return -1; + } + } + + return 0; +} + + +int +sl_pp_token_get(struct sl_pp_context *context, + struct sl_pp_token_info *out) +{ + int c = _pure_getc(context); + + switch (c) { + case ' ': + case '\t': + out->token = SL_PP_WHITESPACE; + break; + + case '\n': + out->token = SL_PP_NEWLINE; + break; + + case '#': + out->token = SL_PP_HASH; + break; + + case ',': + out->token = SL_PP_COMMA; + break; + + case ';': + out->token = SL_PP_SEMICOLON; + break; + + case '{': + out->token = SL_PP_LBRACE; + break; + + case '}': + out->token = SL_PP_RBRACE; + break; + + case '(': + out->token = SL_PP_LPAREN; + break; + + case ')': + out->token = SL_PP_RPAREN; + break; + + case '[': + out->token = SL_PP_LBRACKET; + break; + + case ']': + out->token = SL_PP_RBRACKET; + break; + + case '.': + { + int c2 = _pure_getc(context); + + if (c2 == PURE_ERROR) { + return -1; + } + if (c2 >= '0' && c2 <= '9') { + _pure_ungetc(context, c2); + _pure_ungetc(context, c); + if (_tokenise_number(context, out)) { + return -1; + } + } else { + _pure_ungetc(context, c2); + out->token = SL_PP_DOT; + } + } + break; + + case '+': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '+') { + out->token = SL_PP_INCREMENT; + } else if (c == '=') { + out->token = SL_PP_ADDASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_PLUS; + } + break; + + case '-': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '-') { + out->token = SL_PP_DECREMENT; + } else if (c == '=') { + out->token = SL_PP_SUBASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_MINUS; + } + break; + + case '~': + out->token = SL_PP_BITNOT; + break; + + case '!': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_NOTEQUAL; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_NOT; + } + break; + + case '*': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_MULASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_STAR; + } + break; + + case '/': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_DIVASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_SLASH; + } + break; + + case '%': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_MODASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_MODULO; + } + break; + + case '<': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '<') { + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_LSHIFTASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_LSHIFT; + } + } else if (c == '=') { + out->token = SL_PP_LESSEQUAL; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_LESS; + } + break; + + case '>': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '>') { + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_RSHIFTASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_RSHIFT; + } + } else if (c == '=') { + out->token = SL_PP_GREATEREQUAL; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_GREATER; + } + break; + + case '=': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '=') { + out->token = SL_PP_EQUAL; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_ASSIGN; + } + break; + + case '&': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '&') { + out->token = SL_PP_AND; + } else if (c == '=') { + out->token = SL_PP_BITANDASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_BITAND; + } + break; + + case '^': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '^') { + out->token = SL_PP_XOR; + } else if (c == '=') { + out->token = SL_PP_BITXORASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_BITXOR; + } + break; + + case '|': + c = _pure_getc(context); + if (c == PURE_ERROR) { + return -1; + } + if (c == '|') { + out->token = SL_PP_OR; + } else if (c == '=') { + out->token = SL_PP_BITORASSIGN; + } else { + _pure_ungetc(context, c); + out->token = SL_PP_BITOR; + } + break; + + case '?': + out->token = SL_PP_QUESTION; + break; + + case ':': + out->token = SL_PP_COLON; + break; + + case '\0': + out->token = SL_PP_EOF; + break; + + case PURE_ERROR: + return -1; + + default: + if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_') { + _pure_ungetc(context, c); + if (_tokenise_identifier(context, out)) { + return -1; + } + } else if (c >= '0' && c <= '9') { + _pure_ungetc(context, c); + if (_tokenise_number(context, out)) { + return -1; + } + } else { + out->data.other = c; + out->token = SL_PP_OTHER; + } + } + + return 0; +} + + +int +sl_pp_tokenise(struct sl_pp_context *context, + struct sl_pp_token_info **output) +{ + struct sl_pp_token_info *out = NULL; + unsigned int out_len = 0; + unsigned int out_max = 0; + + for (;;) { + struct sl_pp_token_info info; + + if (sl_pp_token_buffer_get(&context->tokens, &info)) { + free(out); + return -1; + } + + if (out_len >= out_max) { + unsigned int new_max = out_max; + + if (new_max < 0x100) { + new_max = 0x100; + } else if (new_max < 0x10000) { + new_max *= 2; + } else { + new_max += 0x10000; + } + + out = realloc(out, new_max * sizeof(struct sl_pp_token_info)); + if (!out) { + strcpy(context->error_msg, "out of memory"); + return -1; + } + out_max = new_max; + } + + out[out_len++] = info; + + if (info.token == SL_PP_EOF) { + break; + } + } + + *output = out; + return 0; +} diff --git a/src/glsl/pp/sl_pp_token.h b/src/glsl/pp/sl_pp_token.h new file mode 100644 index 00000000000..a12b1934018 --- /dev/null +++ b/src/glsl/pp/sl_pp_token.h @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_TOKEN_H +#define SL_PP_TOKEN_H + + +struct sl_pp_context; + +enum sl_pp_token { + SL_PP_WHITESPACE, + SL_PP_NEWLINE, + SL_PP_HASH, /* # */ + + SL_PP_COMMA, /* , */ + SL_PP_SEMICOLON, /* ; */ + SL_PP_LBRACE, /* { */ + SL_PP_RBRACE, /* } */ + SL_PP_LPAREN, /* ( */ + SL_PP_RPAREN, /* ) */ + SL_PP_LBRACKET, /* [ */ + SL_PP_RBRACKET, /* ] */ + SL_PP_DOT, /* . */ + SL_PP_INCREMENT, /* ++ */ + SL_PP_ADDASSIGN, /* += */ + SL_PP_PLUS, /* + */ + SL_PP_DECREMENT, /* -- */ + SL_PP_SUBASSIGN, /* -= */ + SL_PP_MINUS, /* - */ + SL_PP_BITNOT, /* ~ */ + SL_PP_NOTEQUAL, /* != */ + SL_PP_NOT, /* ! */ + SL_PP_MULASSIGN, /* *= */ + SL_PP_STAR, /* * */ + SL_PP_DIVASSIGN, /* /= */ + SL_PP_SLASH, /* / */ + SL_PP_MODASSIGN, /* %= */ + SL_PP_MODULO, /* % */ + SL_PP_LSHIFTASSIGN, /* <<= */ + SL_PP_LSHIFT, /* << */ + SL_PP_LESSEQUAL, /* <= */ + SL_PP_LESS, /* < */ + SL_PP_RSHIFTASSIGN, /* >>= */ + SL_PP_RSHIFT, /* >> */ + SL_PP_GREATEREQUAL, /* >= */ + SL_PP_GREATER, /* > */ + SL_PP_EQUAL, /* == */ + SL_PP_ASSIGN, /* = */ + SL_PP_AND, /* && */ + SL_PP_BITANDASSIGN, /* &= */ + SL_PP_BITAND, /* & */ + SL_PP_XOR, /* ^^ */ + SL_PP_BITXORASSIGN, /* ^= */ + SL_PP_BITXOR, /* ^ */ + SL_PP_OR, /* || */ + SL_PP_BITORASSIGN, /* |= */ + SL_PP_BITOR, /* | */ + SL_PP_QUESTION, /* ? */ + SL_PP_COLON, /* : */ + + SL_PP_IDENTIFIER, + + SL_PP_UINT, + SL_PP_FLOAT, + + SL_PP_OTHER, + + SL_PP_PRAGMA_OPTIMIZE, + SL_PP_PRAGMA_DEBUG, + + SL_PP_EXTENSION_REQUIRE, + SL_PP_EXTENSION_ENABLE, + SL_PP_EXTENSION_WARN, + SL_PP_EXTENSION_DISABLE, + + SL_PP_LINE, + + SL_PP_EOF +}; + +union sl_pp_token_data { + int identifier; + int _uint; + int _float; + char other; + int pragma; + int extension; + struct { + unsigned int lineno: 24; + unsigned int fileno: 8; + } line; +}; + +struct sl_pp_token_info { + enum sl_pp_token token; + union sl_pp_token_data data; +}; + +struct sl_pp_purify_options; + +int +sl_pp_token_get(struct sl_pp_context *context, + struct sl_pp_token_info *out); + +int +sl_pp_tokenise(struct sl_pp_context *context, + struct sl_pp_token_info **output); + +#endif /* SL_PP_TOKEN_H */ diff --git a/src/glsl/pp/sl_pp_token_util.h b/src/glsl/pp/sl_pp_token_util.h new file mode 100644 index 00000000000..1685d243a5b --- /dev/null +++ b/src/glsl/pp/sl_pp_token_util.h @@ -0,0 +1,211 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SL_PP_TOKEN_UTIL_H +#define SL_PP_TOKEN_UTIL_H + +#include <assert.h> +#include <stdlib.h> +#include "sl_pp_token.h" + + +struct sl_pp_context; + +/* + * A token buffer allows one to get and unget a token + * from a preprocessor context. + */ +struct sl_pp_token_buffer { + struct sl_pp_context *context; + unsigned int size; + unsigned int capacity; + struct sl_pp_token_info *tokens; +}; + +static int +sl_pp_token_buffer_init(struct sl_pp_token_buffer *buffer, + struct sl_pp_context *context) +{ + buffer->context = context; + buffer->size = 0; + buffer->capacity = 64; + buffer->tokens = malloc(buffer->capacity * sizeof(struct sl_pp_token_info)); + if (!buffer->tokens) { + return -1; + } + return 0; +} + +static void +sl_pp_token_buffer_destroy(struct sl_pp_token_buffer *buffer) +{ + free(buffer->tokens); +} + +static int +sl_pp_token_buffer_get(struct sl_pp_token_buffer *buffer, + struct sl_pp_token_info *out) +{ + /* Pop from stack first if not empty. */ + if (buffer->size) { + *out = buffer->tokens[--buffer->size]; + return 0; + } + + assert(buffer->context); + return sl_pp_token_get(buffer->context, out); +} + +static void +sl_pp_token_buffer_unget(struct sl_pp_token_buffer *buffer, + const struct sl_pp_token_info *in) +{ + /* Resize if needed. */ + if (buffer->size == buffer->capacity) { + buffer->capacity += 64; + buffer->tokens = realloc(buffer->tokens, + buffer->capacity * sizeof(struct sl_pp_token_info)); + assert(buffer->tokens); + } + + /* Push token on stack. */ + buffer->tokens[buffer->size++] = *in; +} + +static int +sl_pp_token_buffer_skip_white(struct sl_pp_token_buffer *buffer, + struct sl_pp_token_info *out) +{ + if (sl_pp_token_buffer_get(buffer, out)) { + return -1; + } + + while (out->token == SL_PP_WHITESPACE) { + if (sl_pp_token_buffer_get(buffer, out)) { + return -1; + } + } + + return 0; +} + + +/* + * A token peek allows one to get a number of tokens from a buffer + * and then either commit the operation or abort it, + * effectively ungetting the peeked tokens. + */ +struct sl_pp_token_peek { + struct sl_pp_token_buffer *buffer; + unsigned int size; + unsigned int capacity; + struct sl_pp_token_info *tokens; +}; + +static int +sl_pp_token_peek_init(struct sl_pp_token_peek *peek, + struct sl_pp_token_buffer *buffer) +{ + peek->buffer = buffer; + peek->size = 0; + peek->capacity = 64; + peek->tokens = malloc(peek->capacity * sizeof(struct sl_pp_token_info)); + if (!peek->tokens) { + return -1; + } + return 0; +} + +static void +sl_pp_token_peek_destroy(struct sl_pp_token_peek *peek) +{ + /* Abort. */ + while (peek->size) { + sl_pp_token_buffer_unget(peek->buffer, &peek->tokens[--peek->size]); + } + free(peek->tokens); +} + +static int +sl_pp_token_peek_get(struct sl_pp_token_peek *peek, + struct sl_pp_token_info *out) +{ + /* Get token from buffer. */ + if (sl_pp_token_buffer_get(peek->buffer, out)) { + return -1; + } + + /* Save it. */ + if (peek->size == peek->capacity) { + peek->capacity += 64; + peek->tokens = realloc(peek->tokens, + peek->capacity * sizeof(struct sl_pp_token_info)); + assert(peek->tokens); + } + peek->tokens[peek->size++] = *out; + return 0; +} + +static void +sl_pp_token_peek_commit(struct sl_pp_token_peek *peek) +{ + peek->size = 0; +} + +static int +sl_pp_token_peek_to_buffer(const struct sl_pp_token_peek *peek, + struct sl_pp_token_buffer *buffer) +{ + unsigned int i; + + if (sl_pp_token_buffer_init(buffer, NULL)) { + return -1; + } + for (i = peek->size; i > 0; i--) { + sl_pp_token_buffer_unget(buffer, &peek->tokens[i - 1]); + } + return 0; +} + +static int +sl_pp_token_peek_skip_white(struct sl_pp_token_peek *peek, + struct sl_pp_token_info *out) +{ + if (sl_pp_token_peek_get(peek, out)) { + return -1; + } + + while (out->token == SL_PP_WHITESPACE) { + if (sl_pp_token_peek_get(peek, out)) { + return -1; + } + } + + return 0; +} + +#endif /* SL_PP_TOKEN_UTIL_H */ diff --git a/src/glsl/pp/sl_pp_version.c b/src/glsl/pp/sl_pp_version.c new file mode 100644 index 00000000000..3c995b77501 --- /dev/null +++ b/src/glsl/pp/sl_pp_version.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <stdlib.h> +#include <string.h> +#include "sl_pp_public.h" +#include "sl_pp_context.h" + + +int +sl_pp_version(struct sl_pp_context *context, + unsigned int *version) +{ + struct sl_pp_token_peek peek; + unsigned int line = context->line; + + /* Default values if `#version' is not present. */ + *version = 110; + + if (sl_pp_token_peek_init(&peek, &context->tokens)) { + return -1; + } + + /* There can be multiple `#version' directives present. + * Accept the value of the last one. + */ + for (;;) { + struct sl_pp_token_info input; + int found_hash = 0; + int found_version = 0; + int found_number = 0; + int found_end = 0; + + /* Skip whitespace and newlines and seek for hash. */ + while (!found_hash) { + if (sl_pp_token_peek_get(&peek, &input)) { + sl_pp_token_peek_destroy(&peek); + return -1; + } + + switch (input.token) { + case SL_PP_NEWLINE: + line++; + break; + + case SL_PP_WHITESPACE: + break; + + case SL_PP_HASH: + found_hash = 1; + break; + + default: + sl_pp_token_peek_destroy(&peek); + return 0; + } + } + + /* Skip whitespace and seek for `version'. */ + while (!found_version) { + if (sl_pp_token_peek_get(&peek, &input)) { + sl_pp_token_peek_destroy(&peek); + return -1; + } + + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_IDENTIFIER: + if (input.data.identifier != context->dict.version) { + sl_pp_token_peek_destroy(&peek); + return 0; + } + found_version = 1; + break; + + default: + sl_pp_token_peek_destroy(&peek); + return 0; + } + } + + sl_pp_token_peek_commit(&peek); + + /* Skip whitespace and seek for version number. */ + while (!found_number) { + if (sl_pp_token_buffer_get(&context->tokens, &input)) { + sl_pp_token_peek_destroy(&peek); + return -1; + } + + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_UINT: + *version = atoi(sl_pp_context_cstr(context, input.data._uint)); + found_number = 1; + break; + + default: + strcpy(context->error_msg, "expected version number after `#version'"); + sl_pp_token_peek_destroy(&peek); + return -1; + } + } + + /* Skip whitespace and seek for either newline or eof. */ + while (!found_end) { + if (sl_pp_token_buffer_get(&context->tokens, &input)) { + sl_pp_token_peek_destroy(&peek); + return -1; + } + + switch (input.token) { + case SL_PP_WHITESPACE: + break; + + case SL_PP_NEWLINE: + line++; + /* pass thru */ + case SL_PP_EOF: + context->line = line; + found_end = 1; + break; + + default: + strcpy(context->error_msg, "expected end of line after version number"); + sl_pp_token_peek_destroy(&peek); + return -1; + } + } + } + + /* Should not get here. */ +} diff --git a/src/glut/glx/glut_menu.c b/src/glut/glx/glut_menu.c index 4c4a5ae7503..d136823c54f 100644 --- a/src/glut/glx/glut_menu.c +++ b/src/glut/glx/glut_menu.c @@ -225,6 +225,7 @@ menuVisualSetup(void) if (!status) { XFreeColormap(__glutDisplay, menuColormap); free(placeHolders); + placeHolders = NULL; continue; } } @@ -241,6 +242,7 @@ menuVisualSetup(void) XFreeColormap(__glutDisplay, menuColormap); if (placeHolders) { free(placeHolders); + placeHolders = NULL; } continue; } @@ -252,6 +254,7 @@ menuVisualSetup(void) XFreeColormap(__glutDisplay, menuColormap); if (placeHolders) { free(placeHolders); + placeHolders = NULL; } continue; } @@ -263,6 +266,7 @@ menuVisualSetup(void) XFreeColormap(__glutDisplay, menuColormap); if (placeHolders) { free(placeHolders); + placeHolders = NULL; } continue; } @@ -271,6 +275,7 @@ menuVisualSetup(void) XFreeColors(__glutDisplay, menuColormap, placeHolders, numPlaceHolders, 0); free(placeHolders); + placeHolders = NULL; } menuWhite = color.pixel; menuVisual = visual->vinfo.visual; diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 6f58ad61617..a815f46b4a1 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -18,11 +18,10 @@ include sources.mak - -# Default: build dependencies, then asm_subdirs, then convenience -# libs (.a) and finally the device drivers: -default: depend asm_subdirs libmesa.a libmesagallium.a libglapi.a \ - driver_subdirs +# Default: build dependencies, then asm_subdirs, GLSL built-in lib, +# then convenience libs (.a) and finally the device drivers: +default: depend asm_subdirs glsl_builtin libmesa.a libmesagallium.a \ + libglapi.a driver_subdirs @@ -30,12 +29,12 @@ default: depend asm_subdirs libmesa.a libmesagallium.a libglapi.a \ # Helper libraries used by many drivers: # Make archive of core mesa object files -libmesa.a: $(MESA_OBJECTS) - @ $(MKLIB) -o mesa -static $(MESA_OBJECTS) +libmesa.a: $(MESA_OBJECTS) $(GLSL_LIBS) + @ $(MKLIB) -o mesa -static $(MESA_OBJECTS) $(GLSL_LIBS) # Make archive of subset of core mesa object files for gallium -libmesagallium.a: $(MESA_GALLIUM_OBJECTS) - @ $(MKLIB) -o mesagallium -static $(MESA_GALLIUM_OBJECTS) +libmesagallium.a: $(MESA_GALLIUM_OBJECTS) $(GLSL_LIBS) + @ $(MKLIB) -o mesagallium -static $(MESA_GALLIUM_OBJECTS) $(GLSL_LIBS) # Make archive of gl* API dispatcher functions only libglapi.a: $(GLAPI_OBJECTS) @@ -60,6 +59,12 @@ asm_subdirs: ###################################################################### +# GLSL built-in library +glsl_builtin: + (cd shader/slang/library && $(MAKE)) || exit 1 ; + + +###################################################################### # Dependency generation depend: $(ALL_SOURCES) @@ -152,6 +157,7 @@ clean: -rm -f depend depend.bak libmesa.a libglapi.a -rm -f drivers/*/*.o -rm -f *.pc + -rm -f shader/slang/library/*_gc.h -@cd drivers/dri && $(MAKE) clean -@cd drivers/x11 && $(MAKE) clean -@cd drivers/osmesa && $(MAKE) clean diff --git a/src/mesa/Makefile.mgw b/src/mesa/Makefile.mgw index 097c390a83e..e894c6277d1 100644 --- a/src/mesa/Makefile.mgw +++ b/src/mesa/Makefile.mgw @@ -218,7 +218,6 @@ clean: -$(call UNLINK,vbo/*.o) -$(call UNLINK,shader/*.o) -$(call UNLINK,shader/slang/*.o) - -$(call UNLINK,shader/grammar/*.o) -$(call UNLINK,sparc/*.o) -$(call UNLINK,ppc/*.o) -$(call UNLINK,swrast/*.o) diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 309e0e54d07..f4e0b98570d 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -164,7 +164,6 @@ if env['platform'] != 'winddk': 'state_tracker/st_cb_flush.c', 'state_tracker/st_cb_drawpixels.c', 'state_tracker/st_cb_fbo.c', - 'state_tracker/st_cb_get.c', 'state_tracker/st_cb_feedback.c', 'state_tracker/st_cb_program.c', 'state_tracker/st_cb_queryobj.c', @@ -190,7 +189,6 @@ if env['platform'] != 'winddk': 'shader/arbprogparse.c', 'shader/arbprogram.c', 'shader/atifragshader.c', - 'shader/grammar/grammar_mesa.c', 'shader/hash_table.c', 'shader/lex.yy.c', 'shader/nvfragparse.c', @@ -228,7 +226,6 @@ if env['platform'] != 'winddk': 'shader/slang/slang_link.c', 'shader/slang/slang_log.c', 'shader/slang/slang_mem.c', - 'shader/slang/slang_preprocess.c', 'shader/slang/slang_print.c', 'shader/slang/slang_simplify.c', 'shader/slang/slang_storage.c', @@ -339,7 +336,9 @@ if env['platform'] != 'winddk': # Add the dir containing the generated header (somewhere inside the # build dir) to the include path env.Append(CPPPATH = [matypes[0].dir]) - + + SConscript('shader/slang/library/SConscript') + # # Libraries # diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 4ca0e7bcc37..9b271f85e92 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -124,7 +124,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver) driver->UnmapTexture = NULL; driver->TextureMemCpy = _mesa_memcpy; driver->IsTextureResident = NULL; - driver->ActiveTexture = NULL; driver->UpdateTexturePalette = NULL; /* imaging */ diff --git a/src/mesa/drivers/common/meta.c b/src/mesa/drivers/common/meta.c index a4315191434..cd9075b3936 100644 --- a/src/mesa/drivers/common/meta.c +++ b/src/mesa/drivers/common/meta.c @@ -2149,6 +2149,7 @@ _mesa_meta_GenerateMipmap(GLcontext *ctx, GLenum target, const GLenum wrapTSave = texObj->WrapT; const GLenum wrapRSave = texObj->WrapR; const GLuint fboSave = ctx->DrawBuffer->Name; + const GLuint original_active_unit = ctx->Texture.CurrentUnit; GLenum faceTarget; GLuint dstLevel; GLuint border = 0; @@ -2169,6 +2170,9 @@ _mesa_meta_GenerateMipmap(GLcontext *ctx, GLenum target, _mesa_meta_begin(ctx, META_ALL); + if (original_active_unit != 0) + _mesa_BindTexture(target, texObj->Name); + if (mipmap->ArrayObj == 0) { /* one-time setup */ diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index da81ec9de51..547f18a009e 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -167,7 +167,7 @@ static int driBindContext(__DRIcontext *pcp, __DRIdrawable *pdp, __DRIdrawable *prp) { - __DRIscreenPrivate *psp; + __DRIscreenPrivate *psp = NULL; /* Bind the drawable to the context */ diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index 95f97414a98..447f3d15b95 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -356,6 +356,63 @@ } while (0) # endif +#elif (SPANTMP_PIXEL_FMT == GL_BGR) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV) + +/** + ** GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV + ** + ** This is really for MESA_FORMAT_XRGB8888. The spantmp code needs to be + ** kicked to the curb, and we need to just code-gen this. + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) ( buf + (_x) * 4 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLuint *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLuint *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +# define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_8888(0xff, color[0], color[1], color[2]) + +# define WRITE_RGBA(_x, _y, r, g, b, a) \ + PUT_VALUE(_x, _y, ((r << 16) | \ + (g << 8) | \ + (b << 0) | \ + (0xff << 24))) + +#define WRITE_PIXEL(_x, _y, p) PUT_VALUE(_x, _y, p) + +# if defined( USE_X86_ASM ) +# define READ_RGBA(rgba, _x, _y) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + __asm__ __volatile__( "bswap %0; rorl $8, %0" \ + : "=r" (p) : "0" (p) ); \ + ((GLuint *)rgba)[0] = p | 0xff000000; \ + } while (0) +# elif defined( MESA_BIG_ENDIAN ) + /* On PowerPC with GCC 3.4.2 the shift madness below becomes a single + * rotlwi instruction. It also produces good code on SPARC. + */ +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + *((uint32_t *) rgba) = (t << 8) | 0xff; \ + } while (0) +# else +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + rgba[0] = (p >> 16) & 0xff; \ + rgba[1] = (p >> 8) & 0xff; \ + rgba[2] = (p >> 0) & 0xff; \ + rgba[3] = 0xff; \ + } while (0) +# endif + #else #error SPANTMP_PIXEL_FMT must be set to a valid value! #endif diff --git a/src/mesa/drivers/dri/i810/i810tris.c b/src/mesa/drivers/dri/i810/i810tris.c index b508496fb69..213ba541cee 100644 --- a/src/mesa/drivers/dri/i810/i810tris.c +++ b/src/mesa/drivers/dri/i810/i810tris.c @@ -270,7 +270,8 @@ do { \ #define LOCAL_VARS(n) \ i810ContextPtr imesa = I810_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint coloroffset = (imesa->vertex_size == 4 ? 3 : 4); \ GLboolean havespec = (imesa->vertex_size > 4); \ (void) color; (void) spec; (void) coloroffset; (void) havespec; diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index c62281d341e..27c5aa1e085 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -56,10 +56,7 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format) case MESA_FORMAT_ARGB4444: return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return MAPSURF_32BIT | MT_32BIT_XRGB8888; - else - return MAPSURF_32BIT | MT_32BIT_ARGB8888; + return MAPSURF_32BIT | MT_32BIT_ARGB8888; case MESA_FORMAT_XRGB8888: return MAPSURF_32BIT | MT_32BIT_XRGB8888; case MESA_FORMAT_YCBCR_REV: diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index c05c7759ac5..1e3c8301d8d 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -645,7 +645,7 @@ i830_state_draw_region(struct intel_context *intel, DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z); /* .5 */ if (irb != NULL) { - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: value |= DV_PF_8888; @@ -661,7 +661,7 @@ i830_state_draw_region(struct intel_context *intel, break; default: _mesa_problem(ctx, "Bad renderbuffer format: %d\n", - irb->texformat); + irb->Base.Format); } } diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 7d4c7cfbabb..0485be2cc1f 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -143,6 +143,9 @@ i915CreateContext(const __GLcontextModes * mesaVis, ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS; ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS; ctx->Const.MaxVarying = I915_TEX_UNITS; + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Advertise the full hardware capabilities. The new memory * manager should cope much better with overload situations: diff --git a/src/mesa/drivers/dri/i915/i915_debug.c b/src/mesa/drivers/dri/i915/i915_debug.c index f7bb7ea44c9..fecfac30339 100644 --- a/src/mesa/drivers/dri/i915/i915_debug.c +++ b/src/mesa/drivers/dri/i915/i915_debug.c @@ -806,6 +806,7 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) default: return debug(stream, "", 0); } + break; default: assert(0); return 0; diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index 1bacd51aec5..221bf033327 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -57,10 +57,7 @@ translate_texture_format(gl_format mesa_format, GLuint internal_format, case MESA_FORMAT_ARGB4444: return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return MAPSURF_32BIT | MT_32BIT_XRGB8888; - else - return MAPSURF_32BIT | MT_32BIT_ARGB8888; + return MAPSURF_32BIT | MT_32BIT_ARGB8888; case MESA_FORMAT_XRGB8888: return MAPSURF_32BIT | MT_32BIT_XRGB8888; case MESA_FORMAT_YCBCR_REV: @@ -142,6 +139,7 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) GLuint *state = i915->state.Tex[unit], format, pitch; GLint lodbias, aniso = 0; GLubyte border[4]; + GLfloat maxlod; memset(state, 0, sizeof(state)); @@ -179,18 +177,9 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) pitch = intelObj->pitchOverride; } else { - GLuint dst_x, dst_y; - - intel_miptree_get_image_offset(intelObj->mt, intelObj->firstLevel, 0, 0, - &dst_x, &dst_y); - dri_bo_reference(intelObj->mt->region->buffer); i915->state.tex_buffer[unit] = intelObj->mt->region->buffer; - /* XXX: This calculation is probably broken for tiled images with - * a non-page-aligned offset. - */ - i915->state.tex_offset[unit] = (dst_x + dst_y * intelObj->mt->pitch) * - intelObj->mt->cpp; + i915->state.tex_offset[unit] = 0; /* Always the origin of the miptree */ format = translate_texture_format(firstImage->TexFormat, firstImage->InternalFormat, @@ -208,10 +197,14 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) state[I915_TEXREG_MS3] |= MS3_TILE_WALK; } + /* We get one field with fraction bits to cover the maximum addressable (smallest + * resolution) LOD. Use it to cover both MAX_LEVEL and MAX_LOD. + */ + maxlod = MIN2(tObj->MaxLod, tObj->MaxLevel - tObj->BaseLevel); state[I915_TEXREG_MS4] = ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK | - (U_FIXED(CLAMP(tObj->MaxLod, 0.0, 11.0), 2) << MS4_MAX_LOD_SHIFT) | + (U_FIXED(CLAMP(maxlod, 0.0, 11.0), 2) << MS4_MAX_LOD_SHIFT) | ((firstImage->Depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index 3e7b5101cca..ba6be9796e1 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -587,7 +587,7 @@ i915_state_draw_region(struct intel_context *intel, DSTORG_VERT_BIAS(0x8) | /* .5 */ LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL); if (irb != NULL) { - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: value |= DV_PF_8888; @@ -603,7 +603,7 @@ i915_state_draw_region(struct intel_context *intel, break; default: _mesa_problem(ctx, "Bad renderbuffer format: %d\n", - irb->texformat); + irb->Base.Format); } } diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c index d4ccd28c9e8..bac1c3a49c3 100644 --- a/src/mesa/drivers/dri/i965/brw_cc.c +++ b/src/mesa/drivers/dri/i965/brw_cc.c @@ -34,6 +34,7 @@ #include "brw_state.h" #include "brw_defines.h" #include "brw_util.h" +#include "intel_fbo.h" #include "main/macros.h" #include "main/enums.h" @@ -89,6 +90,28 @@ struct brw_cc_unit_key { GLenum depth_func; }; +/** + * Modify blend function to force destination alpha to 1.0 + * + * If \c function specifies a blend function that uses destination alpha, + * replace it with a function that hard-wires destination alpha to 1.0. This + * is used when rendering to xRGB targets. + */ +static GLenum +fix_xRGB_alpha(GLenum function) +{ + switch (function) { + case GL_DST_ALPHA: + return GL_ONE; + + case GL_ONE_MINUS_DST_ALPHA: + case GL_SRC_ALPHA_SATURATE: + return GL_ZERO; + } + + return function; +} + static void cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key) { @@ -132,6 +155,17 @@ cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key) key->blend_dst_rgb = ctx->Color.BlendDstRGB; key->blend_src_a = ctx->Color.BlendSrcA; key->blend_dst_a = ctx->Color.BlendDstA; + + /* If the renderbuffer is XRGB, we have to frob the blend function to + * force the destination alpha to 1.0. This means replacing GL_DST_ALPHA + * with GL_ONE and GL_ONE_MINUS_DST_ALPHA with GL_ZERO. + */ + if (ctx->DrawBuffer->Visual.alphaBits == 0) { + key->blend_src_rgb = fix_xRGB_alpha(key->blend_src_rgb); + key->blend_src_a = fix_xRGB_alpha(key->blend_src_a); + key->blend_dst_rgb = fix_xRGB_alpha(key->blend_dst_rgb); + key->blend_dst_a = fix_xRGB_alpha(key->blend_dst_a); + } } key->alpha_enabled = ctx->Color.AlphaEnabled; diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 8bdda60697b..78bea829493 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -111,7 +111,9 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits, ctx->Const.MaxTextureImageUnits); ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */ - ctx->Const.MaxCombinedTextureImageUnits = 0; + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Mesa limits textures to 4kx4k; it would be nice to fix that someday */ diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 00efd3443d1..27aac8b54a9 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -392,6 +392,17 @@ static void emit_sge( struct brw_vs_compile *c, emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_GE); } +static void emit_cmp( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + struct brw_reg arg2 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, brw_imm_f(0)); + brw_SEL(p, dst, arg1, arg2); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + static void emit_max( struct brw_compile *p, struct brw_reg dst, struct brw_reg arg0, @@ -1485,6 +1496,9 @@ void brw_vs_emit(struct brw_vs_compile *c ) brw_MOV(p, brw_acc_reg(), args[2]); brw_MAC(p, dst, args[0], args[1]); break; + case OPCODE_CMP: + emit_cmp(p, dst, args[0], args[1], args[2]); + break; case OPCODE_MAX: emit_max(p, dst, args[0], args[1]); break; diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index 47035cc6fc1..8335e5a650e 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -94,20 +94,14 @@ static GLuint translate_tex_format( gl_format mesa_format, return BRW_SURFACEFORMAT_R8G8B8_UNORM; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return BRW_SURFACEFORMAT_B8G8R8X8_UNORM; - else - return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; case MESA_FORMAT_XRGB8888: return BRW_SURFACEFORMAT_B8G8R8X8_UNORM; case MESA_FORMAT_RGBA8888_REV: _mesa_problem(NULL, "unexpected format in i965:translate_tex_format()"); - if (internal_format == GL_RGB) - return BRW_SURFACEFORMAT_R8G8B8X8_UNORM; - else - return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; case MESA_FORMAT_RGB565: return BRW_SURFACEFORMAT_B5G6R5_UNORM; @@ -537,12 +531,16 @@ brw_update_renderbuffer_surface(struct brw_context *brw, region_bo = region->buffer; key.surface_type = BRW_SURFACE_2D; - switch (irb->texformat) { + switch (irb->Base.Format) { + /* XRGB and ARGB are treated the same here because the chips in this + * family cannot render to XRGB targets. This means that we have to + * mask writes to alpha (ala glColorMask) and reconfigure the alpha + * blending hardware to use GL_ONE (or GL_ZERO) for cases where + * GL_DST_ALPHA (or GL_ONE_MINUS_DST_ALPHA) is used. + */ case MESA_FORMAT_ARGB8888: - key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - break; case MESA_FORMAT_XRGB8888: - key.surface_format = BRW_SURFACEFORMAT_B8G8R8X8_UNORM; + key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; break; case MESA_FORMAT_RGB565: key.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; @@ -554,7 +552,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw, key.surface_format = BRW_SURFACEFORMAT_B4G4R4A4_UNORM; break; default: - _mesa_problem(ctx, "Bad renderbuffer format: %d\n", irb->texformat); + _mesa_problem(ctx, "Bad renderbuffer format: %d\n", irb->Base.Format); } key.tiling = region->tiling; if (brw->intel.intelScreen->driScrnPriv->dri2.enabled) { @@ -579,6 +577,13 @@ brw_update_renderbuffer_surface(struct brw_context *brw, /* _NEW_COLOR */ memcpy(key.color_mask, ctx->Color.ColorMask, sizeof(key.color_mask)); + + /* As mentioned above, disable writes to the alpha component when the + * renderbuffer is XRGB. + */ + if (ctx->DrawBuffer->Visual.alphaBits == 0) + key.color_mask[3] = GL_FALSE; + key.color_blend = (!ctx->Color._LogicOpEnabled && ctx->Color.BlendEnabled); diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index f14854602b6..cdf1408cd33 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -496,7 +496,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: clearVal = PACK_COLOR_8888(clear[3], clear[0], @@ -515,7 +515,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) break; default: _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n", - irb->texformat); + irb->Base.Format); clearVal = 0; } } diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c index 6b12d484d85..05643189a21 100644 --- a/src/mesa/drivers/dri/intel/intel_buffers.c +++ b/src/mesa/drivers/dri/intel/intel_buffers.c @@ -191,10 +191,15 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) return; } - /* - * How many color buffers are we drawing into? + /* How many color buffers are we drawing into? + * + * If there are zero buffers or the buffer is too big, don't configure any + * regions for hardware drawing. We'll fallback to software below. Not + * having regions set makes some of the software fallback paths faster. */ - if (fb->_NumColorDrawBuffers == 0) { + if ((fb->Width > ctx->Const.MaxRenderbufferSize) + || (fb->Height > ctx->Const.MaxRenderbufferSize) + || (fb->_NumColorDrawBuffers == 0)) { /* writing to 0 */ colorRegions[0] = NULL; intel->constant_cliprect = GL_TRUE; diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 1434ae530be..c616d32ebbf 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -68,7 +68,7 @@ int INTEL_DEBUG = (0); #endif -#define DRIVER_DATE "20090712 2009Q2 RC3" +#define DRIVER_DATE "20091221 DEVELOPMENT" #define DRIVER_DATE_GEM "GEM " DRIVER_DATE diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c index 5615040946f..608f75b8240 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.c +++ b/src/mesa/drivers/dri/intel/intel_fbo.c @@ -37,6 +37,7 @@ #include "drivers/common/meta.h" #include "intel_context.h" +#include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" @@ -105,8 +106,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct intel_context *intel = intel_context(ctx); struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; int cpp; + GLuint pitch; ASSERT(rb->Name != 0); @@ -116,18 +117,14 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB5: rb->Format = MESA_FORMAT_RGB565; rb->DataType = GL_UNSIGNED_BYTE; - irb->texformat = MESA_FORMAT_RGB565; - cpp = 2; break; case GL_RGB: case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - rb->Format = MESA_FORMAT_ARGB8888; + rb->Format = MESA_FORMAT_XRGB8888; rb->DataType = GL_UNSIGNED_BYTE; - irb->texformat = MESA_FORMAT_ARGB8888; /* XXX: Need xrgb8888 */ - cpp = 4; break; case GL_RGBA: case GL_RGBA2: @@ -139,8 +136,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGBA16: rb->Format = MESA_FORMAT_ARGB8888; rb->DataType = GL_UNSIGNED_BYTE; - irb->texformat = MESA_FORMAT_ARGB8888; - cpp = 4; break; case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: @@ -150,29 +145,21 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, /* alloc a depth+stencil buffer */ rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - cpp = 4; - irb->texformat = MESA_FORMAT_S8_Z24; break; case GL_DEPTH_COMPONENT16: rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_SHORT; - cpp = 2; - irb->texformat = MESA_FORMAT_Z16; break; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - cpp = 4; - irb->texformat = MESA_FORMAT_S8_Z24; break; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - cpp = 4; - irb->texformat = MESA_FORMAT_S8_Z24; break; default: _mesa_problem(ctx, @@ -181,6 +168,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat); + cpp = _mesa_get_format_bytes(rb->Format); intelFlush(ctx); @@ -190,32 +178,25 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); + /* Choose a pitch to match hardware requirements: + */ + pitch = ((cpp * width + 63) & ~63) / cpp; - irb->region = intel_region_alloc(intel, I915_TILING_NONE, - cpp, width, height, pitch, - GL_TRUE); - if (!irb->region) - return GL_FALSE; /* out of memory? */ + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - ASSERT(irb->region->buffer); + irb->region = intel_region_alloc(intel, I915_TILING_NONE, cpp, + width, height, pitch, GL_TRUE); + if (!irb->region) + return GL_FALSE; /* out of memory? */ - rb->Width = width; - rb->Height = height; + ASSERT(irb->region->buffer); - return GL_TRUE; - } + rb->Width = width; + rb->Height = height; + + return GL_TRUE; } @@ -297,7 +278,6 @@ intel_create_renderbuffer(gl_format format) GET_CURRENT_CONTEXT(ctx); struct intel_renderbuffer *irb; - const GLuint name = 0; irb = CALLOC_STRUCT(intel_renderbuffer); if (!irb) { @@ -305,7 +285,7 @@ intel_create_renderbuffer(gl_format format) return NULL; } - _mesa_init_renderbuffer(&irb->Base, name); + _mesa_init_renderbuffer(&irb->Base, 0); irb->Base.ClassID = INTEL_RB_CLASS; switch (format) { @@ -314,10 +294,6 @@ intel_create_renderbuffer(gl_format format) irb->Base.DataType = GL_UNSIGNED_BYTE; break; case MESA_FORMAT_XRGB8888: - /* XXX this is a hack since XRGB surfaces don't seem to work - * properly yet. Reading the alpha channel returns 0 instead of 1. - */ - format = MESA_FORMAT_ARGB8888; irb->Base._BaseFormat = GL_RGB; irb->Base.DataType = GL_UNSIGNED_BYTE; break; @@ -346,7 +322,6 @@ intel_create_renderbuffer(gl_format format) irb->Base.Format = format; irb->Base.InternalFormat = irb->Base._BaseFormat; - irb->texformat = format; /* intel-specific methods */ irb->Base.Delete = intel_delete_renderbuffer; @@ -423,7 +398,6 @@ static GLboolean intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, struct gl_texture_image *texImage) { - irb->texformat = texImage->TexFormat; gl_format texFormat; if (texImage->TexFormat == MESA_FORMAT_ARGB8888) { @@ -591,6 +565,7 @@ static void intel_finish_render_texture(GLcontext * ctx, struct gl_renderbuffer_attachment *att) { + struct intel_context *intel = intel_context(ctx); struct gl_texture_object *tex_obj = att->Texture; struct gl_texture_image *image = tex_obj->Image[att->CubeMapFace][att->TextureLevel]; @@ -598,8 +573,14 @@ intel_finish_render_texture(GLcontext * ctx, /* Flag that this image may now be validated into the object's miptree. */ intel_image->used_as_render_target = GL_FALSE; -} + /* Since we've (probably) rendered to the texture and will (likely) use + * it in the texture domain later on in this batchbuffer, flush the + * batch. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer like GEM does in the kernel. + */ + intel_batchbuffer_emit_mi_flush(intel->batch); +} /** * Do additional "completeness" testing of a framebuffer object. @@ -632,7 +613,7 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) continue; } - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_XRGB8888: case MESA_FORMAT_RGB565: diff --git a/src/mesa/drivers/dri/intel/intel_fbo.h b/src/mesa/drivers/dri/intel/intel_fbo.h index 50a8a959858..fa43077d6a7 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.h +++ b/src/mesa/drivers/dri/intel/intel_fbo.h @@ -62,8 +62,6 @@ struct intel_renderbuffer struct gl_renderbuffer Base; struct intel_region *region; - gl_format texformat; - GLuint vbl_pending; /**< vblank sequence number of pending flip */ uint8_t *span_cache; diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c index 47075001801..20424e2e589 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_read.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c @@ -285,11 +285,11 @@ intelReadPixels(GLcontext * ctx, intelFlush(ctx); -#ifdef I915 if (do_blit_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; +#ifdef I915 if (do_texture_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 2c89a66a95f..d1681e9088a 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -334,7 +334,7 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #include "intel_spantmp.h" /* x8r8g8b8 color span and pixel functions */ -#define INTEL_PIXEL_FMT GL_BGRA +#define INTEL_PIXEL_FMT GL_BGR #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV #define INTEL_READ_VALUE(offset) pread_xrgb8888(irb, offset) #define INTEL_WRITE_VALUE(offset, v) pwrite_xrgb8888(irb, offset, v) @@ -616,7 +616,7 @@ intel_set_span_functions(struct intel_context *intel, uint32_t tiling = irb->region->tiling; if (intel->intelScreen->kernel_exec_fencing) { - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_RGB565: intel_gttmap_InitPointers_RGB565(rb); break; @@ -630,13 +630,7 @@ intel_set_span_functions(struct intel_context *intel, intel_gttmap_InitPointers_xRGB8888(rb); break; case MESA_FORMAT_ARGB8888: - if (rb->_BaseFormat == GL_RGB) { - /* XXX remove this code someday when we enable XRGB surfaces */ - /* 8888 RGBx */ - intel_gttmap_InitPointers_xRGB8888(rb); - } else { - intel_gttmap_InitPointers_ARGB8888(rb); - } + intel_gttmap_InitPointers_ARGB8888(rb); break; case MESA_FORMAT_Z16: intel_gttmap_InitDepthPointers_z16(rb); @@ -659,7 +653,7 @@ intel_set_span_functions(struct intel_context *intel, default: _mesa_problem(NULL, "Unexpected MesaFormat %d in intelSetSpanFunctions", - irb->texformat); + irb->Base.Format); break; } return; @@ -668,7 +662,7 @@ intel_set_span_functions(struct intel_context *intel, /* If in GEM mode, we need to do the tile address swizzling ourselves, * instead of the fence registers handling it. */ - switch (irb->texformat) { + switch (irb->Base.Format) { case MESA_FORMAT_RGB565: switch (tiling) { case I915_TILING_NONE: @@ -726,35 +720,18 @@ intel_set_span_functions(struct intel_context *intel, } break; case MESA_FORMAT_ARGB8888: - if (rb->_BaseFormat == GL_RGB) { - /* XXX remove this code someday when we enable XRGB surfaces */ - /* 8888 RGBx */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_xRGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_xRGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_xRGB8888(rb); - break; - } - } else { - /* 8888 RGBA */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_ARGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_ARGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_ARGB8888(rb); - break; - } + /* 8888 RGBA */ + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_ARGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_ARGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_ARGB8888(rb); + break; } break; case MESA_FORMAT_Z16: diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c index bfa3dba1f5c..87efb72cc51 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_format.c +++ b/src/mesa/drivers/dri/intel/intel_tex_format.c @@ -50,8 +50,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { return MESA_FORMAT_RGB565; } - /* XXX use MESA_FORMAT_XRGB8888 someday */ - return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; + return do32bpt ? MESA_FORMAT_XRGB8888 : MESA_FORMAT_RGB565; case GL_RGBA8: case GL_RGB10_A2: @@ -70,8 +69,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_RGB10: case GL_RGB12: case GL_RGB16: - /* XXX use MESA_FORMAT_XRGB8888 someday */ - return MESA_FORMAT_ARGB8888; + return MESA_FORMAT_XRGB8888; case GL_RGB5: case GL_RGB4: diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c index a757362b11d..72917ee13bf 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.c +++ b/src/mesa/drivers/dri/mach64/mach64_tex.c @@ -565,7 +565,6 @@ void mach64InitTextureFuncs( struct dd_function_table *functions ) functions->IsTextureResident = driIsTextureResident; functions->UpdateTexturePalette = NULL; - functions->ActiveTexture = NULL; driInitTextureFormats(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_tris.c b/src/mesa/drivers/dri/mach64/mach64_tris.c index f2e8e2e3ae8..c2a0adfef02 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tris.c +++ b/src/mesa/drivers/dri/mach64/mach64_tris.c @@ -1297,7 +1297,8 @@ do { \ #define LOCAL_VARS(n) \ mach64ContextPtr mmesa = MACH64_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint vertex_size = mmesa->vertex_size; \ const GLuint xyoffset = 9; \ const GLuint coloroffset = 8; \ diff --git a/src/mesa/drivers/dri/mga/mgatris.c b/src/mesa/drivers/dri/mga/mgatris.c index b93a21c3acf..c1bcd4b8531 100644 --- a/src/mesa/drivers/dri/mga/mgatris.c +++ b/src/mesa/drivers/dri/mga/mgatris.c @@ -397,7 +397,8 @@ do { \ #define LOCAL_VARS(n) \ mgaContextPtr mmesa = MGA_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ (void) color; (void) spec; diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 448e34e047c..86d4717b050 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -216,7 +216,8 @@ do { \ #define LOCAL_VARS(n) \ r128ContextPtr rmesa = R128_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint coloroffset = rmesa->coloroffset; \ GLuint specoffset = rmesa->specoffset; \ GLboolean havespec = (rmesa->specoffset != 0); \ diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile index 776f1e3f3f7..8212dc12031 100644 --- a/src/mesa/drivers/dri/r200/Makefile +++ b/src/mesa/drivers/dri/r200/Makefile @@ -14,7 +14,7 @@ EGL_SOURCES = server/radeon_egl.c endif ifeq ($(RADEON_LDFLAGS),) -CS_SOURCES = radeon_cs_space_drm.c +CS_SOURCES = radeon_cs_space_drm.c radeon_bo.c radeon_cs.c endif RADEON_COMMON_SOURCES = \ diff --git a/src/mesa/drivers/dri/r200/radeon_bo.c b/src/mesa/drivers/dri/r200/radeon_bo.c new file mode 120000 index 00000000000..9448ffee54b --- /dev/null +++ b/src/mesa/drivers/dri/r200/radeon_bo.c @@ -0,0 +1 @@ +../radeon/radeon_bo.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r200/radeon_bo_int_drm.h b/src/mesa/drivers/dri/r200/radeon_bo_int_drm.h new file mode 120000 index 00000000000..029450928be --- /dev/null +++ b/src/mesa/drivers/dri/r200/radeon_bo_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_bo_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r200/radeon_cs.c b/src/mesa/drivers/dri/r200/radeon_cs.c new file mode 120000 index 00000000000..66b7ad1eb03 --- /dev/null +++ b/src/mesa/drivers/dri/r200/radeon_cs.c @@ -0,0 +1 @@ +../radeon/radeon_cs.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r200/radeon_cs_int_drm.h b/src/mesa/drivers/dri/r200/radeon_cs_int_drm.h new file mode 120000 index 00000000000..462f5245d0e --- /dev/null +++ b/src/mesa/drivers/dri/r200/radeon_cs_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_cs_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile index cb0f715fa07..be005bd1641 100644 --- a/src/mesa/drivers/dri/r300/Makefile +++ b/src/mesa/drivers/dri/r300/Makefile @@ -14,7 +14,7 @@ EGL_SOURCES = server/radeon_egl.c endif ifeq ($(RADEON_LDFLAGS),) -CS_SOURCES = radeon_cs_space_drm.c +CS_SOURCES = radeon_cs_space_drm.c radeon_bo.c radeon_cs.c endif COMMON_SOURCES = \ @@ -43,13 +43,14 @@ RADEON_COMMON_SOURCES = \ DRIVER_SOURCES = \ radeon_screen.c \ + r300_blit.c \ r300_context.c \ r300_draw.c \ - r300_ioctl.c \ r300_cmdbuf.c \ r300_state.c \ r300_render.c \ r300_tex.c \ + r300_texcopy.c \ r300_texstate.c \ r300_vertprog.c \ r300_fragprog_common.c \ diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c new file mode 100644 index 00000000000..ea626d942dc --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.c @@ -0,0 +1,607 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "r300_blit.h" +#include "r300_cmdbuf.h" +#include "r300_emit.h" +#include "r300_tex.h" +#include "compiler/radeon_compiler.h" +#include "compiler/radeon_opcodes.h" + +static void vp_ins_outs(struct r300_vertex_program_compiler *c) +{ + c->code->inputs[VERT_ATTRIB_POS] = 0; + c->code->inputs[VERT_ATTRIB_TEX0] = 1; + c->code->outputs[VERT_RESULT_HPOS] = 0; + c->code->outputs[VERT_RESULT_TEX0] = 1; +} + +static void fp_allocate_hw_inputs( + struct r300_fragment_program_compiler * c, + void (*allocate)(void * data, unsigned input, unsigned hwreg), + void * mydata) +{ + allocate(mydata, FRAG_ATTRIB_TEX0, 0); +} + +static void create_vertex_program(struct r300_context *r300) +{ + struct r300_vertex_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_HPOS; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_POS; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_TEX0; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << VERT_ATTRIB_POS) | (1 << VERT_ATTRIB_TEX0); + compiler.RequiredOutputs = compiler.Base.Program.OutputsWritten = (1 << VERT_RESULT_HPOS) | (1 << VERT_RESULT_TEX0); + compiler.SetHwInputOutput = vp_ins_outs; + compiler.code = &r300->blit.vp_code; + + r3xx_compile_vertex_program(&compiler); +} + +static void create_fragment_program(struct r300_context *r300) +{ + struct r300_fragment_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_TEX; + inst->U.I.TexSrcTarget = RC_TEXTURE_2D; + inst->U.I.TexSrcUnit = 0; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = FRAG_RESULT_COLOR; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = FRAG_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << FRAG_ATTRIB_TEX0); + compiler.OutputColor = FRAG_RESULT_COLOR; + compiler.OutputDepth = FRAG_RESULT_DEPTH; + compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515); + compiler.code = &r300->blit.fp_code; + compiler.AllocateHwInputs = fp_allocate_hw_inputs; + + r3xx_compile_fragment_program(&compiler); +} + +void r300_blit_init(struct r300_context *r300) +{ + create_vertex_program(r300); + create_fragment_program(r300); +} + +static void r300_emit_tx_setup(struct r300_context *r300, + gl_format mesa_format, + struct radeon_bo *bo, + intptr_t offset, + unsigned width, + unsigned height, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + + assert(width <= 2048); + assert(height <= 2048); + assert(r300TranslateTexFormat(mesa_format) >= 0); + assert(offset % 32 == 0); + + BEGIN_BATCH(17); + OUT_BATCH_REGVAL(R300_TX_FILTER0_0, + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_S_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_R_SHIFT) | + R300_TX_MIN_FILTER_MIP_NONE | + R300_TX_MIN_FILTER_LINEAR | + R300_TX_MAG_FILTER_LINEAR | + (0 << 28)); + OUT_BATCH_REGVAL(R300_TX_FILTER1_0, 0); + OUT_BATCH_REGVAL(R300_TX_SIZE_0, + ((width-1) << R300_TX_WIDTHMASK_SHIFT) | + ((height-1) << R300_TX_HEIGHTMASK_SHIFT) | + (0 << R300_TX_DEPTHMASK_SHIFT) | + (0 << R300_TX_MAX_MIP_LEVEL_SHIFT) | + R300_TX_SIZE_TXPITCH_EN); + + OUT_BATCH_REGVAL(R300_TX_FORMAT_0, r300TranslateTexFormat(mesa_format)); + OUT_BATCH_REGVAL(R300_TX_FORMAT2_0, pitch - 1); + OUT_BATCH_REGSEQ(R300_TX_OFFSET_0, 1); + OUT_BATCH_RELOC(0, bo, offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + + OUT_BATCH_REGSEQ(R300_TX_INVALTAGS, 2); + OUT_BATCH(0); + OUT_BATCH(1); + + END_BATCH(); +} + +#define EASY_US_FORMAT(FMT, C0, C1, C2, C3, SIGN) \ + (FMT | R500_C0_SEL_##C0 | R500_C1_SEL_##C1 | \ + R500_C2_SEL_##C2 | R500_C3_SEL_##C3 | R500_OUT_SIGN(SIGN)) + +static uint32_t mesa_format_to_us_format(gl_format mesa_format) +{ + switch(mesa_format) + { + case MESA_FORMAT_S8_Z24: + case MESA_FORMAT_X8_Z24: + case MESA_FORMAT_RGBA8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0); + case MESA_FORMAT_RGB565: // x + case MESA_FORMAT_ARGB1555: // x + case MESA_FORMAT_RGBA8888_REV: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0); + case MESA_FORMAT_ARGB8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, B, G, R, A, 0); + case MESA_FORMAT_ARGB8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + case MESA_FORMAT_XRGB8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + + case MESA_FORMAT_RGB332: + return EASY_US_FORMAT(R500_OUT_FMT_C_3_3_2, A, R, G, B, 0); + + case MESA_FORMAT_RGBA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C4_32_FP, R, G, B, A, 0); + case MESA_FORMAT_RGBA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16_FP, R, G, B, A, 0); + case MESA_FORMAT_ALPHA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C_32_FP, A, A, A, A, 0); + case MESA_FORMAT_ALPHA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C_16_FP, A, A, A, A, 0); + + case MESA_FORMAT_SIGNED_RGBA8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0xf); + case MESA_FORMAT_SIGNED_RGBA8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0xf); + case MESA_FORMAT_SIGNED_RGBA_16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16, R, G, B, A, 0xf); + + default: + assert(!"Invalid format for US output\n"); + return 0; + } +} +#undef EASY_US_FORMAT + +static void r500_emit_fp_setup(struct r300_context *r300, + struct r500_fragment_program_code *fp, + gl_format dst_format) +{ + r500_emit_fp(r300, (uint32_t *)fp->inst, (fp->inst_end + 1) * 6, 0, 0, 0); + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(10); + OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); + OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(fp->inst_end)); + OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(fp->inst_end)); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_US_CONFIG, 0); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + OUT_BATCH_REGVAL(R500_US_PIXSIZE, fp->max_temp_idx); + END_BATCH(); +} + +static void r500_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_RS_INST_0, + (0 << R500_RS_INST_TEX_ID_SHIFT) | + (0 << R500_RS_INST_TEX_ADDR_SHIFT) | + R500_RS_INST_TEX_CN_WRITE | + R500_RS_INST_COL_CN_NO_WRITE); + OUT_BATCH_REGVAL(R500_RS_IP_0, + (0 << R500_RS_IP_TEX_PTR_S_SHIFT) | + (1 << R500_RS_IP_TEX_PTR_T_SHIFT) | + (2 << R500_RS_IP_TEX_PTR_R_SHIFT) | + (3 << R500_RS_IP_TEX_PTR_Q_SHIFT)); + END_BATCH(); +} + +static void r300_emit_fp_setup(struct r300_context *r300, + struct r300_fragment_program_code *code, + gl_format dst_format) +{ + unsigned i; + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH((code->alu.length + 1) * 4 + code->tex.length + 1 + 11); + + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_addr); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_addr); + } + + OUT_BATCH_REGSEQ(R300_US_TEX_INST_0, code->tex.length); + OUT_BATCH_TABLE(code->tex.inst, code->tex.length); + + OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); + OUT_BATCH(R300_PFS_CNTL_FIRST_NODE_HAS_TEX); + OUT_BATCH(code->pixsize); + OUT_BATCH(code->code_offset); + OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); + OUT_BATCH_TABLE(code->code_addr, 4); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + END_BATCH(); +} + +static void r300_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R300_RS_INST_0, + R300_RS_INST_TEX_ID(0) | + R300_RS_INST_TEX_ADDR(0) | + R300_RS_INST_TEX_CN_WRITE); + OUT_BATCH_REGVAL(R300_RS_IP_0, + R300_RS_TEX_PTR(0) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_T(R300_RS_SEL_C1) | + R300_RS_SEL_R(R300_RS_SEL_K0) | + R300_RS_SEL_Q(R300_RS_SEL_K1)); + END_BATCH(); +} + +static void emit_pvs_setup(struct r300_context *r300, + uint32_t *vp_code, + unsigned vp_len) +{ + BATCH_LOCALS(&r300->radeon); + + r300_emit_vpu(r300, vp_code, vp_len * 4, R300_PVS_CODE_START); + + BEGIN_BATCH(4); + OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); + OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | + ((vp_len - 1) << R300_PVS_XYZW_VALID_INST_SHIFT) | + ((vp_len - 1)<< R300_PVS_LAST_INST_SHIFT)); + OUT_BATCH(0); + OUT_BATCH((vp_len - 1) << R300_PVS_LAST_VTX_SRC_INST_SHIFT); + END_BATCH(); +} + +static void emit_vap_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(12); + OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); + OUT_BATCH(R300_VTX_XY_FMT | R300_VTX_Z_FMT); + OUT_BATCH(4); + + OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_0, + ((R300_DATA_TYPE_FLOAT_2 | (0 << R300_DST_VEC_LOC_SHIFT)) << 0) | + (((1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_2 | R300_LAST_VEC) << 16)); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, + ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 0) | + (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 16) ) ); + OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS); + END_BATCH(); +} + +static GLboolean validate_buffers(struct r300_context *r300, + struct radeon_bo *src_bo, + struct radeon_bo *dst_bo) +{ + int ret; + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + src_bo, RADEON_GEM_DOMAIN_VRAM, 0); + + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + dst_bo, 0, RADEON_GEM_DOMAIN_VRAM); + + ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, + first_elem(&r300->radeon.dma.reserved)->bo, + RADEON_GEM_DOMAIN_GTT, 0); + if (ret) + return GL_FALSE; + + return GL_TRUE; +} + +/** + * Calculate texcoords for given image region. + * Output values are [minx, maxx, miny, maxy] + */ +static void calc_tex_coords(float img_width, float img_height, + float x, float y, + float reg_width, float reg_height, + unsigned flip_y, float *buf) +{ + buf[0] = x / img_width; + buf[1] = buf[0] + reg_width / img_width; + buf[2] = y / img_height; + buf[3] = buf[2] + reg_height / img_height; + if (flip_y) + { + float tmp = buf[2]; + buf[2] = 1.0 - buf[3]; + buf[3] = 1.0 - tmp; + } +} + +static void emit_draw_packet(struct r300_context *r300, + unsigned src_width, unsigned src_height, + unsigned src_x_offset, unsigned src_y_offset, + unsigned dst_x_offset, unsigned dst_y_offset, + unsigned reg_width, unsigned reg_height, + unsigned flip_y) +{ + float texcoords[4]; + + calc_tex_coords(src_width, src_height, + src_x_offset, src_y_offset, + reg_width, reg_height, + flip_y, texcoords); + + float verts[] = { dst_x_offset, dst_y_offset, + texcoords[0], texcoords[3], + dst_x_offset, dst_y_offset + reg_height, + texcoords[0], texcoords[2], + dst_x_offset + reg_width, dst_y_offset + reg_height, + texcoords[1], texcoords[2], + dst_x_offset + reg_width, dst_y_offset, + texcoords[1], texcoords[3] }; + + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(19); + OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_IMMD_2, 16); + OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED | + (4 << 16) | R300_VAP_VF_CNTL__PRIM_QUADS); + OUT_BATCH_TABLE(verts, 16); + END_BATCH(); +} + +static void other_stuff(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(15); + OUT_BATCH_REGVAL(R300_GA_POLY_MODE, + R300_GA_POLY_MODE_FRONT_PTYPE_TRI | R300_GA_POLY_MODE_BACK_PTYPE_TRI); + OUT_BATCH_REGVAL(R300_SU_CULL_MODE, R300_FRONT_FACE_CCW); + OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); + OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); + OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); + OUT_BATCH(0x0); + OUT_BATCH(0x0); + OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE); + OUT_BATCH_REGVAL(R300_ZB_CNTL, 0); + END_BATCH(); +} + +static void emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + intptr_t offset, + gl_format mesa_format, + unsigned pitch, + unsigned width, + unsigned height) +{ + BATCH_LOCALS(&r300->radeon); + + unsigned x1, y1, x2, y2; + x1 = 0; + y1 = 0; + x2 = width - 1; + y2 = height - 1; + + if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) { + x1 += R300_SCISSORS_OFFSET; + y1 += R300_SCISSORS_OFFSET; + x2 += R300_SCISSORS_OFFSET; + y2 += R300_SCISSORS_OFFSET; + } + + r300_emit_cb_setup(r300, bo, offset, mesa_format, + _mesa_get_format_bytes(mesa_format), + _mesa_format_row_stride(mesa_format, pitch)); + + BEGIN_BATCH_NO_AUTOSTATE(5); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH_REGVAL(R300_RB3D_CCTL, 0); + END_BATCH(); +} + +/** + * Copy a region of [@a width x @a height] pixels from source buffer + * to destination buffer. + * @param[in] r300 r300 context + * @param[in] src_bo source radeon buffer object + * @param[in] src_offset offset of the source image in the @a src_bo + * @param[in] src_mesaformat source image format + * @param[in] src_pitch aligned source image width + * @param[in] src_width source image width + * @param[in] src_height source image height + * @param[in] src_x_offset x offset in the source image + * @param[in] src_y_offset y offset in the source image + * @param[in] dst_bo destination radeon buffer object + * @param[in] dst_offset offset of the destination image in the @a dst_bo + * @param[in] dst_mesaformat destination image format + * @param[in] dst_pitch aligned destination image width + * @param[in] dst_width destination image width + * @param[in] dst_height destination image height + * @param[in] dst_x_offset x offset in the destination image + * @param[in] dst_y_offset y offset in the destination image + * @param[in] width region width + * @param[in] height region height + * @param[in] flip_y set if y coords of the source image need to be flipped + */ +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + unsigned src_x_offset, + unsigned src_y_offset, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_pitch, + unsigned dst_width, + unsigned dst_height, + unsigned dst_x_offset, + unsigned dst_y_offset, + unsigned reg_width, + unsigned reg_height, + unsigned flip_y) +{ + /* Need to clamp the region size to make sure + * we don't read outside of the source buffer + * or write outside of the destination buffer. + */ + if (reg_width + src_x_offset > src_width) + reg_width = src_width - src_x_offset; + if (reg_height + src_y_offset > src_height) + reg_height = src_height - src_y_offset; + if (reg_width + dst_x_offset > dst_width) + reg_width = dst_width - dst_x_offset; + if (reg_height + dst_y_offset > dst_height) + reg_height = dst_height - dst_y_offset; + + if (src_bo == dst_bo) { + return GL_FALSE; + } + + if (0) { + fprintf(stderr, "src: size [%d x %d], pitch %d, " + "offset [%d x %d], format %s, bo %p\n", + src_width, src_height, src_pitch, + src_offset, src_y_offset, + _mesa_get_format_name(src_mesaformat), + src_bo); + fprintf(stderr, "dst: pitch %d, offset[%d x %d], format %s, bo %p\n", + dst_pitch, dst_x_offset, dst_y_offset, + _mesa_get_format_name(dst_mesaformat), dst_bo); + fprintf(stderr, "region: %d x %d\n", reg_width, reg_height); + } + + if (!validate_buffers(r300, src_bo, dst_bo)) + return GL_FALSE; + + rcommonEnsureCmdBufSpace(&r300->radeon, 200, __FUNCTION__); + + other_stuff(r300); + + r300_emit_tx_setup(r300, src_mesaformat, src_bo, src_offset, src_width, src_height, src_pitch); + + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + r500_emit_fp_setup(r300, &r300->blit.fp_code.code.r500, dst_mesaformat); + r500_emit_rs_setup(r300); + } else { + r300_emit_fp_setup(r300, &r300->blit.fp_code.code.r300, dst_mesaformat); + r300_emit_rs_setup(r300); + } + + emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2); + emit_vap_setup(r300); + + emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height); + + emit_draw_packet(r300, src_width, src_height, + src_x_offset, src_y_offset, + dst_x_offset, dst_y_offset, + reg_width, reg_height, + flip_y); + + r300EmitCacheFlush(r300); + + radeonFlush(r300->radeon.glCtx); + + return GL_TRUE; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_blit.h b/src/mesa/drivers/dri/r300/r300_blit.h new file mode 100644 index 00000000000..dc21e880982 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.h @@ -0,0 +1,54 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef R300_BLIT_H +#define R300_BLIT_H + +void r300_blit_init(struct r300_context *r300); + +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + unsigned src_x_offset, + unsigned src_y_offset, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_pitch, + unsigned dst_width, + unsigned dst_height, + unsigned dst_x_offset, + unsigned dst_y_offset, + unsigned width, + unsigned height, + unsigned flip_y); + +#endif // R300_BLIT_H
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index ad8db6e68e0..e1c33bbb2cf 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_drm.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_reg.h" #include "r300_cmdbuf.h" #include "r300_emit.h" @@ -72,7 +71,7 @@ static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt) #define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) #define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count) -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -86,54 +85,73 @@ int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 4) + extra : 0; } +void r300_emit_vpu(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(5 + len); + OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); + OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); + OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); +} -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw; + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, ndw; - cmd.u = atom->cmd[0]; - addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; - ndw = atom->check(ctx, atom); + cmd.u = atom->cmd[0]; + addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; + ndw = atom->check(ctx, atom); - BEGIN_BATCH_NO_AUTOSTATE(ndw); + r300_emit_vpu(r300, &atom->cmd[1], vpu_count(atom->cmd) * 4, addr); +} - ndw -= 5; - OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); - OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp) +{ + BATCH_LOCALS(&r300->radeon); + + addr |= (type << 16); + addr |= (clamp << 17); + + BEGIN_BATCH_NO_AUTOSTATE(len + 3); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); + OUT_BATCH(addr); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); } -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_r500fp_atom(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw, sz; - int type, clamp; - - ndw = atom->check(ctx, atom); - - cmd.u = atom->cmd[0]; - sz = cmd.r500fp.count; - addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; - type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); - clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); - - addr |= (type << 16); - addr |= (clamp << 17); - - BEGIN_BATCH_NO_AUTOSTATE(ndw); - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); - OUT_BATCH(addr); - ndw-=3; - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - END_BATCH(); + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, count; + int type, clamp; + + cmd.u = atom->cmd[0]; + addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; + type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); + clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); + + if (type) { + count = r500fp_count(atom->cmd) * 4; + } else { + count = r500fp_count(atom->cmd) * 6; + } + + r500_emit_fp(r300, &atom->cmd[1], count, addr, type, clamp); } static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) @@ -256,110 +274,136 @@ static int check_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) return dw; } -static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_scissor(struct r300_context *r300, + unsigned width, + unsigned height) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - struct radeon_renderbuffer *rrb; - uint32_t cbpitch; - uint32_t offset = r300->radeon.state.color.draw_offset; - uint32_t dw = 6; - int i; + int i; + BATCH_LOCALS(&r300->radeon); + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH(0); + OUT_BATCH(((width - 1) << R300_SCISSORS_X_SHIFT) | + ((height - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((width - 1) << R300_CLIPRECT_X_SHIFT) | ((height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } else { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | + (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH(((width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | + ((height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((R300_SCISSORS_OFFSET + width - 1) << R300_CLIPRECT_X_SHIFT) | + ((R300_SCISSORS_OFFSET + height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } +} - rrb = radeon_get_colorbuffer(&r300->radeon); - if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); - return; - } +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + uint32_t cbpitch = pitch / cpp; + uint32_t dw = 6; - if (RADEON_DEBUG & RADEON_STATE) - fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else switch (rrb->base.Format) { + assert(offset % 32 == 0); + + switch (format) { case MESA_FORMAT_RGB565: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_RGB565; - break; + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; case MESA_FORMAT_RGB565_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_RGB565; - break; + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; case MESA_FORMAT_ARGB4444: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB4444; - break; + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; case MESA_FORMAT_ARGB4444_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB4444; - break; - case MESA_FORMAT_ARGB1555: - assert(_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB1555; - break; - case MESA_FORMAT_ARGB1555_REV: - assert(!_mesa_little_endian()); - cbpitch |= R300_COLOR_FORMAT_ARGB1555; - break; - default: - _mesa_problem(ctx, "unexpected format in emit_cb_offset()"); - } + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; + case MESA_FORMAT_ARGB1555: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + case MESA_FORMAT_ARGB1555_REV: + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + default: + if (cpp == 4) { + cbpitch |= R300_COLOR_FORMAT_ARGB8888; + } else { + _mesa_problem(r300->radeon.glCtx, "unexpected format in emit_cb_offset()");; + } + break; + } - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) - cbpitch |= R300_COLOR_TILE_ENABLE; + if (bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + cbpitch |= R300_COLOR_TILE_ENABLE; + + if (r300->radeon.radeonScreen->kernel_mm) + dw += 2; + + BEGIN_BATCH_NO_AUTOSTATE(dw); + OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); + OUT_BATCH_RELOC(offset, bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); + if (!r300->radeon.radeonScreen->kernel_mm) + OUT_BATCH(cbpitch); + else + OUT_BATCH_RELOC(cbpitch, bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); + END_BATCH(); +} + +static void emit_cb_offset_atom(GLcontext *ctx, struct radeon_state_atom * atom) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + uint32_t offset = r300->radeon.state.color.draw_offset; + + rrb = radeon_get_colorbuffer(&r300->radeon); + if (!rrb || !rrb->bo) { + fprintf(stderr, "no rrb\n"); + return; + } + + if (RADEON_DEBUG & RADEON_STATE) + fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); + + r300_emit_cb_setup(r300, rrb->bo, offset, rrb->base.Format, rrb->cpp, rrb->pitch); - if (r300->radeon.radeonScreen->kernel_mm) - dw += 2; - BEGIN_BATCH_NO_AUTOSTATE(dw); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(cbpitch); - else - OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); if (r300->radeon.radeonScreen->driScreen->dri2.enabled) { - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH(0); - OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } else { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | - (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | - ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } + emit_scissor(r300, rrb->base.Width, rrb->base.Height); } } @@ -455,7 +499,7 @@ static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? cnt + 1 : 0; } -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -467,7 +511,7 @@ int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 6) + extra : 0; } -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -644,13 +688,13 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.r500fp.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 0, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp.emit = emit_r500fp; + r300->hw.r500fp.emit = emit_r500fp_atom; ALLOC_STATE(r500fp_const, r500fp_const, R500_FPP_CMDSIZE, 0); r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 1, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp_const.emit = emit_r500fp; + r300->hw.r500fp_const.emit = emit_r500fp_atom; } else { ALLOC_STATE(fp, always, R300_FP_CMDSIZE, 0); r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CONFIG, 3); @@ -694,7 +738,7 @@ void r300InitCmdBuf(r300ContextPtr r300) ALLOC_STATE(rop, always, 2, 0); r300->hw.rop.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_ROPCNTL, 1); ALLOC_STATE(cb, cb_offset, R300_CB_CMDSIZE, 0); - r300->hw.cb.emit = &emit_cb_offset; + r300->hw.cb.emit = &emit_cb_offset_atom; ALLOC_STATE(rb3d_dither_ctl, always, 10, 0); r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9); ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0); @@ -758,20 +802,20 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.vpi.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CODE_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpi.emit = emit_vpu; + r300->hw.vpi.emit = emit_vpu_state; if (is_r500) { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -779,20 +823,20 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R500_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } else { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -800,7 +844,7 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R300_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } } diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h index 1b703e518a0..0e68da928ed 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -44,14 +44,26 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define FIREAOS_BUFSZ (3) #define SCISSORS_BUFSZ (3) -extern void r300InitCmdBuf(r300ContextPtr r300); +void r300InitCmdBuf(r300ContextPtr r300); void r300_emit_scissor(GLcontext *ctx); -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom); -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom); +void r300_emit_vpu(struct r300_context *ctx, + uint32_t *data, + unsigned len, + uint32_t addr); -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom); -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom); -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom); +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp); -#endif /* __R300_CMDBUF_H__ */ +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch); + +#endif /* __R300_CMDBUF_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 5f07b956349..3c6ec2a34a8 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -55,13 +55,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "drivers/common/driverfuncs.h" +#include "drivers/common/meta.h" #include "r300_context.h" #include "radeon_context.h" #include "radeon_span.h" +#include "r300_blit.h" #include "r300_cmdbuf.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "r300_tex.h" #include "r300_emit.h" #include "r300_swtcl.h" @@ -92,6 +93,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/remap_helper.h" +void r300_init_texcopy_functions(struct dd_function_table *table); static const struct dri_extension card_extensions[] = { /* *INDENT-OFF* */ @@ -451,6 +453,13 @@ static void r300InitGLExtensions(GLcontext *ctx) } } +static void r300InitIoctlFuncs(struct dd_function_table *functions) +{ + functions->Clear = _mesa_meta_Clear; + functions->Finish = radeonFinish; + functions->Flush = radeonFlush; +} + /* Create the device specific rendering context. */ GLboolean r300CreateContext(const __GLcontextModes * glVisual, @@ -484,6 +493,10 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, radeonInitQueryObjFunctions(&functions); radeonInitBufferObjectFuncs(&functions); + if (r300->radeon.radeonScreen->kernel_mm) { + r300_init_texcopy_functions(&functions); + } + if (!radeonInitContext(&r300->radeon, &functions, glVisual, driContextPriv, sharedContextPrivate)) { @@ -530,6 +543,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, r300InitSwtcl(ctx); } + r300_blit_init(r300); radeon_fbo_init(&r300->radeon); radeonInitSpanFuncs( ctx ); r300InitCmdBuf(r300); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 518d5cdbf4f..54a92a2e447 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -533,6 +533,11 @@ struct r300_context { uint32_t fallback; + struct { + struct r300_vertex_program_code vp_code; + struct rX00_fragment_program_code fp_code; + } blit; + DECLARE_RENDERINPUTS(render_inputs_bitset); }; @@ -549,6 +554,8 @@ extern void r300InitShaderFunctions(r300ContextPtr r300); extern void r300InitDraw(GLcontext *ctx); +extern void r300_init_texcopy_functions(struct dd_function_table *table); + #define r300PackFloat32 radeonPackFloat32 #define r300PackFloat24 radeonPackFloat24 diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 07e62230874..3759ca2bea7 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -49,7 +49,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" #include "r300_emit.h" -#include "r300_ioctl.h" #include "r300_render.h" #include "r300_swtcl.h" diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c deleted file mode 100644 index 5cb04e2bb6d..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.c +++ /dev/null @@ -1,782 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. -Copyright (C) 2004 Nicolai Haehnle. -All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/** - * \file - * - * \author Keith Whitwell <[email protected]> - * - * \author Nicolai Haehnle <[email protected]> - */ - -#include <sched.h> -#include <errno.h> - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/macros.h" -#include "main/context.h" -#include "main/simple_list.h" -#include "swrast/swrast.h" - -#include "radeon_common.h" -#include "radeon_lock.h" -#include "r300_context.h" -#include "r300_ioctl.h" -#include "r300_cmdbuf.h" -#include "r300_state.h" -#include "r300_vertprog.h" -#include "radeon_reg.h" -#include "r300_emit.h" -#include "r300_context.h" - -#include "vblank.h" - -#define R200_3D_DRAW_IMMD_2 0xC0003500 - -#define CLEARBUFFER_COLOR 0x1 -#define CLEARBUFFER_DEPTH 0x2 -#define CLEARBUFFER_STENCIL 0x4 - -#if 1 - -/** - * Fragment program helper macros - */ - -/* Produce unshifted source selectors */ -#define FP_TMP(idx) (idx) -#define FP_CONST(idx) ((idx) | (1 << 5)) - -/* Produce source/dest selector dword */ -#define FP_SELC_MASK_NO 0 -#define FP_SELC_MASK_X 1 -#define FP_SELC_MASK_Y 2 -#define FP_SELC_MASK_XY 3 -#define FP_SELC_MASK_Z 4 -#define FP_SELC_MASK_XZ 5 -#define FP_SELC_MASK_YZ 6 -#define FP_SELC_MASK_XYZ 7 - -#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTC_SHIFT) | \ - (FP_SELC_MASK_##regmask << 23) | \ - (FP_SELC_MASK_##outmask << 26) | \ - ((src0) << R300_ALU_SRC0C_SHIFT) | \ - ((src1) << R300_ALU_SRC1C_SHIFT) | \ - ((src2) << R300_ALU_SRC2C_SHIFT)) - -#define FP_SELA_MASK_NO 0 -#define FP_SELA_MASK_W 1 - -#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTA_SHIFT) | \ - (FP_SELA_MASK_##regmask << 23) | \ - (FP_SELA_MASK_##outmask << 24) | \ - ((src0) << R300_ALU_SRC0A_SHIFT) | \ - ((src1) << R300_ALU_SRC1A_SHIFT) | \ - ((src2) << R300_ALU_SRC2A_SHIFT)) - -/* Produce unshifted argument selectors */ -#define FP_ARGC(source) R300_ALU_ARGC_##source -#define FP_ARGA(source) R300_ALU_ARGA_##source -#define FP_ABS(arg) ((arg) | (1 << 6)) -#define FP_NEG(arg) ((arg) ^ (1 << 5)) - -/* Produce instruction dword */ -#define FP_INSTRC(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTC_##opcode | \ - ((arg0) << R300_ALU_ARG0C_SHIFT) | \ - ((arg1) << R300_ALU_ARG1C_SHIFT) | \ - ((arg2) << R300_ALU_ARG2C_SHIFT)) - -#define FP_INSTRA(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTA_##opcode | \ - ((arg0) << R300_ALU_ARG0A_SHIFT) | \ - ((arg1) << R300_ALU_ARG1A_SHIFT) | \ - ((arg2) << R300_ALU_ARG2A_SHIFT)) - -#endif - -static void r300EmitClearState(GLcontext * ctx); - -static void r300ClearBuffer(r300ContextPtr r300, int flags, - struct radeon_renderbuffer *rrb, - struct radeon_renderbuffer *rrbd) -{ - BATCH_LOCALS(&r300->radeon); - GLcontext *ctx = r300->radeon.glCtx; - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - GLuint cbpitch = 0; - r300ContextPtr rmesa = r300; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "%s: buffer %p (%i,%i %ix%i)\n", - __FUNCTION__, rrb, dPriv->x, dPriv->y, - dPriv->w, dPriv->h); - - if (rrb) { - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else - cbpitch |= R300_COLOR_FORMAT_RGB565; - - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - cbpitch |= R300_COLOR_TILE_ENABLE; - } - } - - /* TODO in bufmgr */ - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - end_3d(&rmesa->radeon); - - if (flags & CLEARBUFFER_COLOR) { - assert(rrb != 0); - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGVAL(R300_RB3D_COLORPITCH0, cbpitch); - END_BATCH(); - } -#if 1 - if (flags & (CLEARBUFFER_DEPTH | CLEARBUFFER_STENCIL)) { - uint32_t zbpitch = (rrbd->pitch / rrbd->cpp); - if (rrbd->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - zbpitch |= R300_DEPTHMACROTILE_ENABLE; - } - if (rrbd->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){ - zbpitch |= R300_DEPTHMICROTILE_TILED; - } - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1); - OUT_BATCH_RELOC(0, rrbd->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(zbpitch); - else - OUT_BATCH_RELOC(zbpitch, rrbd->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); - } -#endif - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(RB3D_COLOR_CHANNEL_MASK, 1); - if (flags & CLEARBUFFER_COLOR) { - OUT_BATCH((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) | - (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) | - (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) | - (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0)); - } else { - OUT_BATCH(0); - } - - - { - uint32_t t1, t2; - - t1 = 0x0; - t2 = 0x0; - - if (flags & CLEARBUFFER_DEPTH) { - t1 |= R300_Z_ENABLE | R300_Z_WRITE_ENABLE; - t2 |= - (R300_ZS_ALWAYS << R300_Z_FUNC_SHIFT); - } - - if (flags & CLEARBUFFER_STENCIL) { - t1 |= R300_STENCIL_ENABLE; - t2 |= - (R300_ZS_ALWAYS << - R300_S_FRONT_FUNC_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_SFAIL_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZPASS_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZFAIL_OP_SHIFT); - } - - OUT_BATCH_REGSEQ(R300_ZB_CNTL, 3); - OUT_BATCH(t1); - OUT_BATCH(t2); - OUT_BATCH(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) << - R300_STENCILWRITEMASK_SHIFT) | - (ctx->Stencil.Clear & R300_STENCILREF_MASK)); - END_BATCH(); - } - - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(9); - OUT_BATCH(cmdpacket3(r300->radeon.radeonScreen, R300_CMD_PACKET3_CLEAR)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - END_BATCH(); - } else { - OUT_BATCH(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8)); - OUT_BATCH(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | - (1 << R300_PRIM_NUM_VERTICES_SHIFT)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - } - - r300EmitCacheFlush(rmesa); - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - - R300_STATECHANGE(r300, cb); - R300_STATECHANGE(r300, cmk); - R300_STATECHANGE(r300, zs); -} - -static void r300EmitClearState(GLcontext * ctx) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - int i; - int has_tcl; - int is_r500 = 0; - GLuint vap_cntl; - - has_tcl = r300->options.hw_tcl_enabled; - - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - is_r500 = 1; - - /* State atom dirty tracking is a little subtle here. - * - * On the one hand, we need to make sure base state is emitted - * here if we start with an empty batch buffer, otherwise clear - * works incorrectly with multiple processes. Therefore, the first - * BEGIN_BATCH cannot be a BEGIN_BATCH_NO_AUTOSTATE. - * - * On the other hand, implicit state emission clears the state atom - * dirty bits, so we have to call R300_STATECHANGE later than the - * first BEGIN_BATCH. - * - * The final trickiness is that, because we change state, we need - * to ensure that any stored swtcl primitives are flushed properly - * before we start changing state. See the R300_NEWPRIM in r300Clear - * for this. - */ - BEGIN_BATCH(31); - OUT_BATCH_REGSEQ(R300_VAP_PROG_STREAM_CNTL_0, 1); - if (!has_tcl) - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - else - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - - OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); - OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, - ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE0_SHIFT) | - (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE1_SHIFT))); - - /* R300_VAP_INPUT_CNTL_0, R300_VAP_INPUT_CNTL_1 */ - OUT_BATCH_REGSEQ(R300_VAP_VTX_STATE_CNTL, 2); - OUT_BATCH((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT)); - OUT_BATCH(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0); - - /* comes from fglrx startup of clear */ - OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); - OUT_BATCH(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA | - R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | - R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | - R300_VPORT_Z_OFFSET_ENA); - OUT_BATCH(0x8); - - OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); - - OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); - OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT | - R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT); - OUT_BATCH(0); /* no textures */ - - OUT_BATCH_REGVAL(R300_TX_ENABLE, 0); - - OUT_BATCH_REGSEQ(R300_SE_VPORT_XSCALE, 6); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->x); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->y); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(0.0); - - OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); - - OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - END_BATCH(); - - R300_STATECHANGE(r300, vir[0]); - R300_STATECHANGE(r300, fogs); - R300_STATECHANGE(r300, vir[1]); - R300_STATECHANGE(r300, vic); - R300_STATECHANGE(r300, vte); - R300_STATECHANGE(r300, vof); - R300_STATECHANGE(r300, txe); - R300_STATECHANGE(r300, vpt); - R300_STATECHANGE(r300, at); - R300_STATECHANGE(r300, bld); - R300_STATECHANGE(r300, ps); - - if (has_tcl) { - R300_STATECHANGE(r300, vap_clip_cntl); - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE); - END_BATCH(); - } - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_GA_POINT_SIZE, - ((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) | - ((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT)); - END_BATCH(); - - if (!is_r500) { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R300_RS_IP_0, 8); - for (i = 0; i < 8; ++i) - OUT_BATCH(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3)); - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE); - END_BATCH(); - } else { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R500_RS_IP_0, 8); - for (i = 0; i < 8; ++i) { - OUT_BATCH((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) | - (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT)); - } - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE); - END_BATCH(); - } - - if (!is_r500) { - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, fpi[0]); - R300_STATECHANGE(r300, fpi[1]); - R300_STATECHANGE(r300, fpi[2]); - R300_STATECHANGE(r300, fpi[3]); - - BEGIN_BATCH(17); - OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(R300_RGBA_OUT); - - OUT_BATCH_REGVAL(R300_US_ALU_RGB_INST_0, - FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_RGB_ADDR_0, - FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0)); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_INST_0, - FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_ADDR_0, - FP_SELA(0, NO, W, FP_TMP(0), 0, 0)); - END_BATCH(); - } else { - struct radeon_state_atom r500fp; - uint32_t _cmd[10]; - - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, r500fp); - - BEGIN_BATCH(7); - OUT_BATCH_REGSEQ(R500_US_CONFIG, 2); - OUT_BATCH(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); - OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1)); - OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1)); - OUT_BATCH(R500_US_CODE_OFFSET_ADDR(0)); - END_BATCH(); - - r500fp.check = check_r500fp; - r500fp.cmd = _cmd; - r500fp.cmd[0] = cmdr500fp(r300->radeon.radeonScreen, 0, 1, 0, 0); - r500fp.cmd[1] = R500_INST_TYPE_OUT | - R500_INST_TEX_SEM_WAIT | - R500_INST_LAST | - R500_INST_RGB_OMASK_R | - R500_INST_RGB_OMASK_G | - R500_INST_RGB_OMASK_B | - R500_INST_ALPHA_OMASK | - R500_INST_RGB_CLAMP | - R500_INST_ALPHA_CLAMP; - r500fp.cmd[2] = R500_RGB_ADDR0(0) | - R500_RGB_ADDR1(0) | - R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | - R500_RGB_ADDR2_CONST; - r500fp.cmd[3] = R500_ALPHA_ADDR0(0) | - R500_ALPHA_ADDR1(0) | - R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | - R500_ALPHA_ADDR2_CONST; - r500fp.cmd[4] = R500_ALU_RGB_SEL_A_SRC0 | - R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | - R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | - R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | - R500_ALU_RGB_G_SWIZ_B_B; - r500fp.cmd[5] = R500_ALPHA_OP_CMP | - R500_ALPHA_SWIZ_A_A | - R500_ALPHA_SWIZ_B_A; - r500fp.cmd[6] = R500_ALU_RGBA_OP_CMP | - R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | - R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0; - - r500fp.cmd[7] = 0; - if (r300->radeon.radeonScreen->kernel_mm) { - emit_r500fp(ctx, &r500fp); - } else { - int dwords = r500fp.check(ctx,&r500fp); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r500fp.cmd, dwords); - END_BATCH(); - } - - } - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); - - if (has_tcl) { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (12 << R300_VF_MAX_VTX_NUM_SHIFT)); - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - vap_cntl |= R500_TCL_STATE_OPTIMIZATION; - } else { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (5 << R300_VF_MAX_VTX_NUM_SHIFT)); - } - - if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV515) - vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV560) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV570)) - vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420)) - vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R520) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R580)) - vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT); - else - vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT); - - R300_STATECHANGE(r300, vap_cntl); - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_CNTL, vap_cntl); - END_BATCH(); - - if (has_tcl) { - struct radeon_state_atom vpu; - uint32_t _cmd[10]; - R300_STATECHANGE(r300, pvs); - R300_STATECHANGE(r300, vap_flush); - R300_STATECHANGE(r300, vpi); - - BEGIN_BATCH(4); - OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); - OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | - (0 << R300_PVS_XYZW_VALID_INST_SHIFT) | - (1 << R300_PVS_LAST_INST_SHIFT)); - OUT_BATCH((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | - (0 << R300_PVS_MAX_CONST_ADDR_SHIFT)); - OUT_BATCH(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT); - END_BATCH(); - - vpu.check = check_vpu; - vpu.cmd = _cmd; - vpu.cmd[0] = cmdvpu(r300->radeon.radeonScreen, 0, 2); - - vpu.cmd[1] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, - 0, 0xf, PVS_DST_REG_OUT); - vpu.cmd[2] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, - PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[3] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[4] = 0x0; - - vpu.cmd[5] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf, - PVS_DST_REG_OUT); - vpu.cmd[6] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X, - PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, - PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, - NEGATE_NONE); - vpu.cmd[7] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[8] = 0x0; - - if (r300->radeon.radeonScreen->kernel_mm) { - int dwords = r300->hw.vap_flush.check(ctx,&r300->hw.vap_flush); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r300->hw.vap_flush.cmd, dwords); - END_BATCH(); - emit_vpu(ctx, &vpu); - } else { - int dwords = vpu.check(ctx,&vpu); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(vpu.cmd, dwords); - END_BATCH(); - } - - } -} - -static int r300KernelClear(GLcontext *ctx, GLuint flags) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - struct radeon_framebuffer *rfb = dPriv->driverPrivate; - struct radeon_renderbuffer *rrb; - struct radeon_renderbuffer *rrbd; - int bits = 0, ret; - - /* Make sure it fits there. */ - radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs); - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd) { - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs); - if (ret) - return -1; - - rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__); - if (flags || bits) - r300EmitClearState(ctx); - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd && (flags & BUFFER_BIT_DEPTH)) - bits |= CLEARBUFFER_DEPTH; - - if (rrbd && (flags & BUFFER_BIT_STENCIL)) - bits |= CLEARBUFFER_STENCIL; - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL); - bits = 0; - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (bits) - r300ClearBuffer(r300, bits, NULL, rrbd); - - COMMIT_BATCH(); - return 0; -} - -/** - * Buffer clear - */ -static void r300Clear(GLcontext * ctx, GLbitfield mask) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); - GLbitfield swrast_mask = 0, tri_mask = 0; - int i, ret; - struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "r300Clear\n"); - - if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) { - LOCK_HARDWARE(&r300->radeon); - UNLOCK_HARDWARE(&r300->radeon); - if (dPriv->numClipRects == 0) - return; - } - - /* Flush swtcl vertices if necessary, because we will change hardware - * state during clear. See also the state-related comment in - * r300EmitClearState. - */ - R300_NEWPRIM(r300); - - if (colorMask == ~0) - tri_mask |= (mask & BUFFER_BITS_COLOR); - else - tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)); - - - /* HW stencil */ - if (mask & BUFFER_BIT_STENCIL) { - tri_mask |= BUFFER_BIT_STENCIL; - } - - /* HW depth */ - if (mask & BUFFER_BIT_DEPTH) { - tri_mask |= BUFFER_BIT_DEPTH; - } - - /* If we're doing a tri pass for depth/stencil, include a likely color - * buffer with it. - */ - - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((tri_mask) & bufBit) { - if (!fb->Attachment[i].Renderbuffer->ClassID) { - tri_mask &= ~bufBit; - swrast_mask |= bufBit; - } - } - } - - /* SW fallback clearing */ - swrast_mask = mask & ~tri_mask; - - ret = 0; - if (tri_mask) { - if (r300->radeon.radeonScreen->kernel_mm) - radeonUserClear(ctx, tri_mask); - else { - /* if kernel clear fails due to size restraints fallback */ - ret = r300KernelClear(ctx, tri_mask); - if (ret < 0) - swrast_mask |= tri_mask; - } - } - - if (swrast_mask) { - if (RADEON_DEBUG & RADEON_FALLBACKS) - fprintf(stderr, "%s: swrast clear, mask: %x\n", - __FUNCTION__, swrast_mask); - _swrast_Clear(ctx, swrast_mask); - } -} - -void r300InitIoctlFuncs(struct dd_function_table *functions) -{ - functions->Clear = r300Clear; - functions->Finish = radeonFinish; - functions->Flush = radeonFlush; -} diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h deleted file mode 100644 index 3abfa71a6e8..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.h +++ /dev/null @@ -1,44 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell <[email protected]> - * Nicolai Haehnle <[email protected]> - */ - -#ifndef __R300_IOCTL_H__ -#define __R300_IOCTL_H__ - -#include "r300_context.h" -#include "radeon_drm.h" - -extern void r300InitIoctlFuncs(struct dd_function_table *functions); - -#endif /* __R300_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index 4ae593cbe79..02c94250a8f 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -68,7 +68,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/tnl.h" #include "tnl/t_vp_build.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_tex.h" diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index ac20c08e201..da0a9dfb4cf 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -55,7 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_emit.h" diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index 726b3ff98e1..ac3d5b1bec3 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -48,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h index 8a653ea2d11..6ede0fe25c9 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.h +++ b/src/mesa/drivers/dri/r300/r300_tex.h @@ -51,4 +51,6 @@ extern GLboolean r300ValidateBuffers(GLcontext * ctx); extern void r300InitTextureFuncs(struct dd_function_table *functions); +int32_t r300TranslateTexFormat(gl_format mesaFormat); + #endif /* __r300_TEX_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_texcopy.c b/src/mesa/drivers/dri/r300/r300_texcopy.c new file mode 100644 index 00000000000..ebc9c05b8a8 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_texcopy.c @@ -0,0 +1,168 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "main/image.h" +#include "main/teximage.h" +#include "main/texstate.h" +#include "drivers/common/meta.h" + +#include "radeon_mipmap_tree.h" +#include "r300_blit.h" +#include <main/debug.h> + +// TODO: +// need to pass correct pitch for small dst textures! +static GLboolean +do_copy_texsubimage(GLcontext *ctx, + GLenum target, GLint level, + struct radeon_tex_obj *tobj, + radeon_texture_image *timg, + GLint dstx, GLint dsty, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct r300_context *r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + + if (_mesa_get_format_bits(timg->base.TexFormat, GL_DEPTH_BITS) > 0) { + rrb = radeon_get_depthbuffer(&r300->radeon); + } else { + rrb = radeon_get_colorbuffer(&r300->radeon); + } + + if (!timg->mt) { + radeon_validate_texture_miptree(ctx, &tobj->base); + } + + assert(rrb && rrb->bo); + assert(timg->mt->bo); + assert(timg->base.Width >= dstx + width); + assert(timg->base.Height >= dsty + height); + + intptr_t src_offset = rrb->draw_offset; + intptr_t dst_offset = radeon_miptree_image_offset(timg->mt, _mesa_tex_target_to_face(target), level); + + if (src_offset % 32 || dst_offset % 32) { + return GL_FALSE; + } + + if (0) { + fprintf(stderr, "%s: copying to face %d, level %d\n", + __FUNCTION__, _mesa_tex_target_to_face(target), level); + fprintf(stderr, "to: x %d, y %d, offset %d\n", dstx, dsty, (uint32_t) dst_offset); + fprintf(stderr, "from (%dx%d) width %d, height %d, offset %d, pitch %d\n", + x, y, rrb->base.Width, rrb->base.Height, (uint32_t) src_offset, rrb->pitch/rrb->cpp); + fprintf(stderr, "src size %d, dst size %d\n", rrb->bo->size, timg->mt->bo->size); + + } + + /* blit from src buffer to texture */ + return r300_blit(r300, rrb->bo, src_offset, rrb->base.Format, rrb->pitch/rrb->cpp, + rrb->base.Width, rrb->base.Height, x, y, + timg->mt->bo, dst_offset, timg->base.TexFormat, + timg->base.Width, timg->base.Width, timg->base.Height, + dstx, dsty, width, height, 1); +} + +static void +r300CopyTexImage2D(GLcontext *ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + int srcx, srcy, dstx, dsty; + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, GL_UNSIGNED_BYTE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + srcx = x; + srcy = y; + dstx = 0; + dsty = 0; + if (!_mesa_clip_copytexsubimage(ctx, + &dstx, &dsty, + &srcx, &srcy, + &width, &height)) { + return; + } + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + 0, 0, x, y, width, height)) { + goto fail; + } + + return; + +fail: + _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, + width, height, border); +} + +static void +r300CopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + xoffset, yoffset, x, y, width, height)) { + + //DEBUG_FALLBACKS + + _mesa_meta_CopyTexSubImage2D(ctx, target, level, + xoffset, yoffset, x, y, width, height); + } +} + + +void r300_init_texcopy_functions(struct dd_function_table *table) +{ + table->CopyTexImage2D = r300CopyTexImage2D; + table->CopyTexSubImage2D = r300CopyTexSubImage2D; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 9eaf390b460..eea3dee7bc6 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -46,19 +46,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" #include "r300_reg.h" -#define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5 \ - || ((f) >= MESA_FORMAT_RGBA_FLOAT32 && \ - (f) <= MESA_FORMAT_INTENSITY_FLOAT16)) \ - && tx_table[f].flag ) - -#define _ASSIGN(entry, format) \ - [ MESA_FORMAT_ ## entry ] = { format, 0, 1} - /* * Note that the _REV formats are the same as the non-REV formats. This is * because the REV and non-REV formats are identical as a byte string, but @@ -68,67 +59,119 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * identically. -- paulus */ -static const struct tx_table { - GLuint format, filter, flag; -} tx_table[] = { - /* *INDENT-OFF* */ +int32_t r300TranslateTexFormat(gl_format mesaFormat) +{ + switch (mesaFormat) + { #ifdef MESA_LITTLE_ENDIAN - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); #else - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); #endif - _ASSIGN(XRGB8888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), - _ASSIGN(RGB888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), - _ASSIGN(RGB565, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(RGB565_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(ARGB4444, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB4444_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB1555, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(ARGB1555_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(AL88, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(AL88_REV, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(RGB332, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2)), - _ASSIGN(A8, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8)), - _ASSIGN(L8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8)), - _ASSIGN(I8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(CI8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(YCBCR, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(YCBCR_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(RGB_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1)), - _ASSIGN(RGBA_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1)), - _ASSIGN(RGBA_DXT3, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3)), - _ASSIGN(RGBA_DXT5, R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5)), - _ASSIGN(RGBA_FLOAT32, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32)), - _ASSIGN(RGBA_FLOAT16, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16)), - _ASSIGN(RGB_FLOAT32, 0xffffffff), - _ASSIGN(RGB_FLOAT16, 0xffffffff), - _ASSIGN(ALPHA_FLOAT32, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32)), - _ASSIGN(ALPHA_FLOAT16, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16)), - _ASSIGN(LUMINANCE_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32)), - _ASSIGN(LUMINANCE_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16)), - _ASSIGN(INTENSITY_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, X, FL_I32)), - _ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)), - _ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)), - _ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)), - _ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)), - _ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)), - /* EXT_texture_sRGB */ - _ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA), - /* *INDENT-ON* */ + case MESA_FORMAT_XRGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB565: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_RGB565_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_ARGB4444: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB4444_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB1555: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_ARGB1555_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_AL88: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_AL88_REV: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_RGB332: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2); + case MESA_FORMAT_A8: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8); + case MESA_FORMAT_L8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8); + case MESA_FORMAT_I8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_CI8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_YCBCR: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_YCBCR_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_RGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1); + case MESA_FORMAT_RGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1); + case MESA_FORMAT_RGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3); + case MESA_FORMAT_RGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5); + case MESA_FORMAT_RGBA_FLOAT32: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32); + case MESA_FORMAT_RGBA_FLOAT16: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16); + case MESA_FORMAT_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32); + case MESA_FORMAT_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16); + case MESA_FORMAT_LUMINANCE_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32); + case MESA_FORMAT_LUMINANCE_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16); + case MESA_FORMAT_INTENSITY_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I32); + case MESA_FORMAT_INTENSITY_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I16); + case MESA_FORMAT_Z16: + return R300_EASY_TX_FORMAT(X, X, X, X, X16); + case MESA_FORMAT_Z24_S8: + return R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8); + case MESA_FORMAT_S8_Z24: + return R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8); + case MESA_FORMAT_Z32: + return R300_EASY_TX_FORMAT(X, X, X, X, X32); + /* EXT_texture_sRGB */ + case MESA_FORMAT_SRGBA8: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SLA8: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SL8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5) | R300_TX_FORMAT_GAMMA; + default: + return -1; + } }; -#undef _ASSIGN - void r300SetDepthTexMode(struct gl_texture_object *tObj) { static const GLuint formats[3][3] = { @@ -138,9 +181,9 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj) R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X16), }, { - R300_EASY_TX_FORMAT(X, X, X, ONE, X24_Y8), - R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8), - R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X24_Y8), + R300_EASY_TX_FORMAT(Y, Y, Y, ONE, X24_Y8), + R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8), + R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, Y, X24_Y8), }, { R300_EASY_TX_FORMAT(X, X, X, ONE, X32), @@ -205,19 +248,18 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) const struct gl_texture_image *firstImage; firstImage = t->base.Image[0][t->minLod]; - if (!t->image_override - && VALID_FORMAT(firstImage->TexFormat)) { + if (!t->image_override) { if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) { r300SetDepthTexMode(&t->base); } else { - t->pp_txformat = tx_table[firstImage->TexFormat].format; + int32_t txformat = r300TranslateTexFormat(firstImage->TexFormat); + if (txformat < 0) { + _mesa_problem(rmesa->radeon.glCtx, "%s: Invalid format %s", + __FUNCTION__, _mesa_get_format_name(firstImage->TexFormat)); + _mesa_exit(1); + } + t->pp_txformat = (uint32_t) txformat; } - - t->pp_txfilter |= tx_table[firstImage->TexFormat].filter; - } else if (!t->image_override) { - _mesa_problem(NULL, "unexpected texture format in %s", - __FUNCTION__); - return; } if (t->image_override && t->bo) @@ -357,18 +399,15 @@ void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname, switch (depth) { case 32: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 24: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 16: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } @@ -447,18 +486,15 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); else t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 3: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 2: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } diff --git a/src/mesa/drivers/dri/r300/radeon_bo.c b/src/mesa/drivers/dri/r300/radeon_bo.c new file mode 120000 index 00000000000..9448ffee54b --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_bo.c @@ -0,0 +1 @@ +../radeon/radeon_bo.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/radeon_bo_int_drm.h b/src/mesa/drivers/dri/r300/radeon_bo_int_drm.h new file mode 120000 index 00000000000..029450928be --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_bo_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_bo_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/radeon_cs.c b/src/mesa/drivers/dri/r300/radeon_cs.c new file mode 120000 index 00000000000..66b7ad1eb03 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_cs.c @@ -0,0 +1 @@ +../radeon/radeon_cs.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/radeon_cs_int_drm.h b/src/mesa/drivers/dri/r300/radeon_cs_int_drm.h new file mode 120000 index 00000000000..462f5245d0e --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_cs_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_cs_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile index 9b7c42042e6..26f47b72687 100644 --- a/src/mesa/drivers/dri/r600/Makefile +++ b/src/mesa/drivers/dri/r600/Makefile @@ -14,7 +14,7 @@ EGL_SOURCES = server/radeon_egl.c endif ifeq ($(RADEON_LDFLAGS),) -CS_SOURCES = radeon_cs_space_drm.c +CS_SOURCES = radeon_cs_space_drm.c radeon_bo.c radeon_cs.c endif COMMON_SOURCES = \ diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index d27a3245a39..370bb04f932 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -52,29 +52,49 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "radeon_reg.h" +#ifdef HAVE_LIBDRM_RADEON +#include "radeon_cs_int.h" +#else +#include "radeon_cs_int_drm.h" +#endif +struct r600_cs_manager_legacy +{ + struct radeon_cs_manager base; + struct radeon_context *ctx; + /* hack for scratch stuff */ + uint32_t pending_age; + uint32_t pending_count; +}; + +struct r600_cs_reloc_legacy { + struct radeon_cs_reloc base; + uint32_t cindices; + uint32_t *indices; + uint32_t *reloc_indices; +}; -static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm, - uint32_t ndw) +static struct radeon_cs_int *r600_cs_create(struct radeon_cs_manager *csm, + uint32_t ndw) { - struct radeon_cs *cs; + struct radeon_cs_int *csi; - cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs)); - if (cs == NULL) { + csi = (struct radeon_cs_int*)calloc(1, sizeof(struct radeon_cs_int)); + if (csi == NULL) { return NULL; } - cs->csm = csm; - cs->ndw = (ndw + 0x3FF) & (~0x3FF); - cs->packets = (uint32_t*)malloc(4*cs->ndw); - if (cs->packets == NULL) { - free(cs); + csi->csm = csm; + csi->ndw = (ndw + 0x3FF) & (~0x3FF); + csi->packets = (uint32_t*)malloc(4*csi->ndw); + if (csi->packets == NULL) { + free(csi); return NULL; } - cs->relocs_total_size = 0; - return cs; + csi->relocs_total_size = 0; + return csi; } -static int r600_cs_write_reloc(struct radeon_cs *cs, +static int r600_cs_write_reloc(struct radeon_cs_int *csi, struct radeon_bo *bo, uint32_t read_domain, uint32_t write_domain, @@ -83,7 +103,7 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, struct r600_cs_reloc_legacy *relocs; int i; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; /* check domains */ if ((read_domain && write_domain) || (!read_domain && !write_domain)) { /* in one CS a bo can only be in read or write domain but not @@ -98,7 +118,7 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, return -EINVAL; } /* check if bo is already referenced */ - for(i = 0; i < cs->crelocs; i++) { + for(i = 0; i < csi->crelocs; i++) { uint32_t *indices; uint32_t *reloc_indices; @@ -129,109 +149,108 @@ static int r600_cs_write_reloc(struct radeon_cs *cs, } relocs[i].indices = indices; relocs[i].reloc_indices = reloc_indices; - relocs[i].indices[relocs[i].cindices - 1] = cs->cdw; - relocs[i].reloc_indices[relocs[i].cindices - 1] = cs->cdw; - cs->section_cdw += 2; - cs->cdw += 2; + relocs[i].indices[relocs[i].cindices - 1] = csi->cdw; + relocs[i].reloc_indices[relocs[i].cindices - 1] = csi->cdw; + csi->section_cdw += 2; + csi->cdw += 2; return 0; } } /* add bo to reloc */ relocs = (struct r600_cs_reloc_legacy*) - realloc(cs->relocs, - sizeof(struct r600_cs_reloc_legacy) * (cs->crelocs + 1)); + realloc(csi->relocs, + sizeof(struct r600_cs_reloc_legacy) * (csi->crelocs + 1)); if (relocs == NULL) { return -ENOMEM; } - cs->relocs = relocs; - relocs[cs->crelocs].base.bo = bo; - relocs[cs->crelocs].base.read_domain = read_domain; - relocs[cs->crelocs].base.write_domain = write_domain; - relocs[cs->crelocs].base.flags = flags; - relocs[cs->crelocs].indices = (uint32_t*)malloc(4); - relocs[cs->crelocs].reloc_indices = (uint32_t*)malloc(4); - if ( (relocs[cs->crelocs].indices == NULL) || (relocs[cs->crelocs].reloc_indices == NULL) ) + csi->relocs = relocs; + relocs[csi->crelocs].base.bo = bo; + relocs[csi->crelocs].base.read_domain = read_domain; + relocs[csi->crelocs].base.write_domain = write_domain; + relocs[csi->crelocs].base.flags = flags; + relocs[csi->crelocs].indices = (uint32_t*)malloc(4); + relocs[csi->crelocs].reloc_indices = (uint32_t*)malloc(4); + if ( (relocs[csi->crelocs].indices == NULL) || (relocs[csi->crelocs].reloc_indices == NULL) ) { return -ENOMEM; } - relocs[cs->crelocs].indices[0] = cs->cdw; - relocs[cs->crelocs].reloc_indices[0] = cs->cdw; - cs->section_cdw += 2; - cs->cdw += 2; - relocs[cs->crelocs].cindices = 1; - cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo); - cs->crelocs++; + relocs[csi->crelocs].indices[0] = csi->cdw; + relocs[csi->crelocs].reloc_indices[0] = csi->cdw; + csi->section_cdw += 2; + csi->cdw += 2; + relocs[csi->crelocs].cindices = 1; + csi->relocs_total_size += radeon_bo_legacy_relocs_size(bo); + csi->crelocs++; radeon_bo_ref(bo); return 0; } -static int r600_cs_begin(struct radeon_cs *cs, +static int r600_cs_begin(struct radeon_cs_int *csi, uint32_t ndw, const char *file, const char *func, int line) { - if (cs->section) { + if (csi->section_ndw) { fprintf(stderr, "CS already in a section(%s,%s,%d)\n", - cs->section_file, cs->section_func, cs->section_line); + csi->section_file, csi->section_func, csi->section_line); fprintf(stderr, "CS can't start section(%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 1; - cs->section_ndw = ndw; - cs->section_cdw = 0; - cs->section_file = file; - cs->section_func = func; - cs->section_line = line; + csi->section_ndw = ndw; + csi->section_cdw = 0; + csi->section_file = file; + csi->section_func = func; + csi->section_line = line; - if (cs->cdw + ndw > cs->ndw) { + if (csi->cdw + ndw > csi->ndw) { uint32_t tmp, *ptr; int num = (ndw > 0x400) ? ndw : 0x400; - tmp = (cs->cdw + num + 0x3FF) & (~0x3FF); - ptr = (uint32_t*)realloc(cs->packets, 4 * tmp); + tmp = (csi->cdw + num + 0x3FF) & (~0x3FF); + ptr = (uint32_t*)realloc(csi->packets, 4 * tmp); if (ptr == NULL) { return -ENOMEM; } - cs->packets = ptr; - cs->ndw = tmp; + csi->packets = ptr; + csi->ndw = tmp; } return 0; } -static int r600_cs_end(struct radeon_cs *cs, +static int r600_cs_end(struct radeon_cs_int *csi, const char *file, const char *func, int line) { - if (!cs->section) { + if (!csi->section_ndw) { fprintf(stderr, "CS no section to end at (%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 0; - if ( cs->section_ndw != cs->section_cdw ) { + if ( csi->section_ndw != csi->section_cdw ) { fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n", - cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw); - fprintf(stderr, "cs->section_ndw = %d, cs->cdw = %d, cs->section_cdw = %d \n", - cs->section_ndw, cs->cdw, cs->section_cdw); + csi->section_file, csi->section_func, csi->section_line, csi->section_ndw, csi->section_cdw); + fprintf(stderr, "csi->section_ndw = %d, csi->cdw = %d, csi->section_cdw = %d \n", + csi->section_ndw, csi->cdw, csi->section_cdw); fprintf(stderr, "CS section end at (%s,%s,%d)\n", file, func, line); return -EPIPE; } + csi->section_ndw = 0; - if (cs->cdw > cs->ndw) { + if (csi->cdw > csi->ndw) { fprintf(stderr, "CS section overflow at (%s,%s,%d) cdw %d ndw %d\n", - cs->section_file, cs->section_func, cs->section_line,cs->cdw,cs->ndw); + csi->section_file, csi->section_func, csi->section_line,csi->cdw,csi->ndw); fprintf(stderr, "CS section end at (%s,%s,%d)\n", file, func, line); assert(0); @@ -240,20 +259,20 @@ static int r600_cs_end(struct radeon_cs *cs, return 0; } -static int r600_cs_process_relocs(struct radeon_cs *cs, +static int r600_cs_process_relocs(struct radeon_cs_int *csi, uint32_t * reloc_chunk, uint32_t * length_dw_reloc_chunk) { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct r600_cs_reloc_legacy *relocs; int i, j, r; uint32_t offset_dw = 0; - csm = (struct r600_cs_manager_legacy*)cs->csm; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + csm = (struct r600_cs_manager_legacy*)csi->csm; + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; restart: - for (i = 0; i < cs->crelocs; i++) { + for (i = 0; i < csi->crelocs; i++) { uint32_t soffset, eoffset; r = radeon_bo_legacy_validate(relocs[i].base.bo, @@ -269,9 +288,9 @@ restart: for (j = 0; j < relocs[i].cindices; j++) { /* pkt3 nop header in ib chunk */ - cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000; + csi->packets[relocs[i].reloc_indices[j]] = 0xC0001000; /* reloc index in ib chunk */ - cs->packets[relocs[i].reloc_indices[j] + 1] = offset_dw; + csi->packets[relocs[i].reloc_indices[j] + 1] = offset_dw; } /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */ @@ -286,14 +305,14 @@ restart: return 0; } -static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */ +static int r600_cs_set_age(struct radeon_cs_int *csi) /* -------------- */ { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct r600_cs_reloc_legacy *relocs; int i; - relocs = (struct r600_cs_reloc_legacy *)cs->relocs; - for (i = 0; i < cs->crelocs; i++) { + relocs = (struct r600_cs_reloc_legacy *)csi->relocs; + for (i = 0; i < csi->crelocs; i++) { radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age); radeon_bo_unref(relocs[i].base.bo); } @@ -301,21 +320,21 @@ static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */ } #if 0 -static void dump_cmdbuf(struct radeon_cs *cs) +static void dump_cmdbuf(struct radeon_cs_int *csi) { int i; fprintf(stderr,"--start--\n"); - for (i = 0; i < cs->cdw; i++){ - fprintf(stderr,"0x%08x\n", cs->packets[i]); + for (i = 0; i < csi->cdw; i++){ + fprintf(stderr,"0x%08x\n", csi->packets[i]); } fprintf(stderr,"--end--\n"); } #endif -static int r600_cs_emit(struct radeon_cs *cs) +static int r600_cs_emit(struct radeon_cs_int *csi) { - struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)csi->csm; struct drm_radeon_cs cs_cmd; struct drm_radeon_cs_chunk cs_chunk[2]; uint32_t length_dw_reloc_chunk; @@ -329,9 +348,9 @@ static int r600_cs_emit(struct radeon_cs *cs) csm->pending_count = 1; - reloc_chunk = (uint32_t*)calloc(1, cs->crelocs * 4 * 4); + reloc_chunk = (uint32_t*)calloc(1, csi->crelocs * 4 * 4); - r = r600_cs_process_relocs(cs, reloc_chunk, &length_dw_reloc_chunk); + r = r600_cs_process_relocs(csi, reloc_chunk, &length_dw_reloc_chunk); if (r) { free(reloc_chunk); return 0; @@ -339,8 +358,8 @@ static int r600_cs_emit(struct radeon_cs *cs) /* raw ib chunk */ cs_chunk[0].chunk_id = RADEON_CHUNK_ID_IB; - cs_chunk[0].length_dw = cs->cdw; - cs_chunk[0].chunk_data = (unsigned long)(cs->packets); + cs_chunk[0].length_dw = csi->cdw; + cs_chunk[0].chunk_data = (unsigned long)(csi->packets); /* reloc chaunk */ cs_chunk[1].chunk_id = RADEON_CHUNK_ID_RELOCS; @@ -358,7 +377,7 @@ static int r600_cs_emit(struct radeon_cs *cs) do { - r = drmCommandWriteRead(cs->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd)); + r = drmCommandWriteRead(csi->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd)); retry++; } while (r == -EAGAIN && retry < 1000); @@ -369,11 +388,11 @@ static int r600_cs_emit(struct radeon_cs *cs) csm->pending_age = cs_cmd.cs_id; - r600_cs_set_age(cs); + r600_cs_set_age(csi); - cs->csm->read_used = 0; - cs->csm->vram_write_used = 0; - cs->csm->gart_write_used = 0; + csi->csm->read_used = 0; + csi->csm->vram_write_used = 0; + csi->csm->gart_write_used = 0; free(reloc_chunk); @@ -393,35 +412,34 @@ static void inline r600_cs_free_reloc(void *relocs_p, int crelocs) } } -static int r600_cs_destroy(struct radeon_cs *cs) +static int r600_cs_destroy(struct radeon_cs_int *csi) { - r600_cs_free_reloc(cs->relocs, cs->crelocs); - free(cs->relocs); - free(cs->packets); - free(cs); + r600_cs_free_reloc(csi->relocs, csi->crelocs); + free(csi->relocs); + free(csi->packets); + free(csi); return 0; } -static int r600_cs_erase(struct radeon_cs *cs) +static int r600_cs_erase(struct radeon_cs_int *csi) { - r600_cs_free_reloc(cs->relocs, cs->crelocs); - free(cs->relocs); - cs->relocs_total_size = 0; - cs->relocs = NULL; - cs->crelocs = 0; - cs->cdw = 0; - cs->section = 0; + r600_cs_free_reloc(csi->relocs, csi->crelocs); + free(csi->relocs); + csi->relocs_total_size = 0; + csi->relocs = NULL; + csi->crelocs = 0; + csi->cdw = 0; return 0; } -static int r600_cs_need_flush(struct radeon_cs *cs) +static int r600_cs_need_flush(struct radeon_cs_int *csi) { /* this function used to flush when the BO usage got to * a certain size, now the higher levels handle this better */ return 0; } -static void r600_cs_print(struct radeon_cs *cs, FILE *file) +static void r600_cs_print(struct radeon_cs_int *csi, FILE *file) { } diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h index eba43d37b6b..dff00096999 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.h +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h @@ -118,22 +118,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R600_IT_SET_CTL_CONST 0x00006F00 #define R600_IT_SURFACE_BASE_UPDATE 0x00007300 -struct r600_cs_manager_legacy -{ - struct radeon_cs_manager base; - struct radeon_context *ctx; - /* hack for scratch stuff */ - uint32_t pending_age; - uint32_t pending_count; -}; - -struct r600_cs_reloc_legacy { - struct radeon_cs_reloc base; - uint32_t cindices; - uint32_t *indices; - uint32_t *reloc_indices; -}; - struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx); /** diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 25314eff563..45bbc3c071e 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -74,7 +74,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -//#define R600_ENABLE_GLSL_TEST 1 +#define R600_ENABLE_GLSL_TEST 1 #define need_GL_VERSION_2_0 #define need_GL_ARB_occlusion_query @@ -317,20 +317,8 @@ static void r600InitGLExtensions(GLcontext *ctx) #ifdef R600_ENABLE_GLSL_TEST driInitExtensions(ctx, gl_20_extension, GL_TRUE); - //_mesa_enable_2_0_extensions(ctx); - //1.5 - ctx->Extensions.ARB_occlusion_query = GL_TRUE; - ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; - ctx->Extensions.EXT_shadow_funcs = GL_TRUE; - //2.0 - ctx->Extensions.ARB_draw_buffers = GL_TRUE; - ctx->Extensions.ARB_point_sprite = GL_TRUE; - ctx->Extensions.ARB_shader_objects = GL_TRUE; - ctx->Extensions.ARB_vertex_shader = GL_TRUE; - ctx->Extensions.ARB_fragment_shader = GL_TRUE; - ctx->Extensions.EXT_blend_equation_separate = GL_TRUE; - ctx->Extensions.ATI_separate_stencil = GL_TRUE; - + _mesa_enable_2_0_extensions(ctx); + /* glsl compiler has problem if this is not GL_TRUE */ ctx->Shader.EmitCondCodes = GL_TRUE; #endif /* R600_ENABLE_GLSL_TEST */ diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index aed84fc3bd3..51692a11ffb 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -32,6 +32,7 @@ #include "main/mtypes.h" #include "main/imports.h" +#include "shader/prog_parameter.h" #include "radeon_debug.h" #include "r600_context.h" @@ -41,6 +42,39 @@ #define USE_CF_FOR_CONTINUE_BREAK 1 #define USE_CF_FOR_POP_AFTER 1 +struct prog_instruction noise1_insts[12] = { + {OPCODE_BGNSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 2, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 4, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 585, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 8, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_SGT , {{0, 0, 585, 0, 0, 0}, {8, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 1, 1, 0, 8, 1672, 0}, 1, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_IF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 7, 0, 0}, 0, 0, 0, 1, 0, 0, 0, 15, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1755, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDIF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0} +}; +float noise1_const[2][4] = { + {0.300000f, 0.900000f, 0.500000f, 0.300000f} +}; + +COMPILED_SUB noise1_presub = { + &(noise1_insts[0]), + 12, + 2, + 1, + 0, + &(noise1_const[0]), + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + {0,0,0}, + 0 +}; + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -330,14 +364,14 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size) return(format); } -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3) { - if(pAsm->D.dst.op3) + if(nIsOp3 > 0) { return 3; } - switch (pAsm->D.dst.opcode) + switch (opcode) { case SQ_OP2_INST_ADD: case SQ_OP2_INST_KILLE: @@ -378,7 +412,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) return 1; default: radeon_error( - "Need instruction operand number for %x.\n", pAsm->D.dst.opcode); + "Need instruction operand number for %x.\n", opcode); }; return 3; @@ -500,12 +534,20 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->unCFflags = 0; + pAsm->presubs = NULL; + pAsm->unPresubArraySize = 0; + pAsm->unNumPresub = 0; + pAsm->unCurNumILInsts = 0; + + pAsm->unVetTexBits = 0; + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) { - if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) ) + if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) || + (OPCODE_DDX==Opcode) || (OPCODE_DDY==Opcode) ) { return GL_TRUE; } @@ -1220,7 +1262,15 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } pAsm->S[fld].src.rtype = SRC_REG_CONSTANT; - pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + if(pILInst->SrcReg[src].Index < 0) + { + WARN_ONCE("Negative register offsets not supported yet!\n"); + pAsm->S[fld].src.reg = 0; + } + else + { + pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; + } break; case PROGRAM_INPUT: setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); @@ -1373,43 +1423,65 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; break; case PROGRAM_INPUT: - switch (pILInst->SrcReg[0].Index) + if(SPT_VP == pAsm->currentShaderType) { - case FRAG_ATTRIB_WPOS: - case FRAG_ATTRIB_COL0: - case FRAG_ATTRIB_COL1: - case FRAG_ATTRIB_FOGC: - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - bValidTexCoord = GL_TRUE; + switch (pILInst->SrcReg[0].Index) + { + case VERT_ATTRIB_TEX0: + case VERT_ATTRIB_TEX1: + case VERT_ATTRIB_TEX2: + case VERT_ATTRIB_TEX3: + case VERT_ATTRIB_TEX4: + case VERT_ATTRIB_TEX5: + case VERT_ATTRIB_TEX6: + case VERT_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->ucVP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + } + } + else + { + switch (pILInst->SrcReg[0].Index) + { + case FRAG_ATTRIB_WPOS: + case FRAG_ATTRIB_COL0: + case FRAG_ATTRIB_COL1: + case FRAG_ATTRIB_FOGC: + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + case FRAG_ATTRIB_FACE: + fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); + break; + case FRAG_ATTRIB_PNTC: + fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); + break; + } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; pAsm->S[0].src.reg = pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; pAsm->S[0].src.rtype = SRC_REG_INPUT; - break; - case FRAG_ATTRIB_FACE: - fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); - break; - case FRAG_ATTRIB_PNTC: - fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); - break; - } - - if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || - (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) - { - bValidTexCoord = GL_TRUE; - pAsm->S[0].src.reg = - pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; - pAsm->S[0].src.rtype = SRC_REG_INPUT; + } } - break; + break; } } @@ -1454,8 +1526,17 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word0.f.tex_inst = pAsm->D.dst.opcode; tex_instruction_ptr->m_Word0.f.bc_frac_mode = 0x0; tex_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0; + tex_instruction_ptr->m_Word0.f.alt_const = 0; - tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + if(SPT_VP == pAsm->currentShaderType) + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg + VERT_ATTRIB_MAX; + pAsm->unVetTexBits |= 1 << texture_unit_source->reg; + } + else + { + tex_instruction_ptr->m_Word0.f.resource_id = texture_unit_source->reg; + } tex_instruction_ptr->m_Word1.f.lod_bias = 0x0; if (normalized) { @@ -1474,7 +1555,6 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize tex_instruction_ptr->m_Word2.f.offset_x = 0x0; tex_instruction_ptr->m_Word2.f.offset_y = 0x0; tex_instruction_ptr->m_Word2.f.offset_z = 0x0; - tex_instruction_ptr->m_Word2.f.sampler_id = texture_unit_source->reg; // dst @@ -2010,7 +2090,7 @@ GLboolean check_scalar(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2099,7 +2179,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -2180,7 +2260,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) int current_source_index; GLuint contiguous_slots_needed; - GLuint uNumSrc = r700GetNumOperands(pAsm); + GLuint uNumSrc = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); //GLuint channel_swizzle, j; //GLuint chan_counter[4] = {0, 0, 0, 0}; //PVSSRC * pSource[3]; @@ -4292,13 +4372,20 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) } - if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB) - { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; - } - else + switch(pAsm->pILInst[pAsm->uiCurInst].Opcode) { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; + case OPCODE_DDX: + /* will these need WQM(1) on CF inst ? */ + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_H; + break; + case OPCODE_DDY: + pAsm->D.dst.opcode = SQ_TEX_INST_GET_GRADIENTS_V; + break; + case OPCODE_TXB: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; + break; + default: + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; } pAsm->is_tex = GL_TRUE; @@ -4968,7 +5055,7 @@ void add_return_inst(r700_AssemblerBase *pAsm) pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; } -GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift) { /* Put in sub */ if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) @@ -4983,7 +5070,7 @@ GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex) pAsm->unSubArraySize += 10; } - pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex; + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex + uiIL_Shift; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; @@ -5074,9 +5161,13 @@ GLboolean assemble_RET(r700_AssemblerBase *pAsm) GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLint nILindex, + GLuint uiIL_Shift, GLuint uiNumberInsts, - struct prog_instruction *pILInst) + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc) { + GLint uiIL_Offset; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; if(GL_FALSE == add_cf_instruction(pAsm) ) @@ -5109,8 +5200,12 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, pAsm->unCallerArraySize += 10; } - pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = nILindex; - pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + uiIL_Offset = nILindex + uiIL_Shift; + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = uiIL_Offset; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->callers[pAsm->unCallerArrayPointer].finale_cf_ptr = NULL; + pAsm->callers[pAsm->unCallerArrayPointer].prelude_cf_ptr = NULL; pAsm->unCallerArrayPointer++; @@ -5120,7 +5215,7 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLboolean bRet; for(j=0; j<pAsm->unSubArrayPointer; j++) { - if(nILindex == pAsm->subs[j].subIL_Offset) + if(uiIL_Offset == pAsm->subs[j].subIL_Offset) { /* compiled before */ max = pAsm->subs[j].unStackDepthMax @@ -5138,7 +5233,7 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; unSubID = pAsm->unSubArrayPointer; - bRet = AssembleInstr(nILindex, uiNumberInsts, pILInst, pAsm); + bRet = AssembleInstr(nILindex, uiIL_Shift, uiNumberInsts, pILInst, pAsm); if(GL_TRUE == bRet) { @@ -5148,6 +5243,8 @@ GLboolean assemble_CAL(r700_AssemblerBase *pAsm, { pAsm->CALLSTACK[pAsm->CALLSP].max = max; } + + pAsm->subs[unSubID].pPresubDesc = pPresubDesc; } return bRet; @@ -5313,6 +5410,7 @@ GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) } GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) @@ -5468,6 +5566,26 @@ GLboolean AssembleInstr(GLuint uiFirstInst, case OPCODE_MUL: if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) return GL_FALSE; + break; + + case OPCODE_NOISE1: + { + callPreSub(pR700AsmCode, + GLSL_NOISE1, + &noise1_presub, + pILInst->DstReg.Index + pR700AsmCode->starting_temp_register_number, + 1); + radeon_error("noise1: not yet supported shader instruction\n"); + }; + break; + case OPCODE_NOISE2: + radeon_error("noise2: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE3: + radeon_error("noise3: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE4: + radeon_error("noise4: not yet supported shader instruction\n"); break; case OPCODE_POW: @@ -5580,7 +5698,8 @@ GLboolean AssembleInstr(GLuint uiFirstInst, } } break; - + case OPCODE_DDX: + case OPCODE_DDY: case OPCODE_TEX: case OPCODE_TXB: case OPCODE_TXP: @@ -5653,7 +5772,7 @@ GLboolean AssembleInstr(GLuint uiFirstInst, break; case OPCODE_BGNSUB: - if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i) ) + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i, uiIL_Shift) ) { return GL_FALSE; } @@ -5668,9 +5787,11 @@ GLboolean AssembleInstr(GLuint uiFirstInst, case OPCODE_CAL: if( GL_FALSE == assemble_CAL(pR700AsmCode, - pILInst[i].BranchTarget, + pILInst[i].BranchTarget, + uiIL_Shift, uiNumberInsts, - pILInst) ) + pILInst, + NULL) ) { return GL_FALSE; } @@ -5707,7 +5828,7 @@ GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) return GL_TRUE; } -GLboolean RelocProgram(r700_AssemblerBase * pAsm) +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg) { GLuint i; GLuint unCFoffset; @@ -5717,6 +5838,12 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) R700ShaderInstruction * pInst; R700ControlFlowGenericClause * pCFInst; + R700ControlFlowALUClause * pCF_ALU; + R700ALUInstruction * pALU; + GLuint unConstOffset = 0; + GLuint unRegOffset; + GLuint unMinRegIndex; + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; /* remove flags init if they are not used */ @@ -5762,6 +5889,11 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) unCFoffset = plstCFmain->uNumOfNode; + if(NULL != pILProg->Parameters) + { + unConstOffset = pILProg->Parameters->NumParameters; + } + /* Reloc subs */ for(i=0; i<pAsm->unSubArrayPointer; i++) { @@ -5799,6 +5931,84 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) pInst = pInst->pNextInst; }; + if(NULL != pAsm->subs[i].pPresubDesc) + { + GLuint uNumSrc; + + unMinRegIndex = pAsm->subs[i].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[i].pPresubDesc->maxStartReg; + unConstOffset += pAsm->subs[i].pPresubDesc->unConstantsStart; + + pInst = plstCFsub->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + + if(pALU->m_Word0.f.src0_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src0_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src0_sel += unConstOffset; + } + + if( ((pALU->m_Word1.val >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT) & 0x0000001F) + >= SQ_OP3_INST_MUL_LIT ) + { /* op3 : 3 srcs */ + if(pALU->m_Word1_OP3.f.src2_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word1_OP3.f.src2_sel = pALU->m_Word1_OP3.f.src2_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word1_OP3.f.src2_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word1_OP3.f.src2_sel += unConstOffset; + } + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + else + { + if(pAsm->bR6xx) + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f6.alu_inst, 0); + } + else + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f.alu_inst, 0); + } + if(2 == uNumSrc) + { /* 2 srcs */ + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + } + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + pInst = pInst->pNextInst; + }; + } + /* Put sub into main */ plstCFmain->pTail->pNextInst = plstCFsub->pHead; plstCFmain->pTail = plstCFsub->pTail; @@ -5812,11 +6022,216 @@ GLboolean RelocProgram(r700_AssemblerBase * pAsm) { pAsm->callers[i].cf_ptr->m_Word0.f.addr = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + + if(NULL != pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc) + { + unMinRegIndex = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->maxStartReg; + + if(NULL != pAsm->callers[i].prelude_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].prelude_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + if(NULL != pAsm->callers[i].finale_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].finale_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + } } return GL_TRUE; } +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc) +{ + /* save assemble context */ + GLuint starting_temp_register_number_save; + GLuint number_used_registers_save; + GLuint uFirstHelpReg_save; + GLuint uHelpReg_save; + GLuint uiCurInst_save; + struct prog_instruction *pILInst_save; + PRESUB_DESC * pPresubDesc; + GLboolean bRet; + int i; + + R700ControlFlowGenericClause* prelude_cf_ptr = NULL; + + /* copy srcs to presub inputs */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + for(i=0; i<uNumValidSrc; i++) + { + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pCompiledSub->srcRegIndex[i]; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 1; + pAsm->D.dst.writew = 1; + + if( GL_FALSE == assemble_src(pAsm, i, 0) ) + { + return GL_FALSE; + } + + next_ins(pAsm); + } + if(uNumValidSrc > 0) + { + prelude_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + /* browse thro existing presubs. */ + for(i=0; i<pAsm->unNumPresub; i++) + { + if(pAsm->presubs[i].sptSigniture == scriptSigniture) + { + break; + } + } + + if(i == pAsm->unNumPresub) + { /* not loaded yet */ + /* save assemble context */ + number_used_registers_save = pAsm->number_used_registers; + uFirstHelpReg_save = pAsm->uFirstHelpReg; + uHelpReg_save = pAsm->uHelpReg; + starting_temp_register_number_save = pAsm->starting_temp_register_number; + pILInst_save = pAsm->pILInst; + uiCurInst_save = pAsm->uiCurInst; + + /* alloc in presub */ + if( (pAsm->unNumPresub + 1) > pAsm->unPresubArraySize ) + { + pAsm->presubs = (PRESUB_DESC*)_mesa_realloc( (void *)pAsm->presubs, + sizeof(PRESUB_DESC) * pAsm->unPresubArraySize, + sizeof(PRESUB_DESC) * (pAsm->unPresubArraySize + 4) ); + if(NULL == pAsm->presubs) + { + radeon_error("No memeory to allocate built in shader function description structures. \n"); + return GL_FALSE; + } + pAsm->unPresubArraySize += 4; + } + + pPresubDesc = &(pAsm->presubs[i]); + pPresubDesc->sptSigniture = scriptSigniture; + + /* constants offsets need to be final resolved at reloc. */ + if(0 == pAsm->unNumPresub) + { + pPresubDesc->unConstantsStart = 0; + } + else + { + pPresubDesc->unConstantsStart = pAsm->presubs[i-1].unConstantsStart + + pAsm->presubs[i-1].pCompiledSub->NumParameters; + } + + pPresubDesc->pCompiledSub = pCompiledSub; + + pPresubDesc->subIL_Shift = pAsm->unCurNumILInsts; + pPresubDesc->maxStartReg = uFirstHelpReg_save; + pAsm->unCurNumILInsts += pCompiledSub->NumInstructions; + + pAsm->unNumPresub++; + + /* setup new assemble context */ + pAsm->starting_temp_register_number = 0; + pAsm->number_used_registers = pCompiledSub->NumTemporaries; + pAsm->uFirstHelpReg = pAsm->number_used_registers; + pAsm->uHelpReg = pAsm->uFirstHelpReg; + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + + + pPresubDesc->number_used_registers = pAsm->number_used_registers; + + /* restore assemble context */ + pAsm->number_used_registers = number_used_registers_save; + pAsm->uFirstHelpReg = uFirstHelpReg_save; + pAsm->uHelpReg = uHelpReg_save; + pAsm->starting_temp_register_number = starting_temp_register_number_save; + pAsm->pILInst = pILInst_save; + pAsm->uiCurInst = uiCurInst_save; + } + else + { /* was loaded */ + pPresubDesc = &(pAsm->presubs[i]); + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + } + + if(GL_FALSE == bRet) + { + radeon_error("Shader presub assemble failed. \n"); + } + else + { + /* copy presub output to real dst */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pCompiledSub->dstRegIndex; + pAsm->S[0].src.swizzlex = pCompiledSub->outputSwizzleX; + pAsm->S[0].src.swizzley = pCompiledSub->outputSwizzleY; + pAsm->S[0].src.swizzlez = pCompiledSub->outputSwizzleZ; + pAsm->S[0].src.swizzlew = pCompiledSub->outputSwizzleW; + + next_ins(pAsm); + + pAsm->callers[pAsm->unCallerArrayPointer - 1].finale_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->callers[pAsm->unCallerArrayPointer - 1].prelude_cf_ptr = prelude_cf_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + if( (pPresubDesc->number_used_registers + pAsm->uFirstHelpReg) > pAsm->number_used_registers ) + { + pAsm->number_used_registers = pPresubDesc->number_used_registers + pAsm->uFirstHelpReg; + } + if(pAsm->uFirstHelpReg > pPresubDesc->maxStartReg) + { + pPresubDesc->maxStartReg = pAsm->uFirstHelpReg; + } + + return bRet; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -6174,6 +6589,11 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) FREE(pR700AsmCode->callers); } + if(NULL != pR700AsmCode->presubs) + { + FREE(pR700AsmCode->presubs); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index 6dc44017eb1..dbd9860f7d0 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -34,6 +34,45 @@ #include "r700_shaderinst.h" #include "r700_shader.h" +typedef enum LOADABLE_SCRIPT_SIGNITURE +{ + GLSL_NOISE1 = 0x10000001, + GLSL_NOISE2 = 0x10000002, + GLSL_NOISE3 = 0x10000003, + GLSL_NOISE4 = 0x10000004 +}LOADABLE_SCRIPT_SIGNITURE; + +typedef struct COMPILED_SUB +{ + struct prog_instruction *Instructions; + GLuint NumInstructions; + GLuint NumTemporaries; + GLuint NumParameters; + GLuint MinRegIndex; + GLfloat (*ParameterValues)[4]; + GLbyte outputSwizzleX; + GLbyte outputSwizzleY; + GLbyte outputSwizzleZ; + GLbyte outputSwizzleW; + GLshort srcRegIndex[3]; + GLushort dstRegIndex; +}COMPILED_SUB; + +typedef struct PRESUB_DESCtag +{ + LOADABLE_SCRIPT_SIGNITURE sptSigniture; + GLint subIL_Shift; + struct prog_src_register InReg[3]; + struct prog_dst_register OutReg; + + GLushort maxStartReg; + GLushort number_used_registers; + + GLuint unConstantsStart; + + COMPILED_SUB * pCompiledSub; +} PRESUB_DESC; + typedef enum SHADER_PIPE_TYPE { SPT_VP = 0, @@ -296,6 +335,7 @@ typedef struct SUB_OFFSET GLint subIL_Offset; GLuint unCFoffset; GLuint unStackDepthMax; + PRESUB_DESC * pPresubDesc; TypedShaderList lstCFInstructions_local; } SUB_OFFSET; @@ -304,6 +344,9 @@ typedef struct CALLER_POINTER GLint subIL_Offset; GLint subDescIndex; R700ControlFlowGenericClause* cf_ptr; + + R700ControlFlowGenericClause* prelude_cf_ptr; + R700ControlFlowGenericClause* finale_cf_ptr; } CALLER_POINTER; #define SQ_MAX_CALL_DEPTH 0x00000020 @@ -437,6 +480,13 @@ typedef struct r700_AssemblerBase GLuint unCFflags; + PRESUB_DESC * presubs; + GLuint unPresubArraySize; + GLuint unNumPresub; + GLuint unCurNumILInsts; + + GLuint unVetTexBits; + } r700_AssemblerBase; //Internal use @@ -458,7 +508,7 @@ BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ; GLboolean is_reduction_opcode(PVSDWORD * dest); GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size); -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm); +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3); GLboolean IsTex(gl_inst_opcode Opcode); GLboolean IsAlu(gl_inst_opcode Opcode); @@ -585,13 +635,15 @@ GLboolean assemble_BRK(r700_AssemblerBase *pAsm); GLboolean assemble_COND(r700_AssemblerBase *pAsm); GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); -GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex); +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift); GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); GLboolean assemble_RET(r700_AssemblerBase *pAsm); GLboolean assemble_CAL(r700_AssemblerBase *pAsm, GLint nILindex, + GLuint uiIL_Offest, GLuint uiNumberInsts, - struct prog_instruction *pILInst); + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc); GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, @@ -602,16 +654,23 @@ GLboolean Process_Export(r700_AssemblerBase* pAsm, GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select); +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + /* struct prog_instruction ** pILInstParent, */ + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc); //Interface GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); -GLboolean RelocProgram(r700_AssemblerBase * pAsm); +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg); GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index d7027400146..c124e02184f 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -45,6 +45,9 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + struct r700_vertex_program *vp = context->selected_vp; + struct radeon_bo *bo = NULL; unsigned int i; BATCH_LOCALS(&context->radeon); @@ -52,7 +55,7 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; uint32_t offset; if (t) { @@ -71,7 +74,16 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) BEGIN_BATCH_NO_AUTOSTATE(9 + 4); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - R600_OUT_BATCH(i * 7); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i + VERT_ATTRIB_MAX + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + } + else + { + R600_OUT_BATCH(i * 7); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2); @@ -95,21 +107,35 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) } } +#define SAMPLER_STRIDE 3 + static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); unsigned int i; + + struct r700_vertex_program *vp = context->selected_vp; + BATCH_LOCALS(&context->radeon); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; if (t) { BEGIN_BATCH_NO_AUTOSTATE(5); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3)); - R600_OUT_BATCH(i * 3); + + if( (1<<i) & vp->r700AsmCode.unVetTexBits ) + { /* vs texture */ + R600_OUT_BATCH((i+SQ_TEX_SAMPLER_VS_OFFSET) * SAMPLER_STRIDE); //work 1 + } + else + { + R600_OUT_BATCH(i * 3); + } + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1); R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2); @@ -1163,7 +1189,11 @@ static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom) count += 3; if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) { - for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) { + /* targets are enabled in r700SetRenderTarget but state + size is calculated before that. Until MRT's are done + hardcode target0 as enabled. */ + count += 3; + for (ui = 1; ui < R700_MAX_RENDER_TARGETS; ui++) { if (r700->render_target[ui].enabled) count += 3; } diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index c6546ab00c2..526d3843d16 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -57,6 +57,10 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x\n", __func__, mask); + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + context->radeon.front_buffer_dirty = GL_TRUE; + } + if( GL_TRUE == r700ClearFast(context, mask) ) { return; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index 8eb439a9519..ce2d9fdf79e 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -34,6 +34,7 @@ #include "main/imports.h" #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" +#include "shader/program.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -42,6 +43,54 @@ #include "r700_debug.h" +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog) +{ + static const gl_state_index winstate[STATE_LENGTH] + = { STATE_INTERNAL, STATE_FB_SIZE, 0, 0, 0}; + struct prog_instruction *newInst, *inst; + GLint win_size; /* state reference */ + GLuint wpos_temp; /* temp register */ + int i, j; + + /* PARAM win_size = STATE_FB_SIZE */ + win_size = _mesa_add_state_reference(fprog->Base.Parameters, winstate); + + wpos_temp = fprog->Base.NumTemporaries++; + + /* scan program where WPOS is used and replace with wpos_temp */ + inst = fprog->Base.Instructions; + for (i = 0; i < fprog->Base.NumInstructions; i++) { + for (j=0; j < 3; j++) { + if(inst->SrcReg[j].File == PROGRAM_INPUT && + inst->SrcReg[j].Index == FRAG_ATTRIB_WPOS) { + inst->SrcReg[j].File = PROGRAM_TEMPORARY; + inst->SrcReg[j].Index = wpos_temp; + } + } + inst++; + } + + _mesa_insert_instructions(&(fprog->Base), 0, 1); + + newInst = fprog->Base.Instructions; + /* invert wpos.y + * wpos_temp.xyzw = wpos.x-yzw + winsize.0y00 */ + newInst[0].Opcode = OPCODE_ADD; + newInst[0].DstReg.File = PROGRAM_TEMPORARY; + newInst[0].DstReg.Index = wpos_temp; + newInst[0].DstReg.WriteMask = WRITEMASK_XYZW; + + newInst[0].SrcReg[0].File = PROGRAM_INPUT; + newInst[0].SrcReg[0].Index = FRAG_ATTRIB_WPOS; + newInst[0].SrcReg[0].Swizzle = SWIZZLE_XYZW; + newInst[0].SrcReg[0].Negate = NEGATE_Y; + + newInst[0].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[0].SrcReg[1].Index = win_size; + newInst[0].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ZERO); + +} + //TODO : Validate FP input with VP output. void Map_Fragment_Program(r700_AssemblerBase *pAsm, struct gl_fragment_program *mesa_fp, @@ -155,6 +204,12 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE] = pAsm->number_used_registers++; } + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC] = pAsm->number_used_registers++; + } + /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -312,7 +367,13 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, //Init_Program Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) ); - Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); + + if(mesa_fp->Base.InputsRead & FRAG_BIT_WPOS) + { + insert_wpos_code(ctx, mesa_fp); + } + + Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) ) { @@ -325,7 +386,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, { fp->r700AsmCode.SamplerUnits[i] = fp->mesa_program.Base.SamplerUnits[i]; } + + fp->r700AsmCode.unCurNumILInsts = mesa_fp->Base.NumInstructions; + if( GL_FALSE == AssembleInstr(0, + 0, mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) @@ -338,7 +403,7 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } - if( GL_FALSE == RelocProgram(&(fp->r700AsmCode)) ) + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode), &(mesa_fp->Base)) ) { return GL_FALSE; } @@ -408,6 +473,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unsigned int unNumOfReg; unsigned int unBit; GLuint exportCount; + GLboolean point_sprite = GL_FALSE; if(GL_FALSE == fp->loaded) { @@ -474,6 +540,36 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); } + /* see if we need any point_sprite replacements */ + for (i = VERT_RESULT_TEX0; i<= VERT_RESULT_TEX7; i++) + { + if(ctx->Point.CoordReplace[i - VERT_RESULT_TEX0] == GL_TRUE) + point_sprite = GL_TRUE; + } + + if ((mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) || point_sprite) + { + /* for FRAG_ATTRIB_PNTC we need to increase num_interp */ + if(mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_PNTC)) + { + ui++; + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask); + } + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_S, PNT_SPRITE_OVRD_X_shift, PNT_SPRITE_OVRD_X_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_T, PNT_SPRITE_OVRD_Y_shift, PNT_SPRITE_OVRD_Y_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_0, PNT_SPRITE_OVRD_Z_shift, PNT_SPRITE_OVRD_Z_mask); + SETfield(r700->SPI_INTERP_CONTROL_0.u32All, SPI_PNT_SPRITE_SEL_1, PNT_SPRITE_OVRD_W_shift, PNT_SPRITE_OVRD_W_mask); + if(ctx->Point.SpriteOrigin == GL_LOWER_LEFT) + SETbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + else + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_TOP_1_bit); + } + else + { + CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, PNT_SPRITE_ENA_bit); + } + ui = (unNumOfReg < ui) ? ui : unNumOfReg; @@ -494,6 +590,10 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) struct r700_vertex_program_cont *vpc = (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + + for(ui = 0; ui < R700_MAX_SHADER_EXPORTS; ui++) + r700->SPI_PS_INPUT_CNTL[ui].u32All = 0; + unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) { @@ -556,6 +656,11 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, SEMANTIC_shift, SEMANTIC_mask); CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + /* ARB_point_sprite */ + if(ctx->Point.CoordReplace[i] == GL_TRUE) + { + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } } } @@ -571,6 +676,22 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) else CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } + unBit = 1 << FRAG_ATTRIB_PNTC; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_PNTC]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, PT_SPRITE_TEX_bit); + } + + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) { @@ -620,6 +741,25 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } else r700->ps.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->ps.num_consts; + for(ui=0; ui<pAsm->unNumPresub; ui++) + { + pCompiledSub = pAsm->presubs[ui].pCompiledSub; + + r700->ps.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->ps.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->ps.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->ps.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->ps.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index e562bfa4789..39c59c9201d 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -48,6 +48,8 @@ struct r700_fragment_program }; /* Internal */ +void insert_wpos_code(GLcontext *ctx, struct gl_fragment_program *fprog); + void Map_Fragment_Program(r700_AssemblerBase *pAsm, struct gl_fragment_program *mesa_fp, GLcontext *ctx); diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.h b/src/mesa/drivers/dri/r600/r700_shaderinst.h index 2829cca0a3c..cdb9a570f7c 100644 --- a/src/mesa/drivers/dri/r600/r700_shaderinst.h +++ b/src/mesa/drivers/dri/r600/r700_shaderinst.h @@ -42,6 +42,13 @@ #define SQ_FETCH_RESOURCE_VS_OFFSET 0x000000a0 #define SQ_FETCH_RESOURCE_VS_COUNT 0x000000b0 +//richard dec.10 glsl +#define SQ_TEX_SAMPLER_PS_OFFSET 0x00000000 +#define SQ_TEX_SAMPLER_PS_COUNT 0x00000012 +#define SQ_TEX_SAMPLER_VS_OFFSET 0x00000012 +#define SQ_TEX_SAMPLER_VS_COUNT 0x00000012 +//------------------- + #define SHADERINST_TYPEMASK_CF 0x10 #define SHADERINST_TYPEMASK_ALU 0x20 #define SHADERINST_TYPEMASK_TEX 0x40 diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index d7420678ff0..16b05d5cd9a 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -705,6 +705,10 @@ static void r700UpdateCulling(GLcontext * ctx) CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); /* default: ccw */ break; } + + /* Winding is inverted when rendering to FBO */ + if (ctx->DrawBuffer && ctx->DrawBuffer->Name) + r700->PA_SU_SC_MODE_CNTL.u32All ^= FACE_bit; } static void r700UpdateLineStipple(GLcontext * ctx) @@ -1227,13 +1231,8 @@ static void r700UpdatePolygonMode(GLcontext * ctx) /* Handle GL_CW (clock wise and GL_CCW (counter clock wise) * correctly by selecting the correct front and back face */ - if (ctx->Polygon.FrontFace == GL_CCW) { - f = ctx->Polygon.FrontMode; - b = ctx->Polygon.BackMode; - } else { - f = ctx->Polygon.BackMode; - b = ctx->Polygon.FrontMode; - } + f = ctx->Polygon.FrontMode; + b = ctx->Polygon.BackMode; /* Enable polygon mode */ SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DUAL_MODE, POLY_MODE_shift, POLY_MODE_mask); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 759b74dc7e9..90fac078ff0 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -341,7 +341,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, { vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i]; } + + vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions; + if(GL_FALSE == AssembleInstr(0, + 0, vp->mesa_program->Base.NumInstructions, &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) @@ -354,7 +358,7 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, return NULL; } - if( GL_FALSE == RelocProgram(&(vp->r700AsmCode)) ) + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) ) { return GL_FALSE; } @@ -671,5 +675,24 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) } else r700->vs.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->vs.num_consts; + for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++) + { + pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub; + + r700->vs.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/radeon_bo.c b/src/mesa/drivers/dri/r600/radeon_bo.c new file mode 120000 index 00000000000..9448ffee54b --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_bo.c @@ -0,0 +1 @@ +../radeon/radeon_bo.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h b/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h new file mode 120000 index 00000000000..029450928be --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_bo_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_bo_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_cs.c b/src/mesa/drivers/dri/r600/radeon_cs.c new file mode 120000 index 00000000000..66b7ad1eb03 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_cs.c @@ -0,0 +1 @@ +../radeon/radeon_cs.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h b/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h new file mode 120000 index 00000000000..462f5245d0e --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_cs_int_drm.h @@ -0,0 +1 @@ +../radeon/radeon_cs_int_drm.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile index ae2e695bfcc..2b2f2c4aa7a 100644 --- a/src/mesa/drivers/dri/radeon/Makefile +++ b/src/mesa/drivers/dri/radeon/Makefile @@ -11,7 +11,7 @@ LIBNAME = radeon_dri.so MINIGLX_SOURCES = server/radeon_dri.c ifeq ($(RADEON_LDFLAGS),) -CS_SOURCES = radeon_cs_space_drm.c +CS_SOURCES = radeon_cs_space_drm.c radeon_bo.c radeon_cs.c endif RADEON_COMMON_SOURCES = \ diff --git a/src/mesa/drivers/dri/radeon/radeon_bo.c b/src/mesa/drivers/dri/radeon/radeon_bo.c new file mode 100644 index 00000000000..393d156cdea --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_bo.c @@ -0,0 +1,110 @@ +#include <radeon_bocs_wrapper.h> +#include <radeon_bo_int_drm.h> + +void radeon_bo_debug(struct radeon_bo *bo, + const char *op) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + + fprintf(stderr, "%s %p 0x%08X 0x%08X 0x%08X\n", + op, bo, bo->handle, boi->size, boi->cref); +} + +struct radeon_bo *radeon_bo_open(struct radeon_bo_manager *bom, + uint32_t handle, + uint32_t size, + uint32_t alignment, + uint32_t domains, + uint32_t flags) +{ + struct radeon_bo *bo; + bo = bom->funcs->bo_open(bom, handle, size, alignment, domains, flags); + return bo; +} + +void radeon_bo_ref(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + boi->cref++; + boi->bom->funcs->bo_ref(boi); +} + +struct radeon_bo *radeon_bo_unref(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + boi->cref--; + return boi->bom->funcs->bo_unref(boi); +} + +int radeon_bo_map(struct radeon_bo *bo, int write) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->bom->funcs->bo_map(boi, write); +} + +int radeon_bo_unmap(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->bom->funcs->bo_unmap(boi); +} + +int radeon_bo_wait(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + if (!boi->bom->funcs->bo_wait) + return 0; + return boi->bom->funcs->bo_wait(boi); +} + +int radeon_bo_is_busy(struct radeon_bo *bo, + uint32_t *domain) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->bom->funcs->bo_is_busy(boi, domain); +} + +int radeon_bo_set_tiling(struct radeon_bo *bo, + uint32_t tiling_flags, uint32_t pitch) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->bom->funcs->bo_set_tiling(boi, tiling_flags, pitch); +} + +int radeon_bo_get_tiling(struct radeon_bo *bo, + uint32_t *tiling_flags, uint32_t *pitch) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->bom->funcs->bo_get_tiling(boi, tiling_flags, pitch); +} + +int radeon_bo_is_static(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + if (boi->bom->funcs->bo_is_static) + return boi->bom->funcs->bo_is_static(boi); + return 0; +} + +int radeon_bo_is_referenced_by_cs(struct radeon_bo *bo, + struct radeon_cs *cs) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + return boi->cref > 1; +} + +uint32_t radeon_bo_get_handle(struct radeon_bo *bo) +{ + return bo->handle; +} + +uint32_t radeon_bo_get_src_domain(struct radeon_bo *bo) +{ + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + uint32_t src_domain; + + src_domain = boi->space_accounted & 0xffff; + if (!src_domain) + src_domain = boi->space_accounted >> 16; + + return src_domain; +} diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h index 46e30b905a0..beb2369880a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h @@ -32,7 +32,6 @@ #include <stdio.h> #include <stdint.h> -//#include "radeon_track.h" /* bo object */ #define RADEON_BO_FLAGS_MACRO_TILE 1 @@ -42,191 +41,35 @@ struct radeon_bo_manager; struct radeon_cs; struct radeon_bo { - uint32_t alignment; + void *ptr; + uint32_t flags; uint32_t handle; uint32_t size; - uint32_t domains; - uint32_t flags; - unsigned cref; -#ifdef RADEON_BO_TRACK - struct radeon_track *track; -#endif - void *ptr; - struct radeon_bo_manager *bom; - uint32_t space_accounted; -}; - -/* bo functions */ -struct radeon_bo_funcs { - struct radeon_bo *(*bo_open)(struct radeon_bo_manager *bom, - uint32_t handle, - uint32_t size, - uint32_t alignment, - uint32_t domains, - uint32_t flags); - void (*bo_ref)(struct radeon_bo *bo); - struct radeon_bo *(*bo_unref)(struct radeon_bo *bo); - int (*bo_map)(struct radeon_bo *bo, int write); - int (*bo_unmap)(struct radeon_bo *bo); - int (*bo_wait)(struct radeon_bo *bo); - int (*bo_is_static)(struct radeon_bo *bo); - int (*bo_set_tiling)(struct radeon_bo *bo, uint32_t tiling_flags, - uint32_t pitch); - int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags, - uint32_t *pitch); - int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain); - int (*bo_is_referenced_by_cs)(struct radeon_bo *bo, struct radeon_cs *cs); }; -struct radeon_bo_manager { - struct radeon_bo_funcs *funcs; - int fd; - -#ifdef RADEON_BO_TRACK - struct radeon_tracker tracker; -#endif -}; - -static inline void _radeon_bo_debug(struct radeon_bo *bo, - const char *op, - const char *file, - const char *func, - int line) -{ - fprintf(stderr, "%s %p 0x%08X 0x%08X 0x%08X [%s %s %d]\n", - op, bo, bo->handle, bo->size, bo->cref, file, func, line); -} - -static inline struct radeon_bo *_radeon_bo_open(struct radeon_bo_manager *bom, - uint32_t handle, - uint32_t size, - uint32_t alignment, - uint32_t domains, - uint32_t flags, - const char *file, - const char *func, - int line) -{ - struct radeon_bo *bo; - - bo = bom->funcs->bo_open(bom, handle, size, alignment, domains, flags); - -#ifdef RADEON_BO_TRACK - if (bo) { - bo->track = radeon_tracker_add_track(&bom->tracker, bo->handle); - radeon_track_add_event(bo->track, file, func, "open", line); - } -#endif - return bo; -} - -static inline void _radeon_bo_ref(struct radeon_bo *bo, - const char *file, - const char *func, - int line) -{ - bo->cref++; -#ifdef RADEON_BO_TRACK - radeon_track_add_event(bo->track, file, func, "ref", line); -#endif - bo->bom->funcs->bo_ref(bo); -} - -static inline struct radeon_bo *_radeon_bo_unref(struct radeon_bo *bo, - const char *file, - const char *func, - int line) -{ - bo->cref--; -#ifdef RADEON_BO_TRACK - radeon_track_add_event(bo->track, file, func, "unref", line); - if (bo->cref <= 0) { - radeon_tracker_remove_track(&bo->bom->tracker, bo->track); - bo->track = NULL; - } -#endif - return bo->bom->funcs->bo_unref(bo); -} - -static inline int _radeon_bo_map(struct radeon_bo *bo, - int write, - const char *file, - const char *func, - int line) -{ - return bo->bom->funcs->bo_map(bo, write); -} - -static inline int _radeon_bo_unmap(struct radeon_bo *bo, - const char *file, - const char *func, - int line) -{ - return bo->bom->funcs->bo_unmap(bo); -} - -static inline int _radeon_bo_wait(struct radeon_bo *bo, - const char *file, - const char *func, - int line) -{ - return bo->bom->funcs->bo_wait(bo); -} - -static inline int _radeon_bo_is_busy(struct radeon_bo *bo, - uint32_t *domain, - const char *file, - const char *func, - int line) -{ - return bo->bom->funcs->bo_is_busy(bo, domain); -} - -static inline int radeon_bo_set_tiling(struct radeon_bo *bo, - uint32_t tiling_flags, uint32_t pitch) -{ - return bo->bom->funcs->bo_set_tiling(bo, tiling_flags, pitch); -} - -static inline int radeon_bo_get_tiling(struct radeon_bo *bo, - uint32_t *tiling_flags, uint32_t *pitch) -{ - return bo->bom->funcs->bo_get_tiling(bo, tiling_flags, pitch); -} - -static inline int radeon_bo_is_static(struct radeon_bo *bo) -{ - if (bo->bom->funcs->bo_is_static) - return bo->bom->funcs->bo_is_static(bo); - return 0; -} - -static inline int _radeon_bo_is_referenced_by_cs(struct radeon_bo *bo, - struct radeon_cs *cs, - const char *file, - const char *func, - unsigned line) -{ - return bo->cref > 1; -} - -#define radeon_bo_open(bom, h, s, a, d, f)\ - _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_ref(bo)\ - _radeon_bo_ref(bo, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_unref(bo)\ - _radeon_bo_unref(bo, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_map(bo, w)\ - _radeon_bo_map(bo, w, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_unmap(bo)\ - _radeon_bo_unmap(bo, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_debug(bo, opcode)\ - _radeon_bo_debug(bo, opcode, __FILE__, __FUNCTION__, __LINE__) -#define radeon_bo_wait(bo) \ - _radeon_bo_wait(bo, __FILE__, __func__, __LINE__) -#define radeon_bo_is_busy(bo, domain) \ - _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__) -#define radeon_bo_is_referenced_by_cs(bo, cs) \ - _radeon_bo_is_referenced_by_cs(bo, cs, __FILE__, __FUNCTION__, __LINE__) +struct radeon_bo_manager; +void radeon_bo_debug(struct radeon_bo *bo, + const char *op); + +struct radeon_bo *radeon_bo_open(struct radeon_bo_manager *bom, + uint32_t handle, + uint32_t size, + uint32_t alignment, + uint32_t domains, + uint32_t flags); + +void radeon_bo_ref(struct radeon_bo *bo); +struct radeon_bo *radeon_bo_unref(struct radeon_bo *bo); +int radeon_bo_map(struct radeon_bo *bo, int write); +int radeon_bo_unmap(struct radeon_bo *bo); +int radeon_bo_wait(struct radeon_bo *bo); +int radeon_bo_is_busy(struct radeon_bo *bo, uint32_t *domain); +int radeon_bo_set_tiling(struct radeon_bo *bo, uint32_t tiling_flags, uint32_t pitch); +int radeon_bo_get_tiling(struct radeon_bo *bo, uint32_t *tiling_flags, uint32_t *pitch); +int radeon_bo_is_static(struct radeon_bo *bo); +int radeon_bo_is_referenced_by_cs(struct radeon_bo *bo, + struct radeon_cs *cs); +uint32_t radeon_bo_get_handle(struct radeon_bo *bo); +uint32_t radeon_bo_get_src_domain(struct radeon_bo *bo); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_int_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_int_drm.h new file mode 100644 index 00000000000..190c332475b --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_bo_int_drm.h @@ -0,0 +1,45 @@ +#ifndef RADEON_BO_INT +#define RADEON_BO_INT + +struct radeon_bo_manager { + struct radeon_bo_funcs *funcs; + int fd; +}; + +struct radeon_bo_int { + void *ptr; + uint32_t flags; + uint32_t handle; + uint32_t size; + /* private members */ + uint32_t alignment; + uint32_t domains; + unsigned cref; + struct radeon_bo_manager *bom; + uint32_t space_accounted; + uint32_t referenced_in_cs; +}; + +/* bo functions */ +struct radeon_bo_funcs { + struct radeon_bo *(*bo_open)(struct radeon_bo_manager *bom, + uint32_t handle, + uint32_t size, + uint32_t alignment, + uint32_t domains, + uint32_t flags); + void (*bo_ref)(struct radeon_bo_int *bo); + struct radeon_bo *(*bo_unref)(struct radeon_bo_int *bo); + int (*bo_map)(struct radeon_bo_int *bo, int write); + int (*bo_unmap)(struct radeon_bo_int *bo); + int (*bo_wait)(struct radeon_bo_int *bo); + int (*bo_is_static)(struct radeon_bo_int *bo); + int (*bo_set_tiling)(struct radeon_bo_int *bo, uint32_t tiling_flags, + uint32_t pitch); + int (*bo_get_tiling)(struct radeon_bo_int *bo, uint32_t *tiling_flags, + uint32_t *pitch); + int (*bo_is_busy)(struct radeon_bo_int *bo, uint32_t *domain); + int (*bo_is_referenced_by_cs)(struct radeon_bo_int *bo, struct radeon_cs *cs); +}; + +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c index ce60a2f7eae..cf12664bacd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c @@ -50,6 +50,12 @@ #include "radeon_bocs_wrapper.h" #include "radeon_macros.h" +#ifdef HAVE_LIBDRM_RADEON +#include "radeon_bo_int.h" +#else +#include "radeon_bo_int_drm.h" +#endif + /* no seriously texmem.c is this screwed up */ struct bo_legacy_texture_object { driTextureObject base; @@ -57,7 +63,7 @@ struct bo_legacy_texture_object { }; struct bo_legacy { - struct radeon_bo base; + struct radeon_bo_int base; int map_count; uint32_t pending; int is_pending; @@ -187,10 +193,10 @@ static void legacy_get_current_age(struct bo_manager_legacy *boml) } } -static int legacy_is_pending(struct radeon_bo *bo) +static int legacy_is_pending(struct radeon_bo_int *boi) { - struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; - struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; + struct bo_manager_legacy *boml = (struct bo_manager_legacy *)boi->bom; + struct bo_legacy *bo_legacy = (struct bo_legacy*)boi; if (bo_legacy->is_pending <= 0) { bo_legacy->is_pending = 0; @@ -204,13 +210,13 @@ static int legacy_is_pending(struct radeon_bo *bo) if (bo_legacy->pnext) { bo_legacy->pnext->pprev = bo_legacy->pprev; } - assert(bo_legacy->is_pending <= bo->cref); + assert(bo_legacy->is_pending <= boi->cref); while (bo_legacy->is_pending--) { - bo = radeon_bo_unref(bo); - if (!bo) + boi = (struct radeon_bo_int *)radeon_bo_unref((struct radeon_bo *)boi); + if (!boi) break; } - if (bo) + if (boi) bo_legacy->is_pending = 0; boml->cpendings--; return 0; @@ -218,7 +224,7 @@ static int legacy_is_pending(struct radeon_bo *bo) return 1; } -static int legacy_wait_pending(struct radeon_bo *bo) +static int legacy_wait_pending(struct radeon_bo_int *bo) { struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; @@ -323,7 +329,7 @@ static struct bo_legacy *bo_allocate(struct bo_manager_legacy *boml, return bo_legacy; } -static int bo_dma_alloc(struct radeon_bo *bo) +static int bo_dma_alloc(struct radeon_bo_int *bo) { struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; @@ -333,7 +339,7 @@ static int bo_dma_alloc(struct radeon_bo *bo) int r; /* align size on 4Kb */ - size = (((4 * 1024) - 1) + bo->size) & ~((4 * 1024) - 1); + size = (((4 * 1024) - 1) + bo_legacy->base.size) & ~((4 * 1024) - 1); alloc.region = RADEON_MEM_REGION_GART; alloc.alignment = bo_legacy->base.alignment; alloc.size = size; @@ -355,7 +361,7 @@ static int bo_dma_alloc(struct radeon_bo *bo) return 0; } -static int bo_dma_free(struct radeon_bo *bo) +static int bo_dma_free(struct radeon_bo_int *bo) { struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; @@ -428,7 +434,7 @@ static struct radeon_bo *bo_open(struct radeon_bo_manager *bom, bo_legacy = boml->bos.next; while (bo_legacy) { if (bo_legacy->base.handle == handle) { - radeon_bo_ref(&(bo_legacy->base)); + radeon_bo_ref((struct radeon_bo *)&(bo_legacy->base)); return (struct radeon_bo*)bo_legacy; } bo_legacy = bo_legacy->next; @@ -468,20 +474,20 @@ retry: return NULL; } } - radeon_bo_ref(&(bo_legacy->base)); + radeon_bo_ref((struct radeon_bo *)&(bo_legacy->base)); return (struct radeon_bo*)bo_legacy; } -static void bo_ref(struct radeon_bo *bo) +static void bo_ref(struct radeon_bo_int *bo) { } -static struct radeon_bo *bo_unref(struct radeon_bo *bo) +static struct radeon_bo *bo_unref(struct radeon_bo_int *boi) { - struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; + struct bo_legacy *bo_legacy = (struct bo_legacy*)boi; - if (bo->cref <= 0) { + if (boi->cref <= 0) { bo_legacy->prev->next = bo_legacy->next; if (bo_legacy->next) { bo_legacy->next->prev = bo_legacy->prev; @@ -491,10 +497,10 @@ static struct radeon_bo *bo_unref(struct radeon_bo *bo) } return NULL; } - return bo; + return (struct radeon_bo *)boi; } -static int bo_map(struct radeon_bo *bo, int write) +static int bo_map(struct radeon_bo_int *bo, int write) { struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; @@ -528,7 +534,7 @@ static int bo_map(struct radeon_bo *bo, int write) return 0; } -static int bo_unmap(struct radeon_bo *bo) +static int bo_unmap(struct radeon_bo_int *bo) { struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; @@ -542,7 +548,7 @@ static int bo_unmap(struct radeon_bo *bo) return 0; } -static int bo_is_busy(struct radeon_bo *bo, uint32_t *domain) +static int bo_is_busy(struct radeon_bo_int *bo, uint32_t *domain) { *domain = 0; if (bo->domains & RADEON_GEM_DOMAIN_GTT) @@ -555,7 +561,7 @@ static int bo_is_busy(struct radeon_bo *bo, uint32_t *domain) return 0; } -static int bo_is_static(struct radeon_bo *bo) +static int bo_is_static(struct radeon_bo_int *bo) { struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; return bo_legacy->static_bo; @@ -574,7 +580,7 @@ static struct radeon_bo_funcs bo_legacy_funcs = { bo_is_busy }; -static int bo_vram_validate(struct radeon_bo *bo, +static int bo_vram_validate(struct radeon_bo_int *bo, uint32_t *soffset, uint32_t *eoffset) { @@ -700,29 +706,30 @@ int radeon_bo_legacy_validate(struct radeon_bo *bo, uint32_t *soffset, uint32_t *eoffset) { - struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + struct bo_manager_legacy *boml = (struct bo_manager_legacy *)boi->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; int r; int retries = 0; if (bo_legacy->map_count) { fprintf(stderr, "bo(%p, %d) is mapped (%d) can't valide it.\n", - bo, bo->size, bo_legacy->map_count); + bo, boi->size, bo_legacy->map_count); return -EINVAL; } - if(bo->size == 0) { + if(boi->size == 0) { fprintf(stderr, "bo(%p) has size 0.\n", bo); return -EINVAL; } if (bo_legacy->static_bo || bo_legacy->validated) { *soffset = bo_legacy->offset; - *eoffset = bo_legacy->offset + bo->size; + *eoffset = bo_legacy->offset + boi->size; return 0; } - if (!(bo->domains & RADEON_GEM_DOMAIN_GTT)) { + if (!(boi->domains & RADEON_GEM_DOMAIN_GTT)) { - r = bo_vram_validate(bo, soffset, eoffset); + r = bo_vram_validate(boi, soffset, eoffset); if (r) { legacy_track_pending(&boml->base, 0); legacy_kick_all_buffers(boml); @@ -736,7 +743,7 @@ int radeon_bo_legacy_validate(struct radeon_bo *bo, } } *soffset = bo_legacy->offset; - *eoffset = bo_legacy->offset + bo->size; + *eoffset = bo_legacy->offset + boi->size; bo_legacy->validated = 1; return 0; @@ -744,7 +751,8 @@ int radeon_bo_legacy_validate(struct radeon_bo *bo, void radeon_bo_legacy_pending(struct radeon_bo *bo, uint32_t pending) { - struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom; + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; + struct bo_manager_legacy *boml = (struct bo_manager_legacy *)boi->bom; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; bo_legacy->pending = pending; @@ -799,7 +807,7 @@ static struct bo_legacy *radeon_legacy_bo_alloc_static(struct bo_manager_legacy if (bo->base.handle > bom->nhandle) { bom->nhandle = bo->base.handle + 1; } - radeon_bo_ref(&(bo->base)); + radeon_bo_ref((struct radeon_bo *)&(bo->base)); return bo; } @@ -894,12 +902,13 @@ void radeon_bo_legacy_texture_age(struct radeon_bo_manager *bom) unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo) { + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; struct bo_legacy *bo_legacy = (struct bo_legacy*)bo; - if (bo_legacy->static_bo || (bo->domains & RADEON_GEM_DOMAIN_GTT)) { + if (bo_legacy->static_bo || (boi->domains & RADEON_GEM_DOMAIN_GTT)) { return 0; } - return bo->size; + return boi->size; } /* @@ -924,7 +933,7 @@ struct radeon_bo *radeon_legacy_bo_alloc_fake(struct radeon_bo_manager *bom, if (bo->base.handle > boml->nhandle) { boml->nhandle = bo->base.handle + 1; } - radeon_bo_ref(&(bo->base)); - return &(bo->base); + radeon_bo_ref((struct radeon_bo *)&(bo->base)); + return (struct radeon_bo *)&(bo->base); } diff --git a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h index 4520a7d7d49..6c2648b6bd8 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h +++ b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h @@ -18,8 +18,11 @@ #define RADEON_TILING_MACRO 0x1 #define RADEON_TILING_MICRO 0x2 #define RADEON_TILING_SWAP 0x4 + +#ifndef RADEON_TILING_SURFACE #define RADEON_TILING_SURFACE 0x8 /* this object requires a surface * when mapped - i.e. front buffer */ +#endif /* to be used to build locally in mesa with no libdrm bits */ #include "../radeon/radeon_bo_drm.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 51fa6189377..2a2b16a54bd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -641,6 +641,27 @@ void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, } } +/** + * Check if we're about to draw into the front color buffer. + * If so, set the intel->front_buffer_dirty field to true. + */ +void +radeon_check_front_buffer_rendering(GLcontext *ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (fb->Name == 0) { + /* drawing to window system buffer */ + if (fb->_NumColorDrawBuffers > 0) { + if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { + radeon->front_buffer_dirty = GL_TRUE; + } + } + } +} + + void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb) { radeonContextPtr radeon = RADEON_CONTEXT(ctx); @@ -1095,7 +1116,7 @@ void radeonFlush(GLcontext *ctx) then no point flushing anything at all. */ if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && is_empty_list(&radeon->dma.reserved)) - return; + goto flush_front; if (radeon->dma.flush) radeon->dma.flush( ctx ); @@ -1103,6 +1124,7 @@ void radeonFlush(GLcontext *ctx) if (radeon->cmdbuf.cs->cdw) rcommonFlushCmdBuf(radeon, __FUNCTION__); +flush_front: if ((ctx->DrawBuffer->Name == 0) && radeon->front_buffer_dirty) { __DRIscreen *const screen = radeon->radeonScreen->driScreen; diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h index 0608fe2418c..faad145cc49 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.h +++ b/src/mesa/drivers/dri/radeon/radeon_common.h @@ -43,6 +43,8 @@ radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb, struct radeon_bo *bo); struct radeon_renderbuffer * radeon_create_renderbuffer(gl_format format, __DRIdrawablePrivate *driDrawPriv); + +void radeon_check_front_buffer_rendering(GLcontext *ctx); static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb) { struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 49a9ec56106..0739496e032 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -406,9 +406,6 @@ struct radeon_state { struct radeon_depthbuffer_state depth; struct radeon_scissor_state scissor; struct radeon_stencilbuffer_state stencil; - - struct radeon_cs_space_check bos[RADEON_MAX_BOS]; - int validated_bo_count; }; /** diff --git a/src/mesa/drivers/dri/radeon/radeon_cs.c b/src/mesa/drivers/dri/radeon/radeon_cs.c new file mode 100644 index 00000000000..17e74333697 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_cs.c @@ -0,0 +1,95 @@ + +#include <stdio.h> +#include <stdint.h> +#include "drm.h" +#include "radeon_drm.h" +#include "radeon_bocs_wrapper.h" +#include "radeon_cs_int_drm.h" + +struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm, + uint32_t ndw) +{ + struct radeon_cs_int *csi = csm->funcs->cs_create(csm, ndw); + return (struct radeon_cs *)csi; +} + +int radeon_cs_write_reloc(struct radeon_cs *cs, + struct radeon_bo *bo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t flags) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + + return csi->csm->funcs->cs_write_reloc(csi, + bo, + read_domain, + write_domain, + flags); +} + +int radeon_cs_begin(struct radeon_cs *cs, + uint32_t ndw, + const char *file, + const char *func, + int line) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_begin(csi, ndw, file, func, line); +} + +int radeon_cs_end(struct radeon_cs *cs, + const char *file, + const char *func, + int line) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_end(csi, file, func, line); +} + +int radeon_cs_emit(struct radeon_cs *cs) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_emit(csi); +} + +int radeon_cs_destroy(struct radeon_cs *cs) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_destroy(csi); +} + +int radeon_cs_erase(struct radeon_cs *cs) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_erase(csi); +} + +int radeon_cs_need_flush(struct radeon_cs *cs) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return csi->csm->funcs->cs_need_flush(csi); +} + +void radeon_cs_print(struct radeon_cs *cs, FILE *file) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + csi->csm->funcs->cs_print(csi, file); +} + +void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + if (domain == RADEON_GEM_DOMAIN_VRAM) + csi->csm->vram_limit = limit; + else + csi->csm->gart_limit = limit; +} + +void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data) +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + csi->space_flush_fn = fn; + csi->space_flush_data = data; +} + diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h index ab4eca31a3c..a3f1750c6ed 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h @@ -36,6 +36,7 @@ #include <string.h> #include "drm.h" #include "radeon_drm.h" +#include "radeon_bo_drm.h" struct radeon_cs_reloc { struct radeon_bo *bo; @@ -49,173 +50,41 @@ struct radeon_cs_reloc { #define RADEON_CS_SPACE_OP_TO_BIG 1 #define RADEON_CS_SPACE_FLUSH 2 -struct radeon_cs_space_check { - struct radeon_bo *bo; - uint32_t read_domains; - uint32_t write_domain; - uint32_t new_accounted; -}; - -#define MAX_SPACE_BOS (32) - -struct radeon_cs_manager; - struct radeon_cs { - struct radeon_cs_manager *csm; - void *relocs; - uint32_t *packets; - unsigned crelocs; - unsigned relocs_total_size; - unsigned cdw; - unsigned ndw; - int section; + uint32_t *packets; + unsigned cdw; + unsigned ndw; unsigned section_ndw; unsigned section_cdw; - const char *section_file; - const char *section_func; - int section_line; - struct radeon_cs_space_check bos[MAX_SPACE_BOS]; - int bo_count; - void (*space_flush_fn)(void *); - void *space_flush_data; -}; - -/* cs functions */ -struct radeon_cs_funcs { - struct radeon_cs *(*cs_create)(struct radeon_cs_manager *csm, - uint32_t ndw); - int (*cs_write_reloc)(struct radeon_cs *cs, - struct radeon_bo *bo, - uint32_t read_domain, - uint32_t write_domain, - uint32_t flags); - int (*cs_begin)(struct radeon_cs *cs, - uint32_t ndw, - const char *file, - const char *func, - int line); - int (*cs_end)(struct radeon_cs *cs, - const char *file, - const char *func, - int line); - int (*cs_emit)(struct radeon_cs *cs); - int (*cs_destroy)(struct radeon_cs *cs); - int (*cs_erase)(struct radeon_cs *cs); - int (*cs_need_flush)(struct radeon_cs *cs); - void (*cs_print)(struct radeon_cs *cs, FILE *file); -}; - -struct radeon_cs_manager { - struct radeon_cs_funcs *funcs; - int fd; - int32_t vram_limit, gart_limit; - int32_t vram_write_used, gart_write_used; - int32_t read_used; }; -static inline struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm, - uint32_t ndw) -{ - return csm->funcs->cs_create(csm, ndw); -} - -static inline int radeon_cs_write_reloc(struct radeon_cs *cs, - struct radeon_bo *bo, - uint32_t read_domain, - uint32_t write_domain, - uint32_t flags) -{ - return cs->csm->funcs->cs_write_reloc(cs, - bo, - read_domain, - write_domain, - flags); -} - -static inline int radeon_cs_begin(struct radeon_cs *cs, - uint32_t ndw, - const char *file, - const char *func, - int line) -{ - return cs->csm->funcs->cs_begin(cs, ndw, file, func, line); -} - -static inline int radeon_cs_end(struct radeon_cs *cs, - const char *file, - const char *func, - int line) -{ - return cs->csm->funcs->cs_end(cs, file, func, line); -} - -static inline int radeon_cs_emit(struct radeon_cs *cs) -{ - return cs->csm->funcs->cs_emit(cs); -} - -static inline int radeon_cs_destroy(struct radeon_cs *cs) -{ - return cs->csm->funcs->cs_destroy(cs); -} - -static inline int radeon_cs_erase(struct radeon_cs *cs) -{ - return cs->csm->funcs->cs_erase(cs); -} - -static inline int radeon_cs_need_flush(struct radeon_cs *cs) -{ - return cs->csm->funcs->cs_need_flush(cs); -} - -static inline void radeon_cs_print(struct radeon_cs *cs, FILE *file) -{ - cs->csm->funcs->cs_print(cs, file); -} - -static inline void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit) -{ - - if (domain == RADEON_GEM_DOMAIN_VRAM) - cs->csm->vram_limit = limit; - else - cs->csm->gart_limit = limit; -} - -static inline void radeon_cs_write_dword(struct radeon_cs *cs, uint32_t dword) -{ - cs->packets[cs->cdw++] = dword; - if (cs->section) { - cs->section_cdw++; - } -} - -static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword) -{ - - memcpy(cs->packets + cs->cdw, &qword, sizeof(qword)); - cs->cdw+=2; - if (cs->section) { - cs->section_cdw+=2; - } -} - -static inline void radeon_cs_write_table(struct radeon_cs *cs, void *data, uint32_t size) -{ - memcpy(cs->packets + cs->cdw, data, size * 4); - cs->cdw += size; - if (cs->section) { - cs->section_cdw += size; - } -} +#define MAX_SPACE_BOS (32) -static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data) -{ - cs->space_flush_fn = fn; - cs->space_flush_data = data; -} +struct radeon_cs_manager; +extern struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm, + uint32_t ndw); + +extern int radeon_cs_begin(struct radeon_cs *cs, + uint32_t ndw, + const char *file, + const char *func, int line); +extern int radeon_cs_end(struct radeon_cs *cs, + const char *file, + const char *func, + int line); +extern int radeon_cs_emit(struct radeon_cs *cs); +extern int radeon_cs_destroy(struct radeon_cs *cs); +extern int radeon_cs_erase(struct radeon_cs *cs); +extern int radeon_cs_need_flush(struct radeon_cs *cs); +extern void radeon_cs_print(struct radeon_cs *cs, FILE *file); +extern void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit); +extern void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data); +extern int radeon_cs_write_reloc(struct radeon_cs *cs, + struct radeon_bo *bo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t flags); /* * add a persistent BO to the list @@ -243,4 +112,30 @@ int radeon_cs_space_check_with_bo(struct radeon_cs *cs, uint32_t read_domains, uint32_t write_domain); +static inline void radeon_cs_write_dword(struct radeon_cs *cs, uint32_t dword) +{ + cs->packets[cs->cdw++] = dword; + if (cs->section_ndw) { + cs->section_cdw++; + } +} + +static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword) +{ + memcpy(cs->packets + cs->cdw, &qword, sizeof(uint64_t)); + cs->cdw += 2; + if (cs->section_ndw) { + cs->section_cdw += 2; + } +} + +static inline void radeon_cs_write_table(struct radeon_cs *cs, + void *data, uint32_t size) +{ + memcpy(cs->packets + cs->cdw, data, size * 4); + cs->cdw += size; + if (cs->section_ndw) { + cs->section_cdw += size; + } +} #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_int_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_int_drm.h new file mode 100644 index 00000000000..8ba76bf9517 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_cs_int_drm.h @@ -0,0 +1,66 @@ + +#ifndef _RADEON_CS_INT_H_ +#define _RADEON_CS_INT_H_ + +struct radeon_cs_space_check { + struct radeon_bo_int *bo; + uint32_t read_domains; + uint32_t write_domain; + uint32_t new_accounted; +}; + +struct radeon_cs_int { + /* keep first two in same place */ + uint32_t *packets; + unsigned cdw; + unsigned ndw; + unsigned section_ndw; + unsigned section_cdw; + /* private members */ + struct radeon_cs_manager *csm; + void *relocs; + unsigned crelocs; + unsigned relocs_total_size; + const char *section_file; + const char *section_func; + int section_line; + struct radeon_cs_space_check bos[MAX_SPACE_BOS]; + int bo_count; + void (*space_flush_fn)(void *); + void *space_flush_data; +}; + +/* cs functions */ +struct radeon_cs_funcs { + struct radeon_cs_int *(*cs_create)(struct radeon_cs_manager *csm, + uint32_t ndw); + int (*cs_write_reloc)(struct radeon_cs_int *cs, + struct radeon_bo *bo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t flags); + int (*cs_begin)(struct radeon_cs_int *cs, + uint32_t ndw, + const char *file, + const char *func, + int line); + int (*cs_end)(struct radeon_cs_int *cs, + const char *file, const char *func, + int line); + + + int (*cs_emit)(struct radeon_cs_int *cs); + int (*cs_destroy)(struct radeon_cs_int *cs); + int (*cs_erase)(struct radeon_cs_int *cs); + int (*cs_need_flush)(struct radeon_cs_int *cs); + void (*cs_print)(struct radeon_cs_int *cs, FILE *file); +}; + +struct radeon_cs_manager { + struct radeon_cs_funcs *funcs; + int fd; + int32_t vram_limit, gart_limit; + int32_t vram_write_used, gart_write_used; + int32_t read_used; +}; +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index f1addb299e2..45b608a1b98 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -30,10 +30,18 @@ * Jérôme Glisse <[email protected]> */ #include <errno.h> +#include <unistd.h> +#include <stdint.h> +#include "drm.h" +#include "radeon_drm.h" #include "radeon_bocs_wrapper.h" #include "radeon_common.h" - +#ifdef HAVE_LIBDRM_RADEON +#include "radeon_cs_int.h" +#else +#include "radeon_cs_int_drm.h" +#endif struct cs_manager_legacy { struct radeon_cs_manager base; struct radeon_context *ctx; @@ -51,27 +59,27 @@ struct cs_reloc_legacy { }; -static struct radeon_cs *cs_create(struct radeon_cs_manager *csm, - uint32_t ndw) +static struct radeon_cs_int *cs_create(struct radeon_cs_manager *csm, + uint32_t ndw) { - struct radeon_cs *cs; + struct radeon_cs_int *csi; - cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs)); - if (cs == NULL) { + csi = (struct radeon_cs_int*)calloc(1, sizeof(struct radeon_cs_int)); + if (csi == NULL) { return NULL; } - cs->csm = csm; - cs->ndw = (ndw + 0x3FF) & (~0x3FF); - cs->packets = (uint32_t*)malloc(4*cs->ndw); - if (cs->packets == NULL) { - free(cs); + csi->csm = csm; + csi->ndw = (ndw + 0x3FF) & (~0x3FF); + csi->packets = (uint32_t*)malloc(4*csi->ndw); + if (csi->packets == NULL) { + free(csi); return NULL; } - cs->relocs_total_size = 0; - return cs; + csi->relocs_total_size = 0; + return csi; } -static int cs_write_reloc(struct radeon_cs *cs, +static int cs_write_reloc(struct radeon_cs_int *cs, struct radeon_bo *bo, uint32_t read_domain, uint32_t write_domain, @@ -150,20 +158,19 @@ static int cs_write_reloc(struct radeon_cs *cs, return 0; } -static int cs_begin(struct radeon_cs *cs, +static int cs_begin(struct radeon_cs_int *cs, uint32_t ndw, const char *file, const char *func, int line) { - if (cs->section) { + if (cs->section_ndw) { fprintf(stderr, "CS already in a section(%s,%s,%d)\n", cs->section_file, cs->section_func, cs->section_line); fprintf(stderr, "CS can't start section(%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 1; cs->section_ndw = ndw; cs->section_cdw = 0; cs->section_file = file; @@ -187,18 +194,17 @@ static int cs_begin(struct radeon_cs *cs, return 0; } -static int cs_end(struct radeon_cs *cs, +static int cs_end(struct radeon_cs_int *cs, const char *file, const char *func, int line) { - if (!cs->section) { + if (!cs->section_ndw) { fprintf(stderr, "CS no section to end at (%s,%s,%d)\n", file, func, line); return -EPIPE; } - cs->section = 0; if (cs->section_ndw != cs->section_cdw) { fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n", cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw); @@ -206,10 +212,12 @@ static int cs_end(struct radeon_cs *cs, file, func, line); return -EPIPE; } + cs->section_ndw = 0; + return 0; } -static int cs_process_relocs(struct radeon_cs *cs) +static int cs_process_relocs(struct radeon_cs_int *cs) { struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm; struct cs_reloc_legacy *relocs; @@ -254,7 +262,7 @@ restart: return 0; } -static int cs_set_age(struct radeon_cs *cs) +static int cs_set_age(struct radeon_cs_int *cs) { struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm; struct cs_reloc_legacy *relocs; @@ -268,7 +276,7 @@ static int cs_set_age(struct radeon_cs *cs) return 0; } -static int cs_emit(struct radeon_cs *cs) +static int cs_emit(struct radeon_cs_int *cs) { struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm; drm_radeon_cmd_buffer_t cmd; @@ -276,7 +284,7 @@ static int cs_emit(struct radeon_cs *cs) uint64_t ull; int r; - csm->ctx->vtbl.emit_cs_header(cs, csm->ctx); + csm->ctx->vtbl.emit_cs_header((struct radeon_cs *)cs, csm->ctx); /* append buffer age */ if ( IS_R300_CLASS(csm->ctx->radeonScreen) ) @@ -289,9 +297,9 @@ static int cs_emit(struct radeon_cs *cs) age.scratch.reg = 2; age.scratch.n_bufs = 1; age.scratch.flags = 0; - radeon_cs_write_dword(cs, age.u); - radeon_cs_write_qword(cs, ull); - radeon_cs_write_dword(cs, 0); + radeon_cs_write_dword((struct radeon_cs *)cs, age.u); + radeon_cs_write_qword((struct radeon_cs *)cs, ull); + radeon_cs_write_dword((struct radeon_cs *)cs, 0); } r = cs_process_relocs(cs); @@ -342,7 +350,7 @@ static void inline cs_free_reloc(void *relocs_p, int crelocs) free(relocs[i].indices); } -static int cs_destroy(struct radeon_cs *cs) +static int cs_destroy(struct radeon_cs_int *cs) { cs_free_reloc(cs->relocs, cs->crelocs); free(cs->relocs); @@ -351,7 +359,7 @@ static int cs_destroy(struct radeon_cs *cs) return 0; } -static int cs_erase(struct radeon_cs *cs) +static int cs_erase(struct radeon_cs_int *cs) { cs_free_reloc(cs->relocs, cs->crelocs); free(cs->relocs); @@ -359,18 +367,18 @@ static int cs_erase(struct radeon_cs *cs) cs->relocs = NULL; cs->crelocs = 0; cs->cdw = 0; - cs->section = 0; + cs->section_ndw = 0; return 0; } -static int cs_need_flush(struct radeon_cs *cs) +static int cs_need_flush(struct radeon_cs_int *cs) { /* this function used to flush when the BO usage got to * a certain size, now the higher levels handle this better */ return 0; } -static void cs_print(struct radeon_cs *cs, FILE *file) +static void cs_print(struct radeon_cs_int *cs, FILE *file) { } diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c index 89cbbb5a6b9..e22b437d561 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c @@ -29,6 +29,8 @@ #include <errno.h> #include <stdlib.h> #include "radeon_bocs_wrapper.h" +#include "radeon_bo_int_drm.h" +#include "radeon_cs_int_drm.h" struct rad_sizes { int32_t op_read; @@ -39,7 +41,7 @@ struct rad_sizes { static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct rad_sizes *sizes) { uint32_t read_domains, write_domain; - struct radeon_bo *bo; + struct radeon_bo_int *bo; bo = sc->bo; sc->new_accounted = 0; @@ -47,7 +49,7 @@ static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct ra write_domain = sc->write_domain; /* legacy needs a static check */ - if (radeon_bo_is_static(bo)) { + if (radeon_bo_is_static((struct radeon_bo *)sc->bo)) { bo->space_accounted = sc->new_accounted = (read_domains << 16) | write_domain; return 0; } @@ -100,11 +102,11 @@ static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct ra return 0; } -static int radeon_cs_do_space_check(struct radeon_cs *cs, struct radeon_cs_space_check *new_tmp) +static int radeon_cs_do_space_check(struct radeon_cs_int *cs, struct radeon_cs_space_check *new_tmp) { struct radeon_cs_manager *csm = cs->csm; int i; - struct radeon_bo *bo; + struct radeon_bo_int *bo; struct rad_sizes sizes; int ret; @@ -158,25 +160,28 @@ static int radeon_cs_do_space_check(struct radeon_cs *cs, struct radeon_cs_space void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain) { + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; int i; - for (i = 0; i < cs->bo_count; i++) { - if (cs->bos[i].bo == bo && - cs->bos[i].read_domains == read_domains && - cs->bos[i].write_domain == write_domain) + for (i = 0; i < csi->bo_count; i++) { + if (csi->bos[i].bo == boi && + csi->bos[i].read_domains == read_domains && + csi->bos[i].write_domain == write_domain) return; } radeon_bo_ref(bo); - i = cs->bo_count; - cs->bos[i].bo = bo; - cs->bos[i].read_domains = read_domains; - cs->bos[i].write_domain = write_domain; - cs->bos[i].new_accounted = 0; - cs->bo_count++; - - assert(cs->bo_count < MAX_SPACE_BOS); + i = csi->bo_count; + csi->bos[i].bo = boi; + csi->bos[i].read_domains = read_domains; + csi->bos[i].write_domain = write_domain; + csi->bos[i].new_accounted = 0; + csi->bo_count++; + + assert(csi->bo_count < MAX_SPACE_BOS); } -static int radeon_cs_check_space_internal(struct radeon_cs *cs, struct radeon_cs_space_check *tmp_bo) +static int radeon_cs_check_space_internal(struct radeon_cs_int *cs, + struct radeon_cs_space_check *tmp_bo) { int ret; int flushed = 0; @@ -198,37 +203,42 @@ again: int radeon_cs_space_check_with_bo(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain) -{ +{ + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + struct radeon_bo_int *boi = (struct radeon_bo_int *)bo; struct radeon_cs_space_check temp_bo; + int ret = 0; if (bo) { - temp_bo.bo = bo; + temp_bo.bo = boi; temp_bo.read_domains = read_domains; temp_bo.write_domain = write_domain; temp_bo.new_accounted = 0; } - ret = radeon_cs_check_space_internal(cs, bo ? &temp_bo : NULL); + ret = radeon_cs_check_space_internal(csi, bo ? &temp_bo : NULL); return ret; } int radeon_cs_space_check(struct radeon_cs *cs) { - return radeon_cs_check_space_internal(cs, NULL); + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; + return radeon_cs_check_space_internal(csi, NULL); } void radeon_cs_space_reset_bos(struct radeon_cs *cs) { + struct radeon_cs_int *csi = (struct radeon_cs_int *)cs; int i; - for (i = 0; i < cs->bo_count; i++) { - radeon_bo_unref(cs->bos[i].bo); - cs->bos[i].bo = NULL; - cs->bos[i].read_domains = 0; - cs->bos[i].write_domain = 0; - cs->bos[i].new_accounted = 0; + for (i = 0; i < csi->bo_count; i++) { + radeon_bo_unref((struct radeon_bo *)csi->bos[i].bo); + csi->bos[i].bo = NULL; + csi->bos[i].read_domains = 0; + csi->bos[i].write_domain = 0; + csi->bos[i].new_accounted = 0; } - cs->bo_count = 0; + csi->bo_count = 0; } diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index b8c65f4ce62..d31e4e47ddb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -205,7 +205,6 @@ again_alloc: counter on unused buffers for later freeing them from begin of list */ dma_bo = last_elem(&rmesa->dma.free); - assert(dma_bo->bo->cref == 1); remove_from_list(dma_bo); insert_at_head(&rmesa->dma.reserved, dma_bo); } diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index fc21069a92c..a536436d55f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -166,8 +166,9 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, uint32_t size; uint32_t pitch = ((cpp * width + 63) & ~63) / cpp; - fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width, - height, pitch); + if (RADEON_DEBUG & RADEON_MEMORY) + fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width, + height, pitch); size = pitch * height * cpp; rrb->pitch = pitch * cpp; diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index a0106d00fa2..13fd6f99719 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -575,6 +575,10 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) GLuint color_mask = 0; GLuint orig_mask = mask; + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + rmesa->radeon.front_buffer_dirty = GL_TRUE; + } + if ( RADEON_DEBUG & RADEON_IOCTL ) { fprintf( stderr, "radeonClear\n"); } diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index a7f347202a1..033f26db2a1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -452,7 +452,10 @@ static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_bo_unmap(image->mt->bo); radeon_miptree_unreference(&image->mt); - } else { + } else if (image->base.Data) { + /* This condition should be removed, it's here to workaround + * a segfault when mapping textures during software fallbacks. + */ const uint32_t srcrowstride = _mesa_format_row_stride(image->base.TexFormat, image->base.Width); uint32_t rows = image->base.Height * image->base.Depth; diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 37904dc8dc9..8db3d2b143b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -811,8 +811,7 @@ static void map_unmap_rb(struct gl_renderbuffer *rb, int flag) return; if (flag) { - if (rrb->bo->bom->funcs->bo_wait) - radeon_bo_wait(rrb->bo); + radeon_bo_wait(rrb->bo); r = radeon_bo_map(rrb->bo, 1); if (r) { fprintf(stderr, "(%s) error(%d) mapping buffer.\n", @@ -828,18 +827,21 @@ static void map_unmap_rb(struct gl_renderbuffer *rb, int flag) } static void -radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map) +radeon_map_unmap_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb, + GLboolean map) { GLuint i, j; /* color draw buffers */ for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++) - map_unmap_rb(ctx->DrawBuffer->_ColorDrawBuffers[j], map); + map_unmap_rb(fb->_ColorDrawBuffers[j], map); + + map_unmap_rb(fb->_ColorReadBuffer, map); /* check for render to textures */ for (i = 0; i < BUFFER_COUNT; i++) { struct gl_renderbuffer_attachment *att = - ctx->DrawBuffer->Attachment + i; + fb->Attachment + i; struct gl_texture_object *tex = att->Texture; if (tex) { /* Render to texture. Note that a mipmapped texture need not @@ -855,15 +857,15 @@ radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map) radeon_teximage_unmap(image); } } - - map_unmap_rb(ctx->ReadBuffer->_ColorReadBuffer, map); - + /* depth buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_DepthBuffer) - map_unmap_rb(ctx->DrawBuffer->_DepthBuffer->Wrapped, map); + if (fb->_DepthBuffer) + map_unmap_rb(fb->_DepthBuffer->Wrapped, map); + + if (fb->_StencilBuffer) + map_unmap_rb(fb->_StencilBuffer->Wrapped, map); - if (ctx->DrawBuffer->_StencilBuffer) - map_unmap_rb(ctx->DrawBuffer->_StencilBuffer->Wrapped, map); + radeon_check_front_buffer_rendering(ctx); } static void radeonSpanRenderStart(GLcontext * ctx) @@ -888,23 +890,30 @@ static void radeonSpanRenderStart(GLcontext * ctx) ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current); } - radeon_map_unmap_buffers(ctx, 1); + radeon_map_unmap_framebuffer(ctx, ctx->DrawBuffer, GL_TRUE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + radeon_map_unmap_framebuffer(ctx, ctx->ReadBuffer, GL_TRUE); } static void radeonSpanRenderFinish(GLcontext * ctx) { radeonContextPtr rmesa = RADEON_CONTEXT(ctx); int i; + _swrast_flush(ctx); - if (!rmesa->radeonScreen->driScreen->dri2.enabled) { - UNLOCK_HARDWARE(rmesa); - } + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current); } - radeon_map_unmap_buffers(ctx, 0); + radeon_map_unmap_framebuffer(ctx, ctx->DrawBuffer, GL_FALSE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + radeon_map_unmap_framebuffer(ctx, ctx->ReadBuffer, GL_FALSE); + + if (!rmesa->radeonScreen->driScreen->dri2.enabled) { + UNLOCK_HARDWARE(rmesa); + } } void radeonInitSpanFuncs(GLcontext * ctx) diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c index 28690325d12..03178116c1a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.c +++ b/src/mesa/drivers/dri/radeon/radeon_texture.c @@ -472,6 +472,19 @@ gl_format radeonChooseTextureFormat(GLcontext * ctx, case GL_RGBA32F_ARB: return MESA_FORMAT_RGBA_FLOAT32; +#ifdef RADEON_R300 + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + return MESA_FORMAT_Z16; + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + if (rmesa->radeonScreen->chip_family >= CHIP_FAMILY_RV515) + return MESA_FORMAT_S8_Z24; + else + return MESA_FORMAT_Z16; +#else case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT16: case GL_DEPTH_COMPONENT24: @@ -479,6 +492,7 @@ gl_format radeonChooseTextureFormat(GLcontext * ctx, case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: return MESA_FORMAT_S8_Z24; +#endif /* EXT_texture_sRGB */ case GL_SRGB: diff --git a/src/mesa/drivers/dri/savage/savageioctl.c b/src/mesa/drivers/dri/savage/savageioctl.c index 948ed18419e..77ab8d16e0a 100644 --- a/src/mesa/drivers/dri/savage/savageioctl.c +++ b/src/mesa/drivers/dri/savage/savageioctl.c @@ -337,6 +337,8 @@ static void savageDDClear( GLcontext *ctx, GLbitfield mask ) GLint ch = ctx->DrawBuffer->_Ymax - cy; /* XXX FIX ME: the cx,cy,cw,ch vars are currently ignored! */ + (void) ch; + (void) cw; if (SAVAGE_DEBUG & DEBUG_VERBOSE_MSG) fprintf (stderr, "%s\n", __FUNCTION__); diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index e9529d19391..9a92541ef7e 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -435,7 +435,8 @@ do { \ #define LOCAL_VARS(n) \ savageContextPtr imesa = SAVAGE_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint coloroffset = \ ((imesa->skip & SAVAGE_SKIP_W) ? 3 : 4); \ GLboolean specoffset = \ diff --git a/src/mesa/drivers/dri/sis/sis_clear.c b/src/mesa/drivers/dri/sis/sis_clear.c index 323383da62a..d358ef62dc7 100644 --- a/src/mesa/drivers/dri/sis/sis_clear.c +++ b/src/mesa/drivers/dri/sis/sis_clear.c @@ -393,7 +393,6 @@ sis_clear_z_stencil_buffer( GLcontext * ctx, GLbitfield mask, GLint x, GLint y, GLint width, GLint height ) { sisContextPtr smesa = SIS_CONTEXT(ctx); - int cmd; mWait3DCmdQueue (8); MMIO(REG_SRC_PITCH, (smesa->zFormat == SiS_ZFORMAT_Z16) ? diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c index 3cf10007b58..4690274c3c0 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.c +++ b/src/mesa/drivers/dri/sis/sis_tris.c @@ -430,7 +430,8 @@ do { \ #define LOCAL_VARS(n) \ sisContextPtr smesa = SIS_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint coloroffset = smesa->coloroffset; \ GLuint specoffset = smesa->specoffset; \ (void) color; (void) spec; (void) coloroffset; (void) specoffset; diff --git a/src/mesa/drivers/dri/unichrome/via_ioctl.c b/src/mesa/drivers/dri/unichrome/via_ioctl.c index b34c133600a..91c94fa377d 100644 --- a/src/mesa/drivers/dri/unichrome/via_ioctl.c +++ b/src/mesa/drivers/dri/unichrome/via_ioctl.c @@ -885,7 +885,6 @@ void viaFlushDmaLocked(struct via_context *vmesa, GLuint flags) } else if (vmesa->numClipRects) { drm_clip_rect_t *pbox = vmesa->pClipRects; - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; for (i = 0; i < vmesa->numClipRects; i++) { drm_clip_rect_t b; diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index ab457d41dc8..01359d51ea6 100644 --- a/src/mesa/drivers/dri/unichrome/via_tris.c +++ b/src/mesa/drivers/dri/unichrome/via_tris.c @@ -330,7 +330,8 @@ do { \ #define LOCAL_VARS(n) \ struct via_context *vmesa = VIA_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = { 0 }; \ + GLuint spec[n] = { 0 }; \ GLuint coloroffset = vmesa->coloroffset; \ GLuint specoffset = vmesa->specoffset; \ (void)color; (void)spec; (void)coloroffset; (void)specoffset; diff --git a/src/mesa/drivers/osmesa/Makefile b/src/mesa/drivers/osmesa/Makefile index 92d41494665..9010bbd130a 100644 --- a/src/mesa/drivers/osmesa/Makefile +++ b/src/mesa/drivers/osmesa/Makefile @@ -21,7 +21,11 @@ INCLUDE_DIRS = \ # Standalone osmesa needs to be linked with core Mesa APIs ifeq ($(DRIVER_DIRS), osmesa) -CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a +CORE_MESA = \ + $(TOP)/src/mesa/libmesa.a \ + $(TOP)/src/mesa/libglapi.a \ + $(TOP)/src/glsl/cl/libglslcl.a \ + $(TOP)/src/glsl/pp/libglslpp.a else CORE_MESA = endif diff --git a/src/mesa/main/compiler.h b/src/mesa/main/compiler.h index 87c3b2e5621..4eb249b4af1 100644 --- a/src/mesa/main/compiler.h +++ b/src/mesa/main/compiler.h @@ -234,7 +234,7 @@ extern "C" { #elif defined(__APPLE__) #include <CoreFoundation/CFByteOrder.h> #define CPU_TO_LE32( x ) CFSwapInt32HostToLittle( x ) -#elif defined(_AIX) +#elif (defined(_AIX) || defined(__blrts)) #define CPU_TO_LE32( x ) x = ((x & 0x000000ff) << 24) | \ ((x & 0x0000ff00) << 8) | \ ((x & 0x00ff0000) >> 8) | \ diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index c5048970cca..2eac1cc2ed9 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -243,7 +243,8 @@ /*@{*/ #define MAX_VERTEX_GENERIC_ATTRIBS 16 #define MAX_VERTEX_TEXTURE_IMAGE_UNITS MAX_TEXTURE_IMAGE_UNITS -#define MAX_COMBINED_TEXTURE_IMAGE_UNITS MAX_TEXTURE_IMAGE_UNITS +#define MAX_COMBINED_TEXTURE_IMAGE_UNITS (MAX_VERTEX_TEXTURE_IMAGE_UNITS + \ + MAX_TEXTURE_IMAGE_UNITS) /*@}*/ diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 03fc57e6654..5c20ce017f7 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -564,10 +564,6 @@ _mesa_init_constants(GLcontext *ctx) /* GL_ARB_draw_buffers */ ctx->Const.MaxDrawBuffers = MAX_DRAW_BUFFERS; - /* GL_OES_read_format */ - ctx->Const.ColorReadFormat = GL_RGBA; - ctx->Const.ColorReadType = GL_UNSIGNED_BYTE; - #if FEATURE_EXT_framebuffer_object ctx->Const.MaxColorAttachments = MAX_COLOR_ATTACHMENTS; ctx->Const.MaxRenderbufferSize = MAX_WIDTH; diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index 27ed921761c..6dadf5c079b 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -182,7 +182,7 @@ struct dd_function_table { * * This is called by the \c _mesa_store_tex[sub]image[123]d() fallback * functions. The driver should examine \p internalFormat and return a - * pointer to an appropriate gl_texture_format. + * gl_format value. */ GLuint (*ChooseTextureFormat)( GLcontext *ctx, GLint internalFormat, GLenum srcFormat, GLenum srcType ); @@ -538,11 +538,6 @@ struct dd_function_table { struct gl_texture_object *t ); /** - * Called by glActiveTextureARB() to set current texture unit. - */ - void (*ActiveTexture)( GLcontext *ctx, GLuint texUnitNumber ); - - /** * Called when the texture's color lookup table is changed. * * If \p tObj is NULL then the shared texture palette @@ -766,13 +761,13 @@ struct dd_function_table { /* May return NULL if MESA_MAP_NOWAIT_BIT is set in access: */ - void * (*MapBufferRange)( GLcontext *ctx, GLenum target, - GLintptr offset, GLsizeiptr length, GLbitfield access, + void * (*MapBufferRange)( GLcontext *ctx, GLenum target, GLintptr offset, + GLsizeiptr length, GLbitfield access, struct gl_buffer_object *obj); - void (*FlushMappedBufferRange) (GLcontext *ctx, GLenum target, - GLintptr offset, GLsizeiptr length, - struct gl_buffer_object *obj); + void (*FlushMappedBufferRange)(GLcontext *ctx, GLenum target, + GLintptr offset, GLsizeiptr length, + struct gl_buffer_object *obj); GLboolean (*UnmapBuffer)( GLcontext *ctx, GLenum target, struct gl_buffer_object *obj ); @@ -787,7 +782,8 @@ struct dd_function_table { struct gl_framebuffer * (*NewFramebuffer)(GLcontext *ctx, GLuint name); struct gl_renderbuffer * (*NewRenderbuffer)(GLcontext *ctx, GLuint name); void (*BindFramebuffer)(GLcontext *ctx, GLenum target, - struct gl_framebuffer *fb, struct gl_framebuffer *fbread); + struct gl_framebuffer *drawFb, + struct gl_framebuffer *readFb); void (*FramebufferRenderbuffer)(GLcontext *ctx, struct gl_framebuffer *fb, GLenum attachment, diff --git a/src/mesa/main/dlopen.c b/src/mesa/main/dlopen.c index 414cfad8e26..81e032081db 100644 --- a/src/mesa/main/dlopen.c +++ b/src/mesa/main/dlopen.c @@ -31,7 +31,7 @@ #include "compiler.h" #include "dlopen.h" -#if defined(_GNU_SOURCE) && !defined(__MINGW32__) +#if defined(_GNU_SOURCE) && !defined(__MINGW32__) && !defined(__blrts) #include <dlfcn.h> #endif #if defined(_WIN32) @@ -46,7 +46,9 @@ void * _mesa_dlopen(const char *libname, int flags) { -#if defined(_GNU_SOURCE) +#if defined(__blrts) + return NULL; +#elif defined(_GNU_SOURCE) flags = RTLD_LAZY | RTLD_GLOBAL; /* Overriding flags at this time */ return dlopen(libname, flags); #elif defined(__MINGW32__) @@ -65,7 +67,9 @@ _mesa_dlopen(const char *libname, int flags) GenericFunc _mesa_dlsym(void *handle, const char *fname) { -#if defined(__DJGPP__) +#if defined(__blrts) + return (GenericFunc) NULL; +#elif defined(__DJGPP__) /* need '_' prefix on symbol names */ char fname2[1000]; fname2[0] = '_'; @@ -88,7 +92,9 @@ _mesa_dlsym(void *handle, const char *fname) void _mesa_dlclose(void *handle) { -#if defined(_GNU_SOURCE) +#if defined(__blrts) + (void) handle; +#elif defined(_GNU_SOURCE) dlclose(handle); #elif defined(__MINGW32__) FreeLibrary(handle); diff --git a/src/mesa/main/fog.c b/src/mesa/main/fog.c index 4323d3db820..269ff3f8b99 100644 --- a/src/mesa/main/fog.c +++ b/src/mesa/main/fog.c @@ -34,15 +34,20 @@ void GLAPIENTRY _mesa_Fogf(GLenum pname, GLfloat param) { - _mesa_Fogfv(pname, ¶m); + GLfloat fparam[4]; + fparam[0] = param; + fparam[1] = fparam[2] = fparam[3] = 0.0F; + _mesa_Fogfv(pname, fparam); } void GLAPIENTRY _mesa_Fogi(GLenum pname, GLint param ) { - GLfloat fparam = (GLfloat) param; - _mesa_Fogfv(pname, &fparam); + GLfloat fparam[4]; + fparam[0] = (GLfloat) param; + fparam[1] = fparam[2] = fparam[3] = 0.0F; + _mesa_Fogfv(pname, fparam); } diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index 154dedacd50..d958dbf7d48 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -969,3 +969,29 @@ _mesa_dest_buffer_exists(GLcontext *ctx, GLenum format) /* OK */ return GL_TRUE; } + +GLenum +_mesa_get_color_read_format(GLcontext *ctx) +{ + switch (ctx->ReadBuffer->_ColorReadBuffer->Format) { + case MESA_FORMAT_ARGB8888: + return GL_BGRA; + case MESA_FORMAT_RGB565: + return GL_BGR; + default: + return GL_RGBA; + } +} + +GLenum +_mesa_get_color_read_type(GLcontext *ctx) +{ + switch (ctx->ReadBuffer->_ColorReadBuffer->Format) { + case MESA_FORMAT_ARGB8888: + return GL_UNSIGNED_BYTE; + case MESA_FORMAT_RGB565: + return GL_UNSIGNED_SHORT_5_6_5_REV; + default: + return GL_UNSIGNED_BYTE; + } +} diff --git a/src/mesa/main/framebuffer.h b/src/mesa/main/framebuffer.h index 45a4703ba99..ef21dd98e83 100644 --- a/src/mesa/main/framebuffer.h +++ b/src/mesa/main/framebuffer.h @@ -81,5 +81,10 @@ _mesa_source_buffer_exists(GLcontext *ctx, GLenum format); extern GLboolean _mesa_dest_buffer_exists(GLcontext *ctx, GLenum format); +extern GLenum +_mesa_get_color_read_type(GLcontext *ctx); + +extern GLenum +_mesa_get_color_read_format(GLcontext *ctx); #endif /* FRAMEBUFFER_H */ diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c index e8932f83b6a..3d32649bade 100644 --- a/src/mesa/main/get.c +++ b/src/mesa/main/get.c @@ -13,6 +13,7 @@ #include "mtypes.h" #include "state.h" #include "texcompress.h" +#include "framebuffer.h" #define FLOAT_TO_BOOLEAN(X) ( (X) ? GL_TRUE : GL_FALSE ) @@ -1767,11 +1768,11 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) break; case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: CHECK_EXT1(OES_read_format, "GetBooleanv"); - params[0] = INT_TO_BOOLEAN(ctx->Const.ColorReadType); + params[0] = INT_TO_BOOLEAN(_mesa_get_color_read_type(ctx)); break; case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: CHECK_EXT1(OES_read_format, "GetBooleanv"); - params[0] = INT_TO_BOOLEAN(ctx->Const.ColorReadFormat); + params[0] = INT_TO_BOOLEAN(_mesa_get_color_read_format(ctx)); break; case GL_NUM_FRAGMENT_REGISTERS_ATI: CHECK_EXT1(ATI_fragment_shader, "GetBooleanv"); @@ -3602,11 +3603,11 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) break; case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: CHECK_EXT1(OES_read_format, "GetFloatv"); - params[0] = (GLfloat)(ctx->Const.ColorReadType); + params[0] = (GLfloat)(_mesa_get_color_read_type(ctx)); break; case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: CHECK_EXT1(OES_read_format, "GetFloatv"); - params[0] = (GLfloat)(ctx->Const.ColorReadFormat); + params[0] = (GLfloat)(_mesa_get_color_read_format(ctx)); break; case GL_NUM_FRAGMENT_REGISTERS_ATI: CHECK_EXT1(ATI_fragment_shader, "GetFloatv"); @@ -5437,11 +5438,11 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) break; case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: CHECK_EXT1(OES_read_format, "GetIntegerv"); - params[0] = ctx->Const.ColorReadType; + params[0] = _mesa_get_color_read_type(ctx); break; case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: CHECK_EXT1(OES_read_format, "GetIntegerv"); - params[0] = ctx->Const.ColorReadFormat; + params[0] = _mesa_get_color_read_format(ctx); break; case GL_NUM_FRAGMENT_REGISTERS_ATI: CHECK_EXT1(ATI_fragment_shader, "GetIntegerv"); @@ -7273,11 +7274,11 @@ _mesa_GetInteger64v( GLenum pname, GLint64 *params ) break; case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: CHECK_EXT1(OES_read_format, "GetInteger64v"); - params[0] = (GLint64)(ctx->Const.ColorReadType); + params[0] = (GLint64)(_mesa_get_color_read_type(ctx)); break; case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: CHECK_EXT1(OES_read_format, "GetInteger64v"); - params[0] = (GLint64)(ctx->Const.ColorReadFormat); + params[0] = (GLint64)(_mesa_get_color_read_format(ctx)); break; case GL_NUM_FRAGMENT_REGISTERS_ATI: CHECK_EXT1(ATI_fragment_shader, "GetInteger64v"); diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py index a29962d3348..01170a42a72 100644 --- a/src/mesa/main/get_gen.py +++ b/src/mesa/main/get_gen.py @@ -942,9 +942,9 @@ StateVars = [ # GL_OES_read_format ( "GL_IMPLEMENTATION_COLOR_READ_TYPE_OES", GLint, - ["ctx->Const.ColorReadType"], "", ["OES_read_format"] ), + ["_mesa_get_color_read_type(ctx)"], "", ["OES_read_format"] ), ( "GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES", GLint, - ["ctx->Const.ColorReadFormat"], "", ["OES_read_format"] ), + ["_mesa_get_color_read_format(ctx)"], "", ["OES_read_format"] ), # GL_ATI_fragment_shader ( "GL_NUM_FRAGMENT_REGISTERS_ATI", GLint, ["6"], "", ["ATI_fragment_shader"] ), @@ -1159,6 +1159,7 @@ def EmitHeader(): #include "mtypes.h" #include "state.h" #include "texcompress.h" +#include "framebuffer.h" #define FLOAT_TO_BOOLEAN(X) ( (X) ? GL_TRUE : GL_FALSE ) diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c index 1c8a081e9a6..654faa5c770 100644 --- a/src/mesa/main/light.c +++ b/src/mesa/main/light.c @@ -206,7 +206,10 @@ _mesa_light(GLcontext *ctx, GLuint lnum, GLenum pname, const GLfloat *params) void GLAPIENTRY _mesa_Lightf( GLenum light, GLenum pname, GLfloat param ) { - _mesa_Lightfv( light, pname, ¶m ); + GLfloat fparam[4]; + fparam[0] = param; + fparam[1] = fparam[2] = fparam[3] = 0.0F; + _mesa_Lightfv( light, pname, fparam ); } @@ -285,7 +288,10 @@ _mesa_Lightfv( GLenum light, GLenum pname, const GLfloat *params ) void GLAPIENTRY _mesa_Lighti( GLenum light, GLenum pname, GLint param ) { - _mesa_Lightiv( light, pname, ¶m ); + GLint iparam[4]; + iparam[0] = param; + iparam[1] = iparam[2] = iparam[3] = 0; + _mesa_Lightiv( light, pname, iparam ); } @@ -537,14 +543,20 @@ _mesa_LightModeliv( GLenum pname, const GLint *params ) void GLAPIENTRY _mesa_LightModeli( GLenum pname, GLint param ) { - _mesa_LightModeliv( pname, ¶m ); + GLint iparam[4]; + iparam[0] = param; + iparam[1] = iparam[2] = iparam[3] = 0; + _mesa_LightModeliv( pname, iparam ); } void GLAPIENTRY _mesa_LightModelf( GLenum pname, GLfloat param ) { - _mesa_LightModelfv( pname, ¶m ); + GLfloat fparam[4]; + fparam[0] = param; + fparam[1] = fparam[2] = fparam[3] = 0.0F; + _mesa_LightModelfv( pname, fparam ); } diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 5f012448275..b52c84b491a 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2356,9 +2356,6 @@ struct gl_constants GLuint MaxDrawBuffers; /**< GL_ARB_draw_buffers */ - GLenum ColorReadFormat; /**< GL_OES_read_format */ - GLenum ColorReadType; /**< GL_OES_read_format */ - GLuint MaxColorAttachments; /**< GL_EXT_framebuffer_object */ GLuint MaxRenderbufferSize; /**< GL_EXT_framebuffer_object */ GLuint MaxSamples; /**< GL_ARB_framebuffer_object */ diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c index b3305448904..dcaeccd90d4 100644 --- a/src/mesa/main/points.c +++ b/src/mesa/main/points.c @@ -69,8 +69,10 @@ _mesa_PointSize( GLfloat size ) void GLAPIENTRY _mesa_PointParameteri( GLenum pname, GLint param ) { - const GLfloat value = (GLfloat) param; - _mesa_PointParameterfv(pname, &value); + GLfloat p[3]; + p[0] = (GLfloat) param; + p[1] = p[2] = 0.0F; + _mesa_PointParameterfv(pname, p); } @@ -90,7 +92,10 @@ _mesa_PointParameteriv( GLenum pname, const GLint *params ) void GLAPIENTRY _mesa_PointParameterf( GLenum pname, GLfloat param) { - _mesa_PointParameterfv(pname, ¶m); + GLfloat p[3]; + p[0] = param; + p[1] = p[2] = 0.0F; + _mesa_PointParameterfv(pname, p); } diff --git a/src/mesa/main/texgen.c b/src/mesa/main/texgen.c index 5abb1ff0ab6..be4e03bc56c 100644 --- a/src/mesa/main/texgen.c +++ b/src/mesa/main/texgen.c @@ -186,8 +186,10 @@ _mesa_TexGeniv(GLenum coord, GLenum pname, const GLint *params ) static void GLAPIENTRY _mesa_TexGend(GLenum coord, GLenum pname, GLdouble param ) { - GLfloat p = (GLfloat) param; - _mesa_TexGenfv( coord, pname, &p ); + GLfloat p[4]; + p[0] = (GLfloat) param; + p[1] = p[2] = p[3] = 0.0F; + _mesa_TexGenfv( coord, pname, p ); } diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c index 7e8a2489ac3..09fe7b85ba0 100644 --- a/src/mesa/main/texobj.c +++ b/src/mesa/main/texobj.c @@ -415,7 +415,7 @@ _mesa_test_texobj_completeness( const GLcontext *ctx, /* Detect cases where the application set the base level to an invalid * value. */ - if ((baseLevel < 0) || (baseLevel > MAX_TEXTURE_LEVELS)) { + if ((baseLevel < 0) || (baseLevel >= MAX_TEXTURE_LEVELS)) { char s[100]; _mesa_sprintf(s, "base level = %d is invalid", baseLevel); incomplete(t, s); diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c index 310d594cd55..db4c7a5edad 100644 --- a/src/mesa/main/texparam.c +++ b/src/mesa/main/texparam.c @@ -595,8 +595,10 @@ _mesa_TexParameterfv(GLenum target, GLenum pname, const GLfloat *params) case GL_DEPTH_TEXTURE_MODE_ARB: { /* convert float param to int */ - GLint p = (GLint) params[0]; - need_update = set_tex_parameteri(ctx, texObj, pname, &p); + GLint p[4]; + p[0] = (GLint) params[0]; + p[1] = p[2] = p[3] = 0; + need_update = set_tex_parameteri(ctx, texObj, pname, p); } break; @@ -645,14 +647,21 @@ _mesa_TexParameteri(GLenum target, GLenum pname, GLint param) case GL_TEXTURE_LOD_BIAS: case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB: { - GLfloat fparam = (GLfloat) param; + GLfloat fparam[4]; + fparam[0] = (GLfloat) param; + fparam[1] = fparam[2] = fparam[3] = 0.0F; /* convert int param to float */ - need_update = set_tex_parameterf(ctx, texObj, pname, &fparam); + need_update = set_tex_parameterf(ctx, texObj, pname, fparam); } break; default: /* this will generate an error if pname is illegal */ - need_update = set_tex_parameteri(ctx, texObj, pname, ¶m); + { + GLint iparam[4]; + iparam[0] = param; + iparam[1] = iparam[2] = iparam[3] = 0; + need_update = set_tex_parameteri(ctx, texObj, pname, iparam); + } } if (ctx->Driver.TexParameter && need_update) { @@ -694,8 +703,10 @@ _mesa_TexParameteriv(GLenum target, GLenum pname, const GLint *params) case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB: { /* convert int param to float */ - GLfloat fparam = (GLfloat) params[0]; - need_update = set_tex_parameterf(ctx, texObj, pname, &fparam); + GLfloat fparams[4]; + fparams[0] = (GLfloat) params[0]; + fparams[1] = fparams[2] = fparams[3] = 0.0F; + need_update = set_tex_parameterf(ctx, texObj, pname, fparams); } break; default: diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index dd732b6666b..a09be71020e 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -56,7 +56,6 @@ having three separate program parameter arrays. #include "main/context.h" #include "main/macros.h" #include "main/mtypes.h" -#include "shader/grammar/grammar_mesa.h" #include "arbprogparse.h" #include "program.h" #include "programopt.h" diff --git a/src/mesa/shader/descrip.mms b/src/mesa/shader/descrip.mms index 19bafd48302..59730020d0c 100644 --- a/src/mesa/shader/descrip.mms +++ b/src/mesa/shader/descrip.mms @@ -16,7 +16,7 @@ VPATH = RCS -INCDIR = [---.include],[.grammar],[-.main],[-.glapi],[.slang] +INCDIR = [---.include],[-.main],[-.glapi],[.slang] LIBDIR = [---.lib] CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1,"__extension__=")/name=(as_is,short)/float=ieee/ieee=denorm @@ -64,8 +64,6 @@ all : $(MMS)$(MMSQUALIFIERS) $(LIBDIR)$(GL_LIB) set def [.slang] $(MMS)$(MMSQUALIFIERS) - set def [-.grammar] - $(MMS)$(MMSQUALIFIERS) set def [-] # Make the library diff --git a/src/mesa/shader/grammar/grammar.c b/src/mesa/shader/grammar/grammar.c deleted file mode 100644 index a9775961d3a..00000000000 --- a/src/mesa/shader/grammar/grammar.c +++ /dev/null @@ -1,3178 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.6 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file grammar.c - * syntax parsing engine - * \author Michal Krol - */ - -#ifndef GRAMMAR_PORT_BUILD -#error Do not build this file directly, build your grammar_XXX.c instead, which includes this file -#endif - -/* -*/ - -/* - INTRODUCTION - ------------ - - The task is to check the syntax of an input string. Input string is a stream of ASCII - characters terminated with a null-character ('\0'). Checking it using C language is - difficult and hard to implement without bugs. It is hard to maintain and make changes when - the syntax changes. - - This is because of a high redundancy of the C code. Large blocks of code are duplicated with - only small changes. Even use of macros does not solve the problem because macros cannot - erase the complexity of the problem. - - The resolution is to create a new language that will be highly oriented to our task. Once - we describe a particular syntax, we are done. We can then focus on the code that implements - the language. The size and complexity of it is relatively small than the code that directly - checks the syntax. - - First, we must implement our new language. Here, the language is implemented in C, but it - could also be implemented in any other language. The code is listed below. We must take - a good care that it is bug free. This is simple because the code is simple and clean. - - Next, we must describe the syntax of our new language in itself. Once created and checked - manually that it is correct, we can use it to check another scripts. - - Note that our new language loading code does not have to check the syntax. It is because we - assume that the script describing itself is correct, and other scripts can be syntactically - checked by the former script. The loading code must only do semantic checking which leads us to - simple resolving references. - - THE LANGUAGE - ------------ - - Here I will describe the syntax of the new language (further called "Synek"). It is mainly a - sequence of declarations terminated by a semicolon. The declaration consists of a symbol, - which is an identifier, and its definition. A definition is in turn a sequence of specifiers - connected with ".and" or ".or" operator. These operators cannot be mixed together in a one - definition. Specifier can be a symbol, string, character, character range or a special - keyword ".true" or ".false". - - On the very beginning of the script there is a declaration of a root symbol and is in the form: - .syntax <root_symbol>; - The <root_symbol> must be on of the symbols in declaration sequence. The syntax is correct if - the root symbol evaluates to true. A symbol evaluates to true if the definition associated with - the symbol evaluates to true. Definition evaluation depends on the operator used to connect - specifiers in the definition. If ".and" operator is used, definition evaluates to true if and - only if all the specifiers evaluate to true. If ".or" operator is used, definition evalutes to - true if any of the specifiers evaluates to true. If definition contains only one specifier, - it is evaluated as if it was connected with ".true" keyword by ".and" operator. - - If specifier is a ".true" keyword, it always evaluates to true. - - If specifier is a ".false" keyword, it always evaluates to false. Specifier evaluates to false - when it does not evaluate to true. - - Character range specifier is in the form: - '<first_character>' - '<second_character>' - If specifier is a character range, it evaluates to true if character in the stream is greater - or equal to <first_character> and less or equal to <second_character>. In that situation - the stream pointer is advanced to point to next character in the stream. All C-style escape - sequences are supported although trigraph sequences are not. The comparisions are performed - on 8-bit unsigned integers. - - Character specifier is in the form: - '<single_character>' - It evaluates to true if the following character range specifier evaluates to true: - '<single_character>' - '<single_character>' - - String specifier is in the form: - "<string>" - Let N be the number of characters in <string>. Let <string>[i] designate i-th character in - <string>. Then the string specifier evaluates to true if and only if for i in the range [0, N) - the following character specifier evaluates to true: - '<string>[i]' - If <string>[i] is a quotation mark, '<string>[i]' is replaced with '\<string>[i]'. - - Symbol specifier can be optionally preceded by a ".loop" keyword in the form: - .loop <symbol> (1) - where <symbol> is defined as follows: - <symbol> <definition>; (2) - Construction (1) is replaced by the following code: - <symbol$1> - and declaration (2) is replaced by the following: - <symbol$1> <symbol$2> .or .true; - <symbol$2> <symbol> .and <symbol$1>; - <symbol> <definition>; - - Synek supports also a register mechanizm. User can, in its SYN file, declare a number of - registers that can be accessed in the syn body. Each reg has its name and a default value. - The register is one byte wide. The C code can change the default value by calling - grammar_set_reg8() with grammar id, register name and a new value. As we know, each rule is - a sequence of specifiers joined with .and or .or operator. And now each specifier can be - prefixed with a condition expression in a form ".if (<reg_name> <operator> <hex_literal>)" - where <operator> can be == or !=. If the condition evaluates to false, the specifier - evaluates to .false. Otherwise it evalutes to the specifier. - - ESCAPE SEQUENCES - ---------------- - - Synek supports all escape sequences in character specifiers. The mapping table is listed below. - All occurences of the characters in the first column are replaced with the corresponding - character in the second column. - - Escape sequence Represents - ------------------------------------------------------------------------------------------------ - \a Bell (alert) - \b Backspace - \f Formfeed - \n New line - \r Carriage return - \t Horizontal tab - \v Vertical tab - \' Single quotation mark - \" Double quotation mark - \\ Backslash - \? Literal question mark - \ooo ASCII character in octal notation - \xhhh ASCII character in hexadecimal notation - ------------------------------------------------------------------------------------------------ - - RAISING ERRORS - -------------- - - Any specifier can be followed by a special construction that is executed when the specifier - evaluates to false. The construction is in the form: - .error <ERROR_TEXT> - <ERROR_TEXT> is an identifier declared earlier by error text declaration. The declaration is - in the form: - .errtext <ERROR_TEXT> "<error_desc>" - When specifier evaluates to false and this construction is present, parsing is stopped - immediately and <error_desc> is returned as a result of parsing. The error position is also - returned and it is meant as an offset from the beggining of the stream to the character that - was valid so far. Example: - - (**** syntax script ****) - - .syntax program; - .errtext MISSING_SEMICOLON "missing ';'" - program declaration .and .loop space .and ';' .error MISSING_SEMICOLON .and - .loop space .and '\0'; - declaration "declare" .and .loop space .and identifier; - space ' '; - - (**** sample code ****) - - declare foo , - - In the example above checking the sample code will result in error message "missing ';'" and - error position 12. The sample code is not correct. Note the presence of '\0' specifier to - assure that there is no code after semicolon - only spaces. - <error_desc> can optionally contain identifier surrounded by dollar signs $. In such a case, - the identifier and dollar signs are replaced by a string retrieved by invoking symbol with - the identifier name. The starting position is the error position. The lenght of the resulting - string is the position after invoking the symbol. - - PRODUCTION - ---------- - - Synek not only checks the syntax but it can also produce (emit) bytes associated with specifiers - that evaluate to true. That is, every specifier and optional error construction can be followed - by a number of emit constructions that are in the form: - .emit <parameter> - <paramater> can be a HEX number, identifier, a star * or a dollar $. HEX number is preceded by - 0x or 0X. If <parameter> is an identifier, it must be earlier declared by emit code declaration - in the form: - .emtcode <identifier> <hex_number> - - When given specifier evaluates to true, all emits associated with the specifier are output - in order they were declared. A star means that last-read character should be output instead - of constant value. Example: - - (**** syntax script ****) - - .syntax foobar; - .emtcode WORD_FOO 0x01 - .emtcode WORD_BAR 0x02 - foobar FOO .emit WORD_FOO .or BAR .emit WORD_BAR .or .true .emit 0x00; - FOO "foo" .and SPACE; - BAR "bar" .and SPACE; - SPACE ' ' .or '\0'; - - (**** sample text 1 ****) - - foo - - (**** sample text 2 ****) - - foobar - - For both samples the result will be one-element array. For first sample text it will be - value 1, for second - 0. Note that every text will be accepted because of presence of - .true as an alternative. - - Another example: - - (**** syntax script ****) - - .syntax declaration; - .emtcode VARIABLE 0x01 - declaration "declare" .and .loop space .and - identifier .emit VARIABLE .and (1) - .true .emit 0x00 .and (2) - .loop space .and ';'; - space ' ' .or '\t'; - identifier .loop id_char .emit *; (3) - id_char 'a'-'z' .or 'A'-'Z' .or '_'; - - (**** sample code ****) - - declare fubar; - - In specifier (1) symbol <identifier> is followed by .emit VARIABLE. If it evaluates to - true, VARIABLE constant and then production of the symbol is output. Specifier (2) is used - to terminate the string with null to signal when the string ends. Specifier (3) outputs - all characters that make declared identifier. The result of sample code will be the - following array: - { 1, 'f', 'u', 'b', 'a', 'r', 0 } - - If .emit is followed by dollar $, it means that current position should be output. Current - position is a 32-bit unsigned integer distance from the very beginning of the parsed string to - first character consumed by the specifier associated with the .emit instruction. Current - position is stored in the output buffer in Little-Endian convention (the lowest byte comes - first). -*/ - -#include <stdio.h> - -static void mem_free (void **); - -/* - internal error messages -*/ -static const byte *OUT_OF_MEMORY = (byte *) "internal error 1001: out of physical memory"; -static const byte *UNRESOLVED_REFERENCE = (byte *) "internal error 1002: unresolved reference '$'"; -static const byte *INVALID_GRAMMAR_ID = (byte *) "internal error 1003: invalid grammar object"; -static const byte *INVALID_REGISTER_NAME = (byte *) "internal error 1004: invalid register name: '$'"; -/*static const byte *DUPLICATE_IDENTIFIER = (byte *) "internal error 1005: identifier '$' already defined";*/ -static const byte *UNREFERENCED_IDENTIFIER =(byte *) "internal error 1006: unreferenced identifier '$'"; - -static const byte *error_message = NULL; /* points to one of the error messages above */ -static byte *error_param = NULL; /* this is inserted into error_message in place of $ */ -static int error_position = -1; - -static byte *unknown = (byte *) "???"; - -static void clear_last_error (void) -{ - /* reset error message */ - error_message = NULL; - - /* free error parameter - if error_param is a "???" don't free it - it's static */ - if (error_param != unknown) - mem_free ((void **) (void *) &error_param); - else - error_param = NULL; - - /* reset error position */ - error_position = -1; -} - -static void set_last_error (const byte *msg, byte *param, int pos) -{ - /* error message can be set only once */ - if (error_message != NULL) - { - mem_free ((void **) (void *) ¶m); - return; - } - - error_message = msg; - - /* if param is NULL, set error_param to unknown ("???") */ - /* note: do not try to strdup the "???" - it may be that we are here because of */ - /* out of memory error so strdup can fail */ - if (param != NULL) - error_param = param; - else - error_param = unknown; - - error_position = pos; -} - -/* - memory management routines -*/ -static void *mem_alloc (size_t size) -{ - void *ptr = grammar_alloc_malloc (size); - if (ptr == NULL) - set_last_error (OUT_OF_MEMORY, NULL, -1); - return ptr; -} - -static void *mem_copy (void *dst, const void *src, size_t size) -{ - return grammar_memory_copy (dst, src, size); -} - -static void mem_free (void **ptr) -{ - grammar_alloc_free (*ptr); - *ptr = NULL; -} - -static void *mem_realloc (void *ptr, size_t old_size, size_t new_size) -{ - void *ptr2 = grammar_alloc_realloc (ptr, old_size, new_size); - if (ptr2 == NULL) - set_last_error (OUT_OF_MEMORY, NULL, -1); - return ptr2; -} - -static byte *str_copy_n (byte *dst, const byte *src, size_t max_len) -{ - return grammar_string_copy_n (dst, src, max_len); -} - -static byte *str_duplicate (const byte *str) -{ - byte *new_str = grammar_string_duplicate (str); - if (new_str == NULL) - set_last_error (OUT_OF_MEMORY, NULL, -1); - return new_str; -} - -static int str_equal (const byte *str1, const byte *str2) -{ - return grammar_string_compare (str1, str2) == 0; -} - -static int str_equal_n (const byte *str1, const byte *str2, unsigned int n) -{ - return grammar_string_compare_n (str1, str2, n) == 0; -} - -static int -str_length (const byte *str) -{ - return (int) (grammar_string_length (str)); -} - -/* - useful macros -*/ -#define GRAMMAR_IMPLEMENT_LIST_APPEND(_Ty)\ - static void _Ty##_append (_Ty **x, _Ty *nx) {\ - while (*x) x = &(**x).next;\ - *x = nx;\ - } - -/* - string to byte map typedef -*/ -typedef struct map_byte_ -{ - byte *key; - byte data; - struct map_byte_ *next; -} map_byte; - -static void map_byte_create (map_byte **ma) -{ - *ma = (map_byte *) mem_alloc (sizeof (map_byte)); - if (*ma) - { - (**ma).key = NULL; - (**ma).data = '\0'; - (**ma).next = NULL; - } -} - -static void map_byte_destroy (map_byte **ma) -{ - if (*ma) - { - map_byte_destroy (&(**ma).next); - mem_free ((void **) &(**ma).key); - mem_free ((void **) ma); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(map_byte) - -/* - searches the map for the specified key, - returns pointer to the element with the specified key if it exists - returns NULL otherwise -*/ -static map_byte *map_byte_locate (map_byte **ma, const byte *key) -{ - while (*ma) - { - if (str_equal ((**ma).key, key)) - return *ma; - - ma = &(**ma).next; - } - - set_last_error (UNRESOLVED_REFERENCE, str_duplicate (key), -1); - return NULL; -} - -/* - searches the map for specified key, - if the key is matched, *data is filled with data associated with the key, - returns 0 if the key is matched, - returns 1 otherwise -*/ -static int map_byte_find (map_byte **ma, const byte *key, byte *data) -{ - map_byte *found = map_byte_locate (ma, key); - if (found != NULL) - { - *data = found->data; - - return 0; - } - - return 1; -} - -/* - regbyte context typedef - - Each regbyte consists of its name and a default value. These are static and created at - grammar script compile-time, for example the following line: - .regbyte vertex_blend 0x00 - adds a new regbyte named "vertex_blend" to the static list and initializes it to 0. - When the script is executed, this regbyte can be accessed by name for read and write. When a - particular regbyte is written, a new regbyte_ctx entry is added to the top of the regbyte_ctx - stack. The new entry contains information abot which regbyte it references and its new value. - When a given regbyte is accessed for read, the stack is searched top-down to find an - entry that references the regbyte. The first matching entry is used to return the current - value it holds. If no entry is found, the default value is returned. -*/ -typedef struct regbyte_ctx_ -{ - map_byte *m_regbyte; - byte m_current_value; - struct regbyte_ctx_ *m_prev; -} regbyte_ctx; - -static void regbyte_ctx_create (regbyte_ctx **re) -{ - *re = (regbyte_ctx *) mem_alloc (sizeof (regbyte_ctx)); - if (*re) - { - (**re).m_regbyte = NULL; - (**re).m_prev = NULL; - } -} - -static void regbyte_ctx_destroy (regbyte_ctx **re) -{ - if (*re) - { - mem_free ((void **) re); - } -} - -static byte regbyte_ctx_extract (regbyte_ctx **re, map_byte *reg) -{ - /* first lookup in the register stack */ - while (*re != NULL) - { - if ((**re).m_regbyte == reg) - return (**re).m_current_value; - - re = &(**re).m_prev; - } - - /* if not found - return the default value */ - return reg->data; -} - -/* - emit type typedef -*/ -typedef enum emit_type_ -{ - et_byte, /* explicit number */ - et_stream, /* eaten character */ - et_position /* current position */ -} emit_type; - -/* - emit destination typedef -*/ -typedef enum emit_dest_ -{ - ed_output, /* write to the output buffer */ - ed_regbyte /* write a particular regbyte */ -} emit_dest; - -/* - emit typedef -*/ -typedef struct emit_ -{ - emit_dest m_emit_dest; - emit_type m_emit_type; /* ed_output */ - byte m_byte; /* et_byte */ - map_byte *m_regbyte; /* ed_regbyte */ - byte *m_regname; /* ed_regbyte - temporary */ - struct emit_ *m_next; -} emit; - -static void emit_create (emit **em) -{ - *em = (emit *) mem_alloc (sizeof (emit)); - if (*em) - { - (**em).m_emit_dest = ed_output; - (**em).m_emit_type = et_byte; - (**em).m_byte = '\0'; - (**em).m_regbyte = NULL; - (**em).m_regname = NULL; - (**em).m_next = NULL; - } -} - -static void emit_destroy (emit **em) -{ - if (*em) - { - emit_destroy (&(**em).m_next); - mem_free ((void **) &(**em).m_regname); - mem_free ((void **) em); - } -} - -static unsigned int emit_size (emit *_E) -{ - unsigned int n = 0; - - while (_E != NULL) - { - if (_E->m_emit_dest == ed_output) - { - if (_E->m_emit_type == et_position) - n += 4; /* position is a 32-bit unsigned integer */ - else - n++; - } - _E = _E->m_next; - } - - return n; -} - -static int emit_push (emit *_E, byte *_P, byte c, unsigned int _Pos, regbyte_ctx **_Ctx) -{ - while (_E != NULL) - { - if (_E->m_emit_dest == ed_output) - { - if (_E->m_emit_type == et_byte) - *_P++ = _E->m_byte; - else if (_E->m_emit_type == et_stream) - *_P++ = c; - else /* _Em->type == et_position */ - { - *_P++ = (byte) (_Pos); - *_P++ = (byte) (_Pos >> 8); - *_P++ = (byte) (_Pos >> 16); - *_P++ = (byte) (_Pos >> 24); - } - } - else - { - regbyte_ctx *new_rbc; - regbyte_ctx_create (&new_rbc); - if (new_rbc == NULL) - return 1; - - new_rbc->m_prev = *_Ctx; - new_rbc->m_regbyte = _E->m_regbyte; - *_Ctx = new_rbc; - - if (_E->m_emit_type == et_byte) - new_rbc->m_current_value = _E->m_byte; - else if (_E->m_emit_type == et_stream) - new_rbc->m_current_value = c; - } - - _E = _E->m_next; - } - - return 0; -} - -/* - error typedef -*/ -typedef struct error_ -{ - byte *m_text; - byte *m_token_name; - struct rule_ *m_token; -} error; - -static void error_create (error **er) -{ - *er = (error *) mem_alloc (sizeof (error)); - if (*er) - { - (**er).m_text = NULL; - (**er).m_token_name = NULL; - (**er).m_token = NULL; - } -} - -static void error_destroy (error **er) -{ - if (*er) - { - mem_free ((void **) &(**er).m_text); - mem_free ((void **) &(**er).m_token_name); - mem_free ((void **) er); - } -} - -struct dict_; - -static byte * -error_get_token (error *, struct dict_ *, const byte *, int); - -/* - condition operand type typedef -*/ -typedef enum cond_oper_type_ -{ - cot_byte, /* constant 8-bit unsigned integer */ - cot_regbyte /* pointer to byte register containing the current value */ -} cond_oper_type; - -/* - condition operand typedef -*/ -typedef struct cond_oper_ -{ - cond_oper_type m_type; - byte m_byte; /* cot_byte */ - map_byte *m_regbyte; /* cot_regbyte */ - byte *m_regname; /* cot_regbyte - temporary */ -} cond_oper; - -/* - condition type typedef -*/ -typedef enum cond_type_ -{ - ct_equal, - ct_not_equal -} cond_type; - -/* - condition typedef -*/ -typedef struct cond_ -{ - cond_type m_type; - cond_oper m_operands[2]; -} cond; - -static void cond_create (cond **co) -{ - *co = (cond *) mem_alloc (sizeof (cond)); - if (*co) - { - (**co).m_operands[0].m_regname = NULL; - (**co).m_operands[1].m_regname = NULL; - } -} - -static void cond_destroy (cond **co) -{ - if (*co) - { - mem_free ((void **) &(**co).m_operands[0].m_regname); - mem_free ((void **) &(**co).m_operands[1].m_regname); - mem_free ((void **) co); - } -} - -/* - specifier type typedef -*/ -typedef enum spec_type_ -{ - st_false, - st_true, - st_byte, - st_byte_range, - st_string, - st_identifier, - st_identifier_loop, - st_debug -} spec_type; - -/* - specifier typedef -*/ -typedef struct spec_ -{ - spec_type m_spec_type; - byte m_byte[2]; /* st_byte, st_byte_range */ - byte *m_string; /* st_string */ - struct rule_ *m_rule; /* st_identifier, st_identifier_loop */ - emit *m_emits; - error *m_errtext; - cond *m_cond; - struct spec_ *next; -} spec; - -static void spec_create (spec **sp) -{ - *sp = (spec *) mem_alloc (sizeof (spec)); - if (*sp) - { - (**sp).m_spec_type = st_false; - (**sp).m_byte[0] = '\0'; - (**sp).m_byte[1] = '\0'; - (**sp).m_string = NULL; - (**sp).m_rule = NULL; - (**sp).m_emits = NULL; - (**sp).m_errtext = NULL; - (**sp).m_cond = NULL; - (**sp).next = NULL; - } -} - -static void spec_destroy (spec **sp) -{ - if (*sp) - { - spec_destroy (&(**sp).next); - emit_destroy (&(**sp).m_emits); - error_destroy (&(**sp).m_errtext); - mem_free ((void **) &(**sp).m_string); - cond_destroy (&(**sp).m_cond); - mem_free ((void **) sp); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(spec) - -/* - operator typedef -*/ -typedef enum oper_ -{ - op_none, - op_and, - op_or -} oper; - -/* - rule typedef -*/ -typedef struct rule_ -{ - oper m_oper; - spec *m_specs; - struct rule_ *next; - int m_referenced; -} rule; - -static void rule_create (rule **ru) -{ - *ru = (rule *) mem_alloc (sizeof (rule)); - if (*ru) - { - (**ru).m_oper = op_none; - (**ru).m_specs = NULL; - (**ru).next = NULL; - (**ru).m_referenced = 0; - } -} - -static void rule_destroy (rule **ru) -{ - if (*ru) - { - rule_destroy (&(**ru).next); - spec_destroy (&(**ru).m_specs); - mem_free ((void **) ru); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(rule) - -/* - returns unique grammar id -*/ -static grammar next_valid_grammar_id (void) -{ - static grammar id = 0; - - return ++id; -} - -/* - dictionary typedef -*/ -typedef struct dict_ -{ - rule *m_rulez; - rule *m_syntax; - rule *m_string; - map_byte *m_regbytes; - grammar m_id; - struct dict_ *next; -} dict; - -static void dict_create (dict **di) -{ - *di = (dict *) mem_alloc (sizeof (dict)); - if (*di) - { - (**di).m_rulez = NULL; - (**di).m_syntax = NULL; - (**di).m_string = NULL; - (**di).m_regbytes = NULL; - (**di).m_id = next_valid_grammar_id (); - (**di).next = NULL; - } -} - -static void dict_destroy (dict **di) -{ - if (*di) - { - rule_destroy (&(**di).m_rulez); - map_byte_destroy (&(**di).m_regbytes); - mem_free ((void **) di); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(dict) - -static void dict_find (dict **di, grammar key, dict **data) -{ - while (*di) - { - if ((**di).m_id == key) - { - *data = *di; - return; - } - - di = &(**di).next; - } - - *data = NULL; -} - -static dict *g_dicts = NULL; - -/* - byte array typedef -*/ -typedef struct barray_ -{ - byte *data; - unsigned int len; -} barray; - -static void barray_create (barray **ba) -{ - *ba = (barray *) mem_alloc (sizeof (barray)); - if (*ba) - { - (**ba).data = NULL; - (**ba).len = 0; - } -} - -static void barray_destroy (barray **ba) -{ - if (*ba) - { - mem_free ((void **) &(**ba).data); - mem_free ((void **) ba); - } -} - -/* - reallocates byte array to requested size, - returns 0 on success, - returns 1 otherwise -*/ -static int barray_resize (barray **ba, unsigned int nlen) -{ - byte *new_pointer; - - if (nlen == 0) - { - mem_free ((void **) &(**ba).data); - (**ba).data = NULL; - (**ba).len = 0; - - return 0; - } - else - { - new_pointer = (byte *) mem_realloc ((**ba).data, (**ba).len * sizeof (byte), - nlen * sizeof (byte)); - if (new_pointer) - { - (**ba).data = new_pointer; - (**ba).len = nlen; - - return 0; - } - } - - return 1; -} - -/* - adds byte array pointed by *nb to the end of array pointed by *ba, - returns 0 on success, - returns 1 otherwise -*/ -static int barray_append (barray **ba, barray **nb) -{ - const unsigned int len = (**ba).len; - - if (barray_resize (ba, (**ba).len + (**nb).len)) - return 1; - - mem_copy ((**ba).data + len, (**nb).data, (**nb).len); - - return 0; -} - -/* - adds emit chain pointed by em to the end of array pointed by *ba, - returns 0 on success, - returns 1 otherwise -*/ -static int barray_push (barray **ba, emit *em, byte c, unsigned int pos, regbyte_ctx **rbc) -{ - unsigned int count = emit_size (em); - - if (barray_resize (ba, (**ba).len + count)) - return 1; - - return emit_push (em, (**ba).data + ((**ba).len - count), c, pos, rbc); -} - -/* - byte pool typedef -*/ -typedef struct bytepool_ -{ - byte *_F; - unsigned int _Siz; -} bytepool; - -static void bytepool_destroy (bytepool **by) -{ - if (*by != NULL) - { - mem_free ((void **) &(**by)._F); - mem_free ((void **) by); - } -} - -static void bytepool_create (bytepool **by, int len) -{ - *by = (bytepool *) (mem_alloc (sizeof (bytepool))); - if (*by != NULL) - { - (**by)._F = (byte *) (mem_alloc (sizeof (byte) * len)); - (**by)._Siz = len; - - if ((**by)._F == NULL) - bytepool_destroy (by); - } -} - -static int bytepool_reserve (bytepool *by, unsigned int n) -{ - byte *_P; - - if (n <= by->_Siz) - return 0; - - /* byte pool can only grow and at least by doubling its size */ - n = n >= by->_Siz * 2 ? n : by->_Siz * 2; - - /* reallocate the memory and adjust pointers to the new memory location */ - _P = (byte *) (mem_realloc (by->_F, sizeof (byte) * by->_Siz, sizeof (byte) * n)); - if (_P != NULL) - { - by->_F = _P; - by->_Siz = n; - return 0; - } - - return 1; -} - -/* - string to string map typedef -*/ -typedef struct map_str_ -{ - byte *key; - byte *data; - struct map_str_ *next; -} map_str; - -static void map_str_create (map_str **ma) -{ - *ma = (map_str *) mem_alloc (sizeof (map_str)); - if (*ma) - { - (**ma).key = NULL; - (**ma).data = NULL; - (**ma).next = NULL; - } -} - -static void map_str_destroy (map_str **ma) -{ - if (*ma) - { - map_str_destroy (&(**ma).next); - mem_free ((void **) &(**ma).key); - mem_free ((void **) &(**ma).data); - mem_free ((void **) ma); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(map_str) - -/* - searches the map for specified key, - if the key is matched, *data is filled with data associated with the key, - returns 0 if the key is matched, - returns 1 otherwise -*/ -static int map_str_find (map_str **ma, const byte *key, byte **data) -{ - while (*ma) - { - if (str_equal ((**ma).key, key)) - { - *data = str_duplicate ((**ma).data); - if (*data == NULL) - return 1; - - return 0; - } - - ma = &(**ma).next; - } - - set_last_error (UNRESOLVED_REFERENCE, str_duplicate (key), -1); - return 1; -} - -/* - string to rule map typedef -*/ -typedef struct map_rule_ -{ - byte *key; - rule *data; - struct map_rule_ *next; -} map_rule; - -static void map_rule_create (map_rule **ma) -{ - *ma = (map_rule *) mem_alloc (sizeof (map_rule)); - if (*ma) - { - (**ma).key = NULL; - (**ma).data = NULL; - (**ma).next = NULL; - } -} - -static void map_rule_destroy (map_rule **ma) -{ - if (*ma) - { - map_rule_destroy (&(**ma).next); - mem_free ((void **) &(**ma).key); - mem_free ((void **) ma); - } -} - -GRAMMAR_IMPLEMENT_LIST_APPEND(map_rule) - -/* - searches the map for specified key, - if the key is matched, *data is filled with data associated with the key, - returns 0 if the is matched, - returns 1 otherwise -*/ -static int map_rule_find (map_rule **ma, const byte *key, rule **data) -{ - while (*ma) - { - if (str_equal ((**ma).key, key)) - { - *data = (**ma).data; - - return 0; - } - - ma = &(**ma).next; - } - - set_last_error (UNRESOLVED_REFERENCE, str_duplicate (key), -1); - return 1; -} - -/* - returns 1 if given character is a white space, - returns 0 otherwise -*/ -static int is_space (byte c) -{ - return c == ' ' || c == '\t' || c == '\n' || c == '\r'; -} - -/* - advances text pointer by 1 if character pointed by *text is a space, - returns 1 if a space has been eaten, - returns 0 otherwise -*/ -static int eat_space (const byte **text) -{ - if (is_space (**text)) - { - (*text)++; - - return 1; - } - - return 0; -} - -/* - returns 1 if text points to C-style comment start string, - returns 0 otherwise -*/ -static int is_comment_start (const byte *text) -{ - return text[0] == '/' && text[1] == '*'; -} - -/* - advances text pointer to first character after C-style comment block - if any, - returns 1 if C-style comment block has been encountered and eaten, - returns 0 otherwise -*/ -static int eat_comment (const byte **text) -{ - if (is_comment_start (*text)) - { - /* *text points to comment block - skip two characters to enter comment body */ - *text += 2; - /* skip any character except consecutive '*' and '/' */ - while (!((*text)[0] == '*' && (*text)[1] == '/')) - (*text)++; - /* skip those two terminating characters */ - *text += 2; - - return 1; - } - - return 0; -} - -/* - advances text pointer to first character that is neither space nor C-style comment block -*/ -static void eat_spaces (const byte **text) -{ - while (eat_space (text) || eat_comment (text)) - ; -} - -/* - resizes string pointed by *ptr to successfully add character c to the end of the string, - returns 0 on success, - returns 1 otherwise -*/ -static int string_grow (byte **ptr, unsigned int *len, byte c) -{ - /* reallocate the string in 16-byte increments */ - if ((*len & 0x0F) == 0x0F || *ptr == NULL) - { - byte *tmp = (byte *) mem_realloc (*ptr, ((*len + 1) & ~0x0F) * sizeof (byte), - ((*len + 1 + 0x10) & ~0x0F) * sizeof (byte)); - if (tmp == NULL) - return 1; - - *ptr = tmp; - } - - if (c) - { - /* append given character */ - (*ptr)[*len] = c; - (*len)++; - } - (*ptr)[*len] = '\0'; - - return 0; -} - -/* - returns 1 if given character is a valid identifier character a-z, A-Z, 0-9 or _ - returns 0 otherwise -*/ -static int is_identifier (byte c) -{ - return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || (c >= '0' && c <= '9') || c == '_'; -} - -/* - copies characters from *text to *id until non-identifier character is encountered, - assumes that *id points to NULL object - caller is responsible for later freeing the string, - text pointer is advanced to point past the copied identifier, - returns 0 if identifier was successfully copied, - returns 1 otherwise -*/ -static int get_identifier (const byte **text, byte **id) -{ - const byte *t = *text; - byte *p = NULL; - unsigned int len = 0; - - if (string_grow (&p, &len, '\0')) - return 1; - - /* loop while next character in buffer is valid for identifiers */ - while (is_identifier (*t)) - { - if (string_grow (&p, &len, *t++)) - { - mem_free ((void **) (void *) &p); - return 1; - } - } - - *text = t; - *id = p; - - return 0; -} - -/* - converts sequence of DEC digits pointed by *text until non-DEC digit is encountered, - advances text pointer past the converted sequence, - returns the converted value -*/ -static unsigned int dec_convert (const byte **text) -{ - unsigned int value = 0; - - while (**text >= '0' && **text <= '9') - { - value = value * 10 + **text - '0'; - (*text)++; - } - - return value; -} - -/* - returns 1 if given character is HEX digit 0-9, A-F or a-f, - returns 0 otherwise -*/ -static int is_hex (byte c) -{ - return (c >= '0' && c <= '9') || (c >= 'A' && c <= 'F') || (c >= 'a' && c <= 'f'); -} - -/* - returns value of passed character as if it was HEX digit -*/ -static unsigned int hex2dec (byte c) -{ - if (c >= '0' && c <= '9') - return c - '0'; - if (c >= 'A' && c <= 'F') - return c - 'A' + 10; - return c - 'a' + 10; -} - -/* - converts sequence of HEX digits pointed by *text until non-HEX digit is encountered, - advances text pointer past the converted sequence, - returns the converted value -*/ -static unsigned int hex_convert (const byte **text) -{ - unsigned int value = 0; - - while (is_hex (**text)) - { - value = value * 0x10 + hex2dec (**text); - (*text)++; - } - - return value; -} - -/* - returns 1 if given character is OCT digit 0-7, - returns 0 otherwise -*/ -static int is_oct (byte c) -{ - return c >= '0' && c <= '7'; -} - -/* - returns value of passed character as if it was OCT digit -*/ -static int oct2dec (byte c) -{ - return c - '0'; -} - -static byte get_escape_sequence (const byte **text) -{ - int value = 0; - - /* skip '\' character */ - (*text)++; - - switch (*(*text)++) - { - case '\'': - return '\''; - case '"': - return '\"'; - case '?': - return '\?'; - case '\\': - return '\\'; - case 'a': - return '\a'; - case 'b': - return '\b'; - case 'f': - return '\f'; - case 'n': - return '\n'; - case 'r': - return '\r'; - case 't': - return '\t'; - case 'v': - return '\v'; - case 'x': - return (byte) hex_convert (text); - } - - (*text)--; - if (is_oct (**text)) - { - value = oct2dec (*(*text)++); - if (is_oct (**text)) - { - value = value * 010 + oct2dec (*(*text)++); - if (is_oct (**text)) - value = value * 010 + oct2dec (*(*text)++); - } - } - - return (byte) value; -} - -/* - copies characters from *text to *str until " or ' character is encountered, - assumes that *str points to NULL object - caller is responsible for later freeing the string, - assumes that *text points to " or ' character that starts the string, - text pointer is advanced to point past the " or ' character, - returns 0 if string was successfully copied, - returns 1 otherwise -*/ -static int get_string (const byte **text, byte **str) -{ - const byte *t = *text; - byte *p = NULL; - unsigned int len = 0; - byte term_char; - - if (string_grow (&p, &len, '\0')) - return 1; - - /* read " or ' character that starts the string */ - term_char = *t++; - /* while next character is not the terminating character */ - while (*t && *t != term_char) - { - byte c; - - if (*t == '\\') - c = get_escape_sequence (&t); - else - c = *t++; - - if (string_grow (&p, &len, c)) - { - mem_free ((void **) (void *) &p); - return 1; - } - } - /* skip " or ' character that ends the string */ - t++; - - *text = t; - *str = p; - return 0; -} - -/* - gets emit code, the syntax is: - ".emtcode" " " <symbol> " " (("0x" | "0X") <hex_value>) | <dec_value> | <character> - assumes that *text already points to <symbol>, - returns 0 if emit code is successfully read, - returns 1 otherwise -*/ -static int get_emtcode (const byte **text, map_byte **ma) -{ - const byte *t = *text; - map_byte *m = NULL; - - map_byte_create (&m); - if (m == NULL) - return 1; - - if (get_identifier (&t, &m->key)) - { - map_byte_destroy (&m); - return 1; - } - eat_spaces (&t); - - if (*t == '\'') - { - byte *c; - - if (get_string (&t, &c)) - { - map_byte_destroy (&m); - return 1; - } - - m->data = (byte) c[0]; - mem_free ((void **) (void *) &c); - } - else if (t[0] == '0' && (t[1] == 'x' || t[1] == 'X')) - { - /* skip HEX "0x" or "0X" prefix */ - t += 2; - m->data = (byte) hex_convert (&t); - } - else - { - m->data = (byte) dec_convert (&t); - } - - eat_spaces (&t); - - *text = t; - *ma = m; - return 0; -} - -/* - gets regbyte declaration, the syntax is: - ".regbyte" " " <symbol> " " (("0x" | "0X") <hex_value>) | <dec_value> | <character> - assumes that *text already points to <symbol>, - returns 0 if regbyte is successfully read, - returns 1 otherwise -*/ -static int get_regbyte (const byte **text, map_byte **ma) -{ - /* pass it to the emtcode parser as it has the same syntax starting at <symbol> */ - return get_emtcode (text, ma); -} - -/* - returns 0 on success, - returns 1 otherwise -*/ -static int get_errtext (const byte **text, map_str **ma) -{ - const byte *t = *text; - map_str *m = NULL; - - map_str_create (&m); - if (m == NULL) - return 1; - - if (get_identifier (&t, &m->key)) - { - map_str_destroy (&m); - return 1; - } - eat_spaces (&t); - - if (get_string (&t, &m->data)) - { - map_str_destroy (&m); - return 1; - } - eat_spaces (&t); - - *text = t; - *ma = m; - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int get_error (const byte **text, error **er, map_str *maps) -{ - const byte *t = *text; - byte *temp = NULL; - - if (*t != '.') - return 0; - - t++; - if (get_identifier (&t, &temp)) - return 1; - eat_spaces (&t); - - if (!str_equal ((byte *) "error", temp)) - { - mem_free ((void **) (void *) &temp); - return 0; - } - - mem_free ((void **) (void *) &temp); - - error_create (er); - if (*er == NULL) - return 1; - - if (*t == '\"') - { - if (get_string (&t, &(**er).m_text)) - { - error_destroy (er); - return 1; - } - eat_spaces (&t); - } - else - { - if (get_identifier (&t, &temp)) - { - error_destroy (er); - return 1; - } - eat_spaces (&t); - - if (map_str_find (&maps, temp, &(**er).m_text)) - { - mem_free ((void **) (void *) &temp); - error_destroy (er); - return 1; - } - - mem_free ((void **) (void *) &temp); - } - - /* try to extract "token" from "...$token$..." */ - { - byte *processed = NULL; - unsigned int len = 0; - int i = 0; - - if (string_grow (&processed, &len, '\0')) - { - error_destroy (er); - return 1; - } - - while (i < str_length ((**er).m_text)) - { - /* check if the dollar sign is repeated - if so skip it */ - if ((**er).m_text[i] == '$' && (**er).m_text[i + 1] == '$') - { - if (string_grow (&processed, &len, '$')) - { - mem_free ((void **) (void *) &processed); - error_destroy (er); - return 1; - } - - i += 2; - } - else if ((**er).m_text[i] != '$') - { - if (string_grow (&processed, &len, (**er).m_text[i])) - { - mem_free ((void **) (void *) &processed); - error_destroy (er); - return 1; - } - - i++; - } - else - { - if (string_grow (&processed, &len, '$')) - { - mem_free ((void **) (void *) &processed); - error_destroy (er); - return 1; - } - - { - /* length of token being extracted */ - unsigned int tlen = 0; - - if (string_grow (&(**er).m_token_name, &tlen, '\0')) - { - mem_free ((void **) (void *) &processed); - error_destroy (er); - return 1; - } - - /* skip the dollar sign */ - i++; - - while ((**er).m_text[i] != '$') - { - if (string_grow (&(**er).m_token_name, &tlen, (**er).m_text[i])) - { - mem_free ((void **) (void *) &processed); - error_destroy (er); - return 1; - } - - i++; - } - - /* skip the dollar sign */ - i++; - } - } - } - - mem_free ((void **) &(**er).m_text); - (**er).m_text = processed; - } - - *text = t; - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int get_emits (const byte **text, emit **em, map_byte *mapb) -{ - const byte *t = *text; - byte *temp = NULL; - emit *e = NULL; - emit_dest dest; - - if (*t != '.') - return 0; - - t++; - if (get_identifier (&t, &temp)) - return 1; - eat_spaces (&t); - - /* .emit */ - if (str_equal ((byte *) "emit", temp)) - dest = ed_output; - /* .load */ - else if (str_equal ((byte *) "load", temp)) - dest = ed_regbyte; - else - { - mem_free ((void **) (void *) &temp); - return 0; - } - - mem_free ((void **) (void *) &temp); - - emit_create (&e); - if (e == NULL) - return 1; - - e->m_emit_dest = dest; - - if (dest == ed_regbyte) - { - if (get_identifier (&t, &e->m_regname)) - { - emit_destroy (&e); - return 1; - } - eat_spaces (&t); - } - - /* 0xNN */ - if (*t == '0' && (t[1] == 'x' || t[1] == 'X')) - { - t += 2; - e->m_byte = (byte) hex_convert (&t); - - e->m_emit_type = et_byte; - } - /* NNN */ - else if (*t >= '0' && *t <= '9') - { - e->m_byte = (byte) dec_convert (&t); - - e->m_emit_type = et_byte; - } - /* * */ - else if (*t == '*') - { - t++; - - e->m_emit_type = et_stream; - } - /* $ */ - else if (*t == '$') - { - t++; - - e->m_emit_type = et_position; - } - /* 'c' */ - else if (*t == '\'') - { - if (get_string (&t, &temp)) - { - emit_destroy (&e); - return 1; - } - e->m_byte = (byte) temp[0]; - - mem_free ((void **) (void *) &temp); - - e->m_emit_type = et_byte; - } - else - { - if (get_identifier (&t, &temp)) - { - emit_destroy (&e); - return 1; - } - - if (map_byte_find (&mapb, temp, &e->m_byte)) - { - mem_free ((void **) (void *) &temp); - emit_destroy (&e); - return 1; - } - - mem_free ((void **) (void *) &temp); - - e->m_emit_type = et_byte; - } - - eat_spaces (&t); - - if (get_emits (&t, &e->m_next, mapb)) - { - emit_destroy (&e); - return 1; - } - - *text = t; - *em = e; - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int get_spec (const byte **text, spec **sp, map_str *maps, map_byte *mapb) -{ - const byte *t = *text; - spec *s = NULL; - - spec_create (&s); - if (s == NULL) - return 1; - - /* first - read optional .if statement */ - if (*t == '.') - { - const byte *u = t; - byte *keyword = NULL; - - /* skip the dot */ - u++; - - if (get_identifier (&u, &keyword)) - { - spec_destroy (&s); - return 1; - } - - /* .if */ - if (str_equal ((byte *) "if", keyword)) - { - cond_create (&s->m_cond); - if (s->m_cond == NULL) - { - spec_destroy (&s); - return 1; - } - - /* skip the left paren */ - eat_spaces (&u); - u++; - - /* get the left operand */ - eat_spaces (&u); - if (get_identifier (&u, &s->m_cond->m_operands[0].m_regname)) - { - spec_destroy (&s); - return 1; - } - s->m_cond->m_operands[0].m_type = cot_regbyte; - - /* get the operator (!= or ==) */ - eat_spaces (&u); - if (*u == '!') - s->m_cond->m_type = ct_not_equal; - else - s->m_cond->m_type = ct_equal; - u += 2; - eat_spaces (&u); - - if (u[0] == '0' && (u[1] == 'x' || u[1] == 'X')) - { - /* skip the 0x prefix */ - u += 2; - - /* get the right operand */ - s->m_cond->m_operands[1].m_byte = hex_convert (&u); - s->m_cond->m_operands[1].m_type = cot_byte; - } - else /*if (*u >= '0' && *u <= '9')*/ - { - /* get the right operand */ - s->m_cond->m_operands[1].m_byte = dec_convert (&u); - s->m_cond->m_operands[1].m_type = cot_byte; - } - - /* skip the right paren */ - eat_spaces (&u); - u++; - - eat_spaces (&u); - - t = u; - } - - mem_free ((void **) (void *) &keyword); - } - - if (*t == '\'') - { - byte *temp = NULL; - - if (get_string (&t, &temp)) - { - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - if (*t == '-') - { - byte *temp2 = NULL; - - /* skip the '-' character */ - t++; - eat_spaces (&t); - - if (get_string (&t, &temp2)) - { - mem_free ((void **) (void *) &temp); - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - s->m_spec_type = st_byte_range; - s->m_byte[0] = *temp; - s->m_byte[1] = *temp2; - - mem_free ((void **) (void *) &temp2); - } - else - { - s->m_spec_type = st_byte; - *s->m_byte = *temp; - } - - mem_free ((void **) (void *) &temp); - } - else if (*t == '"') - { - if (get_string (&t, &s->m_string)) - { - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - s->m_spec_type = st_string; - } - else if (*t == '.') - { - byte *keyword = NULL; - - /* skip the dot */ - t++; - - if (get_identifier (&t, &keyword)) - { - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - /* .true */ - if (str_equal ((byte *) "true", keyword)) - { - s->m_spec_type = st_true; - } - /* .false */ - else if (str_equal ((byte *) "false", keyword)) - { - s->m_spec_type = st_false; - } - /* .debug */ - else if (str_equal ((byte *) "debug", keyword)) - { - s->m_spec_type = st_debug; - } - /* .loop */ - else if (str_equal ((byte *) "loop", keyword)) - { - if (get_identifier (&t, &s->m_string)) - { - mem_free ((void **) (void *) &keyword); - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - s->m_spec_type = st_identifier_loop; - } - mem_free ((void **) (void *) &keyword); - } - else - { - if (get_identifier (&t, &s->m_string)) - { - spec_destroy (&s); - return 1; - } - eat_spaces (&t); - - s->m_spec_type = st_identifier; - } - - if (get_error (&t, &s->m_errtext, maps)) - { - spec_destroy (&s); - return 1; - } - - if (get_emits (&t, &s->m_emits, mapb)) - { - spec_destroy (&s); - return 1; - } - - *text = t; - *sp = s; - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int get_rule (const byte **text, rule **ru, map_str *maps, map_byte *mapb) -{ - const byte *t = *text; - rule *r = NULL; - - rule_create (&r); - if (r == NULL) - return 1; - - if (get_spec (&t, &r->m_specs, maps, mapb)) - { - rule_destroy (&r); - return 1; - } - - while (*t != ';') - { - byte *op = NULL; - spec *sp = NULL; - - /* skip the dot that precedes "and" or "or" */ - t++; - - /* read "and" or "or" keyword */ - if (get_identifier (&t, &op)) - { - rule_destroy (&r); - return 1; - } - eat_spaces (&t); - - if (r->m_oper == op_none) - { - /* .and */ - if (str_equal ((byte *) "and", op)) - r->m_oper = op_and; - /* .or */ - else - r->m_oper = op_or; - } - - mem_free ((void **) (void *) &op); - - if (get_spec (&t, &sp, maps, mapb)) - { - rule_destroy (&r); - return 1; - } - - spec_append (&r->m_specs, sp); - } - - /* skip the semicolon */ - t++; - eat_spaces (&t); - - *text = t; - *ru = r; - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int update_dependency (map_rule *mapr, byte *symbol, rule **ru) -{ - if (map_rule_find (&mapr, symbol, ru)) - return 1; - - (**ru).m_referenced = 1; - - return 0; -} - -/* - returns 0 on success, - returns 1 otherwise, -*/ -static int update_dependencies (dict *di, map_rule *mapr, byte **syntax_symbol, - byte **string_symbol, map_byte *regbytes) -{ - rule *rulez = di->m_rulez; - - /* update dependecies for the root and lexer symbols */ - if (update_dependency (mapr, *syntax_symbol, &di->m_syntax) || - (*string_symbol != NULL && update_dependency (mapr, *string_symbol, &di->m_string))) - return 1; - - mem_free ((void **) syntax_symbol); - mem_free ((void **) string_symbol); - - /* update dependecies for the rest of the rules */ - while (rulez) - { - spec *sp = rulez->m_specs; - - /* iterate through all the specifiers */ - while (sp) - { - /* update dependency for identifier */ - if (sp->m_spec_type == st_identifier || sp->m_spec_type == st_identifier_loop) - { - if (update_dependency (mapr, sp->m_string, &sp->m_rule)) - return 1; - - mem_free ((void **) &sp->m_string); - } - - /* some errtexts reference to a rule */ - if (sp->m_errtext && sp->m_errtext->m_token_name) - { - if (update_dependency (mapr, sp->m_errtext->m_token_name, &sp->m_errtext->m_token)) - return 1; - - mem_free ((void **) &sp->m_errtext->m_token_name); - } - - /* update dependency for condition */ - if (sp->m_cond) - { - int i; - for (i = 0; i < 2; i++) - if (sp->m_cond->m_operands[i].m_type == cot_regbyte) - { - sp->m_cond->m_operands[i].m_regbyte = map_byte_locate (®bytes, - sp->m_cond->m_operands[i].m_regname); - - if (sp->m_cond->m_operands[i].m_regbyte == NULL) - return 1; - - mem_free ((void **) &sp->m_cond->m_operands[i].m_regname); - } - } - - /* update dependency for all .load instructions */ - if (sp->m_emits) - { - emit *em = sp->m_emits; - while (em != NULL) - { - if (em->m_emit_dest == ed_regbyte) - { - em->m_regbyte = map_byte_locate (®bytes, em->m_regname); - - if (em->m_regbyte == NULL) - return 1; - - mem_free ((void **) &em->m_regname); - } - - em = em->m_next; - } - } - - sp = sp->next; - } - - rulez = rulez->next; - } - - /* check for unreferenced symbols */ - rulez = di->m_rulez; - while (rulez != NULL) - { - if (!rulez->m_referenced) - { - map_rule *ma = mapr; - while (ma) - { - if (ma->data == rulez) - { - set_last_error (UNREFERENCED_IDENTIFIER, str_duplicate (ma->key), -1); - return 1; - } - ma = ma->next; - } - } - rulez = rulez->next; - } - - return 0; -} - -static int satisfies_condition (cond *co, regbyte_ctx *ctx) -{ - byte values[2]; - int i; - - if (co == NULL) - return 1; - - for (i = 0; i < 2; i++) - switch (co->m_operands[i].m_type) - { - case cot_byte: - values[i] = co->m_operands[i].m_byte; - break; - case cot_regbyte: - values[i] = regbyte_ctx_extract (&ctx, co->m_operands[i].m_regbyte); - break; - } - - switch (co->m_type) - { - case ct_equal: - return values[0] == values[1]; - case ct_not_equal: - return values[0] != values[1]; - } - - return 0; -} - -static void free_regbyte_ctx_stack (regbyte_ctx *top, regbyte_ctx *limit) -{ - while (top != limit) - { - regbyte_ctx *rbc = top->m_prev; - regbyte_ctx_destroy (&top); - top = rbc; - } -} - -typedef enum match_result_ -{ - mr_not_matched, /* the examined string does not match */ - mr_matched, /* the examined string matches */ - mr_error_raised, /* mr_not_matched + error has been raised */ - mr_dont_emit, /* used by identifier loops only */ - mr_internal_error /* an internal error has occured such as out of memory */ -} match_result; - -/* - * This function does the main job. It parses the text and generates output data. - */ -static match_result -match (dict *di, const byte *text, int *index, rule *ru, barray **ba, int filtering_string, - regbyte_ctx **rbc) -{ - int ind = *index; - match_result status = mr_not_matched; - spec *sp = ru->m_specs; - regbyte_ctx *ctx = *rbc; - - /* for every specifier in the rule */ - while (sp) - { - int i, len, save_ind = ind; - barray *array = NULL; - - if (satisfies_condition (sp->m_cond, ctx)) - { - switch (sp->m_spec_type) - { - case st_identifier: - barray_create (&array); - if (array == NULL) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - status = match (di, text, &ind, sp->m_rule, &array, filtering_string, &ctx); - - if (status == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - return mr_internal_error; - } - break; - case st_string: - len = str_length (sp->m_string); - - /* prefilter the stream */ - if (!filtering_string && di->m_string) - { - barray *ba; - int filter_index = 0; - match_result result; - regbyte_ctx *null_ctx = NULL; - - barray_create (&ba); - if (ba == NULL) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - result = match (di, text + ind, &filter_index, di->m_string, &ba, 1, &null_ctx); - - if (result == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&ba); - return mr_internal_error; - } - - if (result != mr_matched) - { - barray_destroy (&ba); - status = mr_not_matched; - break; - } - - barray_destroy (&ba); - - if (filter_index != len || !str_equal_n (sp->m_string, text + ind, len)) - { - status = mr_not_matched; - break; - } - - status = mr_matched; - ind += len; - } - else - { - status = mr_matched; - for (i = 0; status == mr_matched && i < len; i++) - if (text[ind + i] != sp->m_string[i]) - status = mr_not_matched; - - if (status == mr_matched) - ind += len; - } - break; - case st_byte: - status = text[ind] == *sp->m_byte ? mr_matched : mr_not_matched; - if (status == mr_matched) - ind++; - break; - case st_byte_range: - status = (text[ind] >= sp->m_byte[0] && text[ind] <= sp->m_byte[1]) ? - mr_matched : mr_not_matched; - if (status == mr_matched) - ind++; - break; - case st_true: - status = mr_matched; - break; - case st_false: - status = mr_not_matched; - break; - case st_debug: - status = ru->m_oper == op_and ? mr_matched : mr_not_matched; - break; - case st_identifier_loop: - barray_create (&array); - if (array == NULL) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - status = mr_dont_emit; - for (;;) - { - match_result result; - - save_ind = ind; - result = match (di, text, &ind, sp->m_rule, &array, filtering_string, &ctx); - - if (result == mr_error_raised) - { - status = result; - break; - } - else if (result == mr_matched) - { - if (barray_push (ba, sp->m_emits, text[ind - 1], save_ind, &ctx) || - barray_append (ba, &array)) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - return mr_internal_error; - } - barray_destroy (&array); - barray_create (&array); - if (array == NULL) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - } - else if (result == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - return mr_internal_error; - } - else - break; - } - break; - } - } - else - { - status = mr_not_matched; - } - - if (status == mr_error_raised) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - - return mr_error_raised; - } - - if (ru->m_oper == op_and && status != mr_matched && status != mr_dont_emit) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - - if (sp->m_errtext) - { - set_last_error (sp->m_errtext->m_text, error_get_token (sp->m_errtext, di, text, - ind), ind); - - return mr_error_raised; - } - - return mr_not_matched; - } - - if (status == mr_matched) - { - if (sp->m_emits) - if (barray_push (ba, sp->m_emits, text[ind - 1], save_ind, &ctx)) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - return mr_internal_error; - } - - if (array) - if (barray_append (ba, &array)) - { - free_regbyte_ctx_stack (ctx, *rbc); - barray_destroy (&array); - return mr_internal_error; - } - } - - barray_destroy (&array); - - /* if the rule operator is a logical or, we pick up the first matching specifier */ - if (ru->m_oper == op_or && (status == mr_matched || status == mr_dont_emit)) - { - *index = ind; - *rbc = ctx; - return mr_matched; - } - - sp = sp->next; - } - - /* everything went fine - all specifiers match up */ - if (ru->m_oper == op_and && (status == mr_matched || status == mr_dont_emit)) - { - *index = ind; - *rbc = ctx; - return mr_matched; - } - - free_regbyte_ctx_stack (ctx, *rbc); - return mr_not_matched; -} - -static match_result -fast_match (dict *di, const byte *text, int *index, rule *ru, int *_PP, bytepool *_BP, - int filtering_string, regbyte_ctx **rbc) -{ - int ind = *index; - int _P = filtering_string ? 0 : *_PP; - int _P2; - match_result status = mr_not_matched; - spec *sp = ru->m_specs; - regbyte_ctx *ctx = *rbc; - - /* for every specifier in the rule */ - while (sp) - { - int i, len, save_ind = ind; - - _P2 = _P + (sp->m_emits ? emit_size (sp->m_emits) : 0); - if (bytepool_reserve (_BP, _P2)) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - if (satisfies_condition (sp->m_cond, ctx)) - { - switch (sp->m_spec_type) - { - case st_identifier: - status = fast_match (di, text, &ind, sp->m_rule, &_P2, _BP, filtering_string, &ctx); - - if (status == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - break; - case st_string: - len = str_length (sp->m_string); - - /* prefilter the stream */ - if (!filtering_string && di->m_string) - { - int filter_index = 0; - match_result result; - regbyte_ctx *null_ctx = NULL; - - result = fast_match (di, text + ind, &filter_index, di->m_string, NULL, _BP, 1, &null_ctx); - - if (result == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - if (result != mr_matched) - { - status = mr_not_matched; - break; - } - - if (filter_index != len || !str_equal_n (sp->m_string, text + ind, len)) - { - status = mr_not_matched; - break; - } - - status = mr_matched; - ind += len; - } - else - { - status = mr_matched; - for (i = 0; status == mr_matched && i < len; i++) - if (text[ind + i] != sp->m_string[i]) - status = mr_not_matched; - - if (status == mr_matched) - ind += len; - } - break; - case st_byte: - status = text[ind] == *sp->m_byte ? mr_matched : mr_not_matched; - if (status == mr_matched) - ind++; - break; - case st_byte_range: - status = (text[ind] >= sp->m_byte[0] && text[ind] <= sp->m_byte[1]) ? - mr_matched : mr_not_matched; - if (status == mr_matched) - ind++; - break; - case st_true: - status = mr_matched; - break; - case st_false: - status = mr_not_matched; - break; - case st_debug: - status = ru->m_oper == op_and ? mr_matched : mr_not_matched; - break; - case st_identifier_loop: - status = mr_dont_emit; - for (;;) - { - match_result result; - - save_ind = ind; - result = fast_match (di, text, &ind, sp->m_rule, &_P2, _BP, filtering_string, &ctx); - - if (result == mr_error_raised) - { - status = result; - break; - } - else if (result == mr_matched) - { - if (!filtering_string) - { - if (sp->m_emits != NULL) - { - if (emit_push (sp->m_emits, _BP->_F + _P, text[ind - 1], save_ind, &ctx)) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - } - - _P = _P2; - _P2 += sp->m_emits ? emit_size (sp->m_emits) : 0; - if (bytepool_reserve (_BP, _P2)) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - } - } - else if (result == mr_internal_error) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - else - break; - } - break; - } - } - else - { - status = mr_not_matched; - } - - if (status == mr_error_raised) - { - free_regbyte_ctx_stack (ctx, *rbc); - - return mr_error_raised; - } - - if (ru->m_oper == op_and && status != mr_matched && status != mr_dont_emit) - { - free_regbyte_ctx_stack (ctx, *rbc); - - if (sp->m_errtext) - { - set_last_error (sp->m_errtext->m_text, error_get_token (sp->m_errtext, di, text, - ind), ind); - - return mr_error_raised; - } - - return mr_not_matched; - } - - if (status == mr_matched) - { - if (sp->m_emits != NULL) { - const byte ch = (ind <= 0) ? 0 : text[ind - 1]; - if (emit_push (sp->m_emits, _BP->_F + _P, ch, save_ind, &ctx)) - { - free_regbyte_ctx_stack (ctx, *rbc); - return mr_internal_error; - } - - } - _P = _P2; - } - - /* if the rule operator is a logical or, we pick up the first matching specifier */ - if (ru->m_oper == op_or && (status == mr_matched || status == mr_dont_emit)) - { - *index = ind; - *rbc = ctx; - if (!filtering_string) - *_PP = _P; - return mr_matched; - } - - sp = sp->next; - } - - /* everything went fine - all specifiers match up */ - if (ru->m_oper == op_and && (status == mr_matched || status == mr_dont_emit)) - { - *index = ind; - *rbc = ctx; - if (!filtering_string) - *_PP = _P; - return mr_matched; - } - - free_regbyte_ctx_stack (ctx, *rbc); - return mr_not_matched; -} - -static byte * -error_get_token (error *er, dict *di, const byte *text, int ind) -{ - byte *str = NULL; - - if (er->m_token) - { - barray *ba; - int filter_index = 0; - regbyte_ctx *ctx = NULL; - - barray_create (&ba); - if (ba != NULL) - { - if (match (di, text + ind, &filter_index, er->m_token, &ba, 0, &ctx) == mr_matched && - filter_index) - { - str = (byte *) mem_alloc (filter_index + 1); - if (str != NULL) - { - str_copy_n (str, text + ind, filter_index); - str[filter_index] = '\0'; - } - } - barray_destroy (&ba); - } - } - - return str; -} - -typedef struct grammar_load_state_ -{ - dict *di; - byte *syntax_symbol; - byte *string_symbol; - map_str *maps; - map_byte *mapb; - map_rule *mapr; -} grammar_load_state; - -static void grammar_load_state_create (grammar_load_state **gr) -{ - *gr = (grammar_load_state *) mem_alloc (sizeof (grammar_load_state)); - if (*gr) - { - (**gr).di = NULL; - (**gr).syntax_symbol = NULL; - (**gr).string_symbol = NULL; - (**gr).maps = NULL; - (**gr).mapb = NULL; - (**gr).mapr = NULL; - } -} - -static void grammar_load_state_destroy (grammar_load_state **gr) -{ - if (*gr) - { - dict_destroy (&(**gr).di); - mem_free ((void **) &(**gr).syntax_symbol); - mem_free ((void **) &(**gr).string_symbol); - map_str_destroy (&(**gr).maps); - map_byte_destroy (&(**gr).mapb); - map_rule_destroy (&(**gr).mapr); - mem_free ((void **) gr); - } -} - - -static void error_msg(int line, const char *msg) -{ - fprintf(stderr, "Error in grammar_load_from_text() at line %d: %s\n", line, msg); -} - - -/* - the API -*/ -grammar grammar_load_from_text (const byte *text) -{ - grammar_load_state *g = NULL; - grammar id = 0; - - clear_last_error (); - - grammar_load_state_create (&g); - if (g == NULL) { - error_msg(__LINE__, ""); - return 0; - } - - dict_create (&g->di); - if (g->di == NULL) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - eat_spaces (&text); - - /* skip ".syntax" keyword */ - text += 7; - eat_spaces (&text); - - /* retrieve root symbol */ - if (get_identifier (&text, &g->syntax_symbol)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - eat_spaces (&text); - - /* skip semicolon */ - text++; - eat_spaces (&text); - - while (*text) - { - byte *symbol = NULL; - int is_dot = *text == '.'; - - if (is_dot) - text++; - - if (get_identifier (&text, &symbol)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - eat_spaces (&text); - - /* .emtcode */ - if (is_dot && str_equal (symbol, (byte *) "emtcode")) - { - map_byte *ma = NULL; - - mem_free ((void **) (void *) &symbol); - - if (get_emtcode (&text, &ma)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - map_byte_append (&g->mapb, ma); - } - /* .regbyte */ - else if (is_dot && str_equal (symbol, (byte *) "regbyte")) - { - map_byte *ma = NULL; - - mem_free ((void **) (void *) &symbol); - - if (get_regbyte (&text, &ma)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - map_byte_append (&g->di->m_regbytes, ma); - } - /* .errtext */ - else if (is_dot && str_equal (symbol, (byte *) "errtext")) - { - map_str *ma = NULL; - - mem_free ((void **) (void *) &symbol); - - if (get_errtext (&text, &ma)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - map_str_append (&g->maps, ma); - } - /* .string */ - else if (is_dot && str_equal (symbol, (byte *) "string")) - { - mem_free ((void **) (void *) &symbol); - - if (g->di->m_string != NULL) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - if (get_identifier (&text, &g->string_symbol)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - /* skip semicolon */ - eat_spaces (&text); - text++; - eat_spaces (&text); - } - else - { - rule *ru = NULL; - map_rule *ma = NULL; - - if (get_rule (&text, &ru, g->maps, g->mapb)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - rule_append (&g->di->m_rulez, ru); - - /* if a rule consist of only one specifier, give it an ".and" operator */ - if (ru->m_oper == op_none) - ru->m_oper = op_and; - - map_rule_create (&ma); - if (ma == NULL) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, ""); - return 0; - } - - ma->key = symbol; - ma->data = ru; - map_rule_append (&g->mapr, ma); - } - } - - if (update_dependencies (g->di, g->mapr, &g->syntax_symbol, &g->string_symbol, - g->di->m_regbytes)) - { - grammar_load_state_destroy (&g); - error_msg(__LINE__, "update_dependencies() failed"); - return 0; - } - - dict_append (&g_dicts, g->di); - id = g->di->m_id; - g->di = NULL; - - grammar_load_state_destroy (&g); - - return id; -} - -int grammar_set_reg8 (grammar id, const byte *name, byte value) -{ - dict *di = NULL; - map_byte *reg = NULL; - - clear_last_error (); - - dict_find (&g_dicts, id, &di); - if (di == NULL) - { - set_last_error (INVALID_GRAMMAR_ID, NULL, -1); - return 0; - } - - reg = map_byte_locate (&di->m_regbytes, name); - if (reg == NULL) - { - set_last_error (INVALID_REGISTER_NAME, str_duplicate (name), -1); - return 0; - } - - reg->data = value; - return 1; -} - -/* - internal checking function used by both grammar_check and grammar_fast_check functions -*/ -static int _grammar_check (grammar id, const byte *text, byte **prod, unsigned int *size, - unsigned int estimate_prod_size, int use_fast_path) -{ - dict *di = NULL; - int index = 0; - - clear_last_error (); - - dict_find (&g_dicts, id, &di); - if (di == NULL) - { - set_last_error (INVALID_GRAMMAR_ID, NULL, -1); - return 0; - } - - *prod = NULL; - *size = 0; - - if (use_fast_path) - { - regbyte_ctx *rbc = NULL; - bytepool *bp = NULL; - int _P = 0; - - bytepool_create (&bp, estimate_prod_size); - if (bp == NULL) - return 0; - - if (fast_match (di, text, &index, di->m_syntax, &_P, bp, 0, &rbc) != mr_matched) - { - bytepool_destroy (&bp); - free_regbyte_ctx_stack (rbc, NULL); - return 0; - } - - free_regbyte_ctx_stack (rbc, NULL); - - *prod = bp->_F; - *size = _P; - bp->_F = NULL; - bytepool_destroy (&bp); - } - else - { - regbyte_ctx *rbc = NULL; - barray *ba = NULL; - - barray_create (&ba); - if (ba == NULL) - return 0; - - if (match (di, text, &index, di->m_syntax, &ba, 0, &rbc) != mr_matched) - { - barray_destroy (&ba); - free_regbyte_ctx_stack (rbc, NULL); - return 0; - } - - free_regbyte_ctx_stack (rbc, NULL); - - *prod = (byte *) mem_alloc (ba->len * sizeof (byte)); - if (*prod == NULL) - { - barray_destroy (&ba); - return 0; - } - - mem_copy (*prod, ba->data, ba->len * sizeof (byte)); - *size = ba->len; - barray_destroy (&ba); - } - - return 1; -} - -int grammar_check (grammar id, const byte *text, byte **prod, unsigned int *size) -{ - return _grammar_check (id, text, prod, size, 0, 0); -} - -int grammar_fast_check (grammar id, const byte *text, byte **prod, unsigned int *size, - unsigned int estimate_prod_size) -{ - return _grammar_check (id, text, prod, size, estimate_prod_size, 1); -} - -int grammar_destroy (grammar id) -{ - dict **di = &g_dicts; - - clear_last_error (); - - while (*di != NULL) - { - if ((**di).m_id == id) - { - dict *tmp = *di; - *di = (**di).next; - dict_destroy (&tmp); - return 1; - } - - di = &(**di).next; - } - - set_last_error (INVALID_GRAMMAR_ID, NULL, -1); - return 0; -} - -static void append_character (const char x, byte *text, int *dots_made, int *len, int size) -{ - if (*dots_made == 0) - { - if (*len < size - 1) - { - text[(*len)++] = x; - text[*len] = '\0'; - } - else - { - int i; - for (i = 0; i < 3; i++) - if (--(*len) >= 0) - text[*len] = '.'; - *dots_made = 1; - } - } -} - -void grammar_get_last_error (byte *text, unsigned int size, int *pos) -{ - int len = 0, dots_made = 0; - const byte *p = error_message; - - *text = '\0'; - - if (p) - { - while (*p) - { - if (*p == '$') - { - const byte *r = error_param; - - while (*r) - { - append_character (*r++, text, &dots_made, &len, (int) size); - } - - p++; - } - else - { - append_character (*p++, text, &dots_made, &len, size); - } - } - } - - *pos = error_position; -} diff --git a/src/mesa/shader/grammar/grammar.h b/src/mesa/shader/grammar/grammar.h deleted file mode 100644 index 591e38aefa6..00000000000 --- a/src/mesa/shader/grammar/grammar.h +++ /dev/null @@ -1,103 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.2 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef GRAMMAR_H -#define GRAMMAR_H - - -#ifndef GRAMMAR_PORT_INCLUDE -#error Do not include this file directly, include your grammar_XXX.h instead -#endif - - -#ifdef __cplusplus -extern "C" { -#endif - -void grammar_alloc_free (void *); -void *grammar_alloc_malloc (size_t); -void *grammar_alloc_realloc (void *, size_t, size_t); -void *grammar_memory_copy (void *, const void *, size_t); -int grammar_string_compare (const byte *, const byte *); -int grammar_string_compare_n (const byte *, const byte *, size_t); -byte *grammar_string_copy (byte *, const byte *); -byte *grammar_string_copy_n (byte *, const byte *, size_t); -byte *grammar_string_duplicate (const byte *); -unsigned int grammar_string_length (const byte *); - -/* - loads grammar script from null-terminated ASCII <text> - returns unique grammar id to grammar object - returns 0 if an error occurs (call grammar_get_last_error to retrieve the error text) -*/ -grammar grammar_load_from_text (const byte *text); - -/* - sets a new <value> to a register <name> for grammar <id> - returns 0 on error (call grammar_get_last_error to retrieve the error text) - returns 1 on success -*/ -int grammar_set_reg8 (grammar id, const byte *name, byte value); - -/* - this function is obsolete, use only for debugging purposes - - checks if a null-terminated <text> matches given grammar <id> - returns 0 on error (call grammar_get_last_error to retrieve the error text) - returns 1 on success, the <prod> points to newly allocated buffer with production and <size> - is filled with the production size - call grammar_alloc_free to free the memory block pointed by <prod> -*/ -int grammar_check (grammar id, const byte *text, byte **prod, unsigned int *size); - -/* - does the same what grammar_check does but much more (approx. 4 times) faster - use this function instead of grammar_check - <estimate_prod_size> is a hint - the initial production buffer size will be of this size, - but if more room is needed it will be safely resized; set it to 0x1000 or so -*/ -int grammar_fast_check (grammar id, const byte *text, byte **prod, unsigned int *size, - unsigned int estimate_prod_size); - -/* - destroys grammar object identified by <id> - returns 0 on error (call grammar_get_last_error to retrieve the error text) - returns 1 on success -*/ -int grammar_destroy (grammar id); - -/* - retrieves last grammar error reported either by grammar_load_from_text, grammar_check - or grammar_destroy - the user allocated <text> buffer receives error description, <pos> points to error position, - <size> is the size of the text buffer to fill in - it must be at least 4 bytes long, -*/ -void grammar_get_last_error (byte *text, unsigned int size, int *pos); - -#ifdef __cplusplus -} -#endif - -#endif - diff --git a/src/mesa/shader/grammar/grammar.syn b/src/mesa/shader/grammar/grammar.syn deleted file mode 100644 index 5d99f65bfce..00000000000 --- a/src/mesa/shader/grammar/grammar.syn +++ /dev/null @@ -1,567 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.2 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - /** - * \file grammar.syn - * syntax of .syn script - used to validate other syntax files - * \author Michal Krol - */ - -.syntax grammar; - -/* declaration */ -.emtcode DECLARATION_END 0 -.emtcode DECLARATION_EMITCODE 1 -.emtcode DECLARATION_ERRORTEXT 2 -.emtcode DECLARATION_REGBYTE 3 -.emtcode DECLARATION_LEXER 4 -.emtcode DECLARATION_RULE 5 - -/* specifier */ -.emtcode SPECIFIER_END 0 -.emtcode SPECIFIER_AND_TAG 1 -.emtcode SPECIFIER_OR_TAG 2 -.emtcode SPECIFIER_CHARACTER_RANGE 3 -.emtcode SPECIFIER_CHARACTER 4 -.emtcode SPECIFIER_STRING 5 -.emtcode SPECIFIER_IDENTIFIER 6 -.emtcode SPECIFIER_TRUE 7 -.emtcode SPECIFIER_FALSE 8 -.emtcode SPECIFIER_DEBUG 9 - -/* identifier */ -.emtcode IDENTIFIER_SIMPLE 0 -.emtcode IDENTIFIER_LOOP 1 - -/* error */ -.emtcode ERROR_NOT_PRESENT 0 -.emtcode ERROR_PRESENT 1 - -/* emit */ -.emtcode EMIT_NULL 0 -.emtcode EMIT_INTEGER 1 -.emtcode EMIT_IDENTIFIER 2 -.emtcode EMIT_CHARACTER 3 -.emtcode EMIT_LAST_CHARACTER 4 -.emtcode EMIT_CURRENT_POSITION 5 - -.errtext INVALID_GRAMMAR "internal error 2001: invalid grammar script" -.errtext SYNTAX_EXPECTED "internal error 2002: '.syntax' keyword expected" -.errtext IDENTIFIER_EXPECTED "internal error 2003: identifier expected" -.errtext MISSING_SEMICOLON "internal error 2004: missing ';'" -.errtext INTEGER_EXPECTED "internal error 2005: integer value expected" -.errtext STRING_EXPECTED "internal error 2006: string expected" - -/* - <grammar> ::= ".syntax" <identifier> ";" <declaration_list> -*/ -grammar - grammar_1 .error INVALID_GRAMMAR; -grammar_1 - optional_space .and ".syntax" .error SYNTAX_EXPECTED .and space .and identifier .and - semicolon .and declaration_list .and optional_space .and '\0' .emit DECLARATION_END; - -/* - <optional_space> ::= <space> - | "" -*/ -optional_space - space .or .true; - -/* - <space> ::= <single_space> <single_space>* -*/ -space - single_space .and .loop single_space; - -/* - <single_space> ::= <white_char> - | <comment_block> -*/ -single_space - white_char .or comment_block; - -/* - <white_char> ::= " " - | "\t" - | "\n" - | "\r" -*/ -white_char - ' ' .or '\t' .or '\n' .or '\r'; - -/* - <comment_block> ::= "/" "*" <comment_rest> -*/ -comment_block - '/' .and '*' .and comment_rest; - -/* - <comment_rest> ::= <comment_char_no_star>* <comment_end> - | <comment_char_no_star>* "*" <comment_rest> -*/ -comment_rest - .loop comment_char_no_star .and comment_rest_1; -comment_rest_1 - comment_end .or comment_rest_2; -comment_rest_2 - '*' .and comment_rest; - -/* - <comment_char_no_star> ::= All ASCII characters except "*" and "\0" -*/ -comment_char_no_star - '\x2B'-'\xFF' .or '\x01'-'\x29'; - -/* - <comment_end> ::= "*" "/" -*/ -comment_end - '*' .and '/'; - -/* - <identifier> ::= <identifier> -*/ -identifier - identifier_ne .error IDENTIFIER_EXPECTED; - -/* - <identifier_ne> ::= <first_idchar> <follow_idchar>* -*/ -identifier_ne - first_idchar .emit * .and .loop follow_idchar .emit * .and .true .emit '\0'; - -/* - <first_idchar> ::= "a"-"z" - | "A"-"Z" - | "_" -*/ -first_idchar - 'a'-'z' .or 'A'-'Z' .or '_'; - -/* - <follow_idchar> ::= <first_idchar> - | <digit_dec> -*/ -follow_idchar - first_idchar .or digit_dec; - -/* - <digit_dec> ::= "0"-"9" -*/ -digit_dec - '0'-'9'; - -/* - <semicolon> ::= ";" -*/ -semicolon - optional_space .and ';' .error MISSING_SEMICOLON .and optional_space; - -/* - <declaration_list> ::= <declaration> - | <declaration_list> <declaration> -*/ -declaration_list - declaration .and .loop declaration; - -/* - <declaration> ::= <emitcode_definition> - | <errortext_definition> - | <lexer_definition> - | <rule_definition> -*/ -declaration - emitcode_definition .emit DECLARATION_EMITCODE .or - errortext_definition .emit DECLARATION_ERRORTEXT .or - regbyte_definition .emit DECLARATION_REGBYTE .or - lexer_definition .emit DECLARATION_LEXER .or - rule_definition .emit DECLARATION_RULE; - -/* - <emitcode_definition> ::= ".emtcode" <identifier> <integer> -*/ -emitcode_definition - ".emtcode" .and space .and identifier .and space .and integer .and space_or_null; - -/* - <integer> ::= <integer_ne> -*/ -integer - integer_ne .error INTEGER_EXPECTED; - -/* - <integer_ne> ::= <hex_integer> - | <dec_integer> -*/ -integer_ne - hex_integer .emit 0x10 .or dec_integer .emit 10; - -/* - <hex_integer> :: <hex_prefix> <digit_hex> <digit_hex>* -*/ -hex_integer - hex_prefix .and digit_hex .emit * .and .loop digit_hex .emit * .and .true .emit '\0'; - -/* - <hex_prefix> ::= "0x" - | "0X" -*/ -hex_prefix - '0' .and hex_prefix_1; -hex_prefix_1 - 'x' .or 'X'; - -/* - <digit_hex> ::= "0"-"9" - | "a"-"f" - | "A"-"F" -*/ -digit_hex - '0'-'9' .or 'a'-'f' .or 'A'-'F'; - -/* - <dec_integer> :: <digit_dec> <digit_dec>* -*/ -dec_integer - digit_dec .emit * .and .loop digit_dec .emit * .and .true .emit '\0'; - -/* - <space_or_null> ::= <space> - | "\0" -*/ -space_or_null - space .or '\0'; - -/* - <errortext_definition> ::= ".errtext" <identifier> <string> -*/ -errortext_definition - ".errtext" .and space .and identifier .and space .and string .and space_or_null; - -/* - <string> ::= <string_ne> -*/ -string - string_ne .error STRING_EXPECTED; - -/* - <string_ne> ::= "\"" <string_char_double_quotes> "\"" -*/ -string_ne - '"' .and .loop string_char_double_quotes .and '"' .emit '\0'; - -/* - <string_char_double_quotes> ::= <escape_sequence> - | <string_char> - | "\'" -*/ -string_char_double_quotes - escape_sequence .or string_char .emit * .or '\'' .emit *; - -/* - <string_char> ::= All ASCII characters except "\'", "\"", "\n", "\r", - "\0" and "\\" -*/ -string_char - '\x5D'-'\xFF' .or '\x28'-'\x5B' .or '\x23'-'\x26' .or '\x0E'-'\x21' .or '\x0B'-'\x0C' .or - '\x01'-'\x09'; - -/* - <escape_sequence> ::= "\\" <escape_code> -*/ -escape_sequence - '\\' .emit * .and escape_code; - -/* - <escape_code> ::= <simple_escape_code> - | <hex_escape_code> - | <oct_escape_code> -*/ -escape_code - simple_escape_code .emit * .or hex_escape_code .or oct_escape_code; - -/* - <simple_escape_code> ::= "\'" - | "\"" - | "?" - | "\\" - | "a" - | "b" - | "f" - | "n" - | "r" - | "t" - | "v" -*/ -simple_escape_code - '\'' .or '"' .or '?' .or '\\' .or 'a' .or 'b' .or 'f' .or 'n' .or 'r' .or 't' .or 'v'; - -/* - <hex_escape_code> ::= "x" <digit_hex> <digit_hex>* -*/ -hex_escape_code - 'x' .emit * .and digit_hex .emit * .and .loop digit_hex .emit *; - -/* - <oct_escape_code> ::= <digit_oct> <optional_digit_oct> <optional_digit_oct> -*/ -oct_escape_code - digit_oct .emit * .and optional_digit_oct .and optional_digit_oct; - -/* - <digit_oct> ::= "0"-"7" -*/ -digit_oct - '0'-'7'; - -/* - <optional_digit_oct> ::= <digit_oct> - | "" -*/ -optional_digit_oct - digit_oct .emit * .or .true; - -/* - <regbyte_definition> ::= ".regbyte" <identifier> <integer> -*/ -regbyte_definition - ".regbyte" .and space .and identifier .and space .and integer .and space_or_null; - -/* - <lexer_definition> ::= ".string" <identifier> ";" -*/ -lexer_definition - ".string" .and space .and identifier .and semicolon; - -/* - <rule_definition> ::= <identifier_ne> <definition> -*/ -rule_definition - identifier_ne .and space .and definition; - -/* - <definition> ::= <specifier> <optional_specifiers_and_or> ";" -*/ -definition - specifier .and optional_specifiers_and_or .and semicolon .emit SPECIFIER_END; - -/* - <optional_specifiers_and_or> ::= <and_specifiers> - | <or_specifiers> - | "" -*/ -optional_specifiers_and_or - and_specifiers .emit SPECIFIER_AND_TAG .or or_specifiers .emit SPECIFIER_OR_TAG .or .true; - -/* - <specifier> ::= <specifier_condition> <specifier_rule> -*/ -specifier - specifier_condition .and optional_space .and specifier_rule; - -/* - <specifier_condition> ::= ".if" "(" <left_operand> <operator> <right_operand> ")" -*/ -specifier_condition - specifier_condition_1 .or .true; -specifier_condition_1 - ".if" .and optional_space .and '(' .and optional_space .and left_operand .and operator .and - right_operand .and optional_space .and ')'; - -/* - <left_operand> ::= <identifier> -*/ -left_operand - identifier; - -/* - <operator> ::= "!=" - | "==" -*/ -operator - operator_1 .or operator_2; -operator_1 - optional_space .and '!' .and '=' .and optional_space; -operator_2 - optional_space .and '=' .and '=' .and optional_space; - -/* - <right_operand> ::= <integer> -*/ -right_operand - integer; - -/* - <specifier_rule> ::= <character_range> <optional_error> <emit>* - | <character> <optional_error> <emit>* - | <string> <optional_error> <emit>* - | ".true" <optional_error> <emit>* - | ".false" <optional_error> <emit>* - | ".debug" <optional_error> <emit>* - | <advanced_identifier> <optional_error> <emit>* -*/ -specifier_rule - specifier_rule_1 .and optional_error .and .loop emit .and .true .emit EMIT_NULL; -specifier_rule_1 - character_range .emit SPECIFIER_CHARACTER_RANGE .or - character .emit SPECIFIER_CHARACTER .or - string_ne .emit SPECIFIER_STRING .or - ".true" .emit SPECIFIER_TRUE .or - ".false" .emit SPECIFIER_FALSE .or - ".debug" .emit SPECIFIER_DEBUG .or - advanced_identifier .emit SPECIFIER_IDENTIFIER; - -/* - <character> ::= "\'" <string_char_single_quotes "\'" -*/ -character - '\'' .and string_char_single_quotes .and '\'' .emit '\0'; - -/* - <string_char_single_quotes> ::= <escape_sequence> - | <string_char> - | "\"" -*/ -string_char_single_quotes - escape_sequence .or string_char .emit * .or '"' .emit *; - -/* - <character_range> ::= <character> "-" <character> -*/ -character_range - character .and optional_space .and '-' .and optional_space .and character; - -/* - <advanced_identifier> ::= <optional_loop> <identifier> -*/ -advanced_identifier - optional_loop .and identifier; - -/* - <optional_loop> ::= ".loop" - | "" -*/ -optional_loop - optional_loop_1 .emit IDENTIFIER_LOOP .or .true .emit IDENTIFIER_SIMPLE; -optional_loop_1 - ".loop" .and space; - -/* - <optional_error> ::= <error> - | "" -*/ -optional_error - error .emit ERROR_PRESENT .or .true .emit ERROR_NOT_PRESENT; - -/* - <error> :: ".error" <identifier> -*/ -error - space .and ".error" .and space .and identifier; - -/* - <emit> ::= <emit_output> - | <emit_regbyte> -*/ -emit - emit_output .or emit_regbyte; - -/* - <emit_output> ::= ".emit" <emit_param> -*/ -emit_output - space .and ".emit" .and space .and emit_param; - -/* - <emit_param> ::= <integer> - | <identifier> - | <character> - | "*" - | "$" -*/ -emit_param - integer_ne .emit EMIT_INTEGER .or - identifier_ne .emit EMIT_IDENTIFIER .or - character .emit EMIT_CHARACTER .or - '*' .emit EMIT_LAST_CHARACTER .or - '$' .emit EMIT_CURRENT_POSITION; - -/* - <emit_regbyte> ::= ".load" <identifier> <emit_param> -*/ -emit_regbyte - space .and ".load" .and space .and identifier .and space .and emit_param; - -/* - <and_specifiers> ::= <and_specifier> <and_specifier>* -*/ -and_specifiers - and_specifier .and .loop and_specifier; - -/* - <or_specifiers> ::= <or_specifier> <or_specifier>* -*/ -or_specifiers - or_specifier .and .loop or_specifier; - -/* - <and_specifier> ::= ".and" <specifier> -*/ -and_specifier - space .and ".and" .and space .and specifier; - -/* - <or_specifier> ::= ".or" <specifier> -*/ -or_specifier - space .and ".or" .and space .and specifier; - - -.string __string_filter; - -/* - <__string_filter> ::= <__first_identifier_char> <__next_identifier_char>* -*/ -__string_filter - __first_identifier_char .and .loop __next_identifier_char; - -/* - <__first_identifier_char> ::= "a"-"z" - | "A"-"Z" - | "_" - | "." -*/ -__first_identifier_char - 'a'-'z' .or 'A'-'Z' .or '_' .or '.'; - -/* - <__next_identifier_char> ::= "a"-"z" - | "A"-"Z" - | "_" - | "0"-"9" -*/ -__next_identifier_char - 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9'; - diff --git a/src/mesa/shader/grammar/grammar_crt.c b/src/mesa/shader/grammar/grammar_crt.c deleted file mode 100644 index d2c95d1c8e7..00000000000 --- a/src/mesa/shader/grammar/grammar_crt.c +++ /dev/null @@ -1,64 +0,0 @@ -#include "grammar_crt.h" - -#define GRAMMAR_PORT_BUILD 1 -#include "grammar.c" -#undef GRAMMAR_PORT_BUILD - - -void grammar_alloc_free (void *ptr) -{ - free (ptr); -} - -void *grammar_alloc_malloc (size_t size) -{ - return malloc (size); -} - -void *grammar_alloc_realloc (void *ptr, size_t old_size, size_t size) -{ - return realloc (ptr, size); -} - -void *grammar_memory_copy (void *dst, const void * src, size_t size) -{ - return memcpy (dst, src, size); -} - -int grammar_string_compare (const byte *str1, const byte *str2) -{ - return strcmp ((const char *) str1, (const char *) str2); -} - -int grammar_string_compare_n (const byte *str1, const byte *str2, size_t n) -{ - return strncmp ((const char *) str1, (const char *) str2, n); -} - -byte *grammar_string_copy (byte *dst, const byte *src) -{ - return (byte *) strcpy ((char *) dst, (const char *) src); -} - -byte *grammar_string_copy_n (byte *dst, const byte *src, size_t n) -{ - return (byte *) strncpy ((char *) dst, (const char *) src, n); -} - -unsigned int grammar_string_length (const byte *str) -{ - return strlen ((const char *) str); -} - -byte *grammar_string_duplicate (const byte *src) -{ - const unsigned int size = grammar_string_length (src); - byte *str = grammar_alloc_malloc (size + 1); - if (str != NULL) - { - grammar_memory_copy (str, src, size); - str[size] = '\0'; - } - return str; -} - diff --git a/src/mesa/shader/grammar/grammar_crt.h b/src/mesa/shader/grammar/grammar_crt.h deleted file mode 100644 index 492711e96ae..00000000000 --- a/src/mesa/shader/grammar/grammar_crt.h +++ /dev/null @@ -1,20 +0,0 @@ -#ifndef GRAMMAR_CRT_H -#define GRAMMAR_CRT_H - - -#include <stdlib.h> -#include <malloc.h> -#include <string.h> - - -typedef unsigned long grammar; -typedef unsigned char byte; - - -#define GRAMMAR_PORT_INCLUDE 1 -#include "grammar.h" -#undef GRAMMAR_PORT_INCLUDE - - -#endif - diff --git a/src/mesa/shader/grammar/grammar_mesa.c b/src/mesa/shader/grammar/grammar_mesa.c deleted file mode 100644 index eb962505bfe..00000000000 --- a/src/mesa/shader/grammar/grammar_mesa.c +++ /dev/null @@ -1,87 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.1 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file grammar_mesa.c - * mesa3d port to syntax parsing engine - * \author Michal Krol - */ - -#include "grammar_mesa.h" - -#define GRAMMAR_PORT_BUILD 1 -#include "grammar.c" -#undef GRAMMAR_PORT_BUILD - - -void grammar_alloc_free (void *ptr) -{ - _mesa_free (ptr); -} - -void *grammar_alloc_malloc (size_t size) -{ - return _mesa_malloc (size); -} - -void *grammar_alloc_realloc (void *ptr, size_t old_size, size_t size) -{ - return _mesa_realloc (ptr, old_size, size); -} - -void *grammar_memory_copy (void *dst, const void * src, size_t size) -{ - return _mesa_memcpy (dst, src, size); -} - -int grammar_string_compare (const byte *str1, const byte *str2) -{ - return _mesa_strcmp ((const char *) str1, (const char *) str2); -} - -int grammar_string_compare_n (const byte *str1, const byte *str2, size_t n) -{ - return _mesa_strncmp ((const char *) str1, (const char *) str2, n); -} - -byte *grammar_string_copy (byte *dst, const byte *src) -{ - return (byte *) _mesa_strcpy ((char *) dst, (const char *) src); -} - -byte *grammar_string_copy_n (byte *dst, const byte *src, size_t n) -{ - return (byte *) _mesa_strncpy ((char *) dst, (const char *) src, n); -} - -byte *grammar_string_duplicate (const byte *src) -{ - return (byte *) _mesa_strdup ((const char *) src); -} - -unsigned int grammar_string_length (const byte *str) -{ - return (unsigned int)_mesa_strlen ((const char *) str); -} - diff --git a/src/mesa/shader/grammar/grammar_mesa.h b/src/mesa/shader/grammar/grammar_mesa.h deleted file mode 100644 index 6c92c5812da..00000000000 --- a/src/mesa/shader/grammar/grammar_mesa.h +++ /dev/null @@ -1,43 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.1 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef GRAMMAR_MESA_H -#define GRAMMAR_MESA_H - - -#include "main/imports.h" -/* NOTE: include Mesa 3-D specific headers here */ - - -typedef GLuint grammar; -typedef GLubyte byte; - - -#define GRAMMAR_PORT_INCLUDE 1 -#include "grammar.h" -#undef GRAMMAR_PORT_INCLUDE - - -#endif - diff --git a/src/mesa/shader/grammar/grammar_syn.h b/src/mesa/shader/grammar/grammar_syn.h deleted file mode 100644 index 840a1ab62c2..00000000000 --- a/src/mesa/shader/grammar/grammar_syn.h +++ /dev/null @@ -1,202 +0,0 @@ -".syntax grammar;\n" -".emtcode DECLARATION_END 0\n" -".emtcode DECLARATION_EMITCODE 1\n" -".emtcode DECLARATION_ERRORTEXT 2\n" -".emtcode DECLARATION_REGBYTE 3\n" -".emtcode DECLARATION_LEXER 4\n" -".emtcode DECLARATION_RULE 5\n" -".emtcode SPECIFIER_END 0\n" -".emtcode SPECIFIER_AND_TAG 1\n" -".emtcode SPECIFIER_OR_TAG 2\n" -".emtcode SPECIFIER_CHARACTER_RANGE 3\n" -".emtcode SPECIFIER_CHARACTER 4\n" -".emtcode SPECIFIER_STRING 5\n" -".emtcode SPECIFIER_IDENTIFIER 6\n" -".emtcode SPECIFIER_TRUE 7\n" -".emtcode SPECIFIER_FALSE 8\n" -".emtcode SPECIFIER_DEBUG 9\n" -".emtcode IDENTIFIER_SIMPLE 0\n" -".emtcode IDENTIFIER_LOOP 1\n" -".emtcode ERROR_NOT_PRESENT 0\n" -".emtcode ERROR_PRESENT 1\n" -".emtcode EMIT_NULL 0\n" -".emtcode EMIT_INTEGER 1\n" -".emtcode EMIT_IDENTIFIER 2\n" -".emtcode EMIT_CHARACTER 3\n" -".emtcode EMIT_LAST_CHARACTER 4\n" -".emtcode EMIT_CURRENT_POSITION 5\n" -".errtext INVALID_GRAMMAR \"internal error 2001: invalid grammar script\"\n" -".errtext SYNTAX_EXPECTED \"internal error 2002: '.syntax' keyword expected\"\n" -".errtext IDENTIFIER_EXPECTED \"internal error 2003: identifier expected\"\n" -".errtext MISSING_SEMICOLON \"internal error 2004: missing ';'\"\n" -".errtext INTEGER_EXPECTED \"internal error 2005: integer value expected\"\n" -".errtext STRING_EXPECTED \"internal error 2006: string expected\"\n" -"grammar\n" -" grammar_1 .error INVALID_GRAMMAR;\n" -"grammar_1\n" -" optional_space .and \".syntax\" .error SYNTAX_EXPECTED .and space .and identifier .and\n" -" semicolon .and declaration_list .and optional_space .and '\\0' .emit DECLARATION_END;\n" -"optional_space\n" -" space .or .true;\n" -"space\n" -" single_space .and .loop single_space;\n" -"single_space\n" -" white_char .or comment_block;\n" -"white_char\n" -" ' ' .or '\\t' .or '\\n' .or '\\r';\n" -"comment_block\n" -" '/' .and '*' .and comment_rest;\n" -"comment_rest\n" -" .loop comment_char_no_star .and comment_rest_1;\n" -"comment_rest_1\n" -" comment_end .or comment_rest_2;\n" -"comment_rest_2\n" -" '*' .and comment_rest;\n" -"comment_char_no_star\n" -" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n" -"comment_end\n" -" '*' .and '/';\n" -"identifier\n" -" identifier_ne .error IDENTIFIER_EXPECTED;\n" -"identifier_ne\n" -" first_idchar .emit * .and .loop follow_idchar .emit * .and .true .emit '\\0';\n" -"first_idchar\n" -" 'a'-'z' .or 'A'-'Z' .or '_';\n" -"follow_idchar\n" -" first_idchar .or digit_dec;\n" -"digit_dec\n" -" '0'-'9';\n" -"semicolon\n" -" optional_space .and ';' .error MISSING_SEMICOLON .and optional_space;\n" -"declaration_list\n" -" declaration .and .loop declaration;\n" -"declaration\n" -" emitcode_definition .emit DECLARATION_EMITCODE .or\n" -" errortext_definition .emit DECLARATION_ERRORTEXT .or\n" -" regbyte_definition .emit DECLARATION_REGBYTE .or\n" -" lexer_definition .emit DECLARATION_LEXER .or\n" -" rule_definition .emit DECLARATION_RULE;\n" -"emitcode_definition\n" -" \".emtcode\" .and space .and identifier .and space .and integer .and space_or_null;\n" -"integer\n" -" integer_ne .error INTEGER_EXPECTED;\n" -"integer_ne\n" -" hex_integer .emit 0x10 .or dec_integer .emit 10;\n" -"hex_integer\n" -" hex_prefix .and digit_hex .emit * .and .loop digit_hex .emit * .and .true .emit '\\0';\n" -"hex_prefix\n" -" '0' .and hex_prefix_1;\n" -"hex_prefix_1\n" -" 'x' .or 'X';\n" -"digit_hex\n" -" '0'-'9' .or 'a'-'f' .or 'A'-'F';\n" -"dec_integer\n" -" digit_dec .emit * .and .loop digit_dec .emit * .and .true .emit '\\0';\n" -"space_or_null\n" -" space .or '\\0';\n" -"errortext_definition\n" -" \".errtext\" .and space .and identifier .and space .and string .and space_or_null;\n" -"string\n" -" string_ne .error STRING_EXPECTED;\n" -"string_ne\n" -" '\"' .and .loop string_char_double_quotes .and '\"' .emit '\\0';\n" -"string_char_double_quotes\n" -" escape_sequence .or string_char .emit * .or '\\'' .emit *;\n" -"string_char\n" -" '\\x5D'-'\\xFF' .or '\\x28'-'\\x5B' .or '\\x23'-'\\x26' .or '\\x0E'-'\\x21' .or '\\x0B'-'\\x0C' .or\n" -" '\\x01'-'\\x09';\n" -"escape_sequence\n" -" '\\\\' .emit * .and escape_code;\n" -"escape_code\n" -" simple_escape_code .emit * .or hex_escape_code .or oct_escape_code;\n" -"simple_escape_code\n" -" '\\'' .or '\"' .or '?' .or '\\\\' .or 'a' .or 'b' .or 'f' .or 'n' .or 'r' .or 't' .or 'v';\n" -"hex_escape_code\n" -" 'x' .emit * .and digit_hex .emit * .and .loop digit_hex .emit *;\n" -"oct_escape_code\n" -" digit_oct .emit * .and optional_digit_oct .and optional_digit_oct;\n" -"digit_oct\n" -" '0'-'7';\n" -"optional_digit_oct\n" -" digit_oct .emit * .or .true;\n" -"regbyte_definition\n" -" \".regbyte\" .and space .and identifier .and space .and integer .and space_or_null;\n" -"lexer_definition\n" -" \".string\" .and space .and identifier .and semicolon;\n" -"rule_definition\n" -" identifier_ne .and space .and definition;\n" -"definition\n" -" specifier .and optional_specifiers_and_or .and semicolon .emit SPECIFIER_END;\n" -"optional_specifiers_and_or\n" -" and_specifiers .emit SPECIFIER_AND_TAG .or or_specifiers .emit SPECIFIER_OR_TAG .or .true;\n" -"specifier\n" -" specifier_condition .and optional_space .and specifier_rule;\n" -"specifier_condition\n" -" specifier_condition_1 .or .true;\n" -"specifier_condition_1\n" -" \".if\" .and optional_space .and '(' .and optional_space .and left_operand .and operator .and\n" -" right_operand .and optional_space .and ')';\n" -"left_operand\n" -" identifier;\n" -"operator\n" -" operator_1 .or operator_2;\n" -"operator_1\n" -" optional_space .and '!' .and '=' .and optional_space;\n" -"operator_2\n" -" optional_space .and '=' .and '=' .and optional_space;\n" -"right_operand\n" -" integer;\n" -"specifier_rule\n" -" specifier_rule_1 .and optional_error .and .loop emit .and .true .emit EMIT_NULL;\n" -"specifier_rule_1\n" -" character_range .emit SPECIFIER_CHARACTER_RANGE .or\n" -" character .emit SPECIFIER_CHARACTER .or\n" -" string_ne .emit SPECIFIER_STRING .or\n" -" \".true\" .emit SPECIFIER_TRUE .or\n" -" \".false\" .emit SPECIFIER_FALSE .or\n" -" \".debug\" .emit SPECIFIER_DEBUG .or\n" -" advanced_identifier .emit SPECIFIER_IDENTIFIER;\n" -"character\n" -" '\\'' .and string_char_single_quotes .and '\\'' .emit '\\0';\n" -"string_char_single_quotes\n" -" escape_sequence .or string_char .emit * .or '\"' .emit *;\n" -"character_range\n" -" character .and optional_space .and '-' .and optional_space .and character;\n" -"advanced_identifier\n" -" optional_loop .and identifier;\n" -"optional_loop\n" -" optional_loop_1 .emit IDENTIFIER_LOOP .or .true .emit IDENTIFIER_SIMPLE;\n" -"optional_loop_1\n" -" \".loop\" .and space;\n" -"optional_error\n" -" error .emit ERROR_PRESENT .or .true .emit ERROR_NOT_PRESENT;\n" -"error\n" -" space .and \".error\" .and space .and identifier;\n" -"emit\n" -" emit_output .or emit_regbyte;\n" -"emit_output\n" -" space .and \".emit\" .and space .and emit_param;\n" -"emit_param\n" -" integer_ne .emit EMIT_INTEGER .or\n" -" identifier_ne .emit EMIT_IDENTIFIER .or\n" -" character .emit EMIT_CHARACTER .or\n" -" '*' .emit EMIT_LAST_CHARACTER .or\n" -" '$' .emit EMIT_CURRENT_POSITION;\n" -"emit_regbyte\n" -" space .and \".load\" .and space .and identifier .and space .and emit_param;\n" -"and_specifiers\n" -" and_specifier .and .loop and_specifier;\n" -"or_specifiers\n" -" or_specifier .and .loop or_specifier;\n" -"and_specifier\n" -" space .and \".and\" .and space .and specifier;\n" -"or_specifier\n" -" space .and \".or\" .and space .and specifier;\n" -".string __string_filter;\n" -"__string_filter\n" -" __first_identifier_char .and .loop __next_identifier_char;\n" -"__first_identifier_char\n" -" 'a'-'z' .or 'A'-'Z' .or '_' .or '.';\n" -"__next_identifier_char\n" -" 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9';\n" -"" diff --git a/src/mesa/shader/nvfragparse.c b/src/mesa/shader/nvfragparse.c index 0fd55524abf..b739a6aa07c 100644 --- a/src/mesa/shader/nvfragparse.c +++ b/src/mesa/shader/nvfragparse.c @@ -217,6 +217,12 @@ MatchInstruction(const GLubyte *token) const struct instruction_pattern *inst; struct instruction_pattern result; + result.name = NULL; + result.opcode = MAX_OPCODE; /* i.e. invalid instruction */ + result.inputs = 0; + result.outputs = 0; + result.suffixes = 0; + for (inst = Instructions; inst->name; inst++) { if (_mesa_strncmp((const char *) token, inst->name, 3) == 0) { /* matched! */ @@ -247,7 +253,7 @@ MatchInstruction(const GLubyte *token) return result; } } - result.opcode = MAX_OPCODE; /* i.e. invalid instruction */ + return result; } diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 2f029b02e50..f22492e029e 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -500,7 +500,7 @@ GLfloat * _mesa_lookup_parameter_value(const struct gl_program_parameter_list *paramList, GLsizei nameLen, const char *name) { - GLuint i = _mesa_lookup_parameter_index(paramList, nameLen, name); + GLint i = _mesa_lookup_parameter_index(paramList, nameLen, name); if (i < 0) return NULL; else diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 532adf4d360..6b8d94e6614 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -806,9 +806,17 @@ _mesa_find_free_register(const struct gl_program *prog, GLuint regFile) const struct prog_instruction *inst = prog->Instructions + i; const GLuint n = _mesa_num_inst_src_regs(inst->Opcode); - for (k = 0; k < n; k++) { - if (inst->SrcReg[k].File == regFile) { - used[inst->SrcReg[k].Index] = GL_TRUE; + /* check dst reg first */ + if (inst->DstReg.File == regFile) { + used[inst->DstReg.Index] = GL_TRUE; + } + else { + /* check src regs otherwise */ + for (k = 0; k < n; k++) { + if (inst->SrcReg[k].File == regFile) { + used[inst->SrcReg[k].Index] = GL_TRUE; + break; + } } } } diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index a0daac1b806..9514545709d 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -528,15 +528,11 @@ _mesa_remove_output_reads(struct gl_program *prog, gl_register_file type) /* look for instructions which write to the varying vars identified above */ for (i = 0; i < prog->NumInstructions; i++) { struct prog_instruction *inst = prog->Instructions + i; - const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); - GLuint j; - for (j = 0; j < numSrc; j++) { - if (inst->DstReg.File == type && - outputMap[inst->DstReg.Index] >= 0) { - /* change inst to write to the temp reg, instead of the varying */ - inst->DstReg.File = PROGRAM_TEMPORARY; - inst->DstReg.Index = outputMap[inst->DstReg.Index]; - } + if (inst->DstReg.File == type && + outputMap[inst->DstReg.Index] >= 0) { + /* change inst to write to the temp reg, instead of the varying */ + inst->DstReg.File = PROGRAM_TEMPORARY; + inst->DstReg.Index = outputMap[inst->DstReg.Index]; } } diff --git a/src/mesa/shader/slang/descrip.mms b/src/mesa/shader/slang/descrip.mms index 6eefbcf5bdd..674b786ac08 100644 --- a/src/mesa/shader/slang/descrip.mms +++ b/src/mesa/shader/slang/descrip.mms @@ -17,12 +17,12 @@ VPATH = RCS -INCDIR = [----.include],[--.main],[--.glapi],[-.slang],[-.grammar],[-] +INCDIR = [----.include],[--.main],[--.glapi],[-.slang],[-] LIBDIR = [----.lib] CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm SOURCES = \ - slang_compile.c,slang_preprocess.c + slang_compile.c OBJECTS = slang_builtin.obj,slang_codegen.obj,slang_compile.obj,\ slang_compile_function.obj,slang_compile_operation.obj,\ @@ -59,7 +59,6 @@ slang_library_noise.obj : slang_library_noise.c slang_link.obj : slang_link.c slang_log.obj : slang_log.c slang_mem.obj : slang_mem.c -slang_preprocess.obj : slang_preprocess.c slang_print.obj : slang_print.c slang_simplify.obj : slang_simplify.c slang_storage.obj : slang_storage.c diff --git a/src/mesa/shader/slang/library/.gitignore b/src/mesa/shader/slang/library/.gitignore new file mode 100644 index 00000000000..02a89fc7df7 --- /dev/null +++ b/src/mesa/shader/slang/library/.gitignore @@ -0,0 +1 @@ +*_gc.h diff --git a/src/mesa/shader/slang/library/Makefile b/src/mesa/shader/slang/library/Makefile index 0e03fac2ee1..c6964512bfe 100644 --- a/src/mesa/shader/slang/library/Makefile +++ b/src/mesa/shader/slang/library/Makefile @@ -4,9 +4,7 @@ TOP = ../../../../.. include $(TOP)/configs/current -INCDIR = $(TOP)/include - -LIB_DEP = $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) +GLSL_CL = $(TOP)/src/glsl/apps/compile # # targets @@ -14,42 +12,14 @@ LIB_DEP = $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) .PHONY: default clean -default: syntax builtin +default: builtin clean: - -rm -f syn_to_c gc_to_bin *_syn.h *_gc.h - -syntax: slang_pp_directives_syn.h slang_pp_expression_syn.h slang_shader_syn.h slang_pp_version_syn.h + -rm -f *_gc.h builtin: builtin_110 builtin_120 # -# executables -# - -syn_to_c: syn_to_c.c - $(CC) syn_to_c.c -o syn_to_c - -gc_to_bin: gc_to_bin.c slang_shader_syn.h - $(CC) gc_to_bin.c -o gc_to_bin - -# -# syntax scripts -# - -slang_pp_directives_syn.h: syn_to_c slang_pp_directives.syn - ./syn_to_c slang_pp_directives.syn > slang_pp_directives_syn.h - -slang_pp_expression_syn.h: syn_to_c slang_pp_expression.syn - ./syn_to_c slang_pp_expression.syn > slang_pp_expression_syn.h - -slang_shader_syn.h: syn_to_c slang_shader.syn - ./syn_to_c slang_shader.syn > slang_shader_syn.h - -slang_pp_version_syn.h: syn_to_c slang_pp_version.syn - ./syn_to_c slang_pp_version.syn > slang_pp_version_syn.h - -# # builtin library sources # @@ -58,24 +28,24 @@ builtin_110: slang_common_builtin_gc.h slang_core_gc.h slang_fragment_builtin_gc builtin_120: slang_120_core_gc.h slang_builtin_120_common_gc.h slang_builtin_120_fragment_gc.h -slang_120_core_gc.h: gc_to_bin slang_120_core.gc - ./gc_to_bin 1 slang_120_core.gc slang_120_core_gc.h +slang_120_core_gc.h: slang_120_core.gc + $(GLSL_CL) fragment slang_120_core.gc slang_120_core_gc.h -slang_builtin_120_common_gc.h: gc_to_bin slang_builtin_120_common.gc - ./gc_to_bin 1 slang_builtin_120_common.gc slang_builtin_120_common_gc.h +slang_builtin_120_common_gc.h: slang_builtin_120_common.gc + $(GLSL_CL) fragment slang_builtin_120_common.gc slang_builtin_120_common_gc.h -slang_builtin_120_fragment_gc.h: gc_to_bin slang_builtin_120_fragment.gc - ./gc_to_bin 1 slang_builtin_120_fragment.gc slang_builtin_120_fragment_gc.h +slang_builtin_120_fragment_gc.h: slang_builtin_120_fragment.gc + $(GLSL_CL) fragment slang_builtin_120_fragment.gc slang_builtin_120_fragment_gc.h -slang_common_builtin_gc.h: gc_to_bin slang_common_builtin.gc - ./gc_to_bin 1 slang_common_builtin.gc slang_common_builtin_gc.h +slang_common_builtin_gc.h: slang_common_builtin.gc + $(GLSL_CL) fragment slang_common_builtin.gc slang_common_builtin_gc.h -slang_core_gc.h: gc_to_bin slang_core.gc - ./gc_to_bin 1 slang_core.gc slang_core_gc.h +slang_core_gc.h: slang_core.gc + $(GLSL_CL) fragment slang_core.gc slang_core_gc.h -slang_fragment_builtin_gc.h: gc_to_bin slang_fragment_builtin.gc - ./gc_to_bin 1 slang_fragment_builtin.gc slang_fragment_builtin_gc.h +slang_fragment_builtin_gc.h: slang_fragment_builtin.gc + $(GLSL_CL) fragment slang_fragment_builtin.gc slang_fragment_builtin_gc.h -slang_vertex_builtin_gc.h: gc_to_bin slang_vertex_builtin.gc - ./gc_to_bin 2 slang_vertex_builtin.gc slang_vertex_builtin_gc.h +slang_vertex_builtin_gc.h: slang_vertex_builtin.gc + $(GLSL_CL) vertex slang_vertex_builtin.gc slang_vertex_builtin_gc.h diff --git a/src/mesa/shader/slang/library/SConscript b/src/mesa/shader/slang/library/SConscript new file mode 100644 index 00000000000..ef131146be5 --- /dev/null +++ b/src/mesa/shader/slang/library/SConscript @@ -0,0 +1,52 @@ +####################################################################### +# SConscript for GLSL builtin library + +Import('*') + +env = env.Clone() + +# See also http://www.scons.org/wiki/UsingCodeGenerators + +def glsl_compile_emitter(target, source, env): + env.Depends(target, glsl_compile) + return (target, source) + +bld_frag = Builder( + action = glsl_compile[0].abspath + ' fragment $SOURCE $TARGET', + emitter = glsl_compile_emitter, + suffix = '.gc', + src_suffix = '_gc.h') + +bld_vert = Builder( + action = glsl_compile[0].abspath + ' vertex $SOURCE $TARGET', + emitter = glsl_compile_emitter, + suffix = '.gc', + src_suffix = '_gc.h') + +env['BUILDERS']['bld_frag'] = bld_frag +env['BUILDERS']['bld_vert'] = bld_vert + +# Generate GLSL builtin library binaries +env.bld_frag( + '#src/mesa/shader/slang/library/slang_core_gc.h', + '#src/mesa/shader/slang/library/slang_core.gc') +env.bld_frag( + '#src/mesa/shader/slang/library/slang_common_builtin_gc.h', + '#src/mesa/shader/slang/library/slang_common_builtin.gc') +env.bld_frag( + '#src/mesa/shader/slang/library/slang_fragment_builtin_gc.h', + '#src/mesa/shader/slang/library/slang_fragment_builtin.gc') +env.bld_vert( + '#src/mesa/shader/slang/library/slang_vertex_builtin_gc.h', + '#src/mesa/shader/slang/library/slang_vertex_builtin.gc') + +# Generate GLSL 1.20 builtin library binaries +env.bld_frag( + '#src/mesa/shader/slang/library/slang_120_core_gc.h', + '#src/mesa/shader/slang/library/slang_120_core.gc') +env.bld_frag( + '#src/mesa/shader/slang/library/slang_builtin_120_common_gc.h', + '#src/mesa/shader/slang/library/slang_builtin_120_common.gc') +env.bld_frag( + '#src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h', + '#src/mesa/shader/slang/library/slang_builtin_120_fragment.gc') diff --git a/src/mesa/shader/slang/library/gc_to_bin.c b/src/mesa/shader/slang/library/gc_to_bin.c deleted file mode 100644 index 8aef7b54124..00000000000 --- a/src/mesa/shader/slang/library/gc_to_bin.c +++ /dev/null @@ -1,85 +0,0 @@ -#include "../../grammar/grammar_crt.h" -#include "../../grammar/grammar_crt.c" -#include <stdlib.h> -#include <stdio.h> - -static const char *slang_shader_syn = -#include "slang_shader_syn.h" -; - -static int gc_to_bin (grammar id, const char *in, const char *out) -{ - FILE *f; - byte *source, *prod; - unsigned int size, i, line = 0; - - printf ("Precompiling %s\n", in); - - f = fopen (in, "r"); - if (f == NULL) - return 1; - fseek (f, 0, SEEK_END); - size = ftell (f); - fseek (f, 0, SEEK_SET); - source = (byte *) grammar_alloc_malloc (size + 1); - source[fread (source, 1, size, f)] = '\0'; - fclose (f); - - if (!grammar_fast_check (id, source, &prod, &size, 65536)) - { - grammar_alloc_free (source); - return 1; - } - - f = fopen (out, "w"); - fprintf (f, "\n"); - fprintf (f, "/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */\n"); - fprintf (f, "/* %s */\n", in); - fprintf (f, "\n"); - for (i = 0; i < size; i++) - { - unsigned int a; - if (prod[i] < 10) - a = 1; - else if (prod[i] < 100) - a = 2; - else - a = 3; - if (i < size - 1) - a++; - if (line + a >= 100) - { - fprintf (f, "\n"); - line = 0; - } - line += a; - fprintf (f, "%d", prod[i]); - if (i < size - 1) - fprintf (f, ","); - } - fprintf (f, "\n"); - fclose (f); - grammar_alloc_free (prod); - return 0; -} - -int main (int argc, char *argv[]) -{ - grammar id; - - id = grammar_load_from_text ((const byte *) slang_shader_syn); - if (id == 0) { - fprintf(stderr, "Error loading grammar from text\n"); - return 1; - } - grammar_set_reg8 (id, (const byte *) "parsing_builtin", 1); - grammar_set_reg8 (id, (const byte *) "shader_type", atoi (argv[1])); - if (gc_to_bin (id, argv[2], argv[3])) { - fprintf(stderr, "Error in gc_to_bin %s %s\n", argv[2], argv[3]); - grammar_destroy (id); - return 1; - } - grammar_destroy (id); - return 0; -} - diff --git a/src/mesa/shader/slang/library/slang_120_core_gc.h b/src/mesa/shader/slang/library/slang_120_core_gc.h deleted file mode 100644 index 1fdbddf7c3e..00000000000 --- a/src/mesa/shader/slang/library/slang_120_core_gc.h +++ /dev/null @@ -1,764 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_120_core.gc */ - -5,1,90,95,0,0,26,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0, -1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0, -18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,9,0,102,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,17,48,0,48,0,0,0, -17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0, -5,0,105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0, -1,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0, -11,0,99,48,0,0,1,1,0,0,11,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0, -0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,51,48,0,0, -1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,51, -49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119, -0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,119,0,18,102,51,49,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0, -17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0, -0,28,0,1,1,1,0,0,5,0,105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0, -0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0, -0,28,0,1,1,1,0,0,1,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0, -0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0, -28,0,1,1,1,0,0,12,0,99,48,0,0,1,1,0,0,12,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95, -0,0,27,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9, -0,102,49,49,0,0,1,1,0,0,9,0,102,48,50,0,0,1,1,0,0,9,0,102,49,50,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48, -49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0, -0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,5,0,105,0,0,0, -1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,1,0,98,0,0,0,1, -3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,10,0,99,48,0,0,1, -1,0,0,10,0,99,49,0,0,1,1,0,0,10,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0, -0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,51,48,0,0,1,1,0,0,9,0,102,48,49,0,0, -1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,51,49,0,0,1,1,0,0,9,0,102,48, -50,0,0,1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102,50,50,0,0,1,1,0,0,9,0,102,51,50,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59, -122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119, -0,18,102,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,119,0,18,102,51,50,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0, -17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17, -48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,5,0,105,0,0,0,1,3,2, -90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,1,0,98,0,0,0,1,3, -2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,12,0,99,48,0,0,1, -1,0,0,12,0,99,49,0,0,1,1,0,0,12,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0, -0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,48,50,0,0, -1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102,48,51,0,0,1,1,0,0,9,0,102,49,51,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120, -0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0, -18,102,49,51,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,58,109,97,116,52,120,50,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,16,10,52,0,0,17,48,0,48, -0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,5,0, -105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,1,0, -98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,10, -0,99,48,0,0,1,1,0,0,10,0,99,49,0,0,1,1,0,0,10,0,99,50,0,0,1,1,0,0,10,0,99,51,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,9,0,102,48,48,0,0,1, -1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49, -0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,48,50,0,0,1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102, -50,50,0,0,1,1,0,0,9,0,102,48,51,0,0,1,1,0,0,9,0,102,49,51,0,0,1,1,0,0,9,0,102,50,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48, -49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50, -50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,102,49,51,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,59,122,0,18,102,50,51,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0, -0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17, -48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,5,0,105,0,0,0, -1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,1,0,98,0,0,0,1, -3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,11,0,99,48,0,0,1, -1,0,0,11,0,99,49,0,0,1,1,0,0,11,0,99,50,0,0,1,1,0,0,11,0,99,51,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1, -90,95,0,0,13,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,0, -18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,26,0,109,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0, -18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16, -10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57, -59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121, -0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0, -1,90,95,0,0,13,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0, -0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0, -26,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,26, -0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,109, -0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,31,0,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16, -10,49,0,57,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59, -120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57, -59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0, -57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59, -121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48, -0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17, -48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0, -0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0, -0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1, -90,95,0,0,28,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90, -95,0,0,28,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52, -0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,15,0,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,0,18, -109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59, -121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109, -0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0, -28,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18, -109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17, -48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10, -49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0, -16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57, -59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0, -20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0, -0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57, -59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0, -57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,29,0,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0, -18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0, -0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,0,1, -1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,27,0,1,1,1, -0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8, -48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,14, -0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57, -0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,30,0,109,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120, -0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, -121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,90,95,0,0, -27,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18, -109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0, -18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121, -0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16, -10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0, -1,90,95,0,0,27,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51, -120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,58,118,101,99,50,0,0,17,48,0,48,0,0,0, -0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16, -10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0, -1,90,95,0,0,27,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51, -120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0, -57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95, -0,0,14,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0, -14,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0, -16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0, -0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57, -59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122, -0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0, -18,109,0,16,10,50,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,26,0,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0, -57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,28,0,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0, -18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90, -95,0,0,14,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18, -109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0, -17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0, -0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,13,0,109,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18, -109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0, -30,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,30, -0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109, -0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1, -0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8, -48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0, -0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0, -18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10, -49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1, -0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8, -48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0, -0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0, -16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0, -20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17, -48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0, -30,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18, -109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0, -0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,29, -0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,31,0, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57, -59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109, -0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16, -10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,0,16,10,51,0,57,59,120,121, -0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0, -0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18, -109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0, -30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0, -57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48, -0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,26,0,109,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120, -121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0, -0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,31,0,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,15,0,109,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,59,120, -121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,18, -109,0,16,10,51,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10, -49,0,57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0, -31,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18, -109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50, -0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1, -0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8, -48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0, -0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,26,0,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16, -10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0, -20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18, -109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0, -31,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18, -109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17, -48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1, -0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8, -48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0, -0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,15,0,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,30,0,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0, -57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1, -90,95,0,0,15,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0, -18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0, -57,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,28,0, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,18,109, -0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0, -0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0, -18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48, -0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0, -14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0, -17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48, -0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,26,0,109,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0, -18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,27,0,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0, -0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17, -49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95, -0,0,15,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109, -0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0, -17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0, -17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109, -0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1, -90,95,0,0,0,0,2,1,1,0,2,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2, -0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21, -0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1, -90,95,0,0,0,0,2,2,1,0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2, -0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,27,0,109,0,0,1,1,0,0, -27,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,90,95,0,0,0,0,2, -2,1,0,2,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0, -9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, -57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,109,0,16,8,48,0, -57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90, -95,0,0,0,0,2,2,1,0,2,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8, -48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110, -0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90,95,0,0,0,0,2,4,1, -0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9, -18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,28,0,109,0,0,1,1, -0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,30,0,109,0, -0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49, -0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0, -0,0,0,2,4,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0, -31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109, -0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9, -18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,26,0,109,0,0,1, -1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8, -48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0, -18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59, -120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20, -0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0, -16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18, -109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118, -0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, -18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59, -119,0,48,46,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59, -121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121, -0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0, -57,59,121,0,48,46,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118, -0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0, -48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59, -121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10, -50,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0, -16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0, -18,109,0,16,10,50,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,118,0,59, -120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,18, -118,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,48,46,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,29,0,109, -0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109, -0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122, -0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,120,0,48,46,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48, -18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57, -59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,0,1,90,95,0,0,11,0,2, -21,1,1,0,0,31,0,109,0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, -118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0, -48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51, -0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16, -8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18, -109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18, -118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59, -122,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,48,46,20,0,0,1,90,95,0,0,27,0,2,21,1, -1,0,0,13,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0, -16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50, -0,57,48,20,0,0,1,90,95,0,0,29,0,2,21,1,1,0,0,13,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51, -0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0, -13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1, -90,95,0,0,14,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,29,0,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1, -0,0,28,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0, -16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0, -28,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18, -110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16, -10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20, -0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0,0,27,0,109,0,0,1, -1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90, -95,0,0,29,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, -0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110, -0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,14,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21, -1,1,0,0,14,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16, -10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0, -57,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,30, -0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110, -0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48, -20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, -109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,28,0,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0, -0,27,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18, -110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,29,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,31,0, -109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0, -16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0, -57,48,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,31,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,31,0,109,0,0, -1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0, -28,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,30,0,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,0,0,2, -3,1,0,2,0,26,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0, -0,0,2,3,1,0,2,0,28,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90, -95,0,0,0,0,2,3,1,0,2,0,27,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0, -0,1,90,95,0,0,0,0,2,3,1,0,2,0,30,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0, -48,20,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,29,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,18,109,0,18, -110,0,48,20,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,31,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,18, -109,0,18,110,0,48,20,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,27,0,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10, -50,0,57,0,0,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10, -51,0,57,0,0,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20, -0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0, -20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90, -95,0,0,11,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0, -0,0,0,2,1,1,0,2,0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18, -109,0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1, -0,2,0,27,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49, -0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,30,0,109,0,0, -1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0, -9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10, -50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,31,0,109, -0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0, -21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0, -2,2,1,0,2,0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16, -10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109, -0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,27, -0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18, -97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,30,0,109,0,0,1,1,0,0, -9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109, -0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, -18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,31,0,109,0,0,1,1, -0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18, -109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,3,1,0,2, -0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, -57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,27,0,109, -0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0, -23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,30,0,109,0,0,1,1,0,0,9,0, -97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0, -16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18, -97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,31,0,109,0,0,1,1,0,0, -9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109, -0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0, -26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57, -18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48, -0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,27,0,109,0,0, -1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0, -9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,30,0,109,0,0,1,1,0,0,9,0,97,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10, -50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0, -24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,31,0,109,0,0,1,1,0,0,9,0, -97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0, -16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0, -26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,28,0,2,26,1, -1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,27,0, -2,26,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,30,0,2,26,1,1,0,0,30,0,109,0,0,1,1,0,0,30,0, -110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109, -0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0, -0,1,90,95,0,0,29,0,2,26,1,1,0,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0, -0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, -57,46,0,0,0,0,1,90,95,0,0,31,0,2,26,1,1,0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116, -52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16, -10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0, -16,10,51,0,57,46,0,0,0,0,1,90,95,0,0,26,0,2,27,1,1,0,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58, -109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0, -0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0, -110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109, -0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0, -0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0, -0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,29,0,2,27,1,1,0,0,29,0, -109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8, -48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,31,0,2,27,1,1, -0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57, -18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0, -26,0,2,22,1,1,0,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16, -8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,0,0,0,1,90, -95,0,0,28,0,2,22,1,1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,0, -0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50, -0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,90,95,0,0,30,0,2,22,1,1,0,0,30,0, -109,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8, -48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,49,0,0,0,0,1,90,95,0,0,29,0,2,22,1,1,0,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109, -97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57, -18,110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0, -0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16, -10,51,0,57,18,110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0, -110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16, -10,49,0,57,46,0,0,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109, -97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0, -1,90,95,0,0,28,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18, -97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,28,0,2,26,1, -1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18, -98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,27,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0, -0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110, -0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,27,0,2,26,1,1,0,0,27,0, -109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18, -109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,30,0,2,26,1, -1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,97,0,18,110,0,16,8,48,0, -57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0, -30,0,2,26,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8, -48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0, -1,90,95,0,0,29,0,2,26,1,1,0,0,29,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18, -109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0, -46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,29,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,29, -0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16, -10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,90, -95,0,0,31,0,2,26,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109, -0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46, -0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,31,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,31,0, -110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16, -10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,90, -95,0,0,26,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,97,0, -18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,26,0,2,27,1,1,0,0, -26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,47, -0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0, -110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16, -10,49,0,57,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109, -97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0, -1,90,95,0,0,27,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18, -97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57, -47,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51, -120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10, -50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58, -109,97,116,51,120,52,0,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18, -97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0, -0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18, -98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,29,0,2,27,1,1,0,0,29,0,109,0,0,1,1, -0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10, -49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0, -1,90,95,0,0,29,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18, -97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57, -47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,31,0,2,27,1,1,0,0,31,0,109,0,0,1,1,0,0, -9,0,98,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49, -0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,1, -90,95,0,0,31,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18, -97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57, -47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,26, -0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90, -95,0,0,26,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0, -28,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0, -0,27,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48, -20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,90,95,0,0,29,0,2,21,1, -1,0,0,29,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0, -1,90,95,0,0,29,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, -16,10,51,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0, -0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2, -22,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0, -18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0, -57,48,20,0,0,1,90,95,0,0,26,0,2,22,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1, -105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,28,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,28, -0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,28,0,2,22, -1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18, -98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0, -48,20,0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1, -105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,27,0,109,0,0,1,1,0,0, -9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,30,0,2, -22,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0, -18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0, -57,48,20,0,0,1,90,95,0,0,30,0,2,22,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1, -105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,29,0,2,22,1,1,0,0,29,0,109,0,0,1,1,0,0, -9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,29,0,2, -22,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0, -18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0, -57,48,20,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1, -105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, -109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,31, -0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,27, -1,1,0,0,26,0,109,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10, -49,0,57,54,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18, -109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0, -0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109, -0,16,10,50,0,57,54,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,0,1,8,58,109,97,116,51,120, -52,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,0,0,0,1, -90,95,0,0,29,0,2,27,1,1,0,0,29,0,109,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57, -54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1, -90,95,0,0,31,0,2,27,1,1,0,0,31,0,109,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57, -54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1, -90,95,0,0,0,0,2,25,1,0,2,0,26,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57, -52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,28,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10, -49,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,27,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109, -0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,30,0,109,0,0,0,1, -9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,90,95, -0,0,0,0,2,25,1,0,2,0,29,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9, -18,109,0,16,10,50,0,57,52,0,9,18,109,0,16,10,51,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,31,0,109, -0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9, -18,109,0,16,10,51,0,57,52,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,26,0,109,0,0,0,1,9,18,109,0,16,8,48,0, -57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,28,0,109,0,0,0,1,9,18,109,0, -16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,27,0,109,0,0,0,1,9, -18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,90,95,0, -0,0,0,2,24,1,0,2,0,30,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9, -18,109,0,16,10,50,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,29,0,109,0,0,0,1,9,18,109,0,16,8,48,0, -57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0, -0,1,90,95,0,0,0,0,2,24,1,0,2,0,31,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0, -57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h deleted file mode 100644 index c397b9f0fad..00000000000 --- a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h +++ /dev/null @@ -1,108 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_builtin_120_common.gc */ - -5,1,90,95,0,0,26,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,26,0,109,0,0,1, -0,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,28,0,0,109,97,116,114, -105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,28,0,109,0,0,1,0,0,0,28,0,110,0,0,0,1,8,58,109,97, -116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,27,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0, -1,0,0,0,27,0,109,0,0,1,0,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,30,0,0,109,97,116,114,105,120,67,111,109,112,77, -117,108,116,0,1,0,0,0,30,0,109,0,0,1,0,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109, -0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,29,0,0,109,97,116,114,105,120,67,111, -109,112,77,117,108,116,0,1,0,0,0,29,0,109,0,0,1,0,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0, -0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, -57,48,0,0,0,0,1,90,95,0,0,31,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,31, -0,109,0,0,1,0,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16, -8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0, -16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,0,13,0,0,111, -117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,10,0,114,0,0,0,1,8,58,109, -97,116,50,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18, -99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,14, -0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8, -58,109,97,116,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48, -0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0, -18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0, -48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95, -0,0,15,0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,12,0,114,0, -0,0,1,8,58,109,97,116,52,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59, -120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0, -59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0, -59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18, -99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18, -114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48, -0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0,59,119,0,18,114,0,59,119,0,48,0,0,0,0,1,90,95,0, -0,26,0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,10,0,114,0,0, -0,1,8,58,109,97,116,50,120,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114, -0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18, -99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,27, -0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8, -58,109,97,116,51,120,50,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59, -120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0, -59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,28,0,0, -111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,10,0,114,0,0,0,1,8,58, -109,97,116,50,120,52,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120, -0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59, -120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59, -121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,29,0,0,111,117,116,101,114,80, -114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,12,0,114,0,0,0,1,8,58,109,97,116,52,120,50,0, -0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0, -18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0, -48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59, -121,0,18,114,0,59,119,0,48,0,0,0,0,1,90,95,0,0,30,0,0,111,117,116,101,114,80,114,111,100,117,99, -116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,99,0,59,120,0, -18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0, -48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59, -121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59, -121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0, -59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,31,0,0, -111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,12,0,114,0,0,0,1,8,58, -109,97,116,52,120,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120, -0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59, -121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59, -122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0, -59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0, -59,119,0,48,0,0,0,0,1,90,95,0,0,13,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,13,0,109,0,0,0, -1,8,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109, -0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,90,95,0,0,14,0,0,116,114,97, -110,115,112,111,115,101,0,1,0,0,0,14,0,109,0,0,0,1,8,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,59, -120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57, -59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0, -57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,90,95,0, -0,15,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,15,0,109,0,0,0,1,8,58,109,97,116,52,0,0,18, -109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0, -18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0, -0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122, -0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59, -122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57, -59,119,0,0,18,109,0,16,10,51,0,57,59,119,0,0,0,0,0,1,90,95,0,0,26,0,0,116,114,97,110,115,112,111, -115,101,0,1,0,0,0,27,0,109,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0, -18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0, -0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,0,0,1,90,95,0,0,27,0,0,116, -114,97,110,115,112,111,115,101,0,1,0,0,0,26,0,109,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0, -16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109, -0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,0,0, -0,1,90,95,0,0,28,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,29,0,109,0,0,0,1,8,58,109,97,116, -50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50, -0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10, -49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,0,0,0,1,90, -95,0,0,29,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,28,0,109,0,0,0,1,8,58,109,97,116,52,120, -50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59, -121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57, -59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,0,0,0,1,90,95,0,0,30, -0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,31,0,109,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18, -109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0, -18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0, -0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122, -0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59, -122,0,0,0,0,0,1,90,95,0,0,31,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,30,0,109,0,0,0,1,8, -58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18, -109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0, -18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0, -0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119, -0,0,18,109,0,16,10,50,0,57,59,119,0,0,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h deleted file mode 100644 index add3b5aeaa9..00000000000 --- a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h +++ /dev/null @@ -1,5 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_builtin_120_fragment.gc */ - -5,2,2,90,95,3,0,10,0,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h deleted file mode 100644 index 3c3666e4ea2..00000000000 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ /dev/null @@ -1,880 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_common_builtin.gc */ - -5,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,76,105,103,104,116,115,0,2,16,10,56,0,0,0,2,2,90,95,1,0, -5,0,1,103,108,95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,2,16,10,54,0,0,0,2,2,90,95,1,0,5, -0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,85,110,105,116,115,0,2,16,10,56,0,0,0,2,2,90, -95,1,0,5,0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,2,16,10,56,0, -0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,65,116,116,114,105,98,115,0,2, -16,10,49,54,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,85,110,105,102, -111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,53,49,50,0,0,0,2,2,90,95,1,0,5,0,1, -103,108,95,77,97,120,86,97,114,121,105,110,103,70,108,111,97,116,115,0,2,16,10,51,50,0,0,0,2,2,90, -95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,84,101,120,116,117,114,101,73,109,97,103, -101,85,110,105,116,115,0,2,16,8,48,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,67,111,109,98, -105,110,101,100,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,10,50,0,0,0, -2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105, -116,115,0,2,16,10,50,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,70,114,97,103,109,101,110,116, -85,110,105,102,111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,54,52,0,0,0,2,2,90,95, -1,0,5,0,1,103,108,95,77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,2,16,10,49,0,0,0,2,2,90, -95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,0,0,0,2,2,90,95,4, -0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,90,95,4, -0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97, -116,114,105,120,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105, -120,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95, -4,0,14,0,1,103,108,95,78,111,114,109,97,108,77,97,116,114,105,120,0,0,0,2,2,90,95,4,0,15,0,1,103, -108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2, -2,90,95,4,0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110, -118,101,114,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114, -111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,90,95,4, -0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,101,0, -3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,15, -0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,84,114,97,110,115,112,111, -115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114, -105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108, -86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112, -111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120, -84,114,97,110,115,112,111,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111, -111,114,100,115,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116, -114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0, -1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115, -101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86, -105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101, -84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101, -77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,3,18,103,108, -95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,9,0,1,103,108, -95,78,111,114,109,97,108,83,99,97,108,101,0,0,0,2,2,90,95,0,0,24,103,108,95,68,101,112,116,104,82, -97,110,103,101,80,97,114,97,109,101,116,101,114,115,0,9,0,110,101,97,114,0,0,0,1,9,0,102,97,114,0, -0,0,1,9,0,100,105,102,102,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,68,101,112,116,104,82,97,110, -103,101,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,68,101,112,116,104,82,97,110,103,101, -0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,67,108,105,112,80,108,97,110,101,0,3,18,103,108,95,77,97,120, -67,108,105,112,80,108,97,110,101,115,0,0,0,2,2,90,95,0,0,24,103,108,95,80,111,105,110,116,80,97, -114,97,109,101,116,101,114,115,0,9,0,115,105,122,101,0,0,0,1,9,0,115,105,122,101,77,105,110,0,0,0, -1,9,0,115,105,122,101,77,97,120,0,0,0,1,9,0,102,97,100,101,84,104,114,101,115,104,111,108,100,83, -105,122,101,0,0,0,1,9,0,100,105,115,116,97,110,99,101,67,111,110,115,116,97,110,116,65,116,116,101, -110,117,97,116,105,111,110,0,0,0,1,9,0,100,105,115,116,97,110,99,101,76,105,110,101,97,114,65,116, -116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,100,105,115,116,97,110,99,101,81,117,97,100,114,97, -116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,80, -111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,80,111,105,110,116,0,0,0,2,2, -90,95,0,0,24,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,12,0, -101,109,105,115,115,105,111,110,0,0,0,1,12,0,97,109,98,105,101,110,116,0,0,0,1,12,0,100,105,102, -102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0,0,0,1,9,0,115,104,105,110,105,110,101, -115,115,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109, -101,116,101,114,115,0,0,1,103,108,95,70,114,111,110,116,77,97,116,101,114,105,97,108,0,0,0,2,2,90, -95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,0,1,103, -108,95,66,97,99,107,77,97,116,101,114,105,97,108,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104, -116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,12,0,97,109,98,105,101,110,116,0, -0,0,1,12,0,100,105,102,102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0,0,0,1,12,0,112, -111,115,105,116,105,111,110,0,0,0,1,12,0,104,97,108,102,86,101,99,116,111,114,0,0,0,1,11,0,115,112, -111,116,68,105,114,101,99,116,105,111,110,0,0,0,1,9,0,115,112,111,116,67,111,115,67,117,116,111, -102,102,0,0,0,1,9,0,99,111,110,115,116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0, -1,9,0,108,105,110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,113,117,97,100, -114,97,116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,115,112,111,116,69,120,112, -111,110,101,110,116,0,0,0,1,9,0,115,112,111,116,67,117,116,111,102,102,0,0,0,0,0,0,0,2,2,90,95,4,0, -25,103,108,95,76,105,103,104,116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,0,1, -103,108,95,76,105,103,104,116,83,111,117,114,99,101,0,3,18,103,108,95,77,97,120,76,105,103,104,116, -115,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108,80,97,114,97,109,101, -116,101,114,115,0,12,0,97,109,98,105,101,110,116,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105, -103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,76,105,103, -104,116,77,111,100,101,108,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108, -80,114,111,100,117,99,116,115,0,12,0,115,99,101,110,101,67,111,108,111,114,0,0,0,0,0,0,0,2,2,90,95, -4,0,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,0,1,103, -108,95,70,114,111,110,116,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2, -2,90,95,4,0,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,0, -1,103,108,95,66,97,99,107,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2, -2,90,95,0,0,24,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,12,0,97,109,98,105, -101,110,116,0,0,0,1,12,0,100,105,102,102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0, -0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,0,1,103, -108,95,70,114,111,110,116,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120, -76,105,103,104,116,115,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99, -116,115,0,0,1,103,108,95,66,97,99,107,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108, -95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,84,101,120,116,117,114, -101,69,110,118,67,111,108,111,114,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97, -103,101,85,110,105,116,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,69,121,101,80,108,97,110,101,83,0, -3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12, -0,1,103,108,95,69,121,101,80,108,97,110,101,84,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114, -101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,69,121,101,80,108,97,110,101,82,0, -3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12, -0,1,103,108,95,69,121,101,80,108,97,110,101,81,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114, -101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97, -110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2, -90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,84,0,3,18,103,108,95,77,97,120, -84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106, -101,99,116,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111, -114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,81,0,3,18, -103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,0,0,24,103, -108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,0,12,0,99,111,108,111,114,0,0,0,1,9,0,100, -101,110,115,105,116,121,0,0,0,1,9,0,115,116,97,114,116,0,0,0,1,9,0,101,110,100,0,0,0,1,9,0,115,99, -97,108,101,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,70,111,103,80,97,114,97,109,101,116,101,114, -115,0,0,1,103,108,95,70,111,103,0,0,0,1,90,95,0,0,9,0,0,114,97,100,105,97,110,115,0,1,1,0,0,9,0, -100,101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0, -0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0, -18,100,101,103,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,0,114,97,100,105,97,110,115,0,1,1,0,0,10,0,100, -101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0, -49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,0,114,97, -100,105,97,110,115,0,1,1,0,0,11,0,100,101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49, -53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59, -120,120,120,0,0,0,0,1,90,95,0,0,12,0,0,114,97,100,105,97,110,115,0,1,1,0,0,12,0,100,101,103,0,0,0, -1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,100,101,103,0, -0,18,99,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,9,0,0,100,101,103,114,101,101,115,0,1,1,0,0,9,0, -114,97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0, -0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0, -18,114,97,100,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,0,100,101,103,114,101,101,115,0,1,1,0,0,10,0,114, -97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49, -0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,0,100,101,103, -114,101,101,115,0,1,1,0,0,11,0,114,97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0, -17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120, -120,120,0,0,0,0,1,90,95,0,0,12,0,0,100,101,103,114,101,101,115,0,1,1,0,0,12,0,114,97,100,0,0,0,1,3, -2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99, -0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,9,0,0,115,105,110,0,1,1,0,0,9,0,114,97,100,105,97,110,115, -0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100, -105,97,110,115,0,0,0,0,1,90,95,0,0,10,0,0,115,105,110,0,1,1,0,0,10,0,114,97,100,105,97,110,115,0,0, -0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114, -97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,115, -105,110,0,1,1,0,0,11,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, -111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, -110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,115,105,110,0,1,1,0,0, -12,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115, -105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0, -0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114, -97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,99,111, -115,0,1,1,0,0,9,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101, -0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,90,95,0,0,10,0,0,99, -111,115,0,1,1,0,0,10,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0, -4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,99,111,115,0,1,1,0,0,11,0,114,97,100, -105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115, -105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0, -0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, -18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,99,111,115,0,1,1,0,0,12,0,114,97, -100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111, -115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59, -121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122, -0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,90,95, -0,0,9,0,0,116,97,110,0,1,1,0,0,9,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,9,0,1,115,0,2,58,115, -105,110,0,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,9,0,1,99,0,2,58,99,111,115,0,0,18,97,110, -103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,10,0,0,116,97,110,0,1,1,0,0,10,0,97, -110,103,108,101,0,0,0,1,3,2,90,95,1,0,10,0,1,115,0,2,58,115,105,110,0,0,18,97,110,103,108,101,0,0, -0,0,0,3,2,90,95,1,0,10,0,1,99,0,2,58,99,111,115,0,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18, -99,0,49,0,0,1,90,95,0,0,11,0,0,116,97,110,0,1,1,0,0,11,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0, -11,0,1,115,0,2,58,115,105,110,0,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,11,0,1,99,0,2,58, -99,111,115,0,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,12,0,0,116,97, -110,0,1,1,0,0,12,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,12,0,1,115,0,2,58,115,105,110,0,0,18, -97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,12,0,1,99,0,2,58,99,111,115,0,0,18,97,110,103,108,101,0, -0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,9,0,0,97,115,105,110,0,1,1,0,0,9,0,120,0,0,0,1,3,2, -90,95,1,0,9,0,1,97,48,0,2,17,49,0,53,55,48,55,50,56,56,0,0,0,0,3,2,90,95,1,0,9,0,1,97,49,0,2,17,48, -0,50,49,50,49,49,52,52,0,0,54,0,0,3,2,90,95,1,0,9,0,1,97,50,0,2,17,48,0,48,55,52,50,54,49,48,0,0,0, -0,3,2,90,95,1,0,9,0,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48, -0,0,3,2,90,95,1,0,9,0,1,121,0,2,58,97,98,115,0,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108, -0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97,48,0,18, -121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,0,18,120,0,0,0,48,20,0, -0,1,90,95,0,0,10,0,0,97,115,105,110,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,58,97,115,105,110,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,58,97,115,105,110,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,115,105,110,0,1,1,0, -0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,0,18,118,0,59, -120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,0,18,118,0,59,121,0, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,0,18,118,0,59,122,0,0,0, -20,0,0,1,90,95,0,0,12,0,0,97,115,105,110,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,97,115,105,110,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,97,115,105,110,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -122,0,58,97,115,105,110,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, -58,97,115,105,110,0,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,99,111,115,0,1,1,0,0,9,0, -120,0,0,0,1,3,2,90,95,1,0,9,0,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0, -53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,105,0,58,97,115,105,110,0,0, -18,120,0,0,0,47,20,0,0,1,90,95,0,0,10,0,0,97,99,111,115,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,97,99,111,115,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97, -99,111,115,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115, -0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,0,18, -118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,0,18,118,0, -59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97,99,111,115,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,97,99,111,115,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,97,99,111,115,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,59,122,0,58,97,99,111,115,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -119,0,58,97,99,111,115,0,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,116,97,110,0,1,1,0,0, -9,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,97,115,105,110,0,0,18,120,0,58,105,110,118, -101,114,115,101,115,113,114,116,0,0,18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,90, -95,0,0,10,0,0,97,116,97,110,0,1,1,0,0,10,0,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95, -120,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,116,97,110,0,1,1,0,0,11,0,121,95,111,118,101,114, -95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,121,95,111,118, -101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0, -0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, -58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97, -116,97,110,0,1,1,0,0,12,0,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95, -120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,0,18,121,95, -111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,116,97,110,0,1,1,0,0,9,0,121,0,0, -1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,10,58,97,98,115,0,0,18,120,0,0,0,17,49,0,48, -0,45,52,0,41,0,2,9,18,114,0,58,97,116,97,110,0,0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0, -48,0,0,40,0,2,9,18,114,0,18,114,0,58,115,105,103,110,0,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0, -0,48,46,20,0,0,9,14,0,0,2,9,18,114,0,58,115,105,103,110,0,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54, -53,0,0,48,20,0,0,8,18,114,0,0,0,1,90,95,0,0,10,0,0,97,116,97,110,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10, -0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,117,0,59,120,0,0, -18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,117, -0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,116,97,110,0,1,1,0,0,11,0,117,0,0, -1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,117,0, -59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110, -0,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, -97,116,97,110,0,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97,116,97, -110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58, -97,116,97,110,0,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,59,121,0,58,97,116,97,110,0,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,122,0,58,97,116,97,110,0,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,0,18,117,0,59,119,0,0,18,118,0,59,119,0, -0,0,20,0,0,1,90,95,0,0,9,0,0,112,111,119,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,102,108,111, -97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95, -0,0,10,0,0,112,111,119,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,102,108,111,97,116,95,112, -111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0, -0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,112,111,119,0,1,1,0,0,11,0,97,0,0,1,1,0, -0,11,0,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0, -59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, -18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111, -97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18, -98,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,112,111,119,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4, -102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59, -120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86, -97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119, -101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4, -102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59, -119,0,0,18,98,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,101,120,112,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0, -0,9,0,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120, -112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,116,0,0,0,0,1,90,95,0,0,10,0,0,101,120,112,0,1,1,0, -0,10,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0, -4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59, -120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -116,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,101,120,112,0,1,1,0,0,11,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1, -116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50, -0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101, -120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97, -116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,0,1,90, -95,0,0,12,0,0,101,120,112,0,1,1,0,0,12,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,18,97,0,17,49,0, -52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114, -101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18, -95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112, -50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,116,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,108,111, -103,50,0,1,1,0,0,9,0,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86, -97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,108,111,103,50,0,1,1,0,0,10,0,118,0,0,0,1,4,102,108, -111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4, -102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121, -0,0,0,0,1,90,95,0,0,11,0,0,108,111,103,50,0,1,1,0,0,11,0,118,0,0,0,1,4,102,108,111,97,116,95,108, -111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97, -116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102, -108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0, -0,0,1,90,95,0,0,12,0,0,108,111,103,50,0,1,1,0,0,12,0,118,0,0,0,1,4,102,108,111,97,116,95,108,111, -103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95, -108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111, -97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102, -108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0, -0,0,1,90,95,0,0,9,0,0,108,111,103,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54, -57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,0,18,120,0,0,0,18,99,0,48,0,0,1,90,95,0,0,10, -0,0,108,111,103,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49, -56,49,0,0,0,0,8,58,108,111,103,50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,11,0,0,108,111,103,0, -1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8, -58,108,111,103,50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,12,0,0,108,111,103,0,1,1,0,0,12,0, -118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103, -50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,9,0,0,101,120,112,50,0,1,1,0,0,9,0,97,0,0,0,1,4,102, -108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10, -0,0,101,120,112,50,0,1,1,0,0,10,0,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95, -95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,101,120,112,50,0,1, -1,0,0,11,0,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97, -108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, -116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,101,120,112,50,0,1,1,0,0,12,0, -97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121, -0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0, -59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, -97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,115,113,114,116,0,1,1,0,0,9,0,120,0,0, -0,1,3,2,90,95,1,0,9,0,1,110,120,0,2,18,120,0,54,0,0,3,2,90,95,0,0,9,0,1,114,0,0,0,4,102,108,111,97, -116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,0,18, -114,0,0,0,4,118,101,99,52,95,99,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,110,120,0,0,18,114, -0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,10,0,0,115,113,114,116,0,1,1,0,0,10,0,120,0,0,0,1,3,2,90,95,1, -0,10,0,1,110,120,0,2,18,120,0,54,0,1,1,122,101,114,111,0,2,58,118,101,99,50,0,0,17,48,0,48,0,0,0,0, -0,0,3,2,90,95,0,0,10,0,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,59,120,0,0,18, -120,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,59,121,0,0,18,120,0,59,121,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,120,0,0,18,114,0,59,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,114,0,59,121,0,0,18,114,0,59,121,0,0,0,4,118,101,99,52,95,99,109,112,0,18, -95,95,114,101,116,86,97,108,0,0,18,110,120,0,0,18,114,0,0,18,122,101,114,111,0,0,0,0,1,90,95,0,0, -11,0,0,115,113,114,116,0,1,1,0,0,11,0,120,0,0,0,1,3,2,90,95,1,0,11,0,1,110,120,0,2,18,120,0,54,0,1, -1,122,101,114,111,0,2,58,118,101,99,51,0,0,17,48,0,48,0,0,0,0,0,0,3,2,90,95,0,0,11,0,1,114,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,59,120,0,0,18,120,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,114,0,59,121,0,0,18,120,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,114,0,59,122,0,0,18,120,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,120,0,0, -18,114,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,121,0,0,18,114,0,59,121,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,122,0,0,18,114,0,59,122,0,0,0,4,118,101,99,52, -95,99,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,110,120,0,0,18,114,0,0,18,122,101,114,111,0, -0,0,0,1,90,95,0,0,12,0,0,115,113,114,116,0,1,1,0,0,12,0,120,0,0,0,1,3,2,90,95,1,0,12,0,1,110,120,0, -2,18,120,0,54,0,1,1,122,101,114,111,0,2,58,118,101,99,52,0,0,17,48,0,48,0,0,0,0,0,0,3,2,90,95,0,0, -12,0,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,59,120,0,0,18,120,0,59,120,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,59,121,0,0,18,120,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,114,0,59,122,0,0,18,120,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,114,0,59,119,0,0,18,120,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,120,0,0, -18,114,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,121,0,0,18,114,0,59,121,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,114,0,59,122,0,0,18,114,0,59,122,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,114,0,59,119,0,0,18,114,0,59,119,0,0,0,4,118,101,99,52,95,99,109,112,0,18, -95,95,114,101,116,86,97,108,0,0,18,110,120,0,0,18,114,0,0,18,122,101,114,111,0,0,0,0,1,90,95,0,0,9, -0,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,9,0,120,0,0,0,1,4,102,108,111,97,116,95, -114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,105, -110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,10,0,118,0,0,0,1,4,102,108,111,97,116,95,114,115, -113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,90,95,0,0,11,0, -0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,11,0,118,0,0,0,1,4,102,108,111,97,116,95, -114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97, -116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108, -111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1, -90,95,0,0,12,0,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,12,0,118,0,0,0,1,4,102,108, -111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0, -0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59, -122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118, -0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,9,0,120,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,114,109,97,108, -105,122,101,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,115,0,2,58,105,110,118,101,114,115,101, -115,113,114,116,0,0,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0, -0,1,90,95,0,0,11,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0,9, -0,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0, -18,116,109,112,0,0,0,0,1,90,95,0,0,12,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,12,0,118,0, -0,0,1,3,2,90,95,0,0,9,0,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111,116,0,18,116,109,112,0,0,18, -118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0, -4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,90,95,0,0,9,0,0,97,98,115,0,1,1,0,0,9,0,97,0,0,0,1,4, -118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0, -97,98,115,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,0,97,98,115,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,99, -52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,12, -0,0,97,98,115,0,1,1,0,0,12,0,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97, -108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,0,115,105,103,110,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9, -0,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,17,48,0,48,0,0,0, -0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,90,95,0, -0,10,0,0,115,105,103,110,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0,1,112,0,0,1,1,110,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99, -52,95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0, -0,1,90,95,0,0,11,0,0,115,105,103,110,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0,11,0,1,112,0,0,1,1, -110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0, -0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -112,0,0,18,110,0,0,0,0,1,90,95,0,0,12,0,0,115,105,103,110,0,1,1,0,0,12,0,118,0,0,0,1,3,2,90,95,0,0, -12,0,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0, -0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,90, -95,0,0,9,0,0,102,108,111,111,114,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0,102,108,111,111,114,0,1,1,0,0, -10,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,0,102,108,111,111,114,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101, -99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1, -90,95,0,0,12,0,0,102,108,111,111,114,0,1,1,0,0,12,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111, -114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,0,99,101,105,108,0,1,1,0,0, -9,0,97,0,0,0,1,3,2,90,95,0,0,9,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0, -18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,10,0,0,99, -101,105,108,0,1,1,0,0,10,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52, -95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18, -98,0,54,20,0,0,1,90,95,0,0,11,0,0,99,101,105,108,0,1,1,0,0,11,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,98, -0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,90,95,0,0,12,0,0,99,101,105,108,0,1,1,0,0, -12,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114, -0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,9,0,0,102, -114,97,99,116,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0,102,114,97,99,116,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101, -99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0, -11,0,0,102,114,97,99,116,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,12,0,0,102,114,97,99,116,0,1,1,0, -0,12,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -0,0,1,90,95,0,0,9,0,0,109,111,100,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1, -111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0, -0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,10,0,0,109,111,100,0, -1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,0,18,97,0,18,111,110, -101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,11,0,0,109,111,100,0,1,1,0,0,11,0,97,0,0,1, -1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,18,97,0,18,98,0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101, -114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12,0,0,109,111,100,0,1,1,0,0,12,0,97,0,0,1,1,0,0,9,0,98,0, -0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98, -0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90, -95,0,0,10,0,0,109,111,100,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,3,2,90,95,0,0,10,0,1,111, -110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114, -66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58, -102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0, -11,0,0,109,111,100,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,3,2,90,95,0,0,11,0,1,111,110,101, -79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59, -120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66, -0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102, -108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12, -0,0,109,111,100,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,3,2,90,95,0,0,12,0,1,111,110,101,79, -118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59, -120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66, -0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, -118,101,114,66,0,59,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98, -0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90, -95,0,0,9,0,0,109,105,110,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105, -110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,0,109,105,110,0, -1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,0, -109,105,110,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0, -0,0,0,1,90,95,0,0,12,0,0,109,105,110,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,118,101,99,52, -95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,0,109, -105,110,0,1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,0,109,105,110,0,1,1,0, -0,11,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97, -108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,0,109,105,110,0,1,1,0,0,12,0,97, -0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18, -97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,0,109,97,120,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4, -118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0, -0,10,0,0,109,97,120,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0, -1,90,95,0,0,11,0,0,109,97,120,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118,101,99,52,95,109, -97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59, -120,121,122,0,0,0,0,1,90,95,0,0,12,0,0,109,97,120,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4, -118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0, -0,10,0,0,109,97,120,0,1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,0,109,97, -120,0,1,1,0,0,11,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,0,109,97,120,0,1,1,0,0, -12,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,0,99,108,97,109,112,0,1,1,0,0,9,0,118,97,108,0,0,1,1,0, -0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, -97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, -109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,10,0,0,99,108,97,109,112,0,1,1,0,0,10,0,118,97,108,0,0,1, -1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99, -108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0, -18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,11,0,0,99,108,97,109,112,0,1,1,0,0,11,0,118,97,108,0,0, -1,1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95, -99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108, -0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,12,0,0,99,108,97,109,112,0,1,1,0,0,12,0,118,97,108, -0,0,1,1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52, -95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97, -108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,10,0,0,99,108,97,109,112,0,1,1,0,0,10,0,118,97, -108,0,0,1,1,0,0,10,0,109,105,110,86,97,108,0,0,1,1,0,0,10,0,109,97,120,86,97,108,0,0,0,1,4,118,101, -99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110, -86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,11,0,0,99,108,97,109,112,0,1,1,0,0,11,0, -118,97,108,0,0,1,1,0,0,11,0,109,105,110,86,97,108,0,0,1,1,0,0,11,0,109,97,120,86,97,108,0,0,0,1,4, -118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109, -105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,12,0,0,99,108,97,109,112,0,1,1,0, -0,12,0,118,97,108,0,0,1,1,0,0,12,0,109,105,110,86,97,108,0,0,1,1,0,0,12,0,109,97,120,86,97,108,0,0, -0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18, -109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,9,0,0,109,105,120,0,1,1,0,0, -9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,109,105,120,0,1,1, -0,0,10,0,120,0,0,1,1,0,0,10,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,11,0,0,109,105,120, -0,1,1,0,0,11,0,120,0,0,1,1,0,0,11,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112, -0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,12,0,0,109, -105,120,0,1,1,0,0,12,0,120,0,0,1,1,0,0,12,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108, -114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0, -0,109,105,120,0,1,1,0,0,10,0,120,0,0,1,1,0,0,10,0,121,0,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101,99,52, -95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0, -0,11,0,0,109,105,120,0,1,1,0,0,11,0,120,0,0,1,1,0,0,11,0,121,0,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101, -99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90, -95,0,0,12,0,0,109,105,120,0,1,1,0,0,12,0,120,0,0,1,1,0,0,12,0,121,0,0,1,1,0,0,12,0,97,0,0,0,1,4, -118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0, -0,0,1,90,95,0,0,9,0,0,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,9,0,120,0,0,0,1,4, -118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0, -0,0,0,1,90,95,0,0,10,0,0,115,116,101,112,0,1,1,0,0,10,0,101,100,103,101,0,0,1,1,0,0,10,0,120,0,0,0, -1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18, -101,100,103,101,0,0,0,0,1,90,95,0,0,11,0,0,115,116,101,112,0,1,1,0,0,11,0,101,100,103,101,0,0,1,1, -0,0,11,0,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,12,0,0,115,116,101,112,0,1,1,0,0,12,0, -101,100,103,101,0,0,1,1,0,0,12,0,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116, -86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,10,0,0,115,116,101,112,0,1,1,0,0,9, -0,101,100,103,101,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,11,0,0,115,116, -101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,90, -95,0,0,12,0,0,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101, -99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1, -90,95,0,0,9,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0,1,1,0, -0,9,0,101,100,103,101,49,0,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9,0,1,116,0,2,58,99,108,97,109, -112,0,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49, -0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,90,95,0,0,10,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,10,0,101, -100,103,101,48,0,0,1,1,0,0,10,0,101,100,103,101,49,0,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0, -1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, -101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, -0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,11,0,0,115,109,111,111,116,104,115,116, -101,112,0,1,1,0,0,11,0,101,100,103,101,48,0,0,1,1,0,0,11,0,101,100,103,101,49,0,0,1,1,0,0,11,0,118, -0,0,0,1,3,2,90,95,0,0,11,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47, -18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8, -18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,0,115, -109,111,111,116,104,115,116,101,112,0,1,1,0,0,12,0,101,100,103,101,48,0,0,1,1,0,0,12,0,101,100,103, -101,49,0,0,1,1,0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0, -18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0, -0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0, -0,1,90,95,0,0,10,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0, -1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0,1,116,0,2,58,99,108, -97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0, -47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0, -0,18,116,0,48,47,48,0,0,1,90,95,0,0,11,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0, -101,100,103,101,48,0,0,1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0, -11,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49, -0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48, -17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,0,115,109,111,111,116,104,115, -116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0,1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,12,0, -118,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0, -47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0, -8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,9,0,0,108, -101,110,103,116,104,0,1,1,0,0,9,0,120,0,0,0,1,8,58,97,98,115,0,0,18,120,0,0,0,0,0,1,90,95,0,0,9,0, -0,108,101,110,103,116,104,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9, -0,1,112,0,2,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0,108,101,110,103,116,104,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90, -95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9,0,1,112,0,2,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0,108,101,110,103,116,104,0,1,1, -0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9,0,1,112,0,2,58,100,111,116,0,0, -18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0, -100,105,115,116,97,110,99,101,0,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,0,1,3,2,90,95,1,0,9,0,1, -100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,0, -18,100,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115,116,97,110,99,101,0,1,1,0,0,10,0,118,0,0,1,1,0,0, -10,0,117,0,0,0,1,3,2,90,95,1,0,10,0,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116, -86,97,108,0,58,108,101,110,103,116,104,0,0,18,100,50,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115, -116,97,110,99,101,0,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,3,2,90,95,1,0,11,0,1,100,51,0,2, -18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,0,18,100, -51,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115,116,97,110,99,101,0,1,1,0,0,12,0,118,0,0,1,1,0,0,12, -0,117,0,0,0,1,3,2,90,95,1,0,12,0,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86, -97,108,0,58,108,101,110,103,116,104,0,0,18,100,52,0,0,0,20,0,0,1,90,95,0,0,11,0,0,99,114,111,115, -115,0,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,9,0,0,102,97, -99,101,102,111,114,119,97,114,100,0,1,1,0,0,9,0,78,0,0,1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,114,101, -102,0,0,0,1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3, -2,90,95,0,0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0, -0,0,8,58,109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,10,0,0,102,97,99,101, -102,111,114,119,97,114,100,0,1,1,0,0,10,0,78,0,0,1,1,0,0,10,0,73,0,0,1,1,0,0,10,0,78,114,101,102,0, -0,0,1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90, -95,0,0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8, -58,109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,11,0,0,102,97,99,101,102, -111,114,119,97,114,100,0,1,1,0,0,11,0,78,0,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,114,101,102,0,0,0, -1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0, -0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58, -109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,12,0,0,102,97,99,101,102,111, -114,119,97,114,100,0,1,1,0,0,12,0,78,0,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0,78,114,101,102,0,0,0,1,3, -2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9, -0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109, -105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,9,0,0,114,101,102,108,101,99,116,0, -1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18, -73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,10,0,0,114,101,102,108,101,99,116,0,1,1,0,0,10,0,73,0,0, -1,1,0,0,10,0,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78, -0,48,47,0,0,1,90,95,0,0,11,0,0,114,101,102,108,101,99,116,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,0, -0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90, -95,0,0,12,0,0,114,101,102,108,101,99,116,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0,78,0,0,0,1,8,18,73,0, -17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,9,0,0,114, -101,102,114,97,99,116,0,1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2, -90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90, -95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95, -100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,9,0,1,114,101, -116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,17,48,0,48,0,0, -20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111, -116,95,105,0,48,58,115,113,114,116,0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97, -108,0,0,0,1,90,95,0,0,10,0,0,114,101,102,114,97,99,116,0,1,1,0,0,10,0,73,0,0,1,1,0,0,10,0,78,0,0,1, -1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0, -18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97, -0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0, -3,2,90,95,0,0,10,0,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116, -118,97,108,0,58,118,101,99,50,0,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116, -97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,0,18,107,0, -0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,11,0,0,114,101,102,114,97, -99,116,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0, -1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1, -107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95, -105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,11,0,1,114,101,116,118,97,108,0, -0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,51,0,0,17,48,0,48, -0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95, -100,111,116,95,105,0,48,58,115,113,114,116,0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116, -118,97,108,0,0,0,1,90,95,0,0,12,0,0,114,101,102,114,97,99,116,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0, -78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111, -116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18, -101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47, -48,47,0,0,3,2,90,95,0,0,12,0,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18, -114,101,116,118,97,108,0,58,118,101,99,52,0,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108, -0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116, -0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,13,0,0,109,97, -116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,13,0,109,0,0,1,0,0,0,13,0,110,0,0,0,1,8,58, -109,97,116,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,14,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0, -1,0,0,0,14,0,109,0,0,1,0,0,0,14,0,110,0,0,0,1,8,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18, -110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,15,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108, -116,0,1,0,0,0,15,0,109,0,0,1,0,0,0,15,0,110,0,0,0,1,8,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0, -0,2,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99, -52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90, -95,0,0,3,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118, -101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, -0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0, -0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0, -1,90,95,0,0,2,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4, -118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, -0,0,0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118,0,0,0, -1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0, -18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0, -118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,90,95,0,0,2,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1, -1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,69,113,117,97, -108,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115, -84,104,97,110,69,113,117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95, -115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,108, -101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118, -101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0, -7,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1, -1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86, -97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101,114,84,104,97,110,0, -1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,103,114,101,97,116,101, -114,84,104,97,110,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0, -103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101, -99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2, -0,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118, -101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0, -18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,7, -0,117,0,0,1,1,0,0,7,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104, -97,110,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101,114,84, -104,97,110,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95, -115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0, -0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,11,0,117,0,0,1,1,0,0, -11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113, -117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101, -114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,101,99,52, -95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95, -0,0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0, -7,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117, -97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,10, -0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,11,0,117, -0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,12,0,117, -0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0, -18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0, -118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118, -0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118, -0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, -0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,2,0,117,0,0,1,1,0,0,2,0,118,0,0,0,1,4,118,101, -99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1, -90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,3,0,117,0,0,1,1,0,0,3,0,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1, -90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,4,0,117,0,0,1,1,0,0,4,0,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0, -110,111,116,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0, -0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, -1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90, -95,0,0,2,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,101, -99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1, -90,95,0,0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, -0,0,0,1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0, -1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1, -90,95,0,0,2,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,2,0,117,0,0,1,1,0,0,2,0,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,90,95,0,0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,3,0,117,0,0,1,1,0,0,3,0,118,0,0,0,1,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, -118,0,0,0,0,1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,4,0,117,0,0,1,1,0,0,4,0,118, -0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, -0,0,1,90,95,0,0,1,0,0,97,110,121,0,1,1,0,0,2,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,115,117,109,0,0,0,4, -118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0, -0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59, -120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,110,121,0,1,1,0,0,3,0,118,0,0,0,1,3,2,90,95,0,0, -9,0,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, -120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115, -117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,110, -121,0,1,1,0,0,4,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100, -0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100, -100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99, -52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120, -0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,2,0,118,0,0,0,1,3,2,90,95,0,0,9,0, -1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0, -0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,3, -0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0, -48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,4,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,112,114, -111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0, -59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114, -111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52, -95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1, -90,95,0,0,2,0,0,110,111,116,0,1,1,0,0,2,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,0,110,111, -116,0,1,1,0,0,3,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,0,110,111,116,0,1,1,0,0,4,0,118,0, -0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0, -0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,0,1,1,0,0,16,0,115,97,109,112,108,101, -114,0,0,1,1,0,0,9,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,49,100,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95, -0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,0,115,97,109,112,108,101, -114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,49,100,95,112, -114,111,106,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114, -111,106,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4, -118,101,99,52,95,116,101,120,95,49,100,95,112,114,111,106,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116, -117,114,101,50,68,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0, -0,0,1,4,118,101,99,52,95,116,101,120,95,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, -112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101, -50,68,80,114,111,106,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100, -0,0,0,1,4,118,101,99,52,95,116,101,120,95,50,100,95,112,114,111,106,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,90, -95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,0,17,0,115,97,109,112,108, -101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,50,100,95, -112,114,111,106,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,0,1,1,0,0,18,0,115,97,109, -112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,51, -100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0, -0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,0,18,0,115,97, -109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95, -51,100,95,112,114,111,106,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0, -0,19,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, -116,101,120,95,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0, -0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,0,1,1,0,0,20,0,115, -97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120, -95,49,100,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,80,114,111, -106,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118, -101,99,52,95,116,101,120,95,49,100,95,112,114,111,106,95,115,104,97,100,111,119,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0, -0,115,104,97,100,111,119,50,68,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111, -111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,50,100,95,115,104,97,100,111,119,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95, -0,0,12,0,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0, -0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,50,100,95,112,114,111, -106,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0, -0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,82,101,99,116, -0,1,1,0,0,22,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,0,1,4,118,101,99, -52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,82,101, -99,116,80,114,111,106,0,1,1,0,0,22,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114, -100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,95,112,114,111,106,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0, -0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,22, -0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116, -101,120,95,114,101,99,116,95,112,114,111,106,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, -112,108,101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,50, -68,82,101,99,116,0,1,1,0,0,23,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0, -0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,95,115,104,97,100,111,119,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0, -0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,23,0,115,97,109,112,108,101, -114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116, -95,112,114,111,106,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, -112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,9,0,0,110,111,105,115,101,49,0,1,1,0, -0,9,0,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108, -0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,0,110,111,105,115,101,49,0,1,1,0,0,10,0,120,0,0,0,1,4,102,108, -111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0, -0,9,0,0,110,111,105,115,101,49,0,1,1,0,0,11,0,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115, -101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,0,110,111,105,115,101, -49,0,1,1,0,0,12,0,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101, -116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,110,111,105,115,101,50,0,1,1,0,0,9,0,120,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0, -46,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,105,115,101,50,0,1,1,0,0,10,0,120,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,49,57,0,51, -52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,105,115,101,50,0,1,1,0,0, -11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120, -0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58, -118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1, -90,95,0,0,10,0,0,110,111,105,115,101,50,0,1,1,0,0,12,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,49,57,0,51,52,0,0,0,17,55, -0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,0,110,111, -105,115,101,51,0,1,1,0,0,9,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, -105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105, -115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122, -0,58,110,111,105,115,101,49,0,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,0,110, -111,105,115,101,51,0,1,1,0,0,10,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110, -111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111, -105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118, -101,99,50,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,0,110, -111,105,115,101,51,0,1,1,0,0,11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110, -111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111, -105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51, -0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49, -0,0,18,120,0,58,118,101,99,51,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0, -0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,0,110,111,105,115,101,51,0,1,1,0,0,12,0,120,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,53, -0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,0, -1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,9,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120,0,17,50,51,0,53, -52,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,10,0,120,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110, -111,105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120,0,58, -118,101,99,50,0,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,0, -110,111,105,115,101,52,0,1,1,0,0,11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58, -110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, -111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17, -51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101, -49,0,0,18,120,0,58,118,101,99,51,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0, -0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120, -0,58,118,101,99,51,0,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0, -0,20,0,0,1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,12,0,120,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,49,57,0,51,52,0,0, -0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,53,0,52,55,0,0, -0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,50, -51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20, -0,0,0 diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h deleted file mode 100644 index b3d3e87cf42..00000000000 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ /dev/null @@ -1,869 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_core.gc */ - -5,1,90,95,0,0,5,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18, -95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,5,0,1,1,1,0,0,1,0,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,98,0,20,0,0,1,90,95,0,0,5,0,1,1,1,0,0,5,0,105,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,105,0,20,0,0,1,90,95,0,0,1,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,1, -1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102, -0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,1,1,1,0,0,1,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,98,0,20,0,0,1,90,95,0,0,9,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118, -101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,9,0,1,1,1,0,0,1,0,98,0,0, -0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0, -0,0,1,90,95,0,0,9,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1, -90,95,0,0,10,0,1,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,0,1,90,95,0,0,10,0,1, -1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,102,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101,99,52,95,116, -111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,0,0, -10,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,2,0,98,0,0,0,1,4,105,118, -101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0, -0,0,1,90,95,0,0,10,0,1,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,12,0,118, -0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -118,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0, -122,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,90,95,0, -0,11,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101, -99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0, -0,0,1,90,95,0,0,11,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,3,0,98,0, -0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,109,111, -118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,90,95,0,0,12,0,1,1, -1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0,122,0,0,1,1,0,0,9,0,119,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -119,0,18,119,0,20,0,0,1,90,95,0,0,12,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,5,0,105,0,0,0,1, -4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0, -0,1,90,95,0,0,12,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18, -95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,4,0,98,0,0,0,1,4,105,118, -101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0, -0,12,0,1,1,1,0,0,8,0,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114, -101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,11,0,118,51,0,0,1,1,0,0,9,0,102,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,119,0,18,102,0,20,0,0,1,90,95,0,0,12,0,1,1,1,0,0,10,0,118,50,0,0,1,1,0,0,9,0,102,49, -0,0,1,1,0,0,9,0,102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -119,0,18,102,50,0,20,0,0,1,90,95,0,0,6,0,1,1,1,0,0,5,0,105,0,0,1,1,0,0,5,0,106,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106, -0,20,0,0,1,90,95,0,0,6,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,0,0,6,0,1,1,1,0,0,9,0,102,0,0,0,1,4, -118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -102,0,0,0,0,1,90,95,0,0,6,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99, -52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,1,1,1,0,0,5,0, -105,0,0,1,1,0,0,5,0,106,0,0,1,1,0,0,5,0,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, -105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,18,107,0,20,0,0,1,90,95,0,0,7,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109, -111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,0,0,1,90,95,0,0,7,0,1, -1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,7,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99, -52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95, -0,0,8,0,1,1,1,0,0,5,0,120,0,0,1,1,0,0,5,0,121,0,0,1,1,0,0,5,0,122,0,0,1,1,0,0,5,0,119,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, -18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,119,0,18,119,0,20,0,0,1,90,95,0,0,8,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109, -111,118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,8,0,1,1,1,0,0,9,0,102, -0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18, -102,0,0,0,0,1,90,95,0,0,8,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99, -52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,1,0,98,49,0,0,1,1, -0,0,1,0,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,90,95,0,0,2,0,1,1,1,0,0,5,0,105,49,0,0,1,1,0,0,5, -0,105,50,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -105,49,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -121,0,0,18,105,50,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99, -52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0, -2,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0, -0,1,90,95,0,0,2,0,1,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,6,0,118,0,0,0, -1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,1,0,98,49,0,0,1,1,0,0,1,0,98,50,0,0,1,1,0,0,1,0,98,51, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,90, -95,0,0,3,0,1,1,1,0,0,9,0,102,49,0,0,1,1,0,0,9,0,102,50,0,0,1,1,0,0,9,0,102,51,0,0,0,1,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,17,48,0,48,0,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,17,48,0, -48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51, -0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,109,111,118, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0, -9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90, -95,0,0,3,0,1,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,7,0,118,0,0,0,1, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,1,0,98,49,0,0,1,1,0,0,1,0,98,50,0,0,1,1,0,0,1,0,98,51, -0,0,1,1,0,0,1,0,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18, -98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,90,95,0,0,4,0,1,1,1,0, -0,9,0,102,49,0,0,1,1,0,0,9,0,102,50,0,0,1,1,0,0,9,0,102,51,0,0,1,1,0,0,9,0,102,52,0,0,0,1,3,2,90, -95,1,0,9,0,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,18,122,101,114,111,0,0,0,4,118,101, -99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,18,122,101,114, -111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,102,52,0, -0,18,122,101,114,111,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,109,111, -118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,98,0,0,0,0,1,90,95,0,0,4,0,1, -1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,119,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,105,0,0,17, -48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4, -0,1,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0,0,9,0,109,48,48,0,0,1, -1,0,0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,48,49,0,0,1,1,0,0,9,0,109,49,49,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18, -109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1, -90,95,0,0,13,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109, -97,116,50,0,0,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0,0,1,0,98,0,0, -0,1,8,58,109,97,116,50,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0, -0,10,0,99,48,0,0,1,1,0,0,10,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99, -48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,0,0,14,0,1,1, -1,0,0,9,0,109,48,48,0,0,1,1,0,0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,50,48,0,0,1,1,0,0,9,0,109,48,49, -0,0,1,1,0,0,9,0,109,49,49,0,0,1,1,0,0,9,0,109,50,49,0,0,1,1,0,0,9,0,109,48,50,0,0,1,1,0,0,9,0,109, -49,50,0,0,1,1,0,0,9,0,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109, -50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,9,0,102,0,0,0,1,3,2, -90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118, -0,59,121,121,120,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109,97,116,51,0,0,58,102, -108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,14,0,1,1,1,0,0,1,0,98,0,0,0,1,8,58,109,97,116, -51,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,14,0,1,1,1,0,0,11,0,99,48,0,0,1,1, -0,0,11,0,99,49,0,0,1,1,0,0,11,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,9,0,109,48,48,0,0,1,1,0, -0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,50,48,0,0,1,1,0,0,9,0,109,51,48,0,0,1,1,0,0,9,0,109,48,49,0,0, -1,1,0,0,9,0,109,49,49,0,0,1,1,0,0,9,0,109,50,49,0,0,1,1,0,0,9,0,109,51,49,0,0,1,1,0,0,9,0,109,48, -50,0,0,1,1,0,0,9,0,109,49,50,0,0,1,1,0,0,9,0,109,50,50,0,0,1,1,0,0,9,0,109,51,50,0,0,1,1,0,0,9,0, -109,48,51,0,0,1,1,0,0,9,0,109,49,51,0,0,1,1,0,0,9,0,109,50,51,0,0,1,1,0,0,9,0,109,51,51,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51, -48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48, -50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57, -59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49, -51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,9, -0,102,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0, -9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51, -0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109,97, -116,52,0,0,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,15,0,1,1,1,0,0,1,0,98,0,0,0, -1,8,58,109,97,116,52,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,15,0,1,1,1,0,0, -12,0,99,48,0,0,1,1,0,0,12,0,99,49,0,0,1,1,0,0,12,0,99,50,0,0,1,1,0,0,12,0,99,51,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,5,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0, -0,5,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, -98,0,0,0,0,1,90,95,0,0,5,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0, -5,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,5,0,2,22,1,1,0,0,5,0, -97,0,0,1,1,0,0,5,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99, -52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,6,0,2,26,1,1,0,0,6,0,97,0,0,1,1, -0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, -98,0,0,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,97,0,0,1,1,0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0, -6,0,2,21,1,1,0,0,6,0,97,0,0,1,1,0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,6,0,2,22,1,1,0,0,6,0, -97,0,0,1,1,0,0,6,0,98,0,0,0,1,3,2,90,95,0,0,10,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105, -118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,7,0, -97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0, -18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, -1,90,95,0,0,7,0,2,21,1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,2,22, -1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,3,2,90,95,0,0,11,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95, -105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,8,0,2,26,1,1,0,0, -8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, -1,90,95,0,0,8,0,2,21,1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,0,2,22, -1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,3,2,90,95,0,0,12,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118, -101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,2,26,1,1,0,0,9,0,97, -0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99, -52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1, -90,95,0,0,9,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,2,22, -1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,98,73,110,118,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,90,95,0, -0,10,0,2,26,1,1,0,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10, -0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,10,0, -118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,22,1,1,0,0,10,0, -118,0,0,1,1,0,0,10,0,117,0,0,0,1,3,2,90,95,0,0,10,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, -0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0,0,11,0,118,0,0,1, -1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,27,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0, -0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0, -0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,22,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117, -0,0,0,1,3,2,90,95,0,0,11,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18, -117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0, -18,119,0,0,0,0,1,90,95,0,0,12,0,2,26,1,1,0,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,27, -1,1,0,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,12,0,118,0, -0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,22,1,1,0,0,12,0,118,0,0,1,1,0,0,12, -0,117,0,0,0,1,3,2,90,95,0,0,12,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120, -0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121, -0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,10, -0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,26,1,1, -0,0,10,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,9,0,97,0, -0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10, -0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,2,21,1,1, -0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2, -21,1,1,0,0,10,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0, -0,10,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,3,2,90,95,0,0,10,0,1,105,110,118,85,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0, -18,105,110,118,85,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,22,1,1,0,0,10,0,118,0,0,1,1,0,0,9,0,98,0, -0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0, -0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0,0, -11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,2,27,1,1,0,0,9,0, -97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,27, -1,1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95, -0,0,11,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121, -122,0,0,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59, -120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,3, -2,90,95,0,0,11,0,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0, -59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0, -0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117, -0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,22,1, -1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18, -105,110,118,66,0,0,0,0,1,90,95,0,0,12,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0, -2,26,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0, -117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0, -0,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0, -12,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,2,22,1,1,0, -0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,3,2,90,95,0,0,12,0,1,105,110,118,85,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105,110,118,85,0,0,0,0,1,90,95,0,0,12,0,2,22, -1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,0,0,0,1,90,95,0,0, -6,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,50,0,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,6,0,2,26,1,1,0,0,6,0,118,0,0,1,1,0,0,5,0, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,0,18,98,0,0,0,46,20, -0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,118, -0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,0,18, -98,0,0,0,47,20,0,0,1,90,95,0,0,6,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,6,0,2,21,1, -1,0,0,6,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118, -101,99,50,0,0,18,98,0,0,0,48,20,0,0,1,90,95,0,0,6,0,2,22,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90, -95,0,0,6,0,2,22,1,1,0,0,6,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,58,105,118,101,99,50,0,0,18,98,0,0,0,49,20,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0, -0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,0,18,97,0,0,0,18,117,0, -46,20,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,46,20,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,5, -0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,0,18,97, -0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,47,20,0,0,1,90,95,0,0,7,0, -2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101, -99,51,0,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,7,0,2,21,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,48,20,0,0,1, -90,95,0,0,7,0,2,22,1,1,0,0,5,0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,105,118,101,99,51,0,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,95,0,0,7,0,2,22,1,1,0,0,7,0,118,0,0,1, -1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0, -0,0,49,20,0,0,1,90,95,0,0,8,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,8,0,2,26,1,1,0, -0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99, -52,0,0,18,98,0,0,0,46,20,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0, -8,0,2,27,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58, -105,118,101,99,52,0,0,18,98,0,0,0,47,20,0,0,1,90,95,0,0,8,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0, -117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,48,20, -0,0,1,90,95,0,0,8,0,2,21,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,58,105,118,101,99,52,0,0,18,98,0,0,0,48,20,0,0,1,90,95,0,0,8,0,2,22,1,1,0,0,5,0,97, -0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0, -0,18,117,0,49,20,0,0,1,90,95,0,0,8,0,2,22,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,0,18,98,0,0,0,49,20,0,0,1,90,95,0,0,5,0,2, -27,1,1,0,0,5,0,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,97,0,0,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,118,0,0,0,1,4,118,101,99,52,95,110, -101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,7,0,2,27,1,1,0, -0,7,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0, -18,118,0,0,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,9,0,2,27,1,1,0,0,9,0,97,0,0,0, -1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0, -0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,2,27, -1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,12,0,118,0, -0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0, -0,1,90,95,0,0,13,0,2,27,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,54,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54, -20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,90,95,0,0,9,0,0, -100,111,116,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, -18,98,0,48,20,0,0,1,90,95,0,0,9,0,0,100,111,116,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121, -0,48,46,20,0,0,1,90,95,0,0,9,0,0,100,111,116,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118, -101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0, -9,0,0,100,111,116,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,5,0,2,1,1,0,2,0,5,0,97,0,0, -1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,46,20,0,8,18,97,0,0,0,1,90,95,0,0,5,0,2,2,1,0,2,0, -5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,47,20,0,8,18,97,0,0,0,1,90,95,0,0,5,0, -2,3,1,0,2,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,48,20,0,8,18,97,0,0,0,1,90, -95,0,0,5,0,2,4,1,0,2,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,49,20,0,8,18,97, -0,0,0,1,90,95,0,0,6,0,2,1,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0, -46,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,2,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,118,0, -18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,3,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0, -0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,4,1,0,2,0,6,0,118,0,0, -1,1,0,0,6,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,1,1, -0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90, -95,0,0,7,0,2,2,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8, -18,118,0,0,0,1,90,95,0,0,7,0,2,3,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0, -18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,4,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9, -18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,1,1,0,2,0,8,0,118,0,0,1,1,0,0, -8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,2,1,0,2,0,8, -0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0, -8,0,2,3,1,0,2,0,8,0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18,118, -0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,8,0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0, -49,20,0,8,18,118,0,0,0,1,90,95,0,0,9,0,2,1,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,97,0,18, -97,0,18,98,0,46,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,2,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9, -18,97,0,18,97,0,18,98,0,47,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,3,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0, -98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,48,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,4,1,0,2,0,9,0,97,0,0, -1,1,0,0,9,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,49,20,0,8,18,97,0,0,0,1,90,95,0,0,10,0,2,1,1,0,2, -0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90, -95,0,0,10,0,2,2,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0, -8,18,118,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18, -118,0,18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,4,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0, -0,0,1,9,18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,1,1,0,2,0,11,0,118,0, -0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2, -2,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0, -0,1,90,95,0,0,11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0, -48,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,4,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118, -0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,1,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0, -117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,2,1,0,2,0,12,0, -118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0, -12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18, -118,0,0,0,1,90,95,0,0,12,0,2,4,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0, -18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,1,1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9, -18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,2, -1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0, -0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,3,1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0, -18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,4,1,0,2,0,6, -0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,49,20,0, -8,18,118,0,0,0,1,90,95,0,0,7,0,2,1,1,0,2,0,7,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0, -58,105,118,101,99,51,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,2,1,0,2,0,7,0,118,0, -0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,18,97,0,0,0,47,20,0,8,18,118, -0,0,0,1,90,95,0,0,7,0,2,3,1,0,2,0,7,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105, -118,101,99,51,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,7,0,118,0,0,1,1,0, -0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1, -90,95,0,0,8,0,2,1,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99, -52,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,2,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0, -0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0, -8,0,2,3,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,18, -97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9, -18,118,0,18,118,0,58,105,118,101,99,52,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,1, -1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,18,97,0,0,0, -46,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,2,1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0, -18,118,0,58,118,101,99,50,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10,0, -118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,18,97,0,0,0,48,20,0,8,18, -118,0,0,0,1,90,95,0,0,10,0,2,4,1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,50,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,1,1,0,2,0,11,0,118,0,0,1,1, -0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1, -90,95,0,0,11,0,2,2,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99, -51,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97, -0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,11, -0,2,4,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0, -0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,1,1,0,2,0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18, -118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,2,1,0,2, -0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,47,20,0, -8,18,118,0,0,0,1,90,95,0,0,12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118, -0,58,118,101,99,52,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,4,1,0,2,0,12,0,118,0, -0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0, -0,1,90,95,0,0,13,0,2,26,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,90,95,0,0,13,0,2, -27,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,13,0,109, -0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0, -57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,90,95, -0,0,13,0,2,22,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,14,0,2,26,1,1,0, -0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, -0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,1, -1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, -0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,14,0,109,0,0,1,1,0,0,14,0,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109, -0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49, -0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57, -59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121, -121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,90, -95,0,0,14,0,2,22,1,1,0,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,90,95,0,0,15,0,2,26, -1,1,0,0,15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0, -15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,15,0,109, -0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121, -121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109, -0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0, -16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16, -10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119, -119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0, -16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121, -121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16, -10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16, -10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10, -51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119, -119,0,48,46,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, -49,20,0,0,1,90,95,0,0,13,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,90,95,0,0,13,0,2,26,1,1,0,0,13,0,109,0,0, -1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0, -0,1,90,95,0,0,13,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,90,95,0,0,13,0,2,27,1,1,0,0,13,0,109,0,0,1,1,0, -0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1, -90,95,0,0,13,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,13,0,109,0,0,1,1,0,0,9,0, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0, -0,13,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,13,0,2,22,1,1,0,0,13,0,109,0,0,1,1,0,0,9,0,98,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,90,95,0,0,14,0,2, -26,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, -0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, -0,57,46,20,0,0,1,90,95,0,0,14,0,2,26,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,9,0,97,0,0,1, -1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0, -57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,90,95, -0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,98,0,47,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,14,0, -109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0, -1,90,95,0,0,14,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,49,20,0,0,1,90,95,0,0,14,0,2,22,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,90,95,0,0,15,0,2,26,1,1,0, -0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1, -90,95,0,0,15,0,2,26,1,1,0,0,15,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, -0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51, -0,57,18,98,0,46,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,15,0,109,0,0,1,1, -0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,90,95,0,0,15,0,2, -21,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, -0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20, -0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16, -10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,15,0,109,0,0, -1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98, -0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0,0,1,90,95,0,0,10,0,2, -21,1,1,0,0,13,0,109,0,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8, -48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,90,95, -0,0,10,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,11,0,2, -21,1,1,0,0,14,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8, -48,0,57,18,118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109, -0,16,10,50,0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0, -1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0, -18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18, -118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116, -0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0, -0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120, -120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, -118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119,119,119,0,48,46,20,0,0, -1,90,95,0,0,12,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0, -0,1,90,95,0,0,13,0,2,1,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,2,1,0,2,0,13,0,109,0,0,1,1,0,0,13, -0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,109, -0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,8,18,109,0,0,0,1,90,95,0,0, -13,0,2,3,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,8,18, -109,0,0,0,1,90,95,0,0,13,0,2,4,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,110,0,16,10,49,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,1,1,0,2,0,14,0,109,0,0,1,1,0, -0,14,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9, -18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,109,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,2,1,0,2, -0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16, -8,48,0,57,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9, -18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,8,18,109,0,0,0,1,90, -95,0,0,14,0,2,3,1,0,2,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0, -8,18,109,0,0,0,1,90,95,0,0,14,0,2,4,1,0,2,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16, -10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0, -16,10,50,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,1,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0, -16,10,51,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,2,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0, -16,10,51,0,57,47,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,3,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0, -0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,4,1,0,2,0,15,0,109,0, -0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,109,0,16, -10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,109,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,1,1,0,2,0,13,0,109, -0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18, -109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,118,0,46,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,2,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0, -0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,118,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,47,20, -0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,3,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10, -0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90, -95,0,0,13,0,2,4,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118, -101,99,50,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90, -95,0,0,14,0,2,1,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118, -101,99,51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9,18, -109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,118,0,46,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,2,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0, -97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99,51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,118,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0, -47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,47,20,0,8,18,109,0,0,0,1,90,95,0, -0,14,0,2,3,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99, -51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0, -16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0, -57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,4,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0, -1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99,51,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,8,18,109,0,0, -0,1,90,95,0,0,15,0,2,1,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,118,0,2,58, -118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9, -18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,118,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,118,0,46,20,0,8, -18,109,0,0,0,1,90,95,0,0,15,0,2,2,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1, -118,0,2,58,118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118, -0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,47,20,0,9,18,109,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,118,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,118, -0,47,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,3,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90, -95,0,0,12,0,1,118,0,2,58,118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8, -48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,9,18,109, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51, -0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,4,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0, -0,1,3,2,90,95,0,0,12,0,1,118,0,2,58,118,101,99,52,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,9,18,109,0,16, -10,51,0,57,18,109,0,16,10,51,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10, -0,118,0,0,1,1,0,0,13,0,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0, -11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,14,0,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,8,18, -118,0,0,0,1,90,95,0,0,12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,15,0,109,0,0,0,1,9,18,118,0,18,118,0, -18,109,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,5,0,2,25,1,0,2,0,5,0,97,0,0,0,1,9,18,97,0,18,97,0,16, -10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,6,0,2,25,1,0,2,0,6,0, -118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,7,0,2,25,1,0,2,0,7,0,118,0,0,0,1,9,18,118,0,18,118,0, -58,105,118,101,99,51,0,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0, -1,90,95,0,0,8,0,2,25,1,0,2,0,8,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,16,10,49, -0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,9,0,2,25,1,0,2,0,9,0, -97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20, -0,0,1,90,95,0,0,10,0,2,25,1,0,2,0,10,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49, -0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,11,0,2,25,1,0, -2,0,11,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,12,0,2,25,1,0,2,0,12,0,118,0,0,0,1,9,18,118, -0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,90,95,0,0,13,0,2,25,1,0,2,0,13,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8, -48,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0, -20,0,0,1,90,95,0,0,14,0,2,25,1,0,2,0,14,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1, -90,95,0,0,15,0,2,25,1,0,2,0,15,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118, -101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118, -101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118, -101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, -101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90, -95,0,0,5,0,2,24,1,0,2,0,5,0,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,6,0,2,24,1,0,2,0,6,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105, -118,101,99,50,0,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90, -95,0,0,7,0,2,24,1,0,2,0,7,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,8,0,2,24,1,0,2,0,8,0,118, -0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,9,0,2,24,1,0,2,0,9,0,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0, -48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,10,0,2,24,1,0,2,0,10, -0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,11,0,2,24,1,0,2,0,11,0,118,0,0,0,1,9,18,118,0,18, -118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,0,1,90,95,0,0,12,0,2,24,1,0,2,0,12,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,13,0,2,24,1, -0,2,0,13,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,0,17,49,0,48, -0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,14,0,2,24,1,0,2,0, -14,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0, -0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,0,2,24,1,0,2,0,15,0,109, -0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20, -0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0, -9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,5,0,0,95,95,112,111,115,116,68,101,99, -114,0,1,0,2,0,5,0,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, -10,49,0,47,20,0,0,1,90,95,0,0,6,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,6,0,118,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,16, -10,49,0,0,0,47,20,0,0,1,90,95,0,0,7,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,7,0,118,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0, -16,10,49,0,0,0,47,20,0,0,1,90,95,0,0,8,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,8,0,118,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0, -0,16,10,49,0,0,0,47,20,0,0,1,90,95,0,0,9,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,9,0,97, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0, -1,90,95,0,0,10,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,10,0,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47, -20,0,0,1,90,95,0,0,11,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,11,0,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0, -0,47,20,0,0,1,90,95,0,0,12,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,12,0,118,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,90,95,0,0,13,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,13,0,109,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58, -118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58, -118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,14,0,0,95,95,112,111,115,116,68,101,99, -114,0,1,0,2,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48, -0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,15,0,0,95,95, -112,111,115,116,68,101,99,114,0,1,0,2,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109, -0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20, -0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9, -18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1, -90,95,0,0,9,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,9,0,97,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,95,0,0,10,0,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0,11,0,0,95, -95,112,111,115,116,73,110,99,114,0,1,0,2,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0,12,0, -0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0, -5,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,5,0,97,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,95,0,0,6,0,0,95,95,112,111,115,116, -73,110,99,114,0,1,0,2,0,6,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, -0,18,118,0,58,105,118,101,99,50,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,7,0,0,95,95,112,111,115, -116,73,110,99,114,0,1,0,2,0,7,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18, -118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,8,0,0,95,95,112,111, -115,116,73,110,99,114,0,1,0,2,0,8,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,13,0,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,0,13,0,109,0,0,0,1,3,2,90,95,0,0,13,0,1,110,0,2,18,109,0,0,0,9, -18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18, -109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,8,18, -110,0,0,0,1,90,95,0,0,14,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,14,0,109,0,0,0,1,3,2,90, -95,0,0,14,0,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51, -0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0, -0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,0, -17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,90,95,0,0,15,0,0,95,95,112,111,115,116,73,110,99,114,0, -1,0,2,0,15,0,109,0,0,0,1,3,2,90,95,0,0,15,0,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10, -51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,90,95,0,0,1,0,2,15,1,1,0, -0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,90,95,0,0,1,0,2,15,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0, -0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,40,0,0,1,90,95, -0,0,1,0,2,16,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,1,0,1,99,0,0,0,4,102,108,111, -97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,90,95,0,0,1,0,2,16,1,1, -0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97, -116,0,0,18,98,0,0,0,41,0,0,1,90,95,0,0,1,0,2,18,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90, -95,0,0,1,0,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0, -18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103, -0,18,101,0,32,0,0,1,90,95,0,0,1,0,2,18,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,8,58,102,108,111, -97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,43,0,0,1,90,95,0,0,1,0,2,17,1,1,0,0,9, -0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,1,0,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95, -108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18, -101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,90,95,0,0,1,0,2,17,1,1,0,0,5,0,97,0,0,1, -1,0,0,5,0,98,0,0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0, -42,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,9,0,102,0,0,0,1,4,102,108,111, -97,116,95,112,114,105,110,116,0,18,102,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0, -1,1,0,0,5,0,105,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,90,95,0,0,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,0,1,0,98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0, -18,98,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,10,0,118,0,0,0,1,9,58, -112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,11,0,118,0, -0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0, -0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,12,0,118,0,0,0,1,9,58,112,114, -105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,0,6,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110, -116,77,69,83,65,0,1,1,0,0,7,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120, -0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116, -77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, -0,0,8,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18, -118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0, -0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,2,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69, -83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0, -0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,3,0,118,0,0,0,1,9,58,112,114,105,110, -116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59, -121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,0,4,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0, -18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112, -114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0, -18,118,0,59,119,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,13,0,109,0,0, -0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116, -77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0, -1,1,0,0,14,0,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58, -112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,0,18,109,0,16,10,50,0,57,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0, -15,0,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0, -18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,10,51,0,57,0,0,0, -0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,16,0,101,0,0,0,1,4,105,110,116,95, -112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0, -17,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,0,18,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, -0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,19,0,101,0,0,0,1,4,105,110,116, -95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0, -0,20,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,0,21,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, -0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h b/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h deleted file mode 100644 index c5a1cce2a49..00000000000 --- a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h +++ /dev/null @@ -1,110 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_fragment_builtin.gc */ - -5,2,2,90,95,6,0,12,0,1,103,108,95,70,114,97,103,67,111,111,114,100,0,0,0,2,2,90,95,6,0,1,0,1,103, -108,95,70,114,111,110,116,70,97,99,105,110,103,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,70,114,97,103, -67,111,108,111,114,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,70,114,97,103,68,97,116,97,0,3,18,103,108, -95,77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,0,0,2,2,90,95,5,0,9,0,1,103,108,95,70,114, -97,103,68,101,112,116,104,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95, -3,0,12,0,1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0, -1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101, -67,111,111,114,100,115,0,0,0,2,2,90,95,3,0,9,0,1,103,108,95,70,111,103,70,114,97,103,67,111,111, -114,100,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,0,1,1,0,0,16,0,115,97,109,112, -108,101,114,0,0,1,1,0,0,9,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0, -12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0, -20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120, -95,49,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111, -106,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0, -98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114, -100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112, -99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95, -98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111, -111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0, -16,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115, -0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0, -18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114, -100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105,97,115,0, -18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0, -0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,0,1,1,0,0,17,0,115,97,109,112,108,101,114, -0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99, -111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120, -121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116, -101,120,95,50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80, -114,111,106,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1, -0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111, -111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0, -49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101, -120,95,50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114, -0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80,114, -111,106,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0, -9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111, -114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20, -0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95, -50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, -112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,0,1,1,0,0,18,0, -115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0, -1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0, -18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97, -115,0,20,0,4,118,101,99,52,95,116,101,120,95,51,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116, -101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,0,18,0,115,97,109,112,108,101,114,0,0,1,1,0,0, -12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111, -114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121, -122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97, -115,0,20,0,4,118,101,99,52,95,116,101,120,95,51,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116, -101,120,116,117,114,101,67,117,98,101,0,1,1,0,0,19,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0, -99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100, -52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111, -111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95,99,117,98,101, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0, -0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0, -0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99, -111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20, -0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95, -49,100,95,98,105,97,115,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97, -109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119, -49,68,80,114,111,106,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100, -0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112, -99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49, -20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111, -111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105, -97,115,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,50,68,0,1,1,0, -0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97, -115,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120, -121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20, -0,4,118,101,99,52,95,116,101,120,95,50,100,95,98,105,97,115,95,115,104,97,100,111,119,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90, -95,0,0,12,0,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,0,21,0,115,97,109,112,108,101, -114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1, -112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59, -120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99, -111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4, -118,101,99,52,95,116,101,120,95,50,100,95,98,105,97,115,95,115,104,97,100,111,119,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95, -0,0,9,0,0,100,70,100,120,0,1,1,0,0,9,0,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,10,0,0,100,70,100, -120,0,1,1,0,0,10,0,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,112,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,11,0,0,100,70,100,120,0,1,1,0,0,11, -0,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,112,0,59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,0,100,70,100,120,0,1,1,0,0,12,0,112,0,0,0,1,4, -118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,90,95,0,0,9,0,0, -100,70,100,121,0,1,1,0,0,9,0,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,10,0,0,100,70,100,121,0,1,1,0, -0,10,0,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, -0,18,112,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,11,0,0,100,70,100,121,0,1,1,0,0,11,0,112,0,0,0,1, -4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,59, -120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,0,100,70,100,121,0,1,1,0,0,12,0,112,0,0,0,1,4,118,101,99, -52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,90,95,0,0,9,0,0,102,119, -105,100,116,104,0,1,1,0,0,9,0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0, -0,58,97,98,115,0,0,58,100,70,100,121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,10,0,0,102,119,105, -100,116,104,0,1,1,0,0,10,0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0, -58,97,98,115,0,0,58,100,70,100,121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,11,0,0,102,119,105,100, -116,104,0,1,1,0,0,11,0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97, -98,115,0,0,58,100,70,100,121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,12,0,0,102,119,105,100,116, -104,0,1,1,0,0,12,0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97,98, -115,0,0,58,100,70,100,121,0,0,18,112,0,0,0,0,0,46,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_pp_directives.syn b/src/mesa/shader/slang/library/slang_pp_directives.syn deleted file mode 100644 index b51d168cc03..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_directives.syn +++ /dev/null @@ -1,405 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.6 - * - * Copyright (C) 2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file slang_pp_directives.syn - * slang preprocessor directives parser - * \author Michal Krol - */ - -.syntax source; - -/* - * This syntax script preprocesses a GLSL shader. - * It is assumed, that the #version directive has been parsed. Separate pass for parsing - * version gives better control on behavior depending on the version number given. - * - * The output is a source string with comments and directives removed. White spaces and comments - * are replaced with on or more spaces. All new-lines are preserved and converted to Linux format. - * Directives are escaped with a null character. The end of the source string is marked by - * two consecutive null characters. The consumer is responsible for executing the escaped - * directives, removing dead portions of code and expanding macros. - */ - -.emtcode ESCAPE_TOKEN 0 - -/* - * The TOKEN_* symbols follow the ESCAPE_TOKEN. - * - * NOTE: - * There is no TOKEN_IFDEF and neither is TOKEN_IFNDEF. They are handled with TOKEN_IF and - * operator defined. - * The "#ifdef SYMBOL" is replaced with "#if defined SYMBOL" - * The "#ifndef SYMBOL" is replaced with "#if !defined SYMBOL" - */ -.emtcode TOKEN_END 0 -.emtcode TOKEN_DEFINE 1 -.emtcode TOKEN_UNDEF 2 -.emtcode TOKEN_IF 3 -.emtcode TOKEN_ELSE 4 -.emtcode TOKEN_ELIF 5 -.emtcode TOKEN_ENDIF 6 -.emtcode TOKEN_ERROR 7 -.emtcode TOKEN_PRAGMA 8 -.emtcode TOKEN_EXTENSION 9 -.emtcode TOKEN_LINE 10 - -/* - * The PARAM_* symbols follow the TOKEN_DEFINE. - */ -.emtcode PARAM_END 0 -.emtcode PARAM_PARAMETER 1 - -/* - * The BEHAVIOR_* symbols follow the TOKEN_EXTENSION. - */ -.emtcode BEHAVIOR_REQUIRE 1 -.emtcode BEHAVIOR_ENABLE 2 -.emtcode BEHAVIOR_WARN 3 -.emtcode BEHAVIOR_DISABLE 4 - -/* - * The PRAGMA_* symbols follow TOKEN_PRAGMA - */ -.emtcode PRAGMA_NO_PARAM 0 -.emtcode PRAGMA_PARAM 1 - -source - optional_directive .and .loop source_element .and '\0' .emit ESCAPE_TOKEN .emit TOKEN_END; - -source_element - c_style_comment_block .or cpp_style_comment_block .or new_line_directive .or source_token; - -c_style_comment_block - '/' .and '*' .and c_style_comment_rest .and .true .emit ' '; - -c_style_comment_rest - .loop c_style_comment_body .and c_style_comment_end; - -c_style_comment_body - c_style_comment_char_nostar .or c_style_comment_char_star_noslashstar; - -c_style_comment_char_nostar - new_line .or '\x2B'-'\xFF' .or '\x01'-'\x29'; - -c_style_comment_char_star_noslashstar - '*' .and c_style_comment_char_star_noslashstar_1; -c_style_comment_char_star_noslashstar_1 - c_style_comment_char_noslashstar .or c_style_comment_char_star_noslashstar; - -c_style_comment_char_noslashstar - new_line .or '\x30'-'\xFF' .or '\x01'-'\x29' .or '\x2B'-'\x2E'; - -c_style_comment_end - '*' .and .loop c_style_comment_char_star .and '/'; - -c_style_comment_char_star - '*'; - -cpp_style_comment_block - '/' .and '/' .and cpp_style_comment_block_1; -cpp_style_comment_block_1 - cpp_style_comment_block_2 .or cpp_style_comment_block_3; -cpp_style_comment_block_2 - .loop cpp_style_comment_char .and new_line_directive; -cpp_style_comment_block_3 - .loop cpp_style_comment_char; - -cpp_style_comment_char - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -new_line_directive - new_line .and optional_directive; - -new_line - generic_new_line .emit '\n'; - -generic_new_line - carriage_return_line_feed .or line_feed_carriage_return .or '\n' .or '\r'; - -carriage_return_line_feed - '\r' .and '\n'; - -line_feed_carriage_return - '\n' .and '\r'; - -optional_directive - directive .emit ESCAPE_TOKEN .or .true; - -directive - dir_define .emit TOKEN_DEFINE .or - dir_undef .emit TOKEN_UNDEF .or - dir_if .emit TOKEN_IF .or - dir_ifdef .emit TOKEN_IF .emit 'd' .emit 'e' .emit 'f' .emit 'i' .emit 'n' .emit 'e' .emit 'd' - .emit ' ' .or - dir_ifndef .emit TOKEN_IF .emit '!' .emit 'd' .emit 'e' .emit 'f' .emit 'i' .emit 'n' .emit 'e' - .emit 'd' .emit ' ' .or - dir_else .emit TOKEN_ELSE .or - dir_elif .emit TOKEN_ELIF .or - dir_endif .emit TOKEN_ENDIF .or - dir_ext .emit TOKEN_EXTENSION .or - dir_pragma .emit TOKEN_PRAGMA .or - dir_line .emit TOKEN_LINE; - -dir_define - optional_space .and '#' .and optional_space .and "define" .and symbol .and opt_parameters .and - definition; - -dir_undef - optional_space .and '#' .and optional_space .and "undef" .and symbol; - -dir_if - optional_space .and '#' .and optional_space .and "if" .and expression; - -dir_ifdef - optional_space .and '#' .and optional_space .and "ifdef" .and symbol; - -dir_ifndef - optional_space .and '#' .and optional_space .and "ifndef" .and symbol; - -dir_else - optional_space .and '#' .and optional_space .and "else"; - -dir_elif - optional_space .and '#' .and optional_space .and "elif" .and expression; - -dir_endif - optional_space .and '#' .and optional_space .and "endif"; - -dir_ext - optional_space .and '#' .and optional_space .and "extension" .and space .and extension_name .and - optional_space .and ':' .and optional_space .and extension_behavior; - -dir_line - optional_space .and '#' .and optional_space .and "line" .and expression; - -dir_pragma - optional_space .and '#' .and optional_space .and "pragma" .and symbol .and opt_pragma_param; - - -opt_pragma_param - pragma_param .or .true .emit PRAGMA_NO_PARAM; - -pragma_param - optional_space .and '(' .emit PRAGMA_PARAM .and optional_space .and symbol_no_space .and optional_space .and ')'; - -symbol_no_space - symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\0'; - -symbol - space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\0'; - -opt_parameters - parameters .or .true .emit PARAM_END; - -parameters - '(' .and parameters_1 .and optional_space .and ')' .emit PARAM_END; -parameters_1 - parameters_2 .or .true; -parameters_2 - parameter .emit PARAM_PARAMETER .and .loop parameters_3; -parameters_3 - optional_space .and ',' .and parameter .emit PARAM_PARAMETER; - -parameter - optional_space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and - .true .emit '\0'; - -definition - .loop definition_character .emit * .and .true .emit '\0'; - -definition_character - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -expression - expression_element .and .loop expression_element .and .true .emit '\0'; - -expression_element - expression_character .emit *; - -expression_character - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -extension_name - symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\0'; - -extension_behavior - "require" .emit BEHAVIOR_REQUIRE .or - "enable" .emit BEHAVIOR_ENABLE .or - "warn" .emit BEHAVIOR_WARN .or - "disable" .emit BEHAVIOR_DISABLE; - -optional_space - .loop single_space; - -space - single_space .and .loop single_space; - -single_space - ' ' .or '\t'; - -source_token - space .emit ' ' .or complex_token .or source_token_1; -source_token_1 - simple_token .emit ' ' .and .true .emit ' '; - -/* - * All possible tokens. - */ - -complex_token - identifier .or number; - -simple_token - increment .or decrement .or lequal .or gequal .or equal .or nequal .or and .or xor .or or .or - addto .or subtractfrom .or multiplyto .or divideto .or other; - -identifier - identifier_char1 .emit * .and .loop identifier_char2 .emit *; -identifier_char1 - 'a'-'z' .or 'A'-'Z' .or '_'; -identifier_char2 - 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_'; - -number - float .or integer; - -digit_oct - '0'-'7'; - -digit_dec - '0'-'9'; - -digit_hex - '0'-'9' .or 'A'-'F' .or 'a'-'f'; - -float - float_1 .or float_2; -float_1 - float_fractional_constant .and float_optional_exponent_part; -float_2 - float_digit_sequence .and float_exponent_part; - -float_fractional_constant - float_fractional_constant_1 .or float_fractional_constant_2 .or float_fractional_constant_3; -float_fractional_constant_1 - float_digit_sequence .and '.' .emit '.' .and float_digit_sequence; -float_fractional_constant_2 - float_digit_sequence .and '.' .emit '.'; -float_fractional_constant_3 - '.' .emit '.' .and float_digit_sequence; - -float_optional_exponent_part - float_exponent_part .or .true; - -float_digit_sequence - digit_dec .emit * .and .loop digit_dec .emit *; - -float_exponent_part - float_exponent_part_1 .or float_exponent_part_2; -float_exponent_part_1 - 'e' .emit 'e' .and float_optional_sign .and float_digit_sequence; -float_exponent_part_2 - 'E' .emit 'E' .and float_optional_sign .and float_digit_sequence; - -float_optional_sign - '+' .emit '+' .or '-' .emit '-' .or .true; - -integer - integer_hex .or integer_oct .or integer_dec; - -integer_hex - '0' .emit '0' .and integer_hex_1 .emit * .and digit_hex .emit * .and - .loop digit_hex .emit *; -integer_hex_1 - 'x' .or 'X'; - -integer_oct - '0' .emit '0' .and .loop digit_oct .emit *; - -integer_dec - digit_dec .emit * .and .loop digit_dec .emit *; - -increment - '+' .emit * .and '+' .emit *; - -decrement - '-' .emit * .and '-' .emit *; - -lequal - '<' .emit * .and '=' .emit *; - -gequal - '>' .emit * .and '=' .emit *; - -equal - '=' .emit * .and '=' .emit *; - -nequal - '!' .emit * .and '=' .emit *; - -and - '&' .emit * .and '&' .emit *; - -xor - '^' .emit * .and '^' .emit *; - -or - '|' .emit * .and '|' .emit *; - -addto - '+' .emit * .and '=' .emit *; - -subtractfrom - '-' .emit * .and '=' .emit *; - -multiplyto - '*' .emit * .and '=' .emit *; - -divideto - '/' .emit * .and '=' .emit *; - -/* - * All characters except '\0' and '#'. - */ -other - '\x24'-'\xFF' .emit * .or '\x01'-'\x22' .emit *; - -symbol_character - 'A'-'Z' .or 'a'-'z' .or '_'; - -symbol_character2 - 'A'-'Z' .or 'a'-'z' .or '0'-'9' .or '_'; - -.string string_lexer; - -string_lexer - lex_first_identifier_character .and .loop lex_next_identifier_character; - -lex_first_identifier_character - 'a'-'z' .or 'A'-'Z' .or '_'; - -lex_next_identifier_character - 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_'; - diff --git a/src/mesa/shader/slang/library/slang_pp_directives_syn.h b/src/mesa/shader/slang/library/slang_pp_directives_syn.h deleted file mode 100644 index 430f8d8217c..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_directives_syn.h +++ /dev/null @@ -1,250 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */ - -".syntax source;\n" -".emtcode ESCAPE_TOKEN 0\n" -".emtcode TOKEN_END 0\n" -".emtcode TOKEN_DEFINE 1\n" -".emtcode TOKEN_UNDEF 2\n" -".emtcode TOKEN_IF 3\n" -".emtcode TOKEN_ELSE 4\n" -".emtcode TOKEN_ELIF 5\n" -".emtcode TOKEN_ENDIF 6\n" -".emtcode TOKEN_ERROR 7\n" -".emtcode TOKEN_PRAGMA 8\n" -".emtcode TOKEN_EXTENSION 9\n" -".emtcode TOKEN_LINE 10\n" -".emtcode PARAM_END 0\n" -".emtcode PARAM_PARAMETER 1\n" -".emtcode BEHAVIOR_REQUIRE 1\n" -".emtcode BEHAVIOR_ENABLE 2\n" -".emtcode BEHAVIOR_WARN 3\n" -".emtcode BEHAVIOR_DISABLE 4\n" -".emtcode PRAGMA_NO_PARAM 0\n" -".emtcode PRAGMA_PARAM 1\n" -"source\n" -" optional_directive .and .loop source_element .and '\\0' .emit ESCAPE_TOKEN .emit TOKEN_END;\n" -"source_element\n" -" c_style_comment_block .or cpp_style_comment_block .or new_line_directive .or source_token;\n" -"c_style_comment_block\n" -" '/' .and '*' .and c_style_comment_rest .and .true .emit ' ';\n" -"c_style_comment_rest\n" -" .loop c_style_comment_body .and c_style_comment_end;\n" -"c_style_comment_body\n" -" c_style_comment_char_nostar .or c_style_comment_char_star_noslashstar;\n" -"c_style_comment_char_nostar\n" -" new_line .or '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n" -"c_style_comment_char_star_noslashstar\n" -" '*' .and c_style_comment_char_star_noslashstar_1;\n" -"c_style_comment_char_star_noslashstar_1\n" -" c_style_comment_char_noslashstar .or c_style_comment_char_star_noslashstar;\n" -"c_style_comment_char_noslashstar\n" -" new_line .or '\\x30'-'\\xFF' .or '\\x01'-'\\x29' .or '\\x2B'-'\\x2E';\n" -"c_style_comment_end\n" -" '*' .and .loop c_style_comment_char_star .and '/';\n" -"c_style_comment_char_star\n" -" '*';\n" -"cpp_style_comment_block\n" -" '/' .and '/' .and cpp_style_comment_block_1;\n" -"cpp_style_comment_block_1\n" -" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n" -"cpp_style_comment_block_2\n" -" .loop cpp_style_comment_char .and new_line_directive;\n" -"cpp_style_comment_block_3\n" -" .loop cpp_style_comment_char;\n" -"cpp_style_comment_char\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"new_line_directive\n" -" new_line .and optional_directive;\n" -"new_line\n" -" generic_new_line .emit '\\n';\n" -"generic_new_line\n" -" carriage_return_line_feed .or line_feed_carriage_return .or '\\n' .or '\\r';\n" -"carriage_return_line_feed\n" -" '\\r' .and '\\n';\n" -"line_feed_carriage_return\n" -" '\\n' .and '\\r';\n" -"optional_directive\n" -" directive .emit ESCAPE_TOKEN .or .true;\n" -"directive\n" -" dir_define .emit TOKEN_DEFINE .or\n" -" dir_undef .emit TOKEN_UNDEF .or\n" -" dir_if .emit TOKEN_IF .or\n" -" dir_ifdef .emit TOKEN_IF .emit 'd' .emit 'e' .emit 'f' .emit 'i' .emit 'n' .emit 'e' .emit 'd'\n" -" .emit ' ' .or\n" -" dir_ifndef .emit TOKEN_IF .emit '!' .emit 'd' .emit 'e' .emit 'f' .emit 'i' .emit 'n' .emit 'e'\n" -" .emit 'd' .emit ' ' .or\n" -" dir_else .emit TOKEN_ELSE .or\n" -" dir_elif .emit TOKEN_ELIF .or\n" -" dir_endif .emit TOKEN_ENDIF .or\n" -" dir_ext .emit TOKEN_EXTENSION .or\n" -" dir_pragma .emit TOKEN_PRAGMA .or\n" -" dir_line .emit TOKEN_LINE;\n" -"dir_define\n" -" optional_space .and '#' .and optional_space .and \"define\" .and symbol .and opt_parameters .and\n" -" definition;\n" -"dir_undef\n" -" optional_space .and '#' .and optional_space .and \"undef\" .and symbol;\n" -"dir_if\n" -" optional_space .and '#' .and optional_space .and \"if\" .and expression;\n" -"dir_ifdef\n" -" optional_space .and '#' .and optional_space .and \"ifdef\" .and symbol;\n" -"dir_ifndef\n" -" optional_space .and '#' .and optional_space .and \"ifndef\" .and symbol;\n" -"dir_else\n" -" optional_space .and '#' .and optional_space .and \"else\";\n" -"dir_elif\n" -" optional_space .and '#' .and optional_space .and \"elif\" .and expression;\n" -"dir_endif\n" -" optional_space .and '#' .and optional_space .and \"endif\";\n" -"dir_ext\n" -" optional_space .and '#' .and optional_space .and \"extension\" .and space .and extension_name .and\n" -" optional_space .and ':' .and optional_space .and extension_behavior;\n" -"dir_line\n" -" optional_space .and '#' .and optional_space .and \"line\" .and expression;\n" -"dir_pragma\n" -" optional_space .and '#' .and optional_space .and \"pragma\" .and symbol .and opt_pragma_param;\n" -"opt_pragma_param\n" -" pragma_param .or .true .emit PRAGMA_NO_PARAM;\n" -"pragma_param\n" -" optional_space .and '(' .emit PRAGMA_PARAM .and optional_space .and symbol_no_space .and optional_space .and ')';\n" -"symbol_no_space\n" -" symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\\0';\n" -"symbol\n" -" space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\\0';\n" -"opt_parameters\n" -" parameters .or .true .emit PARAM_END;\n" -"parameters\n" -" '(' .and parameters_1 .and optional_space .and ')' .emit PARAM_END;\n" -"parameters_1\n" -" parameters_2 .or .true;\n" -"parameters_2\n" -" parameter .emit PARAM_PARAMETER .and .loop parameters_3;\n" -"parameters_3\n" -" optional_space .and ',' .and parameter .emit PARAM_PARAMETER;\n" -"parameter\n" -" optional_space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and\n" -" .true .emit '\\0';\n" -"definition\n" -" .loop definition_character .emit * .and .true .emit '\\0';\n" -"definition_character\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"expression\n" -" expression_element .and .loop expression_element .and .true .emit '\\0';\n" -"expression_element\n" -" expression_character .emit *;\n" -"expression_character\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"extension_name\n" -" symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\\0';\n" -"extension_behavior\n" -" \"require\" .emit BEHAVIOR_REQUIRE .or\n" -" \"enable\" .emit BEHAVIOR_ENABLE .or\n" -" \"warn\" .emit BEHAVIOR_WARN .or\n" -" \"disable\" .emit BEHAVIOR_DISABLE;\n" -"optional_space\n" -" .loop single_space;\n" -"space\n" -" single_space .and .loop single_space;\n" -"single_space\n" -" ' ' .or '\\t';\n" -"source_token\n" -" space .emit ' ' .or complex_token .or source_token_1;\n" -"source_token_1\n" -" simple_token .emit ' ' .and .true .emit ' ';\n" -"complex_token\n" -" identifier .or number;\n" -"simple_token\n" -" increment .or decrement .or lequal .or gequal .or equal .or nequal .or and .or xor .or or .or\n" -" addto .or subtractfrom .or multiplyto .or divideto .or other;\n" -"identifier\n" -" identifier_char1 .emit * .and .loop identifier_char2 .emit *;\n" -"identifier_char1\n" -" 'a'-'z' .or 'A'-'Z' .or '_';\n" -"identifier_char2\n" -" 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_';\n" -"number\n" -" float .or integer;\n" -"digit_oct\n" -" '0'-'7';\n" -"digit_dec\n" -" '0'-'9';\n" -"digit_hex\n" -" '0'-'9' .or 'A'-'F' .or 'a'-'f';\n" -"float\n" -" float_1 .or float_2;\n" -"float_1\n" -" float_fractional_constant .and float_optional_exponent_part;\n" -"float_2\n" -" float_digit_sequence .and float_exponent_part;\n" -"float_fractional_constant\n" -" float_fractional_constant_1 .or float_fractional_constant_2 .or float_fractional_constant_3;\n" -"float_fractional_constant_1\n" -" float_digit_sequence .and '.' .emit '.' .and float_digit_sequence;\n" -"float_fractional_constant_2\n" -" float_digit_sequence .and '.' .emit '.';\n" -"float_fractional_constant_3\n" -" '.' .emit '.' .and float_digit_sequence;\n" -"float_optional_exponent_part\n" -" float_exponent_part .or .true;\n" -"float_digit_sequence\n" -" digit_dec .emit * .and .loop digit_dec .emit *;\n" -"float_exponent_part\n" -" float_exponent_part_1 .or float_exponent_part_2;\n" -"float_exponent_part_1\n" -" 'e' .emit 'e' .and float_optional_sign .and float_digit_sequence;\n" -"float_exponent_part_2\n" -" 'E' .emit 'E' .and float_optional_sign .and float_digit_sequence;\n" -"float_optional_sign\n" -" '+' .emit '+' .or '-' .emit '-' .or .true;\n" -"integer\n" -" integer_hex .or integer_oct .or integer_dec;\n" -"integer_hex\n" -" '0' .emit '0' .and integer_hex_1 .emit * .and digit_hex .emit * .and\n" -" .loop digit_hex .emit *;\n" -"integer_hex_1\n" -" 'x' .or 'X';\n" -"integer_oct\n" -" '0' .emit '0' .and .loop digit_oct .emit *;\n" -"integer_dec\n" -" digit_dec .emit * .and .loop digit_dec .emit *;\n" -"increment\n" -" '+' .emit * .and '+' .emit *;\n" -"decrement\n" -" '-' .emit * .and '-' .emit *;\n" -"lequal\n" -" '<' .emit * .and '=' .emit *;\n" -"gequal\n" -" '>' .emit * .and '=' .emit *;\n" -"equal\n" -" '=' .emit * .and '=' .emit *;\n" -"nequal\n" -" '!' .emit * .and '=' .emit *;\n" -"and\n" -" '&' .emit * .and '&' .emit *;\n" -"xor\n" -" '^' .emit * .and '^' .emit *;\n" -"or\n" -" '|' .emit * .and '|' .emit *;\n" -"addto\n" -" '+' .emit * .and '=' .emit *;\n" -"subtractfrom\n" -" '-' .emit * .and '=' .emit *;\n" -"multiplyto\n" -" '*' .emit * .and '=' .emit *;\n" -"divideto\n" -" '/' .emit * .and '=' .emit *;\n" -"other\n" -" '\\x24'-'\\xFF' .emit * .or '\\x01'-'\\x22' .emit *;\n" -"symbol_character\n" -" 'A'-'Z' .or 'a'-'z' .or '_';\n" -"symbol_character2\n" -" 'A'-'Z' .or 'a'-'z' .or '0'-'9' .or '_';\n" -".string string_lexer;\n" -"string_lexer\n" -" lex_first_identifier_character .and .loop lex_next_identifier_character;\n" -"lex_first_identifier_character\n" -" 'a'-'z' .or 'A'-'Z' .or '_';\n" -"lex_next_identifier_character\n" -" 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_';\n" -"" diff --git a/src/mesa/shader/slang/library/slang_pp_expression.syn b/src/mesa/shader/slang/library/slang_pp_expression.syn deleted file mode 100644 index bfdb220bf5c..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_expression.syn +++ /dev/null @@ -1,265 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.6 - * - * Copyright (C) 2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file slang_pp_expression.syn - * slang preprocessor expression parser - * \author Michal Krol - */ - -/* - * Parses one or two (optional) expressions on literal integer constants. Those expressions come - * from #if #elif and #line directives. The preprocessor already parsed those directives and - * expanded the expression (expressions). All occurences of the operator "defined" are already - * replaced with either "0" or "1" literals. - */ - -.syntax expression; - -/* - * Those separate individual expressions. - * For #if/#elif case it is: EXP_EXPRESSION ... EXP_END - * For #line case it may be: EXP_EXPRESSION ... EXP_EXPRESSION ... EXP_END - */ -.emtcode EXP_END 0 -.emtcode EXP_EXPRESSION 1 - -.emtcode OP_END 0 -.emtcode OP_PUSHINT 1 -.emtcode OP_LOGICALOR 2 -.emtcode OP_LOGICALAND 3 -.emtcode OP_OR 4 -.emtcode OP_XOR 5 -.emtcode OP_AND 6 -.emtcode OP_EQUAL 7 -.emtcode OP_NOTEQUAL 8 -.emtcode OP_LESSEQUAL 9 -.emtcode OP_GREATEREQUAL 10 -.emtcode OP_LESS 11 -.emtcode OP_GREATER 12 -.emtcode OP_LEFTSHIFT 13 -.emtcode OP_RIGHTSHIFT 14 -.emtcode OP_ADD 15 -.emtcode OP_SUBTRACT 16 -.emtcode OP_MULTIPLY 17 -.emtcode OP_DIVIDE 18 -.emtcode OP_MODULUS 19 -.emtcode OP_PLUS 20 -.emtcode OP_MINUS 21 -.emtcode OP_NEGATE 22 -.emtcode OP_COMPLEMENT 23 - -expression - first_expression .and optional_second_expression .and optional_space .and '\0' .emit EXP_END; - -first_expression - optional_space .and logical_or_expression .emit EXP_EXPRESSION .and .true .emit OP_END; - -optional_second_expression - second_expression .or .true; - -second_expression - space .and logical_or_expression .emit EXP_EXPRESSION .and .true .emit OP_END; - -logical_or_expression - logical_and_expression .and .loop logical_or_expression_1; -logical_or_expression_1 - barbar .and logical_and_expression .and .true .emit OP_LOGICALOR; - -logical_and_expression - or_expression .and .loop logical_and_expression_1; -logical_and_expression_1 - ampersandampersand .and or_expression .and .true .emit OP_LOGICALAND; - -or_expression - xor_expression .and .loop or_expression_1; -or_expression_1 - bar .and xor_expression .and .true .emit OP_OR; - -xor_expression - and_expression .and .loop xor_expression_1; -xor_expression_1 - caret .and and_expression .and .true .emit OP_XOR; - -and_expression - equality_expression .and .loop and_expression_1; -and_expression_1 - ampersand .and equality_expression .and .true .emit OP_AND; - -equality_expression - relational_expression .and .loop equality_expression_1; -equality_expression_1 - equality_expression_2 .or equality_expression_3; -equality_expression_2 - equalsequals .and relational_expression .and .true .emit OP_EQUAL; -equality_expression_3 - bangequals .and relational_expression .and .true .emit OP_NOTEQUAL; - -relational_expression - shift_expression .and .loop relational_expression_1; -relational_expression_1 - relational_expression_2 .or relational_expression_3 .or relational_expression_4 .or - relational_expression_5; -relational_expression_2 - lessequals .and shift_expression .and .true .emit OP_LESSEQUAL; -relational_expression_3 - greaterequals .and shift_expression .and .true .emit OP_GREATEREQUAL; -relational_expression_4 - less .and shift_expression .and .true .emit OP_LESS; -relational_expression_5 - greater .and shift_expression .and .true .emit OP_GREATER; - -shift_expression - additive_expression .and .loop shift_expression_1; -shift_expression_1 - shift_expression_2 .or shift_expression_3; -shift_expression_2 - lessless .and additive_expression .and .true .emit OP_LEFTSHIFT; -shift_expression_3 - greatergreater .and additive_expression .and .true .emit OP_RIGHTSHIFT; - -additive_expression - multiplicative_expression .and .loop additive_expression_1; -additive_expression_1 - additive_expression_2 .or additive_expression_3; -additive_expression_2 - plus .and multiplicative_expression .and .true .emit OP_ADD; -additive_expression_3 - dash .and multiplicative_expression .and .true .emit OP_SUBTRACT; - -multiplicative_expression - unary_expression .and .loop multiplicative_expression_1; -multiplicative_expression_1 - multiplicative_expression_2 .or multiplicative_expression_3 .or multiplicative_expression_4; -multiplicative_expression_2 - star .and unary_expression .and .true .emit OP_MULTIPLY; -multiplicative_expression_3 - slash .and unary_expression .and .true .emit OP_DIVIDE; -multiplicative_expression_4 - percent .and unary_expression .and .true .emit OP_MODULUS; - -unary_expression - primary_expression .or unary_expression_1 .or unary_expression_2 .or unary_expression_3 .or - unary_expression_4; -unary_expression_1 - plus .and unary_expression .and .true .emit OP_PLUS; -unary_expression_2 - dash .and unary_expression .and .true .emit OP_MINUS; -unary_expression_3 - bang .and unary_expression .and .true .emit OP_NEGATE; -unary_expression_4 - tilda .and unary_expression .and .true .emit OP_COMPLEMENT; - -primary_expression - intconstant .or primary_expression_1; -primary_expression_1 - lparen .and logical_or_expression .and rparen; - -intconstant - integer .emit OP_PUSHINT; - -integer - integer_dec; - -integer_dec - digit_dec .emit 10 .emit * .and .loop digit_dec .emit * .and .true .emit '\0'; - -digit_dec - '0'-'9'; - -optional_space - .loop single_space; - -space - single_space .and .loop single_space; - -single_space - ' ' .or '\t'; - -ampersand - optional_space .and '&' .and optional_space; - -ampersandampersand - optional_space .and '&' .and '&' .and optional_space; - -bang - optional_space .and '!' .and optional_space; - -bangequals - optional_space .and '!' .and '=' .and optional_space; - -bar - optional_space .and '|' .and optional_space; - -barbar - optional_space .and '|' .and '|' .and optional_space; - -caret - optional_space .and '^' .and optional_space; - -dash - optional_space .and '-' .and optional_space; - -equalsequals - optional_space .and '=' .and '=' .and optional_space; - -greater - optional_space .and '>' .and optional_space; - -greaterequals - optional_space .and '>' .and '=' .and optional_space; - -greatergreater - optional_space .and '>' .and '>' .and optional_space; - -less - optional_space .and '<' .and optional_space; - -lessequals - optional_space .and '<' .and '=' .and optional_space; - -lessless - optional_space .and '<' .and '<' .and optional_space; - -lparen - optional_space .and '(' .and optional_space; - -percent - optional_space .and '%' .and optional_space; - -plus - optional_space .and '+' .and optional_space; - -rparen - optional_space .and ')' .and optional_space; - -slash - optional_space .and '/' .and optional_space; - -star - optional_space .and '*' .and optional_space; - -tilda - optional_space .and '~' .and optional_space; - diff --git a/src/mesa/shader/slang/library/slang_pp_expression_syn.h b/src/mesa/shader/slang/library/slang_pp_expression_syn.h deleted file mode 100644 index f3e9ef6b229..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_expression_syn.h +++ /dev/null @@ -1,179 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */ - -".syntax expression;\n" -".emtcode EXP_END 0\n" -".emtcode EXP_EXPRESSION 1\n" -".emtcode OP_END 0\n" -".emtcode OP_PUSHINT 1\n" -".emtcode OP_LOGICALOR 2\n" -".emtcode OP_LOGICALAND 3\n" -".emtcode OP_OR 4\n" -".emtcode OP_XOR 5\n" -".emtcode OP_AND 6\n" -".emtcode OP_EQUAL 7\n" -".emtcode OP_NOTEQUAL 8\n" -".emtcode OP_LESSEQUAL 9\n" -".emtcode OP_GREATEREQUAL 10\n" -".emtcode OP_LESS 11\n" -".emtcode OP_GREATER 12\n" -".emtcode OP_LEFTSHIFT 13\n" -".emtcode OP_RIGHTSHIFT 14\n" -".emtcode OP_ADD 15\n" -".emtcode OP_SUBTRACT 16\n" -".emtcode OP_MULTIPLY 17\n" -".emtcode OP_DIVIDE 18\n" -".emtcode OP_MODULUS 19\n" -".emtcode OP_PLUS 20\n" -".emtcode OP_MINUS 21\n" -".emtcode OP_NEGATE 22\n" -".emtcode OP_COMPLEMENT 23\n" -"expression\n" -" first_expression .and optional_second_expression .and optional_space .and '\\0' .emit EXP_END;\n" -"first_expression\n" -" optional_space .and logical_or_expression .emit EXP_EXPRESSION .and .true .emit OP_END;\n" -"optional_second_expression\n" -" second_expression .or .true;\n" -"second_expression\n" -" space .and logical_or_expression .emit EXP_EXPRESSION .and .true .emit OP_END;\n" -"logical_or_expression\n" -" logical_and_expression .and .loop logical_or_expression_1;\n" -"logical_or_expression_1\n" -" barbar .and logical_and_expression .and .true .emit OP_LOGICALOR;\n" -"logical_and_expression\n" -" or_expression .and .loop logical_and_expression_1;\n" -"logical_and_expression_1\n" -" ampersandampersand .and or_expression .and .true .emit OP_LOGICALAND;\n" -"or_expression\n" -" xor_expression .and .loop or_expression_1;\n" -"or_expression_1\n" -" bar .and xor_expression .and .true .emit OP_OR;\n" -"xor_expression\n" -" and_expression .and .loop xor_expression_1;\n" -"xor_expression_1\n" -" caret .and and_expression .and .true .emit OP_XOR;\n" -"and_expression\n" -" equality_expression .and .loop and_expression_1;\n" -"and_expression_1\n" -" ampersand .and equality_expression .and .true .emit OP_AND;\n" -"equality_expression\n" -" relational_expression .and .loop equality_expression_1;\n" -"equality_expression_1\n" -" equality_expression_2 .or equality_expression_3;\n" -"equality_expression_2\n" -" equalsequals .and relational_expression .and .true .emit OP_EQUAL;\n" -"equality_expression_3\n" -" bangequals .and relational_expression .and .true .emit OP_NOTEQUAL;\n" -"relational_expression\n" -" shift_expression .and .loop relational_expression_1;\n" -"relational_expression_1\n" -" relational_expression_2 .or relational_expression_3 .or relational_expression_4 .or\n" -" relational_expression_5;\n" -"relational_expression_2\n" -" lessequals .and shift_expression .and .true .emit OP_LESSEQUAL;\n" -"relational_expression_3\n" -" greaterequals .and shift_expression .and .true .emit OP_GREATEREQUAL;\n" -"relational_expression_4\n" -" less .and shift_expression .and .true .emit OP_LESS;\n" -"relational_expression_5\n" -" greater .and shift_expression .and .true .emit OP_GREATER;\n" -"shift_expression\n" -" additive_expression .and .loop shift_expression_1;\n" -"shift_expression_1\n" -" shift_expression_2 .or shift_expression_3;\n" -"shift_expression_2\n" -" lessless .and additive_expression .and .true .emit OP_LEFTSHIFT;\n" -"shift_expression_3\n" -" greatergreater .and additive_expression .and .true .emit OP_RIGHTSHIFT;\n" -"additive_expression\n" -" multiplicative_expression .and .loop additive_expression_1;\n" -"additive_expression_1\n" -" additive_expression_2 .or additive_expression_3;\n" -"additive_expression_2\n" -" plus .and multiplicative_expression .and .true .emit OP_ADD;\n" -"additive_expression_3\n" -" dash .and multiplicative_expression .and .true .emit OP_SUBTRACT;\n" -"multiplicative_expression\n" -" unary_expression .and .loop multiplicative_expression_1;\n" -"multiplicative_expression_1\n" -" multiplicative_expression_2 .or multiplicative_expression_3 .or multiplicative_expression_4;\n" -"multiplicative_expression_2\n" -" star .and unary_expression .and .true .emit OP_MULTIPLY;\n" -"multiplicative_expression_3\n" -" slash .and unary_expression .and .true .emit OP_DIVIDE;\n" -"multiplicative_expression_4\n" -" percent .and unary_expression .and .true .emit OP_MODULUS;\n" -"unary_expression\n" -" primary_expression .or unary_expression_1 .or unary_expression_2 .or unary_expression_3 .or\n" -" unary_expression_4;\n" -"unary_expression_1\n" -" plus .and unary_expression .and .true .emit OP_PLUS;\n" -"unary_expression_2\n" -" dash .and unary_expression .and .true .emit OP_MINUS;\n" -"unary_expression_3\n" -" bang .and unary_expression .and .true .emit OP_NEGATE;\n" -"unary_expression_4\n" -" tilda .and unary_expression .and .true .emit OP_COMPLEMENT;\n" -"primary_expression\n" -" intconstant .or primary_expression_1;\n" -"primary_expression_1\n" -" lparen .and logical_or_expression .and rparen;\n" -"intconstant\n" -" integer .emit OP_PUSHINT;\n" -"integer\n" -" integer_dec;\n" -"integer_dec\n" -" digit_dec .emit 10 .emit * .and .loop digit_dec .emit * .and .true .emit '\\0';\n" -"digit_dec\n" -" '0'-'9';\n" -"optional_space\n" -" .loop single_space;\n" -"space\n" -" single_space .and .loop single_space;\n" -"single_space\n" -" ' ' .or '\\t';\n" -"ampersand\n" -" optional_space .and '&' .and optional_space;\n" -"ampersandampersand\n" -" optional_space .and '&' .and '&' .and optional_space;\n" -"bang\n" -" optional_space .and '!' .and optional_space;\n" -"bangequals\n" -" optional_space .and '!' .and '=' .and optional_space;\n" -"bar\n" -" optional_space .and '|' .and optional_space;\n" -"barbar\n" -" optional_space .and '|' .and '|' .and optional_space;\n" -"caret\n" -" optional_space .and '^' .and optional_space;\n" -"dash\n" -" optional_space .and '-' .and optional_space;\n" -"equalsequals\n" -" optional_space .and '=' .and '=' .and optional_space;\n" -"greater\n" -" optional_space .and '>' .and optional_space;\n" -"greaterequals\n" -" optional_space .and '>' .and '=' .and optional_space;\n" -"greatergreater\n" -" optional_space .and '>' .and '>' .and optional_space;\n" -"less\n" -" optional_space .and '<' .and optional_space;\n" -"lessequals\n" -" optional_space .and '<' .and '=' .and optional_space;\n" -"lessless\n" -" optional_space .and '<' .and '<' .and optional_space;\n" -"lparen\n" -" optional_space .and '(' .and optional_space;\n" -"percent\n" -" optional_space .and '%' .and optional_space;\n" -"plus\n" -" optional_space .and '+' .and optional_space;\n" -"rparen\n" -" optional_space .and ')' .and optional_space;\n" -"slash\n" -" optional_space .and '/' .and optional_space;\n" -"star\n" -" optional_space .and '*' .and optional_space;\n" -"tilda\n" -" optional_space .and '~' .and optional_space;\n" -"" diff --git a/src/mesa/shader/slang/library/slang_pp_version.syn b/src/mesa/shader/slang/library/slang_pp_version.syn deleted file mode 100644 index 3fe1a57ba2c..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_version.syn +++ /dev/null @@ -1,122 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.6 - * - * Copyright (C) 2005-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file slang_pp_version.syn - * slang #version directive syntax - * \author Michal Krol - */ - -.syntax version_directive; - -version_directive - version_directive_1; -version_directive_1 - prior_optional_spaces .and optional_version_directive .and .true .emit $; - -optional_version_directive - version_directive_body .or .true .emit 10 .emit 1; - -version_directive_body - '#' .and optional_space .and "version" .and space .and version_number .and optional_space .and - new_line; - -version_number - version_number_100 .or version_number_110 .or version_number_120; - -version_number_100 - leading_zeroes .and "100" .emit 0 .emit 1; - -version_number_110 - leading_zeroes .and "110" .emit 10 .emit 1; - -version_number_120 - leading_zeroes .and "120" .emit 20 .emit 1; - -leading_zeroes - .loop zero; - -zero - '0'; - -space - single_space .and .loop single_space; - -optional_space - .loop single_space; - -single_space - ' ' .or '\t'; - -prior_optional_spaces - .loop prior_space; - -prior_space - c_style_comment_block .or cpp_style_comment_block .or space .or new_line; - -c_style_comment_block - '/' .and '*' .and c_style_comment_rest; - -c_style_comment_rest - .loop c_style_comment_char_no_star .and c_style_comment_rest_1; -c_style_comment_rest_1 - c_style_comment_end .or c_style_comment_rest_2; -c_style_comment_rest_2 - '*' .and c_style_comment_rest; - -c_style_comment_char_no_star - '\x2B'-'\xFF' .or '\x01'-'\x29'; - -c_style_comment_end - '*' .and '/'; - -cpp_style_comment_block - '/' .and '/' .and cpp_style_comment_block_1; -cpp_style_comment_block_1 - cpp_style_comment_block_2 .or cpp_style_comment_block_3; -cpp_style_comment_block_2 - .loop cpp_style_comment_char .and new_line; -cpp_style_comment_block_3 - .loop cpp_style_comment_char; - -cpp_style_comment_char - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -new_line - cr_lf .or lf_cr .or '\n' .or '\r'; - -cr_lf - '\r' .and '\n'; - -lf_cr - '\n' .and '\r'; - -.string __string_filter; - -__string_filter - .loop __identifier_char; - -__identifier_char - 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9'; - diff --git a/src/mesa/shader/slang/library/slang_pp_version_syn.h b/src/mesa/shader/slang/library/slang_pp_version_syn.h deleted file mode 100644 index 54aee3ff282..00000000000 --- a/src/mesa/shader/slang/library/slang_pp_version_syn.h +++ /dev/null @@ -1,69 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */ - -".syntax version_directive;\n" -"version_directive\n" -" version_directive_1;\n" -"version_directive_1\n" -" prior_optional_spaces .and optional_version_directive .and .true .emit $;\n" -"optional_version_directive\n" -" version_directive_body .or .true .emit 10 .emit 1;\n" -"version_directive_body\n" -" '#' .and optional_space .and \"version\" .and space .and version_number .and optional_space .and\n" -" new_line;\n" -"version_number\n" -" version_number_100 .or version_number_110 .or version_number_120;\n" -"version_number_100\n" -" leading_zeroes .and \"100\" .emit 0 .emit 1;\n" -"version_number_110\n" -" leading_zeroes .and \"110\" .emit 10 .emit 1;\n" -"version_number_120\n" -" leading_zeroes .and \"120\" .emit 20 .emit 1;\n" -"leading_zeroes\n" -" .loop zero;\n" -"zero\n" -" '0';\n" -"space\n" -" single_space .and .loop single_space;\n" -"optional_space\n" -" .loop single_space;\n" -"single_space\n" -" ' ' .or '\\t';\n" -"prior_optional_spaces\n" -" .loop prior_space;\n" -"prior_space\n" -" c_style_comment_block .or cpp_style_comment_block .or space .or new_line;\n" -"c_style_comment_block\n" -" '/' .and '*' .and c_style_comment_rest;\n" -"c_style_comment_rest\n" -" .loop c_style_comment_char_no_star .and c_style_comment_rest_1;\n" -"c_style_comment_rest_1\n" -" c_style_comment_end .or c_style_comment_rest_2;\n" -"c_style_comment_rest_2\n" -" '*' .and c_style_comment_rest;\n" -"c_style_comment_char_no_star\n" -" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n" -"c_style_comment_end\n" -" '*' .and '/';\n" -"cpp_style_comment_block\n" -" '/' .and '/' .and cpp_style_comment_block_1;\n" -"cpp_style_comment_block_1\n" -" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n" -"cpp_style_comment_block_2\n" -" .loop cpp_style_comment_char .and new_line;\n" -"cpp_style_comment_block_3\n" -" .loop cpp_style_comment_char;\n" -"cpp_style_comment_char\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"new_line\n" -" cr_lf .or lf_cr .or '\\n' .or '\\r';\n" -"cr_lf\n" -" '\\r' .and '\\n';\n" -"lf_cr\n" -" '\\n' .and '\\r';\n" -".string __string_filter;\n" -"__string_filter\n" -" .loop __identifier_char;\n" -"__identifier_char\n" -" 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9';\n" -"" diff --git a/src/mesa/shader/slang/library/slang_shader.syn b/src/mesa/shader/slang/library/slang_shader.syn deleted file mode 100644 index cc5c70a02f8..00000000000 --- a/src/mesa/shader/slang/library/slang_shader.syn +++ /dev/null @@ -1,1716 +0,0 @@ -/* - * Mesa 3-D graphics library - * - * Copyright (C) 2004-2006 Brian Paul All Rights Reserved. - * Copyright (C) 2008 VMware, Inc. All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* - * \file slang_shader.syn - * slang vertex/fragment shader syntax - * \author Michal Krol - */ - -/* - * usage: - * syn2c slang_shader.syn > slang_shader_syn.h - * - * when modifying or extending this file, several things must be taken into - * consideration: - * - * - when adding new operators that were marked as reserved in the - * initial specification, one must only uncomment particular lines of - * code that refer to operators being added; - * - * - when adding new shader targets, one must reserve a new value for - * shader_type register and use it in .if constructs for symbols that - * are exclusive for that shader; - * - * - some symbols mimic output of other symbols - the best example is - * the "for" construct: expression "for (foo(); ; bar())" is seen as - * "for (foo(); true; bar())" by the output processor - hence, special - * care must be taken when rearranging output of essential symbols; - * - * - order of single-quoted tokens does matter in alternatives - so do not - * parse "<" operator before "<<" and "<<" before "<<="; - * - * - all double-quoted tokens are internally preprocessed to eliminate - * problems with parsing strings that are prefixes of other strings, - * like "sampler1D" and "sampler1DShadow"; - */ - -.syntax translation_unit; - -/* revision number - increment after each change affecting emitted output */ -.emtcode REVISION 5 - -/* external declaration (or precision or invariant stmt) */ -.emtcode EXTERNAL_NULL 0 -.emtcode EXTERNAL_FUNCTION_DEFINITION 1 -.emtcode EXTERNAL_DECLARATION 2 -.emtcode DEFAULT_PRECISION 3 -.emtcode INVARIANT_STMT 4 - -/* precision */ -.emtcode PRECISION_DEFAULT 0 -.emtcode PRECISION_LOW 1 -.emtcode PRECISION_MEDIUM 2 -.emtcode PRECISION_HIGH 3 - -/* declaration */ -.emtcode DECLARATION_FUNCTION_PROTOTYPE 1 -.emtcode DECLARATION_INIT_DECLARATOR_LIST 2 - -/* function type */ -.emtcode FUNCTION_ORDINARY 0 -.emtcode FUNCTION_CONSTRUCTOR 1 -.emtcode FUNCTION_OPERATOR 2 - -/* function call type */ -.emtcode FUNCTION_CALL_NONARRAY 0 -.emtcode FUNCTION_CALL_ARRAY 1 - -/* operator type */ -.emtcode OPERATOR_ADDASSIGN 1 -.emtcode OPERATOR_SUBASSIGN 2 -.emtcode OPERATOR_MULASSIGN 3 -.emtcode OPERATOR_DIVASSIGN 4 -/*.emtcode OPERATOR_MODASSIGN 5*/ -/*.emtcode OPERATOR_LSHASSIGN 6*/ -/*.emtcode OPERATOR_RSHASSIGN 7*/ -/*.emtcode OPERATOR_ORASSIGN 8*/ -/*.emtcode OPERATOR_XORASSIGN 9*/ -/*.emtcode OPERATOR_ANDASSIGN 10*/ -.emtcode OPERATOR_LOGICALXOR 11 -/*.emtcode OPERATOR_BITOR 12*/ -/*.emtcode OPERATOR_BITXOR 13*/ -/*.emtcode OPERATOR_BITAND 14*/ -.emtcode OPERATOR_LESS 15 -.emtcode OPERATOR_GREATER 16 -.emtcode OPERATOR_LESSEQUAL 17 -.emtcode OPERATOR_GREATEREQUAL 18 -/*.emtcode OPERATOR_LSHIFT 19*/ -/*.emtcode OPERATOR_RSHIFT 20*/ -.emtcode OPERATOR_MULTIPLY 21 -.emtcode OPERATOR_DIVIDE 22 -/*.emtcode OPERATOR_MODULUS 23*/ -.emtcode OPERATOR_INCREMENT 24 -.emtcode OPERATOR_DECREMENT 25 -.emtcode OPERATOR_PLUS 26 -.emtcode OPERATOR_MINUS 27 -/*.emtcode OPERATOR_COMPLEMENT 28*/ -.emtcode OPERATOR_NOT 29 - -/* init declarator list */ -.emtcode DECLARATOR_NONE 0 -.emtcode DECLARATOR_NEXT 1 - -/* variable declaration */ -.emtcode VARIABLE_NONE 0 -.emtcode VARIABLE_IDENTIFIER 1 -.emtcode VARIABLE_INITIALIZER 2 -.emtcode VARIABLE_ARRAY_EXPLICIT 3 -.emtcode VARIABLE_ARRAY_UNKNOWN 4 - -/* type qualifier */ -.emtcode TYPE_QUALIFIER_NONE 0 -.emtcode TYPE_QUALIFIER_CONST 1 -.emtcode TYPE_QUALIFIER_ATTRIBUTE 2 -.emtcode TYPE_QUALIFIER_VARYING 3 -.emtcode TYPE_QUALIFIER_UNIFORM 4 -.emtcode TYPE_QUALIFIER_FIXEDOUTPUT 5 -.emtcode TYPE_QUALIFIER_FIXEDINPUT 6 - -/* invariant qualifier */ -.emtcode TYPE_VARIANT 90 -.emtcode TYPE_INVARIANT 91 - -/* centroid qualifier */ -.emtcode TYPE_CENTER 95 -.emtcode TYPE_CENTROID 96 - -/* type specifier */ -.emtcode TYPE_SPECIFIER_VOID 0 -.emtcode TYPE_SPECIFIER_BOOL 1 -.emtcode TYPE_SPECIFIER_BVEC2 2 -.emtcode TYPE_SPECIFIER_BVEC3 3 -.emtcode TYPE_SPECIFIER_BVEC4 4 -.emtcode TYPE_SPECIFIER_INT 5 -.emtcode TYPE_SPECIFIER_IVEC2 6 -.emtcode TYPE_SPECIFIER_IVEC3 7 -.emtcode TYPE_SPECIFIER_IVEC4 8 -.emtcode TYPE_SPECIFIER_FLOAT 9 -.emtcode TYPE_SPECIFIER_VEC2 10 -.emtcode TYPE_SPECIFIER_VEC3 11 -.emtcode TYPE_SPECIFIER_VEC4 12 -.emtcode TYPE_SPECIFIER_MAT2 13 -.emtcode TYPE_SPECIFIER_MAT3 14 -.emtcode TYPE_SPECIFIER_MAT4 15 -.emtcode TYPE_SPECIFIER_SAMPLER1D 16 -.emtcode TYPE_SPECIFIER_SAMPLER2D 17 -.emtcode TYPE_SPECIFIER_SAMPLER3D 18 -.emtcode TYPE_SPECIFIER_SAMPLERCUBE 19 -.emtcode TYPE_SPECIFIER_SAMPLER1DSHADOW 20 -.emtcode TYPE_SPECIFIER_SAMPLER2DSHADOW 21 -.emtcode TYPE_SPECIFIER_SAMPLER2DRECT 22 -.emtcode TYPE_SPECIFIER_SAMPLER2DRECTSHADOW 23 -.emtcode TYPE_SPECIFIER_STRUCT 24 -.emtcode TYPE_SPECIFIER_TYPENAME 25 - -/* OpenGL 2.1 */ -.emtcode TYPE_SPECIFIER_MAT23 26 -.emtcode TYPE_SPECIFIER_MAT32 27 -.emtcode TYPE_SPECIFIER_MAT24 28 -.emtcode TYPE_SPECIFIER_MAT42 29 -.emtcode TYPE_SPECIFIER_MAT34 30 -.emtcode TYPE_SPECIFIER_MAT43 31 - -/* type specifier array */ -.emtcode TYPE_SPECIFIER_NONARRAY 0 -.emtcode TYPE_SPECIFIER_ARRAY 1 - -/* structure field */ -.emtcode FIELD_NONE 0 -.emtcode FIELD_NEXT 1 -.emtcode FIELD_ARRAY 2 - -/* operation */ -.emtcode OP_END 0 -.emtcode OP_BLOCK_BEGIN_NO_NEW_SCOPE 1 -.emtcode OP_BLOCK_BEGIN_NEW_SCOPE 2 -.emtcode OP_DECLARE 3 -.emtcode OP_ASM 4 -.emtcode OP_BREAK 5 -.emtcode OP_CONTINUE 6 -.emtcode OP_DISCARD 7 -.emtcode OP_RETURN 8 -.emtcode OP_EXPRESSION 9 -.emtcode OP_IF 10 -.emtcode OP_WHILE 11 -.emtcode OP_DO 12 -.emtcode OP_FOR 13 -.emtcode OP_PUSH_VOID 14 -.emtcode OP_PUSH_BOOL 15 -.emtcode OP_PUSH_INT 16 -.emtcode OP_PUSH_FLOAT 17 -.emtcode OP_PUSH_IDENTIFIER 18 -.emtcode OP_SEQUENCE 19 -.emtcode OP_ASSIGN 20 -.emtcode OP_ADDASSIGN 21 -.emtcode OP_SUBASSIGN 22 -.emtcode OP_MULASSIGN 23 -.emtcode OP_DIVASSIGN 24 -/*.emtcode OP_MODASSIGN 25*/ -/*.emtcode OP_LSHASSIGN 26*/ -/*.emtcode OP_RSHASSIGN 27*/ -/*.emtcode OP_ORASSIGN 28*/ -/*.emtcode OP_XORASSIGN 29*/ -/*.emtcode OP_ANDASSIGN 30*/ -.emtcode OP_SELECT 31 -.emtcode OP_LOGICALOR 32 -.emtcode OP_LOGICALXOR 33 -.emtcode OP_LOGICALAND 34 -/*.emtcode OP_BITOR 35*/ -/*.emtcode OP_BITXOR 36*/ -/*.emtcode OP_BITAND 37*/ -.emtcode OP_EQUAL 38 -.emtcode OP_NOTEQUAL 39 -.emtcode OP_LESS 40 -.emtcode OP_GREATER 41 -.emtcode OP_LESSEQUAL 42 -.emtcode OP_GREATEREQUAL 43 -/*.emtcode OP_LSHIFT 44*/ -/*.emtcode OP_RSHIFT 45*/ -.emtcode OP_ADD 46 -.emtcode OP_SUBTRACT 47 -.emtcode OP_MULTIPLY 48 -.emtcode OP_DIVIDE 49 -/*.emtcode OP_MODULUS 50*/ -.emtcode OP_PREINCREMENT 51 -.emtcode OP_PREDECREMENT 52 -.emtcode OP_PLUS 53 -.emtcode OP_MINUS 54 -/*.emtcode OP_COMPLEMENT 55*/ -.emtcode OP_NOT 56 -.emtcode OP_SUBSCRIPT 57 -.emtcode OP_CALL 58 -.emtcode OP_FIELD 59 -.emtcode OP_POSTINCREMENT 60 -.emtcode OP_POSTDECREMENT 61 -.emtcode OP_PRECISION 62 -.emtcode OP_METHOD 63 - -/* parameter qualifier */ -.emtcode PARAM_QUALIFIER_IN 0 -.emtcode PARAM_QUALIFIER_OUT 1 -.emtcode PARAM_QUALIFIER_INOUT 2 - -/* function parameter */ -.emtcode PARAMETER_NONE 0 -.emtcode PARAMETER_NEXT 1 - -/* function parameter array presence */ -.emtcode PARAMETER_ARRAY_NOT_PRESENT 0 -.emtcode PARAMETER_ARRAY_PRESENT 1 - -/* INVALID_EXTERNAL_DECLARATION seems to be reported when there's */ -/* any syntax errors... */ -.errtext INVALID_EXTERNAL_DECLARATION "2001: Syntax error." -.errtext INVALID_OPERATOR_OVERRIDE "2002: Invalid operator override." -.errtext LBRACE_EXPECTED "2003: '{' expected but '$err_token$' found." -.errtext LPAREN_EXPECTED "2004: '(' expected but '$err_token$' found." -.errtext RPAREN_EXPECTED "2005: ')' expected but '$err_token$' found." -.errtext INVALID_PRECISION "2006: Invalid precision specifier '$err_token$'." -.errtext INVALID_PRECISION_TYPE "2007: Invalid precision type '$err_token$'." - - -/* - * tells whether the shader that is being parsed is a built-in shader or not - * 0 - normal behaviour - * 1 - accepts constructor and operator definitions and __asm statements - * the implementation will set it to 1 when compiling internal built-in shaders - */ -.regbyte parsing_builtin 0 - -/* - * holds the type of the shader being parsed; possible values are - * listed below. - * FRAGMENT_SHADER 1 - * VERTEX_SHADER 2 - * shader type is set by the caller before parsing - */ -.regbyte shader_type 0 - -/* - * <variable_identifier> ::= <identifier> - */ -variable_identifier - identifier .emit OP_PUSH_IDENTIFIER; - -/* - * <primary_expression> ::= <variable_identifier> - * | <intconstant> - * | <floatconstant> - * | <boolconstant> - * | "(" <expression> ")" - */ -primary_expression - floatconstant .or boolconstant .or intconstant .or variable_identifier .or primary_expression_1; -primary_expression_1 - lparen .and expression .and rparen; - -/* - * <postfix_expression> ::= <primary_expression> - * | <postfix_expression> "[" <integer_expression> "]" - * | <function_call> - * | <postfix_expression> "." <field_selection> - * | <postfix_expression> "++" - * | <postfix_expression> "--" - */ -postfix_expression - postfix_expression_1 .and .loop postfix_expression_2; -postfix_expression_1 - function_call .or primary_expression; -postfix_expression_2 - postfix_expression_3 .or postfix_expression_4 .or - plusplus .emit OP_POSTINCREMENT .or - minusminus .emit OP_POSTDECREMENT; -postfix_expression_3 - lbracket .and integer_expression .and rbracket .emit OP_SUBSCRIPT; -postfix_expression_4 - dot .and field_selection .emit OP_FIELD; - -/* - * <integer_expression> ::= <expression> - */ -integer_expression - expression; - -/* - * <function_call> ::= <function_call_generic> - */ -function_call - function_call_or_method; - -/* - * <function_call_or_method> ::= <regular_function_call> - * | <postfix_expression> "." <function_call_generic> - */ -function_call_or_method - regular_function_call .or method_call; - -/* - * <method_call> ::= <identifier> "." <function_call_generic> - */ -method_call - identifier .emit OP_METHOD .and dot .and function_call_generic .and .true .emit OP_END; - -/* - * <regular_function_call> ::= <function_call_generic> - */ -regular_function_call - function_call_generic .emit OP_CALL .and .true .emit OP_END; - -/* - * <function_call_generic> ::= <function_call_header_with_parameters> ")" - * | <function_call_header_no_parameters> ")" - */ -function_call_generic - function_call_generic_1 .or function_call_generic_2; -function_call_generic_1 - function_call_header_with_parameters .and rparen .error RPAREN_EXPECTED; -function_call_generic_2 - function_call_header_no_parameters .and rparen .error RPAREN_EXPECTED; - -/* - * <function_call_header_no_parameters>::= <function_call_header> "void" - * | <function_call_header> - */ -function_call_header_no_parameters - function_call_header .and function_call_header_no_parameters_1; -function_call_header_no_parameters_1 - "void" .or .true; - -/* - * <function_call_header_with_parameters> ::= <function_call_header> <assignment_expression> - * | <function_call_header_with_parameters> "," <assignment_expression> - */ -function_call_header_with_parameters - function_call_header .and assignment_expression .and .true .emit OP_END .and - .loop function_call_header_with_parameters_1; -function_call_header_with_parameters_1 - comma .and assignment_expression .and .true .emit OP_END; - -/* - * <function_call_header> ::= <function_identifier> "(" - */ -function_call_header - function_identifier .and lparen; - -/* - * <function_identifier> ::= <constructor_identifier> - * | <identifier> - * | <type_specifier> - * - * note: <constructor_identifier> and <type_specifier> have been deleted - */ -function_identifier - identifier .and function_identifier_opt_array; -function_identifier_opt_array - function_identifier_array .emit FUNCTION_CALL_ARRAY .or - .true .emit FUNCTION_CALL_NONARRAY; -function_identifier_array - lbracket .and constant_expression .and rbracket; - -/* - * <unary_expression> ::= <postfix_expression> - * | "++" <unary_expression> - * | "--" <unary_expression> - * | <unary_operator> <unary_expression> - * - * <unary_operator> ::= "+" - * | "-" - * | "!" - * | "~" // reserved - */ -unary_expression - postfix_expression .or unary_expression_1 .or unary_expression_2 .or unary_expression_3 .or - unary_expression_4 .or unary_expression_5/* .or unary_expression_6*/; -unary_expression_1 - plusplus .and unary_expression .and .true .emit OP_PREINCREMENT; -unary_expression_2 - minusminus .and unary_expression .and .true .emit OP_PREDECREMENT; -unary_expression_3 - plus .and unary_expression .and .true .emit OP_PLUS; -unary_expression_4 - minus .and unary_expression .and .true .emit OP_MINUS; -unary_expression_5 - bang .and unary_expression .and .true .emit OP_NOT; -/*unary_expression_6 - tilde .and unary_expression .and .true .emit OP_COMPLEMENT;*/ - -/* - * <multiplicative_expression> ::= <unary_expression> - * | <multiplicative_expression> "*" <unary_expression> - * | <multiplicative_expression> "/" <unary_expression> - * | <multiplicative_expression> "%" <unary_expression> // reserved - */ -multiplicative_expression - unary_expression .and .loop multiplicative_expression_1; -multiplicative_expression_1 - multiplicative_expression_2 .or multiplicative_expression_3/* .or multiplicative_expression_4*/; -multiplicative_expression_2 - star .and unary_expression .and .true .emit OP_MULTIPLY; -multiplicative_expression_3 - slash .and unary_expression .and .true .emit OP_DIVIDE; -/*multiplicative_expression_4 - percent .and unary_expression .and .true .emit OP_MODULUS;*/ - -/* - * <additive_expression> ::= <multiplicative_expression> - * | <additive_expression> "+" <multiplicative_expression> - * | <additive_expression> "-" <multiplicative_expression> - */ -additive_expression - multiplicative_expression .and .loop additive_expression_1; -additive_expression_1 - additive_expression_2 .or additive_expression_3; -additive_expression_2 - plus .and multiplicative_expression .and .true .emit OP_ADD; -additive_expression_3 - minus .and multiplicative_expression .and .true .emit OP_SUBTRACT; - -/* - * <shift_expression> ::= <additive_expression> - * | <shift_expression> "<<" <additive_expression> // reserved - * | <shift_expression> ">>" <additive_expression> // reserved - */ -shift_expression - additive_expression/* .and .loop shift_expression_1*/; -/*shift_expression_1 - shift_expression_2 .or shift_expression_3;*/ -/*shift_expression_2 - lessless .and additive_expression .and .true .emit OP_LSHIFT;*/ -/*shift_expression_3 - greatergreater .and additive_expression .and .true .emit OP_RSHIFT;*/ - -/* - * <relational_expression> ::= <shift_expression> - * | <relational_expression> "<" <shift_expression> - * | <relational_expression> ">" <shift_expression> - * | <relational_expression> "<=" <shift_expression> - * | <relational_expression> ">=" <shift_expression> - */ -relational_expression - shift_expression .and .loop relational_expression_1; -relational_expression_1 - relational_expression_2 .or relational_expression_3 .or relational_expression_4 .or - relational_expression_5; -relational_expression_2 - lessequals .and shift_expression .and .true .emit OP_LESSEQUAL; -relational_expression_3 - greaterequals .and shift_expression .and .true .emit OP_GREATEREQUAL; -relational_expression_4 - less .and shift_expression .and .true .emit OP_LESS; -relational_expression_5 - greater .and shift_expression .and .true .emit OP_GREATER; - -/* - * <equality_expression> ::= <relational_expression> - * | <equality_expression> "==" <relational_expression> - * | <equality_expression> "!=" <relational_expression> - */ -equality_expression - relational_expression .and .loop equality_expression_1; -equality_expression_1 - equality_expression_2 .or equality_expression_3; -equality_expression_2 - equalsequals .and relational_expression .and .true .emit OP_EQUAL; -equality_expression_3 - bangequals .and relational_expression .and .true .emit OP_NOTEQUAL; - -/* - * <and_expression> ::= <equality_expression> - * | <and_expression> "&" <equality_expression> // reserved - */ -and_expression - equality_expression/* .and .loop and_expression_1*/; -/*and_expression_1 - ampersand .and equality_expression .and .true .emit OP_BITAND;*/ - -/* - * <exclusive_or_expression> ::= <and_expression> - * | <exclusive_or_expression> "^" <and_expression> // reserved - */ -exclusive_or_expression - and_expression/* .and .loop exclusive_or_expression_1*/; -/*exclusive_or_expression_1 - caret .and and_expression .and .true .emit OP_BITXOR;*/ - -/* - * <inclusive_or_expression> ::= <exclusive_or_expression> - * | <inclusive_or_expression> "|" <exclusive_or_expression> // reserved - */ -inclusive_or_expression - exclusive_or_expression/* .and .loop inclusive_or_expression_1*/; -/*inclusive_or_expression_1 - bar .and exclusive_or_expression .and .true .emit OP_BITOR;*/ - -/* - * <logical_and_expression> ::= <inclusive_or_expression> - * | <logical_and_expression> "&&" <inclusive_or_expression> - */ -logical_and_expression - inclusive_or_expression .and .loop logical_and_expression_1; -logical_and_expression_1 - ampersandampersand .and inclusive_or_expression .and .true .emit OP_LOGICALAND; - -/* - * <logical_xor_expression> ::= <logical_and_expression> - * | <logical_xor_expression> "^^" <logical_and_expression> - */ -logical_xor_expression - logical_and_expression .and .loop logical_xor_expression_1; -logical_xor_expression_1 - caretcaret .and logical_and_expression .and .true .emit OP_LOGICALXOR; - -/* - * <logical_or_expression> ::= <logical_xor_expression> - * | <logical_or_expression> "||" <logical_xor_expression> - */ -logical_or_expression - logical_xor_expression .and .loop logical_or_expression_1; -logical_or_expression_1 - barbar .and logical_xor_expression .and .true .emit OP_LOGICALOR; - -/* - * <conditional_expression> ::= <logical_or_expression> - * | <logical_or_expression> "?" <expression> ":" <conditional_expression> - */ -conditional_expression - logical_or_expression .and .loop conditional_expression_1; -conditional_expression_1 - question .and expression .and colon .and conditional_expression .and .true .emit OP_SELECT; - -/* - * <assignment_expression> ::= <conditional_expression> - * | <unary_expression> <assignment_operator> <assignment_expression> - * - * <assignment_operator> ::= "=" - * | "*=" - * | "/=" - * | "+=" - * | "-=" - * | "%=" // reserved - * | "<<=" // reserved - * | ">>=" // reserved - * | "&=" // reserved - * | "^=" // reserved - * | "|=" // reserved - */ -assignment_expression - assignment_expression_1 .or assignment_expression_2 .or assignment_expression_3 .or - assignment_expression_4 .or assignment_expression_5/* .or assignment_expression_6 .or - assignment_expression_7 .or assignment_expression_8 .or assignment_expression_9 .or - assignment_expression_10 .or assignment_expression_11*/ .or conditional_expression; -assignment_expression_1 - unary_expression .and equals .and assignment_expression .and .true .emit OP_ASSIGN; -assignment_expression_2 - unary_expression .and starequals .and assignment_expression .and .true .emit OP_MULASSIGN; -assignment_expression_3 - unary_expression .and slashequals .and assignment_expression .and .true .emit OP_DIVASSIGN; -assignment_expression_4 - unary_expression .and plusequals .and assignment_expression .and .true .emit OP_ADDASSIGN; -assignment_expression_5 - unary_expression .and minusequals .and assignment_expression .and .true .emit OP_SUBASSIGN; -/*assignment_expression_6 - unary_expression .and percentequals .and assignment_expression .and .true .emit OP_MODASSIGN;*/ -/*assignment_expression_7 - unary_expression .and lesslessequals .and assignment_expression .and .true .emit OP_LSHASSIGN;*/ -/*assignment_expression_8 - unary_expression .and greatergreaterequals .and assignment_expression .and - .true .emit OP_RSHASSIGN;*/ -/*assignment_expression_9 - unary_expression .and ampersandequals .and assignment_expression .and .true .emit OP_ANDASSIGN;*/ -/*assignment_expression_10 - unary_expression .and caretequals .and assignment_expression .and .true .emit OP_XORASSIGN;*/ -/*assignment_expression_11 - unary_expression .and barequals .and assignment_expression .and .true .emit OP_ORASSIGN;*/ - -/* - * <expression> ::= <assignment_expression> - * | <expression> "," <assignment_expression> - */ -expression - assignment_expression .and .loop expression_1; -expression_1 - comma .and assignment_expression .and .true .emit OP_SEQUENCE; - -/* - * <constant_expression> ::= <conditional_expression> - */ -constant_expression - conditional_expression .and .true .emit OP_END; - -/* - * <declaration> ::= <function_prototype> ";" - * | <init_declarator_list> ";" - */ -declaration - declaration_1 .or declaration_2; -declaration_1 - function_prototype .emit DECLARATION_FUNCTION_PROTOTYPE .and semicolon; -declaration_2 - init_declarator_list .emit DECLARATION_INIT_DECLARATOR_LIST .and semicolon; - -/* - * <function_prototype> ::= <function_header> "void" ")" - * | <function_declarator> ")" - */ -function_prototype - function_prototype_1 .or function_prototype_2; -function_prototype_1 - function_header .and "void" .and rparen .error RPAREN_EXPECTED .emit PARAMETER_NONE; -function_prototype_2 - function_declarator .and rparen .error RPAREN_EXPECTED .emit PARAMETER_NONE; - -/* - * <function_declarator> ::= <function_header> - * | <function_header_with_parameters> - */ -function_declarator - function_header_with_parameters .or function_header; - -/* - * <function_header_with_parameters> ::= <function_header> <parameter_declaration> - * | <function_header_with_parameters> "," - * <parameter_declaration> - */ -function_header_with_parameters - function_header .and parameter_declaration .and .loop function_header_with_parameters_1; -function_header_with_parameters_1 - comma .and parameter_declaration; - -/* - * <function_header> ::= <fully_specified_type> <identifier> "(" - */ -function_header - function_header_nospace .or function_header_space; -function_header_space - fully_specified_type_space .and space .and function_decl_identifier .and lparen; -function_header_nospace - fully_specified_type_nospace .and function_decl_identifier .and lparen; - -/* - * <function_decl_identifier> ::= "__constructor" - * | <__operator> - * | <identifier> - * - * note: this is an extension to the standard language specification. - * normally slang disallows operator and constructor prototypes and definitions - */ -function_decl_identifier - .if (parsing_builtin != 0) __operator .emit FUNCTION_OPERATOR .or - .if (parsing_builtin != 0) "__constructor" .emit FUNCTION_CONSTRUCTOR .or - identifier .emit FUNCTION_ORDINARY; - -/* - * <__operator> ::= "__operator" <overriden_op> - * - * note: this is an extension to the standard language specification. - * normally slang disallows operator prototypes and definitions - */ -__operator - "__operator" .and overriden_operator .error INVALID_OPERATOR_OVERRIDE; - -/* - * <overriden_op> ::= "=" - * | "+=" - * | "-=" - * | "*=" - * | "/=" - * | "%=" // reserved - * | "<<=" // reserved - * | ">>=" // reserved - * | "&=" // reserved - * | "^=" // reserved - * | "|=" // reserved - * | "^^" - * | "|" // reserved - * | "^" // reserved - * | "&" // reserved - * | "==" - * | "!=" - * | "<" - * | ">" - * | "<=" - * | ">=" - * | "<<" // reserved - * | ">>" // reserved - * | "*" - * | "/" - * | "%" // reserved - * | "++" - * | "--" - * | "+" - * | "-" - * | "~" // reserved - * | "!" - * - * note: this is an extension to the standard language specification. - * normally slang disallows operator prototypes and definitions - */ -overriden_operator - plusplus .emit OPERATOR_INCREMENT .or - plusequals .emit OPERATOR_ADDASSIGN .or - plus .emit OPERATOR_PLUS .or - minusminus .emit OPERATOR_DECREMENT .or - minusequals .emit OPERATOR_SUBASSIGN .or - minus .emit OPERATOR_MINUS .or - bang .emit OPERATOR_NOT .or - starequals .emit OPERATOR_MULASSIGN .or - star .emit OPERATOR_MULTIPLY .or - slashequals .emit OPERATOR_DIVASSIGN .or - slash .emit OPERATOR_DIVIDE .or - lessequals .emit OPERATOR_LESSEQUAL .or - /*lesslessequals .emit OPERATOR_LSHASSIGN .or*/ - /*lessless .emit OPERATOR_LSHIFT .or*/ - less .emit OPERATOR_LESS .or - greaterequals .emit OPERATOR_GREATEREQUAL .or - /*greatergreaterequals .emit OPERATOR_RSHASSIGN .or*/ - /*greatergreater .emit OPERATOR_RSHIFT .or*/ - greater .emit OPERATOR_GREATER .or - /*percentequals .emit OPERATOR_MODASSIGN .or*/ - /*percent .emit OPERATOR_MODULUS .or*/ - /*ampersandequals .emit OPERATOR_ANDASSIGN */ - /*ampersand .emit OPERATOR_BITAND .or*/ - /*barequals .emit OPERATOR_ORASSIGN .or*/ - /*bar .emit OPERATOR_BITOR .or*/ - /*tilde .emit OPERATOR_COMPLEMENT .or*/ - /*caretequals .emit OPERATOR_XORASSIGN .or*/ - caretcaret .emit OPERATOR_LOGICALXOR /*.or - caret .emit OPERATOR_BITXOR*/; - -/* - * <parameter_declarator> ::= <type_specifier> <identifier> - * | <type_specifier> <identifier> "[" <constant_expression> "]" - */ -parameter_declarator - parameter_declarator_nospace .or parameter_declarator_space; -parameter_declarator_nospace - type_specifier_nospace .and identifier .and parameter_declarator_1; -parameter_declarator_space - type_specifier_space .and space .and identifier .and parameter_declarator_1; -parameter_declarator_1 - parameter_declarator_2 .emit PARAMETER_ARRAY_PRESENT .or - .true .emit PARAMETER_ARRAY_NOT_PRESENT; -parameter_declarator_2 - lbracket .and constant_expression .and rbracket; - -/* - * <parameter_declaration> ::= <type_qualifier> <parameter_qualifier> - * <precision> <parameter_declarator> - * | <type_qualifier> <parameter_qualifier> - * <precision> <parameter_type_specifier> - * | <type_qualifier> <parameter_qualifier> - * <parameter_declarator> - * | <type_qualifier> <parameter_qualifier> - * <parameter_type_specifier> - * | <parameter_qualifier> <precision> - * <parameter_declarator> - * | <parameter_qualifier> <precision> - * <parameter_type_specifier> - * | <parameter_qualifier> <parameter_declarator> - * | <parameter_qualifier> <parameter_type_specifier> - */ -parameter_declaration - parameter_declaration_1 .emit PARAMETER_NEXT; -parameter_declaration_1 - parameter_declaration_2 .or parameter_declaration_3; -parameter_declaration_2 - type_qualifier .and space .and parameter_qualifier .and parameter_declaration_4; -parameter_declaration_3 - parameter_qualifier .emit TYPE_QUALIFIER_NONE .and parameter_declaration_4; -parameter_declaration_4 - parameter_declaration_optprec .and parameter_declaration_rest; -parameter_declaration_optprec - parameter_declaration_prec .or .true .emit PRECISION_DEFAULT; -parameter_declaration_prec - precision .and space; -parameter_declaration_rest - parameter_declarator .or parameter_type_specifier; - -/* - * <parameter_qualifier> ::= "in" - * | "out" - * | "inout" - * | "" - */ -parameter_qualifier - parameter_qualifier_1 .or .true .emit PARAM_QUALIFIER_IN; -parameter_qualifier_1 - parameter_qualifier_2 .and space; -parameter_qualifier_2 - "in" .emit PARAM_QUALIFIER_IN .or - "out" .emit PARAM_QUALIFIER_OUT .or - "inout" .emit PARAM_QUALIFIER_INOUT; - -/* - * <parameter_type_specifier> ::= <type_specifier> - * | <type_specifier> "[" <constant_expression> "]" - */ -parameter_type_specifier - parameter_type_specifier_1 .and .true .emit '\0' .and parameter_type_specifier_2; -parameter_type_specifier_1 - type_specifier_nospace .or type_specifier_space; -parameter_type_specifier_2 - parameter_type_specifier_3 .emit PARAMETER_ARRAY_PRESENT .or - .true .emit PARAMETER_ARRAY_NOT_PRESENT; -parameter_type_specifier_3 - lbracket .and constant_expression .and rbracket; - -/* - * <init_declarator_list> ::= <single_declaration> - * | <init_declarator_list> "," <identifier> - * | <init_declarator_list> "," <identifier> "[" "]" - * | <init_declarator_list> "," <identifier> "[" <constant_expression> "]" - * | <init_declarator_list> "," <identifier> "=" <initializer> - */ -init_declarator_list - single_declaration .and .loop init_declarator_list_1 .emit DECLARATOR_NEXT .and - .true .emit DECLARATOR_NONE; -init_declarator_list_1 - comma .and identifier .emit VARIABLE_IDENTIFIER .and init_declarator_list_2; -init_declarator_list_2 - init_declarator_list_3 .or init_declarator_list_4 .or .true .emit VARIABLE_NONE; -init_declarator_list_3 - equals .and initializer .emit VARIABLE_INITIALIZER; -init_declarator_list_4 - lbracket .and init_declarator_list_5 .and rbracket; -init_declarator_list_5 - constant_expression .emit VARIABLE_ARRAY_EXPLICIT .or .true .emit VARIABLE_ARRAY_UNKNOWN; - -/* - * <single_declaration> ::= <fully_specified_type> - * | <fully_specified_type> <identifier> - * | <fully_specified_type> <identifier> "[" "]" - * | <fully_specified_type> <identifier> "[" <constant_expression> "]" - * | <fully_specified_type> <identifier> "=" <initializer> - */ -single_declaration - single_declaration_nospace .or single_declaration_space; -single_declaration_space - fully_specified_type_space .and single_declaration_space_1; -single_declaration_nospace - fully_specified_type_nospace .and single_declaration_nospace_1; -single_declaration_space_1 - single_declaration_space_2 .emit VARIABLE_IDENTIFIER .or .true .emit VARIABLE_NONE; -single_declaration_nospace_1 - single_declaration_nospace_2 .emit VARIABLE_IDENTIFIER .or .true .emit VARIABLE_NONE; -single_declaration_space_2 - space .and identifier .and single_declaration_3; -single_declaration_nospace_2 - identifier .and single_declaration_3; -single_declaration_3 - single_declaration_4 .or single_declaration_5 .or .true .emit VARIABLE_NONE; -single_declaration_4 - equals .and initializer .emit VARIABLE_INITIALIZER; -single_declaration_5 - lbracket .and single_declaration_6 .and rbracket; -single_declaration_6 - constant_expression .emit VARIABLE_ARRAY_EXPLICIT .or .true .emit VARIABLE_ARRAY_UNKNOWN; - -/* - * <fully_specified_type> ::= <opt_invariant> <opt_centroid> <opt_qualifer> <opt_precision> <type_specifier> - * - * Example: "invariant varying highp vec3" - */ -fully_specified_type_space - fully_specified_type_optinvariant .and fully_specified_type_optcentroid .and fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_space; -fully_specified_type_nospace - fully_specified_type_optinvariant .and fully_specified_type_optcentroid .and fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_nospace; -fully_specified_type_optinvariant - fully_specified_type_invariant .or .true .emit TYPE_VARIANT; -fully_specified_type_invariant - invariant_qualifier .and space; -fully_specified_type_optcentroid - fully_specified_type_centroid .or .true .emit TYPE_CENTER; -fully_specified_type_centroid - centroid_qualifier .and space; -fully_specified_type_optqual - fully_specified_type_qual .or .true .emit TYPE_QUALIFIER_NONE; -fully_specified_type_qual - type_qualifier .and space; -fully_specified_type_optprec - fully_specified_type_prec .or .true .emit PRECISION_DEFAULT; -fully_specified_type_prec - precision .and space; - -/* - * <invariant_qualifier> ::= "invariant" - */ -invariant_qualifier - "invariant" .emit TYPE_INVARIANT; - -centroid_qualifier - "centroid" .emit TYPE_CENTROID; - - -/* - * <type_qualifier> ::= "const" - * | "attribute" // Vertex only. - * | "varying" - * | "uniform" - * | "__fixed_output" - * | "__fixed_input" - * - * note: this is an extension to the standard language specification, - * normally slang disallows __fixed_output and __fixed_input type qualifiers - */ -type_qualifier - "const" .emit TYPE_QUALIFIER_CONST .or - .if (shader_type == 2) "attribute" .emit TYPE_QUALIFIER_ATTRIBUTE .or - "varying" .emit TYPE_QUALIFIER_VARYING .or - "uniform" .emit TYPE_QUALIFIER_UNIFORM .or - .if (parsing_builtin != 0) "__fixed_output" .emit TYPE_QUALIFIER_FIXEDOUTPUT .or - .if (parsing_builtin != 0) "__fixed_input" .emit TYPE_QUALIFIER_FIXEDINPUT; - -/* - * <type_specifier_nonarray> ::= "void" - * | "float" - * | "int" - * | "bool" - * | "vec2" - * | "vec3" - * | "vec4" - * | "bvec2" - * | "bvec3" - * | "bvec4" - * | "ivec2" - * | "ivec3" - * | "ivec4" - * | "mat2" - * | "mat3" - * | "mat4" - * | "mat2x3" - * | "mat3x2" - * | "mat2x4" - * | "mat4x2" - * | "mat3x4" - * | "mat4x3" - * | "sampler1D" - * | "sampler2D" - * | "sampler3D" - * | "samplerCube" - * | "sampler1DShadow" - * | "sampler2DShadow" - * | "sampler2DRect" - * | "sampler2DRectShadow" - * | <struct_specifier> - * | <type_name> - */ -type_specifier_nonarray_space - "void" .emit TYPE_SPECIFIER_VOID .or - "float" .emit TYPE_SPECIFIER_FLOAT .or - "int" .emit TYPE_SPECIFIER_INT .or - "bool" .emit TYPE_SPECIFIER_BOOL .or - "vec2" .emit TYPE_SPECIFIER_VEC2 .or - "vec3" .emit TYPE_SPECIFIER_VEC3 .or - "vec4" .emit TYPE_SPECIFIER_VEC4 .or - "bvec2" .emit TYPE_SPECIFIER_BVEC2 .or - "bvec3" .emit TYPE_SPECIFIER_BVEC3 .or - "bvec4" .emit TYPE_SPECIFIER_BVEC4 .or - "ivec2" .emit TYPE_SPECIFIER_IVEC2 .or - "ivec3" .emit TYPE_SPECIFIER_IVEC3 .or - "ivec4" .emit TYPE_SPECIFIER_IVEC4 .or - "mat2" .emit TYPE_SPECIFIER_MAT2 .or - "mat3" .emit TYPE_SPECIFIER_MAT3 .or - "mat4" .emit TYPE_SPECIFIER_MAT4 .or - "mat2x3" .emit TYPE_SPECIFIER_MAT23 .or - "mat3x2" .emit TYPE_SPECIFIER_MAT32 .or - "mat2x4" .emit TYPE_SPECIFIER_MAT24 .or - "mat4x2" .emit TYPE_SPECIFIER_MAT42 .or - "mat3x4" .emit TYPE_SPECIFIER_MAT34 .or - "mat4x3" .emit TYPE_SPECIFIER_MAT43 .or - "sampler1D" .emit TYPE_SPECIFIER_SAMPLER1D .or - "sampler2D" .emit TYPE_SPECIFIER_SAMPLER2D .or - "sampler3D" .emit TYPE_SPECIFIER_SAMPLER3D .or - "samplerCube" .emit TYPE_SPECIFIER_SAMPLERCUBE .or - "sampler1DShadow" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or - "sampler2DShadow" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or - "sampler2DRect" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or - "sampler2DRectShadow" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW .or - type_name .emit TYPE_SPECIFIER_TYPENAME; -type_specifier_nonarray_nospace - struct_specifier .emit TYPE_SPECIFIER_STRUCT; -type_specifier_nonarray - type_specifier_nonarray_nospace .or type_specifier_nonarray_space; - -/* - * <type_specifier> ::= <type_specifier_nonarray> - * | <type_specifier_nonarray> "[" <constant_expression> "]" - */ -type_specifier_space - type_specifier_nonarray_space .and .true .emit TYPE_SPECIFIER_NONARRAY; -type_specifier_nospace - type_specifier_nospace_array .or type_specifier_nospace_1; -type_specifier_nospace_1 - type_specifier_nonarray_nospace .and .true .emit TYPE_SPECIFIER_NONARRAY; -type_specifier_nospace_array - type_specifier_nonarray .and lbracket .emit TYPE_SPECIFIER_ARRAY .and constant_expression .and rbracket; - -/* - * <struct_specifier> ::= "struct" <identifier> "{" <struct_declaration_list> "}" - * | "struct" "{" <struct_declaration_list> "}" - */ -struct_specifier - "struct" .and struct_specifier_1 .and optional_space .and lbrace .error LBRACE_EXPECTED .and - struct_declaration_list .and rbrace .emit FIELD_NONE; -struct_specifier_1 - struct_specifier_2 .or .true .emit '\0'; -struct_specifier_2 - space .and identifier; - -/* - * <struct_declaration_list> ::= <struct_declaration> - * | <struct_declaration_list> <struct_declaration> - */ -struct_declaration_list - struct_declaration .and .loop struct_declaration .emit FIELD_NEXT; - -/* - * <struct_declaration> ::= <type_specifier> <struct_declarator_list> ";" - */ -struct_declaration - struct_declaration_nospace .or struct_declaration_space; -struct_declaration_space - type_specifier_space .and space .and struct_declarator_list .and semicolon .emit FIELD_NONE; -struct_declaration_nospace - type_specifier_nospace .and struct_declarator_list .and semicolon .emit FIELD_NONE; - -/* - * <struct_declarator_list> ::= <struct_declarator> - * | <struct_declarator_list> "," <struct_declarator> - */ -struct_declarator_list - struct_declarator .and .loop struct_declarator_list_1 .emit FIELD_NEXT; -struct_declarator_list_1 - comma .and struct_declarator; - -/* - * <struct_declarator> ::= <identifier> - * | <identifier> "[" <constant_expression> "]" - */ -struct_declarator - identifier .and struct_declarator_1; -struct_declarator_1 - struct_declarator_2 .emit FIELD_ARRAY .or .true .emit FIELD_NONE; -struct_declarator_2 - lbracket .and constant_expression .and rbracket; - -/* - * <initializer> ::= <assignment_expression> - */ -initializer - assignment_expression .and .true .emit OP_END; - -/* - * <declaration_statement> ::= <declaration> - */ -declaration_statement - declaration; - -/* - * <statement> ::= <compound_statement> - * | <simple_statement> - */ -statement - compound_statement .or simple_statement; -statement_space - compound_statement .or statement_space_1; -statement_space_1 - space .and simple_statement; - -/* - * <simple_statement> ::= <__asm_statement> - * | <selection_statement> - * | <iteration_statement> - * | <jump_statement> - * | <expression_statement> - * | <declaration_statement> - * - * note: this is an extension to the standard language specification. - * normally slang disallows use of __asm statements - */ -simple_statement - .if (parsing_builtin != 0) __asm_statement .emit OP_ASM .or - selection_statement .or - iteration_statement .or - precision_stmt .emit OP_PRECISION .or - jump_statement .or - expression_statement .emit OP_EXPRESSION .or - declaration_statement .emit OP_DECLARE; - -/* - * <compound_statement> ::= "{" "}" - * | "{" <statement_list> "}" - */ -compound_statement - compound_statement_1 .emit OP_BLOCK_BEGIN_NEW_SCOPE .and .true .emit OP_END; -compound_statement_1 - compound_statement_2 .or compound_statement_3; -compound_statement_2 - lbrace .and rbrace; -compound_statement_3 - lbrace .and statement_list .and rbrace; - -/* - * <compound_statement_no_new_scope> ::= "{" "}" - * | "{" <statement_list> "}" - */ -compound_statement_no_new_scope - compound_statement_no_new_scope_1 .emit OP_BLOCK_BEGIN_NO_NEW_SCOPE .and .true .emit OP_END; -compound_statement_no_new_scope_1 - compound_statement_no_new_scope_2 .or compound_statement_no_new_scope_3; -compound_statement_no_new_scope_2 - lbrace .and rbrace; -compound_statement_no_new_scope_3 - lbrace .and statement_list .and rbrace; - - -/* - * <statement_list> ::= <statement> - * | <statement_list> <statement> - */ -statement_list - statement .and .loop statement; - -/* - * <expression_statement> ::= ";" - * | <expression> ";" - */ -expression_statement - expression_statement_1 .or expression_statement_2; -expression_statement_1 - semicolon .emit OP_PUSH_VOID .emit OP_END; -expression_statement_2 - expression .and semicolon .emit OP_END; - -/* - * <selection_statement> ::= "if" "(" <expression> ")" <selection_rest_statement> - */ -selection_statement - "if" .emit OP_IF .and lparen .error LPAREN_EXPECTED .and expression .and - rparen .error RPAREN_EXPECTED .emit OP_END .and selection_rest_statement; - -/* - * <selection_rest_statement> ::= <statement> "else" <statement> - * | <statement> - */ -selection_rest_statement - statement .and selection_rest_statement_1; -selection_rest_statement_1 - selection_rest_statement_2 .or .true .emit OP_EXPRESSION .emit OP_PUSH_VOID .emit OP_END; -selection_rest_statement_2 - "else" .and optional_space .and statement; - -/* - * <condition> ::= <expression> - * | <fully_specified_type> <identifier> "=" <initializer> - * - * note: if <condition_1> is executed, the emit format must - * match <declaration> emit format - */ -condition - condition_1 .emit OP_DECLARE .emit DECLARATION_INIT_DECLARATOR_LIST .or - condition_3 .emit OP_EXPRESSION; -condition_1 - condition_1_nospace .or condition_1_space; -condition_1_nospace - fully_specified_type_nospace .and condition_2; -condition_1_space - fully_specified_type_space .and space .and condition_2; -condition_2 - identifier .emit VARIABLE_IDENTIFIER .and equals .emit VARIABLE_INITIALIZER .and - initializer .and .true .emit DECLARATOR_NONE; -condition_3 - expression .and .true .emit OP_END; - -/* - * <iteration_statement> ::= "while" "(" <condition> ")" <statement> - * | "do" <statement> "while" "(" <expression> ")" ";" - * | "for" "(" <for_init_statement> <for_rest_statement> ")" <statement> - */ -iteration_statement - iteration_statement_1 .or iteration_statement_2 .or iteration_statement_3; -iteration_statement_1 - "while" .emit OP_WHILE .and lparen .error LPAREN_EXPECTED .and condition .and - rparen .error RPAREN_EXPECTED .and statement; -iteration_statement_2 - "do" .emit OP_DO .and statement_space .and "while" .and lparen .error LPAREN_EXPECTED .and - expression .and rparen .error RPAREN_EXPECTED .emit OP_END .and semicolon; -iteration_statement_3 - "for" .emit OP_FOR .and lparen .error LPAREN_EXPECTED .and for_init_statement .and - for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement; - -/* - * <for_init_statement> ::= <expression_statement> - * | <declaration_statement> - */ -for_init_statement - expression_statement .emit OP_EXPRESSION .or declaration_statement .emit OP_DECLARE; - -/* - * <conditionopt> ::= <condition> - * | "" - * - * note: <conditionopt> is used only by "for" statement. - * if <condition> is ommitted, parser simulates default behaviour, - * that is simulates "true" expression - */ -conditionopt - condition .or - .true .emit OP_EXPRESSION .emit OP_PUSH_BOOL .emit 2 .emit '1' .emit '\0' .emit OP_END; - -/* - * <for_rest_statement> ::= <conditionopt> ";" - * | <conditionopt> ";" <expression> - */ -for_rest_statement - conditionopt .and semicolon .and for_rest_statement_1; -for_rest_statement_1 - for_rest_statement_2 .or .true .emit OP_PUSH_VOID .emit OP_END; -for_rest_statement_2 - expression .and .true .emit OP_END; - -/* - * <jump_statement> ::= "continue" ";" - * | "break" ";" - * | "return" ";" - * | "return" <expression> ";" - * | "discard" ";" // Fragment shader only. - */ -jump_statement - jump_statement_1 .or jump_statement_2 .or jump_statement_3 .or jump_statement_4 .or - .if (shader_type == 1) jump_statement_5; -jump_statement_1 - "continue" .and semicolon .emit OP_CONTINUE; -jump_statement_2 - "break" .and semicolon .emit OP_BREAK; -jump_statement_3 - "return" .emit OP_RETURN .and optional_space .and expression .and semicolon .emit OP_END; -jump_statement_4 - "return" .emit OP_RETURN .and semicolon .emit OP_PUSH_VOID .emit OP_END; -jump_statement_5 - "discard" .and semicolon .emit OP_DISCARD; - -/* - * <__asm_statement> ::= "__asm" <identifier> <asm_arguments> ";" - * - * note: this is an extension to the standard language specification. - * normally slang disallows __asm statements - */ -__asm_statement - "__asm" .and space .and identifier .and space .and asm_arguments .and semicolon .emit OP_END; - -/* - * <asm_arguments> ::= <asm_argument> - * | <asm_arguments> "," <asm_argument> - * - * note: this is an extension to the standard language specification. - * normally slang disallows __asm statements - */ -asm_arguments - asm_argument .and .true .emit OP_END .and .loop asm_arguments_1; -asm_arguments_1 - comma .and asm_argument .and .true .emit OP_END; - -/* - * <asm_argument> ::= <variable_identifier> - * | <floatconstant> - * - * note: this is an extension to the standard language specification. - * normally slang disallows __asm statements - */ -asm_argument - var_with_field .or - variable_identifier .or - floatconstant; - -var_with_field - variable_identifier .and dot .and field_selection .emit OP_FIELD; - - -/* - * <translation_unit> ::= <external_declaration> - * | <translation_unit> <external_declaration> - */ -translation_unit - optional_space .emit REVISION .and external_declaration .error INVALID_EXTERNAL_DECLARATION .and - .loop external_declaration .and optional_space .and - '\0' .error INVALID_EXTERNAL_DECLARATION .emit EXTERNAL_NULL; - - -/* - * <external_declaration> ::= <function_definition> - * | <declaration> - */ -external_declaration - precision_stmt .emit DEFAULT_PRECISION .or - function_definition .emit EXTERNAL_FUNCTION_DEFINITION .or - invariant_stmt .emit INVARIANT_STMT .or - declaration .emit EXTERNAL_DECLARATION; - - -/* - * <precision_stmt> ::= "precision" <precision> <prectype> - */ -precision_stmt - "precision" .and space .and precision .error INVALID_PRECISION .and space .and prectype .error INVALID_PRECISION_TYPE .and semicolon; - -/* - * <precision> ::= "lowp" - * | "mediump" - * | "highp" - */ -precision - "lowp" .emit PRECISION_LOW .or - "mediump" .emit PRECISION_MEDIUM .or - "highp" .emit PRECISION_HIGH; - -/* - * <prectype> ::= "int" - * | "float" - * | "a sampler type" - */ -prectype - "int" .emit TYPE_SPECIFIER_INT .or - "float" .emit TYPE_SPECIFIER_FLOAT .or - "sampler1D" .emit TYPE_SPECIFIER_SAMPLER1D .or - "sampler2D" .emit TYPE_SPECIFIER_SAMPLER2D .or - "sampler3D" .emit TYPE_SPECIFIER_SAMPLER3D .or - "samplerCube" .emit TYPE_SPECIFIER_SAMPLERCUBE .or - "sampler1DShadow" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or - "sampler2DShadow" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or - "sampler2DRect" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or - "sampler2DRectShadow" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW; - - -/* - * <invariant_stmt> ::= "invariant" identifier; - */ -invariant_stmt - "invariant" .and space .and identifier .and semicolon; - - -/* - * <function_definition> :: <function_prototype> <compound_statement_no_new_scope> - */ -function_definition - function_prototype .and compound_statement_no_new_scope; - - - -/* - * helper rules, not part of the official language syntax - */ - -digit_oct - '0'-'7'; - -digit_dec - '0'-'9'; - -digit_hex - '0'-'9' .or 'A'-'F' .or 'a'-'f'; - -id_character_first - 'a'-'z' .or 'A'-'Z' .or '_'; - -id_character_next - id_character_first .or digit_dec; - -identifier - id_character_first .emit * .and .loop id_character_next .emit * .and .true .emit '\0'; - -float - float_1 .or float_2 .or float_3; -float_1 - float_fractional_constant .and float_optional_exponent_part .and optional_f_suffix; -float_2 - float_digit_sequence .and .true .emit '\0' .and float_exponent_part .and optional_f_suffix; -float_3 - float_digit_sequence .and .true .emit '\0' .and 'f' .emit '\0'; - -float_fractional_constant - float_fractional_constant_1 .or float_fractional_constant_2 .or float_fractional_constant_3; -float_fractional_constant_1 - float_digit_sequence .and '.' .and float_digit_sequence; -float_fractional_constant_2 - float_digit_sequence .and '.' .and .true .emit '\0'; -float_fractional_constant_3 - '.' .emit '\0' .and float_digit_sequence; - -float_optional_exponent_part - float_exponent_part .or .true .emit '\0'; - -float_digit_sequence - digit_dec .emit * .and .loop digit_dec .emit * .and .true .emit '\0'; - -float_exponent_part - float_exponent_part_1 .or float_exponent_part_2; -float_exponent_part_1 - 'e' .and float_optional_sign .and float_digit_sequence; -float_exponent_part_2 - 'E' .and float_optional_sign .and float_digit_sequence; - -float_optional_sign - float_sign .or .true; - -float_sign - '+' .or '-' .emit '-'; - -optional_f_suffix - 'f' .or .true; - - -integer - integer_hex .or integer_oct .or integer_dec; - -integer_hex - '0' .and integer_hex_1 .emit 0x10 .and digit_hex .emit * .and .loop digit_hex .emit * .and - .true .emit '\0'; -integer_hex_1 - 'x' .or 'X'; - -integer_oct - '0' .emit 8 .emit * .and .loop digit_oct .emit * .and .true .emit '\0'; - -integer_dec - digit_dec .emit 10 .emit * .and .loop digit_dec .emit * .and .true .emit '\0'; - -boolean - "true" .emit 2 .emit '1' .emit '\0' .or - "false" .emit 2 .emit '0' .emit '\0'; - -type_name - identifier; - -field_selection - identifier; - -floatconstant - float .emit OP_PUSH_FLOAT; - -intconstant - integer .emit OP_PUSH_INT; - -boolconstant - boolean .emit OP_PUSH_BOOL; - -optional_space - .loop single_space; - -space - single_space .and .loop single_space; - -single_space - white_char .or c_style_comment_block .or cpp_style_comment_block; - -white_char - ' ' .or '\t' .or new_line .or '\v' .or '\f'; - -new_line - cr_lf .or lf_cr .or '\n' .or '\r'; - -cr_lf - '\r' .and '\n'; - -lf_cr - '\n' .and '\r'; - -c_style_comment_block - '/' .and '*' .and c_style_comment_rest; - -c_style_comment_rest - .loop c_style_comment_char_no_star .and c_style_comment_rest_1; -c_style_comment_rest_1 - c_style_comment_end .or c_style_comment_rest_2; -c_style_comment_rest_2 - '*' .and c_style_comment_rest; - -c_style_comment_char_no_star - '\x2B'-'\xFF' .or '\x01'-'\x29'; - -c_style_comment_end - '*' .and '/'; - -cpp_style_comment_block - '/' .and '/' .and cpp_style_comment_block_1; -cpp_style_comment_block_1 - cpp_style_comment_block_2 .or cpp_style_comment_block_3; -cpp_style_comment_block_2 - .loop cpp_style_comment_char .and new_line; -cpp_style_comment_block_3 - .loop cpp_style_comment_char; - -cpp_style_comment_char - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -/* lexical rules */ - -/*ampersand - optional_space .and '&' .and optional_space;*/ - -ampersandampersand - optional_space .and '&' .and '&' .and optional_space; - -/*ampersandequals - optional_space .and '&' .and '=' .and optional_space;*/ - -/*bar - optional_space .and '|' .and optional_space;*/ - -barbar - optional_space .and '|' .and '|' .and optional_space; - -/*barequals - optional_space .and '|' .and '=' .and optional_space;*/ - -bang - optional_space .and '!' .and optional_space; - -bangequals - optional_space .and '!' .and '=' .and optional_space; - -/*caret - optional_space .and '^' .and optional_space;*/ - -caretcaret - optional_space .and '^' .and '^' .and optional_space; - -/*caretequals - optional_space .and '^' .and '=' .and optional_space;*/ - -colon - optional_space .and ':' .and optional_space; - -comma - optional_space .and ',' .and optional_space; - -dot - optional_space .and '.' .and optional_space; - -equals - optional_space .and '=' .and optional_space; - -equalsequals - optional_space .and '=' .and '=' .and optional_space; - -greater - optional_space .and '>' .and optional_space; - -greaterequals - optional_space .and '>' .and '=' .and optional_space; - -/*greatergreater - optional_space .and '>' .and '>' .and optional_space;*/ - -/*greatergreaterequals - optional_space .and '>' .and '>' .and '=' .and optional_space;*/ - -lbrace - optional_space .and '{' .and optional_space; - -lbracket - optional_space .and '[' .and optional_space; - -less - optional_space .and '<' .and optional_space; - -lessequals - optional_space .and '<' .and '=' .and optional_space; - -/*lessless - optional_space .and '<' .and '<' .and optional_space;*/ - -/*lesslessequals - optional_space .and '<' .and '<' .and '=' .and optional_space;*/ - -lparen - optional_space .and '(' .and optional_space; - -minus - optional_space .and '-' .and optional_space; - -minusequals - optional_space .and '-' .and '=' .and optional_space; - -minusminus - optional_space .and '-' .and '-' .and optional_space; - -/*percent - optional_space .and '%' .and optional_space;*/ - -/*percentequals - optional_space .and '%' .and '=' .and optional_space;*/ - -plus - optional_space .and '+' .and optional_space; - -plusequals - optional_space .and '+' .and '=' .and optional_space; - -plusplus - optional_space .and '+' .and '+' .and optional_space; - -question - optional_space .and '?' .and optional_space; - -rbrace - optional_space .and '}' .and optional_space; - -rbracket - optional_space .and ']' .and optional_space; - -rparen - optional_space .and ')' .and optional_space; - -semicolon - optional_space .and ';' .and optional_space; - -slash - optional_space .and '/' .and optional_space; - -slashequals - optional_space .and '/' .and '=' .and optional_space; - -star - optional_space .and '*' .and optional_space; - -starequals - optional_space .and '*' .and '=' .and optional_space; - -/*tilde - optional_space .and '~' .and optional_space;*/ - -/* string rules - these are used internally by the parser when parsing quoted strings */ - -.string string_lexer; - -string_lexer - lex_first_identifier_character .and .loop lex_next_identifier_character; - -lex_first_identifier_character - 'a'-'z' .or 'A'-'Z' .or '_'; - -lex_next_identifier_character - 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_'; - -/* error rules - these are used by error messages */ - -err_token - '~' .or '`' .or '!' .or '@' .or '#' .or '$' .or '%' .or '^' .or '&' .or '*' .or '(' .or ')' .or - '-' .or '+' .or '=' .or '|' .or '\\' .or '[' .or ']' .or '{' .or '}' .or ':' .or ';' .or '"' .or - '\'' .or '<' .or ',' .or '>' .or '.' .or '/' .or '?' .or err_identifier; - -err_identifier - id_character_first .and .loop id_character_next; - diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h deleted file mode 100644 index 6a382970e1a..00000000000 --- a/src/mesa/shader/slang/library/slang_shader_syn.h +++ /dev/null @@ -1,866 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */ - -".syntax translation_unit;\n" -".emtcode REVISION 5\n" -".emtcode EXTERNAL_NULL 0\n" -".emtcode EXTERNAL_FUNCTION_DEFINITION 1\n" -".emtcode EXTERNAL_DECLARATION 2\n" -".emtcode DEFAULT_PRECISION 3\n" -".emtcode INVARIANT_STMT 4\n" -".emtcode PRECISION_DEFAULT 0\n" -".emtcode PRECISION_LOW 1\n" -".emtcode PRECISION_MEDIUM 2\n" -".emtcode PRECISION_HIGH 3\n" -".emtcode DECLARATION_FUNCTION_PROTOTYPE 1\n" -".emtcode DECLARATION_INIT_DECLARATOR_LIST 2\n" -".emtcode FUNCTION_ORDINARY 0\n" -".emtcode FUNCTION_CONSTRUCTOR 1\n" -".emtcode FUNCTION_OPERATOR 2\n" -".emtcode FUNCTION_CALL_NONARRAY 0\n" -".emtcode FUNCTION_CALL_ARRAY 1\n" -".emtcode OPERATOR_ADDASSIGN 1\n" -".emtcode OPERATOR_SUBASSIGN 2\n" -".emtcode OPERATOR_MULASSIGN 3\n" -".emtcode OPERATOR_DIVASSIGN 4\n" -".emtcode OPERATOR_LOGICALXOR 11\n" -".emtcode OPERATOR_LESS 15\n" -".emtcode OPERATOR_GREATER 16\n" -".emtcode OPERATOR_LESSEQUAL 17\n" -".emtcode OPERATOR_GREATEREQUAL 18\n" -".emtcode OPERATOR_MULTIPLY 21\n" -".emtcode OPERATOR_DIVIDE 22\n" -".emtcode OPERATOR_INCREMENT 24\n" -".emtcode OPERATOR_DECREMENT 25\n" -".emtcode OPERATOR_PLUS 26\n" -".emtcode OPERATOR_MINUS 27\n" -".emtcode OPERATOR_NOT 29\n" -".emtcode DECLARATOR_NONE 0\n" -".emtcode DECLARATOR_NEXT 1\n" -".emtcode VARIABLE_NONE 0\n" -".emtcode VARIABLE_IDENTIFIER 1\n" -".emtcode VARIABLE_INITIALIZER 2\n" -".emtcode VARIABLE_ARRAY_EXPLICIT 3\n" -".emtcode VARIABLE_ARRAY_UNKNOWN 4\n" -".emtcode TYPE_QUALIFIER_NONE 0\n" -".emtcode TYPE_QUALIFIER_CONST 1\n" -".emtcode TYPE_QUALIFIER_ATTRIBUTE 2\n" -".emtcode TYPE_QUALIFIER_VARYING 3\n" -".emtcode TYPE_QUALIFIER_UNIFORM 4\n" -".emtcode TYPE_QUALIFIER_FIXEDOUTPUT 5\n" -".emtcode TYPE_QUALIFIER_FIXEDINPUT 6\n" -".emtcode TYPE_VARIANT 90\n" -".emtcode TYPE_INVARIANT 91\n" -".emtcode TYPE_CENTER 95\n" -".emtcode TYPE_CENTROID 96\n" -".emtcode TYPE_SPECIFIER_VOID 0\n" -".emtcode TYPE_SPECIFIER_BOOL 1\n" -".emtcode TYPE_SPECIFIER_BVEC2 2\n" -".emtcode TYPE_SPECIFIER_BVEC3 3\n" -".emtcode TYPE_SPECIFIER_BVEC4 4\n" -".emtcode TYPE_SPECIFIER_INT 5\n" -".emtcode TYPE_SPECIFIER_IVEC2 6\n" -".emtcode TYPE_SPECIFIER_IVEC3 7\n" -".emtcode TYPE_SPECIFIER_IVEC4 8\n" -".emtcode TYPE_SPECIFIER_FLOAT 9\n" -".emtcode TYPE_SPECIFIER_VEC2 10\n" -".emtcode TYPE_SPECIFIER_VEC3 11\n" -".emtcode TYPE_SPECIFIER_VEC4 12\n" -".emtcode TYPE_SPECIFIER_MAT2 13\n" -".emtcode TYPE_SPECIFIER_MAT3 14\n" -".emtcode TYPE_SPECIFIER_MAT4 15\n" -".emtcode TYPE_SPECIFIER_SAMPLER1D 16\n" -".emtcode TYPE_SPECIFIER_SAMPLER2D 17\n" -".emtcode TYPE_SPECIFIER_SAMPLER3D 18\n" -".emtcode TYPE_SPECIFIER_SAMPLERCUBE 19\n" -".emtcode TYPE_SPECIFIER_SAMPLER1DSHADOW 20\n" -".emtcode TYPE_SPECIFIER_SAMPLER2DSHADOW 21\n" -".emtcode TYPE_SPECIFIER_SAMPLER2DRECT 22\n" -".emtcode TYPE_SPECIFIER_SAMPLER2DRECTSHADOW 23\n" -".emtcode TYPE_SPECIFIER_STRUCT 24\n" -".emtcode TYPE_SPECIFIER_TYPENAME 25\n" -".emtcode TYPE_SPECIFIER_MAT23 26\n" -".emtcode TYPE_SPECIFIER_MAT32 27\n" -".emtcode TYPE_SPECIFIER_MAT24 28\n" -".emtcode TYPE_SPECIFIER_MAT42 29\n" -".emtcode TYPE_SPECIFIER_MAT34 30\n" -".emtcode TYPE_SPECIFIER_MAT43 31\n" -".emtcode TYPE_SPECIFIER_NONARRAY 0\n" -".emtcode TYPE_SPECIFIER_ARRAY 1\n" -".emtcode FIELD_NONE 0\n" -".emtcode FIELD_NEXT 1\n" -".emtcode FIELD_ARRAY 2\n" -".emtcode OP_END 0\n" -".emtcode OP_BLOCK_BEGIN_NO_NEW_SCOPE 1\n" -".emtcode OP_BLOCK_BEGIN_NEW_SCOPE 2\n" -".emtcode OP_DECLARE 3\n" -".emtcode OP_ASM 4\n" -".emtcode OP_BREAK 5\n" -".emtcode OP_CONTINUE 6\n" -".emtcode OP_DISCARD 7\n" -".emtcode OP_RETURN 8\n" -".emtcode OP_EXPRESSION 9\n" -".emtcode OP_IF 10\n" -".emtcode OP_WHILE 11\n" -".emtcode OP_DO 12\n" -".emtcode OP_FOR 13\n" -".emtcode OP_PUSH_VOID 14\n" -".emtcode OP_PUSH_BOOL 15\n" -".emtcode OP_PUSH_INT 16\n" -".emtcode OP_PUSH_FLOAT 17\n" -".emtcode OP_PUSH_IDENTIFIER 18\n" -".emtcode OP_SEQUENCE 19\n" -".emtcode OP_ASSIGN 20\n" -".emtcode OP_ADDASSIGN 21\n" -".emtcode OP_SUBASSIGN 22\n" -".emtcode OP_MULASSIGN 23\n" -".emtcode OP_DIVASSIGN 24\n" -".emtcode OP_SELECT 31\n" -".emtcode OP_LOGICALOR 32\n" -".emtcode OP_LOGICALXOR 33\n" -".emtcode OP_LOGICALAND 34\n" -".emtcode OP_EQUAL 38\n" -".emtcode OP_NOTEQUAL 39\n" -".emtcode OP_LESS 40\n" -".emtcode OP_GREATER 41\n" -".emtcode OP_LESSEQUAL 42\n" -".emtcode OP_GREATEREQUAL 43\n" -".emtcode OP_ADD 46\n" -".emtcode OP_SUBTRACT 47\n" -".emtcode OP_MULTIPLY 48\n" -".emtcode OP_DIVIDE 49\n" -".emtcode OP_PREINCREMENT 51\n" -".emtcode OP_PREDECREMENT 52\n" -".emtcode OP_PLUS 53\n" -".emtcode OP_MINUS 54\n" -".emtcode OP_NOT 56\n" -".emtcode OP_SUBSCRIPT 57\n" -".emtcode OP_CALL 58\n" -".emtcode OP_FIELD 59\n" -".emtcode OP_POSTINCREMENT 60\n" -".emtcode OP_POSTDECREMENT 61\n" -".emtcode OP_PRECISION 62\n" -".emtcode OP_METHOD 63\n" -".emtcode PARAM_QUALIFIER_IN 0\n" -".emtcode PARAM_QUALIFIER_OUT 1\n" -".emtcode PARAM_QUALIFIER_INOUT 2\n" -".emtcode PARAMETER_NONE 0\n" -".emtcode PARAMETER_NEXT 1\n" -".emtcode PARAMETER_ARRAY_NOT_PRESENT 0\n" -".emtcode PARAMETER_ARRAY_PRESENT 1\n" -".errtext INVALID_EXTERNAL_DECLARATION \"2001: Syntax error.\"\n" -".errtext INVALID_OPERATOR_OVERRIDE \"2002: Invalid operator override.\"\n" -".errtext LBRACE_EXPECTED \"2003: '{' expected but '$err_token$' found.\"\n" -".errtext LPAREN_EXPECTED \"2004: '(' expected but '$err_token$' found.\"\n" -".errtext RPAREN_EXPECTED \"2005: ')' expected but '$err_token$' found.\"\n" -".errtext INVALID_PRECISION \"2006: Invalid precision specifier '$err_token$'.\"\n" -".errtext INVALID_PRECISION_TYPE \"2007: Invalid precision type '$err_token$'.\"\n" -".regbyte parsing_builtin 0\n" -".regbyte shader_type 0\n" -"variable_identifier\n" -" identifier .emit OP_PUSH_IDENTIFIER;\n" -"primary_expression\n" -" floatconstant .or boolconstant .or intconstant .or variable_identifier .or primary_expression_1;\n" -"primary_expression_1\n" -" lparen .and expression .and rparen;\n" -"postfix_expression\n" -" postfix_expression_1 .and .loop postfix_expression_2;\n" -"postfix_expression_1\n" -" function_call .or primary_expression;\n" -"postfix_expression_2\n" -" postfix_expression_3 .or postfix_expression_4 .or\n" -" plusplus .emit OP_POSTINCREMENT .or\n" -" minusminus .emit OP_POSTDECREMENT;\n" -"postfix_expression_3\n" -" lbracket .and integer_expression .and rbracket .emit OP_SUBSCRIPT;\n" -"postfix_expression_4\n" -" dot .and field_selection .emit OP_FIELD;\n" -"integer_expression\n" -" expression;\n" -"function_call\n" -" function_call_or_method;\n" -"function_call_or_method\n" -" regular_function_call .or method_call;\n" -"method_call\n" -" identifier .emit OP_METHOD .and dot .and function_call_generic .and .true .emit OP_END;\n" -"regular_function_call\n" -" function_call_generic .emit OP_CALL .and .true .emit OP_END;\n" -"function_call_generic\n" -" function_call_generic_1 .or function_call_generic_2;\n" -"function_call_generic_1\n" -" function_call_header_with_parameters .and rparen .error RPAREN_EXPECTED;\n" -"function_call_generic_2\n" -" function_call_header_no_parameters .and rparen .error RPAREN_EXPECTED;\n" -"function_call_header_no_parameters\n" -" function_call_header .and function_call_header_no_parameters_1;\n" -"function_call_header_no_parameters_1\n" -" \"void\" .or .true;\n" -"function_call_header_with_parameters\n" -" function_call_header .and assignment_expression .and .true .emit OP_END .and\n" -" .loop function_call_header_with_parameters_1;\n" -"function_call_header_with_parameters_1\n" -" comma .and assignment_expression .and .true .emit OP_END;\n" -"function_call_header\n" -" function_identifier .and lparen;\n" -"function_identifier\n" -" identifier .and function_identifier_opt_array;\n" -"function_identifier_opt_array\n" -" function_identifier_array .emit FUNCTION_CALL_ARRAY .or\n" -" .true .emit FUNCTION_CALL_NONARRAY;\n" -"function_identifier_array\n" -" lbracket .and constant_expression .and rbracket;\n" -"unary_expression\n" -" postfix_expression .or unary_expression_1 .or unary_expression_2 .or unary_expression_3 .or\n" -" unary_expression_4 .or unary_expression_5;\n" -"unary_expression_1\n" -" plusplus .and unary_expression .and .true .emit OP_PREINCREMENT;\n" -"unary_expression_2\n" -" minusminus .and unary_expression .and .true .emit OP_PREDECREMENT;\n" -"unary_expression_3\n" -" plus .and unary_expression .and .true .emit OP_PLUS;\n" -"unary_expression_4\n" -" minus .and unary_expression .and .true .emit OP_MINUS;\n" -"unary_expression_5\n" -" bang .and unary_expression .and .true .emit OP_NOT;\n" -"multiplicative_expression\n" -" unary_expression .and .loop multiplicative_expression_1;\n" -"multiplicative_expression_1\n" -" multiplicative_expression_2 .or multiplicative_expression_3;\n" -"multiplicative_expression_2\n" -" star .and unary_expression .and .true .emit OP_MULTIPLY;\n" -"multiplicative_expression_3\n" -" slash .and unary_expression .and .true .emit OP_DIVIDE;\n" -"additive_expression\n" -" multiplicative_expression .and .loop additive_expression_1;\n" -"additive_expression_1\n" -" additive_expression_2 .or additive_expression_3;\n" -"additive_expression_2\n" -" plus .and multiplicative_expression .and .true .emit OP_ADD;\n" -"additive_expression_3\n" -" minus .and multiplicative_expression .and .true .emit OP_SUBTRACT;\n" -"shift_expression\n" -" additive_expression;\n" -"relational_expression\n" -" shift_expression .and .loop relational_expression_1;\n" -"relational_expression_1\n" -" relational_expression_2 .or relational_expression_3 .or relational_expression_4 .or\n" -" relational_expression_5;\n" -"relational_expression_2\n" -" lessequals .and shift_expression .and .true .emit OP_LESSEQUAL;\n" -"relational_expression_3\n" -" greaterequals .and shift_expression .and .true .emit OP_GREATEREQUAL;\n" -"relational_expression_4\n" -" less .and shift_expression .and .true .emit OP_LESS;\n" -"relational_expression_5\n" -" greater .and shift_expression .and .true .emit OP_GREATER;\n" -"equality_expression\n" -" relational_expression .and .loop equality_expression_1;\n" -"equality_expression_1\n" -" equality_expression_2 .or equality_expression_3;\n" -"equality_expression_2\n" -" equalsequals .and relational_expression .and .true .emit OP_EQUAL;\n" -"equality_expression_3\n" -" bangequals .and relational_expression .and .true .emit OP_NOTEQUAL;\n" -"and_expression\n" -" equality_expression;\n" -"exclusive_or_expression\n" -" and_expression;\n" -"inclusive_or_expression\n" -" exclusive_or_expression;\n" -"logical_and_expression\n" -" inclusive_or_expression .and .loop logical_and_expression_1;\n" -"logical_and_expression_1\n" -" ampersandampersand .and inclusive_or_expression .and .true .emit OP_LOGICALAND;\n" -"logical_xor_expression\n" -" logical_and_expression .and .loop logical_xor_expression_1;\n" -"logical_xor_expression_1\n" -" caretcaret .and logical_and_expression .and .true .emit OP_LOGICALXOR;\n" -"logical_or_expression\n" -" logical_xor_expression .and .loop logical_or_expression_1;\n" -"logical_or_expression_1\n" -" barbar .and logical_xor_expression .and .true .emit OP_LOGICALOR;\n" -"conditional_expression\n" -" logical_or_expression .and .loop conditional_expression_1;\n" -"conditional_expression_1\n" -" question .and expression .and colon .and conditional_expression .and .true .emit OP_SELECT;\n" -"assignment_expression\n" -" assignment_expression_1 .or assignment_expression_2 .or assignment_expression_3 .or\n" -" assignment_expression_4 .or assignment_expression_5 .or conditional_expression;\n" -"assignment_expression_1\n" -" unary_expression .and equals .and assignment_expression .and .true .emit OP_ASSIGN;\n" -"assignment_expression_2\n" -" unary_expression .and starequals .and assignment_expression .and .true .emit OP_MULASSIGN;\n" -"assignment_expression_3\n" -" unary_expression .and slashequals .and assignment_expression .and .true .emit OP_DIVASSIGN;\n" -"assignment_expression_4\n" -" unary_expression .and plusequals .and assignment_expression .and .true .emit OP_ADDASSIGN;\n" -"assignment_expression_5\n" -" unary_expression .and minusequals .and assignment_expression .and .true .emit OP_SUBASSIGN;\n" -"expression\n" -" assignment_expression .and .loop expression_1;\n" -"expression_1\n" -" comma .and assignment_expression .and .true .emit OP_SEQUENCE;\n" -"constant_expression\n" -" conditional_expression .and .true .emit OP_END;\n" -"declaration\n" -" declaration_1 .or declaration_2;\n" -"declaration_1\n" -" function_prototype .emit DECLARATION_FUNCTION_PROTOTYPE .and semicolon;\n" -"declaration_2\n" -" init_declarator_list .emit DECLARATION_INIT_DECLARATOR_LIST .and semicolon;\n" -"function_prototype\n" -" function_prototype_1 .or function_prototype_2;\n" -"function_prototype_1\n" -" function_header .and \"void\" .and rparen .error RPAREN_EXPECTED .emit PARAMETER_NONE;\n" -"function_prototype_2\n" -" function_declarator .and rparen .error RPAREN_EXPECTED .emit PARAMETER_NONE;\n" -"function_declarator\n" -" function_header_with_parameters .or function_header;\n" -"function_header_with_parameters\n" -" function_header .and parameter_declaration .and .loop function_header_with_parameters_1;\n" -"function_header_with_parameters_1\n" -" comma .and parameter_declaration;\n" -"function_header\n" -" function_header_nospace .or function_header_space;\n" -"function_header_space\n" -" fully_specified_type_space .and space .and function_decl_identifier .and lparen;\n" -"function_header_nospace\n" -" fully_specified_type_nospace .and function_decl_identifier .and lparen;\n" -"function_decl_identifier\n" -" .if (parsing_builtin != 0) __operator .emit FUNCTION_OPERATOR .or\n" -" .if (parsing_builtin != 0) \"__constructor\" .emit FUNCTION_CONSTRUCTOR .or\n" -" identifier .emit FUNCTION_ORDINARY;\n" -"__operator\n" -" \"__operator\" .and overriden_operator .error INVALID_OPERATOR_OVERRIDE;\n" -"overriden_operator\n" -" plusplus .emit OPERATOR_INCREMENT .or\n" -" plusequals .emit OPERATOR_ADDASSIGN .or\n" -" plus .emit OPERATOR_PLUS .or\n" -" minusminus .emit OPERATOR_DECREMENT .or\n" -" minusequals .emit OPERATOR_SUBASSIGN .or\n" -" minus .emit OPERATOR_MINUS .or\n" -" bang .emit OPERATOR_NOT .or\n" -" starequals .emit OPERATOR_MULASSIGN .or\n" -" star .emit OPERATOR_MULTIPLY .or\n" -" slashequals .emit OPERATOR_DIVASSIGN .or\n" -" slash .emit OPERATOR_DIVIDE .or\n" -" lessequals .emit OPERATOR_LESSEQUAL .or\n" -" \n" -" \n" -" less .emit OPERATOR_LESS .or\n" -" greaterequals .emit OPERATOR_GREATEREQUAL .or\n" -" \n" -" \n" -" greater .emit OPERATOR_GREATER .or\n" -" \n" -" \n" -" \n" -" \n" -" \n" -" \n" -" \n" -" \n" -" caretcaret .emit OPERATOR_LOGICALXOR ;\n" -"parameter_declarator\n" -" parameter_declarator_nospace .or parameter_declarator_space;\n" -"parameter_declarator_nospace\n" -" type_specifier_nospace .and identifier .and parameter_declarator_1;\n" -"parameter_declarator_space\n" -" type_specifier_space .and space .and identifier .and parameter_declarator_1;\n" -"parameter_declarator_1\n" -" parameter_declarator_2 .emit PARAMETER_ARRAY_PRESENT .or\n" -" .true .emit PARAMETER_ARRAY_NOT_PRESENT;\n" -"parameter_declarator_2\n" -" lbracket .and constant_expression .and rbracket;\n" -"parameter_declaration\n" -" parameter_declaration_1 .emit PARAMETER_NEXT;\n" -"parameter_declaration_1\n" -" parameter_declaration_2 .or parameter_declaration_3;\n" -"parameter_declaration_2\n" -" type_qualifier .and space .and parameter_qualifier .and parameter_declaration_4;\n" -"parameter_declaration_3\n" -" parameter_qualifier .emit TYPE_QUALIFIER_NONE .and parameter_declaration_4;\n" -"parameter_declaration_4\n" -" parameter_declaration_optprec .and parameter_declaration_rest;\n" -"parameter_declaration_optprec\n" -" parameter_declaration_prec .or .true .emit PRECISION_DEFAULT;\n" -"parameter_declaration_prec\n" -" precision .and space;\n" -"parameter_declaration_rest\n" -" parameter_declarator .or parameter_type_specifier;\n" -"parameter_qualifier\n" -" parameter_qualifier_1 .or .true .emit PARAM_QUALIFIER_IN;\n" -"parameter_qualifier_1\n" -" parameter_qualifier_2 .and space;\n" -"parameter_qualifier_2\n" -" \"in\" .emit PARAM_QUALIFIER_IN .or\n" -" \"out\" .emit PARAM_QUALIFIER_OUT .or\n" -" \"inout\" .emit PARAM_QUALIFIER_INOUT;\n" -"parameter_type_specifier\n" -" parameter_type_specifier_1 .and .true .emit '\\0' .and parameter_type_specifier_2;\n" -"parameter_type_specifier_1\n" -" type_specifier_nospace .or type_specifier_space;\n" -"parameter_type_specifier_2\n" -" parameter_type_specifier_3 .emit PARAMETER_ARRAY_PRESENT .or\n" -" .true .emit PARAMETER_ARRAY_NOT_PRESENT;\n" -"parameter_type_specifier_3\n" -" lbracket .and constant_expression .and rbracket;\n" -"init_declarator_list\n" -" single_declaration .and .loop init_declarator_list_1 .emit DECLARATOR_NEXT .and\n" -" .true .emit DECLARATOR_NONE;\n" -"init_declarator_list_1\n" -" comma .and identifier .emit VARIABLE_IDENTIFIER .and init_declarator_list_2;\n" -"init_declarator_list_2\n" -" init_declarator_list_3 .or init_declarator_list_4 .or .true .emit VARIABLE_NONE;\n" -"init_declarator_list_3\n" -" equals .and initializer .emit VARIABLE_INITIALIZER;\n" -"init_declarator_list_4\n" -" lbracket .and init_declarator_list_5 .and rbracket;\n" -"init_declarator_list_5\n" -" constant_expression .emit VARIABLE_ARRAY_EXPLICIT .or .true .emit VARIABLE_ARRAY_UNKNOWN;\n" -"single_declaration\n" -" single_declaration_nospace .or single_declaration_space;\n" -"single_declaration_space\n" -" fully_specified_type_space .and single_declaration_space_1;\n" -"single_declaration_nospace\n" -" fully_specified_type_nospace .and single_declaration_nospace_1;\n" -"single_declaration_space_1\n" -" single_declaration_space_2 .emit VARIABLE_IDENTIFIER .or .true .emit VARIABLE_NONE;\n" -"single_declaration_nospace_1\n" -" single_declaration_nospace_2 .emit VARIABLE_IDENTIFIER .or .true .emit VARIABLE_NONE;\n" -"single_declaration_space_2\n" -" space .and identifier .and single_declaration_3;\n" -"single_declaration_nospace_2\n" -" identifier .and single_declaration_3;\n" -"single_declaration_3\n" -" single_declaration_4 .or single_declaration_5 .or .true .emit VARIABLE_NONE;\n" -"single_declaration_4\n" -" equals .and initializer .emit VARIABLE_INITIALIZER;\n" -"single_declaration_5\n" -" lbracket .and single_declaration_6 .and rbracket;\n" -"single_declaration_6\n" -" constant_expression .emit VARIABLE_ARRAY_EXPLICIT .or .true .emit VARIABLE_ARRAY_UNKNOWN;\n" -"fully_specified_type_space\n" -" fully_specified_type_optinvariant .and fully_specified_type_optcentroid .and fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_space;\n" -"fully_specified_type_nospace\n" -" fully_specified_type_optinvariant .and fully_specified_type_optcentroid .and fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_nospace;\n" -"fully_specified_type_optinvariant\n" -" fully_specified_type_invariant .or .true .emit TYPE_VARIANT;\n" -"fully_specified_type_invariant\n" -" invariant_qualifier .and space;\n" -"fully_specified_type_optcentroid\n" -" fully_specified_type_centroid .or .true .emit TYPE_CENTER;\n" -"fully_specified_type_centroid\n" -" centroid_qualifier .and space;\n" -"fully_specified_type_optqual\n" -" fully_specified_type_qual .or .true .emit TYPE_QUALIFIER_NONE;\n" -"fully_specified_type_qual\n" -" type_qualifier .and space;\n" -"fully_specified_type_optprec\n" -" fully_specified_type_prec .or .true .emit PRECISION_DEFAULT;\n" -"fully_specified_type_prec\n" -" precision .and space;\n" -"invariant_qualifier\n" -" \"invariant\" .emit TYPE_INVARIANT;\n" -"centroid_qualifier\n" -" \"centroid\" .emit TYPE_CENTROID;\n" -"type_qualifier\n" -" \"const\" .emit TYPE_QUALIFIER_CONST .or\n" -" .if (shader_type == 2) \"attribute\" .emit TYPE_QUALIFIER_ATTRIBUTE .or\n" -" \"varying\" .emit TYPE_QUALIFIER_VARYING .or\n" -" \"uniform\" .emit TYPE_QUALIFIER_UNIFORM .or\n" -" .if (parsing_builtin != 0) \"__fixed_output\" .emit TYPE_QUALIFIER_FIXEDOUTPUT .or\n" -" .if (parsing_builtin != 0) \"__fixed_input\" .emit TYPE_QUALIFIER_FIXEDINPUT;\n" -"type_specifier_nonarray_space\n" -" \"void\" .emit TYPE_SPECIFIER_VOID .or\n" -" \"float\" .emit TYPE_SPECIFIER_FLOAT .or\n" -" \"int\" .emit TYPE_SPECIFIER_INT .or\n" -" \"bool\" .emit TYPE_SPECIFIER_BOOL .or\n" -" \"vec2\" .emit TYPE_SPECIFIER_VEC2 .or\n" -" \"vec3\" .emit TYPE_SPECIFIER_VEC3 .or\n" -" \"vec4\" .emit TYPE_SPECIFIER_VEC4 .or\n" -" \"bvec2\" .emit TYPE_SPECIFIER_BVEC2 .or\n" -" \"bvec3\" .emit TYPE_SPECIFIER_BVEC3 .or\n" -" \"bvec4\" .emit TYPE_SPECIFIER_BVEC4 .or\n" -" \"ivec2\" .emit TYPE_SPECIFIER_IVEC2 .or\n" -" \"ivec3\" .emit TYPE_SPECIFIER_IVEC3 .or\n" -" \"ivec4\" .emit TYPE_SPECIFIER_IVEC4 .or\n" -" \"mat2\" .emit TYPE_SPECIFIER_MAT2 .or\n" -" \"mat3\" .emit TYPE_SPECIFIER_MAT3 .or\n" -" \"mat4\" .emit TYPE_SPECIFIER_MAT4 .or\n" -" \"mat2x3\" .emit TYPE_SPECIFIER_MAT23 .or\n" -" \"mat3x2\" .emit TYPE_SPECIFIER_MAT32 .or\n" -" \"mat2x4\" .emit TYPE_SPECIFIER_MAT24 .or\n" -" \"mat4x2\" .emit TYPE_SPECIFIER_MAT42 .or\n" -" \"mat3x4\" .emit TYPE_SPECIFIER_MAT34 .or\n" -" \"mat4x3\" .emit TYPE_SPECIFIER_MAT43 .or\n" -" \"sampler1D\" .emit TYPE_SPECIFIER_SAMPLER1D .or\n" -" \"sampler2D\" .emit TYPE_SPECIFIER_SAMPLER2D .or\n" -" \"sampler3D\" .emit TYPE_SPECIFIER_SAMPLER3D .or\n" -" \"samplerCube\" .emit TYPE_SPECIFIER_SAMPLERCUBE .or\n" -" \"sampler1DShadow\" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or\n" -" \"sampler2DShadow\" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or\n" -" \"sampler2DRect\" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or\n" -" \"sampler2DRectShadow\" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW .or\n" -" type_name .emit TYPE_SPECIFIER_TYPENAME;\n" -"type_specifier_nonarray_nospace\n" -" struct_specifier .emit TYPE_SPECIFIER_STRUCT;\n" -"type_specifier_nonarray\n" -" type_specifier_nonarray_nospace .or type_specifier_nonarray_space;\n" -"type_specifier_space\n" -" type_specifier_nonarray_space .and .true .emit TYPE_SPECIFIER_NONARRAY;\n" -"type_specifier_nospace\n" -" type_specifier_nospace_array .or type_specifier_nospace_1;\n" -"type_specifier_nospace_1\n" -" type_specifier_nonarray_nospace .and .true .emit TYPE_SPECIFIER_NONARRAY;\n" -"type_specifier_nospace_array\n" -" type_specifier_nonarray .and lbracket .emit TYPE_SPECIFIER_ARRAY .and constant_expression .and rbracket;\n" -"struct_specifier\n" -" \"struct\" .and struct_specifier_1 .and optional_space .and lbrace .error LBRACE_EXPECTED .and\n" -" struct_declaration_list .and rbrace .emit FIELD_NONE;\n" -"struct_specifier_1\n" -" struct_specifier_2 .or .true .emit '\\0';\n" -"struct_specifier_2\n" -" space .and identifier;\n" -"struct_declaration_list\n" -" struct_declaration .and .loop struct_declaration .emit FIELD_NEXT;\n" -"struct_declaration\n" -" struct_declaration_nospace .or struct_declaration_space;\n" -"struct_declaration_space\n" -" type_specifier_space .and space .and struct_declarator_list .and semicolon .emit FIELD_NONE;\n" -"struct_declaration_nospace\n" -" type_specifier_nospace .and struct_declarator_list .and semicolon .emit FIELD_NONE;\n" -"struct_declarator_list\n" -" struct_declarator .and .loop struct_declarator_list_1 .emit FIELD_NEXT;\n" -"struct_declarator_list_1\n" -" comma .and struct_declarator;\n" -"struct_declarator\n" -" identifier .and struct_declarator_1;\n" -"struct_declarator_1\n" -" struct_declarator_2 .emit FIELD_ARRAY .or .true .emit FIELD_NONE;\n" -"struct_declarator_2\n" -" lbracket .and constant_expression .and rbracket;\n" -"initializer\n" -" assignment_expression .and .true .emit OP_END;\n" -"declaration_statement\n" -" declaration;\n" -"statement\n" -" compound_statement .or simple_statement;\n" -"statement_space\n" -" compound_statement .or statement_space_1;\n" -"statement_space_1\n" -" space .and simple_statement;\n" -"simple_statement\n" -" .if (parsing_builtin != 0) __asm_statement .emit OP_ASM .or\n" -" selection_statement .or\n" -" iteration_statement .or\n" -" precision_stmt .emit OP_PRECISION .or\n" -" jump_statement .or\n" -" expression_statement .emit OP_EXPRESSION .or\n" -" declaration_statement .emit OP_DECLARE;\n" -"compound_statement\n" -" compound_statement_1 .emit OP_BLOCK_BEGIN_NEW_SCOPE .and .true .emit OP_END;\n" -"compound_statement_1\n" -" compound_statement_2 .or compound_statement_3;\n" -"compound_statement_2\n" -" lbrace .and rbrace;\n" -"compound_statement_3\n" -" lbrace .and statement_list .and rbrace;\n" -"compound_statement_no_new_scope\n" -" compound_statement_no_new_scope_1 .emit OP_BLOCK_BEGIN_NO_NEW_SCOPE .and .true .emit OP_END;\n" -"compound_statement_no_new_scope_1\n" -" compound_statement_no_new_scope_2 .or compound_statement_no_new_scope_3;\n" -"compound_statement_no_new_scope_2\n" -" lbrace .and rbrace;\n" -"compound_statement_no_new_scope_3\n" -" lbrace .and statement_list .and rbrace;\n" -"statement_list\n" -" statement .and .loop statement;\n" -"expression_statement\n" -" expression_statement_1 .or expression_statement_2;\n" -"expression_statement_1\n" -" semicolon .emit OP_PUSH_VOID .emit OP_END;\n" -"expression_statement_2\n" -" expression .and semicolon .emit OP_END;\n" -"selection_statement\n" -" \"if\" .emit OP_IF .and lparen .error LPAREN_EXPECTED .and expression .and\n" -" rparen .error RPAREN_EXPECTED .emit OP_END .and selection_rest_statement;\n" -"selection_rest_statement\n" -" statement .and selection_rest_statement_1;\n" -"selection_rest_statement_1\n" -" selection_rest_statement_2 .or .true .emit OP_EXPRESSION .emit OP_PUSH_VOID .emit OP_END;\n" -"selection_rest_statement_2\n" -" \"else\" .and optional_space .and statement;\n" -"condition\n" -" condition_1 .emit OP_DECLARE .emit DECLARATION_INIT_DECLARATOR_LIST .or\n" -" condition_3 .emit OP_EXPRESSION;\n" -"condition_1\n" -" condition_1_nospace .or condition_1_space;\n" -"condition_1_nospace\n" -" fully_specified_type_nospace .and condition_2;\n" -"condition_1_space\n" -" fully_specified_type_space .and space .and condition_2;\n" -"condition_2\n" -" identifier .emit VARIABLE_IDENTIFIER .and equals .emit VARIABLE_INITIALIZER .and\n" -" initializer .and .true .emit DECLARATOR_NONE;\n" -"condition_3\n" -" expression .and .true .emit OP_END;\n" -"iteration_statement\n" -" iteration_statement_1 .or iteration_statement_2 .or iteration_statement_3;\n" -"iteration_statement_1\n" -" \"while\" .emit OP_WHILE .and lparen .error LPAREN_EXPECTED .and condition .and\n" -" rparen .error RPAREN_EXPECTED .and statement;\n" -"iteration_statement_2\n" -" \"do\" .emit OP_DO .and statement_space .and \"while\" .and lparen .error LPAREN_EXPECTED .and\n" -" expression .and rparen .error RPAREN_EXPECTED .emit OP_END .and semicolon;\n" -"iteration_statement_3\n" -" \"for\" .emit OP_FOR .and lparen .error LPAREN_EXPECTED .and for_init_statement .and\n" -" for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement;\n" -"for_init_statement\n" -" expression_statement .emit OP_EXPRESSION .or declaration_statement .emit OP_DECLARE;\n" -"conditionopt\n" -" condition .or\n" -" .true .emit OP_EXPRESSION .emit OP_PUSH_BOOL .emit 2 .emit '1' .emit '\\0' .emit OP_END;\n" -"for_rest_statement\n" -" conditionopt .and semicolon .and for_rest_statement_1;\n" -"for_rest_statement_1\n" -" for_rest_statement_2 .or .true .emit OP_PUSH_VOID .emit OP_END;\n" -"for_rest_statement_2\n" -" expression .and .true .emit OP_END;\n" -"jump_statement\n" -" jump_statement_1 .or jump_statement_2 .or jump_statement_3 .or jump_statement_4 .or\n" -" .if (shader_type == 1) jump_statement_5;\n" -"jump_statement_1\n" -" \"continue\" .and semicolon .emit OP_CONTINUE;\n" -"jump_statement_2\n" -" \"break\" .and semicolon .emit OP_BREAK;\n" -"jump_statement_3\n" -" \"return\" .emit OP_RETURN .and optional_space .and expression .and semicolon .emit OP_END;\n" -"jump_statement_4\n" -" \"return\" .emit OP_RETURN .and semicolon .emit OP_PUSH_VOID .emit OP_END;\n" -"jump_statement_5\n" -" \"discard\" .and semicolon .emit OP_DISCARD;\n" -"__asm_statement\n" -" \"__asm\" .and space .and identifier .and space .and asm_arguments .and semicolon .emit OP_END;\n" -"asm_arguments\n" -" asm_argument .and .true .emit OP_END .and .loop asm_arguments_1;\n" -"asm_arguments_1\n" -" comma .and asm_argument .and .true .emit OP_END;\n" -"asm_argument\n" -" var_with_field .or\n" -" variable_identifier .or\n" -" floatconstant;\n" -"var_with_field\n" -" variable_identifier .and dot .and field_selection .emit OP_FIELD;\n" -"translation_unit\n" -" optional_space .emit REVISION .and external_declaration .error INVALID_EXTERNAL_DECLARATION .and\n" -" .loop external_declaration .and optional_space .and\n" -" '\\0' .error INVALID_EXTERNAL_DECLARATION .emit EXTERNAL_NULL;\n" -"external_declaration\n" -" precision_stmt .emit DEFAULT_PRECISION .or\n" -" function_definition .emit EXTERNAL_FUNCTION_DEFINITION .or\n" -" invariant_stmt .emit INVARIANT_STMT .or\n" -" declaration .emit EXTERNAL_DECLARATION;\n" -"precision_stmt\n" -" \"precision\" .and space .and precision .error INVALID_PRECISION .and space .and prectype .error INVALID_PRECISION_TYPE .and semicolon;\n" -"precision\n" -" \"lowp\" .emit PRECISION_LOW .or\n" -" \"mediump\" .emit PRECISION_MEDIUM .or\n" -" \"highp\" .emit PRECISION_HIGH;\n" -"prectype\n" -" \"int\" .emit TYPE_SPECIFIER_INT .or\n" -" \"float\" .emit TYPE_SPECIFIER_FLOAT .or\n" -" \"sampler1D\" .emit TYPE_SPECIFIER_SAMPLER1D .or\n" -" \"sampler2D\" .emit TYPE_SPECIFIER_SAMPLER2D .or\n" -" \"sampler3D\" .emit TYPE_SPECIFIER_SAMPLER3D .or\n" -" \"samplerCube\" .emit TYPE_SPECIFIER_SAMPLERCUBE .or\n" -" \"sampler1DShadow\" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or\n" -" \"sampler2DShadow\" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or\n" -" \"sampler2DRect\" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or\n" -" \"sampler2DRectShadow\" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW;\n" -"invariant_stmt\n" -" \"invariant\" .and space .and identifier .and semicolon;\n" -"function_definition\n" -" function_prototype .and compound_statement_no_new_scope;\n" -"digit_oct\n" -" '0'-'7';\n" -"digit_dec\n" -" '0'-'9';\n" -"digit_hex\n" -" '0'-'9' .or 'A'-'F' .or 'a'-'f';\n" -"id_character_first\n" -" 'a'-'z' .or 'A'-'Z' .or '_';\n" -"id_character_next\n" -" id_character_first .or digit_dec;\n" -"identifier\n" -" id_character_first .emit * .and .loop id_character_next .emit * .and .true .emit '\\0';\n" -"float\n" -" float_1 .or float_2 .or float_3;\n" -"float_1\n" -" float_fractional_constant .and float_optional_exponent_part .and optional_f_suffix;\n" -"float_2\n" -" float_digit_sequence .and .true .emit '\\0' .and float_exponent_part .and optional_f_suffix;\n" -"float_3\n" -" float_digit_sequence .and .true .emit '\\0' .and 'f' .emit '\\0';\n" -"float_fractional_constant\n" -" float_fractional_constant_1 .or float_fractional_constant_2 .or float_fractional_constant_3;\n" -"float_fractional_constant_1\n" -" float_digit_sequence .and '.' .and float_digit_sequence;\n" -"float_fractional_constant_2\n" -" float_digit_sequence .and '.' .and .true .emit '\\0';\n" -"float_fractional_constant_3\n" -" '.' .emit '\\0' .and float_digit_sequence;\n" -"float_optional_exponent_part\n" -" float_exponent_part .or .true .emit '\\0';\n" -"float_digit_sequence\n" -" digit_dec .emit * .and .loop digit_dec .emit * .and .true .emit '\\0';\n" -"float_exponent_part\n" -" float_exponent_part_1 .or float_exponent_part_2;\n" -"float_exponent_part_1\n" -" 'e' .and float_optional_sign .and float_digit_sequence;\n" -"float_exponent_part_2\n" -" 'E' .and float_optional_sign .and float_digit_sequence;\n" -"float_optional_sign\n" -" float_sign .or .true;\n" -"float_sign\n" -" '+' .or '-' .emit '-';\n" -"optional_f_suffix\n" -" 'f' .or .true;\n" -"integer\n" -" integer_hex .or integer_oct .or integer_dec;\n" -"integer_hex\n" -" '0' .and integer_hex_1 .emit 0x10 .and digit_hex .emit * .and .loop digit_hex .emit * .and\n" -" .true .emit '\\0';\n" -"integer_hex_1\n" -" 'x' .or 'X';\n" -"integer_oct\n" -" '0' .emit 8 .emit * .and .loop digit_oct .emit * .and .true .emit '\\0';\n" -"integer_dec\n" -" digit_dec .emit 10 .emit * .and .loop digit_dec .emit * .and .true .emit '\\0';\n" -"boolean\n" -" \"true\" .emit 2 .emit '1' .emit '\\0' .or\n" -" \"false\" .emit 2 .emit '0' .emit '\\0';\n" -"type_name\n" -" identifier;\n" -"field_selection\n" -" identifier;\n" -"floatconstant\n" -" float .emit OP_PUSH_FLOAT;\n" -"intconstant\n" -" integer .emit OP_PUSH_INT;\n" -"boolconstant\n" -" boolean .emit OP_PUSH_BOOL;\n" -"optional_space\n" -" .loop single_space;\n" -"space\n" -" single_space .and .loop single_space;\n" -"single_space\n" -" white_char .or c_style_comment_block .or cpp_style_comment_block;\n" -"white_char\n" -" ' ' .or '\\t' .or new_line .or '\\v' .or '\\f';\n" -"new_line\n" -" cr_lf .or lf_cr .or '\\n' .or '\\r';\n" -"cr_lf\n" -" '\\r' .and '\\n';\n" -"lf_cr\n" -" '\\n' .and '\\r';\n" -"c_style_comment_block\n" -" '/' .and '*' .and c_style_comment_rest;\n" -"c_style_comment_rest\n" -" .loop c_style_comment_char_no_star .and c_style_comment_rest_1;\n" -"c_style_comment_rest_1\n" -" c_style_comment_end .or c_style_comment_rest_2;\n" -"c_style_comment_rest_2\n" -" '*' .and c_style_comment_rest;\n" -"c_style_comment_char_no_star\n" -" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n" -"c_style_comment_end\n" -" '*' .and '/';\n" -"cpp_style_comment_block\n" -" '/' .and '/' .and cpp_style_comment_block_1;\n" -"cpp_style_comment_block_1\n" -" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n" -"cpp_style_comment_block_2\n" -" .loop cpp_style_comment_char .and new_line;\n" -"cpp_style_comment_block_3\n" -" .loop cpp_style_comment_char;\n" -"cpp_style_comment_char\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"ampersandampersand\n" -" optional_space .and '&' .and '&' .and optional_space;\n" -"barbar\n" -" optional_space .and '|' .and '|' .and optional_space;\n" -"bang\n" -" optional_space .and '!' .and optional_space;\n" -"bangequals\n" -" optional_space .and '!' .and '=' .and optional_space;\n" -"caretcaret\n" -" optional_space .and '^' .and '^' .and optional_space;\n" -"colon\n" -" optional_space .and ':' .and optional_space;\n" -"comma\n" -" optional_space .and ',' .and optional_space;\n" -"dot\n" -" optional_space .and '.' .and optional_space;\n" -"equals\n" -" optional_space .and '=' .and optional_space;\n" -"equalsequals\n" -" optional_space .and '=' .and '=' .and optional_space;\n" -"greater\n" -" optional_space .and '>' .and optional_space;\n" -"greaterequals\n" -" optional_space .and '>' .and '=' .and optional_space;\n" -"lbrace\n" -" optional_space .and '{' .and optional_space;\n" -"lbracket\n" -" optional_space .and '[' .and optional_space;\n" -"less\n" -" optional_space .and '<' .and optional_space;\n" -"lessequals\n" -" optional_space .and '<' .and '=' .and optional_space;\n" -"lparen\n" -" optional_space .and '(' .and optional_space;\n" -"minus\n" -" optional_space .and '-' .and optional_space;\n" -"minusequals\n" -" optional_space .and '-' .and '=' .and optional_space;\n" -"minusminus\n" -" optional_space .and '-' .and '-' .and optional_space;\n" -"plus\n" -" optional_space .and '+' .and optional_space;\n" -"plusequals\n" -" optional_space .and '+' .and '=' .and optional_space;\n" -"plusplus\n" -" optional_space .and '+' .and '+' .and optional_space;\n" -"question\n" -" optional_space .and '?' .and optional_space;\n" -"rbrace\n" -" optional_space .and '}' .and optional_space;\n" -"rbracket\n" -" optional_space .and ']' .and optional_space;\n" -"rparen\n" -" optional_space .and ')' .and optional_space;\n" -"semicolon\n" -" optional_space .and ';' .and optional_space;\n" -"slash\n" -" optional_space .and '/' .and optional_space;\n" -"slashequals\n" -" optional_space .and '/' .and '=' .and optional_space;\n" -"star\n" -" optional_space .and '*' .and optional_space;\n" -"starequals\n" -" optional_space .and '*' .and '=' .and optional_space;\n" -".string string_lexer;\n" -"string_lexer\n" -" lex_first_identifier_character .and .loop lex_next_identifier_character;\n" -"lex_first_identifier_character\n" -" 'a'-'z' .or 'A'-'Z' .or '_';\n" -"lex_next_identifier_character\n" -" 'a'-'z' .or 'A'-'Z' .or '0'-'9' .or '_';\n" -"err_token\n" -" '~' .or '`' .or '!' .or '@' .or '#' .or '$' .or '%' .or '^' .or '&' .or '*' .or '(' .or ')' .or\n" -" '-' .or '+' .or '=' .or '|' .or '\\\\' .or '[' .or ']' .or '{' .or '}' .or ':' .or ';' .or '\"' .or\n" -" '\\'' .or '<' .or ',' .or '>' .or '.' .or '/' .or '?' .or err_identifier;\n" -"err_identifier\n" -" id_character_first .and .loop id_character_next;\n" -"" diff --git a/src/mesa/shader/slang/library/slang_version.syn b/src/mesa/shader/slang/library/slang_version.syn deleted file mode 100644 index aaf8bef342f..00000000000 --- a/src/mesa/shader/slang/library/slang_version.syn +++ /dev/null @@ -1,118 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.3 - * - * Copyright (C) 2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file slang_version.syn - * slang #version directive syntax - * \author Michal Krol - */ - -.syntax version_directive; - -version_directive - version_directive_1 .and .loop version_directive_2; -version_directive_1 - prior_optional_spaces .and optional_version_directive .and .true .emit $; -version_directive_2 - prior_optional_spaces .and version_directive_body .and .true .emit $; - -optional_version_directive - version_directive_body .or .true .emit 10 .emit 1; - -version_directive_body - '#' .and optional_space .and "version" .and space .and version_number .and optional_space .and - new_line; - -version_number - version_number_110; - -version_number_110 - leading_zeroes .and "110" .emit 10 .emit 1; - -leading_zeroes - .loop zero; - -zero - '0'; - -space - single_space .and .loop single_space; - -optional_space - .loop single_space; - -single_space - ' ' .or '\t'; - -prior_optional_spaces - .loop prior_space; - -prior_space - c_style_comment_block .or cpp_style_comment_block .or space .or new_line; - -c_style_comment_block - '/' .and '*' .and c_style_comment_rest; - -c_style_comment_rest - .loop c_style_comment_char_no_star .and c_style_comment_rest_1; -c_style_comment_rest_1 - c_style_comment_end .or c_style_comment_rest_2; -c_style_comment_rest_2 - '*' .and c_style_comment_rest; - -c_style_comment_char_no_star - '\x2B'-'\xFF' .or '\x01'-'\x29'; - -c_style_comment_end - '*' .and '/'; - -cpp_style_comment_block - '/' .and '/' .and cpp_style_comment_block_1; -cpp_style_comment_block_1 - cpp_style_comment_block_2 .or cpp_style_comment_block_3; -cpp_style_comment_block_2 - .loop cpp_style_comment_char .and new_line; -cpp_style_comment_block_3 - .loop cpp_style_comment_char; - -cpp_style_comment_char - '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; - -new_line - cr_lf .or lf_cr .or '\n' .or '\r'; - -cr_lf - '\r' .and '\n'; - -lf_cr - '\n' .and '\r'; - -.string __string_filter; - -__string_filter - .loop __identifier_char; - -__identifier_char - 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9'; - diff --git a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h deleted file mode 100644 index e5a252b0196..00000000000 --- a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h +++ /dev/null @@ -1,109 +0,0 @@ - -/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ -/* slang_vertex_builtin.gc */ - -5,2,2,90,95,5,0,12,0,1,103,108,95,80,111,115,105,116,105,111,110,0,0,0,2,2,90,95,5,0,9,0,1,103,108, -95,80,111,105,110,116,83,105,122,101,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,67,108,105,112,86,101, -114,116,101,120,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95,2,0,12,0, -1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,2,0,11,0,1,103, -108,95,78,111,114,109,97,108,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,86,101,114,116,101,120,0,0,0,2, -2,90,95,2,0,12,0,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,48,0,0,0,2,2,90,95, -2,0,12,0,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,49,0,0,0,2,2,90,95,2,0,12,0, -1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,50,0,0,0,2,2,90,95,2,0,12,0,1,103, -108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,51,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95, -77,117,108,116,105,84,101,120,67,111,111,114,100,52,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117, -108,116,105,84,101,120,67,111,111,114,100,53,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117,108,116, -105,84,101,120,67,111,111,114,100,54,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117,108,116,105,84, -101,120,67,111,111,114,100,55,0,0,0,2,2,90,95,2,0,9,0,1,103,108,95,70,111,103,67,111,111,114,100,0, -0,0,2,2,90,95,3,0,12,0,1,103,108,95,70,114,111,110,116,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0, -1,103,108,95,66,97,99,107,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,70,114,111,110, -116,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,66, -97,99,107,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108, -95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111, -114,100,115,0,0,0,2,2,90,95,3,0,9,0,1,103,108,95,70,111,103,70,114,97,103,67,111,111,114,100,0,0,0, -1,90,95,0,0,12,0,0,102,116,114,97,110,115,102,111,114,109,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116, -114,105,120,0,16,8,48,0,57,18,103,108,95,86,101,114,116,101,120,0,59,120,120,120,120,0,48,18,103, -108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105, -120,0,16,10,49,0,57,18,103,108,95,86,101,114,116,101,120,0,59,121,121,121,121,0,48,46,18,103,108, -95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0, -16,10,50,0,57,18,103,108,95,86,101,114,116,101,120,0,59,122,122,122,122,0,48,46,18,103,108,95,77, -111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,10, -51,0,57,18,103,108,95,86,101,114,116,101,120,0,59,119,119,119,119,0,48,46,20,0,0,1,90,95,0,0,12,0, -0,116,101,120,116,117,114,101,49,68,76,111,100,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,1,1,0, -0,9,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114, -100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111, -114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105,97, -115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100, -52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0, -0,16,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100, -0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0, -18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114, -100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105,97,115,0, -18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0, -0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,0,16,0, -115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1, -3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111, -111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59, -119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105,97,115,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90, -95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101, -114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1, -99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59, -120,121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, -101,120,95,50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80, -114,111,106,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114, -100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112, -99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59, -122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, -101,120,95,50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80, -114,111,106,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114, -100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112, -99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59, -122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, -101,120,95,50,100,95,98,105,97,115,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,76, -111,100,0,1,1,0,0,18,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0, -9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100, -52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0, -59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,51,100,95,98,105,97,115,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1, -90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,0,18,0,115,97, -109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90, -95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99, -111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111, -114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,51,100,95,98,105,97,115, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0, -0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,67,117,98,101,76,111,100,0,1,1,0,0,19,0,115, -97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2, -90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99, -111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52, -95,116,101,120,95,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114, -0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,76,111,100,0, -1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108, -111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59, -120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0, -20,0,4,118,101,99,52,95,116,101,120,95,49,100,95,98,105,97,115,95,115,104,97,100,111,119,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1, -90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,80,114,111,106,76,111,100,0,1,1,0,0,20,0,115,97,109, -112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95, -0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114, -100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18, -99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4, -118,101,99,52,95,116,101,120,95,49,100,95,98,105,97,115,95,115,104,97,100,111,119,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95, -0,0,12,0,0,115,104,97,100,111,119,50,68,76,111,100,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1, -1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111, -114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18, -99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,50,100,95, -98,105,97,115,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,50,68,80, -114,111,106,76,111,100,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114, -100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112, -99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59, -119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18, -112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,95,50,100, -95,98,105,97,115,95,115,104,97,100,111,119,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0 diff --git a/src/mesa/shader/slang/library/syn_to_c.c b/src/mesa/shader/slang/library/syn_to_c.c deleted file mode 100644 index f997edfd8b5..00000000000 --- a/src/mesa/shader/slang/library/syn_to_c.c +++ /dev/null @@ -1,72 +0,0 @@ -#include <stdio.h> - -static int was_space = 0; -static int first_char = 1; - -static void put_char (int c) -{ - if (c == '\n') { - if (!first_char) { - fputs ("\\n\"\n\"", stdout); - first_char = 1; - } - } - else { - first_char = 0; - if (c == '\\') - fputs ("\\\\", stdout); - else if (c == '\"') - fputs ("\\\"", stdout); - else if (!was_space || !(c == ' ' || c == '\t')) - fputc (c, stdout); - was_space = (c == ' ' || c == '\t'); - } -} - -int main (int argc, char *argv[]) -{ - int c; - FILE *f; - - if (argc == 1) - return 1; - f = fopen (argv[1], "r"); - if (f == NULL) - return 1; - - fputs ("\n", stdout); - fputs ("/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */\n", stdout); - fputs ("\n", stdout); - fputs ("\"", stdout); - c = getc (f); - while (c != EOF) { - if (c == '/') { - int c2 = getc (f); - if (c2 == '*') { - was_space = 0; - c = getc (f); - for (;;) { - if (c == '*') { - c2 = getc (f); - if (c2 == '/') - break; - } - c = getc (f); - } - } - else { - put_char (c); - put_char (c2); - } - } - else { - put_char (c); - } - c = getc (f); - } - fputs ("\"\n", stdout); - - fclose (f); - return 0; -} - diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index a2708884437..f60ad9ee385 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -35,10 +35,10 @@ #include "shader/prog_optimize.h" #include "shader/prog_print.h" #include "shader/prog_parameter.h" -#include "shader/grammar/grammar_mesa.h" +#include "../../glsl/pp/sl_pp_public.h" +#include "../../glsl/cl/sl_cl_parse.h" #include "slang_codegen.h" #include "slang_compile.h" -#include "slang_preprocess.h" #include "slang_storage.h" #include "slang_emit.h" #include "slang_log.h" @@ -128,7 +128,7 @@ _slang_code_object_dtr(slang_code_object * self) typedef struct slang_parse_ctx_ { - const byte *I; + const unsigned char *I; slang_info_log *L; int parsing_builtin; GLboolean global_scope; /**< Is object being declared a global? */ @@ -184,22 +184,102 @@ parse_identifier(slang_parse_ctx * C) } static int +is_hex_digit(char c) +{ + return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F'); +} + +static int +parse_general_number(slang_parse_ctx *ctx, float *number) +{ + char *flt = NULL; + + if (*ctx->I == '0') { + int value = 0; + const unsigned char *pi; + + if (ctx->I[1] == 'x' || ctx->I[1] == 'X') { + ctx->I += 2; + if (!is_hex_digit(*ctx->I)) { + return 0; + } + do { + int digit; + + if (*ctx->I >= '0' && *ctx->I <= '9') { + digit = (int)(*ctx->I - '0'); + } else if (*ctx->I >= 'a' && *ctx->I <= 'f') { + digit = (int)(*ctx->I - 'a') + 10; + } else { + digit = (int)(*ctx->I - 'A') + 10; + } + value = value * 0x10 + digit; + ctx->I++; + } while (is_hex_digit(*ctx->I)); + if (*ctx->I != '\0') { + return 0; + } + ctx->I++; + *number = (float)value; + return 1; + } + + pi = ctx->I; + pi++; + while (*pi >= '0' && *pi <= '7') { + int digit; + + digit = (int)(*pi - '0'); + value = value * 010 + digit; + pi++; + } + if (*pi == '\0') { + pi++; + ctx->I = pi; + *number = (float)value; + return 1; + } + } + + parse_identifier_str(ctx, &flt); + flt = strdup(flt); + if (!flt) { + return 0; + } + if (flt[strlen(flt) - 1] == 'f' || flt[strlen(flt) - 1] == 'F') { + flt[strlen(flt) - 1] = '\0'; + } + *number = (float)_mesa_strtod(flt, (char **)NULL); + free(flt); + + return 1; +} + +static int parse_number(slang_parse_ctx * C, int *number) { const int radix = (int) (*C->I++); - *number = 0; - while (*C->I != '\0') { - int digit; - if (*C->I >= '0' && *C->I <= '9') - digit = (int) (*C->I - '0'); - else if (*C->I >= 'A' && *C->I <= 'Z') - digit = (int) (*C->I - 'A') + 10; - else - digit = (int) (*C->I - 'a') + 10; - *number = *number * radix + digit; + + if (radix == 1) { + float f = 0.0f; + + parse_general_number(C, &f); + *number = (int)f; + } else { + *number = 0; + while (*C->I != '\0') { + int digit; + if (*C->I >= '0' && *C->I <= '9') + digit = (int) (*C->I - '0'); + else if (*C->I >= 'A' && *C->I <= 'Z') + digit = (int) (*C->I - 'A') + 10; + else + digit = (int) (*C->I - 'a') + 10; + *number = *number * radix + digit; + C->I++; + } C->I++; } - C->I++; if (*number > 65535) slang_info_log_warning(C->L, "%d: literal integer overflow.", *number); return 1; @@ -208,32 +288,37 @@ parse_number(slang_parse_ctx * C, int *number) static int parse_float(slang_parse_ctx * C, float *number) { - char *integral = NULL; - char *fractional = NULL; - char *exponent = NULL; - char *whole = NULL; - - parse_identifier_str(C, &integral); - parse_identifier_str(C, &fractional); - parse_identifier_str(C, &exponent); - - whole = (char *) _slang_alloc((_mesa_strlen(integral) + - _mesa_strlen(fractional) + - _mesa_strlen(exponent) + 3) * sizeof(char)); - if (whole == NULL) { - slang_info_log_memory(C->L); - RETURN0; - } + if (*C->I == 1) { + C->I++; + parse_general_number(C, number); + } else { + char *integral = NULL; + char *fractional = NULL; + char *exponent = NULL; + char *whole = NULL; + + parse_identifier_str(C, &integral); + parse_identifier_str(C, &fractional); + parse_identifier_str(C, &exponent); + + whole = (char *) _slang_alloc((_mesa_strlen(integral) + + _mesa_strlen(fractional) + + _mesa_strlen(exponent) + 3) * sizeof(char)); + if (whole == NULL) { + slang_info_log_memory(C->L); + RETURN0; + } - slang_string_copy(whole, integral); - slang_string_concat(whole, "."); - slang_string_concat(whole, fractional); - slang_string_concat(whole, "E"); - slang_string_concat(whole, exponent); + slang_string_copy(whole, integral); + slang_string_concat(whole, "."); + slang_string_concat(whole, fractional); + slang_string_concat(whole, "E"); + slang_string_concat(whole, exponent); - *number = (float) (_mesa_strtod(whole, (char **) NULL)); + *number = (float) (_mesa_strtod(whole, (char **) NULL)); - _slang_free(whole); + _slang_free(whole); + } return 1; } @@ -2455,7 +2540,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, } static GLboolean -compile_binary(const byte * prod, slang_code_unit * unit, +compile_binary(const unsigned char * prod, slang_code_unit * unit, GLuint version, slang_unit_type type, slang_info_log * infolog, slang_code_unit * builtin, slang_code_unit * downlink, @@ -2488,17 +2573,56 @@ compile_binary(const byte * prod, slang_code_unit * unit, } static GLboolean -compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, - slang_unit_type type, slang_info_log * infolog, - slang_code_unit * builtin, +compile_with_grammar(const char *source, + slang_code_unit *unit, + slang_unit_type type, + slang_info_log *infolog, + slang_code_unit *builtin, struct gl_shader *shader, - const struct gl_extensions *extensions, - struct gl_sl_pragmas *pragmas) + struct gl_sl_pragmas *pragmas, + unsigned int shader_type, + unsigned int parsing_builtin) { - byte *prod; - GLuint size, start, version; - slang_string preprocessed; - GLuint maxVersion; + struct sl_pp_purify_options options; + struct sl_pp_context *context; + unsigned char *prod; + GLuint size; + unsigned int version; + unsigned int maxVersion; + int result; + char errmsg[200] = ""; + + assert(shader_type == 1 || shader_type == 2); + + memset(&options, 0, sizeof(options)); + + context = sl_pp_context_create(source, &options); + if (!context) { + slang_info_log_error(infolog, "out of memory"); + return GL_FALSE; + } + + if (sl_pp_version(context, &version)) { + slang_info_log_error(infolog, "%s", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + return GL_FALSE; + } + + if (sl_pp_context_add_extension(context, "ARB_draw_buffers", "GL_ARB_draw_buffers") || + sl_pp_context_add_extension(context, "ARB_texture_rectangle", "GL_ARB_texture_rectangle")) { + slang_info_log_error(infolog, "%s", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + return GL_FALSE; + } + +#if FEATURE_es2_glsl + if (sl_pp_context_add_predefined(context, "GL_ES", "1") || + sl_pp_context_add_predefined(context, "GL_FRAGMENT_PRECISION_HIGH", "1")) { + slang_info_log_error(infolog, "%s", sl_pp_context_error_message(context)); + sl_pp_context_destroy(context); + return GL_FALSE; + } +#endif #if FEATURE_ARB_shading_language_120 maxVersion = 120; @@ -2508,36 +2632,30 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, maxVersion = 110; #endif - /* First retrieve the version number. */ - if (!_slang_preprocess_version(source, &version, &start, infolog)) - return GL_FALSE; - - if (version > maxVersion) { + if (version > maxVersion || + (version != 100 && version != 110 && version != 120)) { slang_info_log_error(infolog, "language version %.2f is not supported.", version * 0.01); - return GL_FALSE; - } - - /* Now preprocess the source string. */ - slang_string_init(&preprocessed); - if (!_slang_preprocess_directives(&preprocessed, &source[start], - infolog, extensions, pragmas)) { - slang_string_free(&preprocessed); - slang_info_log_error(infolog, "failed to preprocess the source."); + sl_pp_context_destroy(context); return GL_FALSE; } /* Finally check the syntax and generate its binary representation. */ - if (!grammar_fast_check(id, - (const byte *) (slang_string_cstr(&preprocessed)), - &prod, &size, 65536)) { - char buf[1024]; - GLint pos; - - slang_string_free(&preprocessed); - grammar_get_last_error((byte *) (buf), sizeof(buf), &pos); - slang_info_log_error(infolog, buf); + result = sl_cl_compile(context, + shader_type, + parsing_builtin, + &prod, + &size, + errmsg, + sizeof(errmsg)); + + sl_pp_context_destroy(context); + + if (result) { + /*GLint pos;*/ + + slang_info_log_error(infolog, errmsg); /* syntax error (possibly in library code) */ #if 0 { @@ -2551,77 +2669,64 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, #endif return GL_FALSE; } - slang_string_free(&preprocessed); /* Syntax is okay - translate it to internal representation. */ if (!compile_binary(prod, unit, version, type, infolog, builtin, &builtin[SLANG_BUILTIN_TOTAL - 1], shader)) { - grammar_alloc_free(prod); + free(prod); return GL_FALSE; } - grammar_alloc_free(prod); + free(prod); return GL_TRUE; } -LONGSTRING static const char *slang_shader_syn = -#include "library/slang_shader_syn.h" - ; - -static const byte slang_core_gc[] = { +static const unsigned char slang_core_gc[] = { #include "library/slang_core_gc.h" }; -static const byte slang_120_core_gc[] = { +static const unsigned char slang_120_core_gc[] = { #include "library/slang_120_core_gc.h" }; -static const byte slang_120_fragment_gc[] = { +static const unsigned char slang_120_fragment_gc[] = { #include "library/slang_builtin_120_fragment_gc.h" }; -static const byte slang_common_builtin_gc[] = { +static const unsigned char slang_common_builtin_gc[] = { #include "library/slang_common_builtin_gc.h" }; -static const byte slang_fragment_builtin_gc[] = { +static const unsigned char slang_fragment_builtin_gc[] = { #include "library/slang_fragment_builtin_gc.h" }; -static const byte slang_vertex_builtin_gc[] = { +static const unsigned char slang_vertex_builtin_gc[] = { #include "library/slang_vertex_builtin_gc.h" }; static GLboolean -compile_object(grammar * id, const char *source, slang_code_object * object, - slang_unit_type type, slang_info_log * infolog, +compile_object(const char *source, + slang_code_object *object, + slang_unit_type type, + slang_info_log *infolog, struct gl_shader *shader, - const struct gl_extensions *extensions, struct gl_sl_pragmas *pragmas) { slang_code_unit *builtins = NULL; GLuint base_version = 110; - - /* load GLSL grammar */ - *id = grammar_load_from_text((const byte *) (slang_shader_syn)); - if (*id == 0) { - byte buf[1024]; - int pos; - - grammar_get_last_error(buf, 1024, &pos); - slang_info_log_error(infolog, (const char *) (buf)); - return GL_FALSE; - } + unsigned int shader_type; + unsigned int parsing_builtin; /* set shader type - the syntax is slightly different for different shaders */ - if (type == SLANG_UNIT_FRAGMENT_SHADER - || type == SLANG_UNIT_FRAGMENT_BUILTIN) - grammar_set_reg8(*id, (const byte *) "shader_type", 1); - else - grammar_set_reg8(*id, (const byte *) "shader_type", 2); + if (type == SLANG_UNIT_FRAGMENT_SHADER || type == SLANG_UNIT_FRAGMENT_BUILTIN) { + shader_type = 1; + } else { + shader_type = 2; + } /* enable language extensions */ - grammar_set_reg8(*id, (const byte *) "parsing_builtin", 1); + parsing_builtin = 1; /* if parsing user-specified shader, load built-in library */ if (type == SLANG_UNIT_FRAGMENT_SHADER || type == SLANG_UNIT_VERTEX_SHADER) { @@ -2686,51 +2791,24 @@ compile_object(grammar * id, const char *source, slang_code_object * object, } /* disable language extensions */ -#if NEW_SLANG /* allow-built-ins */ - grammar_set_reg8(*id, (const byte *) "parsing_builtin", 1); -#else - grammar_set_reg8(*id, (const byte *) "parsing_builtin", 0); -#endif + parsing_builtin = 0; + builtins = object->builtin; } /* compile the actual shader - pass-in built-in library for external shader */ - return compile_with_grammar(*id, source, &object->unit, type, infolog, - builtins, shader, extensions, pragmas); -} - - -static GLboolean -compile_shader(GLcontext *ctx, slang_code_object * object, - slang_unit_type type, slang_info_log * infolog, - struct gl_shader *shader) -{ - GLboolean success; - grammar id = 0; - -#if 0 /* for debug */ - _mesa_printf("********* COMPILE SHADER ***********\n"); - _mesa_printf("%s\n", shader->Source); - _mesa_printf("************************************\n"); -#endif - - assert(shader->Program); - - _slang_code_object_dtr(object); - _slang_code_object_ctr(object); - - success = compile_object(&id, shader->Source, object, type, infolog, shader, - &ctx->Extensions, &shader->Pragmas); - if (id != 0) - grammar_destroy(id); - if (!success) - return GL_FALSE; - - return GL_TRUE; + return compile_with_grammar(source, + &object->unit, + type, + infolog, + builtins, + shader, + pragmas, + shader_type, + parsing_builtin); } - GLboolean _slang_compile(GLcontext *ctx, struct gl_shader *shader) { @@ -2738,6 +2816,7 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) slang_info_log info_log; slang_code_object obj; slang_unit_type type; + GLenum progTarget; if (shader->Type == GL_VERTEX_SHADER) { type = SLANG_UNIT_VERTEX_SHADER; @@ -2754,22 +2833,28 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) shader->Main = GL_FALSE; - if (!shader->Program) { - GLenum progTarget; - if (shader->Type == GL_VERTEX_SHADER) - progTarget = GL_VERTEX_PROGRAM_ARB; - else - progTarget = GL_FRAGMENT_PROGRAM_ARB; - shader->Program = ctx->Driver.NewProgram(ctx, progTarget, 1); - shader->Program->Parameters = _mesa_new_parameter_list(); - shader->Program->Varying = _mesa_new_parameter_list(); - shader->Program->Attributes = _mesa_new_parameter_list(); - } + /* free the shader's old instructions, etc */ + _mesa_reference_program(ctx, &shader->Program, NULL); + + /* allocate new GPU program, parameter lists, etc. */ + if (shader->Type == GL_VERTEX_SHADER) + progTarget = GL_VERTEX_PROGRAM_ARB; + else + progTarget = GL_FRAGMENT_PROGRAM_ARB; + shader->Program = ctx->Driver.NewProgram(ctx, progTarget, 1); + shader->Program->Parameters = _mesa_new_parameter_list(); + shader->Program->Varying = _mesa_new_parameter_list(); + shader->Program->Attributes = _mesa_new_parameter_list(); slang_info_log_construct(&info_log); _slang_code_object_ctr(&obj); - success = compile_shader(ctx, &obj, type, &info_log, shader); + success = compile_object(shader->Source, + &obj, + type, + &info_log, + shader, + &shader->Pragmas); /* free shader's prev info log */ if (shader->InfoLog) { diff --git a/src/mesa/shader/slang/slang_compile_operation.c b/src/mesa/shader/slang/slang_compile_operation.c index 3e2bdbc91ff..be73094ca04 100644 --- a/src/mesa/shader/slang/slang_compile_operation.c +++ b/src/mesa/shader/slang/slang_compile_operation.c @@ -52,6 +52,7 @@ slang_operation_construct(slang_operation * oper) _slang_variable_scope_ctr(oper->locals); oper->fun = NULL; oper->var = NULL; + oper->label = NULL; return GL_TRUE; } diff --git a/src/mesa/shader/slang/slang_compile_operation.h b/src/mesa/shader/slang/slang_compile_operation.h index 58f1edeed85..1f15c198963 100644 --- a/src/mesa/shader/slang/slang_compile_operation.h +++ b/src/mesa/shader/slang/slang_compile_operation.h @@ -127,7 +127,6 @@ typedef struct slang_operation_ * indicate such. num_children indicates number of elements. */ GLboolean array_constructor; - double x; } slang_operation; diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c deleted file mode 100644 index e9a24cc009a..00000000000 --- a/src/mesa/shader/slang/slang_preprocess.c +++ /dev/null @@ -1,1406 +0,0 @@ -/* - * Mesa 3-D graphics library - * - * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. - * Copyright (C) 2009 VMware, Inc. All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file slang_preprocess.c - * slang preprocessor - * \author Michal Krol - */ - -#include "main/imports.h" -#include "shader/grammar/grammar_mesa.h" -#include "slang_preprocess.h" - -LONGSTRING static const char *slang_pp_directives_syn = -#include "library/slang_pp_directives_syn.h" -; - -LONGSTRING static const char *slang_pp_expression_syn = -#include "library/slang_pp_expression_syn.h" -; - -LONGSTRING static const char *slang_pp_version_syn = -#include "library/slang_pp_version_syn.h" -; - -static GLvoid -grammar_error_to_log (slang_info_log *log) -{ - char buf[1024]; - GLint pos; - - grammar_get_last_error ((byte *) (buf), sizeof (buf), &pos); - if (buf[0] == 0) { - _mesa_snprintf(buf, sizeof(buf), "Preprocessor error"); - } - slang_info_log_error (log, buf); -} - -GLboolean -_slang_preprocess_version (const char *text, GLuint *version, GLuint *eaten, slang_info_log *log) -{ - grammar id; - byte *prod, *I; - unsigned int size; - - id = grammar_load_from_text ((const byte *) (slang_pp_version_syn)); - if (id == 0) { - grammar_error_to_log (log); - return GL_FALSE; - } - - if (!grammar_fast_check (id, (const byte *) (text), &prod, &size, 8)) { - grammar_error_to_log (log); - grammar_destroy (id); - return GL_FALSE; - } - - /* there can be multiple #version directives - grab the last one */ - I = &prod[size - 6]; - *version = (GLuint) (I[0]) + (GLuint) (I[1]) * 100; - *eaten = (GLuint) (I[2]) + ((GLuint) (I[3]) << 8) + ((GLuint) (I[4]) << 16) + ((GLuint) (I[5]) << 24); - - grammar_destroy (id); - grammar_alloc_free (prod); - return GL_TRUE; -} - -/* - * The preprocessor does the following work. - * 1. Remove comments. Each comment block is replaced with a single space and if the - * block contains new-lines, they are preserved. This ensures that line numbers - * stay the same and if a comment block delimits two tokens, the are delitmited - * by the space after comment removal. - * 2. Remove preprocessor directives from the source string, checking their syntax and - * executing them if appropriate. Again, new-lines are preserved. - * 3. Expand macros. - * 4. Tokenize the source string by ensuring there is at least one space between every - * two adjacent tokens. - */ - -#define PP_ANNOTATE 0 - -static GLvoid -pp_annotate (slang_string *output, const char *fmt, ...) -{ -#if PP_ANNOTATE - va_list va; - char buffer[1024]; - - va_start (va, fmt); - _mesa_vsprintf (buffer, fmt, va); - va_end (va); - slang_string_pushs (output, buffer, _mesa_strlen (buffer)); -#else - (GLvoid) (output); - (GLvoid) (fmt); -#endif -} - - /* - * The expression is executed on a fixed-sized stack. The PUSH macro makes a runtime - * check if the stack is not overflown by too complex expressions. In that situation the - * GLSL preprocessor should report internal compiler error. - * The BINARYDIV makes a runtime check if the divider is not 0. If it is, it reports - * compilation error. - */ - -#define EXECUTION_STACK_SIZE 1024 - -#define PUSH(x)\ - do {\ - if (sp == 0) {\ - slang_info_log_error (elog, "internal compiler error: preprocessor execution stack overflow.");\ - return GL_FALSE;\ - }\ - stack[--sp] = x;\ - } while (GL_FALSE) - -#define POP(x)\ - do {\ - assert (sp < EXECUTION_STACK_SIZE);\ - x = stack[sp++];\ - } while (GL_FALSE) - -#define BINARY(op)\ - do {\ - GLint a, b;\ - POP(b);\ - POP(a);\ - PUSH(a op b);\ - } while (GL_FALSE) - -#define BINARYDIV(op)\ - do {\ - GLint a, b;\ - POP(b);\ - POP(a);\ - if (b == 0) {\ - slang_info_log_error (elog, "division by zero in preprocessor expression.");\ - return GL_FALSE;\ - }\ - PUSH(a op b);\ - } while (GL_FALSE) - -#define UNARY(op)\ - do {\ - GLint a;\ - POP(a);\ - PUSH(op a);\ - } while (GL_FALSE) - -#define OP_END 0 -#define OP_PUSHINT 1 -#define OP_LOGICALOR 2 -#define OP_LOGICALAND 3 -#define OP_OR 4 -#define OP_XOR 5 -#define OP_AND 6 -#define OP_EQUAL 7 -#define OP_NOTEQUAL 8 -#define OP_LESSEQUAL 9 -#define OP_GREATEREQUAL 10 -#define OP_LESS 11 -#define OP_GREATER 12 -#define OP_LEFTSHIFT 13 -#define OP_RIGHTSHIFT 14 -#define OP_ADD 15 -#define OP_SUBTRACT 16 -#define OP_MULTIPLY 17 -#define OP_DIVIDE 18 -#define OP_MODULUS 19 -#define OP_PLUS 20 -#define OP_MINUS 21 -#define OP_NEGATE 22 -#define OP_COMPLEMENT 23 - -static GLboolean -execute_expression (slang_string *output, const byte *code, GLuint *pi, GLint *result, - slang_info_log *elog) -{ - GLuint i = *pi; - GLint stack[EXECUTION_STACK_SIZE]; - GLuint sp = EXECUTION_STACK_SIZE; - - while (code[i] != OP_END) { - switch (code[i++]) { - case OP_PUSHINT: - i++; - PUSH(_mesa_atoi ((const char *) (&code[i]))); - i += _mesa_strlen ((const char *) (&code[i])) + 1; - break; - case OP_LOGICALOR: - BINARY(||); - break; - case OP_LOGICALAND: - BINARY(&&); - break; - case OP_OR: - BINARY(|); - break; - case OP_XOR: - BINARY(^); - break; - case OP_AND: - BINARY(&); - break; - case OP_EQUAL: - BINARY(==); - break; - case OP_NOTEQUAL: - BINARY(!=); - break; - case OP_LESSEQUAL: - BINARY(<=); - break; - case OP_GREATEREQUAL: - BINARY(>=); - break; - case OP_LESS: - BINARY(<); - break; - case OP_GREATER: - BINARY(>); - break; - case OP_LEFTSHIFT: - BINARY(<<); - break; - case OP_RIGHTSHIFT: - BINARY(>>); - break; - case OP_ADD: - BINARY(+); - break; - case OP_SUBTRACT: - BINARY(-); - break; - case OP_MULTIPLY: - BINARY(*); - break; - case OP_DIVIDE: - BINARYDIV(/); - break; - case OP_MODULUS: - BINARYDIV(%); - break; - case OP_PLUS: - UNARY(+); - break; - case OP_MINUS: - UNARY(-); - break; - case OP_NEGATE: - UNARY(!); - break; - case OP_COMPLEMENT: - UNARY(~); - break; - default: - assert (0); - } - } - - /* Write-back the index skipping the OP_END. */ - *pi = i + 1; - - /* There should be exactly one value left on the stack. This is our result. */ - POP(*result); - pp_annotate (output, "%d ", *result); - assert (sp == EXECUTION_STACK_SIZE); - return GL_TRUE; -} - -/* - * Function execute_expressions() executes up to 2 expressions. The second expression is there - * for the #line directive which takes 1 or 2 expressions that indicate line and file numbers. - * If it fails, it returns 0. If it succeeds, it returns the number of executed expressions. - */ - -#define EXP_END 0 -#define EXP_EXPRESSION 1 - -static GLuint -execute_expressions (slang_string *output, grammar eid, const byte *expr, GLint results[2], - slang_info_log *elog) -{ - GLint success; - byte *code; - GLuint size, count = 0; - - success = grammar_fast_check (eid, expr, &code, &size, 64); - if (success) { - GLuint i = 0; - - while (code[i++] == EXP_EXPRESSION) { - assert (count < 2); - - if (!execute_expression (output, code, &i, &results[count], elog)) { - count = 0; - break; - } - count++; - } - grammar_alloc_free (code); - } - else { - slang_info_log_error (elog, "syntax error in preprocessor expression.");\ - } - return count; -} - -/* - * The pp_symbol structure is used to hold macro definitions and macro formal parameters. The - * pp_symbols strcture is a collection of pp_symbol. It is used both for storing macro formal - * parameters and all global macro definitions. Making this unification wastes some memory, - * becuse macro formal parameters don't need further lists of symbols. We lose 8 bytes per - * formal parameter here, but making this we can use the same code to substitute macro parameters - * as well as macros in the source string. - */ - -typedef struct -{ - struct pp_symbol_ *symbols; - GLuint count; -} pp_symbols; - -static GLvoid -pp_symbols_init (pp_symbols *self) -{ - self->symbols = NULL; - self->count = 0; -} - -static GLvoid -pp_symbols_free (pp_symbols *); - -typedef struct pp_symbol_ -{ - slang_string name; - slang_string replacement; - pp_symbols parameters; -} pp_symbol; - -static GLvoid -pp_symbol_init (pp_symbol *self) -{ - slang_string_init (&self->name); - slang_string_init (&self->replacement); - pp_symbols_init (&self->parameters); -} - -static GLvoid -pp_symbol_free (pp_symbol *self) -{ - slang_string_free (&self->name); - slang_string_free (&self->replacement); - pp_symbols_free (&self->parameters); -} - -static GLvoid -pp_symbol_reset (pp_symbol *self) -{ - /* Leave symbol name intact. */ - slang_string_reset (&self->replacement); - pp_symbols_free (&self->parameters); - pp_symbols_init (&self->parameters); -} - -static GLvoid -pp_symbols_free (pp_symbols *self) -{ - GLuint i; - - for (i = 0; i < self->count; i++) - pp_symbol_free (&self->symbols[i]); - _mesa_free (self->symbols); -} - -static pp_symbol * -pp_symbols_push (pp_symbols *self) -{ - self->symbols = (pp_symbol *) (_mesa_realloc (self->symbols, self->count * sizeof (pp_symbol), - (self->count + 1) * sizeof (pp_symbol))); - if (self->symbols == NULL) - return NULL; - pp_symbol_init (&self->symbols[self->count]); - return &self->symbols[self->count++]; -} - -static GLboolean -pp_symbols_erase (pp_symbols *self, pp_symbol *symbol) -{ - assert (symbol >= self->symbols && symbol < self->symbols + self->count); - - self->count--; - pp_symbol_free (symbol); - if (symbol < self->symbols + self->count) - _mesa_memcpy (symbol, symbol + 1, sizeof (pp_symbol) * (self->symbols + self->count - symbol)); - self->symbols = (pp_symbol *) (_mesa_realloc (self->symbols, (self->count + 1) * sizeof (pp_symbol), - self->count * sizeof (pp_symbol))); - return self->symbols != NULL; -} - -static pp_symbol * -pp_symbols_find (pp_symbols *self, const char *name) -{ - GLuint i; - - for (i = 0; i < self->count; i++) - if (_mesa_strcmp (name, slang_string_cstr (&self->symbols[i].name)) == 0) - return &self->symbols[i]; - return NULL; -} - -/* - * The condition context of a single #if/#else/#endif level. Those can be nested, so there - * is a stack of condition contexts. - * There is a special global context on the bottom of the stack. It is there to simplify - * context handling. - */ - -typedef struct -{ - GLboolean current; /* The condition value of this level. */ - GLboolean effective; /* The effective product of current condition, outer level conditions - * and position within #if-#else-#endif sections. */ - GLboolean else_allowed; /* TRUE if in #if-#else section, FALSE if in #else-#endif section - * and for global context. */ - GLboolean endif_required; /* FALSE for global context only. */ -} pp_cond_ctx; - -/* Should be enuff. */ -#define CONDITION_STACK_SIZE 64 - -typedef struct -{ - pp_cond_ctx stack[CONDITION_STACK_SIZE]; - pp_cond_ctx *top; -} pp_cond_stack; - -static GLboolean -pp_cond_stack_push (pp_cond_stack *self, slang_info_log *elog) -{ - if (self->top == self->stack) { - slang_info_log_error (elog, "internal compiler error: preprocessor condition stack overflow."); - return GL_FALSE; - } - self->top--; - return GL_TRUE; -} - -static GLvoid -pp_cond_stack_reevaluate (pp_cond_stack *self) -{ - /* There must be at least 2 conditions on the stack - one global and one being evaluated. */ - assert (self->top <= &self->stack[CONDITION_STACK_SIZE - 2]); - - self->top->effective = self->top->current && self->top[1].effective; -} - - -/** - * Extension enables through #extension directive. - * NOTE: Currently, only enable/disable state is stored. - */ -typedef struct -{ - GLboolean ARB_draw_buffers; - GLboolean ARB_texture_rectangle; -} pp_ext; - - -/** - * Disable all extensions. Called at startup and on #extension all: disable. - */ -static GLvoid -pp_ext_disable_all(pp_ext *self) -{ - _mesa_memset(self, 0, sizeof(self)); -} - - -/** - * Called during preprocessor initialization to set the initial enable/disable - * state of extensions. - */ -static GLvoid -pp_ext_init(pp_ext *self, const struct gl_extensions *extensions) -{ - pp_ext_disable_all (self); - self->ARB_draw_buffers = GL_TRUE; - if (extensions->NV_texture_rectangle) - self->ARB_texture_rectangle = GL_TRUE; -} - -/** - * Called in response to #extension directives to enable/disable - * the named extension. - */ -static GLboolean -pp_ext_set(pp_ext *self, const char *name, GLboolean enable) -{ - if (_mesa_strcmp (name, "GL_ARB_draw_buffers") == 0) - self->ARB_draw_buffers = enable; - else if (_mesa_strcmp (name, "GL_ARB_texture_rectangle") == 0) - self->ARB_texture_rectangle = enable; - else - return GL_FALSE; - return GL_TRUE; -} - - -/** - * Called in response to #pragma. For example, "#pragma debug(on)" would - * call this function as pp_pragma("debug", "on"). - * \return GL_TRUE if pragma is valid, GL_FALSE if invalid - */ -static GLboolean -pp_pragma(struct gl_sl_pragmas *pragmas, const char *pragma, const char *param) -{ -#if 0 - printf("#pragma %s %s\n", pragma, param); -#endif - if (_mesa_strcmp(pragma, "optimize") == 0) { - if (!param) - return GL_FALSE; /* missing required param */ - if (_mesa_strcmp(param, "on") == 0) { - if (!pragmas->IgnoreOptimize) - pragmas->Optimize = GL_TRUE; - } - else if (_mesa_strcmp(param, "off") == 0) { - if (!pragmas->IgnoreOptimize) - pragmas->Optimize = GL_FALSE; - } - else { - return GL_FALSE; /* invalid param */ - } - } - else if (_mesa_strcmp(pragma, "debug") == 0) { - if (!param) - return GL_FALSE; /* missing required param */ - if (_mesa_strcmp(param, "on") == 0) { - if (!pragmas->IgnoreDebug) - pragmas->Debug = GL_TRUE; - } - else if (_mesa_strcmp(param, "off") == 0) { - if (!pragmas->IgnoreDebug) - pragmas->Debug = GL_FALSE; - } - else { - return GL_FALSE; /* invalid param */ - } - } - /* all other pragmas are silently ignored */ - return GL_TRUE; -} - - -/** - * The state of preprocessor: current line, file and version number, list - * of all defined macros and the #if/#endif context. - */ -typedef struct -{ - GLint line; - GLint file; - GLint version; - pp_symbols symbols; - pp_ext ext; - slang_info_log *elog; - pp_cond_stack cond; -} pp_state; - -static GLvoid -pp_state_init (pp_state *self, slang_info_log *elog, - const struct gl_extensions *extensions) -{ - self->line = 0; - self->file = 1; -#if FEATURE_es2_glsl - self->version = 100; -#else - self->version = 110; -#endif - pp_symbols_init (&self->symbols); - pp_ext_init (&self->ext, extensions); - self->elog = elog; - - /* Initialize condition stack and create the global context. */ - self->cond.top = &self->cond.stack[CONDITION_STACK_SIZE - 1]; - self->cond.top->current = GL_TRUE; - self->cond.top->effective = GL_TRUE; - self->cond.top->else_allowed = GL_FALSE; - self->cond.top->endif_required = GL_FALSE; -} - -static GLvoid -pp_state_free (pp_state *self) -{ - pp_symbols_free (&self->symbols); -} - -#define IS_FIRST_ID_CHAR(x) (((x) >= 'a' && (x) <= 'z') || ((x) >= 'A' && (x) <= 'Z') || (x) == '_') -#define IS_NEXT_ID_CHAR(x) (IS_FIRST_ID_CHAR(x) || ((x) >= '0' && (x) <= '9')) -#define IS_WHITE(x) ((x) == ' ' || (x) == '\n') -#define IS_NULL(x) ((x) == '\0') - -#define SKIP_WHITE(x) do { while (IS_WHITE(*(x))) (x)++; } while (GL_FALSE) - -typedef struct -{ - slang_string *output; - const char *input; - pp_state *state; -} expand_state; - -static GLboolean -expand_defined (expand_state *e, slang_string *buffer) -{ - GLboolean in_paren = GL_FALSE; - const char *id; - - /* Parse the optional opening parenthesis. */ - SKIP_WHITE(e->input); - if (*e->input == '(') { - e->input++; - in_paren = GL_TRUE; - SKIP_WHITE(e->input); - } - - /* Parse operand. */ - if (!IS_FIRST_ID_CHAR(*e->input)) { - slang_info_log_error (e->state->elog, - "preprocess error: identifier expected after operator 'defined'."); - return GL_FALSE; - } - slang_string_reset (buffer); - slang_string_pushc (buffer, *e->input++); - while (IS_NEXT_ID_CHAR(*e->input)) - slang_string_pushc (buffer, *e->input++); - id = slang_string_cstr (buffer); - - /* Check if the operand is defined. Output 1 if it is defined, output 0 if not. */ - if (pp_symbols_find (&e->state->symbols, id) == NULL) - slang_string_pushs (e->output, " 0 ", 3); - else - slang_string_pushs (e->output, " 1 ", 3); - - /* Parse the closing parentehesis if the opening one was there. */ - if (in_paren) { - SKIP_WHITE(e->input); - if (*e->input != ')') { - slang_info_log_error (e->state->elog, "preprocess error: ')' expected."); - return GL_FALSE; - } - e->input++; - SKIP_WHITE(e->input); - } - return GL_TRUE; -} - -static GLboolean -expand (expand_state *, pp_symbols *); - -static GLboolean -expand_symbol (expand_state *e, pp_symbol *symbol) -{ - expand_state es; - - /* If the macro has some parameters, we need to parse them. */ - if (symbol->parameters.count != 0) { - GLuint i; - - /* Parse the opening parenthesis. */ - SKIP_WHITE(e->input); - if (*e->input != '(') { - slang_info_log_error (e->state->elog, "preprocess error: '(' expected."); - return GL_FALSE; - } - e->input++; - SKIP_WHITE(e->input); - - /* Parse macro actual parameters. This can be anything, separated by a colon. - */ - for (i = 0; i < symbol->parameters.count; i++) { - GLuint nested_paren_count = 0; /* track number of nested parentheses */ - - if (*e->input == ')') { - slang_info_log_error (e->state->elog, "preprocess error: unexpected ')'."); - return GL_FALSE; - } - - /* Eat all characters up to the comma or closing parentheses. */ - pp_symbol_reset (&symbol->parameters.symbols[i]); - while (!IS_NULL(*e->input)) { - /* Exit loop only when all nested parens have been eaten. */ - if (nested_paren_count == 0 && (*e->input == ',' || *e->input == ')')) - break; - - /* Actually count nested parens here. */ - if (*e->input == '(') - nested_paren_count++; - else if (*e->input == ')') - nested_paren_count--; - - slang_string_pushc (&symbol->parameters.symbols[i].replacement, *e->input++); - } - - /* If it was not the last paremeter, skip the comma. Otherwise, skip the - * closing parentheses. */ - if (i + 1 == symbol->parameters.count) { - /* This is the last paremeter - skip the closing parentheses. */ - if (*e->input != ')') { - slang_info_log_error (e->state->elog, "preprocess error: ')' expected."); - return GL_FALSE; - } - e->input++; - SKIP_WHITE(e->input); - } - else { - /* Skip the separating comma. */ - if (*e->input != ',') { - slang_info_log_error (e->state->elog, "preprocess error: ',' expected."); - return GL_FALSE; - } - e->input++; - SKIP_WHITE(e->input); - } - } - } - - /* Expand the macro. Use its parameters as a priority symbol list to expand - * macro parameters correctly. */ - es.output = e->output; - es.input = slang_string_cstr (&symbol->replacement); - es.state = e->state; - slang_string_pushc (e->output, ' '); - if (!expand (&es, &symbol->parameters)) - return GL_FALSE; - slang_string_pushc (e->output, ' '); - return GL_TRUE; -} - -/* - * Function expand() expands source text from <input> to <output>. The expansion is made using - * the list passed in <symbols> parameter. It allows us to expand macro formal parameters with - * actual parameters. The global list of symbols from pp state is used when doing a recursive - * call of expand(). - */ - -static GLboolean -expand (expand_state *e, pp_symbols *symbols) -{ - while (!IS_NULL(*e->input)) { - if (IS_FIRST_ID_CHAR(*e->input)) { - slang_string buffer; - const char *id; - - /* Parse the identifier. */ - slang_string_init (&buffer); - slang_string_pushc (&buffer, *e->input++); - while (IS_NEXT_ID_CHAR(*e->input)) - slang_string_pushc (&buffer, *e->input++); - id = slang_string_cstr (&buffer); - - /* Now check if the identifier is special in some way. The "defined" identifier is - * actually an operator that we must handle here and expand it either to " 0 " or " 1 ". - * The other identifiers start with "__" and we expand it to appropriate values - * taken from the preprocessor state. */ - if (_mesa_strcmp (id, "defined") == 0) { - if (!expand_defined (e, &buffer)) - return GL_FALSE; - } - else if (_mesa_strcmp (id, "__LINE__") == 0) { - slang_string_pushc (e->output, ' '); - slang_string_pushi (e->output, e->state->line); - slang_string_pushc (e->output, ' '); - } - else if (_mesa_strcmp (id, "__FILE__") == 0) { - slang_string_pushc (e->output, ' '); - slang_string_pushi (e->output, e->state->file); - slang_string_pushc (e->output, ' '); - } - else if (_mesa_strcmp (id, "__VERSION__") == 0) { - slang_string_pushc (e->output, ' '); - slang_string_pushi (e->output, e->state->version); - slang_string_pushc (e->output, ' '); - } -#if FEATURE_es2_glsl - else if (_mesa_strcmp (id, "GL_ES") == 0 || - _mesa_strcmp (id, "GL_FRAGMENT_PRECISION_HIGH") == 0) { - slang_string_pushc (e->output, ' '); - slang_string_pushi (e->output, '1'); - slang_string_pushc (e->output, ' '); - } -#endif - else { - pp_symbol *symbol; - - /* The list of symbols from <symbols> take precedence over the list from <state>. - * Note that in some cases this is the same list so avoid double look-up. */ - symbol = pp_symbols_find (symbols, id); - if (symbol == NULL && symbols != &e->state->symbols) - symbol = pp_symbols_find (&e->state->symbols, id); - - /* If the symbol was found, recursively expand its definition. */ - if (symbol != NULL) { - if (!expand_symbol (e, symbol)) { - slang_string_free (&buffer); - return GL_FALSE; - } - } - else { - slang_string_push (e->output, &buffer); - } - } - slang_string_free (&buffer); - } - else if (IS_WHITE(*e->input)) { - slang_string_pushc (e->output, *e->input++); - } - else { - while (!IS_WHITE(*e->input) && !IS_NULL(*e->input) && !IS_FIRST_ID_CHAR(*e->input)) - slang_string_pushc (e->output, *e->input++); - } - } - return GL_TRUE; -} - -static GLboolean -parse_if (slang_string *output, const byte *prod, GLuint *pi, GLint *result, pp_state *state, - grammar eid) -{ - const char *text; - GLuint len; - - text = (const char *) (&prod[*pi]); - len = _mesa_strlen (text); - - if (state->cond.top->effective) { - slang_string expr; - GLuint count; - GLint results[2]; - expand_state es; - - /* Expand the expression. */ - slang_string_init (&expr); - es.output = &expr; - es.input = text; - es.state = state; - if (!expand (&es, &state->symbols)) - return GL_FALSE; - - /* Execute the expression. */ - count = execute_expressions (output, eid, (const byte *) (slang_string_cstr (&expr)), - results, state->elog); - slang_string_free (&expr); - if (count != 1) - return GL_FALSE; - *result = results[0]; - } - else { - /* The directive is dead. */ - *result = 0; - } - - *pi += len + 1; - return GL_TRUE; -} - -#define ESCAPE_TOKEN 0 - -#define TOKEN_END 0 -#define TOKEN_DEFINE 1 -#define TOKEN_UNDEF 2 -#define TOKEN_IF 3 -#define TOKEN_ELSE 4 -#define TOKEN_ELIF 5 -#define TOKEN_ENDIF 6 -#define TOKEN_ERROR 7 -#define TOKEN_PRAGMA 8 -#define TOKEN_EXTENSION 9 -#define TOKEN_LINE 10 - -#define PARAM_END 0 -#define PARAM_PARAMETER 1 - -#define BEHAVIOR_REQUIRE 1 -#define BEHAVIOR_ENABLE 2 -#define BEHAVIOR_WARN 3 -#define BEHAVIOR_DISABLE 4 - -#define PRAGMA_NO_PARAM 0 -#define PRAGMA_PARAM 1 - - -static GLboolean -preprocess_source (slang_string *output, const char *source, - grammar pid, grammar eid, - slang_info_log *elog, - const struct gl_extensions *extensions, - struct gl_sl_pragmas *pragmas) -{ - static const char *predefined[] = { - "__FILE__", - "__LINE__", - "__VERSION__", -#if FEATURE_es2_glsl - "GL_ES", - "GL_FRAGMENT_PRECISION_HIGH", -#endif - NULL - }; - byte *prod; - GLuint size, i; - pp_state state; - - if (!grammar_fast_check (pid, (const byte *) (source), &prod, &size, 65536)) { - grammar_error_to_log (elog); - return GL_FALSE; - } - - pp_state_init (&state, elog, extensions); - - /* add the predefined symbols to the symbol table */ - for (i = 0; predefined[i]; i++) { - pp_symbol *symbol = NULL; - symbol = pp_symbols_push(&state.symbols); - assert(symbol); - slang_string_pushs(&symbol->name, - predefined[i], _mesa_strlen(predefined[i])); - } - - i = 0; - while (i < size) { - if (prod[i] != ESCAPE_TOKEN) { - if (state.cond.top->effective) { - slang_string input; - expand_state es; - - /* Eat only one line of source code to expand it. - * FIXME: This approach has one drawback. If a macro with parameters spans across - * multiple lines, the preprocessor will raise an error. */ - slang_string_init (&input); - while (prod[i] != '\0' && prod[i] != '\n') - slang_string_pushc (&input, prod[i++]); - if (prod[i] != '\0') - slang_string_pushc (&input, prod[i++]); - - /* Increment line number. */ - state.line++; - - es.output = output; - es.input = slang_string_cstr (&input); - es.state = &state; - if (!expand (&es, &state.symbols)) - goto error; - - slang_string_free (&input); - } - else { - /* Condition stack is disabled - keep track on line numbers and output only newlines. */ - if (prod[i] == '\n') { - state.line++; - /*pp_annotate (output, "%c", prod[i]);*/ - } - else { - /*pp_annotate (output, "%c", prod[i]);*/ - } - i++; - } - } - else { - const char *id; - GLuint idlen; - GLubyte token; - - i++; - token = prod[i++]; - switch (token) { - - case TOKEN_END: - /* End of source string. - * Check if all #ifs have been terminated by matching #endifs. - * On condition stack there should be only the global condition context. */ - if (state.cond.top->endif_required) { - slang_info_log_error (elog, "end of source without matching #endif."); - return GL_FALSE; - } - break; - - case TOKEN_DEFINE: - { - pp_symbol *symbol = NULL; - - /* Parse macro name. */ - id = (const char *) (&prod[i]); - idlen = _mesa_strlen (id); - if (state.cond.top->effective) { - pp_annotate (output, "// #define %s(", id); - - /* If the symbol is already defined, override it. */ - symbol = pp_symbols_find (&state.symbols, id); - if (symbol == NULL) { - symbol = pp_symbols_push (&state.symbols); - if (symbol == NULL) - goto error; - slang_string_pushs (&symbol->name, id, idlen); - } - else { - pp_symbol_reset (symbol); - } - } - i += idlen + 1; - - /* Parse optional macro parameters. */ - while (prod[i++] != PARAM_END) { - pp_symbol *param; - - id = (const char *) (&prod[i]); - idlen = _mesa_strlen (id); - if (state.cond.top->effective) { - pp_annotate (output, "%s, ", id); - param = pp_symbols_push (&symbol->parameters); - if (param == NULL) - goto error; - slang_string_pushs (¶m->name, id, idlen); - } - i += idlen + 1; - } - - /* Parse macro replacement. */ - id = (const char *) (&prod[i]); - idlen = _mesa_strlen (id); - if (state.cond.top->effective) { - slang_string replacement; - expand_state es; - - pp_annotate (output, ") %s", id); - - slang_string_init(&replacement); - slang_string_pushs(&replacement, id, idlen); - - /* Expand macro replacement. */ - es.output = &symbol->replacement; - es.input = slang_string_cstr(&replacement); - es.state = &state; - if (!expand(&es, &state.symbols)) { - slang_string_free(&replacement); - goto error; - } - slang_string_free(&replacement); - } - i += idlen + 1; - } - break; - - case TOKEN_UNDEF: - id = (const char *) (&prod[i]); - i += _mesa_strlen (id) + 1; - if (state.cond.top->effective) { - pp_symbol *symbol; - - pp_annotate (output, "// #undef %s", id); - /* Try to find symbol with given name and remove it. */ - symbol = pp_symbols_find (&state.symbols, id); - if (symbol != NULL) - if (!pp_symbols_erase (&state.symbols, symbol)) - goto error; - } - break; - - case TOKEN_IF: - { - GLint result; - - /* Parse #if expression end execute it. */ - pp_annotate (output, "// #if "); - if (!parse_if (output, prod, &i, &result, &state, eid)) - goto error; - - /* Push new condition on the stack. */ - if (!pp_cond_stack_push (&state.cond, state.elog)) - goto error; - state.cond.top->current = result ? GL_TRUE : GL_FALSE; - state.cond.top->else_allowed = GL_TRUE; - state.cond.top->endif_required = GL_TRUE; - pp_cond_stack_reevaluate (&state.cond); - } - break; - - case TOKEN_ELSE: - /* Check if #else is alloved here. */ - if (!state.cond.top->else_allowed) { - slang_info_log_error (elog, "#else without matching #if."); - goto error; - } - - /* Negate current condition and reevaluate it. */ - state.cond.top->current = !state.cond.top->current; - state.cond.top->else_allowed = GL_FALSE; - pp_cond_stack_reevaluate (&state.cond); - if (state.cond.top->effective) - pp_annotate (output, "// #else"); - break; - - case TOKEN_ELIF: - /* Check if #elif is alloved here. */ - if (!state.cond.top->else_allowed) { - slang_info_log_error (elog, "#elif without matching #if."); - goto error; - } - - /* Negate current condition and reevaluate it. */ - state.cond.top->current = !state.cond.top->current; - pp_cond_stack_reevaluate (&state.cond); - - if (state.cond.top->effective) - pp_annotate (output, "// #elif "); - - { - GLint result; - - /* Parse #elif expression end execute it. */ - if (!parse_if (output, prod, &i, &result, &state, eid)) - goto error; - - /* Update current condition and reevaluate it. */ - state.cond.top->current = result ? GL_TRUE : GL_FALSE; - pp_cond_stack_reevaluate (&state.cond); - } - break; - - case TOKEN_ENDIF: - /* Check if #endif is alloved here. */ - if (!state.cond.top->endif_required) { - slang_info_log_error (elog, "#endif without matching #if."); - goto error; - } - - /* Pop the condition off the stack. */ - state.cond.top++; - if (state.cond.top->effective) - pp_annotate (output, "// #endif"); - break; - - case TOKEN_EXTENSION: - /* Parse the extension name. */ - id = (const char *) (&prod[i]); - i += _mesa_strlen (id) + 1; - if (state.cond.top->effective) - pp_annotate (output, "// #extension %s: ", id); - - /* Parse and apply extension behavior. */ - if (state.cond.top->effective) { - switch (prod[i++]) { - - case BEHAVIOR_REQUIRE: - pp_annotate (output, "require"); - if (!pp_ext_set (&state.ext, id, GL_TRUE)) { - if (_mesa_strcmp (id, "all") == 0) { - slang_info_log_error (elog, "require: bad behavior for #extension all."); - goto error; - } - else { - slang_info_log_error (elog, "%s: required extension is not supported.", id); - goto error; - } - } - break; - - case BEHAVIOR_ENABLE: - pp_annotate (output, "enable"); - if (!pp_ext_set (&state.ext, id, GL_TRUE)) { - if (_mesa_strcmp (id, "all") == 0) { - slang_info_log_error (elog, "enable: bad behavior for #extension all."); - goto error; - } - else { - slang_info_log_warning (elog, "%s: enabled extension is not supported.", id); - } - } - break; - - case BEHAVIOR_WARN: - pp_annotate (output, "warn"); - if (!pp_ext_set (&state.ext, id, GL_TRUE)) { - if (_mesa_strcmp (id, "all") != 0) { - slang_info_log_warning (elog, "%s: enabled extension is not supported.", id); - } - } - break; - - case BEHAVIOR_DISABLE: - pp_annotate (output, "disable"); - if (!pp_ext_set (&state.ext, id, GL_FALSE)) { - if (_mesa_strcmp (id, "all") == 0) { - pp_ext_disable_all (&state.ext); - } - else { - slang_info_log_warning (elog, "%s: disabled extension is not supported.", id); - } - } - break; - - default: - assert (0); - } - } - break; - - case TOKEN_PRAGMA: - { - GLint have_param; - const char *pragma, *param; - - pragma = (const char *) (&prod[i]); - i += _mesa_strlen(pragma) + 1; - have_param = (prod[i++] == PRAGMA_PARAM); - if (have_param) { - param = (const char *) (&prod[i]); - i += _mesa_strlen(param) + 1; - } - else { - param = NULL; - } - pp_pragma(pragmas, pragma, param); - } - break; - - case TOKEN_LINE: - id = (const char *) (&prod[i]); - i += _mesa_strlen (id) + 1; - - if (state.cond.top->effective) { - slang_string buffer; - GLuint count; - GLint results[2]; - expand_state es; - - slang_string_init (&buffer); - state.line++; - es.output = &buffer; - es.input = id; - es.state = &state; - if (!expand (&es, &state.symbols)) - goto error; - - pp_annotate (output, "// #line "); - count = execute_expressions (output, eid, - (const byte *) (slang_string_cstr (&buffer)), - results, state.elog); - slang_string_free (&buffer); - if (count == 0) - goto error; - - state.line = results[0] - 1; - if (count == 2) - state.file = results[1]; - } - break; - } - } - } - - /* Check for missing #endifs. */ - if (state.cond.top->endif_required) { - slang_info_log_error (elog, "#endif expected but end of source found."); - goto error; - } - - grammar_alloc_free(prod); - pp_state_free (&state); - return GL_TRUE; - -error: - grammar_alloc_free(prod); - pp_state_free (&state); - return GL_FALSE; -} - - -/** - * Remove the continuation characters from the input string. - * This is the very first step in preprocessing and is effective - * even inside comment blocks. - * If there is a whitespace between a backslash and a newline, - * this is not considered as a line continuation. - * \return GL_TRUE for success, GL_FALSE otherwise. - */ -static GLboolean -_slang_preprocess_backslashes(slang_string *output, - const char *input) -{ - while (*input) { - if (input[0] == '\\') { - /* If a newline follows, eat the backslash and the newline. */ - if (input[1] == '\r') { - if (input[2] == '\n') { - input += 3; - } else { - input += 2; - } - } else if (input[1] == '\n') { - if (input[2] == '\r') { - input += 3; - } else { - input += 2; - } - } else { - /* Leave the backslash alone. */ - slang_string_pushc(output, *input++); - } - } else { - slang_string_pushc(output, *input++); - } - } - return GL_TRUE; -} - - -/** - * Run preprocessor on source code. - * \param extensions indicates which GL extensions are enabled - * \param output the post-process results - * \param input the input text - * \param elog log to record warnings, errors - * \param extensions out extension settings - * \param pragmas in/out #pragma settings - * \return GL_TRUE for success, GL_FALSE for error - */ -GLboolean -_slang_preprocess_directives(slang_string *output, - const char *input, - slang_info_log *elog, - const struct gl_extensions *extensions, - struct gl_sl_pragmas *pragmas) -{ - grammar pid, eid; - GLboolean success; - slang_string without_backslashes; - - pid = grammar_load_from_text ((const byte *) (slang_pp_directives_syn)); - if (pid == 0) { - grammar_error_to_log (elog); - return GL_FALSE; - } - eid = grammar_load_from_text ((const byte *) (slang_pp_expression_syn)); - if (eid == 0) { - grammar_error_to_log (elog); - grammar_destroy (pid); - return GL_FALSE; - } - - slang_string_init(&without_backslashes); - success = _slang_preprocess_backslashes(&without_backslashes, input); - - if (0) { - _mesa_printf("Pre-processed shader:\n"); - _mesa_printf("%s", slang_string_cstr(&without_backslashes)); - _mesa_printf("----------------------\n"); - } - - if (success) { - success = preprocess_source(output, - slang_string_cstr(&without_backslashes), - pid, - eid, - elog, - extensions, - pragmas); - } - - slang_string_free(&without_backslashes); - grammar_destroy (eid); - grammar_destroy (pid); - - if (0) { - _mesa_printf("Post-processed shader:\n"); - _mesa_printf("%s", slang_string_cstr(output)); - _mesa_printf("----------------------\n"); - } - - return success; -} - diff --git a/src/mesa/shader/slang/slang_preprocess.h b/src/mesa/shader/slang/slang_preprocess.h deleted file mode 100644 index f344820daef..00000000000 --- a/src/mesa/shader/slang/slang_preprocess.h +++ /dev/null @@ -1,41 +0,0 @@ -/* - * Mesa 3-D graphics library - * - * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. - * Copyright (C) 2009 VMware, Inc. All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef SLANG_PREPROCESS_H -#define SLANG_PREPROCESS_H - -#include "slang_compile.h" -#include "slang_log.h" - - -extern GLboolean -_slang_preprocess_version (const char *, GLuint *, GLuint *, slang_info_log *); - -extern GLboolean -_slang_preprocess_directives(slang_string *output, const char *input, - slang_info_log *, - const struct gl_extensions *extensions, - struct gl_sl_pragmas *pragmas); - -#endif /* SLANG_PREPROCESS_H */ diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index b8a21f642cb..13b9ca3c877 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -84,10 +84,11 @@ _slang_lookup_constant(const char *name) for (i = 0; info[i].Name; i++) { if (strcmp(info[i].Name, name) == 0) { /* found */ - GLint value = -1; - _mesa_GetIntegerv(info[i].Token, &value); - ASSERT(value >= 0); /* sanity check that glGetFloatv worked */ - return value; + GLint values[16]; + values[0] = -1; + _mesa_GetIntegerv(info[i].Token, values); + ASSERT(values[0] >= 0); /* sanity check that glGetFloatv worked */ + return values[0]; } } return -1; diff --git a/src/mesa/sources.mak b/src/mesa/sources.mak index 615a5588eae..a7a3b9a7f97 100644 --- a/src/mesa/sources.mak +++ b/src/mesa/sources.mak @@ -191,7 +191,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ - state_tracker/st_cb_get.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_feedback.c \ @@ -219,7 +218,6 @@ SHADER_SOURCES = \ shader/arbprogparse.c \ shader/arbprogram.c \ shader/atifragshader.c \ - shader/grammar/grammar_mesa.c \ shader/hash_table.c \ shader/lex.yy.c \ shader/nvfragparse.c \ @@ -256,7 +254,6 @@ SLANG_SOURCES = \ shader/slang/slang_link.c \ shader/slang/slang_log.c \ shader/slang/slang_mem.c \ - shader/slang/slang_preprocess.c \ shader/slang/slang_print.c \ shader/slang/slang_simplify.c \ shader/slang/slang_storage.c \ @@ -366,6 +363,12 @@ GLAPI_OBJECTS = \ COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) +### Other archives/libraries + +GLSL_LIBS = \ + $(TOP)/src/glsl/pp/libglslpp.a \ + $(TOP)/src/glsl/cl/libglslcl.a + ### Include directories diff --git a/src/mesa/sparc/xform.S b/src/mesa/sparc/xform.S index f2b9674bf2d..2a7cce41e5a 100644 --- a/src/mesa/sparc/xform.S +++ b/src/mesa/sparc/xform.S @@ -17,7 +17,7 @@ #include "sparc_matrix.h" -#if defined(SVR4) || defined(__SVR4) || defined(__svr4__) +#if defined(SVR4) || defined(__SVR4) || defined(__svr4__) || defined(__arch64__) /* Solaris requires this for 64-bit. */ .register %g2, #scratch .register %g3, #scratch diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index e83b6c92efb..72b30e7c043 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -52,6 +52,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" +#include "util/u_format.h" #include "util/u_pack_color.h" #include "util/u_simple_shaders.h" #include "util/u_draw_quad.h" @@ -341,7 +342,7 @@ static INLINE GLboolean check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) { const struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format); + const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format); if (ctx->Scissor.Enabled && (ctx->Scissor.X != 0 || @@ -365,7 +366,7 @@ static INLINE GLboolean check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) { const struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format); + const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format); const GLuint stencilMax = 0xff; const GLboolean maskStencil = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 1baff190409..7c664267d4e 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -63,6 +63,7 @@ #include "tgsi/tgsi_ureg.h" #include "util/u_tile.h" #include "util/u_draw_quad.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_rect.h" #include "shader/prog_instruction.h" @@ -647,7 +648,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } if(format != GL_DEPTH_STENCIL && - pf_get_component_bits( strb->format, PIPE_FORMAT_COMP_Z ) != 0) + util_format_get_component_bits(strb->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0) usage = PIPE_TRANSFER_READ_WRITE; else usage = PIPE_TRANSFER_WRITE; @@ -842,7 +843,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GL_STENCIL_INDEX, GL_UNSIGNED_BYTE, &ctx->DefaultPacking, buffer); - if(pf_get_component_bits( rbDraw->format, PIPE_FORMAT_COMP_Z ) != 0) + if(util_format_get_component_bits(rbDraw->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0) usage = PIPE_TRANSFER_READ_WRITE; else usage = PIPE_TRANSFER_WRITE; @@ -856,8 +857,8 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, usage, dstx, dsty, width, height); - assert(pf_get_blockwidth(ptDraw->texture->format) == 1); - assert(pf_get_blockheight(ptDraw->texture->format) == 1); + assert(util_format_get_blockwidth(ptDraw->texture->format) == 1); + assert(util_format_get_blockheight(ptDraw->texture->format) == 1); /* map the stencil buffer */ drawMap = screen->transfer_map(screen, ptDraw); @@ -1083,7 +1084,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (ST_DEBUG & DEBUG_FALLBACK) debug_printf("%s: fallback processing\n", __FUNCTION__); - if (type == GL_DEPTH && pf_is_depth_and_stencil(pt->format)) + if (type == GL_DEPTH && util_format_is_depth_and_stencil(pt->format)) transfer_usage = PIPE_TRANSFER_READ_WRITE; else transfer_usage = PIPE_TRANSFER_WRITE; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index ead8e228887..45ce34a85f7 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -49,6 +49,7 @@ #include "st_public.h" #include "st_texture.h" +#include "util/u_format.h" #include "util/u_rect.h" @@ -104,8 +105,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, assert(strb->format != PIPE_FORMAT_NONE); - strb->stride = pf_get_stride(strb->format, width); - size = pf_get_2d_size(strb->format, strb->stride, height); + strb->stride = util_format_get_stride(strb->format, width); + size = util_format_get_2d_size(strb->format, strb->stride, height); strb->data = _mesa_malloc(size); @@ -130,7 +131,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.depth0 = 1; template.last_level = 0; template.nr_samples = rb->NumSamples; - if (pf_is_depth_stencil(format)) { + if (util_format_is_depth_or_stencil(format)) { template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; } else { diff --git a/src/mesa/state_tracker/st_cb_get.c b/src/mesa/state_tracker/st_cb_get.c deleted file mode 100644 index e7d7f03bc9b..00000000000 --- a/src/mesa/state_tracker/st_cb_get.c +++ /dev/null @@ -1,97 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * glGet functions - * - * \author Brian Paul - */ - -#include "main/imports.h" -#include "main/context.h" - -#include "pipe/p_defines.h" - -#include "st_cb_fbo.h" -#include "st_cb_get.h" - - - -/** - * Examine the current color read buffer format to determine - * which GL pixel format/type combo is the best match. - */ -static void -get_preferred_read_format_type(GLcontext *ctx, GLint *format, GLint *type) -{ - struct gl_framebuffer *fb = ctx->ReadBuffer; - struct st_renderbuffer *strb = st_renderbuffer(fb->_ColorReadBuffer); - - /* defaults */ - *format = ctx->Const.ColorReadFormat; - *type = ctx->Const.ColorReadType; - - if (strb) { - /* XXX could add more cases here... */ - if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM) { - *format = GL_BGRA; - if (_mesa_little_endian()) - *type = GL_UNSIGNED_INT_8_8_8_8_REV; - else - *type = GL_UNSIGNED_INT_8_8_8_8; - } - } -} - - -/** - * We only intercept the OES preferred ReadPixels format/type. - * Everything else goes to the default _mesa_GetIntegerv. - */ -static GLboolean -st_GetIntegerv(GLcontext *ctx, GLenum pname, GLint *params) -{ - GLint dummy; - - switch (pname) { - case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: - get_preferred_read_format_type(ctx, &dummy, params); - return GL_TRUE; - case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: - get_preferred_read_format_type(ctx, params, &dummy); - return GL_TRUE; - default: - return GL_FALSE; - } -} - - -void st_init_get_functions(struct dd_function_table *functions) -{ - functions->GetIntegerv = st_GetIntegerv; -} diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 6d136f5abf3..6e1ecb1c502 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -62,6 +62,7 @@ #include "pipe/p_shader_tokens.h" #include "util/u_tile.h" #include "util/u_blit.h" +#include "util/u_format.h" #include "util/u_surface.h" #include "util/u_math.h" @@ -204,7 +205,7 @@ static GLuint default_usage(enum pipe_format fmt) { GLuint usage = PIPE_TEXTURE_USAGE_SAMPLER; - if (pf_is_depth_stencil(fmt)) + if (util_format_is_depth_or_stencil(fmt)) usage |= PIPE_TEXTURE_USAGE_DEPTH_STENCIL; else usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; @@ -633,7 +634,7 @@ st_TexImage(GLcontext * ctx, if (stImage->pt) { if (format == GL_DEPTH_COMPONENT && - pf_is_depth_and_stencil(stImage->pt->format)) + util_format_is_depth_and_stencil(stImage->pt->format)) transfer_usage = PIPE_TRANSFER_READ_WRITE; else transfer_usage = PIPE_TRANSFER_WRITE; @@ -832,7 +833,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level, /* copy/pack data into user buffer */ if (st_equal_formats(stImage->pt->format, format, type)) { /* memcpy */ - const uint bytesPerRow = width * pf_get_blocksize(stImage->pt->format); + const uint bytesPerRow = width * util_format_get_blocksize(stImage->pt->format); ubyte *map = screen->transfer_map(screen, tex_xfer); GLuint row; for (row = 0; row < height; row++) { @@ -889,7 +890,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, GLubyte *dest; if (stImage->pt && - pf_is_compressed(stImage->pt->format) && + util_format_is_compressed(stImage->pt->format) && !compressed_dst) { /* Need to decompress the texture. * We'll do this by rendering a textured quad. @@ -914,7 +915,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, PIPE_TRANSFER_READ, 0, 0, stImage->base.Width, stImage->base.Height); - texImage->RowStride = stImage->transfer->stride / pf_get_blocksize(stImage->pt->format); + texImage->RowStride = stImage->transfer->stride / util_format_get_blocksize(stImage->pt->format); } else { /* Otherwise, the image should actually be stored in @@ -1040,7 +1041,7 @@ st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level, unsigned face = _mesa_tex_target_to_face(target); if (format == GL_DEPTH_COMPONENT && - pf_is_depth_and_stencil(stImage->pt->format)) + util_format_is_depth_and_stencil(stImage->pt->format)) transfer_usage = PIPE_TRANSFER_READ_WRITE; else transfer_usage = PIPE_TRANSFER_WRITE; @@ -1177,7 +1178,7 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, xoffset, yoffset, width, height); - srcBlockStride = pf_get_stride(pformat, width); + srcBlockStride = util_format_get_stride(pformat, width); dstBlockStride = stImage->transfer->stride; } else { assert(stImage->pt); @@ -1191,16 +1192,16 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, return; } - assert(xoffset % pf_get_blockwidth(pformat) == 0); - assert(yoffset % pf_get_blockheight(pformat) == 0); - assert(width % pf_get_blockwidth(pformat) == 0); - assert(height % pf_get_blockheight(pformat) == 0); + assert(xoffset % util_format_get_blockwidth(pformat) == 0); + assert(yoffset % util_format_get_blockheight(pformat) == 0); + assert(width % util_format_get_blockwidth(pformat) == 0); + assert(height % util_format_get_blockheight(pformat) == 0); - for (y = 0; y < height; y += pf_get_blockheight(pformat)) { + for (y = 0; y < height; y += util_format_get_blockheight(pformat)) { /* don't need to adjust for xoffset and yoffset as st_texture_image_map does that */ - const char *src = (const char*)data + srcBlockStride * pf_get_nblocksy(pformat, y); - char *dst = (char*)texImage->Data + dstBlockStride * pf_get_nblocksy(pformat, y); - memcpy(dst, src, pf_get_stride(pformat, width)); + const char *src = (const char*)data + srcBlockStride * util_format_get_nblocksy(pformat, y); + char *dst = (char*)texImage->Data + dstBlockStride * util_format_get_nblocksy(pformat, y); + memcpy(dst, src, util_format_get_stride(pformat, width)); } if (stImage->pt) { @@ -1264,7 +1265,7 @@ fallback_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level, if ((baseFormat == GL_DEPTH_COMPONENT || baseFormat == GL_DEPTH_STENCIL) && - pf_is_depth_and_stencil(stImage->pt->format)) + util_format_is_depth_and_stencil(stImage->pt->format)) transfer_usage = PIPE_TRANSFER_READ_WRITE; else transfer_usage = PIPE_TRANSFER_WRITE; @@ -1690,10 +1691,10 @@ copy_image_data_to_texture(struct st_context *st, dstLevel, stImage->base.Data, stImage->base.RowStride * - pf_get_blocksize(stObj->pt->format), + util_format_get_blocksize(stObj->pt->format), stImage->base.RowStride * stImage->base.Height * - pf_get_blocksize(stObj->pt->format)); + util_format_get_blocksize(stObj->pt->format)); _mesa_align_free(stImage->base.Data); stImage->base.Data = NULL; } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index f0eddafd331..d18a25ab514 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -51,7 +51,6 @@ #include "st_cb_drawtex.h" #endif #include "st_cb_fbo.h" -#include "st_cb_get.h" #if FEATURE_feedback #include "st_cb_feedback.h" #endif @@ -331,7 +330,6 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_rasterpos_functions(functions); #endif st_init_fbo_functions(functions); - st_init_get_functions(functions); #if FEATURE_feedback st_init_feedback_functions(functions); #endif diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 2d287ef4ef5..e54f21be600 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -332,7 +332,7 @@ setup_interleaved_attribs(GLcontext *ctx, { struct pipe_context *pipe = ctx->st->pipe; GLuint attr; - const GLubyte *offset0; + const GLubyte *offset0 = NULL; for (attr = 0; attr < vpv->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; @@ -490,6 +490,20 @@ check_uniforms(GLcontext *ctx) } +static unsigned translate_prim( GLcontext *ctx, + unsigned prim ) +{ + /* Avoid quadstrips if it's easy to do so: + */ + if (prim == GL_QUAD_STRIP && + ctx->Light.ShadeModel != GL_FLAT && + ctx->Polygon.FrontMode == GL_FILL && + ctx->Polygon.BackMode == GL_FILL) + prim = GL_TRIANGLE_STRIP; + + return prim; +} + /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -513,12 +527,13 @@ st_draw_vbo(GLcontext *ctx, GLuint attr; struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; unsigned num_vbuffers, num_velements; - GLboolean userSpace; + GLboolean userSpace = GL_FALSE; GLboolean vertDataEdgeFlags; /* Gallium probably doesn't want this in some cases. */ if (!index_bounds_valid) - vbo_get_minmax_index(ctx, prims, ib, &min_index, &max_index); + if (!vbo_all_varyings_in_vbos(arrays)) + vbo_get_minmax_index(ctx, prims, ib, &min_index, &max_index); /* sanity check for pointer arithmetic below */ assert(sizeof(arrays[0]->Ptr[0]) == 1); @@ -595,6 +610,7 @@ st_draw_vbo(GLcontext *ctx, struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer *indexBuf = NULL; unsigned indexSize, indexOffset, i; + unsigned prim; switch (ib->type) { case GL_UNSIGNED_INT: @@ -633,16 +649,20 @@ st_draw_vbo(GLcontext *ctx, * through to driver & draw module. These interfaces still * need a bit of work... */ + prim = translate_prim( ctx, prims[i].mode ); + pipe->draw_range_elements(pipe, indexBuf, indexSize, min_index, max_index, - prims[i].mode, + prim, prims[i].start + indexOffset, prims[i].count); } else { for (i = 0; i < nr_prims; i++) { + prim = translate_prim( ctx, prims[i].mode ); + pipe->draw_elements(pipe, indexBuf, indexSize, - prims[i].mode, + prim, prims[i].start + indexOffset, prims[i].count); } } @@ -652,8 +672,12 @@ st_draw_vbo(GLcontext *ctx, else { /* non-indexed */ GLuint i; + GLuint prim; + for (i = 0; i < nr_prims; i++) { - pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); + prim = translate_prim( ctx, prims[i].mode ); + + pipe->draw_arrays(pipe, prim, prims[i].start, prims[i].count); } } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 93125afe9e1..a06621d2b73 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -42,42 +42,34 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "st_context.h" #include "st_format.h" -static GLuint -format_bits( - pipe_format_rgbazs_t info, - GLuint comp ) -{ - return pf_get_component_bits( (enum pipe_format) info, comp ); -} static GLuint -format_max_bits( - pipe_format_rgbazs_t info ) +format_max_bits(enum pipe_format format) { - GLuint size = format_bits( info, PIPE_FORMAT_COMP_R ); + GLuint size = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0); - size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_G ) ); - size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_B ) ); - size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_A ) ); - size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_Z ) ); - size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_S ) ); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)); return size; } static GLuint -format_size( - pipe_format_rgbazs_t info ) +format_size(enum pipe_format format) { return - format_bits( info, PIPE_FORMAT_COMP_R ) + - format_bits( info, PIPE_FORMAT_COMP_G ) + - format_bits( info, PIPE_FORMAT_COMP_B ) + - format_bits( info, PIPE_FORMAT_COMP_A ) + - format_bits( info, PIPE_FORMAT_COMP_Z ) + - format_bits( info, PIPE_FORMAT_COMP_S ); + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1); } /* @@ -86,11 +78,13 @@ format_size( GLboolean st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) { - if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { - pipe_format_rgbazs_t info; + const struct util_format_description *desc; - info = format; + desc = util_format_description(format); + assert(desc); + if (desc->layout == UTIL_FORMAT_LAYOUT_ARITH || + desc->layout == UTIL_FORMAT_LAYOUT_ARRAY) { #if 0 printf("%s\n", pf_name( format ) ); #endif @@ -103,22 +97,22 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->datatype = GL_UNSIGNED_INT_24_8; } else { - const GLuint size = format_max_bits( info ); + const GLuint size = format_max_bits(format); if (size == 8) { - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_BYTE; else pinfo->datatype = GL_BYTE; } else if (size == 16) { - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_SHORT; else pinfo->datatype = GL_SHORT; } else { assert( size <= 32 ); - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_INT; else pinfo->datatype = GL_INT; @@ -126,23 +120,23 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) } /* Component bits */ - pinfo->red_bits = format_bits( info, PIPE_FORMAT_COMP_R ); - pinfo->green_bits = format_bits( info, PIPE_FORMAT_COMP_G ); - pinfo->blue_bits = format_bits( info, PIPE_FORMAT_COMP_B ); - pinfo->alpha_bits = format_bits( info, PIPE_FORMAT_COMP_A ); - pinfo->depth_bits = format_bits( info, PIPE_FORMAT_COMP_Z ); - pinfo->stencil_bits = format_bits( info, PIPE_FORMAT_COMP_S ); + pinfo->red_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0); + pinfo->green_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1); + pinfo->blue_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2); + pinfo->alpha_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3); + pinfo->depth_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0); + pinfo->stencil_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1); pinfo->luminance_bits = 0; pinfo->intensity_bits = 0; /* Format size */ - pinfo->size = format_size( info ) / 8; + pinfo->size = format_size(format) / 8; /* Luminance & Intensity bits */ - if( pf_swizzle_x(info) == PIPE_FORMAT_COMP_R && - pf_swizzle_y(info) == PIPE_FORMAT_COMP_R && - pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) { - if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) { + if (desc->swizzle[0] == UTIL_FORMAT_SWIZZLE_X && + desc->swizzle[1] == UTIL_FORMAT_SWIZZLE_X && + desc->swizzle[2] == UTIL_FORMAT_SWIZZLE_X) { + if (desc->swizzle[3] == UTIL_FORMAT_SWIZZLE_X) { pinfo->intensity_bits = pinfo->red_bits; } else { @@ -153,7 +147,7 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->mesa_format = st_pipe_format_to_mesa_format(format); } - else if (pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR) { + else if (desc->layout == UTIL_FORMAT_LAYOUT_YUV) { pinfo->mesa_format = MESA_FORMAT_YCBCR; pinfo->datatype = GL_UNSIGNED_SHORT; pinfo->size = 2; /* two bytes per "texel" */ diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 77005518302..2c283d464ae 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -37,6 +37,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_gen_mipmap.h" #include "util/u_math.h" @@ -146,8 +147,8 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, srcData = (ubyte *) screen->transfer_map(screen, srcTrans); dstData = (ubyte *) screen->transfer_map(screen, dstTrans); - srcStride = srcTrans->stride / pf_get_blocksize(srcTrans->texture->format); - dstStride = dstTrans->stride / pf_get_blocksize(dstTrans->texture->format); + srcStride = srcTrans->stride / util_format_get_blocksize(srcTrans->texture->format); + dstStride = dstTrans->stride / util_format_get_blocksize(dstTrans->texture->format); _mesa_generate_mipmap_level(target, datatype, comps, 0 /*border*/, @@ -224,6 +225,9 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, /* find expected last mipmap level */ lastLevel = compute_num_levels(ctx, texObj, target) - 1; + if (lastLevel == 0) + return; + if (pt->last_level < lastLevel) { /* The current gallium texture doesn't have space for all the * mipmap levels we need to generate. So allocate a new texture. diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index feaac013c1a..6a869fae904 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -102,6 +102,8 @@ st_prepare_vertex_program(struct st_context *st, if (stvp->Base.IsPositionInvariant) _mesa_insert_mvp_code(st->ctx, &stvp->Base); + assert(stvp->Base.Base.NumInstructions > 1); + /* * Determine number of inputs, the mappings between VERT_ATTRIB_x * and TGSI generic input indexes, plus input attrib semantic info. diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 3035d78b611..8a3e4cd3ac3 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -43,6 +43,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_rect.h" #include "util/u_math.h" @@ -405,7 +406,7 @@ st_bind_texture_surface(struct pipe_surface *ps, int target, int level, } /* map pipe format to base format for now */ - if (pf_get_component_bits(format, PIPE_FORMAT_COMP_A) > 0) + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3) > 0) internalFormat = GL_RGBA; else internalFormat = GL_RGB; diff --git a/src/mesa/swrast/s_atifragshader.c b/src/mesa/swrast/s_atifragshader.c index 5fefae6c42b..e88ff191239 100644 --- a/src/mesa/swrast/s_atifragshader.c +++ b/src/mesa/swrast/s_atifragshader.c @@ -279,7 +279,7 @@ handle_sample_op(GLcontext * ctx, struct atifs_machine *machine, /* sample from unit idx using texinst->src as coords */ GLuint swizzle = texinst->swizzle; GLuint coord_source = texinst->src; - GLfloat tex_coords[4]; + GLfloat tex_coords[4] = { 0 }; if (coord_source >= GL_TEXTURE0_ARB && coord_source <= GL_TEXTURE7_ARB) { coord_source -= GL_TEXTURE0_ARB; diff --git a/src/mesa/swrast/s_triangle.c b/src/mesa/swrast/s_triangle.c index d80a6761f40..5bec6066967 100644 --- a/src/mesa/swrast/s_triangle.c +++ b/src/mesa/swrast/s_triangle.c @@ -553,6 +553,9 @@ affine_span(GLcontext *ctx, SWspan *span, info.format = texImg->TexFormat; \ info.filter = obj->MinFilter; \ info.envmode = unit->EnvMode; \ + info.er = 0; \ + info.eg = 0; \ + info.eb = 0; \ span.arrayMask |= SPAN_RGBA; \ \ if (info.envmode == GL_BLEND) { \ @@ -815,6 +818,9 @@ fast_persp_span(GLcontext *ctx, SWspan *span, info.format = texImg->TexFormat; \ info.filter = obj->MinFilter; \ info.envmode = unit->EnvMode; \ + info.er = 0; \ + info.eg = 0; \ + info.eb = 0; \ \ if (info.envmode == GL_BLEND) { \ /* potential off-by-one error here? (1.0f -> 2048 -> 0) */ \ diff --git a/src/mesa/swrast_setup/ss_tritmp.h b/src/mesa/swrast_setup/ss_tritmp.h index 17f38639563..8484aab5a93 100644 --- a/src/mesa/swrast_setup/ss_tritmp.h +++ b/src/mesa/swrast_setup/ss_tritmp.h @@ -41,9 +41,9 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) GLenum mode = GL_FILL; GLuint facing = 0; GLchan saved_color[3][4]; - GLfloat saved_col0[3][4]; - GLfloat saved_spec[3][4]; - GLfloat saved_index[3]; + GLfloat saved_col0[3][4] = { { 0 } }; + GLfloat saved_spec[3][4] = { { 0 } }; + GLfloat saved_index[3] = { 0 }; v[0] = &verts[e0]; v[1] = &verts[e1]; diff --git a/src/mesa/tnl_dd/t_dd_dmatmp.h b/src/mesa/tnl_dd/t_dd_dmatmp.h index e5885782c7a..2424204b886 100644 --- a/src/mesa/tnl_dd/t_dd_dmatmp.h +++ b/src/mesa/tnl_dd/t_dd_dmatmp.h @@ -241,6 +241,7 @@ static void TAG(render_line_loop_verts)( GLcontext *ctx, tmp = ALLOC_VERTS(nr+1); tmp = TAG(emit_verts)( ctx, j, nr, tmp ); tmp = TAG(emit_verts)( ctx, start, 1, tmp ); + (void) tmp; } else { TAG(emit_verts)( ctx, j, nr, ALLOC_VERTS(nr) ); @@ -254,6 +255,7 @@ static void TAG(render_line_loop_verts)( GLcontext *ctx, tmp = ALLOC_VERTS(2); tmp = TAG(emit_verts)( ctx, start+1, 1, tmp ); tmp = TAG(emit_verts)( ctx, start, 1, tmp ); + (void) tmp; } FLUSH(); @@ -358,6 +360,7 @@ static void TAG(render_tri_fan_verts)( GLcontext *ctx, tmp = ALLOC_VERTS( nr ); tmp = TAG(emit_verts)( ctx, start, 1, tmp ); tmp = TAG(emit_verts)( ctx, j, nr - 1, tmp ); + (void) tmp; currentsz = dmasz; } @@ -397,6 +400,7 @@ static void TAG(render_poly_verts)( GLcontext *ctx, tmp = ALLOC_VERTS( nr ); tmp = TAG(emit_verts)( ctx, start, 1, tmp ); tmp = TAG(emit_verts)( ctx, j, nr - 1, tmp ); + (void) tmp; currentsz = dmasz; } @@ -634,6 +638,7 @@ static void TAG(render_quads_verts)( GLcontext *ctx, /* Send v1, v2, v3 */ tmp = EMIT_VERTS(ctx, j + 1, 3, tmp); + (void) tmp; } } else { @@ -820,6 +825,7 @@ static void TAG(render_line_loop_elts)( GLcontext *ctx, tmp = ALLOC_ELTS(nr+1); tmp = TAG(emit_elts)( ctx, elts+j, nr, tmp ); tmp = TAG(emit_elts)( ctx, elts+start, 1, tmp ); + (void) tmp; } else { TAG(emit_elts)( ctx, elts+j, nr, ALLOC_ELTS(nr) ); @@ -833,6 +839,7 @@ static void TAG(render_line_loop_elts)( GLcontext *ctx, tmp = ALLOC_ELTS(2); tmp = TAG(emit_elts)( ctx, elts+start+1, 1, tmp ); tmp = TAG(emit_elts)( ctx, elts+start, 1, tmp ); + (void) tmp; } FLUSH(); @@ -946,6 +953,7 @@ static void TAG(render_tri_fan_elts)( GLcontext *ctx, tmp = ALLOC_ELTS( nr ); tmp = TAG(emit_elts)( ctx, elts+start, 1, tmp ); tmp = TAG(emit_elts)( ctx, elts+j, nr - 1, tmp ); + (void) tmp; FLUSH(); currentsz = dmasz; } @@ -983,6 +991,7 @@ static void TAG(render_poly_elts)( GLcontext *ctx, tmp = ALLOC_ELTS( nr ); tmp = TAG(emit_elts)( ctx, elts+start, 1, tmp ); tmp = TAG(emit_elts)( ctx, elts+j, nr - 1, tmp ); + (void) tmp; FLUSH(); currentsz = dmasz; } diff --git a/src/mesa/tnl_dd/t_dd_dmatmp2.h b/src/mesa/tnl_dd/t_dd_dmatmp2.h index 2380c49fdf4..cd225b6343e 100644 --- a/src/mesa/tnl_dd/t_dd_dmatmp2.h +++ b/src/mesa/tnl_dd/t_dd_dmatmp2.h @@ -266,6 +266,7 @@ static void TAG(render_line_loop_verts)( GLcontext *ctx, if (j + nr < count) { ELT_TYPE *dest = ALLOC_ELTS( nr ); dest = TAG(emit_consecutive_elts)( ctx, dest, j, nr ); + (void) dest; j += nr - 1; CLOSE_ELTS(); } @@ -273,6 +274,7 @@ static void TAG(render_line_loop_verts)( GLcontext *ctx, ELT_TYPE *dest = ALLOC_ELTS( nr + 1 ); dest = TAG(emit_consecutive_elts)( ctx, dest, j, nr ); dest = TAG(emit_consecutive_elts)( ctx, dest, start, 1 ); + (void) dest; j += nr; CLOSE_ELTS(); } @@ -554,6 +556,7 @@ static void TAG(render_points_elts)( GLcontext *ctx, nr = MIN2( dmasz, count - j ); dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } } @@ -590,6 +593,7 @@ static void TAG(render_lines_elts)( GLcontext *ctx, nr = MIN2( dmasz, count - j ); dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } @@ -621,6 +625,7 @@ static void TAG(render_line_strip_elts)( GLcontext *ctx, nr = MIN2( dmasz, count - j ); dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } } @@ -671,6 +676,7 @@ static void TAG(render_line_loop_elts)( GLcontext *ctx, j += nr - 1; if (j + 1 >= count && (flags & PRIM_END)) { dest = TAG(emit_elts)( ctx, dest, elts+start, 1 ); + (void) dest; } CLOSE_ELTS(); } @@ -703,6 +709,7 @@ static void TAG(render_triangles_elts)( GLcontext *ctx, nr = MIN2( dmasz, count - j ); dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } } @@ -734,6 +741,7 @@ static void TAG(render_tri_strip_elts)( GLcontext *ctx, dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } } @@ -759,6 +767,7 @@ static void TAG(render_tri_fan_elts)( GLcontext *ctx, dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+start, 1 ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr - 1 ); + (void) dest; CLOSE_ELTS(); } } @@ -785,6 +794,7 @@ static void TAG(render_poly_elts)( GLcontext *ctx, dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+start, 1 ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr - 1 ); + (void) dest; CLOSE_ELTS(); } } @@ -843,6 +853,7 @@ static void TAG(render_quad_strip_elts)( GLcontext *ctx, nr = MIN2( dmasz, count - j ); dest = ALLOC_ELTS( nr ); dest = TAG(emit_elts)( ctx, dest, elts+j, nr ); + (void) dest; CLOSE_ELTS(); } } diff --git a/src/mesa/tnl_dd/t_dd_tritmp.h b/src/mesa/tnl_dd/t_dd_tritmp.h index 8574fe618b5..6acd837317f 100644 --- a/src/mesa/tnl_dd/t_dd_tritmp.h +++ b/src/mesa/tnl_dd/t_dd_tritmp.h @@ -132,7 +132,7 @@ static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 ) struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb; VERTEX *v[3]; GLfloat offset = 0; - GLfloat z[3]; + GLfloat z[3] = { 0 }; GLenum mode = GL_FILL; GLuint facing = 0; LOCAL_VARS(3); @@ -395,7 +395,7 @@ static void TAG(quadr)( GLcontext *ctx, struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb; VERTEX *v[4]; GLfloat offset = 0; - GLfloat z[4]; + GLfloat z[4] = { 0 }; GLenum mode = GL_FILL; GLuint facing = 0; LOCAL_VARS(4); diff --git a/src/mesa/vbo/vbo_exec_eval.c b/src/mesa/vbo/vbo_exec_eval.c index 0c691b3a5cd..a7846213d0c 100644 --- a/src/mesa/vbo/vbo_exec_eval.c +++ b/src/mesa/vbo/vbo_exec_eval.c @@ -35,17 +35,20 @@ static void clear_active_eval1( struct vbo_exec_context *exec, GLuint attr ) { + assert(attr < Elements(exec->eval.map1)); exec->eval.map1[attr].map = NULL; } static void clear_active_eval2( struct vbo_exec_context *exec, GLuint attr ) { + assert(attr < Elements(exec->eval.map2)); exec->eval.map2[attr].map = NULL; } static void set_active_eval1( struct vbo_exec_context *exec, GLuint attr, GLuint dim, struct gl_1d_map *map ) { + assert(attr < Elements(exec->eval.map1)); if (!exec->eval.map1[attr].map) { exec->eval.map1[attr].map = map; exec->eval.map1[attr].sz = dim; @@ -55,6 +58,7 @@ static void set_active_eval1( struct vbo_exec_context *exec, GLuint attr, GLuint static void set_active_eval2( struct vbo_exec_context *exec, GLuint attr, GLuint dim, struct gl_2d_map *map ) { + assert(attr < Elements(exec->eval.map2)); if (!exec->eval.map2[attr].map) { exec->eval.map2[attr].map = map; exec->eval.map2[attr].sz = dim; @@ -73,18 +77,6 @@ void vbo_exec_eval_update( struct vbo_exec_context *exec ) clear_active_eval2( exec, attr ); } - /* _NEW_PROGRAM */ - if (ctx->VertexProgram._Enabled) { - for (attr = 0; attr < VBO_ATTRIB_FIRST_MATERIAL; attr++) { - /* _NEW_EVAL */ - if (ctx->Eval.Map1Attrib[attr]) - set_active_eval1( exec, attr, 4, &ctx->EvalMap.Map1Attrib[attr] ); - - if (ctx->Eval.Map2Attrib[attr]) - set_active_eval2( exec, attr, 4, &ctx->EvalMap.Map2Attrib[attr] ); - } - } - if (ctx->Eval.Map1Color4) set_active_eval1( exec, VBO_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 ); @@ -125,6 +117,23 @@ void vbo_exec_eval_update( struct vbo_exec_context *exec ) else if (ctx->Eval.Map2Vertex3) set_active_eval2( exec, VBO_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 ); + /* _NEW_PROGRAM */ + if (ctx->VertexProgram._Enabled) { + /* These are the 16 evaluators which GL_NV_vertex_program defines. + * They alias and override the conventional vertex attributs. + */ + for (attr = 0; attr < 16; attr++) { + /* _NEW_EVAL */ + assert(attr < Elements(ctx->Eval.Map1Attrib)); + if (ctx->Eval.Map1Attrib[attr]) + set_active_eval1( exec, attr, 4, &ctx->EvalMap.Map1Attrib[attr] ); + + assert(attr < Elements(ctx->Eval.Map2Attrib)); + if (ctx->Eval.Map2Attrib[attr]) + set_active_eval2( exec, attr, 4, &ctx->EvalMap.Map2Attrib[attr] ); + } + } + exec->eval.recalculate_maps = 0; } diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 5aedf5b04be..647be995c12 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -7,10 +7,12 @@ #define DISASSEM 0 #define X86_TWOB 0x0f +#if 0 static unsigned char *cptr( void (*label)() ) { return (unsigned char *)(unsigned long)label; } +#endif static void do_realloc( struct x86_function *p ) |