diff options
92 files changed, 3127 insertions, 604 deletions
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 d09ab925656..ba1357697d1 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[] = @@ -149,6 +150,27 @@ 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( @@ -273,6 +295,50 @@ 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 ) @@ -492,6 +558,7 @@ 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; @@ -546,6 +613,7 @@ 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..717358620c5 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; 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 a5d2db04ec1..5f5c95bfbdb 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; @@ -160,6 +161,19 @@ 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 ); @@ -212,6 +226,7 @@ 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 || @@ -235,6 +250,8 @@ 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 8a7ee0c7e4f..a1e8a4f6bb7 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 */ @@ -61,8 +60,13 @@ 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..c9050ca864c 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -126,7 +126,8 @@ 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 +421,9 @@ 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..895af2c8d00 --- /dev/null +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -0,0 +1,718 @@ +/************************************************************************** + * + * 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_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 = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_Z) != 0; + is_stencil = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_S) != 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 (pf_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_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_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_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 83263d9fe64..69ff3b9dd9e 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -46,6 +46,7 @@ #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" @@ -1317,7 +1318,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D); /* vertex data that doesn't change */ for (i = 0; i < 4; i++) { @@ -1383,59 +1384,9 @@ 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 */ 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_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/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_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/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..3020806c5d6 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -133,6 +133,9 @@ 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); @@ -202,7 +205,7 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, unsigned src_offset = src->offset + sy * src_pitch + sx * pf_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); @@ -250,7 +253,7 @@ 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 +318,7 @@ 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/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 8d49366dfcb..6f44b1c7fe1 100644 --- a/src/gallium/drivers/nv30/nv30_context.h +++ b/src/gallium/drivers/nv30/nv30_context.h @@ -184,6 +184,7 @@ 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_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/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 a3d594167aa..cf33b64a86d 100644 --- a/src/gallium/drivers/nv40/nv40_context.h +++ b/src/gallium/drivers/nv40/nv40_context.h @@ -204,6 +204,7 @@ 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_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c index 198692965dc..ba0fbcb26a9 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/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..9e083b662dd 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -55,6 +55,20 @@ 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) { @@ -130,6 +144,8 @@ 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 +185,10 @@ 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 +202,10 @@ 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 +221,12 @@ 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 = pf_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..e496cf4cad8 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -119,7 +119,7 @@ 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; @@ -131,6 +131,9 @@ 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; @@ -150,6 +153,20 @@ 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) { @@ -342,25 +359,34 @@ 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 +394,11 @@ 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 +490,24 @@ 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 @@ -644,7 +672,7 @@ 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 +687,9 @@ 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 +704,17 @@ 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 +834,33 @@ 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 +964,6 @@ 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 +1032,6 @@ 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; @@ -1257,9 +1321,68 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src) emit(pc, e); } +static struct nv50_program_exec * +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); + return pc->p->exec_tail; +} + +#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 +1399,10 @@ 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 +1411,214 @@ 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. + */ + emit_quadop(pc, NULL, 0, 0, tlod, tlod, 0x55); + emit_branch(pc, 0, 2, &join_at)->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, NULL)->param.index = target; + } + + emit_mov(pc, tlod, src); /* target */ + emit(pc, tex); /* texlod */ + + join_at->param.index = target + 2 * 2; + emit_nop(pc); + pc->p->exec_tail->inst[1] |= 2; /* 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++]); @@ -1389,38 +1641,6 @@ 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); @@ -1515,8 +1735,6 @@ 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_DDY: case TGSI_OPCODE_DP3: @@ -1526,29 +1744,14 @@ 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 +1779,13 @@ 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 +1794,17 @@ 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 +1887,7 @@ 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 +1941,8 @@ 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 +1998,8 @@ 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; @@ -2187,11 +2402,19 @@ 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 +2468,9 @@ 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 +2753,10 @@ 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: @@ -3024,7 +3236,7 @@ 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 +3284,8 @@ 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..120aa6f362b 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -68,6 +68,7 @@ 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), @@ -85,7 +86,7 @@ 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; @@ -96,7 +97,7 @@ 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); @@ -140,48 +141,78 @@ 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); + p * (32 * 8 * 4), nv50->miptree_nr[p] * 8 * 4); - for (i = 0, unit = 0; unit < nv50->miptree_nr; ++unit) { - struct nv50_miptree *mt = nv50->miptree[unit]; - - 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); + } + + nv50->state.miptree_nr[p] = nv50->miptree_nr[p]; + return TRUE; +} - 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); +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); + + if (nv50_validate_textures(nv50, so, PIPE_SHADER_VERTEX) == FALSE || + nv50_validate_textures(nv50, so, PIPE_SHADER_FRAGMENT) == FALSE) { + so_ref(NULL, &so); - 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); + 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 +220,4 @@ 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..6240a0c757a 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -16,6 +16,7 @@ struct nv50_transfer { int level_depth; int level_x; int level_y; + int level_z; unsigned nblocksx; unsigned nblocksy; }; @@ -24,10 +25,10 @@ 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 +57,7 @@ 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 +76,13 @@ 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 +119,6 @@ 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, @@ -166,6 +153,7 @@ 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_z = zslice; tx->level_x = pf_get_nblocksx(pt->format, x); tx->level_y = pf_get_nblocksy(pt->format, y); ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, @@ -175,23 +163,18 @@ 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); 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, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, @@ -216,11 +199,11 @@ nv50_transfer_del(struct pipe_transfer *ptx) 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, @@ -282,7 +265,7 @@ 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/r300/Makefile b/src/gallium/drivers/r300/Makefile index 9c9fc6f64b3..8cfd4147c20 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 \ 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..6bd2766730f 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" @@ -248,6 +250,8 @@ 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_emit.c b/src/gallium/drivers/r300/r300_emit.c index f784e1fa8e5..55c8aa07bdf 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -41,9 +41,16 @@ 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; @@ -335,7 +342,7 @@ void r300_emit_fb_state(struct r300_context* r300, assert(fb->nr_cbufs <= 4); BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) + - (fb->zsbuf ? 10 : 0) + 4); + (fb->zsbuf ? 10 : 0) + 6); /* Flush and free renderbuffer caches. */ OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, @@ -345,6 +352,9 @@ 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]; @@ -873,10 +883,21 @@ 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; } diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index c1ea87d11e9..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) diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 35b335df6a1..4b210f72db2 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -82,8 +82,9 @@ static void r300_emit_draw_arrays(struct r300_context *r300, { CS_LOCALS(r300); - BEGIN_CS(4); - OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); + BEGIN_CS(6); + 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 +109,8 @@ static void r300_emit_draw_elements(struct r300_context *r300, assert((start * indexSize) % 4 == 0); assert(offset_dwords == 0); - BEGIN_CS(10); + BEGIN_CS(12); + 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) { diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index c0d9797020d..feb571a23dd 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -220,12 +220,18 @@ 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. */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index edf7114bbb5..91cf972edee 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -339,6 +339,7 @@ static void r300->dirty_state |= R300_NEW_SCISSOR; } r300->dirty_state |= R300_NEW_FRAMEBUFFERS; + r300->dirty_state |= R300_NEW_BLEND; } /* Create fragment shader state. */ @@ -626,8 +627,6 @@ 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_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/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_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 588ca5e026d..79f3d3f0566 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 */ }; @@ -151,6 +153,22 @@ 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/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 2a8daed051d..76acc99ad73 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -81,6 +81,7 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws, domain |= RADEON_GEM_DOMAIN_GTT; } + radeon_buffer->ws = radeon_ws; radeon_buffer->bo = radeon_bo_open(radeon_ws->priv->bom, 0, size, alignment, domain, 0); if (radeon_buffer->bo == NULL) { @@ -131,6 +132,11 @@ static void radeon_buffer_del(struct pipe_buffer *buffer) { struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer; + struct radeon_winsys_priv *priv = radeon_buffer->ws->priv; + + if (radeon_bo_is_referenced_by_cs(radeon_buffer->bo, priv->cs)) { + priv->cs->space_flush_fn(priv->cs->space_flush_data); + } radeon_bo_unref(radeon_buffer->bo); free(radeon_buffer); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h index bfe2221d1ed..1e91e18927a 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h @@ -50,6 +50,7 @@ struct radeon_pipe_buffer { struct pipe_buffer base; struct radeon_bo *bo; + struct radeon_winsys *ws; boolean flinked; uint32_t flink; }; diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c index dec7c065036..2f7fbc72423 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c @@ -171,6 +171,7 @@ struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api, radeon_buffer->base.screen = screen; radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL; radeon_buffer->bo = bo; + radeon_buffer->ws = (struct radeon_winsys*)screen->winsys; return &radeon_buffer->base; } diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c index 10e1b3c9124..ca6dd3bcf8e 100644 --- a/src/mesa/drivers/dri/r300/r300_blit.c +++ b/src/mesa/drivers/dri/r300/r300_blit.c @@ -141,7 +141,7 @@ static void r300_emit_tx_setup(struct r300_context *r300, assert(width <= 2048); assert(height <= 2048); - assert(r300TranslateTexFormat(mesa_format) != 0); + assert(r300TranslateTexFormat(mesa_format) >= 0); assert(offset % 32 == 0); BEGIN_BATCH(17); @@ -315,8 +315,8 @@ static void r300_emit_rs_setup(struct r300_context *r300) OUT_BATCH_REGVAL(R300_RS_IP_0, R300_RS_TEX_PTR(0) | R300_RS_SEL_S(R300_RS_SEL_C0) | - R300_RS_SEL_R(R300_RS_SEL_C1) | - R300_RS_SEL_T(R300_RS_SEL_K0) | + 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(); } @@ -428,6 +428,7 @@ 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) { @@ -448,7 +449,7 @@ static void emit_cb_setup(struct r300_context *r300, r300_emit_cb_setup(r300, bo, offset, mesa_format, _mesa_get_format_bytes(mesa_format), - _mesa_format_row_stride(mesa_format, width)); + _mesa_format_row_stride(mesa_format, pitch)); BEGIN_BATCH_NO_AUTOSTATE(5); OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); @@ -468,18 +469,14 @@ GLboolean r300_blit(struct r300_context *r300, struct radeon_bo *dst_bo, intptr_t dst_offset, gl_format dst_mesaformat, + unsigned dst_pitch, unsigned dst_width, unsigned dst_height) { - //assert(src_width == dst_width); - //assert(src_height == dst_height); - if (src_bo == dst_bo) { return GL_FALSE; } - //return GL_FALSE; - if (0) { fprintf(stderr, "src: width %d, height %d, pitch %d vs %d, format %s\n", src_width, src_height, src_pitch, @@ -511,14 +508,13 @@ GLboolean r300_blit(struct r300_context *r300, emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2); emit_vap_setup(r300, dst_width, dst_height); - emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_width, dst_height); + emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height); emit_draw_packet(r300, dst_width, dst_height); r300EmitCacheFlush(r300); radeonFlush(r300->radeon.glCtx); - //r300ResetHwState(r300); 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 index 29c5aa95149..28ffd4ea421 100644 --- a/src/mesa/drivers/dri/r300/r300_blit.h +++ b/src/mesa/drivers/dri/r300/r300_blit.h @@ -40,6 +40,7 @@ GLboolean r300_blit(struct r300_context *r300, struct radeon_bo *dst_bo, intptr_t dst_offset, gl_format dst_mesaformat, + unsigned dst_pitch, unsigned dst_width, unsigned dst_height); diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 05005f61c30..3c6ec2a34a8 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -486,7 +486,6 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, r300_init_vtbl(&r300->radeon); _mesa_init_driver_functions(&functions); - r300_init_texcopy_functions(&functions); r300InitIoctlFuncs(&functions); r300InitStateFuncs(&functions); r300InitTextureFuncs(&functions); @@ -494,6 +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)) { diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h index beb10072e9c..6ede0fe25c9 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.h +++ b/src/mesa/drivers/dri/r300/r300_tex.h @@ -51,6 +51,6 @@ extern GLboolean r300ValidateBuffers(GLcontext * ctx); extern void r300InitTextureFuncs(struct dd_function_table *functions); -uint32_t r300TranslateTexFormat(gl_format mesaFormat); +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 index 5e3a724d4e6..7702a1d67df 100644 --- a/src/mesa/drivers/dri/r300/r300_texcopy.c +++ b/src/mesa/drivers/dri/r300/r300_texcopy.c @@ -63,7 +63,6 @@ do_copy_texsubimage(GLcontext *ctx, assert(timg->mt->bo); assert(timg->base.Width >= dstx + width); assert(timg->base.Height >= dsty + height); - //assert(tobj->mt == timg->mt); intptr_t src_offset = rrb->draw_offset + x * rrb->cpp + y * rrb->pitch; intptr_t dst_offset = radeon_miptree_image_offset(timg->mt, _mesa_tex_target_to_face(target), level); @@ -87,7 +86,7 @@ do_copy_texsubimage(GLcontext *ctx, /* blit from src buffer to texture */ return r300_blit(r300, rrb->bo, src_offset, rrb->base.Format, rrb->pitch, rrb->base.Width, rrb->base.Height, timg->mt->bo ? timg->mt->bo : timg->bo, dst_offset, - timg->base.TexFormat, width, height); + timg->base.TexFormat, timg->base.Width, width, height); } static void diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 6db56ba618f..d4a728381e4 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -59,7 +59,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * identically. -- paulus */ -uint32_t r300TranslateTexFormat(gl_format mesaFormat) +int32_t r300TranslateTexFormat(gl_format mesaFormat) { switch (mesaFormat) { @@ -168,7 +168,7 @@ uint32_t r300TranslateTexFormat(gl_format mesaFormat) case MESA_FORMAT_SRGBA_DXT5: return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5) | R300_TX_FORMAT_GAMMA; default: - return 0; + return -1; } }; @@ -252,12 +252,13 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) { r300SetDepthTexMode(&t->base); } else { - t->pp_txformat = r300TranslateTexFormat(firstImage->TexFormat); - if (t->pp_txformat == 0) { + 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; } } diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index d27a3245a39..5e1504872d6 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -52,6 +52,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "radeon_reg.h" +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, 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..b66fe78ac3b 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -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 e84f5245253..e10b23b97f1 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -539,12 +539,15 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 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; } @@ -1412,43 +1415,65 @@ 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) + { + 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 { - 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 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; } } @@ -1493,8 +1518,17 @@ 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) { @@ -1513,7 +1547,6 @@ 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 @@ -4331,13 +4364,20 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) } - if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB) + switch(pAsm->pILInst[pAsm->uiCurInst].Opcode) { - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; - } - else - { - 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; @@ -5650,7 +5690,8 @@ GLboolean AssembleInstr(GLuint uiFirstInst, } } break; - + case OPCODE_DDX: + case OPCODE_DDY: case OPCODE_TEX: case OPCODE_TXB: case OPCODE_TXP: diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index 6ef945dfda3..dbd9860f7d0 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -485,6 +485,8 @@ typedef struct r700_AssemblerBase GLuint unNumPresub; GLuint unCurNumILInsts; + GLuint unVetTexBits; + } r700_AssemblerBase; //Internal use diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index ee2a0a4c8a6..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); diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index d15f0137107..ca0710b6813 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) ) { @@ -479,6 +540,21 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } + 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); + //SETbit(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; SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); @@ -498,6 +574,10 @@ 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) { @@ -575,6 +655,22 @@ 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++) { 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/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_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); } |