diff options
Diffstat (limited to 'src/gallium/drivers')
43 files changed, 1977 insertions, 1715 deletions
diff --git a/src/gallium/drivers/i915/Makefile b/src/gallium/drivers/i915/Makefile index 778124728bb..36197fbc93b 100644 --- a/src/gallium/drivers/i915/Makefile +++ b/src/gallium/drivers/i915/Makefile @@ -27,6 +27,7 @@ C_SOURCES = \ i915_resource_buffer.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ + i915_fpc_optimize.c \ i915_surface.c include ../../Makefile.template diff --git a/src/gallium/drivers/i915/SConscript b/src/gallium/drivers/i915/SConscript index 98370601b7f..76f597001fe 100644 --- a/src/gallium/drivers/i915/SConscript +++ b/src/gallium/drivers/i915/SConscript @@ -14,6 +14,7 @@ i915 = env.ConvenienceLibrary( 'i915_flush.c', 'i915_fpc_emit.c', 'i915_fpc_translate.c', + 'i915_fpc_optimize.c', 'i915_prim_emit.c', 'i915_prim_vbuf.c', 'i915_query.c', diff --git a/src/gallium/drivers/i915/i915_batch.h b/src/gallium/drivers/i915/i915_batch.h index ce2691b2fd7..a1f8bcae802 100644 --- a/src/gallium/drivers/i915/i915_batch.h +++ b/src/gallium/drivers/i915/i915_batch.h @@ -29,6 +29,7 @@ #define I915_BATCH_H #include "i915_batchbuffer.h" +#include "i915_context.h" #define BEGIN_BATCH(dwords) \ @@ -49,11 +50,26 @@ #define FLUSH_BATCH(fence) \ i915_flush(i915, fence) - /************************************************************************ * i915_flush.c */ void i915_flush(struct i915_context *i915, struct pipe_fence_handle **fence); +/* + * Flush if the current color buf is idle and we have more than 256 vertices + * queued, or if the current color buf is busy and we have more than 4096 + * vertices queued. + */ +static INLINE void i915_flush_heuristically(struct i915_context* i915, + int num_vertex) +{ + struct i915_winsys *iws = i915->iws; + i915->vertices_since_last_flush += num_vertex; + if ( i915->vertices_since_last_flush > 4096 + || ( i915->vertices_since_last_flush > 256 && + !iws->buffer_is_busy(iws, i915->current.cbuf_bo)) ) + FLUSH_BATCH(NULL); +} + #endif diff --git a/src/gallium/drivers/i915/i915_clear.c b/src/gallium/drivers/i915/i915_clear.c index fcb208d6dae..e1d6a749cdc 100644 --- a/src/gallium/drivers/i915/i915_clear.c +++ b/src/gallium/drivers/i915/i915_clear.c @@ -120,6 +120,11 @@ i915_clear_emit(struct pipe_context *pipe, unsigned buffers, const float *rgba, OUT_BATCH_F(desty + height); OUT_BATCH_F(destx); OUT_BATCH_F(desty); + + /* Flush after clear, its expected to be a costly operation. + * This is not required, just a heuristic + */ + FLUSH_BATCH(NULL); } /** diff --git a/src/gallium/drivers/i915/i915_context.h b/src/gallium/drivers/i915/i915_context.h index c964208fedd..84862351ffe 100644 --- a/src/gallium/drivers/i915/i915_context.h +++ b/src/gallium/drivers/i915/i915_context.h @@ -264,6 +264,8 @@ struct i915_context { struct util_slab_mempool transfer_pool; struct util_slab_mempool texture_transfer_pool; + int vertices_since_last_flush; + /** blitter/hw-clear */ struct blitter_context* blitter; diff --git a/src/gallium/drivers/i915/i915_flush.c b/src/gallium/drivers/i915/i915_flush.c index b4e81147c4f..6d76afa9dbc 100644 --- a/src/gallium/drivers/i915/i915_flush.c +++ b/src/gallium/drivers/i915/i915_flush.c @@ -77,4 +77,5 @@ void i915_flush(struct i915_context *i915, struct pipe_fence_handle **fence) i915->static_dirty = ~0; /* kernel emits flushes in between batchbuffers */ i915->flush_dirty = 0; + i915->vertices_since_last_flush = 0; } diff --git a/src/gallium/drivers/i915/i915_fpc.h b/src/gallium/drivers/i915/i915_fpc.h index 509395cf1f5..b760bc461a1 100644 --- a/src/gallium/drivers/i915/i915_fpc.h +++ b/src/gallium/drivers/i915/i915_fpc.h @@ -33,7 +33,9 @@ #include "i915_context.h" #include "i915_reg.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" #define I915_PROGRAM_SIZE 192 @@ -207,4 +209,90 @@ extern void i915_program_error(struct i915_fp_compile *p, const char *msg, ...); +/*====================================================================== + * i915_fpc_optimize.c + */ + + +struct i915_src_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned SwizzleX : 3; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 3; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 3; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 3; /* TGSI_SWIZZLE_ */ + unsigned Absolute : 1; /* BOOL */ + unsigned Negate : 1; /* BOOL */ +}; + +/* Additional swizzle supported in i915 */ +#define TGSI_SWIZZLE_ZERO 4 +#define TGSI_SWIZZLE_ONE 5 + +struct i915_dst_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 6; +}; + + +struct i915_full_dst_register +{ + struct i915_dst_register Register; +/* + struct tgsi_src_register Indirect; + struct tgsi_dimension Dimension; + struct tgsi_src_register DimIndirect; +*/ +}; + +struct i915_full_src_register +{ + struct i915_src_register Register; +/* + struct tgsi_src_register Indirect; + struct tgsi_dimension Dimension; + struct tgsi_src_register DimIndirect; +*/ +}; + +struct i915_full_instruction +{ + struct tgsi_instruction Instruction; +/* + struct tgsi_instruction_predicate Predicate; + struct tgsi_instruction_label Label; +*/ + struct tgsi_instruction_texture Texture; + struct i915_full_dst_register Dst[1]; + struct i915_full_src_register Src[3]; +}; + + +union i915_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct i915_full_instruction FullInstruction; + struct tgsi_full_property FullProperty; +}; + +struct i915_token_list +{ + union i915_full_token* Tokens; + unsigned NumTokens; +}; + +extern struct i915_token_list* i915_optimize(const struct tgsi_token *tokens); + +extern void i915_optimize_free(struct i915_token_list* tokens); + #endif diff --git a/src/gallium/drivers/i915/i915_fpc_emit.c b/src/gallium/drivers/i915/i915_fpc_emit.c index d28595e0fd3..c4a42df7882 100644 --- a/src/gallium/drivers/i915/i915_fpc_emit.c +++ b/src/gallium/drivers/i915/i915_fpc_emit.c @@ -369,7 +369,6 @@ i915_emit_const4f(struct i915_fp_compile * p, // XXX emit swizzle here for 0, 1, -1 and any combination thereof // we can use swizzle + neg for that - printf("const %f %f %f %f\n",c0,c1,c2,c3); for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (ifs->constant_flags[reg] == 0xf && ifs->constants[reg][0] == c0 && diff --git a/src/gallium/drivers/i915/i915_fpc_optimize.c b/src/gallium/drivers/i915/i915_fpc_optimize.c new file mode 100644 index 00000000000..2b739e9ccb8 --- /dev/null +++ b/src/gallium/drivers/i915/i915_fpc_optimize.c @@ -0,0 +1,259 @@ +/************************************************************************** + * + * Copyright 2011 The Chromium OS authors. + * 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 GOOGLE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_fpc.h" + +#include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "util/u_string.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" + +static boolean same_dst_reg(struct i915_full_dst_register* d1, struct i915_full_dst_register* d2) +{ + return (d1->Register.File == d2->Register.File && + d1->Register.Indirect == d2->Register.Indirect && + d1->Register.Dimension == d2->Register.Dimension && + d1->Register.Index == d2->Register.Index); +} + +static boolean same_src_reg(struct i915_full_src_register* d1, struct i915_full_src_register* d2) +{ + return (d1->Register.File == d2->Register.File && + d1->Register.Indirect == d2->Register.Indirect && + d1->Register.Dimension == d2->Register.Dimension && + d1->Register.Index == d2->Register.Index && + d1->Register.Absolute == d2->Register.Absolute && + d1->Register.Negate == d2->Register.Negate); +} + +static boolean is_unswizzled(struct i915_full_src_register* r, + unsigned write_mask) +{ + if ( write_mask & TGSI_WRITEMASK_X && r->Register.SwizzleX != TGSI_SWIZZLE_X) + return FALSE; + if ( write_mask & TGSI_WRITEMASK_Y && r->Register.SwizzleY != TGSI_SWIZZLE_Y) + return FALSE; + if ( write_mask & TGSI_WRITEMASK_Z && r->Register.SwizzleZ != TGSI_SWIZZLE_Z) + return FALSE; + if ( write_mask & TGSI_WRITEMASK_W && r->Register.SwizzleW != TGSI_SWIZZLE_W) + return FALSE; + return TRUE; +} + +static boolean op_commutes(unsigned opcode) +{ + if (opcode == TGSI_OPCODE_ADD) return TRUE; + if (opcode == TGSI_OPCODE_MUL) return TRUE; + return FALSE; +} + +static unsigned op_neutral_element(unsigned opcode) +{ + if (opcode == TGSI_OPCODE_ADD) + return TGSI_SWIZZLE_ZERO; + if (opcode == TGSI_OPCODE_MUL) + return TGSI_SWIZZLE_ONE; + + debug_printf("Unknown opcode %d\n",opcode); + return TGSI_SWIZZLE_ZERO; +} + +/* + * Sets the swizzle to the neutral element for the operation for the bits + * of writemask which are set, swizzle to identity otherwise. + */ +static void set_neutral_element_swizzle(struct i915_full_src_register* r, + unsigned write_mask, + unsigned neutral) +{ + if ( write_mask & TGSI_WRITEMASK_X ) + r->Register.SwizzleX = neutral; + else + r->Register.SwizzleX = TGSI_SWIZZLE_X; + + if ( write_mask & TGSI_WRITEMASK_Y ) + r->Register.SwizzleY = neutral; + else + r->Register.SwizzleY = TGSI_SWIZZLE_Y; + + if ( write_mask & TGSI_WRITEMASK_Z ) + r->Register.SwizzleZ = neutral; + else + r->Register.SwizzleZ = TGSI_SWIZZLE_Z; + + if ( write_mask & TGSI_WRITEMASK_W ) + r->Register.SwizzleW = neutral; + else + r->Register.SwizzleW = TGSI_SWIZZLE_W; +} + +/* + * Optimize away things like: + * MUL OUT[0].xyz, TEMP[1], TEMP[2] + * MOV OUT[0].w, TEMP[2] + * into: + * MUL OUT[0].xyzw, TEMP[1].xyz1, TEMP[2] + * This is useful for optimizing texenv. + */ +static void i915_fpc_optimize_mov_after_alu(union i915_full_token* current, union i915_full_token* next) +{ + if ( current->Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION && + next->Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION && + op_commutes(current->FullInstruction.Instruction.Opcode) && + current->FullInstruction.Instruction.Saturate == next->FullInstruction.Instruction.Saturate && + next->FullInstruction.Instruction.Opcode == TGSI_OPCODE_MOV && + same_dst_reg(&next->FullInstruction.Dst[0], &next->FullInstruction.Dst[0]) && + same_src_reg(&next->FullInstruction.Src[0], ¤t->FullInstruction.Src[1]) && + is_unswizzled(¤t->FullInstruction.Src[0], current->FullInstruction.Dst[0].Register.WriteMask) && + is_unswizzled(¤t->FullInstruction.Src[1], current->FullInstruction.Dst[0].Register.WriteMask) && + is_unswizzled(&next->FullInstruction.Src[0], next->FullInstruction.Dst[0].Register.WriteMask) ) + { + next->FullInstruction.Instruction.Opcode = TGSI_OPCODE_NOP; + + set_neutral_element_swizzle(¤t->FullInstruction.Src[1], 0, 0); + set_neutral_element_swizzle(¤t->FullInstruction.Src[0], + next->FullInstruction.Dst[0].Register.WriteMask, + op_neutral_element(current->FullInstruction.Instruction.Opcode)); + + current->FullInstruction.Dst[0].Register.WriteMask = current->FullInstruction.Dst[0].Register.WriteMask | + next->FullInstruction.Dst[0].Register.WriteMask; + return; + } + + if ( current->Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION && + next->Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION && + op_commutes(current->FullInstruction.Instruction.Opcode) && + current->FullInstruction.Instruction.Saturate == next->FullInstruction.Instruction.Saturate && + next->FullInstruction.Instruction.Opcode == TGSI_OPCODE_MOV && + same_dst_reg(&next->FullInstruction.Dst[0], &next->FullInstruction.Dst[0]) && + same_src_reg(&next->FullInstruction.Src[0], ¤t->FullInstruction.Src[0]) && + is_unswizzled(¤t->FullInstruction.Src[0], current->FullInstruction.Dst[0].Register.WriteMask) && + is_unswizzled(¤t->FullInstruction.Src[1], current->FullInstruction.Dst[0].Register.WriteMask) && + is_unswizzled(&next->FullInstruction.Src[0], next->FullInstruction.Dst[0].Register.WriteMask) ) + { + next->FullInstruction.Instruction.Opcode = TGSI_OPCODE_NOP; + + set_neutral_element_swizzle(¤t->FullInstruction.Src[0], 0, 0); + set_neutral_element_swizzle(¤t->FullInstruction.Src[1], + next->FullInstruction.Dst[0].Register.WriteMask, + op_neutral_element(current->FullInstruction.Instruction.Opcode)); + + current->FullInstruction.Dst[0].Register.WriteMask = current->FullInstruction.Dst[0].Register.WriteMask | + next->FullInstruction.Dst[0].Register.WriteMask; + return; + } +} + +static void copy_src_reg(struct i915_src_register* o, const struct tgsi_src_register* i) +{ + o->File = i->File; + o->Indirect = i->Indirect; + o->Dimension = i->Dimension; + o->Index = i->Index; + o->SwizzleX = i->SwizzleX; + o->SwizzleY = i->SwizzleY; + o->SwizzleZ = i->SwizzleZ; + o->SwizzleW = i->SwizzleW; + o->Absolute = i->Absolute; + o->Negate = i->Negate; +} + +static void copy_dst_reg(struct i915_dst_register* o, const struct tgsi_dst_register* i) +{ + o->File = i->File; + o->WriteMask = i->WriteMask; + o->Indirect = i->Indirect; + o->Dimension = i->Dimension; + o->Index = i->Index; +} + +static void copy_instruction(struct i915_full_instruction* o, const struct tgsi_full_instruction* i) +{ + memcpy(&o->Instruction, &i->Instruction, sizeof(o->Instruction)); + memcpy(&o->Texture, &i->Texture, sizeof(o->Texture)); + + copy_dst_reg(&o->Dst[0].Register, &i->Dst[0].Register); + + copy_src_reg(&o->Src[0].Register, &i->Src[0].Register); + copy_src_reg(&o->Src[1].Register, &i->Src[1].Register); + copy_src_reg(&o->Src[2].Register, &i->Src[2].Register); +} + +static void copy_token(union i915_full_token* o, union tgsi_full_token* i) +{ + if (i->Token.Type != TGSI_TOKEN_TYPE_INSTRUCTION) + memcpy(o, i, sizeof(*o)); + else + copy_instruction(&o->FullInstruction, &i->FullInstruction); + +} + +struct i915_token_list* i915_optimize(const struct tgsi_token *tokens) +{ + struct i915_token_list *out_tokens = MALLOC(sizeof(struct i915_token_list)); + struct tgsi_parse_context parse; + int i = 0; + + out_tokens->NumTokens = 0; + + /* Count the tokens */ + tgsi_parse_init( &parse, tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + out_tokens->NumTokens++; + } + tgsi_parse_free (&parse); + + /* Allocate our tokens */ + out_tokens->Tokens = MALLOC(sizeof(union i915_full_token) * out_tokens->NumTokens); + + tgsi_parse_init( &parse, tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + copy_token(&out_tokens->Tokens[i] , &parse.FullToken); + + if (i > 0) + i915_fpc_optimize_mov_after_alu(&out_tokens->Tokens[i-1], &out_tokens->Tokens[i]); + + i++; + } + tgsi_parse_free (&parse); + + return out_tokens; +} + +void i915_optimize_free(struct i915_token_list* tokens) +{ + free(tokens->Tokens); + free(tokens); +} + + diff --git a/src/gallium/drivers/i915/i915_fpc_translate.c b/src/gallium/drivers/i915/i915_fpc_translate.c index 0cbd4f2d748..e19d9be04de 100644 --- a/src/gallium/drivers/i915/i915_fpc_translate.c +++ b/src/gallium/drivers/i915/i915_fpc_translate.c @@ -172,7 +172,7 @@ static uint get_mapping(struct i915_fragment_shader* fs, int unit) */ static uint src_vector(struct i915_fp_compile *p, - const struct tgsi_full_src_register *source, + const struct i915_full_src_register *source, struct i915_fragment_shader* fs) { uint index = source->Register.Index; @@ -287,7 +287,7 @@ src_vector(struct i915_fp_compile *p, */ static uint get_result_vector(struct i915_fp_compile *p, - const struct tgsi_full_dst_register *dest) + const struct i915_full_dst_register *dest) { switch (dest->Register.File) { case TGSI_FILE_OUTPUT: @@ -316,7 +316,7 @@ get_result_vector(struct i915_fp_compile *p, * Compute flags for saturation and writemask. */ static uint -get_result_flags(const struct tgsi_full_instruction *inst) +get_result_flags(const struct i915_full_instruction *inst) { const uint writeMask = inst->Dst[0].Register.WriteMask; @@ -378,7 +378,7 @@ translate_tex_src_target(struct i915_fp_compile *p, uint tex) */ static void emit_tex(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, + const struct i915_full_instruction *inst, uint opcode, struct i915_fragment_shader* fs) { @@ -404,7 +404,7 @@ emit_tex(struct i915_fp_compile *p, */ static void emit_simple_arith(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, + const struct i915_full_instruction *inst, uint opcode, uint numArgs, struct i915_fragment_shader* fs) { @@ -429,11 +429,11 @@ emit_simple_arith(struct i915_fp_compile *p, /** As above, but swap the first two src regs */ static void emit_simple_arith_swap2(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, + const struct i915_full_instruction *inst, uint opcode, uint numArgs, struct i915_fragment_shader* fs) { - struct tgsi_full_instruction inst2; + struct i915_full_instruction inst2; assert(numArgs == 2); @@ -450,13 +450,14 @@ emit_simple_arith_swap2(struct i915_fp_compile *p, * * Possible concerns: * + * DDX, DDY -- return 0 * SIN, COS -- could use another taylor step? * LIT -- results seem a little different to sw mesa * LOG -- different to mesa on negative numbers, but this is conformant. */ static void i915_translate_instruction(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, + const struct i915_full_instruction *inst, struct i915_fragment_shader *fs) { uint writemask; @@ -727,6 +728,9 @@ i915_translate_instruction(struct i915_fp_compile *p, emit_simple_arith(p, inst, A0_MUL, 2, fs); break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_POW: src0 = src_vector(p, &inst->Src[0], fs); src1 = src_vector(p, &inst->Src[1], fs); @@ -1043,107 +1047,107 @@ i915_translate_instruction(struct i915_fp_compile *p, } -/** - * Translate TGSI fragment shader into i915 hardware instructions. - * \param p the translation state - * \param tokens the TGSI token array - */ -static void -i915_translate_instructions(struct i915_fp_compile *p, - const struct tgsi_token *tokens, - struct i915_fragment_shader *fs) +static void i915_translate_token(struct i915_fp_compile *p, + const union i915_full_token* token, + struct i915_fragment_shader *fs) { struct i915_fragment_shader *ifs = p->shader; - struct tgsi_parse_context parse; - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - tgsi_parse_token( &parse ); + switch( token->Token.Type ) { + case TGSI_TOKEN_TYPE_PROPERTY: + /* + * We only support one cbuf, but we still need to ignore the property + * correctly so we don't hit the assert at the end of the switch case. + */ + assert(token->FullProperty.Property.PropertyName == + TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS); + break; - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_PROPERTY: - /* - * We only support one cbuf, but we still need to ignore the property - * correctly so we don't hit the assert at the end of the switch case. - */ - assert(parse.FullToken.FullProperty.Property.PropertyName == - TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS); - break; - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_CONSTANT) { - uint i; - for (i = parse.FullToken.FullDeclaration.Range.First; - i <= parse.FullToken.FullDeclaration.Range.Last; - i++) { - assert(ifs->constant_flags[i] == 0x0); - ifs->constant_flags[i] = I915_CONSTFLAG_USER; - ifs->num_constants = MAX2(ifs->num_constants, i + 1); - } + case TGSI_TOKEN_TYPE_DECLARATION: + if (token->FullDeclaration.Declaration.File + == TGSI_FILE_CONSTANT) { + uint i; + for (i = token->FullDeclaration.Range.First; + i <= token->FullDeclaration.Range.Last; + i++) { + assert(ifs->constant_flags[i] == 0x0); + ifs->constant_flags[i] = I915_CONSTFLAG_USER; + ifs->num_constants = MAX2(ifs->num_constants, i + 1); } - else if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_TEMPORARY) { - uint i; - for (i = parse.FullToken.FullDeclaration.Range.First; - i <= parse.FullToken.FullDeclaration.Range.Last; - i++) { - if (i >= I915_MAX_TEMPORARY) - debug_printf("Too many temps (%d)\n",i); - else - /* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */ - p->temp_flag |= (1 << i); /* mark temp as used */ - } + } + else if (token->FullDeclaration.Declaration.File + == TGSI_FILE_TEMPORARY) { + uint i; + for (i = token->FullDeclaration.Range.First; + i <= token->FullDeclaration.Range.Last; + i++) { + if (i >= I915_MAX_TEMPORARY) + debug_printf("Too many temps (%d)\n",i); + else + /* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */ + p->temp_flag |= (1 << i); /* mark temp as used */ } - break; + } + break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - const struct tgsi_full_immediate *imm - = &parse.FullToken.FullImmediate; - const uint pos = p->num_immediates++; - uint j; - assert( imm->Immediate.NrTokens <= 4 + 1 ); - for (j = 0; j < imm->Immediate.NrTokens - 1; j++) { - p->immediates[pos][j] = imm->u[j].Float; - } + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + const struct tgsi_full_immediate *imm + = &token->FullImmediate; + const uint pos = p->num_immediates++; + uint j; + assert( imm->Immediate.NrTokens <= 4 + 1 ); + for (j = 0; j < imm->Immediate.NrTokens - 1; j++) { + p->immediates[pos][j] = imm->u[j].Float; } - break; + } + break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - if (p->first_instruction) { - /* resolve location of immediates */ - uint i, j; - for (i = 0; i < p->num_immediates; i++) { - /* find constant slot for this immediate */ - for (j = 0; j < I915_MAX_CONSTANT; j++) { - if (ifs->constant_flags[j] == 0x0) { - memcpy(ifs->constants[j], - p->immediates[i], - 4 * sizeof(float)); - /*printf("immediate %d maps to const %d\n", i, j);*/ - ifs->constant_flags[j] = 0xf; /* all four comps used */ - p->immediates_map[i] = j; - ifs->num_constants = MAX2(ifs->num_constants, j + 1); - break; - } + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (p->first_instruction) { + /* resolve location of immediates */ + uint i, j; + for (i = 0; i < p->num_immediates; i++) { + /* find constant slot for this immediate */ + for (j = 0; j < I915_MAX_CONSTANT; j++) { + if (ifs->constant_flags[j] == 0x0) { + memcpy(ifs->constants[j], + p->immediates[i], + 4 * sizeof(float)); + /*printf("immediate %d maps to const %d\n", i, j);*/ + ifs->constant_flags[j] = 0xf; /* all four comps used */ + p->immediates_map[i] = j; + ifs->num_constants = MAX2(ifs->num_constants, j + 1); + break; } } - - p->first_instruction = FALSE; } - i915_translate_instruction(p, &parse.FullToken.FullInstruction, fs); - break; - - default: - assert( 0 ); + p->first_instruction = FALSE; } - } /* while */ + i915_translate_instruction(p, &token->FullInstruction, fs); + break; + + default: + assert( 0 ); + } - tgsi_parse_free (&parse); +} + +/** + * Translate TGSI fragment shader into i915 hardware instructions. + * \param p the translation state + * \param tokens the TGSI token array + */ +static void +i915_translate_instructions(struct i915_fp_compile *p, + const struct i915_token_list *tokens, + struct i915_fragment_shader *fs) +{ + int i; + for(i = 0; i<tokens->NumTokens; i++) { + i915_translate_token(p, &tokens->Tokens[i], fs); + } } @@ -1302,8 +1306,10 @@ i915_translate_fragment_program( struct i915_context *i915, p = i915_init_compile(i915, fs); - i915_translate_instructions(p, tokens, fs); + struct i915_token_list* i_tokens = i915_optimize(tokens); + i915_translate_instructions(p, i_tokens, fs); i915_fixup_depth_write(p); i915_fini_compile(i915, p); + i915_optimize_free(i_tokens); } diff --git a/src/gallium/drivers/i915/i915_prim_emit.c b/src/gallium/drivers/i915/i915_prim_emit.c index 85656cd7846..1acde97d4bd 100644 --- a/src/gallium/drivers/i915/i915_prim_emit.c +++ b/src/gallium/drivers/i915/i915_prim_emit.c @@ -166,6 +166,8 @@ emit_prim( struct draw_stage *stage, for (i = 0; i < nr; i++) emit_hw_vertex(i915, prim->v[i]); + + i915_flush_heuristically(i915, nr); } diff --git a/src/gallium/drivers/i915/i915_prim_vbuf.c b/src/gallium/drivers/i915/i915_prim_vbuf.c index 79db3b650eb..d8ae1de2963 100644 --- a/src/gallium/drivers/i915/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915/i915_prim_vbuf.c @@ -487,6 +487,7 @@ draw_arrays_fallback(struct vbuf_render *render, draw_arrays_generate_indices(render, start, nr, i915_render->fallback); + i915_flush_heuristically(i915, nr_indices); out: return; } @@ -534,6 +535,7 @@ i915_vbuf_render_draw_arrays(struct vbuf_render *render, nr); OUT_BATCH(start); /* Beginning vertex index */ + i915_flush_heuristically(i915, nr); out: return; } @@ -657,6 +659,7 @@ i915_vbuf_render_draw_elements(struct vbuf_render *render, save_nr_indices, i915_render->fallback); + i915_flush_heuristically(i915, nr_indices); out: return; } diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c index f412626955d..2812de1fe80 100644 --- a/src/gallium/drivers/i915/i915_state.c +++ b/src/gallium/drivers/i915/i915_state.c @@ -244,7 +244,7 @@ i915_create_sampler_state(struct pipe_context *pipe, /* Shadow: */ - if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { cso->state[0] |= (SS2_SHADOW_ENABLE | i915_translate_shadow_compare_func(sampler->compare_func)); diff --git a/src/gallium/drivers/i915/i915_state_emit.c b/src/gallium/drivers/i915/i915_state_emit.c index 39fb13aec7e..4f447962bb9 100644 --- a/src/gallium/drivers/i915/i915_state_emit.c +++ b/src/gallium/drivers/i915/i915_state_emit.c @@ -346,97 +346,80 @@ emit_constants(struct i915_context *i915) static const struct { enum pipe_format format; - uint hw_shift_R; - uint hw_shift_G; - uint hw_shift_B; - uint hw_shift_A; + uint hw_swizzle; } fixup_formats[] = { - { PIPE_FORMAT_R8G8B8A8_UNORM, 20, 24, 28, 16 /* BGRA */}, - { PIPE_FORMAT_L8_UNORM, 28, 28, 28, 16 /* RRRA */}, - { PIPE_FORMAT_I8_UNORM, 28, 28, 28, 16 /* RRRA */}, - { PIPE_FORMAT_A8_UNORM, 16, 16, 16, 16 /* AAAA */}, - { PIPE_FORMAT_NONE, 0, 0, 0, 0}, + { PIPE_FORMAT_R8G8B8A8_UNORM, 0x21030000 /* BGRA */}, + { PIPE_FORMAT_L8_UNORM, 0x00030000 /* RRRA */}, + { PIPE_FORMAT_I8_UNORM, 0x00030000 /* RRRA */}, + { PIPE_FORMAT_A8_UNORM, 0x33330000 /* AAAA */}, + { PIPE_FORMAT_NONE, 0x00000000}, }; -static boolean need_fixup(struct pipe_surface* p) +static uint need_target_fixup(struct pipe_surface* p) { enum pipe_format f; - /* if we don't have a surface bound yet, we don't need to fixup the shader */ if (!p) - return FALSE; + return 0; f = p->format; for(int i=0; fixup_formats[i].format != PIPE_FORMAT_NONE; i++) if (fixup_formats[i].format == f) - return TRUE; + return 1; - return FALSE; + return 0; } -static uint fixup_swizzle(enum pipe_format f, uint v) +static uint fixup_swizzle(enum pipe_format f) { - int i; - - for(i=0; fixup_formats[i].format != PIPE_FORMAT_NONE; i++) + for(int i=0; fixup_formats[i].format != PIPE_FORMAT_NONE; i++) if (fixup_formats[i].format == f) - break; - - if (fixup_formats[i].format == PIPE_FORMAT_NONE) - return v; - - uint rgba = v & 0xFFFF0000; + return fixup_formats[i].hw_swizzle; - v &= 0xFFFF; - v |= ((rgba >> fixup_formats[i].hw_shift_R) & 0xF) << 28; - v |= ((rgba >> fixup_formats[i].hw_shift_G) & 0xF) << 24; - v |= ((rgba >> fixup_formats[i].hw_shift_B) & 0xF) << 20; - v |= ((rgba >> fixup_formats[i].hw_shift_A) & 0xF) << 16; - - return v; + return 0; } static void validate_program(struct i915_context *i915, unsigned *batch_space) { - *batch_space = i915->fs->program_len; + struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; + uint additional_size = need_target_fixup(cbuf_surface); + + /* we need more batch space if we want to emulate rgba framebuffers */ + *batch_space = i915->fs->program_len + 3 * additional_size; } static void emit_program(struct i915_context *i915) { struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; - boolean need_format_fixup = need_fixup(cbuf_surface); - int i; - int fixup_offset = -1; + uint target_fixup = need_target_fixup(cbuf_surface); + uint i; /* we should always have, at least, a pass-through program */ assert(i915->fs->program_len > 0); - if (need_format_fixup) { - /* Find where we emit the output color */ - for (i = i915->fs->program_len - 3; i>0; i-=3) { - uint instr = i915->fs->program[i]; - if ((instr & (REG_NR_MASK << A0_DEST_TYPE_SHIFT)) == - (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) ) { - /* Found it! */ - fixup_offset = i + 1; - break; - } - } - if (fixup_offset == -1) { - need_format_fixup = FALSE; - debug_printf("couldn't find fixup offset\n"); - } + { + /* first word has the size, we have to adjust that */ + uint size = (i915->fs->program[0]); + size += target_fixup * 3; + OUT_BATCH(size); } - /* emit the program to the hw */ - for (i = 0; i < i915->fs->program_len; i++) { - if (need_format_fixup && (i == fixup_offset) ) { - uint v = fixup_swizzle(cbuf_surface->format, i915->fs->program[i]); - OUT_BATCH(v); - } else - OUT_BATCH(i915->fs->program[i]); + /* output the declarations of the program */ + for (i=1 ; i < i915->fs->program_len; i++) + OUT_BATCH(i915->fs->program[i]); + + /* we emit an additional mov with swizzle to fake RGBA framebuffers */ + if (target_fixup) { + /* mov out_color, out_color.zyxw */ + OUT_BATCH(A0_MOV | + (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | + A0_DEST_CHANNEL_ALL | + (REG_TYPE_OC << A0_SRC0_TYPE_SHIFT) | + (T_DIFFUSE << A0_SRC0_NR_SHIFT)); + OUT_BATCH(fixup_swizzle(cbuf_surface->format)); + OUT_BATCH(0); } } diff --git a/src/gallium/drivers/i915/i915_winsys.h b/src/gallium/drivers/i915/i915_winsys.h index 21cfdc9613e..20438609e07 100644 --- a/src/gallium/drivers/i915/i915_winsys.h +++ b/src/gallium/drivers/i915/i915_winsys.h @@ -207,6 +207,12 @@ struct i915_winsys { void (*buffer_destroy)(struct i915_winsys *iws, struct i915_winsys_buffer *buffer); + + /** + * Check if a buffer is busy. + */ + boolean (*buffer_is_busy)(struct i915_winsys *iws, + struct i915_winsys_buffer *buffer); /*@}*/ diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c index ceb83f6e684..ac3e361a446 100644 --- a/src/gallium/drivers/nv50/nv50_context.c +++ b/src/gallium/drivers/nv50/nv50_context.c @@ -60,13 +60,13 @@ nv50_texture_barrier(struct pipe_context *pipe) void nv50_default_flush_notify(struct nouveau_channel *chan) { - struct nv50_context *nv50 = chan->user_private; + struct nv50_screen *screen = chan->user_private; - if (!nv50) + if (!screen) return; - nouveau_fence_update(&nv50->screen->base, TRUE); - nouveau_fence_next(&nv50->screen->base); + nouveau_fence_update(&screen->base, TRUE); + nouveau_fence_next(&screen->base); } static void @@ -100,10 +100,8 @@ nv50_destroy(struct pipe_context *pipe) draw_destroy(nv50->draw); - if (nv50->screen->cur_ctx == nv50) { - nv50->screen->base.channel->user_private = NULL; + if (nv50->screen->cur_ctx == nv50) nv50->screen->cur_ctx = NULL; - } FREE(nv50); } @@ -140,7 +138,6 @@ nv50_create(struct pipe_screen *pscreen, void *priv) if (!screen->cur_ctx) screen->cur_ctx = nv50; - screen->base.channel->user_private = nv50; screen->base.channel->flush_notify = nv50_default_flush_notify; nv50_init_query_functions(nv50); diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index cc921d08666..4cda303c44a 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -215,6 +215,7 @@ nv50_screen_destroy(struct pipe_screen *pscreen) nouveau_fence_wait(screen->base.fence.current); nouveau_fence_ref (NULL, &screen->base.fence.current); } + screen->base.channel->user_private = NULL; nouveau_bo_ref(NULL, &screen->code); nouveau_bo_ref(NULL, &screen->tls_bo); @@ -300,6 +301,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) FAIL_SCREEN_INIT("nouveau_screen_init failed: %d\n", ret); chan = screen->base.channel; + chan->user_private = screen; pscreen->winsys = ws; pscreen->destroy = nv50_screen_destroy; diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 11561f5a8e6..d29c1e9723f 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -282,8 +282,7 @@ nv50_switch_pipe_context(struct nv50_context *ctx_to) if (!ctx_to->zsa) ctx_to->dirty &= ~NV50_NEW_ZSA; - ctx_to->screen->base.channel->user_private = ctx_to->screen->cur_ctx = - ctx_to; + ctx_to->screen->cur_ctx = ctx_to; } static struct state_validate { diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 3d7e880ccce..fb51db84ac6 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -34,25 +34,16 @@ #include "nv50_defs.xml.h" +#define NV50_ENG2D_SUPPORTED_FORMATS 0xff0843e080608409ULL + /* return TRUE for formats that can be converted among each other by NV50_2D */ static INLINE boolean nv50_2d_format_faithful(enum pipe_format format) { - switch (format) { - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_B8G8R8X8_UNORM: - case PIPE_FORMAT_B8G8R8A8_SRGB: - case PIPE_FORMAT_B8G8R8X8_SRGB: - case PIPE_FORMAT_B5G6R5_UNORM: - case PIPE_FORMAT_B5G5R5A1_UNORM: - case PIPE_FORMAT_B10G10R10A2_UNORM: - case PIPE_FORMAT_R8_UNORM: - case PIPE_FORMAT_R32G32B32A32_FLOAT: - case PIPE_FORMAT_R32G32B32_FLOAT: - return TRUE; - default: - return FALSE; - } + uint8_t id = nv50_format_table[format].rt; + + return (id >= 0xc0) && + (NV50_ENG2D_SUPPORTED_FORMATS & (1ULL << (id - 0xc0))); } static INLINE uint8_t @@ -63,7 +54,7 @@ nv50_2d_format(enum pipe_format format) /* Hardware values for color formats range from 0xc0 to 0xff, * but the 2D engine doesn't support all of them. */ - if ((id >= 0xc0) && (0xff0843e080608409ULL & (1ULL << (id - 0xc0)))) + if ((id >= 0xc0) && (NV50_ENG2D_SUPPORTED_FORMATS & (1ULL << (id - 0xc0)))) return id; switch (util_format_get_blocksize(format)) { diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index bb08941c243..f23008ae4cf 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -389,11 +389,11 @@ nv50_prim_gl(unsigned prim) static void nv50_draw_vbo_flush_notify(struct nouveau_channel *chan) { - struct nv50_context *nv50 = chan->user_private; + struct nv50_screen *screen = chan->user_private; - nouveau_fence_update(&nv50->screen->base, TRUE); + nouveau_fence_update(&screen->base, TRUE); - nv50_bufctx_emit_relocs(nv50); + nv50_bufctx_emit_relocs(screen->cur_ctx); } static void @@ -650,7 +650,6 @@ nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) nv50_state_validate(nv50); chan->flush_notify = nv50_draw_vbo_flush_notify; - chan->user_private = nv50; if (nv50->vbo_fifo) { nv50_push_vbo(nv50, info); diff --git a/src/gallium/drivers/nvc0/nvc0_context.c b/src/gallium/drivers/nvc0/nvc0_context.c index 2679b7f86aa..983db23eedb 100644 --- a/src/gallium/drivers/nvc0/nvc0_context.c +++ b/src/gallium/drivers/nvc0/nvc0_context.c @@ -89,10 +89,8 @@ nvc0_destroy(struct pipe_context *pipe) draw_destroy(nvc0->draw); - if (nvc0->screen->cur_ctx == nvc0) { - nvc0->screen->base.channel->user_private = NULL; + if (nvc0->screen->cur_ctx == nvc0) nvc0->screen->cur_ctx = NULL; - } FREE(nvc0); } @@ -100,13 +98,13 @@ nvc0_destroy(struct pipe_context *pipe) void nvc0_default_flush_notify(struct nouveau_channel *chan) { - struct nvc0_context *nvc0 = chan->user_private; + struct nvc0_screen *screen = chan->user_private; - if (!nvc0) + if (!screen) return; - nouveau_fence_update(&nvc0->screen->base, TRUE); - nouveau_fence_next(&nvc0->screen->base); + nouveau_fence_update(&screen->base, TRUE); + nouveau_fence_next(&screen->base); } struct pipe_context * @@ -141,7 +139,6 @@ nvc0_create(struct pipe_screen *pscreen, void *priv) if (!screen->cur_ctx) screen->cur_ctx = nvc0; - screen->base.channel->user_private = nvc0; screen->base.channel->flush_notify = nvc0_default_flush_notify; nvc0_init_query_functions(nvc0); diff --git a/src/gallium/drivers/nvc0/nvc0_screen.c b/src/gallium/drivers/nvc0/nvc0_screen.c index 34bf0f0a2ad..1bd7fa9f0ea 100644 --- a/src/gallium/drivers/nvc0/nvc0_screen.c +++ b/src/gallium/drivers/nvc0/nvc0_screen.c @@ -198,8 +198,11 @@ nvc0_screen_destroy(struct pipe_screen *pscreen) { struct nvc0_screen *screen = nvc0_screen(pscreen); - nouveau_fence_wait(screen->base.fence.current); - nouveau_fence_ref(NULL, &screen->base.fence.current); + if (screen->base.fence.current) { + nouveau_fence_wait(screen->base.fence.current); + nouveau_fence_ref(NULL, &screen->base.fence.current); + } + screen->base.channel->user_private = NULL; nouveau_bo_ref(NULL, &screen->text); nouveau_bo_ref(NULL, &screen->tls); @@ -358,6 +361,7 @@ nvc0_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) return NULL; } chan = screen->base.channel; + chan->user_private = screen; pscreen->winsys = ws; pscreen->destroy = nvc0_screen_destroy; diff --git a/src/gallium/drivers/nvc0/nvc0_state_validate.c b/src/gallium/drivers/nvc0/nvc0_state_validate.c index 9b2a28150b1..f300f37fb7b 100644 --- a/src/gallium/drivers/nvc0/nvc0_state_validate.c +++ b/src/gallium/drivers/nvc0/nvc0_state_validate.c @@ -428,8 +428,7 @@ nvc0_switch_pipe_context(struct nvc0_context *ctx_to) if (!ctx_to->zsa) ctx_to->dirty &= ~NVC0_NEW_ZSA; - ctx_to->screen->base.channel->user_private = ctx_to->screen->cur_ctx = - ctx_to; + ctx_to->screen->cur_ctx = ctx_to; } static struct state_validate { diff --git a/src/gallium/drivers/nvc0/nvc0_vbo.c b/src/gallium/drivers/nvc0/nvc0_vbo.c index 41079104b39..8a5bf8dc582 100644 --- a/src/gallium/drivers/nvc0/nvc0_vbo.c +++ b/src/gallium/drivers/nvc0/nvc0_vbo.c @@ -367,11 +367,11 @@ nvc0_prim_gl(unsigned prim) static void nvc0_draw_vbo_flush_notify(struct nouveau_channel *chan) { - struct nvc0_context *nvc0 = chan->user_private; + struct nvc0_screen *screen = chan->user_private; - nouveau_fence_update(&nvc0->screen->base, TRUE); + nouveau_fence_update(&screen->base, TRUE); - nvc0_bufctx_emit_relocs(nvc0); + nvc0_bufctx_emit_relocs(screen->cur_ctx); } static void @@ -587,7 +587,6 @@ nvc0_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) nvc0_state_validate(nvc0); chan->flush_notify = nvc0_draw_vbo_flush_notify; - chan->user_private = nvc0; if (nvc0->vbo_fifo) { nvc0_push_vbo(nvc0, info); diff --git a/src/gallium/drivers/r600/eg_asm.c b/src/gallium/drivers/r600/eg_asm.c index fb0b0f104bf..c95872b0809 100644 --- a/src/gallium/drivers/r600/eg_asm.c +++ b/src/gallium/drivers/r600/eg_asm.c @@ -69,7 +69,7 @@ int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W(cf->output.swizzle_w) | S_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(cf->output.barrier) | S_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(cf->output.inst); - if (bc->chiprev == CHIPREV_EVERGREEN) /* no EOP on cayman */ + if (bc->chip_class == EVERGREEN) /* no EOP on cayman */ bc->bytecode[id] |= S_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(cf->output.end_of_program); id++; diff --git a/src/gallium/drivers/r600/eg_state_inlines.h b/src/gallium/drivers/r600/eg_state_inlines.h deleted file mode 100644 index b5590116e8f..00000000000 --- a/src/gallium/drivers/r600/eg_state_inlines.h +++ /dev/null @@ -1,587 +0,0 @@ -/* - * Copyright 2010 Red Hat Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ -#ifndef EG_STATE_INLINES_H -#define EG_STATE_INLINES_H - -#include "util/u_format.h" -#include "evergreend.h" -#include "r600_formats.h" - -static INLINE uint32_t r600_translate_blend_function(int blend_func) -{ - switch (blend_func) { - case PIPE_BLEND_ADD: - return V_028780_COMB_DST_PLUS_SRC; - case PIPE_BLEND_SUBTRACT: - return V_028780_COMB_SRC_MINUS_DST; - case PIPE_BLEND_REVERSE_SUBTRACT: - return V_028780_COMB_DST_MINUS_SRC; - case PIPE_BLEND_MIN: - return V_028780_COMB_MIN_DST_SRC; - case PIPE_BLEND_MAX: - return V_028780_COMB_MAX_DST_SRC; - default: - R600_ERR("Unknown blend function %d\n", blend_func); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_blend_factor(int blend_fact) -{ - switch (blend_fact) { - case PIPE_BLENDFACTOR_ONE: - return V_028780_BLEND_ONE; - case PIPE_BLENDFACTOR_SRC_COLOR: - return V_028780_BLEND_SRC_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA: - return V_028780_BLEND_SRC_ALPHA; - case PIPE_BLENDFACTOR_DST_ALPHA: - return V_028780_BLEND_DST_ALPHA; - case PIPE_BLENDFACTOR_DST_COLOR: - return V_028780_BLEND_DST_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - return V_028780_BLEND_SRC_ALPHA_SATURATE; - case PIPE_BLENDFACTOR_CONST_COLOR: - return V_028780_BLEND_CONST_COLOR; - case PIPE_BLENDFACTOR_CONST_ALPHA: - return V_028780_BLEND_CONST_ALPHA; - case PIPE_BLENDFACTOR_ZERO: - return V_028780_BLEND_ZERO; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - return V_028780_BLEND_ONE_MINUS_SRC_COLOR; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - return V_028780_BLEND_ONE_MINUS_SRC_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - return V_028780_BLEND_ONE_MINUS_DST_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - return V_028780_BLEND_ONE_MINUS_DST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - return V_028780_BLEND_ONE_MINUS_CONST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - return V_028780_BLEND_ONE_MINUS_CONST_ALPHA; - case PIPE_BLENDFACTOR_SRC1_COLOR: - return V_028780_BLEND_SRC1_COLOR; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - return V_028780_BLEND_SRC1_ALPHA; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - return V_028780_BLEND_INV_SRC1_COLOR; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - return V_028780_BLEND_INV_SRC1_ALPHA; - default: - R600_ERR("Bad blend factor %d not supported!\n", blend_fact); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_stencil_op(int s_op) -{ - switch (s_op) { - case PIPE_STENCIL_OP_KEEP: - return V_028800_STENCIL_KEEP; - case PIPE_STENCIL_OP_ZERO: - return V_028800_STENCIL_ZERO; - case PIPE_STENCIL_OP_REPLACE: - return V_028800_STENCIL_REPLACE; - case PIPE_STENCIL_OP_INCR: - return V_028800_STENCIL_INCR; - case PIPE_STENCIL_OP_DECR: - return V_028800_STENCIL_DECR; - case PIPE_STENCIL_OP_INCR_WRAP: - return V_028800_STENCIL_INCR_WRAP; - case PIPE_STENCIL_OP_DECR_WRAP: - return V_028800_STENCIL_DECR_WRAP; - case PIPE_STENCIL_OP_INVERT: - return V_028800_STENCIL_INVERT; - default: - R600_ERR("Unknown stencil op %d", s_op); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_fill(uint32_t func) -{ - switch(func) { - case PIPE_POLYGON_MODE_FILL: - return 2; - case PIPE_POLYGON_MODE_LINE: - return 1; - case PIPE_POLYGON_MODE_POINT: - return 0; - default: - assert(0); - return 0; - } -} - -/* translates straight */ -static INLINE uint32_t r600_translate_ds_func(int func) -{ - return func; -} - -static inline unsigned r600_tex_wrap(unsigned wrap) -{ - switch (wrap) { - default: - case PIPE_TEX_WRAP_REPEAT: - return V_03C000_SQ_TEX_WRAP; - case PIPE_TEX_WRAP_CLAMP: - return V_03C000_SQ_TEX_CLAMP_HALF_BORDER; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return V_03C000_SQ_TEX_CLAMP_LAST_TEXEL; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return V_03C000_SQ_TEX_CLAMP_BORDER; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - return V_03C000_SQ_TEX_MIRROR; - case PIPE_TEX_WRAP_MIRROR_CLAMP: - return V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - return V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: - return V_03C000_SQ_TEX_MIRROR_ONCE_BORDER; - } -} - -static inline unsigned r600_tex_filter(unsigned filter) -{ - switch (filter) { - default: - case PIPE_TEX_FILTER_NEAREST: - return V_03C000_SQ_TEX_XY_FILTER_POINT; - case PIPE_TEX_FILTER_LINEAR: - return V_03C000_SQ_TEX_XY_FILTER_BILINEAR; - } -} - -static inline unsigned r600_tex_mipfilter(unsigned filter) -{ - switch (filter) { - case PIPE_TEX_MIPFILTER_NEAREST: - return V_03C000_SQ_TEX_Z_FILTER_POINT; - case PIPE_TEX_MIPFILTER_LINEAR: - return V_03C000_SQ_TEX_Z_FILTER_LINEAR; - default: - case PIPE_TEX_MIPFILTER_NONE: - return V_03C000_SQ_TEX_Z_FILTER_NONE; - } -} - -static inline unsigned r600_tex_compare(unsigned compare) -{ - switch (compare) { - default: - case PIPE_FUNC_NEVER: - return V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER; - case PIPE_FUNC_LESS: - return V_03C000_SQ_TEX_DEPTH_COMPARE_LESS; - case PIPE_FUNC_EQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL; - case PIPE_FUNC_LEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL; - case PIPE_FUNC_GREATER: - return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER; - case PIPE_FUNC_NOTEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL; - case PIPE_FUNC_GEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL; - case PIPE_FUNC_ALWAYS: - return V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS; - } -} - -static inline unsigned r600_tex_swizzle(unsigned swizzle) -{ - switch (swizzle) { - case PIPE_SWIZZLE_RED: - return V_030010_SQ_SEL_X; - case PIPE_SWIZZLE_GREEN: - return V_030010_SQ_SEL_Y; - case PIPE_SWIZZLE_BLUE: - return V_030010_SQ_SEL_Z; - case PIPE_SWIZZLE_ALPHA: - return V_030010_SQ_SEL_W; - case PIPE_SWIZZLE_ZERO: - return V_030010_SQ_SEL_0; - default: - case PIPE_SWIZZLE_ONE: - return V_030010_SQ_SEL_1; - } -} - -static inline unsigned r600_format_type(unsigned format_type) -{ - switch (format_type) { - default: - case UTIL_FORMAT_TYPE_UNSIGNED: - return V_030010_SQ_FORMAT_COMP_UNSIGNED; - case UTIL_FORMAT_TYPE_SIGNED: - return V_030010_SQ_FORMAT_COMP_SIGNED; - case UTIL_FORMAT_TYPE_FIXED: - return V_030010_SQ_FORMAT_COMP_UNSIGNED_BIASED; - } -} - -static inline unsigned r600_tex_dim(unsigned dim) -{ - switch (dim) { - default: - case PIPE_TEXTURE_1D: - return V_030000_SQ_TEX_DIM_1D; - case PIPE_TEXTURE_1D_ARRAY: - return V_030000_SQ_TEX_DIM_1D_ARRAY; - case PIPE_TEXTURE_2D: - case PIPE_TEXTURE_RECT: - return V_030000_SQ_TEX_DIM_2D; - case PIPE_TEXTURE_2D_ARRAY: - return V_030000_SQ_TEX_DIM_2D_ARRAY; - case PIPE_TEXTURE_3D: - return V_030000_SQ_TEX_DIM_3D; - case PIPE_TEXTURE_CUBE: - return V_030000_SQ_TEX_DIM_CUBEMAP; - } -} - -static inline uint32_t r600_translate_dbformat(enum pipe_format format) -{ - switch (format) { - case PIPE_FORMAT_Z16_UNORM: - return V_028040_Z_16; - case PIPE_FORMAT_Z24X8_UNORM: - return V_028040_Z_24; - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_028040_Z_24; - default: - return ~0; - } -} - -static inline uint32_t r600_translate_stencilformat(enum pipe_format format) -{ - if (format == PIPE_FORMAT_Z24_UNORM_S8_USCALED) - return 1; - else - return 0; -} - -static inline uint32_t r600_translate_colorswap(enum pipe_format format) -{ - switch (format) { - /* 8-bit buffers. */ - case PIPE_FORMAT_L4A4_UNORM: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_A8_UNORM: - return V_028C70_SWAP_ALT_REV; - case PIPE_FORMAT_I8_UNORM: - case PIPE_FORMAT_L8_UNORM: - case PIPE_FORMAT_L8_SRGB: - case PIPE_FORMAT_R8_UNORM: - case PIPE_FORMAT_R8_SNORM: - return V_028C70_SWAP_STD; - - /* 16-bit buffers. */ - case PIPE_FORMAT_B5G6R5_UNORM: - return V_028C70_SWAP_STD_REV; - - case PIPE_FORMAT_B5G5R5A1_UNORM: - case PIPE_FORMAT_B5G5R5X1_UNORM: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_B4G4R4A4_UNORM: - case PIPE_FORMAT_B4G4R4X4_UNORM: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_Z16_UNORM: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_L8A8_UNORM: - case PIPE_FORMAT_L8A8_SRGB: - return V_028C70_SWAP_ALT; - case PIPE_FORMAT_R8G8_UNORM: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_R16_UNORM: - case PIPE_FORMAT_R16_FLOAT: - return V_028C70_SWAP_STD; - - /* 32-bit buffers. */ - case PIPE_FORMAT_A8B8G8R8_SRGB: - return V_028C70_SWAP_STD_REV; - case PIPE_FORMAT_B8G8R8A8_SRGB: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_B8G8R8X8_UNORM: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_X8R8G8B8_UNORM: - return V_028C70_SWAP_ALT_REV; - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_UNORM: - case PIPE_FORMAT_R8G8B8X8_UNORM: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_A8B8G8R8_UNORM: - case PIPE_FORMAT_X8B8G8R8_UNORM: - /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */ - return V_028C70_SWAP_STD_REV; - - case PIPE_FORMAT_Z24X8_UNORM: - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_X8Z24_UNORM: - case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_R10G10B10A2_UNORM: - case PIPE_FORMAT_R10G10B10X2_SNORM: - case PIPE_FORMAT_R10SG10SB10SA2U_NORM: - return V_028C70_SWAP_STD; - - case PIPE_FORMAT_B10G10R10A2_UNORM: - return V_028C70_SWAP_ALT; - - case PIPE_FORMAT_R11G11B10_FLOAT: - case PIPE_FORMAT_R32_FLOAT: - case PIPE_FORMAT_R16G16_FLOAT: - case PIPE_FORMAT_R16G16_UNORM: - return V_028C70_SWAP_STD; - - /* 64-bit buffers. */ - case PIPE_FORMAT_R32G32_FLOAT: - case PIPE_FORMAT_R16G16B16A16_UNORM: - case PIPE_FORMAT_R16G16B16A16_SNORM: - case PIPE_FORMAT_R16G16B16A16_FLOAT: - - /* 128-bit buffers. */ - case PIPE_FORMAT_R32G32B32A32_FLOAT: - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_UNORM: - return V_028C70_SWAP_STD; - default: - R600_ERR("unsupported colorswap format %d\n", format); - return ~0; - } - return ~0; -} - -static INLINE uint32_t r600_translate_colorformat(enum pipe_format format) -{ - switch (format) { - /* 8-bit buffers. */ - case PIPE_FORMAT_L4A4_UNORM: - return V_028C70_COLOR_4_4; - - case PIPE_FORMAT_A8_UNORM: - case PIPE_FORMAT_I8_UNORM: - case PIPE_FORMAT_L8_UNORM: - case PIPE_FORMAT_L8_SRGB: - case PIPE_FORMAT_R8_UNORM: - case PIPE_FORMAT_R8_SNORM: - return V_028C70_COLOR_8; - - /* 16-bit buffers. */ - case PIPE_FORMAT_B5G6R5_UNORM: - return V_028C70_COLOR_5_6_5; - - case PIPE_FORMAT_B5G5R5A1_UNORM: - case PIPE_FORMAT_B5G5R5X1_UNORM: - return V_028C70_COLOR_1_5_5_5; - - case PIPE_FORMAT_B4G4R4A4_UNORM: - case PIPE_FORMAT_B4G4R4X4_UNORM: - return V_028C70_COLOR_4_4_4_4; - - case PIPE_FORMAT_Z16_UNORM: - return V_028C70_COLOR_16; - - case PIPE_FORMAT_L8A8_UNORM: - case PIPE_FORMAT_L8A8_SRGB: - case PIPE_FORMAT_R8G8_UNORM: - return V_028C70_COLOR_8_8; - - case PIPE_FORMAT_R16_UNORM: - return V_028C70_COLOR_16; - - case PIPE_FORMAT_R16_FLOAT: - return V_028C70_COLOR_16_FLOAT; - - /* 32-bit buffers. */ - case PIPE_FORMAT_A8B8G8R8_SRGB: - case PIPE_FORMAT_A8B8G8R8_UNORM: - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_B8G8R8A8_SRGB: - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_B8G8R8X8_UNORM: - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_UNORM: - case PIPE_FORMAT_R8G8B8X8_UNORM: - case PIPE_FORMAT_R8SG8SB8UX8U_NORM: - case PIPE_FORMAT_X8B8G8R8_UNORM: - case PIPE_FORMAT_X8R8G8B8_UNORM: - case PIPE_FORMAT_R8G8B8_UNORM: - return V_028C70_COLOR_8_8_8_8; - - case PIPE_FORMAT_R10G10B10A2_UNORM: - case PIPE_FORMAT_R10G10B10X2_SNORM: - case PIPE_FORMAT_B10G10R10A2_UNORM: - case PIPE_FORMAT_R10SG10SB10SA2U_NORM: - return V_028C70_COLOR_2_10_10_10; - - case PIPE_FORMAT_Z24X8_UNORM: - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_028C70_COLOR_8_24; - - case PIPE_FORMAT_X8Z24_UNORM: - case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - return V_028C70_COLOR_24_8; - - case PIPE_FORMAT_R32_FLOAT: - return V_028C70_COLOR_32_FLOAT; - - case PIPE_FORMAT_R16G16_FLOAT: - return V_028C70_COLOR_16_16_FLOAT; - - case PIPE_FORMAT_R16G16_SSCALED: - case PIPE_FORMAT_R16G16_UNORM: - return V_028C70_COLOR_16_16; - - case PIPE_FORMAT_R11G11B10_FLOAT: - return V_028C70_COLOR_10_11_11_FLOAT; - - /* 64-bit buffers. */ - case PIPE_FORMAT_R16G16B16_USCALED: - case PIPE_FORMAT_R16G16B16A16_USCALED: - case PIPE_FORMAT_R16G16B16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_UNORM: - case PIPE_FORMAT_R16G16B16A16_SNORM: - return V_028C70_COLOR_16_16_16_16; - - case PIPE_FORMAT_R16G16B16_FLOAT: - case PIPE_FORMAT_R16G16B16A16_FLOAT: - return V_028C70_COLOR_16_16_16_16_FLOAT; - - case PIPE_FORMAT_R32G32_FLOAT: - return V_028C70_COLOR_32_32_FLOAT; - - case PIPE_FORMAT_R32G32_USCALED: - case PIPE_FORMAT_R32G32_SSCALED: - return V_028C70_COLOR_32_32; - - /* 96-bit buffers. */ - case PIPE_FORMAT_R32G32B32_FLOAT: - return V_028C70_COLOR_32_32_32_FLOAT; - - /* 128-bit buffers. */ - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_UNORM: - return V_028C70_COLOR_32_32_32_32; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return V_028C70_COLOR_32_32_32_32_FLOAT; - - /* YUV buffers. */ - case PIPE_FORMAT_UYVY: - case PIPE_FORMAT_YUYV: - default: - return ~0; /* Unsupported. */ - } -} - -static INLINE uint32_t r600_colorformat_endian_swap(uint32_t colorformat) -{ - if (R600_BIG_ENDIAN) { - switch(colorformat) { - case V_028C70_COLOR_4_4: - return(ENDIAN_NONE); - - /* 8-bit buffers. */ - case V_028C70_COLOR_8: - return(ENDIAN_NONE); - - /* 16-bit buffers. */ - case V_028C70_COLOR_5_6_5: - case V_028C70_COLOR_1_5_5_5: - case V_028C70_COLOR_4_4_4_4: - case V_028C70_COLOR_16: - case V_028C70_COLOR_8_8: - return(ENDIAN_8IN16); - - /* 32-bit buffers. */ - case V_028C70_COLOR_8_8_8_8: - case V_028C70_COLOR_2_10_10_10: - case V_028C70_COLOR_8_24: - case V_028C70_COLOR_24_8: - case V_028C70_COLOR_32_FLOAT: - case V_028C70_COLOR_16_16_FLOAT: - case V_028C70_COLOR_16_16: - return(ENDIAN_8IN32); - - /* 64-bit buffers. */ - case V_028C70_COLOR_16_16_16_16: - case V_028C70_COLOR_16_16_16_16_FLOAT: - return(ENDIAN_8IN16); - - case V_028C70_COLOR_32_32_FLOAT: - case V_028C70_COLOR_32_32: - return(ENDIAN_8IN32); - - /* 96-bit buffers. */ - case V_028C70_COLOR_32_32_32_FLOAT: - /* 128-bit buffers. */ - case V_028C70_COLOR_32_32_32_32_FLOAT: - case V_028C70_COLOR_32_32_32_32: - return(ENDIAN_8IN32); - default: - return ENDIAN_NONE; /* Unsupported. */ - } - } else { - return ENDIAN_NONE; - } -} - -static INLINE boolean r600_is_sampler_format_supported(struct pipe_screen *screen, enum pipe_format format) -{ - return r600_translate_texformat(screen, format, NULL, NULL, NULL) != ~0; -} - -static INLINE boolean r600_is_colorbuffer_format_supported(enum pipe_format format) -{ - return r600_translate_colorformat(format) != ~0 && - r600_translate_colorswap(format) != ~0; -} - -static INLINE boolean r600_is_zs_format_supported(enum pipe_format format) -{ - return r600_translate_dbformat(format) != ~0; -} - -#endif diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c index dc182611482..fbf25feaf20 100644 --- a/src/gallium/drivers/r600/evergreen_state.c +++ b/src/gallium/drivers/r600/evergreen_state.c @@ -46,7 +46,587 @@ #include "r600_resource.h" #include "r600_shader.h" #include "r600_pipe.h" -#include "eg_state_inlines.h" +#include "r600_formats.h" + +static uint32_t r600_translate_blend_function(int blend_func) +{ + switch (blend_func) { + case PIPE_BLEND_ADD: + return V_028780_COMB_DST_PLUS_SRC; + case PIPE_BLEND_SUBTRACT: + return V_028780_COMB_SRC_MINUS_DST; + case PIPE_BLEND_REVERSE_SUBTRACT: + return V_028780_COMB_DST_MINUS_SRC; + case PIPE_BLEND_MIN: + return V_028780_COMB_MIN_DST_SRC; + case PIPE_BLEND_MAX: + return V_028780_COMB_MAX_DST_SRC; + default: + R600_ERR("Unknown blend function %d\n", blend_func); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_blend_factor(int blend_fact) +{ + switch (blend_fact) { + case PIPE_BLENDFACTOR_ONE: + return V_028780_BLEND_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return V_028780_BLEND_SRC_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return V_028780_BLEND_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return V_028780_BLEND_DST_ALPHA; + case PIPE_BLENDFACTOR_DST_COLOR: + return V_028780_BLEND_DST_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return V_028780_BLEND_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return V_028780_BLEND_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return V_028780_BLEND_CONST_ALPHA; + case PIPE_BLENDFACTOR_ZERO: + return V_028780_BLEND_ZERO; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return V_028780_BLEND_ONE_MINUS_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return V_028780_BLEND_ONE_MINUS_SRC_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return V_028780_BLEND_ONE_MINUS_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return V_028780_BLEND_ONE_MINUS_DST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return V_028780_BLEND_ONE_MINUS_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return V_028780_BLEND_ONE_MINUS_CONST_ALPHA; + case PIPE_BLENDFACTOR_SRC1_COLOR: + return V_028780_BLEND_SRC1_COLOR; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + return V_028780_BLEND_SRC1_ALPHA; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + return V_028780_BLEND_INV_SRC1_COLOR; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + return V_028780_BLEND_INV_SRC1_ALPHA; + default: + R600_ERR("Bad blend factor %d not supported!\n", blend_fact); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_stencil_op(int s_op) +{ + switch (s_op) { + case PIPE_STENCIL_OP_KEEP: + return V_028800_STENCIL_KEEP; + case PIPE_STENCIL_OP_ZERO: + return V_028800_STENCIL_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return V_028800_STENCIL_REPLACE; + case PIPE_STENCIL_OP_INCR: + return V_028800_STENCIL_INCR; + case PIPE_STENCIL_OP_DECR: + return V_028800_STENCIL_DECR; + case PIPE_STENCIL_OP_INCR_WRAP: + return V_028800_STENCIL_INCR_WRAP; + case PIPE_STENCIL_OP_DECR_WRAP: + return V_028800_STENCIL_DECR_WRAP; + case PIPE_STENCIL_OP_INVERT: + return V_028800_STENCIL_INVERT; + default: + R600_ERR("Unknown stencil op %d", s_op); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_fill(uint32_t func) +{ + switch(func) { + case PIPE_POLYGON_MODE_FILL: + return 2; + case PIPE_POLYGON_MODE_LINE: + return 1; + case PIPE_POLYGON_MODE_POINT: + return 0; + default: + assert(0); + return 0; + } +} + +/* translates straight */ +static uint32_t r600_translate_ds_func(int func) +{ + return func; +} + +static unsigned r600_tex_wrap(unsigned wrap) +{ + switch (wrap) { + default: + case PIPE_TEX_WRAP_REPEAT: + return V_03C000_SQ_TEX_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return V_03C000_SQ_TEX_CLAMP_HALF_BORDER; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return V_03C000_SQ_TEX_CLAMP_LAST_TEXEL; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return V_03C000_SQ_TEX_CLAMP_BORDER; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return V_03C000_SQ_TEX_MIRROR; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + return V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + return V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + return V_03C000_SQ_TEX_MIRROR_ONCE_BORDER; + } +} + +static unsigned r600_tex_filter(unsigned filter) +{ + switch (filter) { + default: + case PIPE_TEX_FILTER_NEAREST: + return V_03C000_SQ_TEX_XY_FILTER_POINT; + case PIPE_TEX_FILTER_LINEAR: + return V_03C000_SQ_TEX_XY_FILTER_BILINEAR; + } +} + +static unsigned r600_tex_mipfilter(unsigned filter) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + return V_03C000_SQ_TEX_Z_FILTER_POINT; + case PIPE_TEX_MIPFILTER_LINEAR: + return V_03C000_SQ_TEX_Z_FILTER_LINEAR; + default: + case PIPE_TEX_MIPFILTER_NONE: + return V_03C000_SQ_TEX_Z_FILTER_NONE; + } +} + +static unsigned r600_tex_compare(unsigned compare) +{ + switch (compare) { + default: + case PIPE_FUNC_NEVER: + return V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER; + case PIPE_FUNC_LESS: + return V_03C000_SQ_TEX_DEPTH_COMPARE_LESS; + case PIPE_FUNC_EQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL; + case PIPE_FUNC_LEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL; + case PIPE_FUNC_GREATER: + return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER; + case PIPE_FUNC_NOTEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL; + case PIPE_FUNC_GEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL; + case PIPE_FUNC_ALWAYS: + return V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS; + } +} + +static unsigned r600_tex_dim(unsigned dim) +{ + switch (dim) { + default: + case PIPE_TEXTURE_1D: + return V_030000_SQ_TEX_DIM_1D; + case PIPE_TEXTURE_1D_ARRAY: + return V_030000_SQ_TEX_DIM_1D_ARRAY; + case PIPE_TEXTURE_2D: + case PIPE_TEXTURE_RECT: + return V_030000_SQ_TEX_DIM_2D; + case PIPE_TEXTURE_2D_ARRAY: + return V_030000_SQ_TEX_DIM_2D_ARRAY; + case PIPE_TEXTURE_3D: + return V_030000_SQ_TEX_DIM_3D; + case PIPE_TEXTURE_CUBE: + return V_030000_SQ_TEX_DIM_CUBEMAP; + } +} + +static uint32_t r600_translate_dbformat(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + return V_028040_Z_16; + case PIPE_FORMAT_Z24X8_UNORM: + return V_028040_Z_24; + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_028040_Z_24; + default: + return ~0U; + } +} + +static uint32_t r600_translate_stencilformat(enum pipe_format format) +{ + if (format == PIPE_FORMAT_Z24_UNORM_S8_USCALED) + return 1; + else + return 0; +} + +static uint32_t r600_translate_colorswap(enum pipe_format format) +{ + switch (format) { + /* 8-bit buffers. */ + case PIPE_FORMAT_L4A4_UNORM: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_A8_UNORM: + return V_028C70_SWAP_ALT_REV; + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_L8_SRGB: + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8_SNORM: + return V_028C70_SWAP_STD; + + /* 16-bit buffers. */ + case PIPE_FORMAT_B5G6R5_UNORM: + return V_028C70_SWAP_STD_REV; + + case PIPE_FORMAT_B5G5R5A1_UNORM: + case PIPE_FORMAT_B5G5R5X1_UNORM: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_B4G4R4A4_UNORM: + case PIPE_FORMAT_B4G4R4X4_UNORM: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_Z16_UNORM: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_L8A8_UNORM: + case PIPE_FORMAT_L8A8_SRGB: + return V_028C70_SWAP_ALT; + case PIPE_FORMAT_R8G8_UNORM: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_R16_UNORM: + case PIPE_FORMAT_R16_FLOAT: + return V_028C70_SWAP_STD; + + /* 32-bit buffers. */ + case PIPE_FORMAT_A8B8G8R8_SRGB: + return V_028C70_SWAP_STD_REV; + case PIPE_FORMAT_B8G8R8A8_SRGB: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + return V_028C70_SWAP_ALT_REV; + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R8G8B8X8_UNORM: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_A8B8G8R8_UNORM: + case PIPE_FORMAT_X8B8G8R8_UNORM: + /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */ + return V_028C70_SWAP_STD_REV; + + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8_USCALED_Z24_UNORM: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_R10G10B10A2_UNORM: + case PIPE_FORMAT_R10G10B10X2_SNORM: + case PIPE_FORMAT_R10SG10SB10SA2U_NORM: + return V_028C70_SWAP_STD; + + case PIPE_FORMAT_B10G10R10A2_UNORM: + return V_028C70_SWAP_ALT; + + case PIPE_FORMAT_R11G11B10_FLOAT: + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_R16G16_FLOAT: + case PIPE_FORMAT_R16G16_UNORM: + return V_028C70_SWAP_STD; + + /* 64-bit buffers. */ + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R16G16B16A16_UNORM: + case PIPE_FORMAT_R16G16B16A16_SNORM: + case PIPE_FORMAT_R16G16B16A16_FLOAT: + + /* 128-bit buffers. */ + case PIPE_FORMAT_R32G32B32A32_FLOAT: + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_UNORM: + return V_028C70_SWAP_STD; + default: + R600_ERR("unsupported colorswap format %d\n", format); + return ~0U; + } + return ~0U; +} + +static uint32_t r600_translate_colorformat(enum pipe_format format) +{ + switch (format) { + /* 8-bit buffers. */ + case PIPE_FORMAT_L4A4_UNORM: + return V_028C70_COLOR_4_4; + + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_L8_SRGB: + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8_SNORM: + return V_028C70_COLOR_8; + + /* 16-bit buffers. */ + case PIPE_FORMAT_B5G6R5_UNORM: + return V_028C70_COLOR_5_6_5; + + case PIPE_FORMAT_B5G5R5A1_UNORM: + case PIPE_FORMAT_B5G5R5X1_UNORM: + return V_028C70_COLOR_1_5_5_5; + + case PIPE_FORMAT_B4G4R4A4_UNORM: + case PIPE_FORMAT_B4G4R4X4_UNORM: + return V_028C70_COLOR_4_4_4_4; + + case PIPE_FORMAT_Z16_UNORM: + return V_028C70_COLOR_16; + + case PIPE_FORMAT_L8A8_UNORM: + case PIPE_FORMAT_L8A8_SRGB: + case PIPE_FORMAT_R8G8_UNORM: + return V_028C70_COLOR_8_8; + + case PIPE_FORMAT_R16_UNORM: + return V_028C70_COLOR_16; + + case PIPE_FORMAT_R16_FLOAT: + return V_028C70_COLOR_16_FLOAT; + + /* 32-bit buffers. */ + case PIPE_FORMAT_A8B8G8R8_SRGB: + case PIPE_FORMAT_A8B8G8R8_UNORM: + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_SRGB: + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R8G8B8X8_UNORM: + case PIPE_FORMAT_R8SG8SB8UX8U_NORM: + case PIPE_FORMAT_X8B8G8R8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8_UNORM: + return V_028C70_COLOR_8_8_8_8; + + case PIPE_FORMAT_R10G10B10A2_UNORM: + case PIPE_FORMAT_R10G10B10X2_SNORM: + case PIPE_FORMAT_B10G10R10A2_UNORM: + case PIPE_FORMAT_R10SG10SB10SA2U_NORM: + return V_028C70_COLOR_2_10_10_10; + + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_028C70_COLOR_8_24; + + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8_USCALED_Z24_UNORM: + return V_028C70_COLOR_24_8; + + case PIPE_FORMAT_R32_FLOAT: + return V_028C70_COLOR_32_FLOAT; + + case PIPE_FORMAT_R16G16_FLOAT: + return V_028C70_COLOR_16_16_FLOAT; + + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16_UNORM: + return V_028C70_COLOR_16_16; + + case PIPE_FORMAT_R11G11B10_FLOAT: + return V_028C70_COLOR_10_11_11_FLOAT; + + /* 64-bit buffers. */ + case PIPE_FORMAT_R16G16B16_USCALED: + case PIPE_FORMAT_R16G16B16A16_USCALED: + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_UNORM: + case PIPE_FORMAT_R16G16B16A16_SNORM: + return V_028C70_COLOR_16_16_16_16; + + case PIPE_FORMAT_R16G16B16_FLOAT: + case PIPE_FORMAT_R16G16B16A16_FLOAT: + return V_028C70_COLOR_16_16_16_16_FLOAT; + + case PIPE_FORMAT_R32G32_FLOAT: + return V_028C70_COLOR_32_32_FLOAT; + + case PIPE_FORMAT_R32G32_USCALED: + case PIPE_FORMAT_R32G32_SSCALED: + return V_028C70_COLOR_32_32; + + /* 96-bit buffers. */ + case PIPE_FORMAT_R32G32B32_FLOAT: + return V_028C70_COLOR_32_32_32_FLOAT; + + /* 128-bit buffers. */ + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_UNORM: + return V_028C70_COLOR_32_32_32_32; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return V_028C70_COLOR_32_32_32_32_FLOAT; + + /* YUV buffers. */ + case PIPE_FORMAT_UYVY: + case PIPE_FORMAT_YUYV: + default: + return ~0U; /* Unsupported. */ + } +} + +static uint32_t r600_colorformat_endian_swap(uint32_t colorformat) +{ + if (R600_BIG_ENDIAN) { + switch(colorformat) { + case V_028C70_COLOR_4_4: + return ENDIAN_NONE; + + /* 8-bit buffers. */ + case V_028C70_COLOR_8: + return ENDIAN_NONE; + + /* 16-bit buffers. */ + case V_028C70_COLOR_5_6_5: + case V_028C70_COLOR_1_5_5_5: + case V_028C70_COLOR_4_4_4_4: + case V_028C70_COLOR_16: + case V_028C70_COLOR_8_8: + return ENDIAN_8IN16; + + /* 32-bit buffers. */ + case V_028C70_COLOR_8_8_8_8: + case V_028C70_COLOR_2_10_10_10: + case V_028C70_COLOR_8_24: + case V_028C70_COLOR_24_8: + case V_028C70_COLOR_32_FLOAT: + case V_028C70_COLOR_16_16_FLOAT: + case V_028C70_COLOR_16_16: + return ENDIAN_8IN32; + + /* 64-bit buffers. */ + case V_028C70_COLOR_16_16_16_16: + case V_028C70_COLOR_16_16_16_16_FLOAT: + return ENDIAN_8IN16; + + case V_028C70_COLOR_32_32_FLOAT: + case V_028C70_COLOR_32_32: + return ENDIAN_8IN32; + + /* 96-bit buffers. */ + case V_028C70_COLOR_32_32_32_FLOAT: + /* 128-bit buffers. */ + case V_028C70_COLOR_32_32_32_32_FLOAT: + case V_028C70_COLOR_32_32_32_32: + return ENDIAN_8IN32; + default: + return ENDIAN_NONE; /* Unsupported. */ + } + } else { + return ENDIAN_NONE; + } +} + +static bool r600_is_sampler_format_supported(struct pipe_screen *screen, enum pipe_format format) +{ + return r600_translate_texformat(screen, format, NULL, NULL, NULL) != ~0U; +} + +static bool r600_is_colorbuffer_format_supported(enum pipe_format format) +{ + return r600_translate_colorformat(format) != ~0U && + r600_translate_colorswap(format) != ~0U; +} + +static bool r600_is_zs_format_supported(enum pipe_format format) +{ + return r600_translate_dbformat(format) != ~0U; +} + +boolean evergreen_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned sample_count, + unsigned usage) +{ + unsigned retval = 0; + + if (target >= PIPE_MAX_TEXTURE_TYPES) { + R600_ERR("r600: unsupported texture type %d\n", target); + return FALSE; + } + + if (!util_format_is_supported(format, usage)) + return FALSE; + + /* Multisample */ + if (sample_count > 1) + return FALSE; + + if ((usage & PIPE_BIND_SAMPLER_VIEW) && + r600_is_sampler_format_supported(screen, format)) { + retval |= PIPE_BIND_SAMPLER_VIEW; + } + + if ((usage & (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED)) && + r600_is_colorbuffer_format_supported(format)) { + retval |= usage & + (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED); + } + + if ((usage & PIPE_BIND_DEPTH_STENCIL) && + r600_is_zs_format_supported(format)) { + retval |= PIPE_BIND_DEPTH_STENCIL; + } + + if ((usage & PIPE_BIND_VERTEX_BUFFER) && + r600_is_vertex_format_supported(format)) { + retval |= PIPE_BIND_VERTEX_BUFFER; + } + + if (usage & PIPE_BIND_TRANSFER_READ) + retval |= PIPE_BIND_TRANSFER_READ; + if (usage & PIPE_BIND_TRANSFER_WRITE) + retval |= PIPE_BIND_TRANSFER_WRITE; + + return retval == usage; +} static void evergreen_set_blend_color(struct pipe_context *ctx, const struct pipe_blend_color *state) @@ -77,13 +657,11 @@ static void *evergreen_create_blend_state(struct pipe_context *ctx, u32 color_control, target_mask; /* FIXME there is more then 8 framebuffer */ unsigned blend_cntl[8]; - enum radeon_family family; if (blend == NULL) { return NULL; } - family = r600_get_family(rctx->radeon); rstate = &blend->rstate; rstate->id = R600_PIPE_STATE_BLEND; @@ -110,7 +688,7 @@ static void *evergreen_create_blend_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028808_CB_COLOR_CONTROL, color_control, 0xFFFFFFFD, NULL); - if (family != CHIP_CAYMAN) + if (rctx->chip_class != CAYMAN) r600_pipe_state_add_reg(rstate, R_028C3C_PA_SC_AA_MASK, 0xFFFFFFFF, 0xFFFFFFFF, NULL); else { r600_pipe_state_add_reg(rstate, CM_R_028C38_PA_SC_AA_MASK_X0Y0_X1Y0, 0xFFFFFFFF, 0xFFFFFFFF, NULL); @@ -247,9 +825,6 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, unsigned tmp; unsigned prov_vtx = 1, polygon_dual_mode; unsigned clip_rule; - enum radeon_family family; - - family = r600_get_family(rctx->radeon); if (rs == NULL) { return NULL; @@ -308,7 +883,7 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, tmp = (unsigned)state->line_width * 8; r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL); - if (family == CHIP_CAYMAN) { + if (rctx->chip_class == CAYMAN) { r600_pipe_state_add_reg(rstate, CM_R_028BDC_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, CM_R_028BE4_PA_SU_VTX_CNTL, S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules), @@ -867,14 +1442,11 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state *rstate = CALLOC_STRUCT(r600_pipe_state); u32 shader_mask, tl, br, target_mask; - enum radeon_family family; int tl_x, tl_y, br_x, br_y; if (rstate == NULL) return; - family = r600_get_family(rctx->radeon); - evergreen_context_flush_dest_caches(&rctx->ctx); rctx->ctx.num_dest_buffers = state->nr_cbufs; @@ -911,7 +1483,7 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, if (br_y == 0) tl_y = 1; /* cayman hw workaround */ - if (family == CHIP_CAYMAN) { + if (rctx->chip_class == CAYMAN) { if (br_x == 1 && br_y == 1) br_x = 2; } @@ -955,7 +1527,7 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, shader_mask, 0xFFFFFFFF, NULL); - if (family == CHIP_CAYMAN) { + if (rctx->chip_class == CAYMAN) { r600_pipe_state_add_reg(rstate, CM_R_028BE0_PA_SC_AA_CONFIG, 0x00000000, 0xFFFFFFFF, NULL); } else { @@ -1142,9 +1714,9 @@ void evergreen_init_config(struct r600_pipe_context *rctx) enum radeon_family family; unsigned tmp; - family = r600_get_family(rctx->radeon); + family = rctx->family; - if (family == CHIP_CAYMAN) { + if (rctx->chip_class == CAYMAN) { cayman_init_config(rctx); return; } diff --git a/src/gallium/drivers/r600/r600.h b/src/gallium/drivers/r600/r600.h index 151e831e5c6..2af4d311f60 100644 --- a/src/gallium/drivers/r600/r600.h +++ b/src/gallium/drivers/r600/r600.h @@ -228,6 +228,7 @@ struct r600_query { #define R600_QUERY_STATE_STARTED (1 << 0) #define R600_QUERY_STATE_ENDED (1 << 1) #define R600_QUERY_STATE_SUSPENDED (1 << 2) +#define R600_QUERY_STATE_FLUSHED (1 << 3) #define R600_CONTEXT_DRAW_PENDING (1 << 0) #define R600_CONTEXT_DST_CACHES_DIRTY (1 << 1) @@ -294,7 +295,7 @@ boolean r600_context_query_result(struct r600_context *ctx, void r600_query_begin(struct r600_context *ctx, struct r600_query *query); void r600_query_end(struct r600_context *ctx, struct r600_query *query); void r600_context_queries_suspend(struct r600_context *ctx); -void r600_context_queries_resume(struct r600_context *ctx); +void r600_context_queries_resume(struct r600_context *ctx, boolean flushed); void r600_query_predication(struct r600_context *ctx, struct r600_query *query, int operation, int flag_wait); void r600_context_emit_fence(struct r600_context *ctx, struct r600_bo *fence, diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c index 065f955ebcb..5fae2b00c8b 100644 --- a/src/gallium/drivers/r600/r600_asm.c +++ b/src/gallium/drivers/r600/r600_asm.c @@ -41,9 +41,9 @@ static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r if(alu->is_op3) return 3; - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: switch (alu->inst) { case V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP: return 0; @@ -93,8 +93,8 @@ static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r "Need instruction operand number for 0x%x.\n", alu->inst); } break; - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: switch (alu->inst) { case EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP: return 0; @@ -195,48 +195,10 @@ static struct r600_bc_tex *r600_bc_tex(void) return tex; } -int r600_bc_init(struct r600_bc *bc, enum radeon_family family) +void r600_bc_init(struct r600_bc *bc, enum chip_class chip_class) { LIST_INITHEAD(&bc->cf); - bc->family = family; - switch (bc->family) { - case CHIP_R600: - case CHIP_RV610: - case CHIP_RV630: - case CHIP_RV670: - case CHIP_RV620: - case CHIP_RV635: - case CHIP_RS780: - case CHIP_RS880: - bc->chiprev = CHIPREV_R600; - break; - case CHIP_RV770: - case CHIP_RV730: - case CHIP_RV710: - case CHIP_RV740: - bc->chiprev = CHIPREV_R700; - break; - case CHIP_CEDAR: - case CHIP_REDWOOD: - case CHIP_JUNIPER: - case CHIP_CYPRESS: - case CHIP_HEMLOCK: - case CHIP_PALM: - case CHIP_SUMO: - case CHIP_SUMO2: - case CHIP_BARTS: - case CHIP_TURKS: - case CHIP_CAICOS: - bc->chiprev = CHIPREV_EVERGREEN; - break; - case CHIP_CAYMAN: - bc->chiprev = CHIPREV_CAYMAN; - break; - default: - R600_ERR("unknown family %d\n", bc->family); - return -EINVAL; - } - return 0; + bc->chip_class = chip_class; } static int r600_bc_add_cf(struct r600_bc *bc) @@ -301,9 +263,9 @@ int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output) /* alu instructions that can ony exits once per group */ static int is_alu_once_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: return !alu->is_op3 && ( alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLE || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT || @@ -339,8 +301,8 @@ static int is_alu_once_inst(struct r600_bc *bc, struct r600_bc_alu *alu) alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE_PUSH_INT || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETLT_PUSH_INT || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETLE_PUSH_INT); - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: default: return !alu->is_op3 && ( alu->inst == EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLE || @@ -382,16 +344,16 @@ static int is_alu_once_inst(struct r600_bc *bc, struct r600_bc_alu *alu) static int is_alu_reduction_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: return !alu->is_op3 && ( alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4 || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4_IEEE || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX4); - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: default: return !alu->is_op3 && ( alu->inst == EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE || @@ -403,13 +365,13 @@ static int is_alu_reduction_inst(struct r600_bc *bc, struct r600_bc_alu *alu) static int is_alu_cube_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: return !alu->is_op3 && alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE; - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: default: return !alu->is_op3 && alu->inst == EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE; @@ -418,15 +380,15 @@ static int is_alu_cube_inst(struct r600_bc *bc, struct r600_bc_alu *alu) static int is_alu_mova_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: return !alu->is_op3 && ( alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT); - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: default: return !alu->is_op3 && ( alu->inst == EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT); @@ -438,16 +400,16 @@ static int is_alu_vec_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { return is_alu_reduction_inst(bc, alu) || is_alu_mova_inst(bc, alu) || - (bc->chiprev == CHIPREV_EVERGREEN && + (bc->chip_class == EVERGREEN && alu->inst == EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR); } /* alu instructions that can only execute on the trans unit */ static int is_alu_trans_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) { - switch (bc->chiprev) { - case CHIPREV_R600: - case CHIPREV_R700: + switch (bc->chip_class) { + case R600: + case R700: if (!alu->is_op3) return alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ASHR_INT || alu->inst == V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT || @@ -478,8 +440,8 @@ static int is_alu_trans_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) alu->inst == V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_D2 || alu->inst == V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_M2 || alu->inst == V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_M4; - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: default: if (!alu->is_op3) /* Note that FLT_TO_INT_* instructions are vector-only instructions @@ -525,7 +487,7 @@ static int assign_alu_units(struct r600_bc *bc, struct r600_bc_alu *alu_first, { struct r600_bc_alu *alu; unsigned i, chan, trans; - int max_slots = bc->chiprev == CHIPREV_CAYMAN ? 4 : 5; + int max_slots = bc->chip_class == CAYMAN ? 4 : 5; for (i = 0; i < max_slots; i++) assignment[i] = NULL; @@ -612,7 +574,7 @@ static int reserve_gpr(struct alu_bank_swizzle *bs, unsigned sel, unsigned chan, static int reserve_cfile(struct r600_bc *bc, struct alu_bank_swizzle *bs, unsigned sel, unsigned chan) { int res, num_res = 4; - if (bc->chiprev >= CHIPREV_R700) { + if (bc->chip_class >= R700) { num_res = 2; chan /= 2; } @@ -733,8 +695,8 @@ static int check_and_set_bank_swizzle(struct r600_bc *bc, struct alu_bank_swizzle bs; int bank_swizzle[5]; int i, r = 0, forced = 0; - boolean scalar_only = bc->chiprev == CHIPREV_CAYMAN ? false : true; - int max_slots = bc->chiprev == CHIPREV_CAYMAN ? 4 : 5; + boolean scalar_only = bc->chip_class == CAYMAN ? false : true; + int max_slots = bc->chip_class == CAYMAN ? 4 : 5; for (i = 0; i < max_slots; i++) { if (slots[i] && slots[i]->bank_swizzle_force) { @@ -806,7 +768,7 @@ static int replace_gpr_with_pv_ps(struct r600_bc *bc, struct r600_bc_alu *prev[5]; int gpr[5], chan[5]; int i, j, r, src, num_src; - int max_slots = bc->chiprev == CHIPREV_CAYMAN ? 4 : 5; + int max_slots = bc->chip_class == CAYMAN ? 4 : 5; r = assign_alu_units(bc, alu_prev, prev); if (r) @@ -834,7 +796,7 @@ static int replace_gpr_with_pv_ps(struct r600_bc *bc, if (!is_gpr(alu->src[src].sel) || alu->src[src].rel) continue; - if (bc->chiprev < CHIPREV_CAYMAN) { + if (bc->chip_class < CAYMAN) { if (alu->src[src].sel == gpr[4] && alu->src[src].chan == chan[4]) { alu->src[src].sel = V_SQ_ALU_SRC_PS; @@ -948,7 +910,7 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], int i, j, r, src, num_src; int num_once_inst = 0; int have_mova = 0, have_rel = 0; - int max_slots = bc->chiprev == CHIPREV_CAYMAN ? 4 : 5; + int max_slots = bc->chip_class == CAYMAN ? 4 : 5; r = assign_alu_units(bc, alu_prev, prev); if (r) @@ -1252,7 +1214,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int uint32_t literal[4]; unsigned nliteral; struct r600_bc_alu *slots[5]; - int max_slots = bc->chiprev == CHIPREV_CAYMAN ? 4 : 5; + int max_slots = bc->chip_class == CAYMAN ? 4 : 5; r = assign_alu_units(bc, bc->cf_last->curr_bs_head, slots); if (r) return r; @@ -1302,26 +1264,26 @@ int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu) static unsigned r600_bc_num_tex_and_vtx_instructions(const struct r600_bc *bc) { - switch (bc->chiprev) { - case CHIPREV_R600: + switch (bc->chip_class) { + case R600: return 8; - case CHIPREV_R700: + case R700: return 16; - case CHIPREV_EVERGREEN: - case CHIPREV_CAYMAN: + case EVERGREEN: + case CAYMAN: return 64; default: - R600_ERR("Unknown chiprev %d.\n", bc->chiprev); + R600_ERR("Unknown chip class %d.\n", bc->chip_class); return 8; } } static inline boolean last_inst_was_vtx_fetch(struct r600_bc *bc) { - if (bc->chiprev == CHIPREV_CAYMAN) { + if (bc->chip_class == CAYMAN) { if (bc->cf_last->inst != CM_V_SQ_CF_WORD1_SQ_CF_INST_TC) return TRUE; } else { @@ -1350,7 +1312,7 @@ int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx) free(nvtx); return r; } - if (bc->chiprev == CHIPREV_CAYMAN) + if (bc->chip_class == CAYMAN) bc->cf_last->inst = CM_V_SQ_CF_WORD1_SQ_CF_INST_TC; else bc->cf_last->inst = V_SQ_CF_WORD1_SQ_CF_INST_VTX; @@ -1438,7 +1400,7 @@ static int r600_bc_vtx_build(struct r600_bc *bc, struct r600_bc_vtx *vtx, unsign S_SQ_VTX_WORD0_FETCH_TYPE(vtx->fetch_type) | S_SQ_VTX_WORD0_SRC_GPR(vtx->src_gpr) | S_SQ_VTX_WORD0_SRC_SEL_X(vtx->src_sel_x); - if (bc->chiprev < CHIPREV_CAYMAN) + if (bc->chip_class < CAYMAN) bc->bytecode[id] |= S_SQ_VTX_WORD0_MEGA_FETCH_COUNT(vtx->mega_fetch_count); id++; bc->bytecode[id++] = S_SQ_VTX_WORD1_DST_SEL_X(vtx->dst_sel_x) | @@ -1453,7 +1415,7 @@ static int r600_bc_vtx_build(struct r600_bc *bc, struct r600_bc_vtx *vtx, unsign S_SQ_VTX_WORD1_GPR_DST_GPR(vtx->dst_gpr); bc->bytecode[id] = S_SQ_VTX_WORD2_OFFSET(vtx->offset)| S_SQ_VTX_WORD2_ENDIAN_SWAP(vtx->endian); - if (bc->chiprev < CHIPREV_CAYMAN) + if (bc->chip_class < CAYMAN) bc->bytecode[id] |= S_SQ_VTX_WORD2_MEGA_FETCH(1); id++; bc->bytecode[id++] = 0; @@ -1560,13 +1522,13 @@ static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) S_SQ_CF_ALU_WORD1_KCACHE_ADDR0(cf->kcache[0].addr) | S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(cf->kcache[1].addr) | S_SQ_CF_ALU_WORD1_BARRIER(1) | - S_SQ_CF_ALU_WORD1_USES_WATERFALL(bc->chiprev == CHIPREV_R600 ? cf->r6xx_uses_waterfall : 0) | + S_SQ_CF_ALU_WORD1_USES_WATERFALL(bc->chip_class == R600 ? cf->r6xx_uses_waterfall : 0) | S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1); break; case V_SQ_CF_WORD1_SQ_CF_INST_TEX: case V_SQ_CF_WORD1_SQ_CF_INST_VTX: case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC: - if (bc->chiprev == CHIPREV_R700) + if (bc->chip_class == R700) r700_bc_cf_vtx_build(&bc->bytecode[id], cf); else r600_bc_cf_vtx_build(&bc->bytecode[id], cf); @@ -1673,7 +1635,7 @@ int r600_bc_build(struct r600_bc *bc) return -ENOMEM; LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) { addr = cf->addr; - if (bc->chiprev >= CHIPREV_EVERGREEN) + if (bc->chip_class >= EVERGREEN) r = eg_bc_cf_build(bc, cf); else r = r600_bc_cf_build(bc, cf); @@ -1691,17 +1653,17 @@ int r600_bc_build(struct r600_bc *bc) if (r) return r; r600_bc_alu_adjust_literals(bc, alu, literal, nliteral); - switch(bc->chiprev) { - case CHIPREV_R600: + switch(bc->chip_class) { + case R600: r = r600_bc_alu_build(bc, alu, addr); break; - case CHIPREV_R700: - case CHIPREV_EVERGREEN: /* eg alu is same encoding as r700 */ - case CHIPREV_CAYMAN: /* eg alu is same encoding as r700 */ + case R700: + case EVERGREEN: /* eg alu is same encoding as r700 */ + case CAYMAN: /* eg alu is same encoding as r700 */ r = r700_bc_alu_build(bc, alu, addr); break; default: - R600_ERR("unknown family %d\n", bc->family); + R600_ERR("unknown chip class %d.\n", bc->chip_class); return -EINVAL; } if (r) @@ -1726,7 +1688,7 @@ int r600_bc_build(struct r600_bc *bc) } break; case V_SQ_CF_WORD1_SQ_CF_INST_TEX: - if (bc->chiprev == CHIPREV_CAYMAN) { + if (bc->chip_class == CAYMAN) { LIST_FOR_EACH_ENTRY(vtx, &cf->vtx, list) { r = r600_bc_vtx_build(bc, vtx, addr); if (r) @@ -1812,17 +1774,17 @@ void r600_bc_dump(struct r600_bc *bc) unsigned nliteral; char chip = '6'; - switch (bc->chiprev) { - case 1: + switch (bc->chip_class) { + case R700: chip = '7'; break; - case 2: + case EVERGREEN: chip = 'E'; break; - case 3: + case CAYMAN: chip = 'C'; break; - case 0: + case R600: default: chip = '6'; break; @@ -1993,7 +1955,7 @@ void r600_bc_dump(struct r600_bc *bc) fprintf(stderr, "%04d %08X ", id, bc->bytecode[id]); fprintf(stderr, "SRC(GPR:%d ", vtx->src_gpr); fprintf(stderr, "SEL_X:%d) ", vtx->src_sel_x); - if (bc->chiprev < CHIPREV_CAYMAN) + if (bc->chip_class < CAYMAN) fprintf(stderr, "MEGA_FETCH_COUNT:%d ", vtx->mega_fetch_count); else fprintf(stderr, "SEL_Y:%d) ", 0); @@ -2162,7 +2124,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru struct r600_bc_vtx vtx; struct pipe_vertex_element *elements = ve->elements; const struct util_format_description *desc; - unsigned fetch_resource_start = rctx->family >= CHIP_CEDAR ? 0 : 160; + unsigned fetch_resource_start = rctx->chip_class >= EVERGREEN ? 0 : 160; unsigned format, num_format, format_comp, endian; u32 *bytecode; int i, r; @@ -2180,9 +2142,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru } memset(&bc, 0, sizeof(bc)); - r = r600_bc_init(&bc, r600_get_family(rctx->radeon)); - if (r) - return r; + r600_bc_init(&bc, rctx->chip_class); for (i = 0; i < ve->count; i++) { if (elements[i].instance_divisor > 1) { @@ -2287,7 +2247,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru r600_bo_unmap(rctx->radeon, ve->fetch_shader); r600_bc_clear(&bc); - if (rctx->family >= CHIP_CEDAR) + if (rctx->chip_class >= EVERGREEN) evergreen_fetch_shader(&rctx->context, ve); else r600_fetch_shader(&rctx->context, ve); diff --git a/src/gallium/drivers/r600/r600_asm.h b/src/gallium/drivers/r600/r600_asm.h index 540f45bbd06..cbdaacf7178 100644 --- a/src/gallium/drivers/r600/r600_asm.h +++ b/src/gallium/drivers/r600/r600_asm.h @@ -171,8 +171,7 @@ struct r600_cf_callstack { }; struct r600_bc { - enum radeon_family family; - int chiprev; /* 0 - r600, 1 - r700, 2 - evergreen */ + enum chip_class chip_class; int type; struct list_head cf; struct r600_bc_cf *cf_last; @@ -193,7 +192,7 @@ struct r600_bc { int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf); /* r600_asm.c */ -int r600_bc_init(struct r600_bc *bc, enum radeon_family family); +void r600_bc_init(struct r600_bc *bc, enum chip_class chip_class); void r600_bc_clear(struct r600_bc *bc); int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu); int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx); diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c index 6171d285bb9..35e68b6e222 100644 --- a/src/gallium/drivers/r600/r600_blit.c +++ b/src/gallium/drivers/r600/r600_blit.c @@ -97,7 +97,7 @@ static void r600_blitter_end(struct pipe_context *ctx) rctx->saved_render_cond_mode); rctx->saved_render_cond = NULL; } - r600_context_queries_resume(&rctx->ctx); + r600_context_queries_resume(&rctx->ctx, FALSE); rctx->blit = false; } diff --git a/src/gallium/drivers/r600/r600_formats.h b/src/gallium/drivers/r600/r600_formats.h index ae0bc432ad2..1c1089d89d2 100644 --- a/src/gallium/drivers/r600/r600_formats.h +++ b/src/gallium/drivers/r600/r600_formats.h @@ -81,4 +81,36 @@ static INLINE unsigned r600_endian_swap(unsigned size) } } +static INLINE bool r600_is_vertex_format_supported(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + unsigned i; + + if (!desc) + return false; + + /* Find the first non-VOID channel. */ + for (i = 0; i < 4; i++) { + if (desc->channel[i].type != UTIL_FORMAT_TYPE_VOID) + break; + } + if (i == 4) + return false; + + /* No fixed, no double. */ + if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN || + desc->channel[i].type == UTIL_FORMAT_TYPE_FIXED || + (desc->channel[i].size == 64 && + desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT)) + return false; + + /* No scaled/norm formats with 32 bits per channel. */ + if (desc->channel[i].size == 32 && + (desc->channel[i].type == UTIL_FORMAT_TYPE_SIGNED || + desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED)) + return false; + + return true; +} + #endif diff --git a/src/gallium/drivers/r600/r600_opcodes.h b/src/gallium/drivers/r600/r600_opcodes.h index 184f32c9960..7ae091ea5cd 100644 --- a/src/gallium/drivers/r600/r600_opcodes.h +++ b/src/gallium/drivers/r600/r600_opcodes.h @@ -409,14 +409,8 @@ #define EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_EXPORT_COMBINED 0x0000005B #define EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_RAT_COMBINED_CACHELESS 0x0000005C +#define BC_INST(bc, x) ((bc)->chip_class >= EVERGREEN ? EG_##x : x) -#define CHIPREV_R600 0 -#define CHIPREV_R700 1 -#define CHIPREV_EVERGREEN 2 -#define CHIPREV_CAYMAN 3 - -#define BC_INST(bc, x) ((bc)->chiprev >= CHIPREV_EVERGREEN ? EG_##x : x) - -#define CTX_INST(x) (ctx->bc->chiprev >= CHIPREV_EVERGREEN ? EG_##x : x) +#define CTX_INST(x) (ctx->bc->chip_class >= EVERGREEN ? EG_##x : x) #endif diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c index 76bb1883ede..8e492787235 100644 --- a/src/gallium/drivers/r600/r600_pipe.c +++ b/src/gallium/drivers/r600/r600_pipe.c @@ -47,7 +47,6 @@ #include "r600_resource.h" #include "r600_shader.h" #include "r600_pipe.h" -#include "r600_state_inlines.h" /* * pipe_context @@ -197,7 +196,6 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void { struct r600_pipe_context *rctx = CALLOC_STRUCT(r600_pipe_context); struct r600_screen* rscreen = (struct r600_screen *)screen; - enum chip_class class; if (rctx == NULL) return NULL; @@ -214,6 +212,7 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void rctx->screen = rscreen; rctx->radeon = rscreen->radeon; rctx->family = r600_get_family(rctx->radeon); + rctx->chip_class = r600_get_family_class(rctx->radeon); rctx->fences.bo = NULL; rctx->fences.data = NULL; @@ -230,47 +229,29 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void rctx->context.create_video_decoder = vl_create_decoder; rctx->context.create_video_buffer = vl_video_buffer_create; - switch (r600_get_family(rctx->radeon)) { - case CHIP_R600: - case CHIP_RV610: - case CHIP_RV630: - case CHIP_RV670: - case CHIP_RV620: - case CHIP_RV635: - case CHIP_RS780: - case CHIP_RS880: - case CHIP_RV770: - case CHIP_RV730: - case CHIP_RV710: - case CHIP_RV740: + switch (rctx->chip_class) { + case R600: + case R700: r600_init_state_functions(rctx); if (r600_context_init(&rctx->ctx, rctx->radeon)) { r600_destroy_context(&rctx->context); return NULL; } r600_init_config(rctx); + rctx->custom_dsa_flush = r600_create_db_flush_dsa(rctx); break; - case CHIP_CEDAR: - case CHIP_REDWOOD: - case CHIP_JUNIPER: - case CHIP_CYPRESS: - case CHIP_HEMLOCK: - case CHIP_PALM: - case CHIP_SUMO: - case CHIP_SUMO2: - case CHIP_BARTS: - case CHIP_TURKS: - case CHIP_CAICOS: - case CHIP_CAYMAN: + case EVERGREEN: + case CAYMAN: evergreen_init_state_functions(rctx); if (evergreen_context_init(&rctx->ctx, rctx->radeon)) { r600_destroy_context(&rctx->context); return NULL; } evergreen_init_config(rctx); + rctx->custom_dsa_flush = evergreen_create_db_flush_dsa(rctx); break; default: - R600_ERR("unsupported family %d\n", r600_get_family(rctx->radeon)); + R600_ERR("Unsupported chip class %d.\n", rctx->chip_class); r600_destroy_context(&rctx->context); return NULL; } @@ -295,12 +276,6 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void return NULL; } - class = r600_get_family_class(rctx->radeon); - if (class == R600 || class == R700) - rctx->custom_dsa_flush = r600_create_db_flush_dsa(rctx); - else - rctx->custom_dsa_flush = evergreen_create_db_flush_dsa(rctx); - return &rctx->context; } @@ -523,64 +498,6 @@ static int r600_get_video_param(struct pipe_screen *screen, } } -static boolean r600_is_format_supported(struct pipe_screen* screen, - enum pipe_format format, - enum pipe_texture_target target, - unsigned sample_count, - unsigned usage) -{ - unsigned retval = 0; - if (target >= PIPE_MAX_TEXTURE_TYPES) { - R600_ERR("r600: unsupported texture type %d\n", target); - return FALSE; - } - - if (!util_format_is_supported(format, usage)) - return FALSE; - - /* Multisample */ - if (sample_count > 1) - return FALSE; - - if ((usage & PIPE_BIND_SAMPLER_VIEW) && - r600_is_sampler_format_supported(screen, format)) { - retval |= PIPE_BIND_SAMPLER_VIEW; - } - - if ((usage & (PIPE_BIND_RENDER_TARGET | - PIPE_BIND_DISPLAY_TARGET | - PIPE_BIND_SCANOUT | - PIPE_BIND_SHARED)) && - r600_is_colorbuffer_format_supported(format)) { - retval |= usage & - (PIPE_BIND_RENDER_TARGET | - PIPE_BIND_DISPLAY_TARGET | - PIPE_BIND_SCANOUT | - PIPE_BIND_SHARED); - } - - if ((usage & PIPE_BIND_DEPTH_STENCIL) && - r600_is_zs_format_supported(format)) { - retval |= PIPE_BIND_DEPTH_STENCIL; - } - - if (usage & PIPE_BIND_VERTEX_BUFFER) { - struct r600_screen *rscreen = (struct r600_screen *)screen; - enum radeon_family family = r600_get_family(rscreen->radeon); - - if (r600_is_vertex_format_supported(format, family)) { - retval |= PIPE_BIND_VERTEX_BUFFER; - } - } - - if (usage & PIPE_BIND_TRANSFER_READ) - retval |= PIPE_BIND_TRANSFER_READ; - if (usage & PIPE_BIND_TRANSFER_WRITE) - retval |= PIPE_BIND_TRANSFER_WRITE; - - return retval == usage; -} - static void r600_destroy_screen(struct pipe_screen* pscreen) { struct r600_screen *rscreen = (struct r600_screen *)pscreen; @@ -670,7 +587,11 @@ struct pipe_screen *r600_screen_create(struct radeon *radeon) rscreen->screen.get_shader_param = r600_get_shader_param; rscreen->screen.get_paramf = r600_get_paramf; rscreen->screen.get_video_param = r600_get_video_param; - rscreen->screen.is_format_supported = r600_is_format_supported; + if (r600_get_family_class(radeon) >= EVERGREEN) { + rscreen->screen.is_format_supported = evergreen_is_format_supported; + } else { + rscreen->screen.is_format_supported = r600_is_format_supported; + } rscreen->screen.is_video_format_supported = vl_video_buffer_is_format_supported; rscreen->screen.context_create = r600_create_context; rscreen->screen.fence_reference = r600_fence_reference; diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h index 2667c80bcef..6f399ed43b0 100644 --- a/src/gallium/drivers/r600/r600_pipe.h +++ b/src/gallium/drivers/r600/r600_pipe.h @@ -175,7 +175,8 @@ struct r600_pipe_fences { struct r600_pipe_context { struct pipe_context context; struct blitter_context *blitter; - unsigned family; + enum radeon_family family; + enum chip_class chip_class; void *custom_dsa_flush; struct r600_screen *screen; struct radeon *radeon; @@ -247,6 +248,11 @@ void evergreen_pipe_init_buffer_resource(struct r600_pipe_context *rctx, void evergreen_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, unsigned offset, unsigned stride); +boolean evergreen_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned sample_count, + unsigned usage); /* r600_blit.c */ void r600_init_blit_functions(struct r600_pipe_context *rctx); @@ -290,6 +296,11 @@ void r600_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, unsigned offset, unsigned stride); void r600_adjust_gprs(struct r600_pipe_context *rctx); +boolean r600_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned sample_count, + unsigned usage); /* r600_texture.c */ void r600_init_screen_texture_functions(struct pipe_screen *screen); diff --git a/src/gallium/drivers/r600/r600_query.c b/src/gallium/drivers/r600/r600_query.c index bedb48b6031..174505c75e9 100644 --- a/src/gallium/drivers/r600/r600_query.c +++ b/src/gallium/drivers/r600/r600_query.c @@ -61,10 +61,7 @@ static boolean r600_get_query_result(struct pipe_context *ctx, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_query *rquery = (struct r600_query *)query; - if (rquery->num_results) { - ctx->flush(ctx, NULL); - } - return r600_context_query_result(&rctx->ctx, (struct r600_query *)query, wait, vresult); + return r600_context_query_result(&rctx->ctx, rquery, wait, vresult); } static void r600_render_condition(struct pipe_context *ctx, diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c index f83d7079b29..3e21ad1fdc6 100644 --- a/src/gallium/drivers/r600/r600_shader.c +++ b/src/gallium/drivers/r600/r600_shader.c @@ -99,14 +99,14 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *s /* build state */ switch (rshader->processor_type) { case TGSI_PROCESSOR_VERTEX: - if (rshader->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_shader_vs(ctx, shader); } else { r600_pipe_shader_vs(ctx, shader); } break; case TGSI_PROCESSOR_FRAGMENT: - if (rshader->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_shader_ps(ctx, shader); } else { r600_pipe_shader_ps(ctx, shader); @@ -135,7 +135,6 @@ int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *s fprintf(stderr, "--------------------------------------------------------------\n"); tgsi_dump(shader->tokens, 0); } - shader->shader.family = r600_get_family(rctx->radeon); r = r600_shader_from_tgsi(rctx, shader); if (r) { R600_ERR("translation from TGSI failed !\n"); @@ -317,7 +316,7 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) ctx->shader->input[i].interpolate = d->Declaration.Interpolate; ctx->shader->input[i].centroid = d->Declaration.Centroid; ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + i; - if (ctx->type == TGSI_PROCESSOR_FRAGMENT && ctx->bc->chiprev >= CHIPREV_EVERGREEN) { + if (ctx->type == TGSI_PROCESSOR_FRAGMENT && ctx->bc->chip_class >= EVERGREEN) { /* turn input into interpolate on EG */ if (ctx->shader->input[i].name != TGSI_SEMANTIC_POSITION) { if (ctx->shader->input[i].interpolate > 0) { @@ -610,9 +609,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi ctx.bc = &shader->bc; ctx.shader = shader; - r = r600_bc_init(ctx.bc, shader->family); - if (r) - return r; + r600_bc_init(ctx.bc, rctx->chip_class); ctx.tokens = tokens; tgsi_scan_shader(tokens, &ctx.info); tgsi_parse_init(&ctx.parse, tokens); @@ -651,13 +648,13 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi } if (ctx.type == TGSI_PROCESSOR_VERTEX) { ctx.file_offset[TGSI_FILE_INPUT] = 1; - if (ctx.bc->chiprev >= CHIPREV_EVERGREEN) { + if (ctx.bc->chip_class >= EVERGREEN) { r600_bc_add_cfinst(ctx.bc, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); } else { r600_bc_add_cfinst(ctx.bc, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); } } - if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chiprev >= CHIPREV_EVERGREEN) { + if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chip_class >= EVERGREEN) { ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx); } ctx.file_offset[TGSI_FILE_OUTPUT] = ctx.file_offset[TGSI_FILE_INPUT] + @@ -711,9 +708,9 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi goto out_err; if ((r = tgsi_split_literal_constant(&ctx))) goto out_err; - if (ctx.bc->chiprev == CHIPREV_CAYMAN) + if (ctx.bc->chip_class == CAYMAN) ctx.inst_info = &cm_shader_tgsi_instruction[opcode]; - else if (ctx.bc->chiprev >= CHIPREV_EVERGREEN) + else if (ctx.bc->chip_class >= EVERGREEN) ctx.inst_info = &eg_shader_tgsi_instruction[opcode]; else ctx.inst_info = &r600_shader_tgsi_instruction[opcode]; @@ -802,7 +799,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi if (shader->output[i].name == TGSI_SEMANTIC_COLOR) { output[i + j].array_base = shader->output[i].sid; output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL; - if (shader->fs_write_all && (shader->family >= CHIP_CEDAR)) { + if (shader->fs_write_all && (rctx->chip_class >= EVERGREEN)) { for (j = 1; j < shader->nr_cbufs; j++) { memset(&output[i + j], 0, sizeof(struct r600_bc_output)); output[i + j].gpr = shader->output[i].gpr; @@ -886,7 +883,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi } /* set export done on last export of each type */ for (i = noutput - 1, output_done = 0; i >= 0; i--) { - if (ctx.bc->chiprev < CHIPREV_CAYMAN) { + if (ctx.bc->chip_class < CAYMAN) { if (i == (noutput - 1)) { output[i].end_of_program = 1; } @@ -903,7 +900,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi goto out_err; } /* add program end */ - if (ctx.bc->chiprev == CHIPREV_CAYMAN) + if (ctx.bc->chip_class == CAYMAN) cm_bc_add_cf_end(ctx.bc); free(ctx.literals); @@ -939,6 +936,17 @@ static void r600_bc_src(struct r600_bc_alu_src *bc_src, bc_src->value = shader_src->value[bc_src->chan]; } +static void r600_bc_src_set_abs(struct r600_bc_alu_src *bc_src) +{ + bc_src->abs = 1; + bc_src->neg = 0; +} + +static void r600_bc_src_toggle_neg(struct r600_bc_alu_src *bc_src) +{ + bc_src->neg = !bc_src->neg; +} + static void tgsi_dst(struct r600_shader_ctx *ctx, const struct tgsi_full_dst_register *tgsi_dst, unsigned swizzle, @@ -995,12 +1003,10 @@ static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap) /* handle some special cases */ switch (ctx->inst_info->tgsi_opcode) { case TGSI_OPCODE_SUB: - alu.src[1].neg = 1; + r600_bc_src_toggle_neg(&alu.src[1]); break; case TGSI_OPCODE_ABS: - alu.src[0].abs = 1; - if (alu.src[0].neg) - alu.src[0].neg = 0; + r600_bc_src_set_abs(&alu.src[0]); break; default: break; @@ -1114,7 +1120,7 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) alu.src[2].sel = V_SQ_ALU_SRC_LITERAL; alu.src[2].chan = 0; - if (ctx->bc->chiprev == CHIPREV_R600) { + if (ctx->bc->chip_class == R600) { alu.src[1].value = *(uint32_t *)&double_pi; alu.src[2].value = *(uint32_t *)&neg_pi; } else { @@ -1221,7 +1227,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) /* dst.x = COS */ if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0 ; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS); @@ -1255,7 +1261,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) /* dst.y = SIN */ if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0 ; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN); @@ -1364,19 +1370,37 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) struct r600_bc_alu alu; int r; + /* tmp.x = max(src.y, 0.0) */ + memset(&alu, 0, sizeof(struct r600_bc_alu)); + alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX); + r600_bc_src(&alu.src[0], &ctx->src[0], 1); + alu.src[1].sel = V_SQ_ALU_SRC_0; /*0.0*/ + alu.src[1].chan = 1; + + alu.dst.sel = ctx->temp_reg; + alu.dst.chan = 0; + alu.dst.write = 1; + + alu.last = 1; + r = r600_bc_add_alu(ctx->bc, &alu); + if (r) + return r; + if (inst->Dst[0].Register.WriteMask & (1 << 2)) { int chan; int sel; int i; - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - /* dst.z = log(src.y) */ + /* tmp.z = log(tmp.x) */ memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED); - r600_bc_src(&alu.src[0], &ctx->src[0], 1); - tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); + alu.src[0].sel = ctx->temp_reg; + alu.src[0].chan = 0; + alu.dst.sel = ctx->temp_reg; + alu.dst.chan = i; if (i == 2) { alu.dst.write = 1; alu.last = 1; @@ -1388,10 +1412,11 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) return r; } } else { - /* dst.z = log(src.y) */ + /* tmp.z = log(tmp.x) */ memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED); - r600_bc_src(&alu.src[0], &ctx->src[0], 1); + alu.src[0].sel = ctx->temp_reg; + alu.src[0].chan = 0; alu.dst.sel = ctx->temp_reg; alu.dst.chan = 2; alu.dst.write = 1; @@ -1404,13 +1429,12 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) chan = alu.dst.chan; sel = alu.dst.sel; - /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */ + /* tmp.x = amd MUL_LIT(tmp.z, src.w, src.x ) */ memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT); - r600_bc_src(&alu.src[0], &ctx->src[0], 3); - alu.src[1].sel = sel; - alu.src[1].chan = chan; - + alu.src[0].sel = sel; + alu.src[0].chan = chan; + r600_bc_src(&alu.src[1], &ctx->src[0], 3); r600_bc_src(&alu.src[2], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 0; @@ -1421,7 +1445,7 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) if (r) return r; - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { /* dst.z = exp(tmp.x) */ memset(&alu, 0, sizeof(struct r600_bc_alu)); @@ -1506,7 +1530,7 @@ static int tgsi_rsq(struct r600_shader_ctx *ctx) for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { r600_bc_src(&alu.src[i], &ctx->src[i], 0); - alu.src[i].abs = 1; + r600_bc_src_set_abs(&alu.src[i]); } alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; @@ -1898,7 +1922,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) } else if (inst->Instruction.Opcode == TGSI_OPCODE_TXP) { int out_chan; /* Add perspective divide */ - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { out_chan = 2; for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); @@ -1980,7 +2004,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) } /* tmp1.z = RCP_e(|tmp1.z|) */ - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); @@ -2192,7 +2216,7 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) alu.src[0].sel = V_SQ_ALU_SRC_1; alu.src[0].chan = 0; r600_bc_src(&alu.src[1], &ctx->src[0], i); - alu.src[1].neg = 1; + r600_bc_src_toggle_neg(&alu.src[1]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == lasti) { @@ -2373,7 +2397,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) if (r) return r; - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; @@ -2429,7 +2453,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) /* result.z = RoughApprox2ToX(tmp);*/ if ((inst->Dst[0].Register.WriteMask >> 2) & 0x1) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); @@ -2489,14 +2513,15 @@ static int tgsi_log(struct r600_shader_ctx *ctx) int r; int i; - /* result.x = floor(log2(src)); */ + /* result.x = floor(log2(|src|)); */ if (inst->Dst[0].Register.WriteMask & 1) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -2514,6 +2539,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 0; @@ -2538,15 +2564,16 @@ static int tgsi_log(struct r600_shader_ctx *ctx) return r; } - /* result.y = src.x / (2 ^ floor(log2(src.x))); */ + /* result.y = |src.x| / (2 ^ floor(log2(|src.x|))); */ if ((inst->Dst[0].Register.WriteMask >> 1) & 1) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -2564,6 +2591,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 1; @@ -2590,7 +2618,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (r) return r; - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); @@ -2624,7 +2652,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) return r; } - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); @@ -2663,6 +2691,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.src[1].sel = ctx->temp_reg; alu.src[1].chan = 1; @@ -2677,14 +2706,15 @@ static int tgsi_log(struct r600_shader_ctx *ctx) return r; } - /* result.z = log2(src);*/ + /* result.z = log2(|src|);*/ if ((inst->Dst[0].Register.WriteMask >> 2) & 1) { - if (ctx->bc->chiprev == CHIPREV_CAYMAN) { + if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; if (i == 2) @@ -2702,6 +2732,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bc_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; diff --git a/src/gallium/drivers/r600/r600_shader.h b/src/gallium/drivers/r600/r600_shader.h index 76aebf2b1ea..3ba84bd8907 100644 --- a/src/gallium/drivers/r600/r600_shader.h +++ b/src/gallium/drivers/r600/r600_shader.h @@ -43,7 +43,6 @@ struct r600_shader { unsigned nlds; struct r600_shader_io input[32]; struct r600_shader_io output[32]; - enum radeon_family family; boolean uses_kill; boolean fs_write_all; boolean clamp_color; diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c index 7c1976f12e0..f8f7c2031db 100644 --- a/src/gallium/drivers/r600/r600_state.c +++ b/src/gallium/drivers/r600/r600_state.c @@ -44,7 +44,590 @@ #include "r600_resource.h" #include "r600_shader.h" #include "r600_pipe.h" -#include "r600_state_inlines.h" +#include "r600_formats.h" + +static uint32_t r600_translate_blend_function(int blend_func) +{ + switch (blend_func) { + case PIPE_BLEND_ADD: + return V_028804_COMB_DST_PLUS_SRC; + case PIPE_BLEND_SUBTRACT: + return V_028804_COMB_SRC_MINUS_DST; + case PIPE_BLEND_REVERSE_SUBTRACT: + return V_028804_COMB_DST_MINUS_SRC; + case PIPE_BLEND_MIN: + return V_028804_COMB_MIN_DST_SRC; + case PIPE_BLEND_MAX: + return V_028804_COMB_MAX_DST_SRC; + default: + R600_ERR("Unknown blend function %d\n", blend_func); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_blend_factor(int blend_fact) +{ + switch (blend_fact) { + case PIPE_BLENDFACTOR_ONE: + return V_028804_BLEND_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return V_028804_BLEND_SRC_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return V_028804_BLEND_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return V_028804_BLEND_DST_ALPHA; + case PIPE_BLENDFACTOR_DST_COLOR: + return V_028804_BLEND_DST_COLOR; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return V_028804_BLEND_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return V_028804_BLEND_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return V_028804_BLEND_CONST_ALPHA; + case PIPE_BLENDFACTOR_ZERO: + return V_028804_BLEND_ZERO; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return V_028804_BLEND_ONE_MINUS_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return V_028804_BLEND_ONE_MINUS_SRC_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return V_028804_BLEND_ONE_MINUS_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return V_028804_BLEND_ONE_MINUS_DST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return V_028804_BLEND_ONE_MINUS_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return V_028804_BLEND_ONE_MINUS_CONST_ALPHA; + case PIPE_BLENDFACTOR_SRC1_COLOR: + return V_028804_BLEND_SRC1_COLOR; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + return V_028804_BLEND_SRC1_ALPHA; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + return V_028804_BLEND_INV_SRC1_COLOR; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + return V_028804_BLEND_INV_SRC1_ALPHA; + default: + R600_ERR("Bad blend factor %d not supported!\n", blend_fact); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_stencil_op(int s_op) +{ + switch (s_op) { + case PIPE_STENCIL_OP_KEEP: + return V_028800_STENCIL_KEEP; + case PIPE_STENCIL_OP_ZERO: + return V_028800_STENCIL_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return V_028800_STENCIL_REPLACE; + case PIPE_STENCIL_OP_INCR: + return V_028800_STENCIL_INCR; + case PIPE_STENCIL_OP_DECR: + return V_028800_STENCIL_DECR; + case PIPE_STENCIL_OP_INCR_WRAP: + return V_028800_STENCIL_INCR_WRAP; + case PIPE_STENCIL_OP_DECR_WRAP: + return V_028800_STENCIL_DECR_WRAP; + case PIPE_STENCIL_OP_INVERT: + return V_028800_STENCIL_INVERT; + default: + R600_ERR("Unknown stencil op %d", s_op); + assert(0); + break; + } + return 0; +} + +static uint32_t r600_translate_fill(uint32_t func) +{ + switch(func) { + case PIPE_POLYGON_MODE_FILL: + return 2; + case PIPE_POLYGON_MODE_LINE: + return 1; + case PIPE_POLYGON_MODE_POINT: + return 0; + default: + assert(0); + return 0; + } +} + +/* translates straight */ +static uint32_t r600_translate_ds_func(int func) +{ + return func; +} + +static unsigned r600_tex_wrap(unsigned wrap) +{ + switch (wrap) { + default: + case PIPE_TEX_WRAP_REPEAT: + return V_03C000_SQ_TEX_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return V_03C000_SQ_TEX_CLAMP_HALF_BORDER; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return V_03C000_SQ_TEX_CLAMP_LAST_TEXEL; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return V_03C000_SQ_TEX_CLAMP_BORDER; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return V_03C000_SQ_TEX_MIRROR; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + return V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + return V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + return V_03C000_SQ_TEX_MIRROR_ONCE_BORDER; + } +} + +static unsigned r600_tex_filter(unsigned filter) +{ + switch (filter) { + default: + case PIPE_TEX_FILTER_NEAREST: + return V_03C000_SQ_TEX_XY_FILTER_POINT; + case PIPE_TEX_FILTER_LINEAR: + return V_03C000_SQ_TEX_XY_FILTER_BILINEAR; + } +} + +static unsigned r600_tex_mipfilter(unsigned filter) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + return V_03C000_SQ_TEX_Z_FILTER_POINT; + case PIPE_TEX_MIPFILTER_LINEAR: + return V_03C000_SQ_TEX_Z_FILTER_LINEAR; + default: + case PIPE_TEX_MIPFILTER_NONE: + return V_03C000_SQ_TEX_Z_FILTER_NONE; + } +} + +static unsigned r600_tex_compare(unsigned compare) +{ + switch (compare) { + default: + case PIPE_FUNC_NEVER: + return V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER; + case PIPE_FUNC_LESS: + return V_03C000_SQ_TEX_DEPTH_COMPARE_LESS; + case PIPE_FUNC_EQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL; + case PIPE_FUNC_LEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL; + case PIPE_FUNC_GREATER: + return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER; + case PIPE_FUNC_NOTEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL; + case PIPE_FUNC_GEQUAL: + return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL; + case PIPE_FUNC_ALWAYS: + return V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS; + } +} + +static unsigned r600_tex_dim(unsigned dim) +{ + switch (dim) { + default: + case PIPE_TEXTURE_1D: + return V_038000_SQ_TEX_DIM_1D; + case PIPE_TEXTURE_1D_ARRAY: + return V_038000_SQ_TEX_DIM_1D_ARRAY; + case PIPE_TEXTURE_2D: + case PIPE_TEXTURE_RECT: + return V_038000_SQ_TEX_DIM_2D; + case PIPE_TEXTURE_2D_ARRAY: + return V_038000_SQ_TEX_DIM_2D_ARRAY; + case PIPE_TEXTURE_3D: + return V_038000_SQ_TEX_DIM_3D; + case PIPE_TEXTURE_CUBE: + return V_038000_SQ_TEX_DIM_CUBEMAP; + } +} + +static uint32_t r600_translate_dbformat(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + return V_028010_DEPTH_16; + case PIPE_FORMAT_Z24X8_UNORM: + return V_028010_DEPTH_X8_24; + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_028010_DEPTH_8_24; + case PIPE_FORMAT_Z32_FLOAT: + return V_028010_DEPTH_32_FLOAT; + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + return V_028010_DEPTH_X24_8_32_FLOAT; + default: + return ~0U; + } +} + +static uint32_t r600_translate_colorswap(enum pipe_format format) +{ + switch (format) { + /* 8-bit buffers. */ + case PIPE_FORMAT_A8_UNORM: + return V_0280A0_SWAP_ALT_REV; + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_L8_SRGB: + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8_SNORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_L4A4_UNORM: + return V_0280A0_SWAP_ALT; + + /* 16-bit buffers. */ + case PIPE_FORMAT_B5G6R5_UNORM: + return V_0280A0_SWAP_STD_REV; + + case PIPE_FORMAT_B5G5R5A1_UNORM: + case PIPE_FORMAT_B5G5R5X1_UNORM: + return V_0280A0_SWAP_ALT; + + case PIPE_FORMAT_B4G4R4A4_UNORM: + case PIPE_FORMAT_B4G4R4X4_UNORM: + return V_0280A0_SWAP_ALT; + + case PIPE_FORMAT_Z16_UNORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_L8A8_UNORM: + case PIPE_FORMAT_L8A8_SRGB: + return V_0280A0_SWAP_ALT; + case PIPE_FORMAT_R8G8_UNORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_R16_UNORM: + case PIPE_FORMAT_R16_FLOAT: + return V_0280A0_SWAP_STD; + + /* 32-bit buffers. */ + + case PIPE_FORMAT_A8B8G8R8_SRGB: + return V_0280A0_SWAP_STD_REV; + case PIPE_FORMAT_B8G8R8A8_SRGB: + return V_0280A0_SWAP_ALT; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: + return V_0280A0_SWAP_ALT; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + return V_0280A0_SWAP_ALT_REV; + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R8G8B8X8_UNORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_A8B8G8R8_UNORM: + case PIPE_FORMAT_X8B8G8R8_UNORM: + /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */ + return V_0280A0_SWAP_STD_REV; + + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8_USCALED_Z24_UNORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_R10G10B10A2_UNORM: + case PIPE_FORMAT_R10G10B10X2_SNORM: + case PIPE_FORMAT_R10SG10SB10SA2U_NORM: + return V_0280A0_SWAP_STD; + + case PIPE_FORMAT_B10G10R10A2_UNORM: + return V_0280A0_SWAP_ALT; + + case PIPE_FORMAT_R11G11B10_FLOAT: + case PIPE_FORMAT_R16G16_UNORM: + case PIPE_FORMAT_R16G16_FLOAT: + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT: + return V_0280A0_SWAP_STD; + + /* 64-bit buffers. */ + case PIPE_FORMAT_R32G32_FLOAT: + case PIPE_FORMAT_R16G16B16A16_UNORM: + case PIPE_FORMAT_R16G16B16A16_SNORM: + case PIPE_FORMAT_R16G16B16A16_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + + /* 128-bit buffers. */ + case PIPE_FORMAT_R32G32B32A32_FLOAT: + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_UNORM: + return V_0280A0_SWAP_STD; + default: + R600_ERR("unsupported colorswap format %d\n", format); + return ~0U; + } + return ~0U; +} + +static uint32_t r600_translate_colorformat(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_L4A4_UNORM: + return V_0280A0_COLOR_4_4; + + /* 8-bit buffers. */ + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_L8_SRGB: + case PIPE_FORMAT_R8_UNORM: + case PIPE_FORMAT_R8_SNORM: + return V_0280A0_COLOR_8; + + /* 16-bit buffers. */ + case PIPE_FORMAT_B5G6R5_UNORM: + return V_0280A0_COLOR_5_6_5; + + case PIPE_FORMAT_B5G5R5A1_UNORM: + case PIPE_FORMAT_B5G5R5X1_UNORM: + return V_0280A0_COLOR_1_5_5_5; + + case PIPE_FORMAT_B4G4R4A4_UNORM: + case PIPE_FORMAT_B4G4R4X4_UNORM: + return V_0280A0_COLOR_4_4_4_4; + + case PIPE_FORMAT_Z16_UNORM: + return V_0280A0_COLOR_16; + + case PIPE_FORMAT_L8A8_UNORM: + case PIPE_FORMAT_L8A8_SRGB: + case PIPE_FORMAT_R8G8_UNORM: + return V_0280A0_COLOR_8_8; + + case PIPE_FORMAT_R16_UNORM: + return V_0280A0_COLOR_16; + + case PIPE_FORMAT_R16_FLOAT: + return V_0280A0_COLOR_16_FLOAT; + + /* 32-bit buffers. */ + case PIPE_FORMAT_A8B8G8R8_SRGB: + case PIPE_FORMAT_A8B8G8R8_UNORM: + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_SRGB: + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + case PIPE_FORMAT_R8G8B8X8_UNORM: + case PIPE_FORMAT_R8SG8SB8UX8U_NORM: + case PIPE_FORMAT_X8B8G8R8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8_UNORM: + return V_0280A0_COLOR_8_8_8_8; + + case PIPE_FORMAT_R10G10B10A2_UNORM: + case PIPE_FORMAT_R10G10B10X2_SNORM: + case PIPE_FORMAT_B10G10R10A2_UNORM: + case PIPE_FORMAT_R10SG10SB10SA2U_NORM: + return V_0280A0_COLOR_2_10_10_10; + + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + return V_0280A0_COLOR_8_24; + + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8_USCALED_Z24_UNORM: + return V_0280A0_COLOR_24_8; + + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + return V_0280A0_COLOR_X24_8_32_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT: + return V_0280A0_COLOR_32_FLOAT; + + case PIPE_FORMAT_R16G16_FLOAT: + return V_0280A0_COLOR_16_16_FLOAT; + + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16_UNORM: + return V_0280A0_COLOR_16_16; + + case PIPE_FORMAT_R11G11B10_FLOAT: + return V_0280A0_COLOR_10_11_11_FLOAT; + + /* 64-bit buffers. */ + case PIPE_FORMAT_R16G16B16_USCALED: + case PIPE_FORMAT_R16G16B16A16_USCALED: + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_UNORM: + case PIPE_FORMAT_R16G16B16A16_SNORM: + return V_0280A0_COLOR_16_16_16_16; + + case PIPE_FORMAT_R16G16B16_FLOAT: + case PIPE_FORMAT_R16G16B16A16_FLOAT: + return V_0280A0_COLOR_16_16_16_16_FLOAT; + + case PIPE_FORMAT_R32G32_FLOAT: + return V_0280A0_COLOR_32_32_FLOAT; + + case PIPE_FORMAT_R32G32_USCALED: + case PIPE_FORMAT_R32G32_SSCALED: + return V_0280A0_COLOR_32_32; + + /* 96-bit buffers. */ + case PIPE_FORMAT_R32G32B32_FLOAT: + return V_0280A0_COLOR_32_32_32_FLOAT; + + /* 128-bit buffers. */ + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return V_0280A0_COLOR_32_32_32_32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_UNORM: + return V_0280A0_COLOR_32_32_32_32; + + /* YUV buffers. */ + case PIPE_FORMAT_UYVY: + case PIPE_FORMAT_YUYV: + default: + return ~0U; /* Unsupported. */ + } +} + +static uint32_t r600_colorformat_endian_swap(uint32_t colorformat) +{ + if (R600_BIG_ENDIAN) { + switch(colorformat) { + case V_0280A0_COLOR_4_4: + return ENDIAN_NONE; + + /* 8-bit buffers. */ + case V_0280A0_COLOR_8: + return ENDIAN_NONE; + + /* 16-bit buffers. */ + case V_0280A0_COLOR_5_6_5: + case V_0280A0_COLOR_1_5_5_5: + case V_0280A0_COLOR_4_4_4_4: + case V_0280A0_COLOR_16: + case V_0280A0_COLOR_8_8: + return ENDIAN_8IN16; + + /* 32-bit buffers. */ + case V_0280A0_COLOR_8_8_8_8: + case V_0280A0_COLOR_2_10_10_10: + case V_0280A0_COLOR_8_24: + case V_0280A0_COLOR_24_8: + case V_0280A0_COLOR_32_FLOAT: + case V_0280A0_COLOR_16_16_FLOAT: + case V_0280A0_COLOR_16_16: + return ENDIAN_8IN32; + + /* 64-bit buffers. */ + case V_0280A0_COLOR_16_16_16_16: + case V_0280A0_COLOR_16_16_16_16_FLOAT: + return ENDIAN_8IN16; + + case V_0280A0_COLOR_32_32_FLOAT: + case V_0280A0_COLOR_32_32: + case V_0280A0_COLOR_X24_8_32_FLOAT: + return ENDIAN_8IN32; + + /* 128-bit buffers. */ + case V_0280A0_COLOR_32_32_32_FLOAT: + case V_0280A0_COLOR_32_32_32_32_FLOAT: + case V_0280A0_COLOR_32_32_32_32: + return ENDIAN_8IN32; + default: + return ENDIAN_NONE; /* Unsupported. */ + } + } else { + return ENDIAN_NONE; + } +} + +static bool r600_is_sampler_format_supported(struct pipe_screen *screen, enum pipe_format format) +{ + return r600_translate_texformat(screen, format, NULL, NULL, NULL) != ~0U; +} + +static bool r600_is_colorbuffer_format_supported(enum pipe_format format) +{ + return r600_translate_colorformat(format) != ~0U && + r600_translate_colorswap(format) != ~0U; +} + +static bool r600_is_zs_format_supported(enum pipe_format format) +{ + return r600_translate_dbformat(format) != ~0U; +} + +boolean r600_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned sample_count, + unsigned usage) +{ + unsigned retval = 0; + + if (target >= PIPE_MAX_TEXTURE_TYPES) { + R600_ERR("r600: unsupported texture type %d\n", target); + return FALSE; + } + + if (!util_format_is_supported(format, usage)) + return FALSE; + + /* Multisample */ + if (sample_count > 1) + return FALSE; + + if ((usage & PIPE_BIND_SAMPLER_VIEW) && + r600_is_sampler_format_supported(screen, format)) { + retval |= PIPE_BIND_SAMPLER_VIEW; + } + + if ((usage & (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED)) && + r600_is_colorbuffer_format_supported(format)) { + retval |= usage & + (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED); + } + + if ((usage & PIPE_BIND_DEPTH_STENCIL) && + r600_is_zs_format_supported(format)) { + retval |= PIPE_BIND_DEPTH_STENCIL; + } + + if ((usage & PIPE_BIND_VERTEX_BUFFER) && + r600_is_vertex_format_supported(format)) { + retval |= PIPE_BIND_VERTEX_BUFFER; + } + + if (usage & PIPE_BIND_TRANSFER_READ) + retval |= PIPE_BIND_TRANSFER_READ; + if (usage & PIPE_BIND_TRANSFER_WRITE) + retval |= PIPE_BIND_TRANSFER_WRITE; + + return retval == usage; +} void r600_polygon_offset_update(struct r600_pipe_context *rctx) { @@ -63,6 +646,7 @@ void r600_polygon_offset_update(struct r600_pipe_context *rctx) offset_units *= 2.0f; break; case PIPE_FORMAT_Z32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: depth = -23; offset_units *= 1.0f; offset_db_fmt_cntl |= S_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT(1); @@ -831,7 +1415,7 @@ static void r600_cb(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta /* EXPORT_NORM is an optimzation that can be enabled for better * performance in certain cases */ - if (rctx->family < CHIP_RV770) { + if (rctx->chip_class == R600) { /* EXPORT_NORM can be enabled if: * - 11-bit or smaller UNORM/SNORM/SRGB * - BLEND_CLAMP is enabled @@ -992,7 +1576,7 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, 0xFFFFFFFF, NULL); - if (rctx->family >= CHIP_RV770) { + if (rctx->chip_class >= R700) { r600_pipe_state_add_reg(rstate, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, 0xFFFFFFFF, NULL); @@ -1086,16 +1670,13 @@ void r600_init_state_functions(struct r600_pipe_context *rctx) void r600_adjust_gprs(struct r600_pipe_context *rctx) { - enum radeon_family family; struct r600_pipe_state rstate; unsigned num_ps_gprs = rctx->default_ps_gprs; unsigned num_vs_gprs = rctx->default_vs_gprs; unsigned tmp; int diff; - family = r600_get_family(rctx->radeon); - - if (family >= CHIP_CEDAR) + if (rctx->chip_class >= EVERGREEN) return; if (!rctx->ps_shader && !rctx->vs_shader) @@ -1147,7 +1728,7 @@ void r600_init_config(struct r600_pipe_context *rctx) struct r600_pipe_state *rstate = &rctx->config; u32 tmp; - family = r600_get_family(rctx->radeon); + family = rctx->family; ps_prio = 0; vs_prio = 1; gs_prio = 2; @@ -1328,7 +1909,7 @@ void r600_init_config(struct r600_pipe_context *rctx) r600_pipe_state_add_reg(rstate, R_009714_VC_ENHANCE, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x00000000, 0xFFFFFFFF, NULL); - if (family >= CHIP_RV770) { + if (rctx->chip_class >= R700) { r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00004000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_009508_TA_CNTL_AUX, S_009508_DISABLE_CUBE_ANISO(1) | diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c index d9140403e5a..408eaed491b 100644 --- a/src/gallium/drivers/r600/r600_state_common.c +++ b/src/gallium/drivers/r600/r600_state_common.c @@ -109,7 +109,7 @@ void r600_bind_rs_state(struct pipe_context *ctx, void *state) rctx->states[rs->rstate.id] = &rs->rstate; r600_context_pipe_state_set(&rctx->ctx, &rs->rstate); - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_polygon_offset_update(rctx); } else { r600_polygon_offset_update(rctx); @@ -212,7 +212,7 @@ void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count, /* Zero states. */ for (i = 0; i < count; i++) { if (!buffers[i].buffer) { - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_context_pipe_state_set_fs_resource(&rctx->ctx, NULL, i); } else { r600_context_pipe_state_set_fs_resource(&rctx->ctx, NULL, i); @@ -220,7 +220,7 @@ void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count, } } for (; i < rctx->vbuf_mgr->nr_real_vertex_buffers; i++) { - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_context_pipe_state_set_fs_resource(&rctx->ctx, NULL, i); } else { r600_context_pipe_state_set_fs_resource(&rctx->ctx, NULL, i); @@ -367,7 +367,7 @@ static void r600_spi_update(struct r600_pipe_context *rctx) for (i = 0; i < rshader->ninput; i++) { if (rshader->input[i].name == TGSI_SEMANTIC_POSITION || rshader->input[i].name == TGSI_SEMANTIC_FACE) - if (rctx->family >= CHIP_CEDAR) + if (rctx->chip_class >= EVERGREEN) continue; else sid=0; @@ -387,7 +387,7 @@ static void r600_spi_update(struct r600_pipe_context *rctx) tmp |= S_028644_PT_SPRITE_TEX(1); } - if (rctx->family < CHIP_CEDAR) { + if (rctx->chip_class < EVERGREEN) { if (rshader->input[i].centroid) tmp |= S_028644_SEL_CENTROID(1); @@ -434,14 +434,14 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, rstate = &rctx->vs_const_buffer_resource[index]; if (!rstate->id) { - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_init_buffer_resource(rctx, rstate); } else { r600_pipe_init_buffer_resource(rctx, rstate); } } - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); evergreen_context_pipe_state_set_vs_resource(&rctx->ctx, rstate, index); } else { @@ -462,13 +462,13 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, rstate = &rctx->ps_const_buffer_resource[index]; if (!rstate->id) { - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_init_buffer_resource(rctx, rstate); } else { r600_pipe_init_buffer_resource(rctx, rstate); } } - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, rstate, index); } else { @@ -521,14 +521,14 @@ static void r600_vertex_buffer_update(struct r600_pipe_context *rctx) offset += vertex_buffer->buffer_offset + r600_bo_offset(rbuffer->bo); if (!rstate->id) { - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_init_buffer_resource(rctx, rstate); } else { r600_pipe_init_buffer_resource(rctx, rstate); } } - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_pipe_mod_buffer_resource(rstate, rbuffer, offset, vertex_buffer->stride); evergreen_context_pipe_state_set_fs_resource(&rctx->ctx, rstate, i); } else { @@ -600,7 +600,7 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) r600_shader_rebuild(ctx, rctx->vs_shader); if ((rctx->ps_shader->shader.clamp_color != rctx->clamp_fragment_color) || - ((rctx->family >= CHIP_CEDAR) && rctx->ps_shader->shader.fs_write_all && + ((rctx->chip_class >= EVERGREEN) && rctx->ps_shader->shader.fs_write_all && (rctx->ps_shader->shader.nr_cbufs != rctx->nr_cbufs))) r600_shader_rebuild(ctx, rctx->ps_shader); @@ -655,7 +655,7 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) rdraw.indices_bo_offset = draw.index_buffer_offset; } - if (rctx->family >= CHIP_CEDAR) { + if (rctx->chip_class >= EVERGREEN) { evergreen_context_draw(&rctx->ctx, &rdraw); } else { r600_context_draw(&rctx->ctx, &rdraw); diff --git a/src/gallium/drivers/r600/r600_state_inlines.h b/src/gallium/drivers/r600/r600_state_inlines.h deleted file mode 100644 index 8711dbf1720..00000000000 --- a/src/gallium/drivers/r600/r600_state_inlines.h +++ /dev/null @@ -1,614 +0,0 @@ -/* - * Copyright 2010 Red Hat Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ -#ifndef R600_STATE_INLINES_H -#define R600_STATE_INLINES_H - -#include "util/u_format.h" -#include "r600d.h" -#include "r600_formats.h" - -static INLINE uint32_t r600_translate_blend_function(int blend_func) -{ - switch (blend_func) { - case PIPE_BLEND_ADD: - return V_028804_COMB_DST_PLUS_SRC; - case PIPE_BLEND_SUBTRACT: - return V_028804_COMB_SRC_MINUS_DST; - case PIPE_BLEND_REVERSE_SUBTRACT: - return V_028804_COMB_DST_MINUS_SRC; - case PIPE_BLEND_MIN: - return V_028804_COMB_MIN_DST_SRC; - case PIPE_BLEND_MAX: - return V_028804_COMB_MAX_DST_SRC; - default: - R600_ERR("Unknown blend function %d\n", blend_func); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_blend_factor(int blend_fact) -{ - switch (blend_fact) { - case PIPE_BLENDFACTOR_ONE: - return V_028804_BLEND_ONE; - case PIPE_BLENDFACTOR_SRC_COLOR: - return V_028804_BLEND_SRC_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA: - return V_028804_BLEND_SRC_ALPHA; - case PIPE_BLENDFACTOR_DST_ALPHA: - return V_028804_BLEND_DST_ALPHA; - case PIPE_BLENDFACTOR_DST_COLOR: - return V_028804_BLEND_DST_COLOR; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - return V_028804_BLEND_SRC_ALPHA_SATURATE; - case PIPE_BLENDFACTOR_CONST_COLOR: - return V_028804_BLEND_CONST_COLOR; - case PIPE_BLENDFACTOR_CONST_ALPHA: - return V_028804_BLEND_CONST_ALPHA; - case PIPE_BLENDFACTOR_ZERO: - return V_028804_BLEND_ZERO; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - return V_028804_BLEND_ONE_MINUS_SRC_COLOR; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - return V_028804_BLEND_ONE_MINUS_SRC_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - return V_028804_BLEND_ONE_MINUS_DST_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - return V_028804_BLEND_ONE_MINUS_DST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - return V_028804_BLEND_ONE_MINUS_CONST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - return V_028804_BLEND_ONE_MINUS_CONST_ALPHA; - case PIPE_BLENDFACTOR_SRC1_COLOR: - return V_028804_BLEND_SRC1_COLOR; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - return V_028804_BLEND_SRC1_ALPHA; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - return V_028804_BLEND_INV_SRC1_COLOR; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - return V_028804_BLEND_INV_SRC1_ALPHA; - default: - R600_ERR("Bad blend factor %d not supported!\n", blend_fact); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_stencil_op(int s_op) -{ - switch (s_op) { - case PIPE_STENCIL_OP_KEEP: - return V_028800_STENCIL_KEEP; - case PIPE_STENCIL_OP_ZERO: - return V_028800_STENCIL_ZERO; - case PIPE_STENCIL_OP_REPLACE: - return V_028800_STENCIL_REPLACE; - case PIPE_STENCIL_OP_INCR: - return V_028800_STENCIL_INCR; - case PIPE_STENCIL_OP_DECR: - return V_028800_STENCIL_DECR; - case PIPE_STENCIL_OP_INCR_WRAP: - return V_028800_STENCIL_INCR_WRAP; - case PIPE_STENCIL_OP_DECR_WRAP: - return V_028800_STENCIL_DECR_WRAP; - case PIPE_STENCIL_OP_INVERT: - return V_028800_STENCIL_INVERT; - default: - R600_ERR("Unknown stencil op %d", s_op); - assert(0); - break; - } - return 0; -} - -static INLINE uint32_t r600_translate_fill(uint32_t func) -{ - switch(func) { - case PIPE_POLYGON_MODE_FILL: - return 2; - case PIPE_POLYGON_MODE_LINE: - return 1; - case PIPE_POLYGON_MODE_POINT: - return 0; - default: - assert(0); - return 0; - } -} - -/* translates straight */ -static INLINE uint32_t r600_translate_ds_func(int func) -{ - return func; -} - -static inline unsigned r600_tex_wrap(unsigned wrap) -{ - switch (wrap) { - default: - case PIPE_TEX_WRAP_REPEAT: - return V_03C000_SQ_TEX_WRAP; - case PIPE_TEX_WRAP_CLAMP: - return V_03C000_SQ_TEX_CLAMP_HALF_BORDER; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return V_03C000_SQ_TEX_CLAMP_LAST_TEXEL; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return V_03C000_SQ_TEX_CLAMP_BORDER; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - return V_03C000_SQ_TEX_MIRROR; - case PIPE_TEX_WRAP_MIRROR_CLAMP: - return V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - return V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: - return V_03C000_SQ_TEX_MIRROR_ONCE_BORDER; - } -} - -static inline unsigned r600_tex_filter(unsigned filter) -{ - switch (filter) { - default: - case PIPE_TEX_FILTER_NEAREST: - return V_03C000_SQ_TEX_XY_FILTER_POINT; - case PIPE_TEX_FILTER_LINEAR: - return V_03C000_SQ_TEX_XY_FILTER_BILINEAR; - } -} - -static inline unsigned r600_tex_mipfilter(unsigned filter) -{ - switch (filter) { - case PIPE_TEX_MIPFILTER_NEAREST: - return V_03C000_SQ_TEX_Z_FILTER_POINT; - case PIPE_TEX_MIPFILTER_LINEAR: - return V_03C000_SQ_TEX_Z_FILTER_LINEAR; - default: - case PIPE_TEX_MIPFILTER_NONE: - return V_03C000_SQ_TEX_Z_FILTER_NONE; - } -} - -static inline unsigned r600_tex_compare(unsigned compare) -{ - switch (compare) { - default: - case PIPE_FUNC_NEVER: - return V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER; - case PIPE_FUNC_LESS: - return V_03C000_SQ_TEX_DEPTH_COMPARE_LESS; - case PIPE_FUNC_EQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL; - case PIPE_FUNC_LEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL; - case PIPE_FUNC_GREATER: - return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER; - case PIPE_FUNC_NOTEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL; - case PIPE_FUNC_GEQUAL: - return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL; - case PIPE_FUNC_ALWAYS: - return V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS; - } -} - -static inline unsigned r600_tex_swizzle(unsigned swizzle) -{ - switch (swizzle) { - case PIPE_SWIZZLE_RED: - return V_038010_SQ_SEL_X; - case PIPE_SWIZZLE_GREEN: - return V_038010_SQ_SEL_Y; - case PIPE_SWIZZLE_BLUE: - return V_038010_SQ_SEL_Z; - case PIPE_SWIZZLE_ALPHA: - return V_038010_SQ_SEL_W; - case PIPE_SWIZZLE_ZERO: - return V_038010_SQ_SEL_0; - default: - case PIPE_SWIZZLE_ONE: - return V_038010_SQ_SEL_1; - } -} - -static inline unsigned r600_format_type(unsigned format_type) -{ - switch (format_type) { - default: - case UTIL_FORMAT_TYPE_UNSIGNED: - return V_038010_SQ_FORMAT_COMP_UNSIGNED; - case UTIL_FORMAT_TYPE_SIGNED: - return V_038010_SQ_FORMAT_COMP_SIGNED; - case UTIL_FORMAT_TYPE_FIXED: - return V_038010_SQ_FORMAT_COMP_UNSIGNED_BIASED; - } -} - -static inline unsigned r600_tex_dim(unsigned dim) -{ - switch (dim) { - default: - case PIPE_TEXTURE_1D: - return V_038000_SQ_TEX_DIM_1D; - case PIPE_TEXTURE_1D_ARRAY: - return V_038000_SQ_TEX_DIM_1D_ARRAY; - case PIPE_TEXTURE_2D: - case PIPE_TEXTURE_RECT: - return V_038000_SQ_TEX_DIM_2D; - case PIPE_TEXTURE_2D_ARRAY: - return V_038000_SQ_TEX_DIM_2D_ARRAY; - case PIPE_TEXTURE_3D: - return V_038000_SQ_TEX_DIM_3D; - case PIPE_TEXTURE_CUBE: - return V_038000_SQ_TEX_DIM_CUBEMAP; - } -} - -static inline uint32_t r600_translate_dbformat(enum pipe_format format) -{ - switch (format) { - case PIPE_FORMAT_Z16_UNORM: - return V_028010_DEPTH_16; - case PIPE_FORMAT_Z24X8_UNORM: - return V_028010_DEPTH_X8_24; - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_028010_DEPTH_8_24; - default: - return ~0; - } -} - -static inline uint32_t r600_translate_colorswap(enum pipe_format format) -{ - switch (format) { - /* 8-bit buffers. */ - case PIPE_FORMAT_A8_UNORM: - return V_0280A0_SWAP_ALT_REV; - case PIPE_FORMAT_I8_UNORM: - case PIPE_FORMAT_L8_UNORM: - case PIPE_FORMAT_L8_SRGB: - case PIPE_FORMAT_R8_UNORM: - case PIPE_FORMAT_R8_SNORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_L4A4_UNORM: - return V_0280A0_SWAP_ALT; - - /* 16-bit buffers. */ - case PIPE_FORMAT_B5G6R5_UNORM: - return V_0280A0_SWAP_STD_REV; - - case PIPE_FORMAT_B5G5R5A1_UNORM: - case PIPE_FORMAT_B5G5R5X1_UNORM: - return V_0280A0_SWAP_ALT; - - case PIPE_FORMAT_B4G4R4A4_UNORM: - case PIPE_FORMAT_B4G4R4X4_UNORM: - return V_0280A0_SWAP_ALT; - - case PIPE_FORMAT_Z16_UNORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_L8A8_UNORM: - case PIPE_FORMAT_L8A8_SRGB: - return V_0280A0_SWAP_ALT; - case PIPE_FORMAT_R8G8_UNORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_R16_UNORM: - case PIPE_FORMAT_R16_SNORM: - case PIPE_FORMAT_R16_FLOAT: - return V_0280A0_SWAP_STD; - - /* 32-bit buffers. */ - - case PIPE_FORMAT_A8B8G8R8_SRGB: - return V_0280A0_SWAP_STD_REV; - case PIPE_FORMAT_B8G8R8A8_SRGB: - return V_0280A0_SWAP_ALT; - - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_B8G8R8X8_UNORM: - return V_0280A0_SWAP_ALT; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_X8R8G8B8_UNORM: - return V_0280A0_SWAP_ALT_REV; - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_UNORM: - case PIPE_FORMAT_R8G8B8X8_UNORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_A8B8G8R8_UNORM: - case PIPE_FORMAT_X8B8G8R8_UNORM: - /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */ - return V_0280A0_SWAP_STD_REV; - - case PIPE_FORMAT_Z24X8_UNORM: - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_X8Z24_UNORM: - case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_R10G10B10A2_UNORM: - case PIPE_FORMAT_R10G10B10X2_SNORM: - case PIPE_FORMAT_R10SG10SB10SA2U_NORM: - return V_0280A0_SWAP_STD; - - case PIPE_FORMAT_B10G10R10A2_UNORM: - return V_0280A0_SWAP_ALT; - - case PIPE_FORMAT_R11G11B10_FLOAT: - case PIPE_FORMAT_R16G16_UNORM: - case PIPE_FORMAT_R16G16_FLOAT: - case PIPE_FORMAT_R32_FLOAT: - return V_0280A0_SWAP_STD; - - /* 64-bit buffers. */ - case PIPE_FORMAT_R32G32_FLOAT: - case PIPE_FORMAT_R16G16B16A16_UNORM: - case PIPE_FORMAT_R16G16B16A16_SNORM: - case PIPE_FORMAT_R16G16B16A16_FLOAT: - - /* 128-bit buffers. */ - case PIPE_FORMAT_R32G32B32A32_FLOAT: - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_UNORM: - return V_0280A0_SWAP_STD; - default: - R600_ERR("unsupported colorswap format %d\n", format); - return ~0; - } - return ~0; -} - -static INLINE uint32_t r600_translate_colorformat(enum pipe_format format) -{ - switch (format) { - case PIPE_FORMAT_L4A4_UNORM: - return V_0280A0_COLOR_4_4; - - /* 8-bit buffers. */ - case PIPE_FORMAT_A8_UNORM: - case PIPE_FORMAT_I8_UNORM: - case PIPE_FORMAT_L8_UNORM: - case PIPE_FORMAT_L8_SRGB: - case PIPE_FORMAT_R8_UNORM: - case PIPE_FORMAT_R8_SNORM: - return V_0280A0_COLOR_8; - - /* 16-bit buffers. */ - case PIPE_FORMAT_B5G6R5_UNORM: - return V_0280A0_COLOR_5_6_5; - - case PIPE_FORMAT_B5G5R5A1_UNORM: - case PIPE_FORMAT_B5G5R5X1_UNORM: - return V_0280A0_COLOR_1_5_5_5; - - case PIPE_FORMAT_B4G4R4A4_UNORM: - case PIPE_FORMAT_B4G4R4X4_UNORM: - return V_0280A0_COLOR_4_4_4_4; - - case PIPE_FORMAT_Z16_UNORM: - return V_0280A0_COLOR_16; - - case PIPE_FORMAT_L8A8_UNORM: - case PIPE_FORMAT_L8A8_SRGB: - case PIPE_FORMAT_R8G8_UNORM: - return V_0280A0_COLOR_8_8; - - case PIPE_FORMAT_R16_UNORM: - case PIPE_FORMAT_R16_SNORM: - return V_0280A0_COLOR_16; - - case PIPE_FORMAT_R16_FLOAT: - return V_0280A0_COLOR_16_FLOAT; - - /* 32-bit buffers. */ - case PIPE_FORMAT_A8B8G8R8_SRGB: - case PIPE_FORMAT_A8B8G8R8_UNORM: - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_B8G8R8A8_SRGB: - case PIPE_FORMAT_B8G8R8A8_UNORM: - case PIPE_FORMAT_B8G8R8X8_UNORM: - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_UNORM: - case PIPE_FORMAT_R8G8B8X8_UNORM: - case PIPE_FORMAT_R8SG8SB8UX8U_NORM: - case PIPE_FORMAT_X8B8G8R8_UNORM: - case PIPE_FORMAT_X8R8G8B8_UNORM: - case PIPE_FORMAT_R8G8B8_UNORM: - return V_0280A0_COLOR_8_8_8_8; - - case PIPE_FORMAT_R10G10B10A2_UNORM: - case PIPE_FORMAT_R10G10B10X2_SNORM: - case PIPE_FORMAT_B10G10R10A2_UNORM: - case PIPE_FORMAT_R10SG10SB10SA2U_NORM: - return V_0280A0_COLOR_2_10_10_10; - - case PIPE_FORMAT_Z24X8_UNORM: - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return V_0280A0_COLOR_8_24; - - case PIPE_FORMAT_X8Z24_UNORM: - case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - return V_0280A0_COLOR_24_8; - - case PIPE_FORMAT_R32_FLOAT: - return V_0280A0_COLOR_32_FLOAT; - - case PIPE_FORMAT_R16G16_FLOAT: - return V_0280A0_COLOR_16_16_FLOAT; - - case PIPE_FORMAT_R16G16_SSCALED: - case PIPE_FORMAT_R16G16_UNORM: - return V_0280A0_COLOR_16_16; - - case PIPE_FORMAT_R11G11B10_FLOAT: - return V_0280A0_COLOR_10_11_11_FLOAT; - - /* 64-bit buffers. */ - case PIPE_FORMAT_R16G16B16_USCALED: - case PIPE_FORMAT_R16G16B16A16_USCALED: - case PIPE_FORMAT_R16G16B16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_UNORM: - case PIPE_FORMAT_R16G16B16A16_SNORM: - return V_0280A0_COLOR_16_16_16_16; - - case PIPE_FORMAT_R16G16B16_FLOAT: - case PIPE_FORMAT_R16G16B16A16_FLOAT: - return V_0280A0_COLOR_16_16_16_16_FLOAT; - - case PIPE_FORMAT_R32G32_FLOAT: - return V_0280A0_COLOR_32_32_FLOAT; - - case PIPE_FORMAT_R32G32_USCALED: - case PIPE_FORMAT_R32G32_SSCALED: - return V_0280A0_COLOR_32_32; - - /* 96-bit buffers. */ - case PIPE_FORMAT_R32G32B32_FLOAT: - return V_0280A0_COLOR_32_32_32_FLOAT; - - /* 128-bit buffers. */ - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return V_0280A0_COLOR_32_32_32_32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_UNORM: - return V_0280A0_COLOR_32_32_32_32; - - /* YUV buffers. */ - case PIPE_FORMAT_UYVY: - case PIPE_FORMAT_YUYV: - default: - return ~0; /* Unsupported. */ - } -} - -static INLINE uint32_t r600_colorformat_endian_swap(uint32_t colorformat) -{ - if (R600_BIG_ENDIAN) { - switch(colorformat) { - case V_0280A0_COLOR_4_4: - return(ENDIAN_NONE); - - /* 8-bit buffers. */ - case V_0280A0_COLOR_8: - return(ENDIAN_NONE); - - /* 16-bit buffers. */ - case V_0280A0_COLOR_5_6_5: - case V_0280A0_COLOR_1_5_5_5: - case V_0280A0_COLOR_4_4_4_4: - case V_0280A0_COLOR_16: - case V_0280A0_COLOR_8_8: - return(ENDIAN_8IN16); - - /* 32-bit buffers. */ - case V_0280A0_COLOR_8_8_8_8: - case V_0280A0_COLOR_2_10_10_10: - case V_0280A0_COLOR_8_24: - case V_0280A0_COLOR_24_8: - case V_0280A0_COLOR_32_FLOAT: - case V_0280A0_COLOR_16_16_FLOAT: - case V_0280A0_COLOR_16_16: - return(ENDIAN_8IN32); - - /* 64-bit buffers. */ - case V_0280A0_COLOR_16_16_16_16: - case V_0280A0_COLOR_16_16_16_16_FLOAT: - return(ENDIAN_8IN16); - - case V_0280A0_COLOR_32_32_FLOAT: - case V_0280A0_COLOR_32_32: - return(ENDIAN_8IN32); - - /* 128-bit buffers. */ - case V_0280A0_COLOR_32_32_32_FLOAT: - case V_0280A0_COLOR_32_32_32_32_FLOAT: - case V_0280A0_COLOR_32_32_32_32: - return(ENDIAN_8IN32); - default: - return ENDIAN_NONE; /* Unsupported. */ - } - } else { - return ENDIAN_NONE; - } -} - -static INLINE boolean r600_is_sampler_format_supported(struct pipe_screen *screen, enum pipe_format format) -{ - return r600_translate_texformat(screen, format, NULL, NULL, NULL) != ~0; -} - -static INLINE boolean r600_is_colorbuffer_format_supported(enum pipe_format format) -{ - return r600_translate_colorformat(format) != ~0 && - r600_translate_colorswap(format) != ~0; -} - -static INLINE boolean r600_is_zs_format_supported(enum pipe_format format) -{ - return r600_translate_dbformat(format) != ~0; -} - -static INLINE boolean r600_is_vertex_format_supported(enum pipe_format format, - enum radeon_family family) -{ - unsigned i; - const struct util_format_description *desc = util_format_description(format); - if (!desc) - return FALSE; - - /* Find the first non-VOID channel. */ - for (i = 0; i < 4; i++) { - if (desc->channel[i].type != UTIL_FORMAT_TYPE_VOID) { - break; - } - } - if (i == 4) - return FALSE; - - /* No fixed, no double. */ - if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN || - desc->channel[i].type == UTIL_FORMAT_TYPE_FIXED || - (desc->channel[i].size == 64 && - desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT)) - return FALSE; - - /* No scaled/norm formats with 32 bits per channel. */ - if (desc->channel[i].size == 32 && - (desc->channel[i].type == UTIL_FORMAT_TYPE_SIGNED || - desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED)) - return FALSE; - - return TRUE; -} - -#endif diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c index 854761d17cb..63cacbbd50c 100644 --- a/src/gallium/drivers/r600/r600_texture.c +++ b/src/gallium/drivers/r600/r600_texture.c @@ -35,7 +35,6 @@ #include "pipebuffer/pb_buffer.h" #include "r600_pipe.h" #include "r600_resource.h" -#include "r600_state_inlines.h" #include "r600d.h" #include "r600_formats.h" @@ -861,6 +860,12 @@ uint32_t r600_translate_texformat(struct pipe_screen *screen, result = FMT_8; word4 |= S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_INT); goto out_word4; + case PIPE_FORMAT_Z32_FLOAT: + result = FMT_32_FLOAT; + goto out_word4; + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + result = FMT_X24_8_32_FLOAT; + goto out_word4; default: goto out_unknown; } diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 89b2a91fc1f..15f3a8fd813 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -189,7 +189,8 @@ convert_quad_depth( struct depth_data *data, /** - * Compute the depth_data::shader_stencil_refs[] values from the float fragment stencil values. + * Compute the depth_data::shader_stencil_refs[] values from the float + * fragment stencil values. */ static void convert_quad_stencil( struct depth_data *data, @@ -205,10 +206,9 @@ convert_quad_stencil( struct depth_data *data, case PIPE_FORMAT_Z24_UNORM_S8_USCALED: case PIPE_FORMAT_X8Z24_UNORM: case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - { - for (j = 0; j < QUAD_SIZE; j++) { - data->shader_stencil_refs[j] = ((unsigned)(quad->output.stencil[j])); - } + case PIPE_FORMAT_S8_USCALED: + for (j = 0; j < QUAD_SIZE; j++) { + data->shader_stencil_refs[j] = ((unsigned)(quad->output.stencil[j])); } break; default: @@ -216,6 +216,7 @@ convert_quad_stencil( struct depth_data *data, } } + /** * Write data->bzzzz[] values and data->stencilVals into the Z/stencil buffer. */ |