From a30560e6f0fc8e3056f48a140c9c6b582f5e2e77 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= Date: Sun, 4 Oct 2009 16:49:53 +0200 Subject: r300/compiler: Refactor the radeon_pair code to support control flow instructions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Nicolai Hähnle --- src/mesa/drivers/dri/r300/compiler/Makefile | 3 + .../drivers/dri/r300/compiler/r300_fragprog_emit.c | 77 +- src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c | 38 + .../drivers/dri/r300/compiler/r500_fragprog_emit.c | 67 +- .../drivers/dri/r300/compiler/radeon_dataflow.c | 132 ++- .../drivers/dri/r300/compiler/radeon_dataflow.h | 11 + .../drivers/dri/r300/compiler/radeon_opcodes.c | 4 + .../drivers/dri/r300/compiler/radeon_opcodes.h | 5 + .../dri/r300/compiler/radeon_pair_regalloc.c | 349 ++++++++ .../dri/r300/compiler/radeon_pair_schedule.c | 479 +++++++++++ .../dri/r300/compiler/radeon_pair_translate.c | 253 ++++++ .../drivers/dri/r300/compiler/radeon_program.c | 12 +- .../drivers/dri/r300/compiler/radeon_program.h | 3 + .../dri/r300/compiler/radeon_program_pair.c | 889 +-------------------- .../dri/r300/compiler/radeon_program_pair.h | 97 +-- .../dri/r300/compiler/radeon_program_print.c | 114 ++- 16 files changed, 1524 insertions(+), 1009 deletions(-) create mode 100644 src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c create mode 100644 src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c create mode 100644 src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c (limited to 'src/mesa/drivers') diff --git a/src/mesa/drivers/dri/r300/compiler/Makefile b/src/mesa/drivers/dri/r300/compiler/Makefile index 0d8dcb9f87..d83888d90a 100644 --- a/src/mesa/drivers/dri/r300/compiler/Makefile +++ b/src/mesa/drivers/dri/r300/compiler/Makefile @@ -13,6 +13,9 @@ C_SOURCES = \ radeon_opcodes.c \ radeon_program_alu.c \ radeon_program_pair.c \ + radeon_pair_translate.c \ + radeon_pair_schedule.c \ + radeon_pair_regalloc.c \ radeon_dataflow.c \ radeon_dataflow_deadcode.c \ radeon_dataflow_swizzles.c \ diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c index 2dad3dba4a..375838d98e 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c @@ -56,7 +56,6 @@ struct r300_emit_state { }; #define PROG_CODE \ - struct r300_emit_state * emit = (struct r300_emit_state*)data; \ struct r300_fragment_program_compiler *c = emit->compiler; \ struct r300_fragment_program_code *code = &c->code->code.r300 @@ -75,6 +74,18 @@ static void use_temporary(struct r300_fragment_program_code *code, unsigned int code->pixsize = index; } +static unsigned int use_source(struct r300_fragment_program_code* code, struct radeon_pair_instruction_source src) +{ + if (src.File == RC_FILE_CONSTANT) { + return src.Index | (1 << 5); + } else if (src.File == RC_FILE_TEMPORARY) { + use_temporary(code, src.Index); + return src.Index; + } + + return 0; +} + static unsigned int translate_rgb_opcode(struct r300_fragment_program_compiler * c, rc_opcode opcode) { @@ -120,7 +131,7 @@ static unsigned int translate_alpha_opcode(struct r300_fragment_program_compiler /** * Emit one paired ALU instruction. */ -static int emit_alu(void* data, struct radeon_pair_instruction* inst) +static int emit_alu(struct r300_emit_state * emit, struct rc_pair_instruction* inst) { PROG_CODE; @@ -136,14 +147,10 @@ static int emit_alu(void* data, struct radeon_pair_instruction* inst) code->alu.inst[ip].alpha_inst = translate_alpha_opcode(c, inst->Alpha.Opcode); for(j = 0; j < 3; ++j) { - unsigned int src = inst->RGB.Src[j].Index | (inst->RGB.Src[j].Constant << 5); - if (!inst->RGB.Src[j].Constant) - use_temporary(code, inst->RGB.Src[j].Index); + unsigned int src = use_source(code, inst->RGB.Src[j]); code->alu.inst[ip].rgb_addr |= src << (6*j); - src = inst->Alpha.Src[j].Index | (inst->Alpha.Src[j].Constant << 5); - if (!inst->Alpha.Src[j].Constant) - use_temporary(code, inst->Alpha.Src[j].Index); + src = use_source(code, inst->Alpha.Src[j]); code->alu.inst[ip].alpha_addr |= src << (6*j); unsigned int arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle); @@ -203,7 +210,7 @@ static int finish_node(struct r300_emit_state * emit) if (code->alu.length == emit->node_first_alu) { /* Generate a single NOP for this node */ - struct radeon_pair_instruction inst; + struct rc_pair_instruction inst; memset(&inst, 0, sizeof(inst)); if (!emit_alu(emit, &inst)) return 0; @@ -248,7 +255,7 @@ static int finish_node(struct r300_emit_state * emit) * Begin a block of texture instructions. * Create the necessary indirection. */ -static int begin_tex(void* data) +static int begin_tex(struct r300_emit_state * emit) { PROG_CODE; @@ -273,7 +280,7 @@ static int begin_tex(void* data) } -static int emit_tex(void* data, struct radeon_pair_texture_instruction* inst) +static int emit_tex(struct r300_emit_state * emit, struct rc_instruction * inst) { PROG_CODE; @@ -282,31 +289,31 @@ static int emit_tex(void* data, struct radeon_pair_texture_instruction* inst) return 0; } - unsigned int unit = inst->TexSrcUnit; - unsigned int dest = inst->DestIndex; + unsigned int unit = inst->U.I.TexSrcUnit; + unsigned int dest = inst->U.I.DstReg.Index; unsigned int opcode; - switch(inst->Opcode) { - case RADEON_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break; - case RADEON_OPCODE_TEX: opcode = R300_TEX_OP_LD; break; - case RADEON_OPCODE_TXB: opcode = R300_TEX_OP_TXB; break; - case RADEON_OPCODE_TXP: opcode = R300_TEX_OP_TXP; break; + switch(inst->U.I.Opcode) { + case RC_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break; + case RC_OPCODE_TEX: opcode = R300_TEX_OP_LD; break; + case RC_OPCODE_TXB: opcode = R300_TEX_OP_TXB; break; + case RC_OPCODE_TXP: opcode = R300_TEX_OP_TXP; break; default: - error("Unknown texture opcode %i", inst->Opcode); + error("Unknown texture opcode %i", inst->U.I.Opcode); return 0; } - if (inst->Opcode == RADEON_OPCODE_KIL) { + if (inst->U.I.Opcode == RC_OPCODE_KIL) { unit = 0; dest = 0; } else { use_temporary(code, dest); } - use_temporary(code, inst->SrcIndex); + use_temporary(code, inst->U.I.SrcReg[0].Index); code->tex.inst[code->tex.length++] = - (inst->SrcIndex << R300_SRC_ADDR_SHIFT) | + (inst->U.I.SrcReg[0].Index << R300_SRC_ADDR_SHIFT) | (dest << R300_DST_ADDR_SHIFT) | (unit << R300_TEX_ID_SHIFT) | (opcode << R300_TEX_INST_SHIFT); @@ -314,13 +321,6 @@ static int emit_tex(void* data, struct radeon_pair_texture_instruction* inst) } -static const struct radeon_pair_handler pair_handler = { - .EmitPaired = &emit_alu, - .EmitTex = &emit_tex, - .BeginTexBlock = &begin_tex, - .MaxHwTemps = R300_PFS_NUM_TEMP_REGS -}; - /** * Final compilation step: Turn the intermediate radeon_program into * machine-readable instructions. @@ -335,7 +335,24 @@ void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi memset(code, 0, sizeof(struct r300_fragment_program_code)); - radeonPairProgram(compiler, &pair_handler, &emit); + for(struct rc_instruction * inst = compiler->Base.Program.Instructions.Next; + inst != &compiler->Base.Program.Instructions && !compiler->Base.Error; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) { + if (inst->U.I.Opcode == RC_OPCODE_BEGIN_TEX) { + begin_tex(&emit); + continue; + } + + emit_tex(&emit, inst); + } else { + emit_alu(&emit, &inst->U.P); + } + } + + if (code->pixsize >= R300_PFS_NUM_TEMP_REGS) + rc_error(&compiler->Base, "Too many hardware temporaries used.\n"); + if (compiler->Base.Error) return; diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c index 746e4495fe..5581f25352 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c @@ -114,6 +114,8 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) } rc_dataflow_deadcode(&c->Base, &dataflow_outputs_mark_use, c); + if (c->Base.Error) + return; if (c->Base.Debug) { fprintf(stderr, "Fragment Program: After deadcode:\n"); @@ -122,6 +124,8 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) } rc_dataflow_swizzles(&c->Base); + if (c->Base.Error) + return; if (c->Base.Debug) { fprintf(stderr, "Compiler: after dataflow passes:\n"); @@ -129,6 +133,40 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) fflush(stderr); } + rc_pair_translate(c); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after pair translate:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + + rc_pair_schedule(c); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after pair scheduling:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + + if (c->is_r500) + rc_pair_regalloc(c, 128); + else + rc_pair_regalloc(c, R300_PFS_NUM_TEMP_REGS); + + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after pair register allocation:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + if (c->is_r500) { r500BuildFragmentProgramHwCode(c); } else { diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c index 2f0c0d5283..8f618d88ad 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c @@ -37,10 +37,6 @@ * * \author Corbin Simpson * - * \todo Depth write, WPOS/FOGC inputs - * - * \todo FogOption - * */ #include "r500_fragprog.h" @@ -51,7 +47,6 @@ #define PROG_CODE \ - struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \ struct r500_fragment_program_code *code = &c->code->code.r500 #define error(fmt, args...) do { \ @@ -114,7 +109,7 @@ static unsigned int fix_hw_swizzle(unsigned int swz) return swz; } -static unsigned int translate_arg_rgb(struct radeon_pair_instruction *inst, int arg) +static unsigned int translate_arg_rgb(struct rc_pair_instruction *inst, int arg) { unsigned int t = inst->RGB.Arg[arg].Source; int comp; @@ -127,7 +122,7 @@ static unsigned int translate_arg_rgb(struct radeon_pair_instruction *inst, int return t; } -static unsigned int translate_arg_alpha(struct radeon_pair_instruction *inst, int i) +static unsigned int translate_arg_alpha(struct rc_pair_instruction *inst, int i) { unsigned int t = inst->Alpha.Arg[i].Source; t |= fix_hw_swizzle(inst->Alpha.Arg[i].Swizzle) << 2; @@ -144,16 +139,21 @@ static void use_temporary(struct r500_fragment_program_code* code, unsigned int static unsigned int use_source(struct r500_fragment_program_code* code, struct radeon_pair_instruction_source src) { - if (!src.Constant) + if (src.File == RC_FILE_CONSTANT) { + return src.Index | 0x100; + } else if (src.File == RC_FILE_TEMPORARY) { use_temporary(code, src.Index); - return src.Index | src.Constant << 8; + return src.Index; + } + + return 0; } /** * Emit a paired ALU instruction. */ -static int emit_paired(void *data, struct radeon_pair_instruction *inst) +static int emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair_instruction *inst) { PROG_CODE; @@ -221,7 +221,7 @@ static unsigned int translate_strq_swizzle(unsigned int swizzle) /** * Emit a single TEX instruction */ -static int emit_tex(void *data, struct radeon_pair_texture_instruction *inst) +static int emit_tex(struct r300_fragment_program_compiler *c, struct rc_sub_instruction *inst) { PROG_CODE; @@ -233,46 +233,44 @@ static int emit_tex(void *data, struct radeon_pair_texture_instruction *inst) int ip = ++code->inst_end; code->inst[ip].inst0 = R500_INST_TYPE_TEX - | (inst->WriteMask << 11) + | (inst->DstReg.WriteMask << 11) | R500_INST_TEX_SEM_WAIT; code->inst[ip].inst1 = R500_TEX_ID(inst->TexSrcUnit) | R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED; if (inst->TexSrcTarget == RC_TEXTURE_RECT) - code->inst[ip].inst1 |= R500_TEX_UNSCALED; + code->inst[ip].inst1 |= R500_TEX_UNSCALED; switch (inst->Opcode) { - case RADEON_OPCODE_KIL: + case RC_OPCODE_KIL: code->inst[ip].inst1 |= R500_TEX_INST_TEXKILL; break; - case RADEON_OPCODE_TEX: + case RC_OPCODE_TEX: code->inst[ip].inst1 |= R500_TEX_INST_LD; break; - case RADEON_OPCODE_TXB: + case RC_OPCODE_TXB: code->inst[ip].inst1 |= R500_TEX_INST_LODBIAS; break; - case RADEON_OPCODE_TXP: + case RC_OPCODE_TXP: code->inst[ip].inst1 |= R500_TEX_INST_PROJ; break; default: error("emit_tex can't handle opcode %x\n", inst->Opcode); } - code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcIndex) - | (translate_strq_swizzle(inst->SrcSwizzle) << 8) - | R500_TEX_DST_ADDR(inst->DestIndex) + use_temporary(code, inst->SrcReg[0].Index); + if (inst->Opcode != RC_OPCODE_KIL) + use_temporary(code, inst->DstReg.Index); + + code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcReg[0].Index) + | (translate_strq_swizzle(inst->SrcReg[0].Swizzle) << 8) + | R500_TEX_DST_ADDR(inst->DstReg.Index) | R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A; return 1; } -static const struct radeon_pair_handler pair_handler = { - .EmitPaired = emit_paired, - .EmitTex = emit_tex, - .MaxHwTemps = 128 -}; - void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler) { struct r500_fragment_program_code *code = &compiler->code->code.r500; @@ -281,7 +279,22 @@ void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi code->max_temp_idx = 1; code->inst_end = -1; - radeonPairProgram(compiler, &pair_handler, compiler); + for(struct rc_instruction * inst = compiler->Base.Program.Instructions.Next; + inst != &compiler->Base.Program.Instructions && !compiler->Base.Error; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) { + if (inst->U.I.Opcode == RC_OPCODE_BEGIN_TEX) + continue; + + emit_tex(compiler, &inst->U.I); + } else { + emit_paired(compiler, &inst->U.P); + } + } + + if (code->max_temp_idx >= 128) + rc_error(&compiler->Base, "Too many hardware temporaries used"); + if (compiler->Base.Error) return; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c index a171d8ab54..58dcb20d29 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c @@ -27,6 +27,136 @@ #include "radeon_dataflow.h" -#include "radeon_compiler.h" +#include "radeon_program.h" +static void reads_normal(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_sub_instruction * inst = &fullinst->U.I; + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + unsigned int refmask = 0; + + if (inst->SrcReg[src].File == RC_FILE_NONE) + return; + + for(unsigned int chan = 0; chan < 4; ++chan) + refmask |= 1 << GET_SWZ(inst->SrcReg[src].Swizzle, chan); + + refmask &= ~RC_MASK_XYZW; + + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_BIT(refmask, chan)) { + cb(userdata, fullinst, inst->SrcReg[src].File, inst->SrcReg[src].Index, chan); + } + } + + if (refmask && inst->SrcReg[src].RelAddr) + cb(userdata, fullinst, RC_FILE_ADDRESS, 0, RC_MASK_X); + } +} + +static void reads_pair(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + unsigned int refmasks[3] = { 0, 0, 0 }; + + if (inst->RGB.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->RGB.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + for(unsigned int chan = 0; chan < 3; ++chan) { + unsigned int swz = GET_SWZ(inst->RGB.Arg[arg].Swizzle, chan); + if (swz < 4) + refmasks[inst->RGB.Arg[arg].Source] |= 1 << swz; + } + } + } + + if (inst->Alpha.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Alpha.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + if (inst->Alpha.Arg[arg].Swizzle < 4) + refmasks[inst->Alpha.Arg[arg].Source] |= 1 << inst->Alpha.Arg[arg].Swizzle; + } + } + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + for(unsigned int chan = 0; chan < 3; ++chan) { + if (GET_BIT(refmasks[src], chan)) + cb(userdata, fullinst, inst->RGB.Src[src].File, inst->RGB.Src[src].Index, chan); + } + } + + if (inst->Alpha.Src[src].Used) { + if (GET_BIT(refmasks[src], 3)) + cb(userdata, fullinst, inst->Alpha.Src[src].File, inst->Alpha.Src[src].Index, 3); + } + } +} + +/** + * Calls a callback function for all sourced register channels. + * + * This is conservative, i.e. channels may be called multiple times, + * and the writemask of the instruction is not taken into account. + */ +void rc_for_all_reads(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + reads_normal(inst, cb, userdata); + } else { + reads_pair(inst, cb, userdata); + } +} + + + +static void writes_normal(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_sub_instruction * inst = &fullinst->U.I; + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + if (opcode->HasDstReg) { + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_BIT(inst->DstReg.WriteMask, chan)) + cb(userdata, fullinst, inst->DstReg.File, inst->DstReg.Index, chan); + } + } + + if (inst->WriteALUResult) + cb(userdata, fullinst, RC_FILE_SPECIAL, RC_SPECIAL_ALU_RESULT, 0); +} + +static void writes_pair(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + + for(unsigned int chan = 0; chan < 3; ++chan) { + if (GET_BIT(inst->RGB.WriteMask, chan)) + cb(userdata, fullinst, RC_FILE_TEMPORARY, inst->RGB.DestIndex, chan); + } + + if (inst->Alpha.WriteMask) + cb(userdata, fullinst, RC_FILE_TEMPORARY, inst->Alpha.DestIndex, 3); + + if (inst->WriteALUResult) + cb(userdata, fullinst, RC_FILE_SPECIAL, RC_SPECIAL_ALU_RESULT, 0); +} + +/** + * Calls a callback function for all written register channels. + * + * \warning Does not report output registers for paired instructions! + */ +void rc_for_all_writes(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + writes_normal(inst, cb, userdata); + } else { + writes_pair(inst, cb, userdata); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h index 76c323d057..5aa4cb64f3 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h @@ -35,6 +35,17 @@ struct rc_instruction; struct rc_swizzle_caps; +/** + * Help analyze the register accesses of instructions. + */ +/*@{*/ +typedef void (*rc_read_write_fn)(void * userdata, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan); +void rc_for_all_reads(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata); +void rc_for_all_writes(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata); +/*@}*/ + + /** * Compiler passes based on dataflow analysis. */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c index e097a62b55..a5072b5e1e 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c @@ -363,6 +363,10 @@ struct rc_opcode_info rc_opcodes[MAX_RC_OPCODE] = { .Opcode = RC_OPCODE_REPL_ALPHA, .Name = "REPL_ALPHA", .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_BEGIN_TEX, + .Name = "BEGIN_TEX" } }; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h index f8ba5255ca..c9c5b9f80f 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h @@ -183,6 +183,11 @@ typedef enum { * across all other channels */ RC_OPCODE_REPL_ALPHA, + /** special instruction, used in R300-R500 fragment programs + * to indicate the start of a block of texture instructions that + * can run simultaneously. */ + RC_OPCODE_BEGIN_TEX, + MAX_RC_OPCODE } rc_opcode; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c new file mode 100644 index 0000000000..e39ac2f510 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c @@ -0,0 +1,349 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include + +#include "radeon_compiler.h" +#include "radeon_dataflow.h" + + +#define VERBOSE 0 + +#define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0) + + +struct live_intervals { + int Start; + int End; + struct live_intervals * Next; +}; + +struct register_info { + struct live_intervals Live; + + unsigned int Used:1; + unsigned int Allocated:1; + rc_register_file File:3; + unsigned int Index:RC_REGISTER_INDEX_BITS; +}; + +struct hardware_register { + struct live_intervals * Used; +}; + +struct regalloc_state { + struct radeon_compiler * C; + + struct register_info Input[RC_REGISTER_MAX_INDEX]; + struct register_info Temporary[RC_REGISTER_MAX_INDEX]; + + struct hardware_register * HwTemporary; + unsigned int NumHwTemporaries; +}; + +static void print_live_intervals(struct live_intervals * src) +{ + if (!src) { + DBG("(null)"); + return; + } + + while(src) { + DBG("(%i,%i)", src->Start, src->End); + src = src->Next; + } +} + +static void add_live_intervals(struct regalloc_state * s, + struct live_intervals ** dst, struct live_intervals * src) +{ + struct live_intervals ** dst_backup = dst; + + if (VERBOSE) { + DBG("add_live_intervals: "); + print_live_intervals(*dst); + DBG(" to "); + print_live_intervals(src); + DBG("\n"); + } + + while(src) { + if (*dst && (*dst)->End < src->Start) { + dst = &(*dst)->Next; + } else if (!*dst || (*dst)->Start > src->End) { + struct live_intervals * li = memory_pool_malloc(&s->C->Pool, sizeof(*li)); + li->Start = src->Start; + li->End = src->End; + li->Next = *dst; + *dst = li; + src = src->Next; + } else { + if (src->End > (*dst)->End) + (*dst)->End = src->End; + if (src->Start < (*dst)->Start) + (*dst)->Start = src->Start; + src = src->Next; + } + } + + if (VERBOSE) { + DBG(" result: "); + print_live_intervals(*dst_backup); + DBG("\n"); + } +} + +static int overlap_live_intervals(struct live_intervals * dst, struct live_intervals * src) +{ + if (VERBOSE) { + DBG("overlap_live_intervals: "); + print_live_intervals(dst); + DBG(" to "); + print_live_intervals(src); + DBG("\n"); + } + + while(src && dst) { + if (dst->End <= src->Start) { + dst = dst->Next; + } else if (dst->End <= src->End) { + DBG(" overlap\n"); + return 1; + } else if (dst->Start < src->End) { + DBG(" overlap\n"); + return 1; + } else { + src = src->Next; + } + } + + DBG(" no overlap\n"); + + return 0; +} + +static int try_add_live_intervals(struct regalloc_state * s, + struct live_intervals ** dst, struct live_intervals * src) +{ + if (overlap_live_intervals(*dst, src)) + return 0; + + add_live_intervals(s, dst, src); + return 1; +} + +static void scan_callback(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct regalloc_state * s = data; + struct register_info * reg; + + if (file == RC_FILE_TEMPORARY) + reg = &s->Temporary[index]; + else if (file == RC_FILE_INPUT) + reg = &s->Input[index]; + else + return; + + if (!reg->Used) { + reg->Used = 1; + if (file == RC_FILE_INPUT) + reg->Live.Start = -1; + else + reg->Live.Start = inst->IP; + } else { + if (inst->IP > reg->Live.End) + reg->Live.End = inst->IP; + } +} + +static void compute_live_intervals(struct regalloc_state * s) +{ + rc_recompute_ips(s->C); + + for(struct rc_instruction * inst = s->C->Program.Instructions.Next; + inst != &s->C->Program.Instructions; + inst = inst->Next) { + rc_for_all_reads(inst, scan_callback, s); + rc_for_all_writes(inst, scan_callback, s); + } +} + +static void rewrite_register(struct regalloc_state * s, + rc_register_file * file, unsigned int * index) +{ + const struct register_info * reg; + + if (*file == RC_FILE_TEMPORARY) + reg = &s->Temporary[*index]; + else if (*file == RC_FILE_INPUT) + reg = &s->Input[*index]; + else + return; + + if (reg->Allocated) { + *file = reg->File; + *index = reg->Index; + } +} + +static void rewrite_normal_instruction(struct regalloc_state * s, struct rc_sub_instruction * inst) +{ + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + if (opcode->HasDstReg) { + rc_register_file file = inst->DstReg.File; + unsigned int index = inst->DstReg.Index; + + rewrite_register(s, &file, &index); + + inst->DstReg.File = file; + inst->DstReg.Index = index; + } + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + rc_register_file file = inst->SrcReg[src].File; + unsigned int index = inst->SrcReg[src].Index; + + rewrite_register(s, &file, &index); + + inst->SrcReg[src].File = file; + inst->SrcReg[src].Index = index; + } +} + +static void rewrite_pair_instruction(struct regalloc_state * s, struct rc_pair_instruction * inst) +{ + if (inst->RGB.WriteMask) { + rc_register_file file = RC_FILE_TEMPORARY; + unsigned int index = inst->RGB.DestIndex; + + rewrite_register(s, &file, &index); + + inst->RGB.DestIndex = index; + } + + if (inst->Alpha.WriteMask) { + rc_register_file file = RC_FILE_TEMPORARY; + unsigned int index = inst->Alpha.DestIndex; + + rewrite_register(s, &file, &index); + + inst->Alpha.DestIndex = index; + } + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + rc_register_file file = inst->RGB.Src[src].File; + unsigned int index = inst->RGB.Src[src].Index; + + rewrite_register(s, &file, &index); + + inst->RGB.Src[src].File = file; + inst->RGB.Src[src].Index = index; + } + + if (inst->Alpha.Src[src].Used) { + rc_register_file file = inst->Alpha.Src[src].File; + unsigned int index = inst->Alpha.Src[src].Index; + + rewrite_register(s, &file, &index); + + inst->Alpha.Src[src].File = file; + inst->Alpha.Src[src].Index = index; + } + } +} + +static void do_regalloc(struct regalloc_state * s) +{ + /* Simple and stupid greedy register allocation */ + for(unsigned int index = 0; index < RC_REGISTER_MAX_INDEX; ++index) { + struct register_info * reg = &s->Temporary[index]; + + if (!reg->Used) + continue; + + for(unsigned int hwreg = 0; hwreg < s->NumHwTemporaries; ++hwreg) { + if (try_add_live_intervals(s, &s->HwTemporary[hwreg].Used, ®->Live)) { + reg->Allocated = 1; + reg->File = RC_FILE_TEMPORARY; + reg->Index = hwreg; + goto success; + } + } + + rc_error(s->C, "Ran out of hardware temporaries\n"); + return; + + success:; + } + + /* Rewrite all instructions based on the translation table we built */ + for(struct rc_instruction * inst = s->C->Program.Instructions.Next; + inst != &s->C->Program.Instructions; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) + rewrite_normal_instruction(s, &inst->U.I); + else + rewrite_pair_instruction(s, &inst->U.P); + } +} + +static void alloc_input(void * data, unsigned int input, unsigned int hwreg) +{ + struct regalloc_state * s = data; + + if (!s->Input[input].Used) + return; + + add_live_intervals(s, &s->HwTemporary[hwreg].Used, &s->Input[input].Live); + + s->Input[input].Allocated = 1; + s->Input[input].File = RC_FILE_TEMPORARY; + s->Input[input].Index = hwreg; + +} + +void rc_pair_regalloc(struct r300_fragment_program_compiler *c, unsigned maxtemps) +{ + struct regalloc_state s; + + memset(&s, 0, sizeof(s)); + s.C = &c->Base; + s.NumHwTemporaries = maxtemps; + s.HwTemporary = memory_pool_malloc(&s.C->Pool, maxtemps*sizeof(struct hardware_register)); + memset(s.HwTemporary, 0, maxtemps*sizeof(struct hardware_register)); + + compute_live_intervals(&s); + + c->AllocateHwInputs(c, &alloc_input, &s); + + do_regalloc(&s); +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c new file mode 100644 index 0000000000..8a4b5ac8a9 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c @@ -0,0 +1,479 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include + +#include "radeon_compiler.h" +#include "radeon_dataflow.h" + + +#define VERBOSE 0 + +#define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0) + +struct schedule_instruction { + struct rc_instruction * Instruction; + + /** Next instruction in the linked list of ready instructions. */ + struct schedule_instruction *NextReady; + + /** Values that this instruction reads and writes */ + struct reg_value * WriteValues[4]; + struct reg_value * ReadValues[12]; + unsigned int NumWriteValues:3; + unsigned int NumReadValues:4; + + /** + * Number of (read and write) dependencies that must be resolved before + * this instruction can be scheduled. + */ + unsigned int NumDependencies:5; +}; + + +/** + * Used to keep track of which instructions read a value. + */ +struct reg_value_reader { + struct schedule_instruction *Reader; + struct reg_value_reader *Next; +}; + +/** + * Used to keep track which values are stored in each component of a + * RC_FILE_TEMPORARY. + */ +struct reg_value { + struct schedule_instruction * Writer; + + /** + * Unordered linked list of instructions that read from this value. + * When this value becomes available, we increase all readers' + * dependency count. + */ + struct reg_value_reader *Readers; + + /** + * Number of readers of this value. This is decremented each time + * a reader of the value is committed. + * When the reader cound reaches zero, the dependency count + * of the instruction writing \ref Next is decremented. + */ + unsigned int NumReaders; + + struct reg_value *Next; /**< Pointer to the next value to be written to the same register */ +}; + +struct register_state { + struct reg_value * Values[4]; +}; + +struct schedule_state { + struct radeon_compiler * C; + struct schedule_instruction * Current; + + struct register_state Temporary[RC_REGISTER_MAX_INDEX]; + + /** + * Linked lists of instructions that can be scheduled right now, + * based on which ALU/TEX resources they require. + */ + /*@{*/ + struct schedule_instruction *ReadyFullALU; + struct schedule_instruction *ReadyRGB; + struct schedule_instruction *ReadyAlpha; + struct schedule_instruction *ReadyTEX; + /*@}*/ +}; + +static struct reg_value ** get_reg_valuep(struct schedule_state * s, + rc_register_file file, unsigned int index, unsigned int chan) +{ + if (file != RC_FILE_TEMPORARY) + return 0; + + if (index >= RC_REGISTER_MAX_INDEX) { + rc_error(s->C, "%s: index %i out of bounds\n", __FUNCTION__, index); + return 0; + } + + return &s->Temporary[index].Values[chan]; +} + +static struct reg_value * get_reg_value(struct schedule_state * s, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct reg_value ** pv = get_reg_valuep(s, file, index, chan); + if (!pv) + return 0; + return *pv; +} + +static void add_inst_to_list(struct schedule_instruction ** list, struct schedule_instruction * inst) +{ + inst->NextReady = *list; + *list = inst; +} + +static void instruction_ready(struct schedule_state * s, struct schedule_instruction * sinst) +{ + DBG("%i is now ready\n", sinst->Instruction->IP); + + if (sinst->Instruction->Type == RC_INSTRUCTION_NORMAL) + add_inst_to_list(&s->ReadyTEX, sinst); + else if (sinst->Instruction->U.P.Alpha.Opcode == RC_OPCODE_NOP) + add_inst_to_list(&s->ReadyRGB, sinst); + else if (sinst->Instruction->U.P.RGB.Opcode == RC_OPCODE_NOP) + add_inst_to_list(&s->ReadyAlpha, sinst); + else + add_inst_to_list(&s->ReadyFullALU, sinst); +} + +static void decrease_dependencies(struct schedule_state * s, struct schedule_instruction * sinst) +{ + assert(sinst->NumDependencies > 0); + sinst->NumDependencies--; + if (!sinst->NumDependencies) + instruction_ready(s, sinst); +} + +static void commit_instruction(struct schedule_state * s, struct schedule_instruction * sinst) +{ + DBG("%i: commit\n", sinst->Instruction->IP); + + for(unsigned int i = 0; i < sinst->NumReadValues; ++i) { + struct reg_value * v = sinst->ReadValues[i]; + assert(v->NumReaders > 0); + v->NumReaders--; + if (!v->NumReaders) { + if (v->Next) + decrease_dependencies(s, v->Next->Writer); + } + } + + for(unsigned int i = 0; i < sinst->NumWriteValues; ++i) { + struct reg_value * v = sinst->WriteValues[i]; + for(struct reg_value_reader * r = v->Readers; r; r = r->Next) { + decrease_dependencies(s, r->Reader); + } + } +} + +/** + * Emit all ready texture instructions in a single block. + * + * Emit as a single block to (hopefully) sample many textures in parallel, + * and to avoid hardware indirections on R300. + */ +static void emit_all_tex(struct schedule_state * s, struct rc_instruction * before) +{ + struct schedule_instruction *readytex; + + assert(s->ReadyTEX); + + /* Don't let the ready list change under us! */ + readytex = s->ReadyTEX; + s->ReadyTEX = 0; + + /* Node marker for R300 */ + struct rc_instruction * inst_begin = rc_insert_new_instruction(s->C, before->Prev); + inst_begin->U.I.Opcode = RC_OPCODE_BEGIN_TEX; + + /* Link texture instructions back in */ + while(readytex) { + struct schedule_instruction * tex = readytex; + readytex = readytex->NextReady; + + rc_insert_instruction(before->Prev, tex->Instruction); + commit_instruction(s, tex); + } +} + + +static int destructive_merge_instructions( + struct rc_pair_instruction * rgb, + struct rc_pair_instruction * alpha) +{ + assert(rgb->Alpha.Opcode == RC_OPCODE_NOP); + assert(alpha->RGB.Opcode == RC_OPCODE_NOP); + + /* Copy alpha args into rgb */ + const struct rc_opcode_info * opcode = rc_get_opcode_info(alpha->Alpha.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + unsigned int oldsrc = alpha->Alpha.Arg[arg].Source; + rc_register_file file = 0; + unsigned int index = 0; + + if (alpha->Alpha.Arg[arg].Swizzle < 3) { + srcrgb = 1; + file = alpha->RGB.Src[oldsrc].File; + index = alpha->RGB.Src[oldsrc].Index; + } else if (alpha->Alpha.Arg[arg].Swizzle < 4) { + srcalpha = 1; + file = alpha->Alpha.Src[oldsrc].File; + index = alpha->Alpha.Src[oldsrc].Index; + } + + int source = rc_pair_alloc_source(rgb, srcrgb, srcalpha, file, index); + if (source < 0) + return 0; + + rgb->Alpha.Arg[arg].Source = source; + rgb->Alpha.Arg[arg].Swizzle = alpha->Alpha.Arg[arg].Swizzle; + rgb->Alpha.Arg[arg].Abs = alpha->Alpha.Arg[arg].Abs; + rgb->Alpha.Arg[arg].Negate = alpha->Alpha.Arg[arg].Negate; + } + + /* Copy alpha opcode into rgb */ + rgb->Alpha.Opcode = alpha->Alpha.Opcode; + rgb->Alpha.DestIndex = alpha->Alpha.DestIndex; + rgb->Alpha.WriteMask = alpha->Alpha.WriteMask; + rgb->Alpha.OutputWriteMask = alpha->Alpha.OutputWriteMask; + rgb->Alpha.DepthWriteMask = alpha->Alpha.DepthWriteMask; + rgb->Alpha.Saturate = alpha->Alpha.Saturate; + + /* Merge ALU result writing */ + if (alpha->WriteALUResult) { + if (rgb->WriteALUResult) + return 0; + + rgb->WriteALUResult = alpha->WriteALUResult; + rgb->ALUResultCompare = alpha->ALUResultCompare; + } + + return 1; +} + +/** + * Try to merge the given instructions into the rgb instructions. + * + * Return true on success; on failure, return false, and keep + * the instructions untouched. + */ +static int merge_instructions(struct rc_pair_instruction * rgb, struct rc_pair_instruction * alpha) +{ + struct rc_pair_instruction backup; + + memcpy(&backup, rgb, sizeof(struct rc_pair_instruction)); + + if (destructive_merge_instructions(rgb, alpha)) + return 1; + + memcpy(rgb, &backup, sizeof(struct rc_pair_instruction)); + return 0; +} + + +/** + * Find a good ALU instruction or pair of ALU instruction and emit it. + * + * Prefer emitting full ALU instructions, so that when we reach a point + * where no full ALU instruction can be emitted, we have more candidates + * for RGB/Alpha pairing. + */ +static void emit_one_alu(struct schedule_state *s, struct rc_instruction * before) +{ + struct schedule_instruction * sinst; + + if (s->ReadyFullALU || !(s->ReadyRGB && s->ReadyAlpha)) { + if (s->ReadyFullALU) { + sinst = s->ReadyFullALU; + s->ReadyFullALU = s->ReadyFullALU->NextReady; + } else if (s->ReadyRGB) { + sinst = s->ReadyRGB; + s->ReadyRGB = s->ReadyRGB->NextReady; + } else { + sinst = s->ReadyAlpha; + s->ReadyAlpha = s->ReadyAlpha->NextReady; + } + + rc_insert_instruction(before->Prev, sinst->Instruction); + commit_instruction(s, sinst); + } else { + struct schedule_instruction **prgb; + struct schedule_instruction **palpha; + + /* Some pairings might fail because they require too + * many source slots; try all possible pairings if necessary */ + for(prgb = &s->ReadyRGB; *prgb; prgb = &(*prgb)->NextReady) { + for(palpha = &s->ReadyAlpha; *palpha; palpha = &(*palpha)->NextReady) { + struct schedule_instruction * psirgb = *prgb; + struct schedule_instruction * psialpha = *palpha; + + if (!merge_instructions(&psirgb->Instruction->U.P, &psialpha->Instruction->U.P)) + continue; + + *prgb = (*prgb)->NextReady; + *palpha = (*palpha)->NextReady; + rc_insert_instruction(before->Prev, psirgb->Instruction); + commit_instruction(s, psirgb); + commit_instruction(s, psialpha); + goto success; + } + } + + /* No success in pairing; just take the first RGB instruction */ + sinst = s->ReadyRGB; + s->ReadyRGB = s->ReadyRGB->NextReady; + + rc_insert_instruction(before->Prev, sinst->Instruction); + commit_instruction(s, sinst); + success: ; + } +} + +static void scan_read(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct schedule_state * s = data; + struct reg_value * v = get_reg_value(s, file, index, chan); + + if (!v) + return; + + DBG("%i: read %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); + + struct reg_value_reader * reader = memory_pool_malloc(&s->C->Pool, sizeof(*reader)); + reader->Reader = s->Current; + reader->Next = v->Readers; + v->Readers = reader; + v->NumReaders++; + + s->Current->NumDependencies++; + + if (s->Current->NumReadValues >= 12) { + rc_error(s->C, "%s: NumReadValues overflow\n", __FUNCTION__); + } else { + s->Current->ReadValues[s->Current->NumReadValues++] = v; + } +} + +static void scan_write(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct schedule_state * s = data; + struct reg_value ** pv = get_reg_valuep(s, file, index, chan); + + if (!pv) + return; + + DBG("%i: write %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); + + struct reg_value * newv = memory_pool_malloc(&s->C->Pool, sizeof(*newv)); + memset(newv, 0, sizeof(*newv)); + + newv->Writer = s->Current; + + if (*pv) { + (*pv)->Next = newv; + s->Current->NumDependencies++; + } + + *pv = newv; + + if (s->Current->NumWriteValues >= 4) { + rc_error(s->C, "%s: NumWriteValues overflow\n", __FUNCTION__); + } else { + s->Current->WriteValues[s->Current->NumWriteValues++] = newv; + } +} + +static void schedule_block(struct r300_fragment_program_compiler * c, + struct rc_instruction * begin, struct rc_instruction * end) +{ + struct schedule_state s; + + memset(&s, 0, sizeof(s)); + s.C = &c->Base; + + /* Scan instructions for data dependencies */ + unsigned int ip = 0; + for(struct rc_instruction * inst = begin; inst != end; inst = inst->Next) { + s.Current = memory_pool_malloc(&c->Base.Pool, sizeof(*s.Current)); + memset(s.Current, 0, sizeof(struct schedule_instruction)); + + s.Current->Instruction = inst; + inst->IP = ip++; + + DBG("%i: Scanning\n", inst->IP); + + rc_for_all_reads(inst, &scan_read, &s); + rc_for_all_writes(inst, &scan_write, &s); + + DBG("%i: Has %i dependencies\n", inst->IP, s.Current->NumDependencies); + + if (!s.Current->NumDependencies) + instruction_ready(&s, s.Current); + } + + /* Temporarily unlink all instructions */ + begin->Prev->Next = end; + end->Prev = begin->Prev; + + /* Schedule instructions back */ + while(!s.C->Error && + (s.ReadyTEX || s.ReadyRGB || s.ReadyAlpha || s.ReadyFullALU)) { + if (s.ReadyTEX) + emit_all_tex(&s, end); + + while(!s.C->Error && (s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha)) + emit_one_alu(&s, end); + } +} + +static int is_controlflow(struct rc_instruction * inst) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + return opcode->IsControlFlow; + } + return 0; +} + +void rc_pair_schedule(struct r300_fragment_program_compiler *c) +{ + struct rc_instruction * inst = c->Base.Program.Instructions.Next; + while(inst != &c->Base.Program.Instructions) { + if (is_controlflow(inst)) { + inst = inst->Next; + continue; + } + + struct rc_instruction * first = inst; + + while(inst != &c->Base.Program.Instructions && !is_controlflow(inst)) + inst = inst->Next; + + DBG("Schedule one block\n"); + schedule_block(c, first, inst); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c new file mode 100644 index 0000000000..c31891a62f --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c @@ -0,0 +1,253 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include "radeon_compiler.h" + + +/** + * Finally rewrite ADD, MOV, MUL as the appropriate native instruction + * and reverse the order of arguments for CMP. + */ +static void final_rewrite(struct rc_sub_instruction *inst) +{ + struct rc_src_register tmp; + + switch(inst->Opcode) { + case RC_OPCODE_ADD: + inst->SrcReg[2] = inst->SrcReg[1]; + inst->SrcReg[1].File = RC_FILE_NONE; + inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; + inst->SrcReg[1].Negate = RC_MASK_NONE; + inst->Opcode = RC_OPCODE_MAD; + break; + case RC_OPCODE_CMP: + tmp = inst->SrcReg[2]; + inst->SrcReg[2] = inst->SrcReg[0]; + inst->SrcReg[0] = tmp; + break; + case RC_OPCODE_MOV: + /* AMD say we should use CMP. + * However, when we transform + * KIL -r0; + * into + * CMP tmp, -r0, -r0, 0; + * KIL tmp; + * we get incorrect behaviour on R500 when r0 == 0.0. + * It appears that the R500 KIL hardware treats -0.0 as less + * than zero. + */ + inst->SrcReg[1].File = RC_FILE_NONE; + inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; + inst->SrcReg[2].File = RC_FILE_NONE; + inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; + inst->Opcode = RC_OPCODE_MAD; + break; + case RC_OPCODE_MUL: + inst->SrcReg[2].File = RC_FILE_NONE; + inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; + inst->Opcode = RC_OPCODE_MAD; + break; + default: + /* nothing to do */ + break; + } +} + + +/** + * Classify an instruction according to which ALUs etc. it needs + */ +static void classify_instruction(struct rc_sub_instruction * inst, + int * needrgb, int * needalpha, int * istranscendent) +{ + *needrgb = (inst->DstReg.WriteMask & RC_MASK_XYZ) ? 1 : 0; + *needalpha = (inst->DstReg.WriteMask & RC_MASK_W) ? 1 : 0; + *istranscendent = 0; + + if (inst->WriteALUResult == RC_ALURESULT_X) + *needrgb = 1; + else if (inst->WriteALUResult == RC_ALURESULT_W) + *needalpha = 1; + + switch(inst->Opcode) { + case RC_OPCODE_ADD: + case RC_OPCODE_CMP: + case RC_OPCODE_DDX: + case RC_OPCODE_DDY: + case RC_OPCODE_FRC: + case RC_OPCODE_MAD: + case RC_OPCODE_MAX: + case RC_OPCODE_MIN: + case RC_OPCODE_MOV: + case RC_OPCODE_MUL: + break; + case RC_OPCODE_COS: + case RC_OPCODE_EX2: + case RC_OPCODE_LG2: + case RC_OPCODE_RCP: + case RC_OPCODE_RSQ: + case RC_OPCODE_SIN: + *istranscendent = 1; + *needalpha = 1; + break; + case RC_OPCODE_DP4: + *needalpha = 1; + /* fall through */ + case RC_OPCODE_DP3: + *needrgb = 1; + break; + default: + break; + } +} + + +/** + * Fill the given ALU instruction's opcodes and source operands into the given pair, + * if possible. + */ +static void set_pair_instruction(struct r300_fragment_program_compiler *c, + struct rc_pair_instruction * pair, + struct rc_sub_instruction * inst) +{ + memset(pair, 0, sizeof(struct rc_pair_instruction)); + + int needrgb, needalpha, istranscendent; + classify_instruction(inst, &needrgb, &needalpha, &istranscendent); + + if (needrgb) { + if (istranscendent) + pair->RGB.Opcode = RC_OPCODE_REPL_ALPHA; + else + pair->RGB.Opcode = inst->Opcode; + if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) + pair->RGB.Saturate = 1; + } + if (needalpha) { + pair->Alpha.Opcode = inst->Opcode; + if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) + pair->Alpha.Saturate = 1; + } + + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + int nargs = opcode->NumSrcRegs; + int i; + + /* Special case for DDX/DDY (MDH/MDV). */ + if (inst->Opcode == RC_OPCODE_DDX || inst->Opcode == RC_OPCODE_DDY) { + nargs++; + } + + for(i = 0; i < opcode->NumSrcRegs; ++i) { + int source; + if (needrgb && !istranscendent) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + int j; + for(j = 0; j < 3; ++j) { + unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); + if (swz < 3) + srcrgb = 1; + else if (swz < 4) + srcalpha = 1; + } + source = rc_pair_alloc_source(pair, srcrgb, srcalpha, + inst->SrcReg[i].File, inst->SrcReg[i].Index); + pair->RGB.Arg[i].Source = source; + pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff; + pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs; + pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (RC_MASK_X | RC_MASK_Y | RC_MASK_Z)); + } + if (needalpha) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, istranscendent ? 0 : 3); + if (swz < 3) + srcrgb = 1; + else if (swz < 4) + srcalpha = 1; + source = rc_pair_alloc_source(pair, srcrgb, srcalpha, + inst->SrcReg[i].File, inst->SrcReg[i].Index); + pair->Alpha.Arg[i].Source = source; + pair->Alpha.Arg[i].Swizzle = swz; + pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs; + pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & RC_MASK_W); + } + } + + /* Destination handling */ + if (inst->DstReg.File == RC_FILE_OUTPUT) { + if (inst->DstReg.Index == c->OutputColor) { + pair->RGB.OutputWriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; + pair->Alpha.OutputWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } else if (inst->DstReg.Index == c->OutputDepth) { + pair->Alpha.DepthWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } + } else { + if (needrgb) { + pair->RGB.DestIndex = inst->DstReg.Index; + pair->RGB.WriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; + } + if (needalpha) { + pair->Alpha.DestIndex = inst->DstReg.Index; + pair->Alpha.WriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } + } + + if (inst->WriteALUResult) { + pair->WriteALUResult = inst->WriteALUResult; + pair->ALUResultCompare = inst->ALUResultCompare; + } +} + + +/** + * Translate all ALU instructions into corresponding pair instructions, + * performing no other changes. + */ +void rc_pair_translate(struct r300_fragment_program_compiler *c) +{ + for(struct rc_instruction * inst = c->Base.Program.Instructions.Next; + inst != &c->Base.Program.Instructions; + inst = inst->Next) { + if (inst->Type != RC_INSTRUCTION_NORMAL) + continue; + + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + + if (opcode->HasTexture || opcode->IsControlFlow || opcode->Opcode == RC_OPCODE_KIL) + continue; + + struct rc_sub_instruction copy = inst->U.I; + + final_rewrite(©); + inst->Type = RC_INSTRUCTION_PAIR; + set_pair_instruction(c, &inst->U.P, ©); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.c b/src/mesa/drivers/dri/r300/compiler/radeon_program.c index 68a093b8c0..0dbc5380bb 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.c @@ -138,16 +138,20 @@ struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c) return inst; } - -struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after) +void rc_insert_instruction(struct rc_instruction * after, struct rc_instruction * inst) { - struct rc_instruction * inst = rc_alloc_instruction(c); - inst->Prev = after; inst->Next = after->Next; inst->Prev->Next = inst; inst->Next->Prev = inst; +} + +struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after) +{ + struct rc_instruction * inst = rc_alloc_instruction(c); + + rc_insert_instruction(after, inst); return inst; } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.h b/src/mesa/drivers/dri/r300/compiler/radeon_program.h index 067cb545fd..33db3ea0ff 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.h @@ -34,6 +34,7 @@ #include "radeon_opcodes.h" #include "radeon_code.h" #include "radeon_program_constants.h" +#include "radeon_program_pair.h" struct radeon_compiler; @@ -121,6 +122,7 @@ struct rc_instruction { rc_instruction_type Type; union { struct rc_sub_instruction I; + struct rc_pair_instruction P; } U; /** @@ -221,6 +223,7 @@ unsigned int rc_find_free_temporary(struct radeon_compiler * c); struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c); struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after); +void rc_insert_instruction(struct rc_instruction * after, struct rc_instruction * inst); void rc_remove_instruction(struct rc_instruction * inst); unsigned int rc_recompute_ips(struct radeon_compiler * c); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c index 0d8d8e0b3b..ee839596aa 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2008 Nicolai Haehnle. + * Copyright (C) 2008-2009 Nicolai Haehnle. * * All Rights Reserved. * @@ -25,581 +25,29 @@ * */ -/** - * @file - * - * Perform temporary register allocation and attempt to pair off instructions - * in RGB and Alpha pairs. Also attempts to optimize the TEX instruction - * vs. ALU instruction scheduling. - */ - #include "radeon_program_pair.h" -#include - -#include "memory_pool.h" -#include "radeon_compiler.h" - -#define error(fmt, args...) do { \ - rc_error(&s->Compiler->Base, "%s::%s(): " fmt "\n", \ - __FILE__, __FUNCTION__, ##args); \ -} while(0) - -struct pair_state_instruction { - struct rc_sub_instruction Instruction; - unsigned int IP; /**< Position of this instruction in original program */ - - unsigned int IsTex:1; /**< Is a texture instruction */ - unsigned int NeedRGB:1; /**< Needs the RGB ALU */ - unsigned int NeedAlpha:1; /**< Needs the Alpha ALU */ - unsigned int IsTranscendent:1; /**< Is a special transcendent instruction */ - - /** - * Number of (read and write) dependencies that must be resolved before - * this instruction can be scheduled. - */ - unsigned int NumDependencies:5; - - /** - * Next instruction in the linked list of ready instructions. - */ - struct pair_state_instruction *NextReady; - - /** - * Values that this instruction writes - */ - struct reg_value *Values[4]; -}; - - -/** - * Used to keep track of which instructions read a value. - */ -struct reg_value_reader { - struct pair_state_instruction *Reader; - struct reg_value_reader *Next; -}; - -/** - * Used to keep track which values are stored in each component of a - * RC_FILE_TEMPORARY. - */ -struct reg_value { - struct pair_state_instruction *Writer; - struct reg_value *Next; /**< Pointer to the next value to be written to the same RC_FILE_TEMPORARY component */ - - /** - * Unordered linked list of instructions that read from this value. - */ - struct reg_value_reader *Readers; - - /** - * Number of readers of this value. This is calculated during @ref scan_instructions - * and continually decremented during code emission. - * When this count reaches zero, the instruction that writes the @ref Next value - * can be scheduled. - */ - unsigned int NumReaders; -}; - -/** - * Used to translate a RC_FILE_INPUT or RC_FILE_TEMPORARY Mesa register - * to the proper hardware temporary. - */ -struct pair_register_translation { - unsigned int Allocated:1; - unsigned int HwIndex:8; - unsigned int RefCount:23; /**< # of times this occurs in an unscheduled instruction SrcReg or DstReg */ - - /** - * Notes the value that is currently contained in each component - * (only used for RC_FILE_TEMPORARY registers). - */ - struct reg_value *Value[4]; -}; - -struct pair_state { - struct r300_fragment_program_compiler * Compiler; - const struct radeon_pair_handler *Handler; - unsigned int Verbose; - void *UserData; - - /** - * Translate Mesa registers to hardware registers - */ - struct pair_register_translation Inputs[RC_REGISTER_MAX_INDEX]; - struct pair_register_translation Temps[RC_REGISTER_MAX_INDEX]; - - struct { - unsigned int RefCount; /**< # of times this occurs in an unscheduled SrcReg or DstReg */ - } HwTemps[128]; - - /** - * Linked list of instructions that can be scheduled right now, - * based on which ALU/TEX resources they require. - */ - struct pair_state_instruction *ReadyFullALU; - struct pair_state_instruction *ReadyRGB; - struct pair_state_instruction *ReadyAlpha; - struct pair_state_instruction *ReadyTEX; -}; - - -static struct pair_register_translation *get_register(struct pair_state *s, rc_register_file file, unsigned int index) -{ - switch(file) { - case RC_FILE_TEMPORARY: return &s->Temps[index]; - case RC_FILE_INPUT: return &s->Inputs[index]; - default: return 0; - } -} - -static void alloc_hw_reg(struct pair_state *s, rc_register_file file, unsigned int index, unsigned int hwindex) -{ - struct pair_register_translation *t = get_register(s, file, index); - assert(!s->HwTemps[hwindex].RefCount); - assert(!t->Allocated); - s->HwTemps[hwindex].RefCount = t->RefCount; - t->Allocated = 1; - t->HwIndex = hwindex; -} - -static unsigned int get_hw_reg(struct pair_state *s, rc_register_file file, unsigned int index) -{ - unsigned int hwindex; - - struct pair_register_translation *t = get_register(s, file, index); - if (!t) { - error("get_hw_reg: %i[%i]\n", file, index); - return 0; - } - - if (t->Allocated) - return t->HwIndex; - - for(hwindex = 0; hwindex < s->Handler->MaxHwTemps; ++hwindex) - if (!s->HwTemps[hwindex].RefCount) - break; - - if (hwindex >= s->Handler->MaxHwTemps) { - error("Ran out of hardware temporaries"); - return 0; - } - - alloc_hw_reg(s, file, index, hwindex); - return hwindex; -} - - -static void deref_hw_reg(struct pair_state *s, unsigned int hwindex) -{ - if (!s->HwTemps[hwindex].RefCount) { - error("Hwindex %i refcount error", hwindex); - return; - } - - s->HwTemps[hwindex].RefCount--; -} - -static void add_pairinst_to_list(struct pair_state_instruction **list, struct pair_state_instruction *pairinst) -{ - pairinst->NextReady = *list; - *list = pairinst; -} - -/** - * The given instruction has become ready. Link it into the ready - * instructions. - */ -static void instruction_ready(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - if (s->Verbose) - fprintf(stderr, "instruction_ready(%i)\n", pairinst->IP); - - if (pairinst->IsTex) - add_pairinst_to_list(&s->ReadyTEX, pairinst); - else if (!pairinst->NeedAlpha) - add_pairinst_to_list(&s->ReadyRGB, pairinst); - else if (!pairinst->NeedRGB) - add_pairinst_to_list(&s->ReadyAlpha, pairinst); - else - add_pairinst_to_list(&s->ReadyFullALU, pairinst); -} - - -/** - * Finally rewrite ADD, MOV, MUL as the appropriate native instruction - * and reverse the order of arguments for CMP. - */ -static void final_rewrite(struct pair_state *s, struct rc_sub_instruction *inst) -{ - struct rc_src_register tmp; - - switch(inst->Opcode) { - case RC_OPCODE_ADD: - inst->SrcReg[2] = inst->SrcReg[1]; - inst->SrcReg[1].File = RC_FILE_NONE; - inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; - inst->SrcReg[1].Negate = RC_MASK_NONE; - inst->Opcode = RC_OPCODE_MAD; - break; - case RC_OPCODE_CMP: - tmp = inst->SrcReg[2]; - inst->SrcReg[2] = inst->SrcReg[0]; - inst->SrcReg[0] = tmp; - break; - case RC_OPCODE_MOV: - /* AMD say we should use CMP. - * However, when we transform - * KIL -r0; - * into - * CMP tmp, -r0, -r0, 0; - * KIL tmp; - * we get incorrect behaviour on R500 when r0 == 0.0. - * It appears that the R500 KIL hardware treats -0.0 as less - * than zero. - */ - inst->SrcReg[1].File = RC_FILE_NONE; - inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; - inst->SrcReg[2].File = RC_FILE_NONE; - inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; - inst->Opcode = RC_OPCODE_MAD; - break; - case RC_OPCODE_MUL: - inst->SrcReg[2].File = RC_FILE_NONE; - inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; - inst->Opcode = RC_OPCODE_MAD; - break; - default: - /* nothing to do */ - break; - } -} - - -/** - * Classify an instruction according to which ALUs etc. it needs - */ -static void classify_instruction(struct pair_state *s, - struct pair_state_instruction *psi) -{ - psi->NeedRGB = (psi->Instruction.DstReg.WriteMask & RC_MASK_XYZ) ? 1 : 0; - psi->NeedAlpha = (psi->Instruction.DstReg.WriteMask & RC_MASK_W) ? 1 : 0; - - switch(psi->Instruction.Opcode) { - case RC_OPCODE_ADD: - case RC_OPCODE_CMP: - case RC_OPCODE_DDX: - case RC_OPCODE_DDY: - case RC_OPCODE_FRC: - case RC_OPCODE_MAD: - case RC_OPCODE_MAX: - case RC_OPCODE_MIN: - case RC_OPCODE_MOV: - case RC_OPCODE_MUL: - break; - case RC_OPCODE_COS: - case RC_OPCODE_EX2: - case RC_OPCODE_LG2: - case RC_OPCODE_RCP: - case RC_OPCODE_RSQ: - case RC_OPCODE_SIN: - psi->IsTranscendent = 1; - psi->NeedAlpha = 1; - break; - case RC_OPCODE_DP4: - psi->NeedAlpha = 1; - /* fall through */ - case RC_OPCODE_DP3: - psi->NeedRGB = 1; - break; - case RC_OPCODE_KIL: - case RC_OPCODE_TEX: - case RC_OPCODE_TXB: - case RC_OPCODE_TXP: - psi->IsTex = 1; - break; - default: - error("Unknown opcode %d\n", psi->Instruction.Opcode); - break; - } -} - - -/** - * Count which (input, temporary) register is read and written how often, - * and scan the instruction stream to find dependencies. - */ -static void scan_instructions(struct pair_state *s) -{ - struct rc_instruction *source; - unsigned int ip; - - for(source = s->Compiler->Base.Program.Instructions.Next, ip = 0; - source != &s->Compiler->Base.Program.Instructions; - source = source->Next, ++ip) { - struct pair_state_instruction *pairinst = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*pairinst)); - memset(pairinst, 0, sizeof(struct pair_state_instruction)); - - pairinst->Instruction = source->U.I; - pairinst->IP = ip; - final_rewrite(s, &pairinst->Instruction); - classify_instruction(s, pairinst); - - const struct rc_opcode_info * opcode = rc_get_opcode_info(pairinst->Instruction.Opcode); - int j; - for(j = 0; j < opcode->NumSrcRegs; j++) { - struct pair_register_translation *t = - get_register(s, pairinst->Instruction.SrcReg[j].File, pairinst->Instruction.SrcReg[j].Index); - if (!t) - continue; - - t->RefCount++; - - if (pairinst->Instruction.SrcReg[j].File == RC_FILE_TEMPORARY) { - int i; - for(i = 0; i < 4; ++i) { - unsigned int swz = GET_SWZ(pairinst->Instruction.SrcReg[j].Swizzle, i); - if (swz >= 4) - continue; /* constant or NIL swizzle */ - if (!t->Value[swz]) - continue; /* this is an undefined read */ - - /* Do not add a dependency if this instruction - * also rewrites the value. The code below adds - * a dependency for the DstReg, which is a superset - * of the SrcReg dependency. */ - if (pairinst->Instruction.DstReg.File == RC_FILE_TEMPORARY && - pairinst->Instruction.DstReg.Index == pairinst->Instruction.SrcReg[j].Index && - GET_BIT(pairinst->Instruction.DstReg.WriteMask, swz)) - continue; - - struct reg_value_reader* r = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*r)); - pairinst->NumDependencies++; - t->Value[swz]->NumReaders++; - r->Reader = pairinst; - r->Next = t->Value[swz]->Readers; - t->Value[swz]->Readers = r; - } - } - } - - if (opcode->HasDstReg) { - struct pair_register_translation *t = - get_register(s, pairinst->Instruction.DstReg.File, pairinst->Instruction.DstReg.Index); - if (t) { - t->RefCount++; - - if (pairinst->Instruction.DstReg.File == RC_FILE_TEMPORARY) { - int j; - for(j = 0; j < 4; ++j) { - if (!GET_BIT(pairinst->Instruction.DstReg.WriteMask, j)) - continue; - - struct reg_value* v = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*v)); - memset(v, 0, sizeof(struct reg_value)); - v->Writer = pairinst; - if (t->Value[j]) { - pairinst->NumDependencies++; - t->Value[j]->Next = v; - } - t->Value[j] = v; - pairinst->Values[j] = v; - } - } - } - } - - if (s->Verbose) - fprintf(stderr, "scan(%i): NumDeps = %i\n", ip, pairinst->NumDependencies); - - if (!pairinst->NumDependencies) - instruction_ready(s, pairinst); - } - - /* Clear the RC_FILE_TEMPORARY state */ - int i, j; - for(i = 0; i < RC_REGISTER_MAX_INDEX; ++i) { - for(j = 0; j < 4; ++j) - s->Temps[i].Value[j] = 0; - } -} - - -static void decrement_dependencies(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - assert(pairinst->NumDependencies > 0); - if (!--pairinst->NumDependencies) - instruction_ready(s, pairinst); -} /** - * Update the dependency tracking state based on what the instruction - * at the given IP does. + * Return the source slot where we installed the given register access, + * or -1 if no slot was free anymore. */ -static void commit_instruction(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - struct rc_sub_instruction *inst = &pairinst->Instruction; - - if (s->Verbose) - fprintf(stderr, "commit_instruction(%i)\n", pairinst->IP); - - if (inst->DstReg.File == RC_FILE_TEMPORARY) { - struct pair_register_translation *t = &s->Temps[inst->DstReg.Index]; - deref_hw_reg(s, t->HwIndex); - - int i; - for(i = 0; i < 4; ++i) { - if (!GET_BIT(inst->DstReg.WriteMask, i)) - continue; - - t->Value[i] = pairinst->Values[i]; - if (t->Value[i]->NumReaders) { - struct reg_value_reader *r; - for(r = pairinst->Values[i]->Readers; r; r = r->Next) - decrement_dependencies(s, r->Reader); - } else if (t->Value[i]->Next) { - /* This happens when the only reader writes - * the register at the same time */ - decrement_dependencies(s, t->Value[i]->Next->Writer); - } - } - } - - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); - int i; - for(i = 0; i < opcode->NumSrcRegs; i++) { - struct pair_register_translation *t = get_register(s, inst->SrcReg[i].File, inst->SrcReg[i].Index); - if (!t) - continue; - - deref_hw_reg(s, get_hw_reg(s, inst->SrcReg[i].File, inst->SrcReg[i].Index)); - - if (inst->SrcReg[i].File != RC_FILE_TEMPORARY) - continue; - - int j; - for(j = 0; j < 4; ++j) { - unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); - if (swz >= 4) - continue; - if (!t->Value[swz]) - continue; - - /* Do not free a dependency if this instruction - * also rewrites the value. See scan_instructions. */ - if (inst->DstReg.File == RC_FILE_TEMPORARY && - inst->DstReg.Index == inst->SrcReg[i].Index && - GET_BIT(inst->DstReg.WriteMask, swz)) - continue; - - if (!--t->Value[swz]->NumReaders) { - if (t->Value[swz]->Next) - decrement_dependencies(s, t->Value[swz]->Next->Writer); - } - } - } -} - - -/** - * Emit all ready texture instructions in a single block. - * - * Emit as a single block to (hopefully) sample many textures in parallel, - * and to avoid hardware indirections on R300. - * - * In R500, we don't really know when the result of a texture instruction - * arrives. So allocate all destinations first, to make sure they do not - * arrive early and overwrite a texture coordinate we're going to use later - * in the block. - */ -static void emit_all_tex(struct pair_state *s) -{ - struct pair_state_instruction *readytex; - struct pair_state_instruction *pairinst; - - assert(s->ReadyTEX); - - // Don't let the ready list change under us! - readytex = s->ReadyTEX; - s->ReadyTEX = 0; - - // Allocate destination hardware registers in one block to avoid conflicts. - for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) { - struct rc_sub_instruction *inst = &pairinst->Instruction; - if (inst->Opcode != RC_OPCODE_KIL) - get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - } - - if (s->Compiler->Base.Debug) - fprintf(stderr, " BEGIN_TEX\n"); - - if (s->Handler->BeginTexBlock) - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->BeginTexBlock(s->UserData); - - for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) { - struct rc_sub_instruction *inst = &pairinst->Instruction; - commit_instruction(s, pairinst); - - if (inst->Opcode != RC_OPCODE_KIL) - inst->DstReg.Index = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - inst->SrcReg[0].Index = get_hw_reg(s, inst->SrcReg[0].File, inst->SrcReg[0].Index); - - if (s->Compiler->Base.Debug) { - /* Should print the TEX instruction here */ - } - - struct radeon_pair_texture_instruction rpti; - - switch(inst->Opcode) { - case RC_OPCODE_TEX: rpti.Opcode = RADEON_OPCODE_TEX; break; - case RC_OPCODE_TXB: rpti.Opcode = RADEON_OPCODE_TXB; break; - case RC_OPCODE_TXP: rpti.Opcode = RADEON_OPCODE_TXP; break; - default: - case RC_OPCODE_KIL: rpti.Opcode = RADEON_OPCODE_KIL; break; - } - - rpti.DestIndex = inst->DstReg.Index; - rpti.WriteMask = inst->DstReg.WriteMask; - rpti.TexSrcUnit = inst->TexSrcUnit; - rpti.TexSrcTarget = inst->TexSrcTarget; - rpti.SrcIndex = inst->SrcReg[0].Index; - rpti.SrcSwizzle = inst->SrcReg[0].Swizzle; - - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitTex(s->UserData, &rpti); - } - - if (s->Compiler->Base.Debug) - fprintf(stderr, " END_TEX\n"); -} - - -static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instruction *pair, - struct rc_src_register src, unsigned int rgb, unsigned int alpha) +int rc_pair_alloc_source(struct rc_pair_instruction *pair, + unsigned int rgb, unsigned int alpha, + rc_register_file file, unsigned int index) { int candidate = -1; int candidate_quality = -1; int i; - if (!rgb && !alpha) + if ((!rgb && !alpha) || file == RC_FILE_NONE) return 0; - unsigned int constant; - unsigned int index; - - if (src.File == RC_FILE_TEMPORARY || src.File == RC_FILE_INPUT) { - constant = 0; - index = get_hw_reg(s, src.File, src.Index); - } else { - constant = 1; - index = src.Index; - } - for(i = 0; i < 3; ++i) { int q = 0; if (rgb) { if (pair->RGB.Src[i].Used) { - if (pair->RGB.Src[i].Constant != constant || + if (pair->RGB.Src[i].File != file || pair->RGB.Src[i].Index != index) continue; q++; @@ -607,7 +55,7 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio } if (alpha) { if (pair->Alpha.Src[i].Used) { - if (pair->Alpha.Src[i].Constant != constant || + if (pair->Alpha.Src[i].File != file || pair->Alpha.Src[i].Index != index) continue; q++; @@ -622,330 +70,15 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio if (candidate >= 0) { if (rgb) { pair->RGB.Src[candidate].Used = 1; - pair->RGB.Src[candidate].Constant = constant; + pair->RGB.Src[candidate].File = file; pair->RGB.Src[candidate].Index = index; } if (alpha) { pair->Alpha.Src[candidate].Used = 1; - pair->Alpha.Src[candidate].Constant = constant; + pair->Alpha.Src[candidate].File = file; pair->Alpha.Src[candidate].Index = index; } } return candidate; } - -/** - * Fill the given ALU instruction's opcodes and source operands into the given pair, - * if possible. - */ -static int fill_instruction_into_pair( - struct pair_state *s, - struct radeon_pair_instruction *pair, - struct pair_state_instruction *pairinst) -{ - struct rc_sub_instruction *inst = &pairinst->Instruction; - - assert(!pairinst->NeedRGB || pair->RGB.Opcode == RC_OPCODE_NOP); - assert(!pairinst->NeedAlpha || pair->Alpha.Opcode == RC_OPCODE_NOP); - - if (pairinst->NeedRGB) { - if (pairinst->IsTranscendent) - pair->RGB.Opcode = RC_OPCODE_REPL_ALPHA; - else - pair->RGB.Opcode = inst->Opcode; - if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) - pair->RGB.Saturate = 1; - } - if (pairinst->NeedAlpha) { - pair->Alpha.Opcode = inst->Opcode; - if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) - pair->Alpha.Saturate = 1; - } - - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); - int nargs = opcode->NumSrcRegs; - int i; - - /* Special case for DDX/DDY (MDH/MDV). */ - if (inst->Opcode == RC_OPCODE_DDX || inst->Opcode == RC_OPCODE_DDY) { - if (pair->RGB.Src[0].Used || pair->Alpha.Src[0].Used) - return 0; - else - nargs++; - } - - for(i = 0; i < nargs; ++i) { - int source; - if (pairinst->NeedRGB && !pairinst->IsTranscendent) { - unsigned int srcrgb = 0; - unsigned int srcalpha = 0; - int j; - for(j = 0; j < 3; ++j) { - unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); - if (swz < 3) - srcrgb = 1; - else if (swz < 4) - srcalpha = 1; - } - source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha); - if (source < 0) - return 0; - pair->RGB.Arg[i].Source = source; - pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff; - pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (RC_MASK_X | RC_MASK_Y | RC_MASK_Z)); - } - if (pairinst->NeedAlpha) { - unsigned int srcrgb = 0; - unsigned int srcalpha = 0; - unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, pairinst->IsTranscendent ? 0 : 3); - if (swz < 3) - srcrgb = 1; - else if (swz < 4) - srcalpha = 1; - source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha); - if (source < 0) - return 0; - pair->Alpha.Arg[i].Source = source; - pair->Alpha.Arg[i].Swizzle = swz; - pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & RC_MASK_W); - } - } - - return 1; -} - - -/** - * Fill in the destination register information. - * - * This is split from filling in source registers because we want - * to avoid allocating hardware temporaries for destinations until - * we are absolutely certain that we're going to emit a certain - * instruction pairing. - */ -static void fill_dest_into_pair( - struct pair_state *s, - struct radeon_pair_instruction *pair, - struct pair_state_instruction *pairinst) -{ - struct rc_sub_instruction *inst = &pairinst->Instruction; - - if (inst->DstReg.File == RC_FILE_OUTPUT) { - if (inst->DstReg.Index == s->Compiler->OutputColor) { - pair->RGB.OutputWriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; - pair->Alpha.OutputWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } else if (inst->DstReg.Index == s->Compiler->OutputDepth) { - pair->Alpha.DepthWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } - } else { - unsigned int hwindex = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - if (pairinst->NeedRGB) { - pair->RGB.DestIndex = hwindex; - pair->RGB.WriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; - } - if (pairinst->NeedAlpha) { - pair->Alpha.DestIndex = hwindex; - pair->Alpha.WriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } - } -} - - -/** - * Find a good ALU instruction or pair of ALU instruction and emit it. - * - * Prefer emitting full ALU instructions, so that when we reach a point - * where no full ALU instruction can be emitted, we have more candidates - * for RGB/Alpha pairing. - */ -static void emit_alu(struct pair_state *s) -{ - struct radeon_pair_instruction pair; - struct pair_state_instruction *psi; - - if (s->ReadyFullALU || !(s->ReadyRGB && s->ReadyAlpha)) { - if (s->ReadyFullALU) { - psi = s->ReadyFullALU; - s->ReadyFullALU = s->ReadyFullALU->NextReady; - } else if (s->ReadyRGB) { - psi = s->ReadyRGB; - s->ReadyRGB = s->ReadyRGB->NextReady; - } else { - psi = s->ReadyAlpha; - s->ReadyAlpha = s->ReadyAlpha->NextReady; - } - - memset(&pair, 0, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psi); - fill_dest_into_pair(s, &pair, psi); - commit_instruction(s, psi); - } else { - struct pair_state_instruction **prgb; - struct pair_state_instruction **palpha; - - /* Some pairings might fail because they require too - * many source slots; try all possible pairings if necessary */ - for(prgb = &s->ReadyRGB; *prgb; prgb = &(*prgb)->NextReady) { - for(palpha = &s->ReadyAlpha; *palpha; palpha = &(*palpha)->NextReady) { - struct pair_state_instruction * psirgb = *prgb; - struct pair_state_instruction * psialpha = *palpha; - memset(&pair, 0, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psirgb); - if (!fill_instruction_into_pair(s, &pair, psialpha)) - continue; - *prgb = (*prgb)->NextReady; - *palpha = (*palpha)->NextReady; - fill_dest_into_pair(s, &pair, psirgb); - fill_dest_into_pair(s, &pair, psialpha); - commit_instruction(s, psirgb); - commit_instruction(s, psialpha); - goto success; - } - } - - /* No success in pairing; just take the first RGB instruction */ - psi = s->ReadyRGB; - s->ReadyRGB = s->ReadyRGB->NextReady; - - memset(&pair, 0, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psi); - fill_dest_into_pair(s, &pair, psi); - commit_instruction(s, psi); - success: ; - } - - if (s->Compiler->Base.Debug) - radeonPrintPairInstruction(&pair); - - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitPaired(s->UserData, &pair); -} - -/* Callback function for assigning input registers to hardware registers */ -static void alloc_helper(void * data, unsigned input, unsigned hwreg) -{ - struct pair_state * s = data; - alloc_hw_reg(s, RC_FILE_INPUT, input, hwreg); -} - -void radeonPairProgram( - struct r300_fragment_program_compiler * compiler, - const struct radeon_pair_handler* handler, void *userdata) -{ - struct pair_state s; - - memset(&s, 0, sizeof(s)); - s.Compiler = compiler; - s.Handler = handler; - s.UserData = userdata; - s.Verbose = 0 && s.Compiler->Base.Debug; - - if (s.Compiler->Base.Debug) - fprintf(stderr, "Emit paired program\n"); - - scan_instructions(&s); - s.Compiler->AllocateHwInputs(s.Compiler, &alloc_helper, &s); - - while(!s.Compiler->Base.Error && - (s.ReadyTEX || s.ReadyRGB || s.ReadyAlpha || s.ReadyFullALU)) { - if (s.ReadyTEX) - emit_all_tex(&s); - - while(s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha) - emit_alu(&s); - } - - if (s.Compiler->Base.Debug) - fprintf(stderr, " END\n"); -} - - -static void print_pair_src(int i, struct radeon_pair_instruction_source* src) -{ - fprintf(stderr, " Src%i = %s[%i]", i, src->Constant ? "CNST" : "TEMP", src->Index); -} - -static const char* opcode_string(rc_opcode opcode) -{ - return rc_get_opcode_info(opcode)->Name; -} - -static int num_pairinst_args(rc_opcode opcode) -{ - return rc_get_opcode_info(opcode)->NumSrcRegs; -} - -static char swizzle_char(rc_swizzle swz) -{ - switch(swz) { - case RC_SWIZZLE_X: return 'x'; - case RC_SWIZZLE_Y: return 'y'; - case RC_SWIZZLE_Z: return 'z'; - case RC_SWIZZLE_W: return 'w'; - case RC_SWIZZLE_ZERO: return '0'; - case RC_SWIZZLE_ONE: return '1'; - case RC_SWIZZLE_HALF: return 'H'; - case RC_SWIZZLE_UNUSED: return '_'; - default: return '?'; - } -} - -void radeonPrintPairInstruction(struct radeon_pair_instruction *inst) -{ - int nargs; - int i; - - fprintf(stderr, " RGB: "); - for(i = 0; i < 3; ++i) { - if (inst->RGB.Src[i].Used) - print_pair_src(i, inst->RGB.Src + i); - } - fprintf(stderr, "\n"); - fprintf(stderr, " Alpha:"); - for(i = 0; i < 3; ++i) { - if (inst->Alpha.Src[i].Used) - print_pair_src(i, inst->Alpha.Src + i); - } - fprintf(stderr, "\n"); - - fprintf(stderr, " %s%s", opcode_string(inst->RGB.Opcode), inst->RGB.Saturate ? "_SAT" : ""); - if (inst->RGB.WriteMask) - fprintf(stderr, " TEMP[%i].%s%s%s", inst->RGB.DestIndex, - (inst->RGB.WriteMask & 1) ? "x" : "", - (inst->RGB.WriteMask & 2) ? "y" : "", - (inst->RGB.WriteMask & 4) ? "z" : ""); - if (inst->RGB.OutputWriteMask) - fprintf(stderr, " COLOR.%s%s%s", - (inst->RGB.OutputWriteMask & 1) ? "x" : "", - (inst->RGB.OutputWriteMask & 2) ? "y" : "", - (inst->RGB.OutputWriteMask & 4) ? "z" : ""); - nargs = num_pairinst_args(inst->RGB.Opcode); - for(i = 0; i < nargs; ++i) { - const char* abs = inst->RGB.Arg[i].Abs ? "|" : ""; - const char* neg = inst->RGB.Arg[i].Negate ? "-" : ""; - fprintf(stderr, ", %s%sSrc%i.%c%c%c%s", neg, abs, inst->RGB.Arg[i].Source, - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 0)), - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 1)), - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 2)), - abs); - } - fprintf(stderr, "\n"); - - fprintf(stderr, " %s%s", opcode_string(inst->Alpha.Opcode), inst->Alpha.Saturate ? "_SAT" : ""); - if (inst->Alpha.WriteMask) - fprintf(stderr, " TEMP[%i].w", inst->Alpha.DestIndex); - if (inst->Alpha.OutputWriteMask) - fprintf(stderr, " COLOR.w"); - if (inst->Alpha.DepthWriteMask) - fprintf(stderr, " DEPTH.w"); - nargs = num_pairinst_args(inst->Alpha.Opcode); - for(i = 0; i < nargs; ++i) { - const char* abs = inst->Alpha.Arg[i].Abs ? "|" : ""; - const char* neg = inst->Alpha.Arg[i].Negate ? "-" : ""; - fprintf(stderr, ", %s%sSrc%i.%c%s", neg, abs, inst->Alpha.Arg[i].Source, - swizzle_char(inst->Alpha.Arg[i].Swizzle), abs); - } - fprintf(stderr, "\n"); -} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h index 440069d558..1600598428 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h @@ -28,22 +28,37 @@ #ifndef __RADEON_PROGRAM_PAIR_H_ #define __RADEON_PROGRAM_PAIR_H_ +#include "radeon_code.h" +#include "radeon_opcodes.h" +#include "radeon_program_constants.h" + struct r300_fragment_program_compiler; /** - * Represents a paired instruction, as found in R300 and R500 + * \file + * Represents a paired ALU instruction, as found in R300 and R500 * fragment programs. + * + * Note that this representation is taking some liberties as far + * as register files are concerned, to allow separate register + * allocation. + * + * Also note that there are some subtleties in that the semantics + * of certain opcodes are implicitly changed in this representation; + * see \ref rc_pair_translate */ + + struct radeon_pair_instruction_source { - unsigned int Index:8; - unsigned int Constant:1; unsigned int Used:1; + rc_register_file File:3; + unsigned int Index:RC_REGISTER_INDEX_BITS; }; struct radeon_pair_instruction_rgb { - unsigned int Opcode:8; - unsigned int DestIndex:8; + rc_opcode Opcode:8; + unsigned int DestIndex:RC_REGISTER_INDEX_BITS; unsigned int WriteMask:3; unsigned int OutputWriteMask:3; unsigned int Saturate:1; @@ -59,8 +74,8 @@ struct radeon_pair_instruction_rgb { }; struct radeon_pair_instruction_alpha { - unsigned int Opcode:8; - unsigned int DestIndex:8; + rc_opcode Opcode:8; + unsigned int DestIndex:RC_REGISTER_INDEX_BITS; unsigned int WriteMask:1; unsigned int OutputWriteMask:1; unsigned int DepthWriteMask:1; @@ -76,66 +91,34 @@ struct radeon_pair_instruction_alpha { } Arg[3]; }; -struct radeon_pair_instruction { +struct rc_pair_instruction { struct radeon_pair_instruction_rgb RGB; struct radeon_pair_instruction_alpha Alpha; -}; - -enum { - RADEON_OPCODE_TEX = 0, - RADEON_OPCODE_TXB, - RADEON_OPCODE_TXP, - RADEON_OPCODE_KIL + rc_write_aluresult WriteALUResult:2; + rc_compare_func ALUResultCompare:3; }; -struct radeon_pair_texture_instruction { - unsigned int Opcode:2; /**< one of RADEON_OPCODE_xxx */ - - unsigned int DestIndex:8; - unsigned int WriteMask:4; - unsigned int TexSrcUnit:5; - unsigned int TexSrcTarget:3; - - unsigned int SrcIndex:8; - unsigned int SrcSwizzle:12; -}; +/** + * General helper functions for dealing with the paired instruction format. + */ +/*@{*/ +int rc_pair_alloc_source(struct rc_pair_instruction *pair, + unsigned int rgb, unsigned int alpha, + rc_register_file file, unsigned int index); +/*@}*/ /** - * + * Compiler passes that operate with the paired format. */ -struct radeon_pair_handler { - /** - * Write a paired instruction to the hardware. - * - * @return 0 on error. - */ - int (*EmitPaired)(void*, struct radeon_pair_instruction*); - - /** - * Write a texture instruction to the hardware. - * Register indices have already been rewritten to the allocated - * hardware register numbers. - * - * @return 0 on error. - */ - int (*EmitTex)(void*, struct radeon_pair_texture_instruction*); - - /** - * Called before a block of contiguous, independent texture - * instructions is emitted. - */ - int (*BeginTexBlock)(void*); - - unsigned MaxHwTemps; -}; - -void radeonPairProgram( - struct r300_fragment_program_compiler * compiler, - const struct radeon_pair_handler*, void *userdata); +/*@{*/ +struct radeon_pair_handler; -void radeonPrintPairInstruction(struct radeon_pair_instruction *inst); +void rc_pair_translate(struct r300_fragment_program_compiler *c); +void rc_pair_schedule(struct r300_fragment_program_compiler *c); +void rc_pair_regalloc(struct r300_fragment_program_compiler *c, unsigned maxtemps); +/*@}*/ #endif /* __RADEON_PROGRAM_PAIR_H_ */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c index fe90a5900e..d863b82d53 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c @@ -99,6 +99,21 @@ static void rc_print_dst_register(FILE * f, struct rc_dst_register dst) } } +static char rc_swizzle_char(unsigned int swz) +{ + switch(swz) { + case RC_SWIZZLE_X: return 'x'; + case RC_SWIZZLE_Y: return 'y'; + case RC_SWIZZLE_Z: return 'z'; + case RC_SWIZZLE_W: return 'w'; + case RC_SWIZZLE_ZERO: return '0'; + case RC_SWIZZLE_ONE: return '1'; + case RC_SWIZZLE_HALF: return 'H'; + case RC_SWIZZLE_UNUSED: return '_'; + } + return '?'; +} + static void rc_print_swizzle(FILE * f, unsigned int swizzle, unsigned int negate) { unsigned int comp; @@ -106,16 +121,7 @@ static void rc_print_swizzle(FILE * f, unsigned int swizzle, unsigned int negate rc_swizzle swz = GET_SWZ(swizzle, comp); if (GET_BIT(negate, comp)) fprintf(f, "-"); - switch(swz) { - case RC_SWIZZLE_X: fprintf(f, "x"); break; - case RC_SWIZZLE_Y: fprintf(f, "y"); break; - case RC_SWIZZLE_Z: fprintf(f, "z"); break; - case RC_SWIZZLE_W: fprintf(f, "w"); break; - case RC_SWIZZLE_ZERO: fprintf(f, "0"); break; - case RC_SWIZZLE_ONE: fprintf(f, "1"); break; - case RC_SWIZZLE_HALF: fprintf(f, "H"); break; - case RC_SWIZZLE_UNUSED: fprintf(f, "_"); break; - } + fprintf(f, "%c", rc_swizzle_char(swz)); } } @@ -142,7 +148,7 @@ static void rc_print_src_register(FILE * f, struct rc_src_register src) fprintf(f, "|"); } -static void rc_print_instruction(FILE * f, struct rc_instruction * inst) +static void rc_print_normal_instruction(FILE * f, struct rc_instruction * inst) { const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); unsigned int reg; @@ -190,6 +196,87 @@ static void rc_print_instruction(FILE * f, struct rc_instruction * inst) fprintf(f, "\n"); } +static void rc_print_pair_instruction(FILE * f, struct rc_instruction * fullinst) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + int printedsrc = 0; + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + if (printedsrc) + fprintf(f, ", "); + fprintf(f, "src%i.xyz = ", src); + rc_print_register(f, inst->RGB.Src[src].File, inst->RGB.Src[src].Index, 0); + printedsrc = 1; + } + if (inst->Alpha.Src[src].Used) { + if (printedsrc) + fprintf(f, ", "); + fprintf(f, "src%i.w = ", src); + rc_print_register(f, inst->Alpha.Src[src].File, inst->Alpha.Src[src].Index, 0); + printedsrc = 1; + } + } + fprintf(f, "\n"); + + if (inst->RGB.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->RGB.Opcode); + + fprintf(f, " %s%s", opcode->Name, inst->RGB.Saturate ? "_SAT" : ""); + if (inst->RGB.WriteMask) + fprintf(f, " temp[%i].%s%s%s", inst->RGB.DestIndex, + (inst->RGB.WriteMask & 1) ? "x" : "", + (inst->RGB.WriteMask & 2) ? "y" : "", + (inst->RGB.WriteMask & 4) ? "z" : ""); + if (inst->RGB.OutputWriteMask) + fprintf(f, " color.%s%s%s", + (inst->RGB.OutputWriteMask & 1) ? "x" : "", + (inst->RGB.OutputWriteMask & 2) ? "y" : "", + (inst->RGB.OutputWriteMask & 4) ? "z" : ""); + if (inst->WriteALUResult == RC_ALURESULT_X) + fprintf(f, " aluresult"); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + const char* abs = inst->RGB.Arg[arg].Abs ? "|" : ""; + const char* neg = inst->RGB.Arg[arg].Negate ? "-" : ""; + fprintf(f, ", %s%ssrc%i.%c%c%c%s", neg, abs, inst->RGB.Arg[arg].Source, + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 0)), + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 1)), + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 2)), + abs); + } + fprintf(f, "\n"); + } + + if (inst->Alpha.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Alpha.Opcode); + + fprintf(f, " %s%s", opcode->Name, inst->Alpha.Saturate ? "_SAT" : ""); + if (inst->Alpha.WriteMask) + fprintf(f, " temp[%i].w", inst->Alpha.DestIndex); + if (inst->Alpha.OutputWriteMask) + fprintf(f, " color.w"); + if (inst->Alpha.DepthWriteMask) + fprintf(f, " depth.w"); + if (inst->WriteALUResult == RC_ALURESULT_W) + fprintf(f, " aluresult"); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + const char* abs = inst->Alpha.Arg[arg].Abs ? "|" : ""; + const char* neg = inst->Alpha.Arg[arg].Negate ? "-" : ""; + fprintf(f, ", %s%ssrc%i.%c%s", neg, abs, inst->Alpha.Arg[arg].Source, + rc_swizzle_char(inst->Alpha.Arg[arg].Swizzle), abs); + } + fprintf(f, "\n"); + } + + if (inst->WriteALUResult) { + fprintf(f, " [aluresult = ("); + rc_print_comparefunc(f, "result", inst->ALUResultCompare, "0"); + fprintf(f, ")]\n"); + } +} + /** * Print program to stderr, default options. */ @@ -203,7 +290,10 @@ void rc_print_program(const struct rc_program *prog) for(inst = prog->Instructions.Next; inst != &prog->Instructions; inst = inst->Next) { fprintf(stderr, "%3d: ", linenum); - rc_print_instruction(stderr, inst); + if (inst->Type == RC_INSTRUCTION_PAIR) + rc_print_pair_instruction(stderr, inst); + else + rc_print_normal_instruction(stderr, inst); linenum++; } -- cgit v1.2.3