diff options
Diffstat (limited to 'src/gallium/auxiliary')
61 files changed, 4539 insertions, 733 deletions
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 2e3f5b2fc0..1c6d657297 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -158,6 +158,60 @@ static void do_triangle( struct draw_context *draw, +#define QUAD(i0,i1,i2,i3) \ + do_triangle( draw, \ + ( DRAW_PIPE_RESET_STIPPLE | \ + DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_2 ), \ + verts + stride * elts[i0], \ + verts + stride * elts[i1], \ + verts + stride * elts[i3]); \ + do_triangle( draw, \ + ( DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_1 ), \ + verts + stride * elts[i1], \ + verts + stride * elts[i2], \ + verts + stride * elts[i3]) + +#define TRIANGLE(flags,i0,i1,i2) \ + do_triangle( draw, \ + elts[i0], /* flags */ \ + verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * elts[i1], \ + verts + stride * elts[i2]) + +#define LINE(flags,i0,i1) \ + do_line( draw, \ + elts[i0], \ + verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * elts[i1]) + +#define POINT(i0) \ + do_point( draw, \ + verts + stride * elts[i0] ) + +#define FUNC pipe_run +#define ARGS \ + struct draw_context *draw, \ + unsigned prim, \ + struct vertex_header *vertices, \ + unsigned stride, \ + const ushort *elts + +#define LOCAL_VARS \ + char *verts = (char *)vertices; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned i; \ + ushort flags + +#define FLUSH + +#include "draw_pt_decompose.h" +#undef ARGS +#undef LOCAL_VARS + + /* Code to run the pipeline on a fairly arbitary collection of vertices. * @@ -178,34 +232,12 @@ void draw_pipeline_run( struct draw_context *draw, unsigned count ) { char *verts = (char *)vertices; - unsigned i; draw->pipeline.verts = verts; draw->pipeline.vertex_stride = stride; draw->pipeline.vertex_count = vertex_count; - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i++) - do_point( draw, - verts + stride * elts[i] ); - break; - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) - do_line( draw, - elts[i+0], /* flags */ - verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), - verts + stride * elts[i+1]); - break; - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) - do_triangle( draw, - elts[i+0], /* flags */ - verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), - verts + stride * elts[i+1], - verts + stride * elts[i+2]); - break; - } + pipe_run(draw, prim, vertices, stride, elts, count); draw->pipeline.verts = NULL; draw->pipeline.vertex_count = 0; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index 014d8c7346..7d76a7dbf3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -28,6 +28,30 @@ /* Authors: Keith Whitwell <keith@tungstengraphics.com> */ +/** + * Notes on wide points and sprite mode: + * + * In wide point/sprite mode we effectively need to convert each incoming + * vertex into four outgoing vertices specifying the corners of a quad. + * Since we don't (yet) have geometry shaders, we have to handle this here + * in the draw module. + * + * For sprites, it also means that this is where we have to handle texcoords + * for the vertices of the quad. OpenGL's GL_COORD_REPLACE state specifies + * if/how enabled texcoords are automatically generated for sprites. We pass + * that info through gallium in the pipe_rasterizer_state::sprite_coord_mode + * array. + * + * Additionally, GLSL's gl_PointCoord fragment attribute has to be handled + * here as well. This is basically an additional texture/generic attribute + * that varies .x from 0 to 1 horizontally across the point and varies .y + * vertically from 0 to 1 down the sprite. + * + * With geometry shaders, the state tracker could create a GS to do + * most/all of this. + */ + + #include "util/u_math.h" #include "util/u_memory.h" #include "pipe/p_defines.h" @@ -52,7 +76,7 @@ struct widepoint_stage { int psize_slot; - int point_coord_fs_input; /**< input for pointcoord (and fog) */ + int point_coord_fs_input; /**< input for pointcoord */ }; @@ -64,8 +88,6 @@ widepoint_stage( struct draw_stage *stage ) } - - /** * Set the vertex texcoords for sprite mode. * Coords may be left untouched or set to a right-side-up or upside-down @@ -89,10 +111,12 @@ static void set_texcoords(const struct widepoint_stage *wide, } if (wide->point_coord_fs_input >= 0) { - /* put gl_PointCoord into extra vertex output's zw components */ - uint k = wide->stage.draw->extra_vp_outputs.slot; - v->data[k][2] = tc[0]; - v->data[k][3] = tc[1]; + /* put gl_PointCoord into the extra vertex slot */ + uint slot = wide->stage.draw->extra_vp_outputs.slot; + v->data[slot][0] = tc[0]; + v->data[slot][1] = tc[1]; + v->data[slot][2] = 0.0F; + v->data[slot][3] = 1.0F; } } @@ -182,10 +206,10 @@ static void widepoint_point( struct draw_stage *stage, static int -find_fog_input_attrib(struct draw_context *draw) +find_pntc_input_attrib(struct draw_context *draw) { - /* Scan the fragment program's input decls to find the fogcoord - * attribute. The z/w components will store the point coord. + /* Scan the fragment program's input decls to find the pointcoord + * attribute. The xy components will store the point coord. */ return 0; /* XXX fix this */ } @@ -229,8 +253,8 @@ static void widepoint_first_point( struct draw_stage *stage, } wide->num_texcoords = j; - /* find fragment shader PointCoord/Fog input */ - wide->point_coord_fs_input = find_fog_input_attrib(draw); + /* find fragment shader PointCoord input */ + wide->point_coord_fs_input = find_pntc_input_attrib(draw); /* setup extra vp output (point coord implemented as a texcoord) */ draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 727977bc3a..b3535c0e48 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -119,7 +119,7 @@ draw_create_vs_llvm(struct draw_context *draw, vs->base.create_varient = draw_vs_varient_generic; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; - vs->machine = &draw->vs.machine; + vs->machine = draw->vs.machine; { struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 2d2af3085e..721b7d2d83 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -128,7 +128,7 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; m_functionsMap[TGSI_OPCODE_MIN] = "min"; m_functionsMap[TGSI_OPCODE_MAX] = "max"; - m_functionsMap[TGSI_OPCODE_POWER] = "pow"; + m_functionsMap[TGSI_OPCODE_POW] = "pow"; m_functionsMap[TGSI_OPCODE_LIT] = "lit"; m_functionsMap[TGSI_OPCODE_RSQ] = "rsq"; m_functionsMap[TGSI_OPCODE_SLT] = "slt"; @@ -311,7 +311,7 @@ std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> i std::vector<llvm::Value*> InstructionsSoa::pow(const std::vector<llvm::Value*> in1, const std::vector<llvm::Value*> in2) { - llvm::Function *func = function(TGSI_OPCODE_POWER); + llvm::Function *func = function(TGSI_OPCODE_POW); return callBuiltin(func, in1, in2); } diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 9c8f89d520..bf84401e11 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -338,7 +338,7 @@ translate_instruction(llvm::Module *module, out = instr->sub(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_LERP: { + case TGSI_OPCODE_LRP: { out = instr->lerp(inputs[0], inputs[1], inputs[2]); } break; @@ -350,17 +350,11 @@ translate_instruction(llvm::Module *module, out = instr->cnd0(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_DOT2ADD: { + case TGSI_OPCODE_DP2A: { out = instr->dot2add(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: { - out = instr->neg(inputs[0]); - } - break; - case TGSI_OPCODE_FRAC: { + case TGSI_OPCODE_FRC: { out = instr->frc(inputs[0]); } break; @@ -368,30 +362,28 @@ translate_instruction(llvm::Module *module, out = instr->clamp(inputs[0]); } break; - case TGSI_OPCODE_FLOOR: { + case TGSI_OPCODE_FLR: { out = instr->floor(inputs[0]); } break; case TGSI_OPCODE_ROUND: break; - case TGSI_OPCODE_EXPBASE2: { + case TGSI_OPCODE_EX2: { out = instr->ex2(inputs[0]); } break; - case TGSI_OPCODE_LOGBASE2: { + case TGSI_OPCODE_LG2: { out = instr->lg2(inputs[0]); } break; - case TGSI_OPCODE_POWER: { + case TGSI_OPCODE_POW: { out = instr->pow(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_CROSSPRODUCT: { + case TGSI_OPCODE_XPD: { out = instr->cross(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; case TGSI_OPCODE_ABS: { out = instr->abs(inputs[0]); } @@ -524,7 +516,7 @@ translate_instruction(llvm::Module *module, return; //just update the state } break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: break; case TGSI_OPCODE_REP: break; @@ -540,7 +532,7 @@ translate_instruction(llvm::Module *module, return; //just update the state } break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: break; case TGSI_OPCODE_ENDREP: break; @@ -582,7 +574,7 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_ENDPRIM: break; - case TGSI_OPCODE_BGNLOOP2: { + case TGSI_OPCODE_BGNLOOP: { instr->beginLoop(); storage->setCurrentBlock(instr->currentBlock()); return; @@ -595,7 +587,7 @@ translate_instruction(llvm::Module *module, return; } break; - case TGSI_OPCODE_ENDLOOP2: { + case TGSI_OPCODE_ENDLOOP: { instr->endLoop(); storage->setCurrentBlock(instr->currentBlock()); return; @@ -619,14 +611,6 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; case TGSI_OPCODE_CALLNZ: break; case TGSI_OPCODE_IFC: @@ -780,44 +764,38 @@ translate_instructionir(llvm::Module *module, out = instr->sub(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_LERP: { + case TGSI_OPCODE_LRP: { } break; case TGSI_OPCODE_CND: break; case TGSI_OPCODE_CND0: break; - case TGSI_OPCODE_DOT2ADD: + case TGSI_OPCODE_DP2A: break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { + case TGSI_OPCODE_FRC: { } break; case TGSI_OPCODE_CLAMP: break; - case TGSI_OPCODE_FLOOR: { + case TGSI_OPCODE_FLR: { } break; case TGSI_OPCODE_ROUND: break; - case TGSI_OPCODE_EXPBASE2: { + case TGSI_OPCODE_EX2: { } break; - case TGSI_OPCODE_LOGBASE2: { + case TGSI_OPCODE_LG2: { } break; - case TGSI_OPCODE_POWER: { + case TGSI_OPCODE_POW: { out = instr->pow(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_CROSSPRODUCT: { + case TGSI_OPCODE_XPD: { } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; case TGSI_OPCODE_ABS: { out = instr->abs(inputs[0]); } @@ -912,7 +890,7 @@ translate_instructionir(llvm::Module *module, case TGSI_OPCODE_IF: { } break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: break; case TGSI_OPCODE_REP: break; @@ -922,7 +900,7 @@ translate_instructionir(llvm::Module *module, case TGSI_OPCODE_ENDIF: { } break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: break; case TGSI_OPCODE_ENDREP: break; @@ -963,13 +941,13 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_ENDPRIM: break; - case TGSI_OPCODE_BGNLOOP2: { + case TGSI_OPCODE_BGNLOOP: { } break; case TGSI_OPCODE_BGNSUB: { } break; - case TGSI_OPCODE_ENDLOOP2: { + case TGSI_OPCODE_ENDLOOP: { } break; case TGSI_OPCODE_ENDSUB: { @@ -985,14 +963,6 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; case TGSI_OPCODE_NRM4: break; case TGSI_OPCODE_CALLNZ: diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index b4900e8dba..5f0a580b09 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -16,6 +16,7 @@ C_SOURCES = \ tgsi_sse2.c \ tgsi_text.c \ tgsi_transform.c \ + tgsi_ureg.c \ tgsi_util.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 8200cce42f..b6bc2924f0 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -16,6 +16,7 @@ tgsi = env.ConvenienceLibrary( 'tgsi_sse2.c', 'tgsi_text.c', 'tgsi_transform.c', + 'tgsi_ureg.c', 'tgsi_util.c', ]) diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt index a3f4947c73..eb492076b7 100644 --- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt +++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt @@ -187,11 +187,7 @@ TGSI Instruction Specification 1.2.6 CND0 - Condition Zero - dst.x = (src2.x >= 0.0) ? src0.x : src1.x - dst.y = (src2.y >= 0.0) ? src0.y : src1.y - dst.z = (src2.z >= 0.0) ? src0.z : src1.z - dst.w = (src2.w >= 0.0) ? src0.w : src1.w - + Removed. Use (CMP src2, src1, src0) instead. 1.2.7 DOT2ADD - 2-component Dot Product And Add @@ -665,9 +661,18 @@ TGSI Instruction Specification TBD -1.9.8 LOOP - Loop +1.9.8 BGNFOR - Begin a For-Loop - TBD + dst.x = floor(src.x) + dst.y = floor(src.y) + dst.z = floor(src.z) + + if (dst.y <= 0) + pc = [matching ENDFOR] + 1 + endif + + Note: The destination must be a loop register. + The source must be a constant register. 1.9.9 REP - Repeat @@ -685,9 +690,16 @@ TGSI Instruction Specification TBD -1.9.12 ENDLOOP - End Loop +1.9.12 ENDFOR - End a For-Loop - TBD + dst.x = dst.x + dst.z + dst.y = dst.y - 1.0 + + if (dst.y > 0) + pc = [matching BGNFOR instruction] + 1 + endif + + Note: The destination must be a loop register. 1.9.13 ENDREP - End Repeat @@ -840,7 +852,7 @@ TGSI Instruction Specification ---------- -1.13.1 BGNLOOP2 - Begin Loop +1.13.1 BGNLOOP - Begin a Loop TBD @@ -850,7 +862,7 @@ TGSI Instruction Specification TBD -1.13.3 ENDLOOP2 - End Loop +1.13.3 ENDLOOP - End a Loop TBD @@ -1015,12 +1027,12 @@ TGSI Instruction Specification 1.18.1 EXPP - Approximate Exponential Base 2 - Alias for EXP. + Use EXP. See also 1.19.3. 1.18.2 LOGP - Logarithm Base 2 - Alias for LG2. + Use LOG. See also 1.19.4. 1.19 vs_2_0 @@ -1037,6 +1049,16 @@ TGSI Instruction Specification Alias for ARR. +1.19.3 EXPP - Approximate Exponential Base 2 + + Use EX2. + + +1.19.4 LOGP - Logarithm Base 2 + + Use LG2. + + 2 Explanation of symbols used ============================== diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 010d501c60..e0cfc54420 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -482,6 +482,8 @@ tgsi_default_full_instruction( void ) full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); } + full_instruction.Flags = 0x0; + return full_instruction; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index e1cd8479cb..111d95b666 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -33,12 +33,20 @@ #include "tgsi_info.h" #include "tgsi_iterate.h" + +/** Number of spaces to indent for IF/LOOP/etc */ +static const int indent_spaces = 3; + + struct dump_ctx { struct tgsi_iterate_context iter; uint instno; + int indent; + uint indentation; + void (*printf)(struct dump_ctx *ctx, const char *format, ...); }; @@ -82,7 +90,7 @@ static const char *processor_type_names[] = "GEOM" }; -static const char *file_names[] = +static const char *file_names[TGSI_FILE_COUNT] = { "NULL", "CONST", @@ -91,7 +99,8 @@ static const char *file_names[] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static const char *interpolate_names[] = @@ -334,13 +343,19 @@ iter_instruction( { struct dump_ctx *ctx = (struct dump_ctx *) iter; uint instno = ctx->instno++; - + const struct tgsi_opcode_info *info = tgsi_get_opcode_info( inst->Instruction.Opcode ); uint i; boolean first_reg = TRUE; INSTID( instno ); TXT( ": " ); - TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic ); + + ctx->indent -= info->pre_dedent; + for(i = 0; (int)i < ctx->indent; ++i) + TXT( " " ); + ctx->indent += info->post_indent; + + TXT( info->mnemonic ); switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: @@ -472,14 +487,22 @@ iter_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_BGNLOOP: + case TGSI_OPCODE_ENDLOOP: case TGSI_OPCODE_CAL: TXT( " :" ); UID( inst->InstructionExtLabel.Label ); break; } + /* update indentation */ + if (inst->Instruction.Opcode == TGSI_OPCODE_IF || + inst->Instruction.Opcode == TGSI_OPCODE_ELSE || + inst->Instruction.Opcode == TGSI_OPCODE_BGNFOR || + inst->Instruction.Opcode == TGSI_OPCODE_BGNLOOP) { + ctx->indentation += indent_spaces; + } + EOL(); return TRUE; @@ -493,7 +516,9 @@ tgsi_dump_instruction( struct dump_ctx ctx; ctx.instno = instno; + ctx.indent = 0; ctx.printf = dump_ctx_printf; + ctx.indentation = 0; iter_instruction( &ctx.iter, (struct tgsi_full_instruction *)inst ); } @@ -525,7 +550,9 @@ tgsi_dump( ctx.iter.epilog = NULL; ctx.instno = 0; + ctx.indent = 0; ctx.printf = dump_ctx_printf; + ctx.indentation = 0; tgsi_iterate_shader( tokens, &ctx.iter ); } @@ -577,7 +604,9 @@ tgsi_dump_str( ctx.base.iter.epilog = NULL; ctx.base.instno = 0; + ctx.base.indent = 0; ctx.base.printf = &str_dump_ctx_printf; + ctx.base.indentation = 0; ctx.str = str; ctx.str[0] = 0; diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c index c944760ca6..4a9c02b141 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -69,7 +69,7 @@ static const char *TGSI_TOKEN_TYPES[] = "TOKEN_TYPE_INSTRUCTION" }; -static const char *TGSI_FILES[] = +static const char *TGSI_FILES[TGSI_FILE_COUNT] = { "FILE_NULL", "FILE_CONSTANT", @@ -78,7 +78,8 @@ static const char *TGSI_FILES[] = "FILE_TEMPORARY", "FILE_SAMPLER", "FILE_ADDRESS", - "FILE_IMMEDIATE" + "FILE_IMMEDIATE", + "FILE_LOOP" }; static const char *TGSI_INTERPOLATES[] = diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index b193fd7a0a..c79c56debd 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -62,6 +62,9 @@ #define FAST_MATH 1 +/** for tgsi_full_instruction::Flags */ +#define SOA_DEPENDENCY_FLAG 0x1 + #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 #define TILE_BOTTOM_LEFT 2 @@ -182,7 +185,7 @@ print_temp(const struct tgsi_exec_machine *mach, uint index) * MOV t3, t2; * The second instruction will have the wrong value for t0 if executed as-is. */ -static boolean +boolean tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst) { uint i, chan; @@ -328,19 +331,24 @@ tgsi_exec_machine_bind_shader( * sizeof(struct tgsi_full_instruction)); maxInstructions += 10; } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); -#if 0 if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) { - debug_printf("SOA dependency in instruction:\n"); - tgsi_dump_instruction(&parse.FullToken.FullInstruction, - numInstructions); + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + parse.FullToken.FullInstruction.Flags = SOA_DEPENDENCY_FLAG; + /* XXX we only handle SOA dependencies properly for MOV/SWZ + * at this time! + */ + if (opcode != TGSI_OPCODE_MOV && opcode != TGSI_OPCODE_SWZ) { + debug_printf("Warning: SOA dependency in instruction" + " is not handled:\n"); + tgsi_dump_instruction(&parse.FullToken.FullInstruction, + numInstructions); + } } -#else - (void) tgsi_check_soa_dependencies; -#endif + + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); numInstructions++; break; @@ -375,15 +383,9 @@ tgsi_exec_machine_create( void ) if (!mach) goto fail; - mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; + memset(mach, 0, sizeof(*mach)); - mach->Samplers = NULL; - mach->Consts = NULL; - mach->Tokens = NULL; - mach->Primitives = NULL; - mach->InterpCoefs = NULL; - mach->Instructions = NULL; - mach->Declarations = NULL; + mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; /* Setup constants. */ for( i = 0; i < 4; i++ ) { @@ -2030,9 +2032,23 @@ exec_instruction( case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); + if (inst->Flags & SOA_DEPENDENCY_FLAG) { + /* Do all fetches into temp regs, then do all stores to avoid + * intermediate/accidental clobbering. This could be done all the + * time for MOV but for other instructions we'll need more temps... + */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[chan_index], 0, chan_index ); + } + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } + } + else { + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } } break; @@ -2313,16 +2329,6 @@ exec_instruction( } break; - case TGSI_OPCODE_CND0: - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - micro_le(&r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[2], &r[0], &r[1]); - STORE(&r[0], 0, chan_index); - } - break; - case TGSI_OPCODE_DP2A: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); @@ -2341,16 +2347,6 @@ exec_instruction( } break; - case TGSI_OPCODE_INDEX: - /* XXX: considered for removal */ - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - /* XXX: considered for removal */ - assert (0); - break; - case TGSI_OPCODE_FRC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); @@ -2454,11 +2450,6 @@ exec_instruction( } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - /* XXX: considered for removal */ - assert (0); - break; - case TGSI_OPCODE_ABS: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); @@ -2765,19 +2756,32 @@ exec_instruction( if (mach->ExecMask) { /* do the call */ - /* push the Cond, Loop, Cont stacks */ + /* First, record the depths of the execution stacks. + * This is important for deeply nested/looped return statements. + * We have to unwind the stacks by the correct amount. For a + * real code generator, we could determine the number of entries + * to pop off each stack with simple static analysis and avoid + * implementing this data structure at run time. + */ + mach->CallStack[mach->CallStackTop].CondStackTop = mach->CondStackTop; + mach->CallStack[mach->CallStackTop].LoopStackTop = mach->LoopStackTop; + mach->CallStack[mach->CallStackTop].ContStackTop = mach->ContStackTop; + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop].ReturnAddr = *pc; + + mach->CallStackTop++; + + /* Second, push the Cond, Loop, Cont, Func stacks */ assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); mach->CondStack[mach->CondStackTop++] = mach->CondMask; assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->ContStack[mach->ContStackTop++] = mach->ContMask; - assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; + /* Finally, jump to the subroutine */ *pc = inst->InstructionExtLabel.Label; } break; @@ -2794,18 +2798,24 @@ exec_instruction( *pc = -1; return; } - *pc = mach->CallStack[--mach->CallStackTop]; - /* pop the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->CallStackTop > 0); + mach->CallStackTop--; + + mach->CondStackTop = mach->CallStack[mach->CallStackTop].CondStackTop; + mach->CondMask = mach->CondStack[mach->CondStackTop]; + + mach->LoopStackTop = mach->CallStack[mach->CallStackTop].LoopStackTop; + mach->LoopMask = mach->LoopStack[mach->LoopStackTop]; + + mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop; + mach->ContMask = mach->ContStack[mach->ContStackTop]; + assert(mach->FuncStackTop > 0); mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + *pc = mach->CallStack[mach->CallStackTop].ReturnAddr; + UPDATE_EXEC_MASK(mach); } break; @@ -3102,26 +3112,81 @@ exec_instruction( mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: + assert(mach->LoopCounterStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + for (chan_index = 0; chan_index < 3; chan_index++) { + FETCH( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[chan_index], 0, chan_index ); + } + STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); + ++mach->LoopCounterStackTop; /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->ContStack[mach->ContStackTop++] = mach->ContMask; + assert(mach->LoopLabelStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopLabelStack[mach->LoopLabelStackTop++] = *pc - 1; + break; + + case TGSI_OPCODE_ENDFOR: + assert(mach->LoopCounterStackTop > 0); + micro_sub( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + /* update LoopMask */ + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[0] <= 0) { + mach->LoopMask &= ~0x1; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[1] <= 0 ) { + mach->LoopMask &= ~0x2; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[2] <= 0 ) { + mach->LoopMask &= ~0x4; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[3] <= 0 ) { + mach->LoopMask &= ~0x8; + } + micro_add( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]); + assert(mach->LoopLabelStackTop > 0); + inst = mach->Instructions + mach->LoopLabelStack[mach->LoopLabelStackTop - 1]; + STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + UPDATE_EXEC_MASK(mach); + if (mach->ExecMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + assert(mach->LoopLabelStackTop > 0); + *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->LoopLabelStackTop > 0); + --mach->LoopLabelStackTop; + assert(mach->LoopCounterStackTop > 0); + --mach->LoopCounterStackTop; + } + UPDATE_EXEC_MASK(mach); break; - + case TGSI_OPCODE_ENDLOOP: - /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: /* Restore ContMask, but don't pop */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; UPDATE_EXEC_MASK(mach); if (mach->ExecMask) { /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; + assert(mach->LoopLabelStackTop > 0); + *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1; } else { /* exit loop: pop LoopMask */ @@ -3130,6 +3195,8 @@ exec_instruction( /* pop ContMask */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->LoopLabelStackTop > 0); + --mach->LoopLabelStackTop; } UPDATE_EXEC_MASK(mach); break; @@ -3197,7 +3264,6 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->FuncMask = 0xf; mach->ExecMask = 0xf; - mach->CondStackTop = 0; /* temporarily subvert this assertion */ assert(mach->CondStackTop == 0); assert(mach->LoopStackTop == 0); assert(mach->ContStackTop == 0); diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index 8a9100f4c3..c72f76809d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -186,6 +186,17 @@ struct tgsi_exec_labels */ #define TGSI_EXEC_MAX_CONST_BUFFER 4096 + +/** function call/activation record */ +struct tgsi_call_record +{ + uint CondStackTop; + uint LoopStackTop; + uint ContStackTop; + uint ReturnAddr; +}; + + /** * Run-time virtual machine state for executing TGSI shader. */ @@ -232,6 +243,14 @@ struct tgsi_exec_machine uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopStackTop; + /** Loop label stack */ + uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopLabelStackTop; + + /** Loop counter stack (x = count, y = current, z = step) */ + struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopCounterStackTop; + /** Loop continue mask stack (see comments in tgsi_exec.c) */ uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; int ContStackTop; @@ -241,7 +260,7 @@ struct tgsi_exec_machine int FuncStackTop; /** Function call stack for saving/restoring the program counter */ - uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + struct tgsi_call_record CallStack[TGSI_EXEC_MAX_CALL_NESTING]; int CallStackTop; struct tgsi_full_instruction *Instructions; @@ -272,6 +291,14 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); + + +boolean +tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst); + + static INLINE void tgsi_set_kill_mask(struct tgsi_exec_machine *mach, unsigned mask) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index 29b8115560..17af4cb7ad 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -31,125 +31,125 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { - { 1, 1, 0, 0, "ARL", TGSI_OPCODE_ARL }, - { 1, 1, 0, 0, "MOV", TGSI_OPCODE_MOV }, - { 1, 1, 0, 0, "LIT", TGSI_OPCODE_LIT }, - { 1, 1, 0, 0, "RCP", TGSI_OPCODE_RCP }, - { 1, 1, 0, 0, "RSQ", TGSI_OPCODE_RSQ }, - { 1, 1, 0, 0, "EXP", TGSI_OPCODE_EXP }, - { 1, 1, 0, 0, "LOG", TGSI_OPCODE_LOG }, - { 1, 2, 0, 0, "MUL", TGSI_OPCODE_MUL }, - { 1, 2, 0, 0, "ADD", TGSI_OPCODE_ADD }, - { 1, 2, 0, 0, "DP3", TGSI_OPCODE_DP3 }, - { 1, 2, 0, 0, "DP4", TGSI_OPCODE_DP4 }, - { 1, 2, 0, 0, "DST", TGSI_OPCODE_DST }, - { 1, 2, 0, 0, "MIN", TGSI_OPCODE_MIN }, - { 1, 2, 0, 0, "MAX", TGSI_OPCODE_MAX }, - { 1, 2, 0, 0, "SLT", TGSI_OPCODE_SLT }, - { 1, 2, 0, 0, "SGE", TGSI_OPCODE_SGE }, - { 1, 3, 0, 0, "MAD", TGSI_OPCODE_MAD }, - { 1, 2, 0, 0, "SUB", TGSI_OPCODE_SUB }, - { 1, 3, 0, 0, "LRP", TGSI_OPCODE_LRP }, - { 1, 3, 0, 0, "CND", TGSI_OPCODE_CND }, - { 1, 3, 0, 0, "CND0", TGSI_OPCODE_CND0 }, - { 1, 3, 0, 0, "DP2A", TGSI_OPCODE_DP2A }, - { 1, 2, 0, 0, "INDEX", TGSI_OPCODE_INDEX }, - { 1, 1, 0, 0, "NEGATE", TGSI_OPCODE_NEGATE }, - { 1, 1, 0, 0, "FRC", TGSI_OPCODE_FRC }, - { 1, 3, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP }, - { 1, 1, 0, 0, "FLR", TGSI_OPCODE_FLR }, - { 1, 1, 0, 0, "ROUND", TGSI_OPCODE_ROUND }, - { 1, 1, 0, 0, "EX2", TGSI_OPCODE_EX2 }, - { 1, 1, 0, 0, "LG2", TGSI_OPCODE_LG2 }, - { 1, 2, 0, 0, "POW", TGSI_OPCODE_POW }, - { 1, 2, 0, 0, "XPD", TGSI_OPCODE_XPD }, - { 1, 2, 0, 0, "M4X4", TGSI_OPCODE_MULTIPLYMATRIX }, - { 1, 1, 0, 0, "ABS", TGSI_OPCODE_ABS }, - { 1, 1, 0, 0, "RCC", TGSI_OPCODE_RCC }, - { 1, 2, 0, 0, "DPH", TGSI_OPCODE_DPH }, - { 1, 1, 0, 0, "COS", TGSI_OPCODE_COS }, - { 1, 1, 0, 0, "DDX", TGSI_OPCODE_DDX }, - { 1, 1, 0, 0, "DDY", TGSI_OPCODE_DDY }, - { 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP }, - { 1, 1, 0, 0, "PK2H", TGSI_OPCODE_PK2H }, - { 1, 1, 0, 0, "PK2US", TGSI_OPCODE_PK2US }, - { 1, 1, 0, 0, "PK4B", TGSI_OPCODE_PK4B }, - { 1, 1, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB }, - { 1, 2, 0, 0, "RFL", TGSI_OPCODE_RFL }, - { 1, 2, 0, 0, "SEQ", TGSI_OPCODE_SEQ }, - { 1, 2, 0, 0, "SFL", TGSI_OPCODE_SFL }, - { 1, 2, 0, 0, "SGT", TGSI_OPCODE_SGT }, - { 1, 1, 0, 0, "SIN", TGSI_OPCODE_SIN }, - { 1, 2, 0, 0, "SLE", TGSI_OPCODE_SLE }, - { 1, 2, 0, 0, "SNE", TGSI_OPCODE_SNE }, - { 1, 2, 0, 0, "STR", TGSI_OPCODE_STR }, - { 1, 2, 1, 0, "TEX", TGSI_OPCODE_TEX }, - { 1, 4, 1, 0, "TXD", TGSI_OPCODE_TXD }, - { 1, 2, 1, 0, "TXP", TGSI_OPCODE_TXP }, - { 1, 1, 0, 0, "UP2H", TGSI_OPCODE_UP2H }, - { 1, 1, 0, 0, "UP2US", TGSI_OPCODE_UP2US }, - { 1, 1, 0, 0, "UP4B", TGSI_OPCODE_UP4B }, - { 1, 1, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB }, - { 1, 3, 0, 0, "X2D", TGSI_OPCODE_X2D }, - { 1, 1, 0, 0, "ARA", TGSI_OPCODE_ARA }, - { 1, 1, 0, 0, "ARR", TGSI_OPCODE_ARR }, - { 0, 1, 0, 0, "BRA", TGSI_OPCODE_BRA }, - { 0, 0, 0, 1, "CAL", TGSI_OPCODE_CAL }, - { 0, 0, 0, 0, "RET", TGSI_OPCODE_RET }, - { 1, 1, 0, 0, "SSG", TGSI_OPCODE_SSG }, - { 1, 3, 0, 0, "CMP", TGSI_OPCODE_CMP }, - { 1, 1, 0, 0, "SCS", TGSI_OPCODE_SCS }, - { 1, 2, 1, 0, "TXB", TGSI_OPCODE_TXB }, - { 1, 1, 0, 0, "NRM", TGSI_OPCODE_NRM }, - { 1, 2, 0, 0, "DIV", TGSI_OPCODE_DIV }, - { 1, 2, 0, 0, "DP2", TGSI_OPCODE_DP2 }, - { 1, 2, 1, 0, "TXL", TGSI_OPCODE_TXL }, - { 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK }, - { 0, 1, 0, 1, "IF", TGSI_OPCODE_IF }, - { 0, 0, 0, 0, "LOOP", TGSI_OPCODE_LOOP }, - { 0, 1, 0, 0, "REP", TGSI_OPCODE_REP }, - { 0, 0, 0, 1, "ELSE", TGSI_OPCODE_ELSE }, - { 0, 0, 0, 0, "ENDIF", TGSI_OPCODE_ENDIF }, - { 0, 0, 0, 0, "ENDLOOP", TGSI_OPCODE_ENDLOOP }, - { 0, 0, 0, 0, "ENDREP", TGSI_OPCODE_ENDREP }, - { 0, 1, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA }, - { 1, 0, 0, 0, "POPA", TGSI_OPCODE_POPA }, - { 1, 1, 0, 0, "CEIL", TGSI_OPCODE_CEIL }, - { 1, 1, 0, 0, "I2F", TGSI_OPCODE_I2F }, - { 1, 1, 0, 0, "NOT", TGSI_OPCODE_NOT }, - { 1, 1, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, - { 1, 2, 0, 0, "SHL", TGSI_OPCODE_SHL }, - { 1, 2, 0, 0, "SHR", TGSI_OPCODE_SHR }, - { 1, 2, 0, 0, "AND", TGSI_OPCODE_AND }, - { 1, 2, 0, 0, "OR", TGSI_OPCODE_OR }, - { 1, 2, 0, 0, "MOD", TGSI_OPCODE_MOD }, - { 1, 2, 0, 0, "XOR", TGSI_OPCODE_XOR }, - { 1, 3, 0, 0, "SAD", TGSI_OPCODE_SAD }, - { 1, 2, 1, 0, "TXF", TGSI_OPCODE_TXF }, - { 1, 2, 1, 0, "TXQ", TGSI_OPCODE_TXQ }, - { 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT }, - { 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT }, - { 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM }, - { 0, 0, 0, 1, "BGNLOOP2", TGSI_OPCODE_BGNLOOP2 }, - { 0, 0, 0, 0, "BGNSUB", TGSI_OPCODE_BGNSUB }, - { 0, 0, 0, 1, "ENDLOOP2", TGSI_OPCODE_ENDLOOP2 }, - { 0, 0, 0, 0, "ENDSUB", TGSI_OPCODE_ENDSUB }, - { 1, 1, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 }, - { 1, 1, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 }, - { 1, 1, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 }, - { 1, 1, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 }, - { 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP }, - { 1, 2, 0, 0, "M4X3", TGSI_OPCODE_M4X3 }, - { 1, 2, 0, 0, "M3X4", TGSI_OPCODE_M3X4 }, - { 1, 2, 0, 0, "M3X3", TGSI_OPCODE_M3X3 }, - { 1, 2, 0, 0, "M3X2", TGSI_OPCODE_M3X2 }, - { 1, 1, 0, 0, "NRM4", TGSI_OPCODE_NRM4 }, - { 0, 1, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ }, - { 0, 1, 0, 0, "IFC", TGSI_OPCODE_IFC }, - { 0, 1, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, - { 0, 1, 0, 0, "KIL", TGSI_OPCODE_KIL }, - { 0, 0, 0, 0, "END", TGSI_OPCODE_END }, - { 1, 1, 0, 0, "SWZ", TGSI_OPCODE_SWZ } + { 1, 1, 0, 0, 0, 0, "ARL", TGSI_OPCODE_ARL }, + { 1, 1, 0, 0, 0, 0, "MOV", TGSI_OPCODE_MOV }, + { 1, 1, 0, 0, 0, 0, "LIT", TGSI_OPCODE_LIT }, + { 1, 1, 0, 0, 0, 0, "RCP", TGSI_OPCODE_RCP }, + { 1, 1, 0, 0, 0, 0, "RSQ", TGSI_OPCODE_RSQ }, + { 1, 1, 0, 0, 0, 0, "EXP", TGSI_OPCODE_EXP }, + { 1, 1, 0, 0, 0, 0, "LOG", TGSI_OPCODE_LOG }, + { 1, 2, 0, 0, 0, 0, "MUL", TGSI_OPCODE_MUL }, + { 1, 2, 0, 0, 0, 0, "ADD", TGSI_OPCODE_ADD }, + { 1, 2, 0, 0, 0, 0, "DP3", TGSI_OPCODE_DP3 }, + { 1, 2, 0, 0, 0, 0, "DP4", TGSI_OPCODE_DP4 }, + { 1, 2, 0, 0, 0, 0, "DST", TGSI_OPCODE_DST }, + { 1, 2, 0, 0, 0, 0, "MIN", TGSI_OPCODE_MIN }, + { 1, 2, 0, 0, 0, 0, "MAX", TGSI_OPCODE_MAX }, + { 1, 2, 0, 0, 0, 0, "SLT", TGSI_OPCODE_SLT }, + { 1, 2, 0, 0, 0, 0, "SGE", TGSI_OPCODE_SGE }, + { 1, 3, 0, 0, 0, 0, "MAD", TGSI_OPCODE_MAD }, + { 1, 2, 0, 0, 0, 0, "SUB", TGSI_OPCODE_SUB }, + { 1, 3, 0, 0, 0, 0, "LRP", TGSI_OPCODE_LRP }, + { 1, 3, 0, 0, 0, 0, "CND", TGSI_OPCODE_CND }, + { 0, 0, 0, 0, 0, 0, "", 20 }, /* removed */ + { 1, 3, 0, 0, 0, 0, "DP2A", TGSI_OPCODE_DP2A }, + { 0, 0, 0, 0, 0, 0, "", 22 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 23 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "FRC", TGSI_OPCODE_FRC }, + { 1, 3, 0, 0, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP }, + { 1, 1, 0, 0, 0, 0, "FLR", TGSI_OPCODE_FLR }, + { 1, 1, 0, 0, 0, 0, "ROUND", TGSI_OPCODE_ROUND }, + { 1, 1, 0, 0, 0, 0, "EX2", TGSI_OPCODE_EX2 }, + { 1, 1, 0, 0, 0, 0, "LG2", TGSI_OPCODE_LG2 }, + { 1, 2, 0, 0, 0, 0, "POW", TGSI_OPCODE_POW }, + { 1, 2, 0, 0, 0, 0, "XPD", TGSI_OPCODE_XPD }, + { 0, 0, 0, 0, 0, 0, "", 32 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "ABS", TGSI_OPCODE_ABS }, + { 1, 1, 0, 0, 0, 0, "RCC", TGSI_OPCODE_RCC }, + { 1, 2, 0, 0, 0, 0, "DPH", TGSI_OPCODE_DPH }, + { 1, 1, 0, 0, 0, 0, "COS", TGSI_OPCODE_COS }, + { 1, 1, 0, 0, 0, 0, "DDX", TGSI_OPCODE_DDX }, + { 1, 1, 0, 0, 0, 0, "DDY", TGSI_OPCODE_DDY }, + { 0, 0, 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP }, + { 1, 1, 0, 0, 0, 0, "PK2H", TGSI_OPCODE_PK2H }, + { 1, 1, 0, 0, 0, 0, "PK2US", TGSI_OPCODE_PK2US }, + { 1, 1, 0, 0, 0, 0, "PK4B", TGSI_OPCODE_PK4B }, + { 1, 1, 0, 0, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB }, + { 1, 2, 0, 0, 0, 0, "RFL", TGSI_OPCODE_RFL }, + { 1, 2, 0, 0, 0, 0, "SEQ", TGSI_OPCODE_SEQ }, + { 1, 2, 0, 0, 0, 0, "SFL", TGSI_OPCODE_SFL }, + { 1, 2, 0, 0, 0, 0, "SGT", TGSI_OPCODE_SGT }, + { 1, 1, 0, 0, 0, 0, "SIN", TGSI_OPCODE_SIN }, + { 1, 2, 0, 0, 0, 0, "SLE", TGSI_OPCODE_SLE }, + { 1, 2, 0, 0, 0, 0, "SNE", TGSI_OPCODE_SNE }, + { 1, 2, 0, 0, 0, 0, "STR", TGSI_OPCODE_STR }, + { 1, 2, 1, 0, 0, 0, "TEX", TGSI_OPCODE_TEX }, + { 1, 4, 1, 0, 0, 0, "TXD", TGSI_OPCODE_TXD }, + { 1, 2, 1, 0, 0, 0, "TXP", TGSI_OPCODE_TXP }, + { 1, 1, 0, 0, 0, 0, "UP2H", TGSI_OPCODE_UP2H }, + { 1, 1, 0, 0, 0, 0, "UP2US", TGSI_OPCODE_UP2US }, + { 1, 1, 0, 0, 0, 0, "UP4B", TGSI_OPCODE_UP4B }, + { 1, 1, 0, 0, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB }, + { 1, 3, 0, 0, 0, 0, "X2D", TGSI_OPCODE_X2D }, + { 1, 1, 0, 0, 0, 0, "ARA", TGSI_OPCODE_ARA }, + { 1, 1, 0, 0, 0, 0, "ARR", TGSI_OPCODE_ARR }, + { 0, 1, 0, 0, 0, 0, "BRA", TGSI_OPCODE_BRA }, + { 0, 0, 0, 1, 0, 0, "CAL", TGSI_OPCODE_CAL }, + { 0, 0, 0, 0, 0, 0, "RET", TGSI_OPCODE_RET }, + { 1, 1, 0, 0, 0, 0, "SSG", TGSI_OPCODE_SSG }, + { 1, 3, 0, 0, 0, 0, "CMP", TGSI_OPCODE_CMP }, + { 1, 1, 0, 0, 0, 0, "SCS", TGSI_OPCODE_SCS }, + { 1, 2, 1, 0, 0, 0, "TXB", TGSI_OPCODE_TXB }, + { 1, 1, 0, 0, 0, 0, "NRM", TGSI_OPCODE_NRM }, + { 1, 2, 0, 0, 0, 0, "DIV", TGSI_OPCODE_DIV }, + { 1, 2, 0, 0, 0, 0, "DP2", TGSI_OPCODE_DP2 }, + { 1, 2, 1, 0, 0, 0, "TXL", TGSI_OPCODE_TXL }, + { 0, 0, 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK }, + { 0, 1, 0, 1, 0, 1, "IF", TGSI_OPCODE_IF }, + { 1, 1, 0, 0, 0, 1, "BGNFOR", TGSI_OPCODE_BGNFOR }, + { 0, 1, 0, 0, 0, 1, "REP", TGSI_OPCODE_REP }, + { 0, 0, 0, 1, 1, 1, "ELSE", TGSI_OPCODE_ELSE }, + { 0, 0, 0, 0, 1, 0, "ENDIF", TGSI_OPCODE_ENDIF }, + { 1, 0, 0, 0, 1, 0, "ENDFOR", TGSI_OPCODE_ENDFOR }, + { 0, 0, 0, 0, 1, 0, "ENDREP", TGSI_OPCODE_ENDREP }, + { 0, 1, 0, 0, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA }, + { 1, 0, 0, 0, 0, 0, "POPA", TGSI_OPCODE_POPA }, + { 1, 1, 0, 0, 0, 0, "CEIL", TGSI_OPCODE_CEIL }, + { 1, 1, 0, 0, 0, 0, "I2F", TGSI_OPCODE_I2F }, + { 1, 1, 0, 0, 0, 0, "NOT", TGSI_OPCODE_NOT }, + { 1, 1, 0, 0, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, + { 1, 2, 0, 0, 0, 0, "SHL", TGSI_OPCODE_SHL }, + { 1, 2, 0, 0, 0, 0, "SHR", TGSI_OPCODE_SHR }, + { 1, 2, 0, 0, 0, 0, "AND", TGSI_OPCODE_AND }, + { 1, 2, 0, 0, 0, 0, "OR", TGSI_OPCODE_OR }, + { 1, 2, 0, 0, 0, 0, "MOD", TGSI_OPCODE_MOD }, + { 1, 2, 0, 0, 0, 0, "XOR", TGSI_OPCODE_XOR }, + { 1, 3, 0, 0, 0, 0, "SAD", TGSI_OPCODE_SAD }, + { 1, 2, 1, 0, 0, 0, "TXF", TGSI_OPCODE_TXF }, + { 1, 2, 1, 0, 0, 0, "TXQ", TGSI_OPCODE_TXQ }, + { 0, 0, 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT }, + { 0, 0, 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT }, + { 0, 0, 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM }, + { 0, 0, 0, 1, 0, 1, "BGNLOOP", TGSI_OPCODE_BGNLOOP }, + { 0, 0, 0, 0, 0, 1, "BGNSUB", TGSI_OPCODE_BGNSUB }, + { 0, 0, 0, 1, 1, 0, "ENDLOOP", TGSI_OPCODE_ENDLOOP }, + { 0, 0, 0, 0, 1, 0, "ENDSUB", TGSI_OPCODE_ENDSUB }, + { 1, 1, 0, 0, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 }, + { 1, 1, 0, 0, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 }, + { 1, 1, 0, 0, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 }, + { 1, 1, 0, 0, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 }, + { 0, 0, 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP }, + { 0, 0, 0, 0, 0, 0, "", 108 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 109 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 110 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 111 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "NRM4", TGSI_OPCODE_NRM4 }, + { 0, 1, 0, 0, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ }, + { 0, 1, 0, 0, 0, 0, "IFC", TGSI_OPCODE_IFC }, + { 0, 1, 0, 0, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, + { 0, 1, 0, 0, 0, 0, "KIL", TGSI_OPCODE_KIL }, + { 0, 0, 0, 0, 0, 0, "END", TGSI_OPCODE_END }, + { 1, 1, 0, 0, 0, 0, "SWZ", TGSI_OPCODE_SWZ } }; const struct tgsi_opcode_info * @@ -170,3 +170,12 @@ tgsi_get_opcode_info( uint opcode ) assert( 0 ); return NULL; } + + +const char * +tgsi_get_opcode_name( uint opcode ) +{ + const struct tgsi_opcode_info *info = tgsi_get_opcode_info(opcode); + return info->mnemonic; +} + diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h index 16577598bb..74713c3b98 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.h +++ b/src/gallium/auxiliary/tgsi/tgsi_info.h @@ -36,10 +36,12 @@ extern "C" { struct tgsi_opcode_info { - uint num_dst; - uint num_src; - boolean is_tex; - boolean is_branch; + unsigned num_dst:3; + unsigned num_src:3; + unsigned is_tex:1; + unsigned is_branch:1; + int pre_dedent:2; + int post_indent:2; const char *mnemonic; uint opcode; }; @@ -47,6 +49,10 @@ struct tgsi_opcode_info const struct tgsi_opcode_info * tgsi_get_opcode_info( uint opcode ); +const char * +tgsi_get_opcode_name( uint opcode ); + + #if defined __cplusplus } #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h new file mode 100644 index 0000000000..e7bcf4bf75 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h @@ -0,0 +1,172 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +#ifndef OP12_TEX +#define OP12_TEX(a) OP12(a) +#endif + +#ifndef OP14_TEX +#define OP14_TEX(a) OP14(a) +#endif + +#ifndef OP00_LBL +#define OP00_LBL(a) OP00(a) +#endif + +#ifndef OP01_LBL +#define OP01_LBL(a) OP01(a) +#endif + +OP11(ARL) +OP11(MOV) +OP11(LIT) +OP11(RCP) +OP11(RSQ) +OP11(EXP) +OP11(LOG) +OP12(MUL) +OP12(ADD) +OP12(DP3) +OP12(DP4) +OP12(DST) +OP12(MIN) +OP12(MAX) +OP12(SLT) +OP12(SGE) +OP13(MAD) +OP12(SUB) +OP13(LRP) +OP13(CND) +OP13(DP2A) +OP11(FRC) +OP13(CLAMP) +OP11(FLR) +OP11(ROUND) +OP11(EX2) +OP11(LG2) +OP12(POW) +OP12(XPD) +OP11(ABS) +OP11(RCC) +OP12(DPH) +OP11(COS) +OP11(DDX) +OP11(DDY) +OP00(KILP) +OP11(PK2H) +OP11(PK2US) +OP11(PK4B) +OP11(PK4UB) +OP12(RFL) +OP12(SEQ) +OP12(SFL) +OP12(SGT) +OP11(SIN) +OP12(SLE) +OP12(SNE) +OP12(STR) +OP12_TEX(TEX) +OP14_TEX(TXD) +OP12_TEX(TXP) +OP11(UP2H) +OP11(UP2US) +OP11(UP4B) +OP11(UP4UB) +OP13(X2D) +OP11(ARA) +OP11(ARR) +OP01(BRA) +OP00_LBL(CAL) +OP00(RET) +OP11(SSG) +OP13(CMP) +OP11(SCS) +OP12_TEX(TXB) +OP11(NRM) +OP12(DIV) +OP12(DP2) +OP12_TEX(TXL) +OP00(BRK) +OP01_LBL(IF) +OP11(BGNFOR) +OP01(REP) +OP00_LBL(ELSE) +OP00(ENDIF) +OP10(ENDFOR) +OP00(ENDREP) +OP01(PUSHA) +OP10(POPA) +OP11(CEIL) +OP11(I2F) +OP11(NOT) +OP11(TRUNC) +OP12(SHL) +OP12(SHR) +OP12(AND) +OP12(OR) +OP12(MOD) +OP12(XOR) +OP13(SAD) +OP12_TEX(TXF) +OP12_TEX(TXQ) +OP00(CONT) +OP00(EMIT) +OP00(ENDPRIM) +OP00_LBL(BGNLOOP) +OP00(BGNSUB) +OP00_LBL(ENDLOOP) +OP00(ENDSUB) +OP11(NOISE1) +OP11(NOISE2) +OP11(NOISE3) +OP11(NOISE4) +OP00(NOP) +OP11(NRM4) +OP01(CALLNZ) +OP01(IFC) +OP01(BREAKC) +OP01(KIL) +OP00(END) +OP11(SWZ) + + +#undef OP00 +#undef OP01 +#undef OP10 +#undef OP11 +#undef OP12 +#undef OP13 + +#ifdef OP14 +#undef OP14 +#endif + +#undef OP00_LBL +#undef OP01_LBL + +#undef OP12_TEX +#undef OP14_TEX + diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h index 1035bda1a8..a26ee5ba86 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h @@ -87,6 +87,7 @@ struct tgsi_full_instruction struct tgsi_instruction_ext_texture InstructionExtTexture; struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; + uint Flags; /**< user-defined usage */ }; union tgsi_full_token diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c index fddf54460a..4b1c7d4e01 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c @@ -38,6 +38,7 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_sse.h" +#include "tgsi/tgsi_info.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_dump.h" @@ -619,17 +620,17 @@ emit_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst) ppc_vandc(gen->f, v1, v0, bit31_vec); /* v1 = v0 & ~bit31 */ } break; - case TGSI_OPCODE_FLOOR: + case TGSI_OPCODE_FLR: ppc_vrfim(gen->f, v1, v0); /* v1 = floor(v0) */ break; - case TGSI_OPCODE_FRAC: + case TGSI_OPCODE_FRC: ppc_vrfim(gen->f, v1, v0); /* tmp = floor(v0) */ ppc_vsubfp(gen->f, v1, v0, v1); /* v1 = v0 - v1 */ break; - case TGSI_OPCODE_EXPBASE2: + case TGSI_OPCODE_EX2: ppc_vexptefp(gen->f, v1, v0); /* v1 = 2^v0 */ break; - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_LG2: /* XXX this may be broken! */ ppc_vlogefp(gen->f, v1, v0); /* v1 = log2(v0) */ break; @@ -1107,14 +1108,23 @@ static int emit_instruction(struct gen_context *gen, struct tgsi_full_instruction *inst) { + + /* we don't handle saturation/clamping yet */ + if (inst->Instruction.Saturate != TGSI_SAT_NONE) + return 0; + + /* need to use extra temps to fix SOA dependencies : */ + if (tgsi_check_soa_dependencies(inst)) + return FALSE; + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: case TGSI_OPCODE_ABS: - case TGSI_OPCODE_FLOOR: - case TGSI_OPCODE_FRAC: - case TGSI_OPCODE_EXPBASE2: - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_FLR: + case TGSI_OPCODE_FRC: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: emit_unaryop(gen, inst); break; case TGSI_OPCODE_RSQ: @@ -1317,8 +1327,10 @@ tgsi_emit_ppc(const struct tgsi_token *tokens, ok = emit_instruction(&gen, &parse.FullToken.FullInstruction); if (!ok) { - debug_printf("failed to translate tgsi opcode %d to PPC (%s)\n", - parse.FullToken.FullInstruction.Instruction.Opcode, + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + debug_printf("failed to translate tgsi opcode %d (%s) to PPC (%s)\n", + opcode, + tgsi_get_opcode_name(opcode), parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? "vertex shader" : "fragment shader"); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 6f1f5c2b4b..53e13b30e6 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -131,7 +131,7 @@ is_register_used( return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } -static const char *file_names[] = +static const char *file_names[TGSI_FILE_COUNT] = { "NULL", "CONST", @@ -140,7 +140,8 @@ static const char *file_names[] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static boolean @@ -198,10 +199,10 @@ iter_instruction( } if (info->num_dst != inst->Instruction.NumDstRegs) { - report_error( ctx, "Invalid number of destination operands, should be %u", info->num_dst ); + report_error( ctx, "%s: Invalid number of destination operands, should be %u", info->mnemonic, info->num_dst ); } if (info->num_src != inst->Instruction.NumSrcRegs) { - report_error( ctx, "Invalid number of source operands, should be %u", info->num_src ); + report_error( ctx, "%s: Invalid number of source operands, should be %u", info->mnemonic, info->num_src ); } /* Check destination and source registers' validity. @@ -234,9 +235,29 @@ iter_instruction( index, "indirect", FALSE ); - if (file != TGSI_FILE_ADDRESS || index != 0) - report_warning( ctx, "Indirect register not ADDR[0]" ); + if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) { + report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]"); + } + } + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_BGNFOR: + case TGSI_OPCODE_ENDFOR: + if (inst->FullDstRegisters[0].DstRegister.File != TGSI_FILE_LOOP || + inst->FullDstRegisters[0].DstRegister.Index != 0) { + report_error(ctx, "Destination register must be LOOP[0]"); + } + break; + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_BGNFOR: + if (inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_CONSTANT && + inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_IMMEDIATE) { + report_error(ctx, "Source register file must be either CONST or IMM"); } + break; } ctx->num_instructions++; @@ -337,7 +358,7 @@ epilog( boolean tgsi_sanity_check( - struct tgsi_token *tokens ) + const struct tgsi_token *tokens ) { struct sanity_check_ctx ctx; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/tgsi_sanity.h index ca45e94c7a..52263ff883 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.h +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.h @@ -40,7 +40,7 @@ extern "C" { */ boolean tgsi_sanity_check( - struct tgsi_token *tokens ); + const struct tgsi_token *tokens ); #if defined __cplusplus } diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c index c535788819..0db4481a3d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c @@ -132,6 +132,7 @@ tgsi_scan_shader(const struct tgsi_token *tokens, if (file == TGSI_FILE_INPUT) { info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.SemanticName; info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.SemanticIndex; + info->input_interpolate[reg] = (ubyte)fulldecl->Declaration.Interpolate; info->num_inputs++; } else if (file == TGSI_FILE_OUTPUT) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h index 2c1a75bc81..8a7ee0c7e4 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h @@ -45,6 +45,7 @@ struct tgsi_shader_info ubyte num_outputs; ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + ubyte input_interpolate[PIPE_MAX_SHADER_INPUTS]; ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index a2d4627da9..501fc05e72 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -36,10 +36,12 @@ #if defined(PIPE_ARCH_SSE) #include "util/u_sse.h" #endif +#include "tgsi/tgsi_info.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" -#include "tgsi_exec.h" -#include "tgsi_sse2.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_sse2.h" #include "rtasm/rtasm_x86sse.h" @@ -1359,6 +1361,32 @@ emit_store( const struct tgsi_full_instruction *inst, unsigned chan_index ) { + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + + case TGSI_SAT_ZERO_ONE: + sse_maxps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + + sse_minps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ) ); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + } + + switch( reg->DstRegister.File ) { case TGSI_FILE_OUTPUT: emit_output( @@ -1387,19 +1415,6 @@ emit_store( default: assert( 0 ); } - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - - case TGSI_SAT_ZERO_ONE: - /* assert( 0 ); */ - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - } } #define STORE( FUNC, INST, XMM, INDEX, CHAN )\ @@ -1467,15 +1482,15 @@ emit_tex( struct x86_function *func, switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: - case TGSI_TEXTURE_SHADOW1D: count = 1; break; case TGSI_TEXTURE_2D: case TGSI_TEXTURE_RECT: - case TGSI_TEXTURE_SHADOW2D: - case TGSI_TEXTURE_SHADOWRECT: count = 2; break; + case TGSI_TEXTURE_SHADOW1D: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: case TGSI_TEXTURE_3D: case TGSI_TEXTURE_CUBE: count = 3; @@ -1759,8 +1774,10 @@ emit_instruction( case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - STORE( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 4 + chan_index, 0, chan_index ); + } + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 4 + chan_index, 0, chan_index ); } break; @@ -2080,10 +2097,6 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_CND0: - return 0; - break; - case TGSI_OPCODE_DP2A: FETCH( func, *inst, 0, 0, CHAN_X ); /* xmm0 = src[0].x */ FETCH( func, *inst, 1, 1, CHAN_X ); /* xmm1 = src[1].x */ @@ -2099,14 +2112,6 @@ emit_instruction( } break; - case TGSI_OPCODE_INDEX: - return 0; - break; - - case TGSI_OPCODE_NEGATE: - return 0; - break; - case TGSI_OPCODE_FRC: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); @@ -2206,10 +2211,6 @@ emit_instruction( } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - return 0; - break; - case TGSI_OPCODE_ABS: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); @@ -2543,7 +2544,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: return 0; break; @@ -2559,7 +2560,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: return 0; break; @@ -2929,11 +2930,29 @@ tgsi_emit_sse2( &parse.FullToken.FullInstruction ); if (!ok) { - debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", - parse.FullToken.FullInstruction.Instruction.Opcode, + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + debug_printf("failed to translate tgsi opcode %d (%s) to SSE (%s)\n", + opcode, + tgsi_get_opcode_name(opcode), parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? "vertex shader" : "fragment shader"); } + + if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) { + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + + /* XXX: we only handle src/dst aliasing in a few opcodes + * currently. Need to use an additional temporay to hold + * the result in the cases where the code is too opaque to + * fix. + */ + if (opcode != TGSI_OPCODE_MOV && + opcode != TGSI_OPCODE_SWZ) { + debug_printf("Warning: src/dst aliasing in instruction" + " is not handled:\n"); + tgsi_dump_instruction(&parse.FullToken.FullInstruction, 1); + } + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index bfcbc40982..d438450b1e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -231,7 +231,8 @@ static const char *file_names[TGSI_FILE_COUNT] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static boolean diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c new file mode 100644 index 0000000000..654426a903 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -0,0 +1,1077 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "tgsi/tgsi_ureg.h" +#include "tgsi/tgsi_info.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_sanity.h" +#include "util/u_memory.h" +#include "util/u_math.h" + +union tgsi_any_token { + struct tgsi_version version; + struct tgsi_header header; + struct tgsi_processor processor; + struct tgsi_token token; + struct tgsi_declaration decl; + struct tgsi_declaration_range decl_range; + struct tgsi_declaration_semantic decl_semantic; + struct tgsi_immediate imm; + union tgsi_immediate_data imm_data; + struct tgsi_instruction insn; + struct tgsi_instruction_ext_nv insn_ext_nv; + struct tgsi_instruction_ext_label insn_ext_label; + struct tgsi_instruction_ext_texture insn_ext_texture; + struct tgsi_instruction_ext_predicate insn_ext_predicate; + struct tgsi_src_register src; + struct tgsi_src_register_ext_swz src_ext_swz; + struct tgsi_src_register_ext_mod src_ext_mod; + struct tgsi_dimension dim; + struct tgsi_dst_register dst; + struct tgsi_dst_register_ext_concode dst_ext_code; + struct tgsi_dst_register_ext_modulate dst_ext_mod; + struct tgsi_dst_register_ext_predicate dst_ext_pred; + unsigned value; +}; + + +struct ureg_tokens { + union tgsi_any_token *tokens; + unsigned size; + unsigned order; + unsigned count; +}; + +#define UREG_MAX_INPUT PIPE_MAX_ATTRIBS +#define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS +#define UREG_MAX_CONSTANT_RANGE 32 +#define UREG_MAX_IMMEDIATE 32 +#define UREG_MAX_TEMP 256 +#define UREG_MAX_ADDR 2 + +#define DOMAIN_DECL 0 +#define DOMAIN_INSN 1 + +struct ureg_program +{ + unsigned processor; + struct pipe_context *pipe; + + struct { + unsigned semantic_name; + unsigned semantic_index; + unsigned interp; + } fs_input[UREG_MAX_INPUT]; + unsigned nr_fs_inputs; + + unsigned vs_inputs[UREG_MAX_INPUT/32]; + + struct { + unsigned semantic_name; + unsigned semantic_index; + } output[UREG_MAX_OUTPUT]; + unsigned nr_outputs; + + struct { + float v[4]; + unsigned nr; + } immediate[UREG_MAX_IMMEDIATE]; + unsigned nr_immediates; + + struct ureg_src sampler[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + + unsigned temps_active[UREG_MAX_TEMP / 32]; + unsigned nr_temps; + + struct { + unsigned first; + unsigned last; + } constant_range[UREG_MAX_CONSTANT_RANGE]; + unsigned nr_constant_ranges; + + unsigned nr_addrs; + unsigned nr_instructions; + + struct ureg_tokens domain[2]; +}; + +static union tgsi_any_token error_tokens[32]; + +static void tokens_error( struct ureg_tokens *tokens ) +{ + if (tokens->tokens && tokens->tokens != error_tokens) + FREE(tokens->tokens); + + tokens->tokens = error_tokens; + tokens->size = Elements(error_tokens); + tokens->count = 0; +} + + +static void tokens_expand( struct ureg_tokens *tokens, + unsigned count ) +{ + unsigned old_size = tokens->size * sizeof(unsigned); + + if (tokens->tokens == error_tokens) + goto fail; + + while (tokens->count + count > tokens->size) { + tokens->size = (1 << ++tokens->order); + } + + tokens->tokens = REALLOC(tokens->tokens, + old_size, + tokens->size * sizeof(unsigned)); + if (tokens->tokens == NULL) + goto fail; + + return; + +fail: + tokens_error(tokens); +} + +static void set_bad( struct ureg_program *ureg ) +{ + tokens_error(&ureg->domain[0]); +} + + + +static union tgsi_any_token *get_tokens( struct ureg_program *ureg, + unsigned domain, + unsigned count ) +{ + struct ureg_tokens *tokens = &ureg->domain[domain]; + union tgsi_any_token *result; + + if (tokens->count + count > tokens->size) + tokens_expand(tokens, count); + + result = &tokens->tokens[tokens->count]; + tokens->count += count; + return result; +} + + +static union tgsi_any_token *retrieve_token( struct ureg_program *ureg, + unsigned domain, + unsigned nr ) +{ + if (ureg->domain[domain].tokens == error_tokens) + return &error_tokens[0]; + + return &ureg->domain[domain].tokens[nr]; +} + + + +static INLINE struct ureg_dst +ureg_dst_register( unsigned file, + unsigned index ) +{ + struct ureg_dst dst; + + dst.File = file; + dst.WriteMask = TGSI_WRITEMASK_XYZW; + dst.Indirect = 0; + dst.IndirectIndex = 0; + dst.IndirectSwizzle = 0; + dst.Saturate = 0; + dst.Index = index; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src_register( unsigned file, + unsigned index ) +{ + struct ureg_src src; + + src.File = file; + src.SwizzleX = TGSI_SWIZZLE_X; + src.SwizzleY = TGSI_SWIZZLE_Y; + src.SwizzleZ = TGSI_SWIZZLE_Z; + src.SwizzleW = TGSI_SWIZZLE_W; + src.Pad = 0; + src.Indirect = 0; + src.IndirectIndex = 0; + src.IndirectSwizzle = 0; + src.Absolute = 0; + src.Index = index; + src.Negate = 0; + + return src; +} + + + + +struct ureg_src +ureg_DECL_fs_input( struct ureg_program *ureg, + unsigned name, + unsigned index, + unsigned interp_mode ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_fs_inputs; i++) { + if (ureg->fs_input[i].semantic_name == name && + ureg->fs_input[i].semantic_index == index) + goto out; + } + + if (ureg->nr_fs_inputs < UREG_MAX_INPUT) { + ureg->fs_input[i].semantic_name = name; + ureg->fs_input[i].semantic_index = index; + ureg->fs_input[i].interp = interp_mode; + ureg->nr_fs_inputs++; + } + else { + set_bad( ureg ); + } + +out: + return ureg_src_register( TGSI_FILE_INPUT, i ); +} + + +struct ureg_src +ureg_DECL_vs_input( struct ureg_program *ureg, + unsigned index ) +{ + assert(ureg->processor == TGSI_PROCESSOR_VERTEX); + + ureg->vs_inputs[index/32] |= 1 << (index % 32); + return ureg_src_register( TGSI_FILE_INPUT, index ); +} + + +struct ureg_dst +ureg_DECL_output( struct ureg_program *ureg, + unsigned name, + unsigned index ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_outputs; i++) { + if (ureg->output[i].semantic_name == name && + ureg->output[i].semantic_index == index) + goto out; + } + + if (ureg->nr_outputs < UREG_MAX_OUTPUT) { + ureg->output[i].semantic_name = name; + ureg->output[i].semantic_index = index; + ureg->nr_outputs++; + } + else { + set_bad( ureg ); + } + +out: + return ureg_dst_register( TGSI_FILE_OUTPUT, i ); +} + + +/* Returns a new constant register. Keep track of which have been + * referred to so that we can emit decls later. + * + * There is nothing in this code to bind this constant to any tracked + * value or manage any constant_buffer contents -- that's the + * resposibility of the calling code. + */ +struct ureg_src ureg_DECL_constant(struct ureg_program *ureg, + unsigned index ) +{ + unsigned minconst = index, maxconst = index; + unsigned i; + + /* Inside existing range? + */ + for (i = 0; i < ureg->nr_constant_ranges; i++) { + if (ureg->constant_range[i].first <= index && + ureg->constant_range[i].last >= index) + goto out; + } + + /* Extend existing range? + */ + for (i = 0; i < ureg->nr_constant_ranges; i++) { + if (ureg->constant_range[i].last == index - 1) { + ureg->constant_range[i].last = index; + goto out; + } + + if (ureg->constant_range[i].first == index + 1) { + ureg->constant_range[i].first = index; + goto out; + } + + minconst = MIN2(minconst, ureg->constant_range[i].first); + maxconst = MAX2(maxconst, ureg->constant_range[i].last); + } + + /* Create new range? + */ + if (ureg->nr_constant_ranges < UREG_MAX_CONSTANT_RANGE) { + i = ureg->nr_constant_ranges++; + ureg->constant_range[i].first = index; + ureg->constant_range[i].last = index; + } + + /* Collapse all ranges down to one: + */ + i = 0; + ureg->constant_range[0].first = minconst; + ureg->constant_range[0].last = maxconst; + ureg->nr_constant_ranges = 1; + +out: + assert(i < ureg->nr_constant_ranges); + assert(ureg->constant_range[i].first <= index); + assert(ureg->constant_range[i].last >= index); + return ureg_src_register( TGSI_FILE_CONSTANT, index ); +} + + +/* Allocate a new temporary. Temporaries greater than UREG_MAX_TEMP + * are legal, but will not be released. + */ +struct ureg_dst ureg_DECL_temporary( struct ureg_program *ureg ) +{ + unsigned i; + + for (i = 0; i < UREG_MAX_TEMP; i += 32) { + int bit = ffs(~ureg->temps_active[i/32]); + if (bit != 0) { + i += bit - 1; + goto out; + } + } + + /* No reusable temps, so allocate a new one: + */ + i = ureg->nr_temps++; + +out: + if (i < UREG_MAX_TEMP) + ureg->temps_active[i/32] |= 1 << (i % 32); + + if (i >= ureg->nr_temps) + ureg->nr_temps = i + 1; + + return ureg_dst_register( TGSI_FILE_TEMPORARY, i ); +} + + +void ureg_release_temporary( struct ureg_program *ureg, + struct ureg_dst tmp ) +{ + if(tmp.File == TGSI_FILE_TEMPORARY) + if (tmp.Index < UREG_MAX_TEMP) + ureg->temps_active[tmp.Index/32] &= ~(1 << (tmp.Index % 32)); +} + + +/* Allocate a new address register. + */ +struct ureg_dst ureg_DECL_address( struct ureg_program *ureg ) +{ + if (ureg->nr_addrs < UREG_MAX_ADDR) + return ureg_dst_register( TGSI_FILE_ADDRESS, ureg->nr_addrs++ ); + + assert( 0 ); + return ureg_dst_register( TGSI_FILE_ADDRESS, 0 ); +} + +/* Allocate a new sampler. + */ +struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg, + unsigned nr ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_samplers; i++) + if (ureg->sampler[i].Index == nr) + return ureg->sampler[i]; + + if (i < PIPE_MAX_SAMPLERS) { + ureg->sampler[i] = ureg_src_register( TGSI_FILE_SAMPLER, nr ); + ureg->nr_samplers++; + return ureg->sampler[i]; + } + + assert( 0 ); + return ureg->sampler[0]; +} + + + + +static int match_or_expand_immediate( const float *v, + unsigned nr, + float *v2, + unsigned *nr2, + unsigned *swizzle ) +{ + unsigned i, j; + + *swizzle = 0; + + for (i = 0; i < nr; i++) { + boolean found = FALSE; + + for (j = 0; j < *nr2 && !found; j++) { + if (v[i] == v2[j]) { + *swizzle |= j << (i * 2); + found = TRUE; + } + } + + if (!found) { + if (*nr2 >= 4) + return FALSE; + + v2[*nr2] = v[i]; + *swizzle |= *nr2 << (i * 2); + (*nr2)++; + } + } + + return TRUE; +} + + + + +struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, + const float *v, + unsigned nr ) +{ + unsigned i, j; + unsigned swizzle; + + /* Could do a first pass where we examine all existing immediates + * without expanding. + */ + + for (i = 0; i < ureg->nr_immediates; i++) { + if (match_or_expand_immediate( v, + nr, + ureg->immediate[i].v, + &ureg->immediate[i].nr, + &swizzle )) + goto out; + } + + if (ureg->nr_immediates < UREG_MAX_IMMEDIATE) { + i = ureg->nr_immediates++; + if (match_or_expand_immediate( v, + nr, + ureg->immediate[i].v, + &ureg->immediate[i].nr, + &swizzle )) + goto out; + } + + set_bad( ureg ); + +out: + /* Make sure that all referenced elements are from this immediate. + * Has the effect of making size-one immediates into scalars. + */ + for (j = nr; j < 4; j++) + swizzle |= (swizzle & 0x3) << (j * 2); + + return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ), + (swizzle >> 0) & 0x3, + (swizzle >> 2) & 0x3, + (swizzle >> 4) & 0x3, + (swizzle >> 6) & 0x3); +} + + +void +ureg_emit_src( struct ureg_program *ureg, + struct ureg_src src ) +{ + unsigned size = (1 + + (src.Absolute ? 1 : 0) + + (src.Indirect ? 1 : 0)); + + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); + unsigned n = 0; + + assert(src.File != TGSI_FILE_NULL); + assert(src.File != TGSI_FILE_OUTPUT); + assert(src.File < TGSI_FILE_COUNT); + + out[n].value = 0; + out[n].src.File = src.File; + out[n].src.SwizzleX = src.SwizzleX; + out[n].src.SwizzleY = src.SwizzleY; + out[n].src.SwizzleZ = src.SwizzleZ; + out[n].src.SwizzleW = src.SwizzleW; + out[n].src.Index = src.Index; + out[n].src.Negate = src.Negate; + n++; + + if (src.Absolute) { + out[0].src.Extended = 1; + out[0].src.Negate = 0; + out[n].value = 0; + out[n].src_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + out[n].src_ext_mod.Absolute = 1; + out[n].src_ext_mod.Negate = src.Negate; + n++; + } + + if (src.Indirect) { + out[0].src.Indirect = 1; + out[n].value = 0; + out[n].src.File = TGSI_FILE_ADDRESS; + out[n].src.SwizzleX = src.IndirectSwizzle; + out[n].src.SwizzleY = src.IndirectSwizzle; + out[n].src.SwizzleZ = src.IndirectSwizzle; + out[n].src.SwizzleW = src.IndirectSwizzle; + out[n].src.Index = src.IndirectIndex; + n++; + } + + assert(n == size); +} + + +void +ureg_emit_dst( struct ureg_program *ureg, + struct ureg_dst dst ) +{ + unsigned size = (1 + + (dst.Indirect ? 1 : 0)); + + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); + unsigned n = 0; + + assert(dst.File != TGSI_FILE_NULL); + assert(dst.File != TGSI_FILE_CONSTANT); + assert(dst.File != TGSI_FILE_INPUT); + assert(dst.File != TGSI_FILE_SAMPLER); + assert(dst.File != TGSI_FILE_IMMEDIATE); + assert(dst.File < TGSI_FILE_COUNT); + + out[n].value = 0; + out[n].dst.File = dst.File; + out[n].dst.WriteMask = dst.WriteMask; + out[n].dst.Indirect = dst.Indirect; + out[n].dst.Index = dst.Index; + n++; + + if (dst.Indirect) { + out[n].value = 0; + out[n].src.File = TGSI_FILE_ADDRESS; + out[n].src.SwizzleX = dst.IndirectSwizzle; + out[n].src.SwizzleY = dst.IndirectSwizzle; + out[n].src.SwizzleZ = dst.IndirectSwizzle; + out[n].src.SwizzleW = dst.IndirectSwizzle; + out[n].src.Index = dst.IndirectIndex; + n++; + } + + assert(n == size); +} + + +static void validate( unsigned opcode, + unsigned nr_dst, + unsigned nr_src ) +{ +#ifdef DEBUG + const struct tgsi_opcode_info *info = tgsi_get_opcode_info( opcode ); + assert(info); + if(info) { + assert(nr_dst == info->num_dst); + assert(nr_src == info->num_src); + } +#endif +} + +unsigned +ureg_emit_insn(struct ureg_program *ureg, + unsigned opcode, + boolean saturate, + unsigned num_dst, + unsigned num_src ) +{ + union tgsi_any_token *out; + + validate( opcode, num_dst, num_src ); + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + out[0].value = 0; + out[0].insn.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + out[0].insn.NrTokens = 0; + out[0].insn.Opcode = opcode; + out[0].insn.Saturate = saturate; + out[0].insn.NumDstRegs = num_dst; + out[0].insn.NumSrcRegs = num_src; + out[0].insn.Padding = 0; + out[0].insn.Extended = 0; + + ureg->nr_instructions++; + + return ureg->domain[DOMAIN_INSN].count - 1; +} + + +void +ureg_emit_label(struct ureg_program *ureg, + unsigned insn_token, + unsigned *label_token ) +{ + union tgsi_any_token *out, *insn; + + if(!label_token) + return; + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + insn = retrieve_token( ureg, DOMAIN_INSN, insn_token ); + + insn->insn.Extended = 1; + + out[0].value = 0; + out[0].insn_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + + *label_token = ureg->domain[DOMAIN_INSN].count - 1; +} + +/* Will return a number which can be used in a label to point to the + * next instruction to be emitted. + */ +unsigned +ureg_get_instruction_number( struct ureg_program *ureg ) +{ + return ureg->nr_instructions; +} + +/* Patch a given label (expressed as a token number) to point to a + * given instruction (expressed as an instruction number). + */ +void +ureg_fixup_label(struct ureg_program *ureg, + unsigned label_token, + unsigned instruction_number ) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, label_token ); + + assert(out->insn_ext_label.Type == TGSI_INSTRUCTION_EXT_TYPE_LABEL); + out->insn_ext_label.Label = instruction_number; +} + + +void +ureg_emit_texture(struct ureg_program *ureg, + unsigned insn_token, + unsigned target ) +{ + union tgsi_any_token *out, *insn; + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + insn = retrieve_token( ureg, DOMAIN_INSN, insn_token ); + + insn->insn.Extended = 1; + + out[0].value = 0; + out[0].insn_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + out[0].insn_ext_texture.Texture = target; +} + + +void +ureg_fixup_insn_size(struct ureg_program *ureg, + unsigned insn ) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, insn ); + + assert(out->insn.Type == TGSI_TOKEN_TYPE_INSTRUCTION); + out->insn.NrTokens = ureg->domain[DOMAIN_INSN].count - insn - 1; +} + + +void +ureg_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + const struct ureg_src *src, + unsigned nr_src ) +{ + unsigned insn, i; + boolean saturate; + + saturate = nr_dst ? dst[0].Saturate : FALSE; + + insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src ); + + for (i = 0; i < nr_dst; i++) + ureg_emit_dst( ureg, dst[i] ); + + for (i = 0; i < nr_src; i++) + ureg_emit_src( ureg, src[i] ); + + ureg_fixup_insn_size( ureg, insn ); +} + +void +ureg_tex_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + unsigned target, + const struct ureg_src *src, + unsigned nr_src ) +{ + unsigned insn, i; + boolean saturate; + + saturate = nr_dst ? dst[0].Saturate : FALSE; + + insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src ); + + ureg_emit_texture( ureg, insn, target ); \ + + for (i = 0; i < nr_dst; i++) + ureg_emit_dst( ureg, dst[i] ); + + for (i = 0; i < nr_src; i++) + ureg_emit_src( ureg, src[i] ); + + ureg_fixup_insn_size( ureg, insn ); +} + + +void +ureg_label_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_src *src, + unsigned nr_src, + unsigned *label_token ) +{ + unsigned insn, i; + + insn = ureg_emit_insn( ureg, opcode, FALSE, 0, nr_src ); + + ureg_emit_label( ureg, insn, label_token ); \ + + for (i = 0; i < nr_src; i++) + ureg_emit_src( ureg, src[i] ); + + ureg_fixup_insn_size( ureg, insn ); +} + + + +static void emit_decl( struct ureg_program *ureg, + unsigned file, + unsigned index, + unsigned semantic_name, + unsigned semantic_index, + unsigned interp ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 ); + + out[0].value = 0; + out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION; + out[0].decl.NrTokens = 3; + out[0].decl.File = file; + out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW; /* FIXME! */ + out[0].decl.Interpolate = interp; + out[0].decl.Semantic = 1; + + out[1].value = 0; + out[1].decl_range.First = + out[1].decl_range.Last = index; + + out[2].value = 0; + out[2].decl_semantic.SemanticName = semantic_name; + out[2].decl_semantic.SemanticIndex = semantic_index; + +} + + +static void emit_decl_range( struct ureg_program *ureg, + unsigned file, + unsigned first, + unsigned count ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 2 ); + + out[0].value = 0; + out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION; + out[0].decl.NrTokens = 2; + out[0].decl.File = file; + out[0].decl.UsageMask = 0xf; + out[0].decl.Interpolate = TGSI_INTERPOLATE_CONSTANT; + out[0].decl.Semantic = 0; + + out[1].value = 0; + out[1].decl_range.First = first; + out[1].decl_range.Last = first + count - 1; +} + +static void emit_immediate( struct ureg_program *ureg, + const float *v ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 5 ); + + out[0].value = 0; + out[0].imm.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + out[0].imm.NrTokens = 5; + out[0].imm.DataType = TGSI_IMM_FLOAT32; + out[0].imm.Padding = 0; + out[0].imm.Extended = 0; + + out[1].imm_data.Float = v[0]; + out[2].imm_data.Float = v[1]; + out[3].imm_data.Float = v[2]; + out[4].imm_data.Float = v[3]; +} + + + + +static void emit_decls( struct ureg_program *ureg ) +{ + unsigned i; + + if (ureg->processor == TGSI_PROCESSOR_VERTEX) { + for (i = 0; i < UREG_MAX_INPUT; i++) { + if (ureg->vs_inputs[i/32] & (1 << (i%32))) { + emit_decl_range( ureg, TGSI_FILE_INPUT, i, 1 ); + } + } + } + else { + for (i = 0; i < ureg->nr_fs_inputs; i++) { + emit_decl( ureg, + TGSI_FILE_INPUT, + i, + ureg->fs_input[i].semantic_name, + ureg->fs_input[i].semantic_index, + ureg->fs_input[i].interp ); + } + } + + for (i = 0; i < ureg->nr_outputs; i++) { + emit_decl( ureg, + TGSI_FILE_OUTPUT, + i, + ureg->output[i].semantic_name, + ureg->output[i].semantic_index, + TGSI_INTERPOLATE_CONSTANT ); + } + + for (i = 0; i < ureg->nr_samplers; i++) { + emit_decl_range( ureg, + TGSI_FILE_SAMPLER, + ureg->sampler[i].Index, 1 ); + } + + if (ureg->nr_constant_ranges) { + for (i = 0; i < ureg->nr_constant_ranges; i++) + emit_decl_range( ureg, + TGSI_FILE_CONSTANT, + ureg->constant_range[i].first, + (ureg->constant_range[i].last + 1 - + ureg->constant_range[i].first) ); + } + + if (ureg->nr_temps) { + emit_decl_range( ureg, + TGSI_FILE_TEMPORARY, + 0, ureg->nr_temps ); + } + + if (ureg->nr_addrs) { + emit_decl_range( ureg, + TGSI_FILE_ADDRESS, + 0, ureg->nr_addrs ); + } + + for (i = 0; i < ureg->nr_immediates; i++) { + emit_immediate( ureg, + ureg->immediate[i].v ); + } +} + +/* Append the instruction tokens onto the declarations to build a + * contiguous stream suitable to send to the driver. + */ +static void copy_instructions( struct ureg_program *ureg ) +{ + unsigned nr_tokens = ureg->domain[DOMAIN_INSN].count; + union tgsi_any_token *out = get_tokens( ureg, + DOMAIN_DECL, + nr_tokens ); + + memcpy(out, + ureg->domain[DOMAIN_INSN].tokens, + nr_tokens * sizeof out[0] ); +} + + +static void +fixup_header_size(struct ureg_program *ureg) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 1 ); + + out->header.BodySize = ureg->domain[DOMAIN_DECL].count - 3; +} + + +static void +emit_header( struct ureg_program *ureg ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 ); + + out[0].version.MajorVersion = 1; + out[0].version.MinorVersion = 1; + out[0].version.Padding = 0; + + out[1].header.HeaderSize = 2; + out[1].header.BodySize = 0; + + out[2].processor.Processor = ureg->processor; + out[2].processor.Padding = 0; +} + + +const struct tgsi_token *ureg_finalize( struct ureg_program *ureg ) +{ + const struct tgsi_token *tokens; + + emit_header( ureg ); + emit_decls( ureg ); + copy_instructions( ureg ); + fixup_header_size( ureg ); + + if (ureg->domain[0].tokens == error_tokens || + ureg->domain[1].tokens == error_tokens) { + debug_printf("%s: error in generated shader\n", __FUNCTION__); + assert(0); + return NULL; + } + + tokens = &ureg->domain[DOMAIN_DECL].tokens[0].token; + + if (0) { + debug_printf("%s: emitted shader %d tokens:\n", __FUNCTION__, + ureg->domain[DOMAIN_DECL].count); + tgsi_dump( tokens, 0 ); + } + +#if DEBUG + if (tokens && !tgsi_sanity_check(tokens)) { + debug_printf("tgsi_ureg.c, sanity check failed on generated tokens:\n"); + tgsi_dump(tokens, 0); + assert(0); + } +#endif + + + return tokens; +} + + +void *ureg_create_shader( struct ureg_program *ureg, + struct pipe_context *pipe ) +{ + struct pipe_shader_state state; + + state.tokens = ureg_finalize(ureg); + if(!state.tokens) + return NULL; + + if (ureg->processor == TGSI_PROCESSOR_VERTEX) + return pipe->create_vs_state( pipe, &state ); + else + return pipe->create_fs_state( pipe, &state ); +} + + +const struct tgsi_token *ureg_get_tokens( struct ureg_program *ureg, + unsigned *nr_tokens ) +{ + const struct tgsi_token *tokens; + + ureg_finalize(ureg); + + tokens = &ureg->domain[DOMAIN_DECL].tokens[0].token; + + if (nr_tokens) + *nr_tokens = ureg->domain[DOMAIN_DECL].size; + + ureg->domain[DOMAIN_DECL].tokens = 0; + ureg->domain[DOMAIN_DECL].size = 0; + ureg->domain[DOMAIN_DECL].order = 0; + ureg->domain[DOMAIN_DECL].count = 0; + + return tokens; +} + + +struct ureg_program *ureg_create( unsigned processor ) +{ + struct ureg_program *ureg = CALLOC_STRUCT( ureg_program ); + if (ureg == NULL) + return NULL; + + ureg->processor = processor; + return ureg; +} + + +void ureg_destroy( struct ureg_program *ureg ) +{ + unsigned i; + + for (i = 0; i < Elements(ureg->domain); i++) { + if (ureg->domain[i].tokens && + ureg->domain[i].tokens != error_tokens) + FREE(ureg->domain[i].tokens); + } + + FREE(ureg); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h new file mode 100644 index 0000000000..f04f443b9e --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -0,0 +1,619 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_UREG_H +#define TGSI_UREG_H + +#include "pipe/p_compiler.h" +#include "pipe/p_shader_tokens.h" + +#ifdef __cplusplus +extern "C" { +#endif + +struct ureg_program; + +/* Almost a tgsi_src_register, but we need to pull in the Absolute + * flag from the _ext token. Indirect flag always implies ADDR[0]. + */ +struct ureg_src +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Pad : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Negate : 1; /* BOOL */ + int IndirectIndex : 16; /* SINT */ + int IndirectSwizzle : 2; /* TGSI_SWIZZLE_ */ +}; + +/* Very similar to a tgsi_dst_register, removing unsupported fields + * and adding a Saturate flag. It's easier to push saturate into the + * destination register than to try and create a _SAT varient of each + * instruction function. + */ +struct ureg_dst +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Saturate : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Pad1 : 5; + unsigned Pad2 : 1; /* BOOL */ + int IndirectIndex : 16; /* SINT */ + int IndirectSwizzle : 2; /* TGSI_SWIZZLE_ */ +}; + +struct pipe_context; + +struct ureg_program * +ureg_create( unsigned processor ); + +const struct tgsi_token * +ureg_finalize( struct ureg_program * ); + +/* Create and return a shader: + */ +void * +ureg_create_shader( struct ureg_program *, + struct pipe_context *pipe ); + + +/* Alternately, return the built token stream and hand ownership of + * that memory to the caller: + */ +const struct tgsi_token * +ureg_get_tokens( struct ureg_program *ureg, + unsigned *nr_tokens ); + + +void +ureg_destroy( struct ureg_program * ); + + +/*********************************************************************** + * Convenience routine: + */ +static INLINE void * +ureg_create_shader_and_destroy( struct ureg_program *p, + struct pipe_context *pipe ) +{ + void *result = ureg_create_shader( p, pipe ); + ureg_destroy( p ); + return result; +} + + + +/*********************************************************************** + * Build shader declarations: + */ + +struct ureg_src +ureg_DECL_fs_input( struct ureg_program *, + unsigned semantic_name, + unsigned semantic_index, + unsigned interp_mode ); + +struct ureg_src +ureg_DECL_vs_input( struct ureg_program *, + unsigned index ); + +struct ureg_dst +ureg_DECL_output( struct ureg_program *, + unsigned semantic_name, + unsigned semantic_index ); + +struct ureg_src +ureg_DECL_immediate( struct ureg_program *, + const float *v, + unsigned nr ); + +struct ureg_src +ureg_DECL_constant( struct ureg_program *, + unsigned index ); + +struct ureg_dst +ureg_DECL_temporary( struct ureg_program * ); + +void +ureg_release_temporary( struct ureg_program *ureg, + struct ureg_dst tmp ); + +struct ureg_dst +ureg_DECL_address( struct ureg_program * ); + +/* Supply an index to the sampler declaration as this is the hook to + * the external pipe_sampler state. Users of this function probably + * don't want just any sampler, but a specific one which they've set + * up state for in the context. + */ +struct ureg_src +ureg_DECL_sampler( struct ureg_program *, + unsigned index ); + + +static INLINE struct ureg_src +ureg_imm4f( struct ureg_program *ureg, + float a, float b, + float c, float d) +{ + float v[4]; + v[0] = a; + v[1] = b; + v[2] = c; + v[3] = d; + return ureg_DECL_immediate( ureg, v, 4 ); +} + +static INLINE struct ureg_src +ureg_imm3f( struct ureg_program *ureg, + float a, float b, + float c) +{ + float v[3]; + v[0] = a; + v[1] = b; + v[2] = c; + return ureg_DECL_immediate( ureg, v, 3 ); +} + +static INLINE struct ureg_src +ureg_imm2f( struct ureg_program *ureg, + float a, float b) +{ + float v[2]; + v[0] = a; + v[1] = b; + return ureg_DECL_immediate( ureg, v, 2 ); +} + +static INLINE struct ureg_src +ureg_imm1f( struct ureg_program *ureg, + float a) +{ + float v[1]; + v[0] = a; + return ureg_DECL_immediate( ureg, v, 1 ); +} + +/*********************************************************************** + * Functions for patching up labels + */ + + +/* Will return a number which can be used in a label to point to the + * next instruction to be emitted. + */ +unsigned +ureg_get_instruction_number( struct ureg_program *ureg ); + + +/* Patch a given label (expressed as a token number) to point to a + * given instruction (expressed as an instruction number). + * + * Labels are obtained from instruction emitters, eg ureg_CAL(). + * Instruction numbers are obtained from ureg_get_instruction_number(), + * above. + */ +void +ureg_fixup_label(struct ureg_program *ureg, + unsigned label_token, + unsigned instruction_number ); + + +/* Generic instruction emitter. Use if you need to pass the opcode as + * a parameter, rather than using the emit_OP() varients below. + */ +void +ureg_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + const struct ureg_src *src, + unsigned nr_src ); + + +void +ureg_tex_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + unsigned target, + const struct ureg_src *src, + unsigned nr_src ); + + +void +ureg_label_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_src *src, + unsigned nr_src, + unsigned *label); + + +/*********************************************************************** + * Internal instruction helpers, don't call these directly: + */ + +unsigned +ureg_emit_insn(struct ureg_program *ureg, + unsigned opcode, + boolean saturate, + unsigned num_dst, + unsigned num_src ); + +void +ureg_emit_label(struct ureg_program *ureg, + unsigned insn_token, + unsigned *label_token ); + +void +ureg_emit_texture(struct ureg_program *ureg, + unsigned insn_token, + unsigned target ); + +void +ureg_emit_dst( struct ureg_program *ureg, + struct ureg_dst dst ); + +void +ureg_emit_src( struct ureg_program *ureg, + struct ureg_src src ); + +void +ureg_fixup_insn_size(struct ureg_program *ureg, + unsigned insn ); + + +#define OP00( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP01( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_src src ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP00_LBL( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + unsigned *label_token ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \ + ureg_emit_label( ureg, insn, label_token ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP01_LBL( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_src src, \ + unsigned *label_token ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \ + ureg_emit_label( ureg, insn, label_token ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP10( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 0 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + + +#define OP11( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 1 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP12( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src0, \ + struct ureg_src src1 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP12_TEX( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + unsigned target, \ + struct ureg_src src0, \ + struct ureg_src src1 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 ); \ + ureg_emit_texture( ureg, insn, target ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP13( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src0, \ + struct ureg_src src1, \ + struct ureg_src src2 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 3 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_emit_src( ureg, src2 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP14_TEX( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + unsigned target, \ + struct ureg_src src0, \ + struct ureg_src src1, \ + struct ureg_src src2, \ + struct ureg_src src3 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 4 ); \ + ureg_emit_texture( ureg, insn, target ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_emit_src( ureg, src2 ); \ + ureg_emit_src( ureg, src3 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + + +/* Use a template include to generate a correctly-typed ureg_OP() + * function for each TGSI opcode: + */ +#include "tgsi_opcode_tmp.h" + + +/*********************************************************************** + * Inline helpers for manipulating register structs: + */ +static INLINE struct ureg_src +ureg_negate( struct ureg_src reg ) +{ + assert(reg.File != TGSI_FILE_NULL); + reg.Negate ^= 1; + return reg; +} + +static INLINE struct ureg_src +ureg_abs( struct ureg_src reg ) +{ + assert(reg.File != TGSI_FILE_NULL); + reg.Absolute = 1; + reg.Negate = 0; + return reg; +} + +static INLINE struct ureg_src +ureg_swizzle( struct ureg_src reg, + int x, int y, int z, int w ) +{ + unsigned swz = ( (reg.SwizzleX << 0) | + (reg.SwizzleY << 2) | + (reg.SwizzleZ << 4) | + (reg.SwizzleW << 6)); + + assert(reg.File != TGSI_FILE_NULL); + assert(x < 4); + assert(y < 4); + assert(z < 4); + assert(w < 4); + + reg.SwizzleX = (swz >> (x*2)) & 0x3; + reg.SwizzleY = (swz >> (y*2)) & 0x3; + reg.SwizzleZ = (swz >> (z*2)) & 0x3; + reg.SwizzleW = (swz >> (w*2)) & 0x3; + return reg; +} + +static INLINE struct ureg_src +ureg_scalar( struct ureg_src reg, int x ) +{ + return ureg_swizzle(reg, x, x, x, x); +} + +static INLINE struct ureg_dst +ureg_writemask( struct ureg_dst reg, + unsigned writemask ) +{ + assert(reg.File != TGSI_FILE_NULL); + reg.WriteMask &= writemask; + return reg; +} + +static INLINE struct ureg_dst +ureg_saturate( struct ureg_dst reg ) +{ + assert(reg.File != TGSI_FILE_NULL); + reg.Saturate = 1; + return reg; +} + +static INLINE struct ureg_dst +ureg_dst_indirect( struct ureg_dst reg, struct ureg_src addr ) +{ + assert(reg.File != TGSI_FILE_NULL); + assert(addr.File == TGSI_FILE_ADDRESS); + reg.Indirect = 1; + reg.IndirectIndex = addr.Index; + reg.IndirectSwizzle = addr.SwizzleX; + return reg; +} + +static INLINE struct ureg_src +ureg_src_indirect( struct ureg_src reg, struct ureg_src addr ) +{ + assert(reg.File != TGSI_FILE_NULL); + assert(addr.File == TGSI_FILE_ADDRESS); + reg.Indirect = 1; + reg.IndirectIndex = addr.Index; + reg.IndirectSwizzle = addr.SwizzleX; + return reg; +} + +static INLINE struct ureg_dst +ureg_dst( struct ureg_src src ) +{ + struct ureg_dst dst; + + dst.File = src.File; + dst.WriteMask = TGSI_WRITEMASK_XYZW; + dst.Indirect = src.Indirect; + dst.IndirectIndex = src.IndirectIndex; + dst.IndirectSwizzle = src.IndirectSwizzle; + dst.Saturate = 0; + dst.Index = src.Index; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src( struct ureg_dst dst ) +{ + struct ureg_src src; + + src.File = dst.File; + src.SwizzleX = TGSI_SWIZZLE_X; + src.SwizzleY = TGSI_SWIZZLE_Y; + src.SwizzleZ = TGSI_SWIZZLE_Z; + src.SwizzleW = TGSI_SWIZZLE_W; + src.Pad = 0; + src.Indirect = dst.Indirect; + src.IndirectIndex = dst.IndirectIndex; + src.IndirectSwizzle = dst.IndirectSwizzle; + src.Absolute = 0; + src.Index = dst.Index; + src.Negate = 0; + + return src; +} + + + +static INLINE struct ureg_dst +ureg_dst_undef( void ) +{ + struct ureg_dst dst; + + dst.File = TGSI_FILE_NULL; + dst.WriteMask = 0; + dst.Indirect = 0; + dst.IndirectIndex = 0; + dst.IndirectSwizzle = 0; + dst.Saturate = 0; + dst.Index = 0; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src_undef( void ) +{ + struct ureg_src src; + + src.File = TGSI_FILE_NULL; + src.SwizzleX = 0; + src.SwizzleY = 0; + src.SwizzleZ = 0; + src.SwizzleW = 0; + src.Pad = 0; + src.Indirect = 0; + src.IndirectIndex = 0; + src.IndirectSwizzle = 0; + src.Absolute = 0; + src.Index = 0; + src.Negate = 0; + + return src; +} + +static INLINE boolean +ureg_src_is_undef( struct ureg_src src ) +{ + return src.File == TGSI_FILE_NULL; +} + +static INLINE boolean +ureg_dst_is_undef( struct ureg_dst dst ) +{ + return dst.File == TGSI_FILE_NULL; +} + + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h index 7877f34558..21eb656327 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_util.h +++ b/src/gallium/auxiliary/tgsi/tgsi_util.h @@ -32,6 +32,10 @@ extern "C" { #endif +struct tgsi_src_register; +struct tgsi_src_register_ext_swz; +struct tgsi_full_src_register; + void * tgsi_align_128bit( void *unaligned ); diff --git a/src/gallium/auxiliary/util/.gitignore b/src/gallium/auxiliary/util/.gitignore new file mode 100644 index 0000000000..29c586c9b5 --- /dev/null +++ b/src/gallium/auxiliary/util/.gitignore @@ -0,0 +1,2 @@ +u_format_access.c +u_format_table.c diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 6a8eb73e84..ae8d330a78 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -5,11 +5,15 @@ LIBNAME = util C_SOURCES = \ u_debug.c \ + u_debug_dump.c \ u_debug_symbol.c \ u_debug_stack.c \ u_blit.c \ u_cache.c \ u_draw_quad.c \ + u_format.c \ + u_format_access.c \ + u_format_table.c \ u_gen_mipmap.c \ u_handle_table.c \ u_hash_table.c \ @@ -32,3 +36,9 @@ C_SOURCES = \ u_simple_screen.c include ../../Makefile.template + +u_format_table.c: u_format_table.py u_format_parse.py u_format.csv + python u_format_table.py u_format.csv > $@ + +u_format_access.c: u_format_access.py u_format_parse.py u_format.csv + python u_format_access.py u_format.csv > $@ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index fb142eebca..28a5ab4256 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -1,5 +1,23 @@ Import('*') +env.Clone() + +env.Append(CPPPATH = ['.']) + +env.CodeGenerate( + target = 'u_format_table.c', + script = 'u_format_table.py', + source = ['u_format.csv'], + command = 'python $SCRIPT $SOURCE > $TARGET' +) + +env.CodeGenerate( + target = 'u_format_access.c', + script = 'u_format_access.py', + source = ['u_format.csv'], + command = 'python $SCRIPT $SOURCE > $TARGET' +) + util = env.ConvenienceLibrary( target = 'util', source = [ @@ -7,11 +25,15 @@ util = env.ConvenienceLibrary( 'u_blit.c', 'u_cache.c', 'u_debug.c', + 'u_debug_dump.c', 'u_debug_memory.c', 'u_debug_profile.c', 'u_debug_stack.c', 'u_debug_symbol.c', 'u_draw_quad.c', + 'u_format.c', + 'u_format_access.c', + 'u_format_table.c', 'u_gen_mipmap.c', 'u_handle_table.c', 'u_hash.c', diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 414cf91025..c516317d70 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -45,6 +45,7 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_simple_shaders.h" +#include "util/u_surface.h" #include "cso_cache/cso_context.h" @@ -61,7 +62,7 @@ struct blit_state struct pipe_viewport_state viewport; void *vs; - void *fs; + void *fs[TGSI_WRITEMASK_XYZW + 1]; struct pipe_buffer *vbuf; /**< quad vertices */ unsigned vbuf_slot; @@ -124,7 +125,7 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs[TGSI_WRITEMASK_XYZW] = util_make_fragment_tex_shader(pipe); ctx->vbuf = NULL; /* init vertex data that doesn't change */ @@ -145,9 +146,13 @@ void util_destroy_blit(struct blit_state *ctx) { struct pipe_context *pipe = ctx->pipe; + unsigned i; pipe->delete_vs_state(pipe, ctx->vs); - pipe->delete_fs_state(pipe, ctx->fs); + + for (i = 0; i < Elements(ctx->fs); i++) + if (ctx->fs[i]) + pipe->delete_fs_state(pipe, ctx->fs[i]); pipe_buffer_reference(&ctx->vbuf, NULL); @@ -155,7 +160,11 @@ util_destroy_blit(struct blit_state *ctx) } -static unsigned get_next_slot( struct blit_state *ctx ) +/** + * Get offset of next free slot in vertex buffer for quad vertices. + */ +static unsigned +get_next_slot( struct blit_state *ctx ) { const unsigned max_slots = 4096 / sizeof ctx->vertices; @@ -173,7 +182,6 @@ static unsigned get_next_slot( struct blit_state *ctx ) } - /** * Setup vertex data for the textured quad we'll draw. * Note: y=0=top @@ -260,20 +268,50 @@ setup_vertex_data_tex(struct blit_state *ctx, return offset; } + + +/** + * \return TRUE if two regions overlap, FALSE otherwise + */ +static boolean +regions_overlap(int srcX0, int srcY0, + int srcX1, int srcY1, + int dstX0, int dstY0, + int dstX1, int dstY1) +{ + if (MAX2(srcX0, srcX1) < MIN2(dstX0, dstX1)) + return FALSE; /* src completely left of dst */ + + if (MAX2(dstX0, dstX1) < MIN2(srcX0, srcX1)) + return FALSE; /* dst completely left of src */ + + if (MAX2(srcY0, srcY1) < MIN2(dstY0, dstY1)) + return FALSE; /* src completely above dst */ + + if (MAX2(dstY0, dstY1) < MIN2(srcY0, srcY1)) + return FALSE; /* dst completely above src */ + + return TRUE; /* some overlap */ +} + + /** * Copy pixel block from src surface to dst surface. * Overlapping regions are acceptable. + * Flipping and stretching are supported. + * XXX what about clipping??? * XXX need some control over blitting Z and/or stencil. */ void -util_blit_pixels(struct blit_state *ctx, - struct pipe_surface *src, - int srcX0, int srcY0, - int srcX1, int srcY1, - struct pipe_surface *dst, - int dstX0, int dstY0, - int dstX1, int dstY1, - float z, uint filter) +util_blit_pixels_writemask(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter, + uint writemask) { struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; @@ -285,10 +323,41 @@ util_blit_pixels(struct blit_state *ctx, const int srcLeft = MIN2(srcX0, srcX1); const int srcTop = MIN2(srcY0, srcY1); unsigned offset; + boolean overlap; assert(filter == PIPE_TEX_MIPFILTER_NEAREST || filter == PIPE_TEX_MIPFILTER_LINEAR); + assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); + + /* do the regions overlap? */ + overlap = util_same_surface(src, dst) && + regions_overlap(srcX0, srcY0, srcX1, srcY1, + dstX0, dstY0, dstX1, dstY1); + + /* + * Check for simple case: no format conversion, no flipping, no stretching, + * no overlapping. + * Filter mode should not matter since there's no stretching. + */ + if (dst->format == src->format && + srcX0 < srcX1 && + dstX0 < dstX1 && + srcY0 < srcY1 && + dstY0 < dstY1 && + (dstX1 - dstX0) == (srcX1 - srcX0) && + (dstY1 - dstY0) == (srcY1 - srcY0) && + !overlap) { + pipe->surface_copy(pipe, + dst, dstX0, dstY0, /* dest */ + src, srcX0, srcY0, /* src */ + srcW, srcH); /* size */ + return; + } + if (srcLeft != srcX0) { /* left-right flip */ int tmp = dstX0; @@ -303,20 +372,6 @@ util_blit_pixels(struct blit_state *ctx, dstY1 = tmp; } - assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_SAMPLER, 0)); - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_SAMPLER, 0)); - - if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { - /* FIXME: this will most surely fail for overlapping rectangles */ - pipe->surface_copy(pipe, - dst, dstX0, dstY0, /* dest */ - src, srcX0, srcY0, /* src */ - srcW, srcH); /* size */ - return; - } - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); @@ -376,8 +431,11 @@ util_blit_pixels(struct blit_state *ctx, /* texture */ cso_set_sampler_textures(ctx->cso, 1, &tex); + if (ctx->fs[writemask] == NULL) + ctx->fs[writemask] = util_make_fragment_tex_shader_writemask(pipe, writemask); + /* shaders */ - cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs[writemask]); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* drawing dest */ @@ -412,6 +470,27 @@ util_blit_pixels(struct blit_state *ctx, } +void +util_blit_pixels(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter ) +{ + util_blit_pixels_writemask( ctx, src, + srcX0, srcY0, + srcX1, srcY1, + dst, + dstX0, dstY0, + dstX1, dstY1, + z, filter, + TGSI_WRITEMASK_XYZW ); +} + + /* Release vertex buffer at end of frame to avoid synchronous * rendering. */ @@ -485,7 +564,7 @@ util_blit_pixels_tex(struct blit_state *ctx, cso_set_sampler_textures(ctx->cso, 1, &tex); /* shaders */ - cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs[TGSI_WRITEMASK_XYZW]); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* drawing dest */ diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index c35beceda8..a102021529 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -60,6 +60,17 @@ util_blit_pixels(struct blit_state *ctx, int dstX1, int dstY1, float z, uint filter); +void +util_blit_pixels_writemask(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter, + uint writemask); + extern void util_blit_pixels_tex(struct blit_state *ctx, struct pipe_texture *tex, diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h index d42b65ce28..b82e7cb4d4 100644 --- a/src/gallium/auxiliary/util/u_debug.h +++ b/src/gallium/auxiliary/util/u_debug.h @@ -65,6 +65,11 @@ extern "C" { #define __FUNCTION__ "???" #endif +#if defined(__GNUC__) +#define _util_printf_format(fmt, list) __attribute__ ((format (printf, fmt, list))) +#else +#define _util_printf_format(fmt, list) +#endif void _debug_vprintf(const char *format, va_list ap); @@ -82,12 +87,16 @@ _debug_printf(const char *format, ...) /** * Print debug messages. * - * The actual channel used to output debug message is platform specific. To - * avoid misformating or truncation, follow these rules of thumb: + * The actual channel used to output debug message is platform specific. To + * avoid misformating or truncation, follow these rules of thumb: * - output whole lines - * - avoid outputing large strings (512 bytes is the current maximum length + * - avoid outputing large strings (512 bytes is the current maximum length * that is guaranteed to be printed in all platforms) */ +#if !defined(PIPE_OS_HAIKU) +static INLINE void +debug_printf(const char *format, ...) _util_printf_format(1,2); + static INLINE void debug_printf(const char *format, ...) { @@ -101,6 +110,7 @@ debug_printf(const char *format, ...) #endif } +#endif /* !PIPE_OS_HAIKU */ /* * ... isn't portable so we need to pass arguments in parentheses. diff --git a/src/gallium/auxiliary/util/u_debug_dump.c b/src/gallium/auxiliary/util/u_debug_dump.c new file mode 100644 index 0000000000..6bdecde048 --- /dev/null +++ b/src/gallium/auxiliary/util/u_debug_dump.c @@ -0,0 +1,189 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "util/u_memory.h" +#include "util/u_debug.h" +#include "util/u_debug_dump.h" + + +#define DEBUG_DUMP_INVALID_NAME "<invalid>" + + +#if 0 +static const char * +debug_dump_strip_prefix(const char *name, + const char *prefix) +{ + const char *stripped; + assert(name); + assert(prefix); + stripped = name; + while(*prefix) { + if(*stripped != *prefix) + return name; + + ++stripped; + ++prefix; + } + return stripped; +} +#endif + +static const char * +debug_dump_enum_continuous(unsigned value, + unsigned num_names, + const char **names) +{ + if (value >= num_names) + return DEBUG_DUMP_INVALID_NAME; + return names[value]; +} + + +#define DEFINE_DEBUG_DUMP_CONTINUOUS(_name) \ + const char * \ + debug_dump_##_name(unsigned value, boolean shortened) \ + { \ + if(shortened) \ + return debug_dump_enum_continuous(value, Elements(debug_dump_##_name##_short_names), debug_dump_##_name##_short_names); \ + else \ + return debug_dump_enum_continuous(value, Elements(debug_dump_##_name##_names), debug_dump_##_name##_names); \ + } + + +static const char * +debug_dump_blend_factor_names[] = { + DEBUG_DUMP_INVALID_NAME, /* 0x0 */ + "PIPE_BLENDFACTOR_ONE", + "PIPE_BLENDFACTOR_SRC_COLOR", + "PIPE_BLENDFACTOR_SRC_ALPHA", + "PIPE_BLENDFACTOR_DST_ALPHA", + "PIPE_BLENDFACTOR_DST_COLOR", + "PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE", + "PIPE_BLENDFACTOR_CONST_COLOR", + "PIPE_BLENDFACTOR_CONST_ALPHA", + "PIPE_BLENDFACTOR_SRC1_COLOR", + "PIPE_BLENDFACTOR_SRC1_ALPHA", + DEBUG_DUMP_INVALID_NAME, /* 0x0b */ + DEBUG_DUMP_INVALID_NAME, /* 0x0c */ + DEBUG_DUMP_INVALID_NAME, /* 0x0d */ + DEBUG_DUMP_INVALID_NAME, /* 0x0e */ + DEBUG_DUMP_INVALID_NAME, /* 0x0f */ + DEBUG_DUMP_INVALID_NAME, /* 0x10 */ + "PIPE_BLENDFACTOR_ZERO", + "PIPE_BLENDFACTOR_INV_SRC_COLOR", + "PIPE_BLENDFACTOR_INV_SRC_ALPHA", + "PIPE_BLENDFACTOR_INV_DST_ALPHA", + "PIPE_BLENDFACTOR_INV_DST_COLOR", + DEBUG_DUMP_INVALID_NAME, /* 0x16 */ + "PIPE_BLENDFACTOR_INV_CONST_COLOR", + "PIPE_BLENDFACTOR_INV_CONST_ALPHA", + "PIPE_BLENDFACTOR_INV_SRC1_COLOR", + "PIPE_BLENDFACTOR_INV_SRC1_ALPHA" +}; + +static const char * +debug_dump_blend_factor_short_names[] = { + DEBUG_DUMP_INVALID_NAME, /* 0x0 */ + "one", + "src_color", + "src_alpha", + "dst_alpha", + "dst_color", + "src_alpha_saturate", + "const_color", + "const_alpha", + "src1_color", + "src1_alpha", + DEBUG_DUMP_INVALID_NAME, /* 0x0b */ + DEBUG_DUMP_INVALID_NAME, /* 0x0c */ + DEBUG_DUMP_INVALID_NAME, /* 0x0d */ + DEBUG_DUMP_INVALID_NAME, /* 0x0e */ + DEBUG_DUMP_INVALID_NAME, /* 0x0f */ + DEBUG_DUMP_INVALID_NAME, /* 0x10 */ + "zero", + "inv_src_color", + "inv_src_alpha", + "inv_dst_alpha", + "inv_dst_color", + DEBUG_DUMP_INVALID_NAME, /* 0x16 */ + "inv_const_color", + "inv_const_alpha", + "inv_src1_color", + "inv_src1_alpha" +}; + +DEFINE_DEBUG_DUMP_CONTINUOUS(blend_factor) + + +static const char * +debug_dump_blend_func_names[] = { + "PIPE_BLEND_ADD", + "PIPE_BLEND_SUBTRACT", + "PIPE_BLEND_REVERSE_SUBTRACT", + "PIPE_BLEND_MIN", + "PIPE_BLEND_MAX" +}; + +static const char * +debug_dump_blend_func_short_names[] = { + "add", + "sub", + "rev_sub", + "min", + "max" +}; + +DEFINE_DEBUG_DUMP_CONTINUOUS(blend_func) + + +static const char * +debug_dump_func_names[] = { + "PIPE_FUNC_NEVER", + "PIPE_FUNC_LESS", + "PIPE_FUNC_EQUAL", + "PIPE_FUNC_LEQUAL", + "PIPE_FUNC_GREATER", + "PIPE_FUNC_NOTEQUAL", + "PIPE_FUNC_GEQUAL", + "PIPE_FUNC_ALWAYS" +}; + +static const char * +debug_dump_func_short_names[] = { + "never", + "less", + "equal", + "less_equal", + "greater", + "not_equal", + "greater_equal", + "always" +}; + +DEFINE_DEBUG_DUMP_CONTINUOUS(func) diff --git a/src/gallium/auxiliary/util/u_debug_dump.h b/src/gallium/auxiliary/util/u_debug_dump.h new file mode 100644 index 0000000000..102935559c --- /dev/null +++ b/src/gallium/auxiliary/util/u_debug_dump.h @@ -0,0 +1,65 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Dump data in human/machine readable format. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ + +#ifndef U_DEBUG_DUMP_H_ +#define U_DEBUG_DUMP_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + + +#ifdef __cplusplus +extern "C" { +#endif + + +const char * +debug_dump_blend_factor(unsigned value, boolean shortened); + +const char * +debug_dump_blend_func(unsigned value, boolean shortened); + +const char * +debug_dump_func(unsigned value, boolean shortened); + + +/* FIXME: Move the other debug_dump_xxx functions out of u_debug.h into here. */ + + +#ifdef __cplusplus +} +#endif + +#endif /* U_DEBUG_H_ */ diff --git a/src/gallium/auxiliary/util/u_fifo.h b/src/gallium/auxiliary/util/u_fifo.h new file mode 100644 index 0000000000..9e007de1ad --- /dev/null +++ b/src/gallium/auxiliary/util/u_fifo.h @@ -0,0 +1,94 @@ +/************************************************************************** + * + * Copyright © 2009 Jakob Bornecrantz + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_FIFO_H +#define U_FIFO_H + +#include "util/u_memory.h" + +struct util_fifo +{ + size_t head; + size_t tail; + size_t num; + size_t size; +}; + +static INLINE struct util_fifo * +u_fifo_create(size_t size) +{ + struct util_fifo *fifo; + fifo = MALLOC(sizeof(*fifo) + size * sizeof(void*)); + + fifo->head = 0; + fifo->tail = 0; + fifo->num = 0; + fifo->size = size; + + return fifo; +} + +static INLINE boolean +u_fifo_add(struct util_fifo *fifo, void *ptr) +{ + void **array = (void**)&fifo[1]; + if (fifo->num >= fifo->size) + return FALSE; + + if (++fifo->head >= fifo->size) + fifo->head = 0; + + array[fifo->head] = ptr; + + ++fifo->num; + + return TRUE; +} + +static INLINE boolean +u_fifo_pop(struct util_fifo *fifo, void **ptr) +{ + void **array = (void**)&fifo[1]; + + if (!fifo->num) + return FALSE; + + if (++fifo->tail >= fifo->size) + fifo->tail = 0; + + *ptr = array[fifo->tail]; + + ++fifo->num; + + return TRUE; +} + +static INLINE void +u_fifo_destroy(struct util_fifo *fifo) +{ + FREE(fifo); +} + +#endif diff --git a/src/gallium/auxiliary/util/u_format.c b/src/gallium/auxiliary/util/u_format.c new file mode 100644 index 0000000000..98ea13b60b --- /dev/null +++ b/src/gallium/auxiliary/util/u_format.c @@ -0,0 +1,46 @@ +/************************************************************************** + * + * Copyright 2009 Vmware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE 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 "u_format.h" + + +const struct util_format_description * +util_format_description(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description_table; + + while(TRUE) { + if(desc->format == format) + return desc; + + if(desc->format == PIPE_FORMAT_NONE) + return NULL; + + ++desc; + }; +} diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv new file mode 100644 index 0000000000..f1bf94f17d --- /dev/null +++ b/src/gallium/auxiliary/util/u_format.csv @@ -0,0 +1,99 @@ +PIPE_FORMAT_A8R8G8B8_UNORM , arith , 1, 1, un8 , un8 , un8 , un8 , zyxw, rgb +PIPE_FORMAT_X8R8G8B8_UNORM , arith , 1, 1, un8 , un8 , un8 , un8 , zyx1, rgb +PIPE_FORMAT_B8G8R8A8_UNORM , arith , 1, 1, un8 , un8 , un8 , un8 , yzwx, rgb +PIPE_FORMAT_B8G8R8X8_UNORM , arith , 1, 1, un8 , un8 , un8 , un8 , yzw1, rgb +PIPE_FORMAT_A1R5G5B5_UNORM , arith , 1, 1, un5 , un5 , un5 , un1 , zyxw, rgb +PIPE_FORMAT_A4R4G4B4_UNORM , arith , 1, 1, un4 , un4 , un4 , un4 , zyxw, rgb +PIPE_FORMAT_R5G6B5_UNORM , arith , 1, 1, un5 , un6 , un5 , , zyx1, rgb +PIPE_FORMAT_A2B10G10R10_UNORM , arith , 1, 1, un10, un10, un10, un2 , xyzw, rgb +PIPE_FORMAT_L8_UNORM , arith , 1, 1, un8 , , , , xxx1, rgb +PIPE_FORMAT_A8_UNORM , arith , 1, 1, un8 , , , , 000x, rgb +PIPE_FORMAT_I8_UNORM , arith , 1, 1, un8 , , , , xxxx, rgb +PIPE_FORMAT_A8L8_UNORM , arith , 1, 1, un8 , un8 , , , xxxy, rgb +PIPE_FORMAT_L16_UNORM , arith , 1, 1, un16, , , , xxx1, rgb +PIPE_FORMAT_Z16_UNORM , array , 1, 1, un16, , , , x___, zs +PIPE_FORMAT_Z32_UNORM , array , 1, 1, un32, , , , x___, zs +PIPE_FORMAT_Z32_FLOAT , array , 1, 1, f32 , , , , x___, zs +PIPE_FORMAT_S8Z24_UNORM , arith , 1, 1, un24, un8 , , , xy__, zs +PIPE_FORMAT_Z24S8_UNORM , arith , 1, 1, un8 , un24, , , yx__, zs +PIPE_FORMAT_X8Z24_UNORM , arith , 1, 1, un24, un8 , , , x___, zs +PIPE_FORMAT_Z24X8_UNORM , arith , 1, 1, un8 , un24, , , y___, zs +PIPE_FORMAT_S8_UNORM , array , 1, 1, un8 , , , , _x__, zs +PIPE_FORMAT_R64_FLOAT , array , 1, 1, f64 , , , , x001, rgb +PIPE_FORMAT_R64G64_FLOAT , array , 1, 1, f64 , f64 , , , xy01, rgb +PIPE_FORMAT_R64G64B64_FLOAT , array , 1, 1, f64 , f64 , f64 , , xyz1, rgb +PIPE_FORMAT_R64G64B64A64_FLOAT , array , 1, 1, f64 , f64 , f64 , f64 , xyzw, rgb +PIPE_FORMAT_R32_FLOAT , array , 1, 1, f32 , , , , x001, rgb +PIPE_FORMAT_R32G32_FLOAT , array , 1, 1, f32 , f32 , , , xy01, rgb +PIPE_FORMAT_R32G32B32_FLOAT , array , 1, 1, f32 , f32 , f32 , , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_FLOAT , array , 1, 1, f32 , f32 , f32 , f32 , xyzw, rgb +PIPE_FORMAT_R32_UNORM , array , 1, 1, un32, , , , x001, rgb +PIPE_FORMAT_R32G32_UNORM , array , 1, 1, un32, un32, , , xy01, rgb +PIPE_FORMAT_R32G32B32_UNORM , array , 1, 1, un32, un32, un32, , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_UNORM , array , 1, 1, un32, un32, un32, un32, xyzw, rgb +PIPE_FORMAT_R32_USCALED , array , 1, 1, u32 , , , , x001, rgb +PIPE_FORMAT_R32G32_USCALED , array , 1, 1, u32 , u32 , , , xy01, rgb +PIPE_FORMAT_R32G32B32_USCALED , array , 1, 1, u32 , u32 , u32 , , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_USCALED , array , 1, 1, u32 , u32 , u32 , u32 , xyzw, rgb +PIPE_FORMAT_R32_SNORM , array , 1, 1, sn32, , , , x001, rgb +PIPE_FORMAT_R32G32_SNORM , array , 1, 1, sn32, sn32, , , xy01, rgb +PIPE_FORMAT_R32G32B32_SNORM , array , 1, 1, sn32, sn32, sn32, , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_SNORM , array , 1, 1, sn32, sn32, sn32, sn32, xyzw, rgb +PIPE_FORMAT_R32_SSCALED , array , 1, 1, s32 , , , , x001, rgb +PIPE_FORMAT_R32G32_SSCALED , array , 1, 1, s32 , s32 , , , xy01, rgb +PIPE_FORMAT_R32G32B32_SSCALED , array , 1, 1, s32 , s32 , s32 , , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_SSCALED , array , 1, 1, s32 , s32 , s32 , s32 , xyzw, rgb +PIPE_FORMAT_R16_UNORM , array , 1, 1, un16, , , , x001, rgb +PIPE_FORMAT_R16G16_UNORM , array , 1, 1, un16, un16, , , xy01, rgb +PIPE_FORMAT_R16G16B16_UNORM , array , 1, 1, un16, un16, un16, , xyz1, rgb +PIPE_FORMAT_R16G16B16A16_UNORM , array , 1, 1, un16, un16, un16, un16, xyzw, rgb +PIPE_FORMAT_R16_USCALED , array , 1, 1, u16 , , , , x001, rgb +PIPE_FORMAT_R16G16_USCALED , array , 1, 1, u16 , u16 , , , xy01, rgb +PIPE_FORMAT_R16G16B16_USCALED , array , 1, 1, u16 , u16 , u16 , , xyz1, rgb +PIPE_FORMAT_R16G16B16A16_USCALED , array , 1, 1, u16 , u16 , u16 , u16 , xyzw, rgb +PIPE_FORMAT_R16_SNORM , array , 1, 1, sn16, , , , x001, rgb +PIPE_FORMAT_R16G16_SNORM , array , 1, 1, sn16, sn16, , , xy01, rgb +PIPE_FORMAT_R16G16B16_SNORM , array , 1, 1, sn16, sn16, sn16, , xyz1, rgb +PIPE_FORMAT_R16G16B16A16_SNORM , array , 1, 1, sn16, sn16, sn16, sn16, xyzw, rgb +PIPE_FORMAT_R16_SSCALED , array , 1, 1, s16 , , , , x001, rgb +PIPE_FORMAT_R16G16_SSCALED , array , 1, 1, s16 , s16 , , , xy01, rgb +PIPE_FORMAT_R16G16B16_SSCALED , array , 1, 1, s16 , s16 , s16 , , xyz1, rgb +PIPE_FORMAT_R16G16B16A16_SSCALED , array , 1, 1, s16 , s16 , s16 , s16 , xyzw, rgb +PIPE_FORMAT_R8_UNORM , array , 1, 1, un8 , , , , x001, rgb +PIPE_FORMAT_R8G8_UNORM , array , 1, 1, un8 , un8 , , , xy01, rgb +PIPE_FORMAT_R8G8B8_UNORM , array , 1, 1, un8 , un8 , un8 , , xyz1, rgb +PIPE_FORMAT_R8G8B8A8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , xyzw, rgb +PIPE_FORMAT_R8G8B8X8_UNORM , array , 1, 1, un8 , un8 , un8 , un8 , xyz1, rgb +PIPE_FORMAT_R8_USCALED , array , 1, 1, u8 , , , , x001, rgb +PIPE_FORMAT_R8G8_USCALED , array , 1, 1, u8 , u8 , , , xy01, rgb +PIPE_FORMAT_R8G8B8_USCALED , array , 1, 1, u8 , u8 , u8 , , xyz1, rgb +PIPE_FORMAT_R8G8B8A8_USCALED , array , 1, 1, u8 , u8 , u8 , u8 , xyzw, rgb +PIPE_FORMAT_R8G8B8X8_USCALED , array , 1, 1, u8 , u8 , u8 , u8 , xyz1, rgb +PIPE_FORMAT_R8_SNORM , array , 1, 1, sn8 , , , , x001, rgb +PIPE_FORMAT_R8G8_SNORM , array , 1, 1, sn8 , sn8 , , , xy01, rgb +PIPE_FORMAT_R8G8B8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , , xyz1, rgb +PIPE_FORMAT_R8G8B8A8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyzw, rgb +PIPE_FORMAT_R8G8B8X8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyz1, rgb +PIPE_FORMAT_B6G5R5_SNORM , arith , 1, 1, sn5 , sn5 , sn6 , , zyx1, rgb +PIPE_FORMAT_A8B8G8R8_SNORM , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyxw, rgb +PIPE_FORMAT_X8B8G8R8_SNORM , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyx1, rgb +PIPE_FORMAT_R8_SSCALED , array , 1, 1, s8 , , , , x001, rgb +PIPE_FORMAT_R8G8_SSCALED , array , 1, 1, s8 , s8 , , , xy01, rgb +PIPE_FORMAT_R8G8B8_SSCALED , array , 1, 1, s8 , s8 , s8 , , xyz1, rgb +PIPE_FORMAT_R8G8B8A8_SSCALED , array , 1, 1, s8 , s8 , s8 , s8 , xyzw, rgb +PIPE_FORMAT_R8G8B8X8_SSCALED , array , 1, 1, s8 , s8 , s8 , s8 , xyz1, rgb +PIPE_FORMAT_R32_FIXED , array , 1, 1, h32 , , , , x001, rgb +PIPE_FORMAT_R32G32_FIXED , array , 1, 1, h32 , h32 , , , xy01, rgb +PIPE_FORMAT_R32G32B32_FIXED , array , 1, 1, h32 , h32 , h32 , , xyz1, rgb +PIPE_FORMAT_R32G32B32A32_FIXED , array , 1, 1, h32 , h32 , h32 , h32 , xyzw, rgb +PIPE_FORMAT_L8_SRGB , arith , 1, 1, u8 , , , , xxx1, srgb +PIPE_FORMAT_A8L8_SRGB , arith , 1, 1, u8 , u8 , , , xxxy, srgb +PIPE_FORMAT_R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , , xyz1, srgb +PIPE_FORMAT_R8G8B8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , xyzw, srgb +PIPE_FORMAT_R8G8B8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , xyz1, srgb +PIPE_FORMAT_A8R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , wxyz, srgb +PIPE_FORMAT_X8R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , 1xyz, srgb +PIPE_FORMAT_B8G8R8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyxw, srgb +PIPE_FORMAT_B8G8R8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyx1, srgb +PIPE_FORMAT_X8UB8UG8SR8S_NORM , arith , 1, 1, sn8 , sn8 , un8 , x8 , 1zyx, rgb +PIPE_FORMAT_B6UG5SR5S_NORM , arith , 1, 1, sn5 , sn5 , un6 , , xyz1, rgb diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h new file mode 100644 index 0000000000..7b5b7fcda5 --- /dev/null +++ b/src/gallium/auxiliary/util/u_format.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2009 Vmware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef U_FORMAT_H +#define U_FORMAT_H + + +#include "pipe/p_format.h" + + +enum util_format_layout { + UTIL_FORMAT_LAYOUT_SCALAR = 0, + UTIL_FORMAT_LAYOUT_ARITH = 1, + UTIL_FORMAT_LAYOUT_ARRAY = 2, + UTIL_FORMAT_LAYOUT_YUV = 3, + UTIL_FORMAT_LAYOUT_DXT = 4 +}; + + +struct util_format_block +{ + /** Block width in pixels */ + unsigned width; + + /** Block height in pixels */ + unsigned height; + + /** Block size in bytes */ + unsigned bits; +}; + + +enum util_format_type { + UTIL_FORMAT_TYPE_VOID = 0, + UTIL_FORMAT_TYPE_UNSIGNED = 1, + UTIL_FORMAT_TYPE_SIGNED = 2, + UTIL_FORMAT_TYPE_FIXED = 3, + UTIL_FORMAT_TYPE_FLOAT = 4 +}; + + +enum util_format_swizzle { + UTIL_FORMAT_SWIZZLE_X = 0, + UTIL_FORMAT_SWIZZLE_Y = 1, + UTIL_FORMAT_SWIZZLE_Z = 2, + UTIL_FORMAT_SWIZZLE_W = 3, + UTIL_FORMAT_SWIZZLE_0 = 4, + UTIL_FORMAT_SWIZZLE_1 = 5, + UTIL_FORMAT_SWIZZLE_NONE = 6 +}; + + +enum util_format_colorspace { + UTIL_FORMAT_COLORSPACE_RGB = 0, + UTIL_FORMAT_COLORSPACE_SRGB = 1, + UTIL_FORMAT_COLORSPACE_YUV = 2, + UTIL_FORMAT_COLORSPACE_ZS = 3, +}; + + +struct util_format_channel_description +{ + unsigned type:6; + unsigned normalized:1; + unsigned size:9; +}; + + +struct util_format_description +{ + enum pipe_format format; + const char *name; + struct util_format_block block; + enum util_format_layout layout; + struct util_format_channel_description channel[4]; + unsigned char swizzle[4]; + enum util_format_colorspace colorspace; +}; + + +extern const struct util_format_description +util_format_description_table[]; + + +const struct util_format_description * +util_format_description(enum pipe_format format); + + +void +util_format_read_4f(enum pipe_format format, + float *dst, unsigned dst_stride, + const void *src, unsigned src_stride, + unsigned x, unsigned y, unsigned w, unsigned h); + +void +util_format_write_4f(enum pipe_format format, + const float *src, unsigned src_stride, + void *dst, unsigned dst_stride, + unsigned x, unsigned y, unsigned w, unsigned h); + +void +util_format_read_4ub(enum pipe_format format, + uint8_t *dst, unsigned dst_stride, + const void *src, unsigned src_stride, + unsigned x, unsigned y, unsigned w, unsigned h); + +void +util_format_write_4ub(enum pipe_format format, + const uint8_t *src, unsigned src_stride, + void *dst, unsigned dst_stride, + unsigned x, unsigned y, unsigned w, unsigned h); + +#endif /* ! U_FORMAT_H */ diff --git a/src/gallium/auxiliary/util/u_format_access.py b/src/gallium/auxiliary/util/u_format_access.py new file mode 100644 index 0000000000..eeb1a9657f --- /dev/null +++ b/src/gallium/auxiliary/util/u_format_access.py @@ -0,0 +1,498 @@ +#!/usr/bin/env python + +''' +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Pixel format accessor functions. + * + * @author Jose Fonseca <jfonseca@vmware.com> + */ +''' + + +import sys + +from u_format_parse import * + + +def short_name(format): + '''Make up a short norm for a format, suitable to be used as suffix in + function names.''' + + name = format.name + if name.startswith('PIPE_FORMAT_'): + name = name[len('PIPE_FORMAT_'):] + name = name.lower() + return name + + +def is_format_supported(format): + '''Determines whether we actually have the plumbing necessary to generate the + to read/write to/from this format.''' + + # FIXME: Ideally we would support any format combination here. + + # XXX: It should be straightforward to support srgb + if format.colorspace not in ('rgb', 'zs'): + return False + + if format.layout not in (ARITH, ARRAY): + return False + + for i in range(4): + type = format.in_types[i] + if type.kind not in (VOID, UNSIGNED, FLOAT): + return False + + # We can only read a color from a depth/stencil format if the depth channel is present + if format.colorspace == 'zs' and format.out_swizzle[0] == SWIZZLE_NONE: + return False + + return True + + +def native_type(format): + '''Get the native appropriate for a format.''' + + if format.layout == ARITH: + # For arithmetic pixel formats return the integer type that matches the whole pixel + return 'uint%u_t' % format.block_size() + elif format.layout == ARRAY: + # For array pixel formats return the integer type that matches the color channel + type = format.in_types[0] + if type.kind == UNSIGNED: + return 'uint%u_t' % type.size + elif type.kind == SIGNED: + return 'int%u_t' % type.size + elif type.kind == FLOAT: + if type.size == 32: + return 'float' + elif type.size == 64: + return 'double' + else: + assert False + else: + assert False + else: + assert False + + +def intermediate_native_type(bits, sign): + '''Find a native type adequate to hold intermediate results of the request bit size.''' + + bytes = 4 # don't use anything smaller than 32bits + while bytes * 8 < bits: + bytes *= 2 + bits = bytes*8 + + if sign: + return 'int%u_t' % bits + else: + return 'uint%u_t' % bits + + +def get_one_shift(type): + '''Get the number of the bit that matches unity for this type.''' + if type.kind == 'FLOAT': + assert False + if not type.norm: + return 0 + if type.kind == UNSIGNED: + return type.size + if type.kind == SIGNED: + return type.size - 1 + if type.kind == FIXED: + return type.size / 2 + assert False + + +def get_one(type): + '''Get the value of unity for this type.''' + if type.kind == 'FLOAT' or not type.norm: + return 1 + else: + return (1 << get_one_shift(type)) - 1 + + +def generate_clamp(): + '''Code generate the clamping functions for each type. + + We don't use a macro so that arguments with side effects, + like *src_pixel++ are correctly handled. + ''' + + for suffix, native_type in [ + ('', 'double'), + ('f', 'float'), + ('ui', 'unsigned int'), + ('si', 'int'), + ]: + print 'static INLINE %s' % native_type + print 'clamp%s(%s value, %s lbound, %s ubound)' % (suffix, native_type, native_type, native_type) + print '{' + print ' if(value < lbound)' + print ' return lbound;' + print ' if(value > ubound)' + print ' return ubound;' + print ' return value;' + print '}' + print + + +def clamp_expr(src_type, dst_type, dst_native_type, value): + '''Generate the expression to clamp the value in the source type to the + destination type range.''' + + if src_type == dst_type: + return value + + # Pick the approriate clamp function + if src_type.kind == FLOAT: + if src_type.size == 32: + func = 'clampf' + elif src_type.size == 64: + func = 'clamp' + else: + assert False + elif src_type.kind == UNSIGNED: + func = 'clampui' + elif src_type.kind == SIGNED: + func = 'clampsi' + else: + assert False + + # Clamp floats to [-1, 1] or [0, 1] range + if src_type.kind == FLOAT and dst_type.norm: + max = 1 + if src_type.sign and dst_type.sign: + min = -1 + else: + min = 0 + return '%s(%s, %s, %s)' % (func, value, min, max) + + # FIXME: Also clamp scaled values + + return value + + +def conversion_expr(src_type, dst_type, dst_native_type, value): + '''Generate the expression to convert a value between two types.''' + + if src_type == dst_type: + return value + + if src_type.kind == FLOAT and dst_type.kind == FLOAT: + return '(%s)%s' % (dst_native_type, value) + + if not src_type.norm and not dst_type.norm: + return '(%s)%s' % (dst_native_type, value) + + value = clamp_expr(src_type, dst_type, dst_native_type, value) + + if dst_type.kind == FLOAT: + if src_type.norm: + one = get_one(src_type) + if src_type.size <= 23: + scale = '(1.0f/0x%x)' % one + else: + # bigger than single precision mantissa, use double + scale = '(1.0/0x%x)' % one + value = '(%s * %s)' % (value, scale) + return '(%s)%s' % (dst_native_type, value) + + if src_type.kind == FLOAT: + if dst_type.norm: + dst_one = get_one(dst_type) + if dst_type.size <= 23: + scale = '0x%x' % dst_one + else: + # bigger than single precision mantissa, use double + scale = '(double)0x%x' % dst_one + value = '(%s * %s)' % (value, scale) + return '(%s)%s' % (dst_native_type, value) + + if src_type.kind == dst_type.kind: + src_one = get_one(src_type) + dst_one = get_one(dst_type) + + if src_one > dst_one and src_type.norm and dst_type.norm: + # We can just bitshift + src_shift = get_one_shift(src_type) + dst_shift = get_one_shift(dst_type) + value = '(%s >> %s)' % (value, src_shift - dst_shift) + else: + # We need to rescale using an intermediate type big enough to hold the multiplication of both + tmp_native_type = intermediate_native_type(src_type.size + dst_type.size, src_type.sign and dst_type.sign) + value = '(%s)%s' % (tmp_native_type, value) + value = '%s * 0x%x / 0x%x' % (value, dst_one, src_one) + value = '(%s)%s' % (dst_native_type, value) + return value + + assert False + + +def generate_format_read(format, dst_type, dst_native_type, dst_suffix): + '''Generate the function to read pixels from a particular format''' + + name = short_name(format) + + src_native_type = native_type(format) + + print 'static void' + print 'util_format_%s_read_%s(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, dst_suffix, dst_native_type) + print '{' + print ' unsigned x, y;' + print ' const uint8_t *src_row = src + y0*src_stride;' + print ' %s *dst_row = dst;' % dst_native_type + print ' for (y = 0; y < h; ++y) {' + print ' const %s *src_pixel = (const %s *)(src_row + x0*%u);' % (src_native_type, src_native_type, format.stride()) + print ' %s *dst_pixel = dst_row;' %dst_native_type + print ' for (x = 0; x < w; ++x) {' + + names = ['']*4 + if format.colorspace == 'rgb': + for i in range(4): + swizzle = format.out_swizzle[i] + if swizzle < 4: + names[swizzle] += 'rgba'[i] + elif format.colorspace == 'zs': + swizzle = format.out_swizzle[0] + if swizzle < 4: + names[swizzle] = 'z' + else: + assert False + else: + assert False + + if format.layout == ARITH: + print ' %s pixel = *src_pixel++;' % src_native_type + shift = 0; + for i in range(4): + src_type = format.in_types[i] + width = src_type.size + if names[i]: + value = 'pixel' + mask = (1 << width) - 1 + if shift: + value = '(%s >> %u)' % (value, shift) + if shift + width < format.block_size(): + value = '(%s & 0x%x)' % (value, mask) + value = conversion_expr(src_type, dst_type, dst_native_type, value) + print ' %s %s = %s;' % (dst_native_type, names[i], value) + shift += width + elif format.layout == ARRAY: + for i in range(4): + src_type = format.in_types[i] + if names[i]: + value = '(*src_pixel++)' + value = conversion_expr(src_type, dst_type, dst_native_type, value) + print ' %s %s = %s;' % (dst_native_type, names[i], value) + else: + assert False + + for i in range(4): + if format.colorspace == 'rgb': + swizzle = format.out_swizzle[i] + if swizzle < 4: + value = names[swizzle] + elif swizzle == SWIZZLE_0: + value = '0' + elif swizzle == SWIZZLE_1: + value = '1' + else: + assert False + elif format.colorspace == 'zs': + if i < 3: + value = 'z' + else: + value = '1' + else: + assert False + print ' *dst_pixel++ = %s; /* %s */' % (value, 'rgba'[i]) + + print ' }' + print ' src_row += src_stride;' + print ' dst_row += dst_stride/sizeof(%s);' % dst_native_type + print ' }' + print '}' + print + + +def generate_format_write(format, src_type, src_native_type, src_suffix): + '''Generate the function to write pixels to a particular format''' + + name = short_name(format) + + dst_native_type = native_type(format) + + print 'static void' + print 'util_format_%s_write_%s(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, src_suffix, src_native_type) + print '{' + print ' unsigned x, y;' + print ' uint8_t *dst_row = dst + y0*dst_stride;' + print ' const %s *src_row = src;' % src_native_type + print ' for (y = 0; y < h; ++y) {' + print ' %s *dst_pixel = (%s *)(dst_row + x0*%u);' % (dst_native_type, dst_native_type, format.stride()) + print ' const %s *src_pixel = src_row;' %src_native_type + print ' for (x = 0; x < w; ++x) {' + + inv_swizzle = [None]*4 + if format.colorspace == 'rgb': + for i in range(4): + swizzle = format.out_swizzle[i] + if swizzle < 4: + inv_swizzle[swizzle] = i + elif format.colorspace == 'zs': + swizzle = format.out_swizzle[0] + if swizzle < 4: + inv_swizzle[swizzle] = 0 + else: + assert False + + if format.layout == ARITH: + print ' %s pixel = 0;' % dst_native_type + shift = 0; + for i in range(4): + dst_type = format.in_types[i] + width = dst_type.size + if inv_swizzle[i] is not None: + value = 'src_pixel[%u]' % inv_swizzle[i] + value = conversion_expr(src_type, dst_type, dst_native_type, value) + if shift: + value = '(%s << %u)' % (value, shift) + print ' pixel |= %s;' % value + shift += width + print ' *dst_pixel++ = pixel;' + elif format.layout == ARRAY: + for i in range(4): + dst_type = format.in_types[i] + if inv_swizzle[i] is not None: + value = 'src_pixel[%u]' % inv_swizzle[i] + value = conversion_expr(src_type, dst_type, dst_native_type, value) + print ' *dst_pixel++ = %s;' % value + else: + assert False + print ' src_pixel += 4;' + + print ' }' + print ' dst_row += dst_stride;' + print ' src_row += src_stride/sizeof(%s);' % src_native_type + print ' }' + print '}' + print + + +def generate_read(formats, dst_type, dst_native_type, dst_suffix): + '''Generate the dispatch function to read pixels from any format''' + + for format in formats: + if is_format_supported(format): + generate_format_read(format, dst_type, dst_native_type, dst_suffix) + + print 'void' + print 'util_format_read_%s(enum pipe_format format, %s *dst, unsigned dst_stride, const void *src, unsigned src_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (dst_suffix, dst_native_type) + print '{' + print ' void (*func)(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % dst_native_type + print ' switch(format) {' + for format in formats: + if is_format_supported(format): + print ' case %s:' % format.name + print ' func = &util_format_%s_read_%s;' % (short_name(format), dst_suffix) + print ' break;' + print ' default:' + print ' debug_printf("unsupported format\\n");' + print ' return;' + print ' }' + print ' func(dst, dst_stride, (const uint8_t *)src, src_stride, x, y, w, h);' + print '}' + print + + +def generate_write(formats, src_type, src_native_type, src_suffix): + '''Generate the dispatch function to write pixels to any format''' + + for format in formats: + if is_format_supported(format): + generate_format_write(format, src_type, src_native_type, src_suffix) + + print 'void' + print 'util_format_write_%s(enum pipe_format format, const %s *src, unsigned src_stride, void *dst, unsigned dst_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (src_suffix, src_native_type) + + print '{' + print ' void (*func)(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % src_native_type + print ' switch(format) {' + for format in formats: + if is_format_supported(format): + print ' case %s:' % format.name + print ' func = &util_format_%s_write_%s;' % (short_name(format), src_suffix) + print ' break;' + print ' default:' + print ' debug_printf("unsupported format\\n");' + print ' return;' + print ' }' + print ' func(src, src_stride, (uint8_t *)dst, dst_stride, x, y, w, h);' + print '}' + print + + +def main(): + formats = [] + for arg in sys.argv[1:]: + formats.extend(parse(arg)) + + print '/* This file is autogenerated by u_format_access.py from u_format.csv. Do not edit directly. */' + print + # This will print the copyright message on the top of this file + print __doc__.strip() + print + print '#include "pipe/p_compiler.h"' + print '#include "u_format.h"' + print '#include "u_math.h"' + print + + generate_clamp() + + type = Type(FLOAT, False, 32) + native_type = 'float' + suffix = '4f' + + generate_read(formats, type, native_type, suffix) + generate_write(formats, type, native_type, suffix) + + type = Type(UNSIGNED, True, 8) + native_type = 'uint8_t' + suffix = '4ub' + + generate_read(formats, type, native_type, suffix) + generate_write(formats, type, native_type, suffix) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/auxiliary/util/u_format_parse.py b/src/gallium/auxiliary/util/u_format_parse.py new file mode 100755 index 0000000000..493aff7112 --- /dev/null +++ b/src/gallium/auxiliary/util/u_format_parse.py @@ -0,0 +1,141 @@ +#!/usr/bin/env python + +''' +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE 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. + * + **************************************************************************/ +''' + + +import sys + + +VOID, UNSIGNED, SIGNED, FIXED, FLOAT = range(5) + +SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_0, SWIZZLE_1, SWIZZLE_NONE, = range(7) + +ARITH = 'arith' +ARRAY = 'array' + + +class Type: + '''Describe the type of a color channel.''' + + def __init__(self, kind, norm, size): + self.kind = kind + self.norm = norm + self.size = size + self.sign = kind in (SIGNED, FIXED, FLOAT) + + def __str__(self): + s = str(self.kind) + if self.norm: + s += 'n' + s += str(self.size) + return s + + def __eq__(self, other): + return self.kind == other.kind and self.norm == other.norm and self.size == other.size + + +class Format: + '''Describe a pixel format.''' + + def __init__(self, name, layout, block_width, block_height, in_types, out_swizzle, colorspace): + self.name = name + self.layout = layout + self.block_width = block_width + self.block_height = block_height + self.in_types = in_types + self.out_swizzle = out_swizzle + self.name = name + self.colorspace = colorspace + + def __str__(self): + return self.name + + def block_size(self): + size = 0 + for type in self.in_types: + size += type.size + return size + + def stride(self): + return self.block_size()/8 + + +_kind_parse_map = { + '': VOID, + 'x': VOID, + 'u': UNSIGNED, + 's': SIGNED, + 'h': FIXED, + 'f': FLOAT, +} + +_swizzle_parse_map = { + 'x': SWIZZLE_X, + 'y': SWIZZLE_Y, + 'z': SWIZZLE_Z, + 'w': SWIZZLE_W, + '0': SWIZZLE_0, + '1': SWIZZLE_1, + '_': SWIZZLE_NONE, +} + +def parse(filename): + '''Parse the format descrition in CSV format in terms of the + Type and Format classes above.''' + + stream = open(filename) + formats = [] + for line in stream: + line = line.rstrip() + fields = [field.strip() for field in line.split(',')] + name = fields[0] + layout = fields[1] + block_width, block_height = map(int, fields[2:4]) + in_types = [] + for field in fields[4:8]: + if field: + kind = _kind_parse_map[field[0]] + if field[1] == 'n': + norm = True + size = int(field[2:]) + else: + norm = False + size = int(field[1:]) + else: + kind = VOID + norm = False + size = 0 + in_type = Type(kind, norm, size) + in_types.append(in_type) + out_swizzle = [_swizzle_parse_map[swizzle] for swizzle in fields[8]] + colorspace = fields[9] + formats.append(Format(name, layout, block_width, block_height, in_types, out_swizzle, colorspace)) + return formats + diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py new file mode 100755 index 0000000000..8834568e8e --- /dev/null +++ b/src/gallium/auxiliary/util/u_format_table.py @@ -0,0 +1,144 @@ +#!/usr/bin/env python + +''' +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE 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. + * + **************************************************************************/ +''' + + +import sys + +from u_format_parse import * + + +def layout_map(layout): + return 'UTIL_FORMAT_LAYOUT_' + str(layout).upper() + + +def colorspace_map(colorspace): + return 'UTIL_FORMAT_COLORSPACE_' + str(colorspace).upper() + + +colorspace_channels_map = { + 'rgb': 'rgba', + 'rgba': 'rgba', + 'zs': 'zs', + 'yuv': ['y1', 'y2', 'u', 'v'], + 'dxt': [] +} + + +kind_map = { + VOID: "UTIL_FORMAT_TYPE_VOID", + UNSIGNED: "UTIL_FORMAT_TYPE_UNSIGNED", + SIGNED: "UTIL_FORMAT_TYPE_SIGNED", + FIXED: "UTIL_FORMAT_TYPE_FIXED", + FLOAT: "UTIL_FORMAT_TYPE_FLOAT", +} + + +def bool_map(value): + if value: + return "TRUE" + else: + return "FALSE" + + +swizzle_map = { + SWIZZLE_X: "UTIL_FORMAT_SWIZZLE_X", + SWIZZLE_Y: "UTIL_FORMAT_SWIZZLE_Y", + SWIZZLE_Z: "UTIL_FORMAT_SWIZZLE_Z", + SWIZZLE_W: "UTIL_FORMAT_SWIZZLE_W", + SWIZZLE_0: "UTIL_FORMAT_SWIZZLE_0", + SWIZZLE_1: "UTIL_FORMAT_SWIZZLE_1", + SWIZZLE_NONE: "UTIL_FORMAT_SWIZZLE_NONE", +} + + +def write_format_table(formats): + print '/* This file is autogenerated by u_format_table.py from u_format.csv. Do not edit directly. */' + print + # This will print the copyright message on the top of this file + print __doc__.strip() + print + print '#include "u_format.h"' + print + print 'const struct util_format_description' + print 'util_format_description_table[] = ' + print "{" + for format in formats: + print " {" + print " %s," % (format.name,) + print " \"%s\"," % (format.name,) + print " {%u, %u, %u}, /* block */" % (format.block_width, format.block_height, format.block_size()) + print " %s," % (layout_map(format.layout),) + print " {" + for i in range(4): + type = format.in_types[i] + if i < 3: + sep = "," + else: + sep = "" + print " {%s, %s, %u}%s /* %s */" % (kind_map[type.kind], bool_map(type.norm), type.size, sep, "xyzw"[i]) + print " }," + print " {" + for i in range(4): + swizzle = format.out_swizzle[i] + if i < 3: + sep = "," + else: + sep = "" + try: + comment = layout_channels_map[format.layout][i] + except: + comment = 'ignored' + print " %s%s /* %s */" % (swizzle_map[swizzle], sep, comment) + print " }," + print " %s," % (colorspace_map(format.colorspace),) + print " }," + print " {" + print " PIPE_FORMAT_NONE," + print " \"PIPE_FORMAT_NONE\"," + print " {0, 0, 0}," + print " 0," + print " {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}}," + print " {0, 0, 0, 0}," + print " 0" + print " }," + print "};" + + +def main(): + + formats = [] + for arg in sys.argv[1:]: + formats.extend(parse(arg)) + write_format_table(formats) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index ca797486a0..f06c0e463d 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -46,10 +46,6 @@ #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" -#include "tgsi/tgsi_build.h" -#include "tgsi/tgsi_dump.h" -#include "tgsi/tgsi_parse.h" - #include "cso_cache/cso_context.h" @@ -921,11 +917,19 @@ static void format_to_type_comps(enum pipe_format pformat, enum dtype *datatype, uint *comps) { + /* XXX I think this could be implemented in terms of the pf_*() functions */ switch (pformat) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_X8R8G8B8_UNORM: case PIPE_FORMAT_B8G8R8A8_UNORM: case PIPE_FORMAT_B8G8R8X8_UNORM: + case PIPE_FORMAT_R8G8B8A8_SRGB: + case PIPE_FORMAT_R8G8B8X8_SRGB: + case PIPE_FORMAT_A8R8G8B8_SRGB: + case PIPE_FORMAT_X8R8G8B8_SRGB: + case PIPE_FORMAT_B8G8R8A8_SRGB: + case PIPE_FORMAT_B8G8R8X8_SRGB: + case PIPE_FORMAT_R8G8B8_SRGB: *datatype = DTYPE_UBYTE; *comps = 4; return; @@ -942,12 +946,14 @@ format_to_type_comps(enum pipe_format pformat, *comps = 3; return; case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_L8_SRGB: case PIPE_FORMAT_A8_UNORM: case PIPE_FORMAT_I8_UNORM: *datatype = DTYPE_UBYTE; *comps = 1; return; case PIPE_FORMAT_A8L8_UNORM: + case PIPE_FORMAT_A8L8_SRGB: *datatype = DTYPE_UBYTE; *comps = 2; return; diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index 163522d3ef..cd6a9fcc09 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -53,11 +53,11 @@ __inline double ceil(double val) { double ceil_val; - if((val - (long) val) == 0) { + if ((val - (long) val) == 0) { ceil_val = val; } else { - if(val > 0) { + if (val > 0) { ceil_val = (long) val + 1; } else { @@ -73,11 +73,11 @@ __inline double floor(double val) { double floor_val; - if((val - (long) val) == 0) { + if ((val - (long) val) == 0) { floor_val = val; } else { - if(val > 0) { + if (val > 0) { floor_val = (long) val; } else { @@ -189,7 +189,10 @@ static INLINE double log2( double x ) extern float pow2_table[POW2_TABLE_SIZE]; - +/** + * Initialize math module. This should be called before using any + * other functions in this module. + */ extern void util_init_math(void); @@ -216,23 +219,24 @@ util_fast_exp2(float x) int32_t ipart; float fpart, mpart; union fi epart; - + if(x > 129.00000f) return 3.402823466e+38f; - - if(x < -126.99999f) + + if (x < -126.99999f) return 0.0f; ipart = (int32_t) x; fpart = x - (float) ipart; - + /* same as * epart.f = (float) (1 << ipart) - * but faster and without integer overflow for ipart > 31 */ + * but faster and without integer overflow for ipart > 31 + */ epart.i = (ipart + 127 ) << 23; - + mpart = pow2_table[POW2_TABLE_OFFSET + (int)(fpart * POW2_TABLE_SCALE)]; - + return epart.f * mpart; } @@ -254,6 +258,9 @@ util_fast_exp(float x) extern float log2_table[LOG2_TABLE_SIZE]; +/** + * Fast approximation to log2(x). + */ static INLINE float util_fast_log2(float x) { @@ -267,6 +274,9 @@ util_fast_log2(float x) } +/** + * Fast approximation to x^y. + */ static INLINE float util_fast_pow(float x, float y) { @@ -282,7 +292,6 @@ util_is_power_of_two( unsigned v ) } - /** * Floor(x), returned as int. */ @@ -292,8 +301,8 @@ util_ifloor(float f) int ai, bi; double af, bf; union fi u; - af = (3 << 22) + 0.5 + (double)f; - bf = (3 << 22) + 0.5 - (double)f; + af = (3 << 22) + 0.5 + (double) f; + bf = (3 << 22) + 0.5 - (double) f; u.f = (float) af; ai = u.i; u.f = (float) bf; bi = u.i; return (ai - bi) >> 1; @@ -313,9 +322,9 @@ util_iround(float f) #elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) int r; _asm { - fld f - fistp r - } + fld f + fistp r + } return r; #else if (f >= 0.0f) @@ -340,15 +349,27 @@ util_is_inf_or_nan(float x) /** + * Test whether x is a power of two. + */ +static INLINE boolean +util_is_pot(unsigned x) +{ + return (x & (x - 1)) == 0; +} + + +/** * Find first bit set in word. Least significant bit is 1. * Return 0 if no bits set. */ -#if defined(_MSC_VER) && _MSC_VER >= 1300 +#if defined(_MSC_VER) && _MSC_VER >= 1300 && (_M_IX86 || _M_AMD64 || _M_IA64) +unsigned char _BitScanForward(unsigned long* Index, unsigned long Mask); +#pragma intrinsic(_BitScanForward) static INLINE unsigned long ffs( unsigned long u ) { unsigned long i; - if(_BitScanForward(&i, u)) + if (_BitScanForward(&i, u)) return i + 1; else return 0; @@ -359,7 +380,7 @@ unsigned ffs( unsigned u ) { unsigned i; - if( u == 0 ) { + if (u == 0) { return 0; } @@ -373,6 +394,10 @@ unsigned ffs( unsigned u ) #define ffs __builtin_ffs #endif +#ifdef __MINGW32__ +#define ffs __builtin_ffs +#endif + /* Could also binary search for the highest bit. */ @@ -398,7 +423,10 @@ fui( float f ) } - +/** + * Convert ubyte to float in [0, 1]. + * XXX a 256-entry lookup table would be slightly faster. + */ static INLINE float ubyte_to_float(ubyte ub) { @@ -429,7 +457,23 @@ float_to_ubyte(float f) } +/** + * Calc log base 2 + */ +static INLINE unsigned +util_logbase2(unsigned n) +{ + unsigned log2 = 0; + while (n >>= 1) + ++log2; + return log2; +} + +/** + * Clamp X to [MIN, MAX]. + * This is a macro to allow float, int, uint, etc. types. + */ #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) @@ -442,6 +486,11 @@ align(int value, int alignment) return (value + alignment - 1) & ~(alignment - 1); } +static INLINE unsigned +minify(unsigned value) +{ + return MAX2(1, value >> 1); +} #ifndef COPY_4V #define COPY_4V( DST, SRC ) \ diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h index 0b18d043ad..c3f8c91833 100644 --- a/src/gallium/auxiliary/util/u_memory.h +++ b/src/gallium/auxiliary/util/u_memory.h @@ -100,8 +100,14 @@ ExFreePool(void *P); #define MALLOC( SIZE ) malloc( SIZE ) #define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) #define FREE( PTR ) free( PTR ) -#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) +static INLINE void * +_REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) +{ + (void) old_size; + return realloc(old_ptr, new_size); +} +#define REALLOC( a, b, c ) _REALLOC( a, b, c ) #endif diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c index 151a480d34..4b75d4ba1d 100644 --- a/src/gallium/auxiliary/util/u_mm.c +++ b/src/gallium/auxiliary/util/u_mm.c @@ -33,30 +33,32 @@ void u_mmDumpMemInfo(const struct mem_block *heap) { - debug_printf("Memory heap %p:\n", (void *)heap); + debug_printf("Memory heap %p:\n", (void *) heap); if (heap == 0) { debug_printf(" heap == 0\n"); - } else { + } + else { const struct mem_block *p; - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); + for (p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); } debug_printf("\nFree list:\n"); - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); + for (p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); } } debug_printf("End of memory blocks\n"); } + struct mem_block * u_mmInit(int ofs, int size) { diff --git a/src/gallium/auxiliary/util/u_mm.h b/src/gallium/auxiliary/util/u_mm.h index ce20e48763..6b158aae6e 100644 --- a/src/gallium/auxiliary/util/u_mm.h +++ b/src/gallium/auxiliary/util/u_mm.h @@ -84,7 +84,7 @@ extern struct mem_block *u_mmFindBlock(struct mem_block *heap, int start); extern void u_mmDestroy(struct mem_block *mmInit); /** - * For debuging purpose. + * For debugging purposes. */ extern void u_mmDumpMemInfo(const struct mem_block *mmInit); diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c index 07d804ecdb..bc4b758406 100644 --- a/src/gallium/auxiliary/util/u_network.c +++ b/src/gallium/auxiliary/util/u_network.c @@ -6,7 +6,7 @@ #if defined(PIPE_SUBSYSTEM_WINDOWS_USER) # include <winsock2.h> # include <windows.h> -#elif defined(PIPE_OS_LINUX) +#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) # include <sys/socket.h> # include <netinet/in.h> # include <unistd.h> @@ -54,7 +54,7 @@ u_socket_close(int s) if (s < 0) return; -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) shutdown(s, SHUT_RDWR); close(s); #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) @@ -169,7 +169,7 @@ u_socket_listen_on_port(uint16_t portnum) void u_socket_block(int s, boolean block) { -#if defined(PIPE_OS_LINUX) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) int old = fcntl(s, F_GETFL, 0); if (old == -1) return; diff --git a/src/gallium/auxiliary/util/u_network.h b/src/gallium/auxiliary/util/u_network.h index 14d3884427..8c778f492c 100644 --- a/src/gallium/auxiliary/util/u_network.h +++ b/src/gallium/auxiliary/util/u_network.h @@ -6,7 +6,7 @@ #if defined(PIPE_SUBSYSTEM_WINDOWS_USER) # define PIPE_HAVE_SOCKETS -#elif defined(PIPE_OS_LINUX) +#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) # define PIPE_HAVE_SOCKETS #endif diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 74259d453b..9866b6fc8a 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -43,7 +43,7 @@ * src_pitch may be negative to do vertical flip of pixels from source. */ void -pipe_copy_rect(ubyte * dst, +util_copy_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, @@ -91,7 +91,7 @@ pipe_copy_rect(ubyte * dst, } void -pipe_fill_rect(ubyte * dst, +util_fill_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, @@ -204,7 +204,7 @@ util_surface_copy(struct pipe_context *pipe, if (src_map && dst_map) { /* If do_flip, invert src_y position and pass negative src stride */ - pipe_copy_rect(dst_map, + util_copy_rect(dst_map, &dst_trans->block, dst_trans->stride, 0, 0, @@ -263,7 +263,7 @@ util_surface_fill(struct pipe_context *pipe, case 1: case 2: case 4: - pipe_fill_rect(dst_map, &dst_trans->block, dst_trans->stride, + util_fill_rect(dst_map, &dst_trans->block, dst_trans->stride, 0, 0, width, height, value); break; case 8: diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h index 59e842e16d..daa50834d3 100644 --- a/src/gallium/auxiliary/util/u_rect.h +++ b/src/gallium/auxiliary/util/u_rect.h @@ -42,13 +42,13 @@ struct pipe_surface; extern void -pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, +util_copy_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, const ubyte * src, int src_stride, unsigned src_x, int src_y); extern void -pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block, +util_fill_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, uint32_t value); diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c index 8114b53cd0..5238299015 100644 --- a/src/gallium/auxiliary/util/u_simple_screen.c +++ b/src/gallium/auxiliary/util/u_simple_screen.c @@ -52,8 +52,7 @@ pass_user_buffer_create(struct pipe_screen *screen, unsigned bytes) { struct pipe_buffer *buffer = - screen->winsys->user_buffer_create(screen->winsys, - ptr, bytes); + screen->winsys->user_buffer_create(screen->winsys, ptr, bytes); buffer->screen = screen; @@ -65,12 +64,12 @@ pass_surface_buffer_create(struct pipe_screen *screen, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { struct pipe_buffer *buffer = - screen->winsys->surface_buffer_create(screen->winsys, - width, height, - format, usage, stride); + screen->winsys->surface_buffer_create(screen->winsys, width, height, + format, usage, tex_usage, stride); buffer->screen = screen; @@ -82,8 +81,7 @@ pass_buffer_map(struct pipe_screen *screen, struct pipe_buffer *buf, unsigned usage) { - return screen->winsys->buffer_map(screen->winsys, - buf, usage); + return screen->winsys->buffer_map(screen->winsys, buf, usage); } static void @@ -105,8 +103,7 @@ pass_flush_frontbuffer(struct pipe_screen *screen, struct pipe_surface *surf, void *context_private) { - screen->winsys->flush_frontbuffer(screen->winsys, - surf, context_private); + screen->winsys->flush_frontbuffer(screen->winsys, surf, context_private); } static void @@ -114,8 +111,7 @@ pass_fence_reference(struct pipe_screen *screen, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence) { - screen->winsys->fence_reference(screen->winsys, - ptr, fence); + screen->winsys->fence_reference(screen->winsys, ptr, fence); } static int @@ -123,8 +119,7 @@ pass_fence_signalled(struct pipe_screen *screen, struct pipe_fence_handle *fence, unsigned flag) { - return screen->winsys->fence_signalled(screen->winsys, - fence, flag); + return screen->winsys->fence_signalled(screen->winsys, fence, flag); } static int @@ -132,11 +127,11 @@ pass_fence_finish(struct pipe_screen *screen, struct pipe_fence_handle *fence, unsigned flag) { - return screen->winsys->fence_finish(screen->winsys, - fence, flag); + return screen->winsys->fence_finish(screen->winsys, fence, flag); } -void u_simple_screen_init(struct pipe_screen *screen) +void +u_simple_screen_init(struct pipe_screen *screen) { screen->buffer_create = pass_buffer_create; screen->user_buffer_create = pass_user_buffer_create; @@ -151,7 +146,8 @@ void u_simple_screen_init(struct pipe_screen *screen) screen->fence_finish = pass_fence_finish; } -const char* u_simple_screen_winsys_name(struct pipe_screen *screen) +const char * +u_simple_screen_winsys_name(struct pipe_screen *screen) { return screen->winsys->get_name(screen->winsys); } diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index e519c354d2..0d706f9449 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -34,17 +34,9 @@ #include "pipe/p_context.h" -#include "util/u_debug.h" -#include "pipe/p_defines.h" -#include "pipe/p_screen.h" #include "pipe/p_shader_tokens.h" - -#include "util/u_memory.h" #include "util/u_simple_shaders.h" - -#include "tgsi/tgsi_build.h" -#include "tgsi/tgsi_dump.h" -#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_ureg.h" @@ -58,93 +50,29 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, const uint *semantic_indexes) { - struct pipe_shader_state shader; - struct tgsi_token tokens[100]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_VERTEX; - uint ti, i; - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); + struct ureg_program *ureg; + uint i; - ti = 3; + ureg = ureg_create( TGSI_PROCESSOR_VERTEX ); + if (ureg == NULL) + return NULL; - /* declare inputs */ for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = semantic_names[i]; - decl.Semantic.SemanticIndex = semantic_indexes[i]; - - decl.DeclarationRange.First = - decl.DeclarationRange.Last = i; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); + struct ureg_src src; + struct ureg_dst dst; + + src = ureg_DECL_vs_input( ureg, i ); + + dst = ureg_DECL_output( ureg, + semantic_names[i], + semantic_indexes[i]); + + ureg_MOV( ureg, dst, src ); } - /* declare outputs */ - for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = semantic_names[i]; - decl.Semantic.SemanticIndex = semantic_indexes[i]; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = i; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - } + ureg_END( ureg ); - /* emit MOV instructions */ - for (i = 0; i < num_attribs; i++) { - /* MOVE out[i], in[i]; */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = i; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = i; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - } - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - - return pipe->create_vs_state(pipe, &shader); + return ureg_create_shader_and_destroy( ureg, pipe ); } @@ -152,108 +80,46 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, /** * Make simple fragment texture shader: - * TEX OUT[0], IN[0], SAMP[0], 2D; + * IMM {0,0,0,1} // (if writemask != 0xf) + * MOV OUT[0], IMM[0] // (if writemask != 0xf) + * TEX OUT[0].writemask, IN[0], SAMP[0], 2D; * END; */ void * -util_make_fragment_tex_shader(struct pipe_context *pipe) +util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, + unsigned writemask ) { - struct pipe_shader_state shader; - struct tgsi_token tokens[100]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti; - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); + struct ureg_program *ureg; + struct ureg_src sampler; + struct ureg_src tex; + struct ureg_dst out; - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; + + sampler = ureg_DECL_sampler( ureg, 0 ); - ti = 3; + tex = ureg_DECL_fs_input( ureg, + TGSI_SEMANTIC_GENERIC, 0, + TGSI_INTERPOLATE_PERSPECTIVE ); - /* declare TEX[0] input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - /* XXX this could be linear... */ - decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); + out = ureg_DECL_output( ureg, + TGSI_SEMANTIC_COLOR, + 0 ); - /* declare color[0] output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); + ureg_TEX( ureg, out, TGSI_TEXTURE_2D, tex, sampler ); + ureg_END( ureg ); - /* declare sampler */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* TEX instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_TEX; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 2; - inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - inst.FullSrcRegisters[1].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - - return pipe->create_fs_state(pipe, &shader); + return ureg_create_shader_and_destroy( ureg, pipe ); } - +void * +util_make_fragment_tex_shader(struct pipe_context *pipe ) +{ + return util_make_fragment_tex_shader_writemask( pipe, + TGSI_WRITEMASK_XYZW ); +} @@ -263,87 +129,23 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) void * util_make_fragment_passthrough_shader(struct pipe_context *pipe) { - struct pipe_shader_state shader; - struct tgsi_token tokens[40]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti; - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* declare output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - - /* MOVE out[0], in[0]; */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); + struct ureg_program *ureg; + struct ureg_src src; + struct ureg_dst dst; - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; - assert(ti < Elements(tokens)); + src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0, + TGSI_INTERPOLATE_PERSPECTIVE ); -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif + dst = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 ); - shader.tokens = tokens; + ureg_MOV( ureg, dst, src ); + ureg_END( ureg ); - return pipe->create_fs_state(pipe, &shader); + return ureg_create_shader_and_destroy( ureg, pipe ); } diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index 6f8d96af9b..d2e80d6eb4 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -50,6 +50,10 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, extern void * +util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, + unsigned writemask ); + +extern void * util_make_fragment_tex_shader(struct pipe_context *pipe); diff --git a/src/gallium/auxiliary/util/u_stream_stdc.c b/src/gallium/auxiliary/util/u_stream_stdc.c index d8f648e5dd..5cd05b2904 100644 --- a/src/gallium/auxiliary/util/u_stream_stdc.c +++ b/src/gallium/auxiliary/util/u_stream_stdc.c @@ -32,7 +32,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU) #include <stdio.h> diff --git a/src/gallium/auxiliary/util/u_surface.h b/src/gallium/auxiliary/util/u_surface.h index a5b73cfc20..ce84ed7ad0 100644 --- a/src/gallium/auxiliary/util/u_surface.h +++ b/src/gallium/auxiliary/util/u_surface.h @@ -37,6 +37,23 @@ struct pipe_texture; struct pipe_surface; +/** + * Are s1 and s2 the same surface? + * Surfaces are basically views into textures so check if the two surfaces + * name the same part of the same texture. + */ +static INLINE boolean +util_same_surface(const struct pipe_surface *s1, const struct pipe_surface *s2) +{ + return (s1->texture == s2->texture && + s1->face == s2->face && + s1->level == s2->level && + s1->zslice == s2->zslice); +} + + + + extern boolean util_create_rgba_surface(struct pipe_screen *screen, uint width, uint height, diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index a0c8ed88f7..0d6489c26e 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -62,7 +62,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt, if(!src) return; - pipe_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y); + util_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y); screen->transfer_unmap(screen, pt); } @@ -90,7 +90,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt, if(!dst) return; - pipe_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0); + util_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0); screen->transfer_unmap(screen, pt); } @@ -153,7 +153,7 @@ a8r8g8b8_put_tile_rgba(unsigned *dst, } -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ +/*** PIPE_FORMAT_X8R8G8B8_UNORM ***/ static void x8r8g8b8_get_tile_rgba(const unsigned *src, @@ -170,7 +170,7 @@ x8r8g8b8_get_tile_rgba(const unsigned *src, pRow[0] = ubyte_to_float((pixel >> 16) & 0xff); pRow[1] = ubyte_to_float((pixel >> 8) & 0xff); pRow[2] = ubyte_to_float((pixel >> 0) & 0xff); - pRow[3] = ubyte_to_float(0xff); + pRow[3] = 1.0F; } p += dst_stride; } @@ -394,6 +394,52 @@ r5g6b5_put_tile_rgba(ushort *dst, +/*** PIPE_FORMAT_R8G8B8_UNORM ***/ + +static void +r8g8b8_get_tile_rgba(const ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = ubyte_to_float(src[0]); + pRow[1] = ubyte_to_float(src[1]); + pRow[2] = ubyte_to_float(src[2]); + pRow[3] = 1.0f; + src += 3; + } + p += dst_stride; + } +} + + +static void +r8g8b8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + dst[0] = float_to_ubyte(pRow[0]); + dst[1] = float_to_ubyte(pRow[1]); + dst[2] = float_to_ubyte(pRow[2]); + dst += 3; + } + p += src_stride; + } +} + + + /*** PIPE_FORMAT_Z16_UNORM ***/ /** @@ -596,6 +642,184 @@ r16g16b16a16_put_tile_rgba(short *dst, } +/*** PIPE_FORMAT_R8G8B8A8_SRGB ***/ + +/** + * Convert an 8-bit sRGB value from non-linear space to a + * linear RGB value in [0, 1]. + * Implemented with a 256-entry lookup table. + */ +static INLINE float +srgb_to_linear(ubyte cs8) +{ + static float table[256]; + static boolean tableReady = FALSE; + if (!tableReady) { + /* compute lookup table now */ + uint i; + for (i = 0; i < 256; i++) { + const float cs = ubyte_to_float(i); + if (cs <= 0.04045) { + table[i] = cs / 12.92f; + } + else { + table[i] = (float) powf((cs + 0.055) / 1.055, 2.4); + } + } + tableReady = TRUE; + } + return table[cs8]; +} + + +/** + * Convert linear float in [0,1] to an srgb ubyte value in [0,255]. + * XXX this hasn't been tested (render to srgb surface). + * XXX this needs optimization. + */ +static INLINE ubyte +linear_to_srgb(float cl) +{ + if (cl >= 1.0F) + return 255; + else if (cl >= 0.0031308F) + return float_to_ubyte(1.055F * powf(cl, 0.41666F) - 0.055F); + else if (cl > 0.0F) + return float_to_ubyte(12.92F * cl); + else + return 0.0; +} + + +static void +a8r8g8b8_srgb_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = srgb_to_linear((pixel >> 16) & 0xff); + pRow[1] = srgb_to_linear((pixel >> 8) & 0xff); + pRow[2] = srgb_to_linear((pixel >> 0) & 0xff); + pRow[3] = ubyte_to_float((pixel >> 24) & 0xff); + } + p += dst_stride; + } +} + +static void +a8r8g8b8_srgb_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + r = linear_to_srgb(pRow[0]); + g = linear_to_srgb(pRow[1]); + b = linear_to_srgb(pRow[2]); + a = float_to_ubyte(pRow[3]); + *dst++ = (a << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A8L8_SRGB ***/ + +static void +a8l8_srgb_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + ushort p = *src++; + pRow[0] = + pRow[1] = + pRow[2] = srgb_to_linear(p & 0xff); + pRow[3] = ubyte_to_float(p >> 8); + } + p += dst_stride; + } +} + +static void +a8l8_srgb_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, a; + r = linear_to_srgb(pRow[0]); + a = float_to_ubyte(pRow[3]); + *dst++ = (a << 8) | r; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_L8_SRGB ***/ + +static void +l8_srgb_get_tile_rgba(const ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = srgb_to_linear(*src); + pRow[3] = 1.0; + } + p += dst_stride; + } +} + +static void +l8_srgb_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r; + r = linear_to_srgb(pRow[0]); + *dst++ = (ubyte) r; + } + p += src_stride; + } +} + /*** PIPE_FORMAT_I8_UNORM ***/ @@ -928,6 +1152,9 @@ pipe_tile_raw_to_rgba(enum pipe_format format, case PIPE_FORMAT_R5G6B5_UNORM: r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; + case PIPE_FORMAT_R8G8B8_UNORM: + r8g8b8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); + break; case PIPE_FORMAT_L8_UNORM: l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); break; @@ -946,6 +1173,15 @@ pipe_tile_raw_to_rgba(enum pipe_format format, case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride); break; + case PIPE_FORMAT_A8R8G8B8_SRGB: + a8r8g8b8_srgb_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_A8L8_SRGB: + a8l8_srgb_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_L8_SRGB: + l8_srgb_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); + break; case PIPE_FORMAT_Z16_UNORM: z16_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; @@ -1035,6 +1271,9 @@ pipe_put_tile_rgba(struct pipe_transfer *pt, case PIPE_FORMAT_R5G6B5_UNORM: r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; + case PIPE_FORMAT_R8G8B8_UNORM: + r8g8b8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); + break; case PIPE_FORMAT_R8G8B8A8_UNORM: assert(0); break; @@ -1059,6 +1298,15 @@ pipe_put_tile_rgba(struct pipe_transfer *pt, case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); break; + case PIPE_FORMAT_A8R8G8B8_SRGB: + a8r8g8b8_srgb_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A8L8_SRGB: + a8l8_srgb_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_L8_SRGB: + l8_srgb_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); + break; case PIPE_FORMAT_Z16_UNORM: /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 5268cbf79c..b958a98635 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -35,7 +35,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) #include <sys/time.h> #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include <windows.h> @@ -77,7 +77,7 @@ util_time_get_frequency(void) void util_time_get(struct util_time *t) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) gettimeofday(&t->tv, NULL); #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) LONGLONG temp; @@ -102,7 +102,7 @@ util_time_add(const struct util_time *t1, int64_t usecs, struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -124,7 +124,7 @@ int64_t util_time_diff(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -144,7 +144,7 @@ util_time_micros( void ) util_time_get(&t1); -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) return t1.tv.tv_usec + t1.tv.tv_sec*1000000LL; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) util_time_get_frequency(); @@ -166,7 +166,7 @@ static INLINE int util_time_compare(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) if (t1->tv.tv_sec < t2->tv.tv_sec) return -1; else if(t1->tv.tv_sec > t2->tv.tv_sec) diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 6bca6077a2..a6189a247b 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -38,11 +38,16 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) #include <time.h> /* timeval */ #include <unistd.h> /* usleep */ #endif +#if defined(PIPE_OS_HAIKU) +#include <sys/time.h> /* timeval */ +#include <unistd.h> +#endif + #include "pipe/p_compiler.h" @@ -58,7 +63,7 @@ extern "C" { */ struct util_time { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) struct timeval tv; #else int64_t counter; @@ -89,7 +94,7 @@ util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr); -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) #define util_time_sleep usleep #else void diff --git a/src/gallium/auxiliary/util/u_timed_winsys.c b/src/gallium/auxiliary/util/u_timed_winsys.c index 77b2a3a1c8..178acdca4d 100644 --- a/src/gallium/auxiliary/util/u_timed_winsys.c +++ b/src/gallium/auxiliary/util/u_timed_winsys.c @@ -212,13 +212,14 @@ timed_surface_buffer_create(struct pipe_winsys *winsys, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { struct pipe_winsys *backend = timed_winsys(winsys)->backend; uint64_t start = time_start(); struct pipe_buffer *ret = backend->surface_buffer_create( backend, width, height, - format, usage, stride ); + format, usage, tex_usage, stride ); time_finish(winsys, start, 7, __FUNCTION__); diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c index c90425f3e5..eb635c9f14 100644 --- a/src/gallium/auxiliary/util/u_upload_mgr.c +++ b/src/gallium/auxiliary/util/u_upload_mgr.c @@ -123,21 +123,25 @@ static enum pipe_error u_upload_alloc_buffer( struct u_upload_mgr *upload, unsigned min_size ) { + unsigned size; + /* Release old buffer, if present: */ u_upload_flush( upload ); /* Allocate a new one: */ - upload->size = align(MAX2(upload->default_size, min_size), 4096); + size = align(MAX2(upload->default_size, min_size), 4096); upload->buffer = pipe_buffer_create( upload->screen, upload->alignment, upload->usage | PIPE_BUFFER_USAGE_CPU_WRITE, - upload->size ); + size ); if (upload->buffer == NULL) goto fail; + upload->size = size; + upload->offset = 0; return 0; |