From cbe5d8ece42a672f12b79eccf708428eddfcdcb4 Mon Sep 17 00:00:00 2001 From: Alistair Leslie-Hughes Date: Mon, 26 May 2025 07:03:34 +1000 Subject: [PATCH] Updated vkd3d to e69f3da089d472b723b016ef67c04c0f954fa792. --- dlls/msado15/tests/msado15.c | 2 +- libs/vkd3d/include/private/spirv_grammar.h | 561 +--- libs/vkd3d/include/private/vkd3d_common.h | 22 +- libs/vkd3d/include/private/vkd3d_version.h | 2 +- libs/vkd3d/include/vkd3d_shader.h | 38 +- libs/vkd3d/libs/vkd3d-common/blob.c | 1 + libs/vkd3d/libs/vkd3d-shader/d3d_asm.c | 64 +- libs/vkd3d/libs/vkd3d-shader/d3dbc.c | 43 +- libs/vkd3d/libs/vkd3d-shader/dxil.c | 690 ++--- libs/vkd3d/libs/vkd3d-shader/fx.c | 2 +- libs/vkd3d/libs/vkd3d-shader/glsl.c | 74 +- libs/vkd3d/libs/vkd3d-shader/hlsl.c | 82 +- libs/vkd3d/libs/vkd3d-shader/hlsl.h | 22 +- libs/vkd3d/libs/vkd3d-shader/hlsl.l | 2 +- libs/vkd3d/libs/vkd3d-shader/hlsl.y | 161 +- libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c | 936 +++++-- .../libs/vkd3d-shader/hlsl_constant_ops.c | 349 ++- libs/vkd3d/libs/vkd3d-shader/ir.c | 2320 +++++++++++------ libs/vkd3d/libs/vkd3d-shader/msl.c | 170 +- libs/vkd3d/libs/vkd3d-shader/preproc.l | 13 +- libs/vkd3d/libs/vkd3d-shader/spirv.c | 775 +++--- libs/vkd3d/libs/vkd3d-shader/tpf.c | 57 +- .../libs/vkd3d-shader/vkd3d_shader_main.c | 250 +- .../libs/vkd3d-shader/vkd3d_shader_private.h | 174 +- .../vkd3d/libs/vkd3d-utils/vkd3d_utils_main.c | 17 +- libs/vkd3d/libs/vkd3d/command.c | 92 +- libs/vkd3d/libs/vkd3d/device.c | 24 +- libs/vkd3d/libs/vkd3d/state.c | 6 +- libs/vkd3d/libs/vkd3d/utils.c | 3 +- libs/vkd3d/libs/vkd3d/vkd3d_private.h | 5 +- 30 files changed, 4257 insertions(+), 2700 deletions(-) diff --git a/dlls/msado15/tests/msado15.c b/dlls/msado15/tests/msado15.c index d81ab7da6cf..6529571a60e 100644 --- a/dlls/msado15/tests/msado15.c +++ b/dlls/msado15/tests/msado15.c @@ -2046,8 +2046,8 @@ START_TEST(msado15) setup_database(); test_Connection(); - test_Connection_Open(); test_ConnectionPoint(); + test_Connection_Open(); test_ADORecordsetConstruction(FALSE); test_ADORecordsetConstruction(TRUE); test_Fields(); diff --git a/libs/vkd3d/include/private/spirv_grammar.h b/libs/vkd3d/include/private/spirv_grammar.h index 34cadd9bd58..2aac5a6558c 100644 --- a/libs/vkd3d/include/private/spirv_grammar.h +++ b/libs/vkd3d/include/private/spirv_grammar.h @@ -43,12 +43,10 @@ enum spirv_parser_operand_type SPIRV_PARSER_OPERAND_TYPE_ADDRESSING_MODEL, SPIRV_PARSER_OPERAND_TYPE_BUILT_IN, SPIRV_PARSER_OPERAND_TYPE_CAPABILITY, - SPIRV_PARSER_OPERAND_TYPE_COMPONENT_TYPE, SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_LAYOUT, SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_OPERANDS, SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_REDUCE, SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_USE, - SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_VECTOR_MATRIX_LAYOUT, SPIRV_PARSER_OPERAND_TYPE_DECORATION, SPIRV_PARSER_OPERAND_TYPE_DIM, SPIRV_PARSER_OPERAND_TYPE_EXECUTION_MODE, @@ -84,7 +82,6 @@ enum spirv_parser_operand_type SPIRV_PARSER_OPERAND_TYPE_LITERAL_STRING, SPIRV_PARSER_OPERAND_TYPE_LOAD_CACHE_CONTROL, SPIRV_PARSER_OPERAND_TYPE_LOOP_CONTROL, - SPIRV_PARSER_OPERAND_TYPE_MATRIX_MULTIPLY_ACCUMULATE_OPERANDS, SPIRV_PARSER_OPERAND_TYPE_MEMORY_ACCESS, SPIRV_PARSER_OPERAND_TYPE_MEMORY_MODEL, SPIRV_PARSER_OPERAND_TYPE_MEMORY_SEMANTICS, @@ -149,7 +146,7 @@ spirv_parser_operand_type_info[] = }, [SPIRV_PARSER_OPERAND_TYPE_BUILT_IN] = { - "BuiltIn", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 126, + "BuiltIn", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 116, (struct spirv_parser_enumerant[]) { {0, "Position"}, @@ -210,9 +207,6 @@ spirv_parser_operand_type_info[] = {0x1156, "DeviceIndex"}, {0x1158, "ViewIndex"}, {0x115c, "ShadingRateKHR"}, - {0x118c, "TileOffsetQCOM"}, - {0x118d, "TileDimensionQCOM"}, - {0x118e, "TileApronSizeQCOM"}, {0x1380, "BaryCoordNoPerspAMD"}, {0x1381, "BaryCoordNoPerspCentroidAMD"}, {0x1382, "BaryCoordNoPerspSampleAMD"}, @@ -264,25 +258,18 @@ spirv_parser_operand_type_info[] = {0x14e0, "HitMicroTriangleVertexBarycentricsNV"}, {0x14e7, "IncomingRayFlagsKHR"}, {0x14e8, "RayGeometryIndexKHR"}, - {0x14ef, "HitIsSphereNV"}, - {0x14f0, "HitIsLSSNV"}, - {0x14f1, "HitSpherePositionNV"}, {0x14fe, "WarpsPerSMNV"}, {0x14ff, "SMCountNV"}, {0x1500, "WarpIDNV"}, {0x1501, "SMIDNV"}, - {0x1514, "HitLSSPositionsNV"}, {0x151d, "HitKindFrontFacingMicroTriangleNV"}, {0x151e, "HitKindBackFacingMicroTriangleNV"}, - {0x152c, "HitSphereRadiusNV"}, - {0x152d, "HitLSSRadiiNV"}, - {0x153c, "ClusterIDNV"}, {0x1785, "CullMaskKHR"}, } }, [SPIRV_PARSER_OPERAND_TYPE_CAPABILITY] = { - "Capability", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 261, + "Capability", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 245, (struct spirv_parser_enumerant[]) { {0, "Matrix"}, @@ -393,7 +380,6 @@ spirv_parser_operand_type_info[] = {0x1184, "TextureSampleWeightedQCOM"}, {0x1185, "TextureBoxFilterQCOM"}, {0x1186, "TextureBlockMatchQCOM"}, - {0x118f, "TileShadingQCOM"}, {0x1192, "TextureBlockMatch2QCOM"}, {0x1390, "Float16ImageAMD"}, {0x1391, "ImageGatherBiasLodAMD"}, @@ -404,9 +390,6 @@ spirv_parser_operand_type_info[] = {0x13bf, "ShaderClockKHR"}, {0x13cb, "ShaderEnqueueAMDX"}, {0x13df, "QuadControlKHR"}, - {0x13fc, "BFloat16TypeKHR"}, - {0x13fd, "BFloat16DotProductKHR"}, - {0x13fe, "BFloat16CooperativeMatrixKHR"}, {0x1481, "SampleMaskOverrideCoverageNV"}, {0x1483, "GeometryShaderPassthroughNV"}, {0x1486, "ShaderViewportIndexLayerEXT"}, @@ -452,19 +435,14 @@ spirv_parser_operand_type_info[] = {0x1507, "ShaderInvocationReorderNV"}, {0x150e, "BindlessTextureNV"}, {0x150f, "RayQueryPositionFetchKHR"}, - {0x1512, "CooperativeVectorNV"}, {0x151c, "AtomicFloat16VectorNV"}, {0x1521, "RayTracingDisplacementMicromapNV"}, {0x1526, "RawAccessChainsNV"}, - {0x152a, "RayTracingSpheresGeometryNV"}, - {0x152b, "RayTracingLinearSweptSpheresGeometryNV"}, {0x1536, "CooperativeMatrixReductionsNV"}, {0x1537, "CooperativeMatrixConversionsNV"}, {0x1538, "CooperativeMatrixPerElementOperationsNV"}, {0x1539, "CooperativeMatrixTensorAddressingNV"}, {0x153a, "CooperativeMatrixBlockLoadsNV"}, - {0x153b, "CooperativeVectorTrainingNV"}, - {0x153d, "RayTracingClusterAccelerationStructureNV"}, {0x153f, "TensorAddressingNV"}, {0x15c0, "SubgroupShuffleINTEL"}, {0x15c1, "SubgroupBufferBlockIOINTEL"}, @@ -529,47 +507,18 @@ spirv_parser_operand_type_info[] = {0x1800, "ArithmeticFenceEXT"}, {0x1806, "FPGAClusterAttributesV2INTEL"}, {0x1811, "FPGAKernelAttributesv2INTEL"}, - {0x1812, "TaskSequenceINTEL"}, {0x1819, "FPMaxErrorINTEL"}, {0x181b, "FPGALatencyControlINTEL"}, {0x181e, "FPGAArgumentInterfacesINTEL"}, {0x182b, "GlobalVariableHostAccessINTEL"}, {0x182d, "GlobalVariableFPGADecorationsINTEL"}, {0x184c, "SubgroupBufferPrefetchINTEL"}, - {0x1854, "Subgroup2DBlockIOINTEL"}, - {0x1855, "Subgroup2DBlockTransformINTEL"}, - {0x1856, "Subgroup2DBlockTransposeINTEL"}, - {0x185c, "SubgroupMatrixMultiplyAccumulateINTEL"}, - {0x1861, "TernaryBitwiseFunctionINTEL"}, {0x1900, "GroupUniformArithmeticKHR"}, - {0x1919, "TensorFloat32RoundingINTEL"}, {0x191b, "MaskedGatherScatterINTEL"}, {0x1929, "CacheControlsINTEL"}, {0x193c, "RegisterLimitsINTEL"}, } }, - [SPIRV_PARSER_OPERAND_TYPE_COMPONENT_TYPE] = - { - "ComponentType", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 15, - (struct spirv_parser_enumerant[]) - { - {0, "Float16NV"}, - {0x1, "Float32NV"}, - {0x2, "Float64NV"}, - {0x3, "SignedInt8NV"}, - {0x4, "SignedInt16NV"}, - {0x5, "SignedInt32NV"}, - {0x6, "SignedInt64NV"}, - {0x7, "UnsignedInt8NV"}, - {0x8, "UnsignedInt16NV"}, - {0x9, "UnsignedInt32NV"}, - {0xa, "UnsignedInt64NV"}, - {0x3ba247f8, "SignedInt8PackedNV"}, - {0x3ba247f9, "UnsignedInt8PackedNV"}, - {0x3ba247fa, "FloatE4M3NV"}, - {0x3ba247fb, "FloatE5M2NV"}, - } - }, [SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_LAYOUT] = { "CooperativeMatrixLayout", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 4, @@ -614,17 +563,6 @@ spirv_parser_operand_type_info[] = {0x2, "MatrixAccumulatorKHR"}, } }, - [SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_VECTOR_MATRIX_LAYOUT] = - { - "CooperativeVectorMatrixLayout", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 4, - (struct spirv_parser_enumerant[]) - { - {0, "RowMajorNV"}, - {0x1, "ColumnMajorNV"}, - {0x2, "InferencingOptimalNV"}, - {0x3, "TrainingOptimalNV"}, - } - }, [SPIRV_PARSER_OPERAND_TYPE_DECORATION] = { "Decoration", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 142, @@ -1240,7 +1178,7 @@ spirv_parser_operand_type_info[] = }, [SPIRV_PARSER_OPERAND_TYPE_EXECUTION_MODE] = { - "ExecutionMode", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 96, + "ExecutionMode", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 94, (struct spirv_parser_enumerant[]) { { @@ -1389,16 +1327,6 @@ spirv_parser_operand_type_info[] = SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER, } }, - {0x1189, "NonCoherentTileAttachmentReadQCOM"}, - { - 0x118a, "TileShadingRateQCOM", 3, - (enum spirv_parser_operand_type[]) - { - SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER, - SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER, - SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER, - } - }, {0x1399, "EarlyAndLateFragmentTestsAMD"}, {0x13a3, "StencilRefReplacingEXT"}, {0x13cd, "CoalescingAMDX"}, @@ -1628,11 +1556,7 @@ spirv_parser_operand_type_info[] = }, [SPIRV_PARSER_OPERAND_TYPE_FPENCODING] = { - "FPEncoding", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 1, - (struct spirv_parser_enumerant[]) - { - {0, "BFloat16KHR"}, - } + "FPEncoding", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM }, [SPIRV_PARSER_OPERAND_TYPE_FPFAST_MATH_MODE] = { @@ -1757,7 +1681,7 @@ spirv_parser_operand_type_info[] = }, [SPIRV_PARSER_OPERAND_TYPE_IMAGE_CHANNEL_DATA_TYPE] = { - "ImageChannelDataType", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 26, + "ImageChannelDataType", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 20, (struct spirv_parser_enumerant[]) { {0, "SnormInt8"}, @@ -1777,15 +1701,9 @@ spirv_parser_operand_type_info[] = {0xe, "Float"}, {0xf, "UnormInt24"}, {0x10, "UnormInt101010_2"}, - {0x11, "UnormInt10X6EXT"}, {0x13, "UnsignedIntRaw10EXT"}, {0x14, "UnsignedIntRaw12EXT"}, {0x15, "UnormInt2_101010EXT"}, - {0x16, "UnsignedInt10X6EXT"}, - {0x17, "UnsignedInt12X4EXT"}, - {0x18, "UnsignedInt14X2EXT"}, - {0x19, "UnormInt12X4EXT"}, - {0x1a, "UnormInt14X2EXT"}, } }, [SPIRV_PARSER_OPERAND_TYPE_IMAGE_CHANNEL_ORDER] = @@ -2146,28 +2064,6 @@ spirv_parser_operand_type_info[] = }, } }, - [SPIRV_PARSER_OPERAND_TYPE_MATRIX_MULTIPLY_ACCUMULATE_OPERANDS] = - { - "MatrixMultiplyAccumulateOperands", SPIRV_PARSER_OPERAND_CATEGORY_BIT_ENUM, 15, - (struct spirv_parser_enumerant[]) - { - {0, "None"}, - {0x1, "MatrixASignedComponentsINTEL"}, - {0x2, "MatrixBSignedComponentsINTEL"}, - {0x4, "MatrixCBFloat16INTEL"}, - {0x8, "MatrixResultBFloat16INTEL"}, - {0x10, "MatrixAPackedInt8INTEL"}, - {0x20, "MatrixBPackedInt8INTEL"}, - {0x40, "MatrixAPackedInt4INTEL"}, - {0x80, "MatrixBPackedInt4INTEL"}, - {0x100, "MatrixATF32INTEL"}, - {0x200, "MatrixBTF32INTEL"}, - {0x400, "MatrixAPackedFloat16INTEL"}, - {0x800, "MatrixBPackedFloat16INTEL"}, - {0x1000, "MatrixAPackedBFloat16INTEL"}, - {0x2000, "MatrixBPackedBFloat16INTEL"}, - } - }, [SPIRV_PARSER_OPERAND_TYPE_MEMORY_ACCESS] = { "MemoryAccess", SPIRV_PARSER_OPERAND_CATEGORY_BIT_ENUM, 9, @@ -2405,7 +2301,7 @@ spirv_parser_operand_type_info[] = }, [SPIRV_PARSER_OPERAND_TYPE_SOURCE_LANGUAGE] = { - "SourceLanguage", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 14, + "SourceLanguage", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 13, (struct spirv_parser_enumerant[]) { {0, "Unknown"}, @@ -2421,12 +2317,11 @@ spirv_parser_operand_type_info[] = {0xa, "WGSL"}, {0xb, "Slang"}, {0xc, "Zig"}, - {0xd, "Rust"}, } }, [SPIRV_PARSER_OPERAND_TYPE_STORAGE_CLASS] = { - "StorageClass", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 28, + "StorageClass", SPIRV_PARSER_OPERAND_CATEGORY_VALUE_ENUM, 27, (struct spirv_parser_enumerant[]) { {0, "UniformConstant"}, @@ -2443,7 +2338,6 @@ spirv_parser_operand_type_info[] = {0xb, "Image"}, {0xc, "StorageBuffer"}, {0x104c, "TileImageEXT"}, - {0x118b, "TileAttachmentQCOM"}, {0x13cc, "NodePayloadAMDX"}, {0x14d0, "CallableDataKHR"}, {0x14d1, "IncomingCallableDataKHR"}, @@ -6996,78 +6890,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_IMAGE_OPERANDS, '?'}, } }, - { - 0x14a8, "OpTypeCooperativeVectorNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x14a9, "OpCooperativeVectorMatrixMulNV", 13, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF, '?'}, - {SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_OPERANDS, '?'}, - } - }, - { - 0x14aa, "OpCooperativeVectorOuterProductAccumulateNV", 7, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF, '?'}, - } - }, - { - 0x14ab, "OpCooperativeVectorReduceSumAccumulateNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x14ac, "OpCooperativeVectorMatrixMulAddNV", 16, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF, '?'}, - {SPIRV_PARSER_OPERAND_TYPE_COOPERATIVE_MATRIX_OPERANDS, '?'}, - } - }, { 0x14ad, "OpCooperativeMatrixConvertNV", 3, (struct spirv_parser_instruction_operand[]) @@ -7138,27 +6960,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, } }, - { - 0x14b6, "OpCooperativeVectorLoadNV", 5, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_MEMORY_ACCESS, '?'}, - } - }, - { - 0x14b7, "OpCooperativeVectorStoreNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_MEMORY_ACCESS, '?'}, - } - }, { 0x14d6, "OpReportIntersectionKHR", 4, (struct spirv_parser_instruction_operand[]) @@ -7249,25 +7050,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, } }, - { - 0x14e1, "OpRayQueryGetClusterIdNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x14e2, "OpHitObjectGetClusterIdNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, { 0x14ee, "OpTypeCooperativeMatrixNV", 5, (struct spirv_parser_instruction_operand[]) @@ -7580,130 +7362,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_RAW_ACCESS_CHAIN_OPERANDS, '?'}, } }, - { - 0x1533, "OpRayQueryGetIntersectionSpherePositionNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1534, "OpRayQueryGetIntersectionSphereRadiusNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1535, "OpRayQueryGetIntersectionLSSPositionsNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1536, "OpRayQueryGetIntersectionLSSRadiiNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1537, "OpRayQueryGetIntersectionLSSHitValueNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1538, "OpHitObjectGetSpherePositionNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1539, "OpHitObjectGetSphereRadiusNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153a, "OpHitObjectGetLSSPositionsNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153b, "OpHitObjectGetLSSRadiiNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153c, "OpHitObjectIsSphereHitNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153d, "OpHitObjectIsLSSHitNV", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153e, "OpRayQueryIsSphereHitNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x153f, "OpRayQueryIsLSSHitNV", 4, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, { 0x15c3, "OpSubgroupShuffleINTEL", 4, (struct spirv_parser_instruction_operand[]) @@ -7962,9 +7620,10 @@ spirv_parser_opcode_info[] = } }, { - 0x15e9, "OpAsmTargetINTEL", 2, + 0x15e9, "OpAsmTargetINTEL", 3, (struct spirv_parser_instruction_operand[]) { + {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_STRING}, } @@ -9237,7 +8896,7 @@ spirv_parser_opcode_info[] = } }, { - 0x16d0, "OpArbitraryFloatSinCosPiINTEL", 8, + 0x16d0, "OpArbitraryFloatSinCosPiINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, @@ -9248,6 +8907,7 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, + {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, } }, { @@ -9279,7 +8939,7 @@ spirv_parser_opcode_info[] = } }, { - 0x16d3, "OpArbitraryFloatCastToIntINTEL", 8, + 0x16d3, "OpArbitraryFloatCastToIntINTEL", 7, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, @@ -9289,7 +8949,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, } }, { @@ -9803,7 +9462,7 @@ spirv_parser_opcode_info[] = } }, { - 0x16fa, "OpArbitraryFloatPowNINTEL", 10, + 0x16fa, "OpArbitraryFloatPowNINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, @@ -9815,7 +9474,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, } }, { @@ -9851,12 +9509,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1723, "OpFixedSqrtINTEL", 8, + 0x1723, "OpFixedSqrtINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9865,12 +9524,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1724, "OpFixedRecipINTEL", 8, + 0x1724, "OpFixedRecipINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9879,12 +9539,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1725, "OpFixedRsqrtINTEL", 8, + 0x1725, "OpFixedRsqrtINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9893,12 +9554,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1726, "OpFixedSinINTEL", 8, + 0x1726, "OpFixedSinINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9907,12 +9569,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1727, "OpFixedCosINTEL", 8, + 0x1727, "OpFixedCosINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9921,12 +9584,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1728, "OpFixedSinCosINTEL", 8, + 0x1728, "OpFixedSinCosINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9935,12 +9599,13 @@ spirv_parser_opcode_info[] = } }, { - 0x1729, "OpFixedSinPiINTEL", 8, + 0x1729, "OpFixedSinPiINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9949,12 +9614,13 @@ spirv_parser_opcode_info[] = } }, { - 0x172a, "OpFixedCosPiINTEL", 8, + 0x172a, "OpFixedCosPiINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9963,12 +9629,13 @@ spirv_parser_opcode_info[] = } }, { - 0x172b, "OpFixedSinCosPiINTEL", 8, + 0x172b, "OpFixedSinCosPiINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9977,12 +9644,13 @@ spirv_parser_opcode_info[] = } }, { - 0x172c, "OpFixedLogINTEL", 8, + 0x172c, "OpFixedLogINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -9991,12 +9659,13 @@ spirv_parser_opcode_info[] = } }, { - 0x172d, "OpFixedExpINTEL", 8, + 0x172d, "OpFixedExpINTEL", 9, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, @@ -10043,12 +9712,13 @@ spirv_parser_opcode_info[] = } }, { - 0x173d, "OpFPGARegINTEL", 3, + 0x173d, "OpFPGARegINTEL", 4, (struct spirv_parser_instruction_operand[]) { {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, + {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, } }, { @@ -10311,50 +9981,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, } }, - { - 0x1813, "OpTaskSequenceCreateINTEL", 7, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - {SPIRV_PARSER_OPERAND_TYPE_LITERAL_INTEGER}, - } - }, - { - 0x1814, "OpTaskSequenceAsyncINTEL", 2, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF, '*'}, - } - }, - { - 0x1815, "OpTaskSequenceGetINTEL", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1816, "OpTaskSequenceReleaseINTEL", 1, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1837, "OpTypeTaskSequenceINTEL", 1, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - } - }, { 0x184d, "OpSubgroupBlockPrefetchINTEL", 3, (struct spirv_parser_instruction_operand[]) @@ -10364,110 +9990,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_MEMORY_ACCESS, '?'}, } }, - { - 0x1857, "OpSubgroup2DBlockLoadINTEL", 10, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1858, "OpSubgroup2DBlockLoadTransformINTEL", 10, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x1859, "OpSubgroup2DBlockLoadTransposeINTEL", 10, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x185a, "OpSubgroup2DBlockPrefetchINTEL", 9, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x185b, "OpSubgroup2DBlockStoreINTEL", 10, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, - { - 0x185d, "OpSubgroupMatrixMultiplyAccumulateINTEL", 7, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_MATRIX_MULTIPLY_ACCUMULATE_OPERANDS, '?'}, - } - }, - { - 0x1862, "OpBitwiseFunctionINTEL", 6, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, { 0x1901, "OpGroupIMulKHR", 5, (struct spirv_parser_instruction_operand[]) @@ -10556,15 +10078,6 @@ spirv_parser_opcode_info[] = {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, } }, - { - 0x191a, "OpRoundFToTF32INTEL", 3, - (struct spirv_parser_instruction_operand[]) - { - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT_TYPE}, - {SPIRV_PARSER_OPERAND_TYPE_ID_RESULT}, - {SPIRV_PARSER_OPERAND_TYPE_ID_REF}, - } - }, { 0x191c, "OpMaskedGatherINTEL", 6, (struct spirv_parser_instruction_operand[]) diff --git a/libs/vkd3d/include/private/vkd3d_common.h b/libs/vkd3d/include/private/vkd3d_common.h index 08341304eea..c2e957a2fea 100644 --- a/libs/vkd3d/include/private/vkd3d_common.h +++ b/libs/vkd3d/include/private/vkd3d_common.h @@ -38,6 +38,8 @@ #include #endif +#define VKD3D_SHADER_API_VERSION_CURRENT VKD3D_SHADER_API_VERSION_1_17 + #ifndef ARRAY_SIZE # define ARRAY_SIZE(x) (sizeof(x) / sizeof(*(x))) #endif @@ -279,7 +281,7 @@ static inline unsigned int vkd3d_popcount(unsigned int v) { #ifdef _MSC_VER return __popcnt(v); -#elif defined(__MINGW32__) +#elif defined(HAVE_BUILTIN_POPCOUNT) return __builtin_popcount(v); #else v -= (v >> 1) & 0x55555555; @@ -341,6 +343,24 @@ static inline unsigned int vkd3d_log2i(unsigned int x) #endif } +static inline unsigned int vkd3d_ctz(uint32_t v) +{ +#ifdef HAVE_BUILTIN_CTZ + return __builtin_ctz(v); +#else + unsigned int c = 31; + + v &= -v; + c = (v & 0x0000ffff) ? c - 16 : c; + c = (v & 0x00ff00ff) ? c - 8 : c; + c = (v & 0x0f0f0f0f) ? c - 4 : c; + c = (v & 0x33333333) ? c - 2 : c; + c = (v & 0x55555555) ? c - 1 : c; + + return c; +#endif +} + static inline void *vkd3d_memmem( const void *haystack, size_t haystack_len, const void *needle, size_t needle_len) { const char *str = haystack; diff --git a/libs/vkd3d/include/private/vkd3d_version.h b/libs/vkd3d/include/private/vkd3d_version.h index 0edc4428022..687751d6a5f 100644 --- a/libs/vkd3d/include/private/vkd3d_version.h +++ b/libs/vkd3d/include/private/vkd3d_version.h @@ -1 +1 @@ -#define VKD3D_VCS_ID " (Wine bundled)" +#define VKD3D_VCS_ID " (git a8ca1f95)" diff --git a/libs/vkd3d/include/vkd3d_shader.h b/libs/vkd3d/include/vkd3d_shader.h index d82869e79ea..3a2f54c8f22 100644 --- a/libs/vkd3d/include/vkd3d_shader.h +++ b/libs/vkd3d/include/vkd3d_shader.h @@ -120,6 +120,11 @@ enum vkd3d_shader_structure_type * \since 1.15 */ VKD3D_SHADER_STRUCTURE_TYPE_SCAN_HULL_SHADER_TESSELLATION_INFO, + /** + * The structure is a vkd3d_shader_scan_thread_group_size_info structure. + * \since 1.18 + */ + VKD3D_SHADER_STRUCTURE_TYPE_SCAN_THREAD_GROUP_SIZE_INFO, VKD3D_FORCE_32_BIT_ENUM(VKD3D_SHADER_STRUCTURE_TYPE), }; @@ -417,10 +422,17 @@ struct vkd3d_shader_code { /** * Pointer to the code. Note that textual formats are not null-terminated. - * Therefore \a size should not include a null terminator, when this - * structure is passed as input to a vkd3d-shader function, and the - * allocated string will not include a null terminator when this structure - * is used as output. + * Therefore \a size should not include a null terminator when this + * structure is passed as input to a vkd3d-shader function, and \a size + * will not include a null terminator when this structure is used as + * output. + * + * For convenience, vkd3d_shader_preprocess() and vkd3d_shader_compile() + * will append a null terminator past the end of their output when + * outputting textual formats like VKD3D_SHADER_TARGET_D3D_ASM. This makes + * it safe to call functions like strlen() on \a code for such output, + * although doing so will obviously not account for any embedded null + * characters that may be present. */ const void *code; /** Size of \a code, in bytes. */ @@ -2282,6 +2294,24 @@ struct vkd3d_shader_scan_hull_shader_tessellation_info enum vkd3d_shader_tessellator_partitioning partitioning; }; +/** + * A chained structure describing the thread group size in a compute shader. + * + * This structure extends vkd3d_shader_compile_info. + * + * \since 1.18 + */ +struct vkd3d_shader_scan_thread_group_size_info +{ + /** Must be set to VKD3D_SHADER_STRUCTURE_TYPE_SCAN_THREAD_GROUP_SIZE_INFO. */ + enum vkd3d_shader_structure_type type; + /** Optional pointer to a structure containing further parameters. */ + const void *next; + + /** The thread group size in the x/y/z direction. */ + unsigned int x, y, z; +}; + /** * Data type of a shader varying, returned as part of struct * vkd3d_shader_signature_element. diff --git a/libs/vkd3d/libs/vkd3d-common/blob.c b/libs/vkd3d/libs/vkd3d-common/blob.c index f60ef7db769..c2c6ad67804 100644 --- a/libs/vkd3d/libs/vkd3d-common/blob.c +++ b/libs/vkd3d/libs/vkd3d-common/blob.c @@ -20,6 +20,7 @@ #define WIDL_C_INLINE_WRAPPERS #endif #define COBJMACROS + #define CONST_VTABLE #include "vkd3d.h" #include "vkd3d_blob.h" diff --git a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c index 6425a8f62d2..6f8fbe84b90 100644 --- a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c +++ b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c @@ -390,27 +390,10 @@ static void shader_print_resource_type(struct vkd3d_d3d_asm_compiler *compiler, static void shader_print_data_type(struct vkd3d_d3d_asm_compiler *compiler, enum vsir_data_type type) { - static const char *const data_type_names[] = - { - [VSIR_DATA_BOOL ] = "bool", - [VSIR_DATA_F16 ] = "half", - [VSIR_DATA_F32 ] = "float", - [VSIR_DATA_F64 ] = "double", - [VSIR_DATA_I32 ] = "int", - [VSIR_DATA_U8 ] = "uint8", - [VSIR_DATA_U16 ] = "uint16", - [VSIR_DATA_U32 ] = "uint", - [VSIR_DATA_U64 ] = "uint64", - [VSIR_DATA_SNORM ] = "snorm", - [VSIR_DATA_UNORM ] = "unorm", - [VSIR_DATA_OPAQUE ] = "opaque", - [VSIR_DATA_MIXED ] = "mixed", - [VSIR_DATA_CONTINUED] = "", - [VSIR_DATA_UNUSED ] = "", - }; + const char *name; - if (type < ARRAY_SIZE(data_type_names)) - vkd3d_string_buffer_printf(&compiler->buffer, "%s", data_type_names[type]); + if ((name = vsir_data_type_get_name(type, NULL))) + vkd3d_string_buffer_printf(&compiler->buffer, "%s", name); else vkd3d_string_buffer_printf(&compiler->buffer, "%s%s", compiler->colours.error, type, compiler->colours.reset); @@ -610,6 +593,18 @@ static void shader_print_uint_literal(struct vkd3d_d3d_asm_compiler *compiler, prefix, compiler->colours.literal, i, compiler->colours.reset, suffix); } +static void shader_print_int64_literal(struct vkd3d_d3d_asm_compiler *compiler, + const char *prefix, int64_t i, const char *suffix) +{ + /* Note that we need to handle INT64_MIN here as well. */ + if (i < 0) + vkd3d_string_buffer_printf(&compiler->buffer, "%s-%s%"PRIu64"%s%s", + prefix, compiler->colours.literal, -(uint64_t)i, compiler->colours.reset, suffix); + else + vkd3d_string_buffer_printf(&compiler->buffer, "%s%s%"PRId64"%s%s", + prefix, compiler->colours.literal, i, compiler->colours.reset, suffix); +} + static void shader_print_uint64_literal(struct vkd3d_d3d_asm_compiler *compiler, const char *prefix, uint64_t i, const char *suffix) { @@ -810,6 +805,12 @@ static void shader_print_register(struct vkd3d_d3d_asm_compiler *compiler, const if (reg->dimension == VSIR_DIMENSION_VEC4) shader_print_double_literal(compiler, ", ", reg->u.immconst_f64[1], ""); } + else if (reg->data_type == VSIR_DATA_I64) + { + shader_print_int64_literal(compiler, "", reg->u.immconst_u64[0], ""); + if (reg->dimension == VSIR_DIMENSION_VEC4) + shader_print_int64_literal(compiler, "", reg->u.immconst_u64[1], ""); + } else if (reg->data_type == VSIR_DATA_U64) { shader_print_uint64_literal(compiler, "", reg->u.immconst_u64[0], ""); @@ -851,7 +852,7 @@ static void shader_print_register(struct vkd3d_d3d_asm_compiler *compiler, const && reg->type != VKD3DSPR_NULL && reg->type != VKD3DSPR_DEPTHOUT) { - if (offset != ~0u) + if (reg->idx_count) { bool is_sm_5_1 = vkd3d_shader_ver_ge(&compiler->shader_version, 5, 1); @@ -879,10 +880,10 @@ static void shader_print_register(struct vkd3d_d3d_asm_compiler *compiler, const /* For descriptors in sm < 5.1 we move the reg->idx values up one slot * to normalise with 5.1. * Here we should ignore it if it's a descriptor in sm < 5.1. */ - if (reg->idx[1].offset != ~0u && (!is_descriptor || is_sm_5_1)) + if (reg->idx_count > 1 && (!is_descriptor || is_sm_5_1)) shader_print_subscript(compiler, reg->idx[1].offset, reg->idx[1].rel_addr); - if (reg->idx[2].offset != ~0u) + if (reg->idx_count > 2) shader_print_subscript(compiler, reg->idx[2].offset, reg->idx[2].rel_addr); } } @@ -974,6 +975,22 @@ static void shader_print_reg_type(struct vkd3d_d3d_asm_compiler *compiler, vkd3d_string_buffer_printf(buffer, ">%s", suffix); } +static void shader_print_indexable_temp_data_type(struct vkd3d_d3d_asm_compiler *compiler, + const struct vkd3d_shader_indexable_temp *t) +{ + struct vkd3d_string_buffer *buffer = &compiler->buffer; + + if (!(compiler->flags & VSIR_ASM_FLAG_DUMP_TYPES)) + return; + + if (t->component_count > 1) + vkd3d_string_buffer_printf(buffer, " component_count); + else + vkd3d_string_buffer_printf(buffer, " data_type); + vkd3d_string_buffer_printf(buffer, ">"); +} + static void shader_print_write_mask(struct vkd3d_d3d_asm_compiler *compiler, const char *prefix, uint32_t mask, const char *suffix) { @@ -1528,6 +1545,7 @@ static void shader_dump_instruction(struct vkd3d_d3d_asm_compiler *compiler, vkd3d_string_buffer_printf(buffer, " %sx%u%s", compiler->colours.reg, ins->declaration.indexable_temp.register_idx, compiler->colours.reset); shader_print_subscript(compiler, ins->declaration.indexable_temp.register_size, NULL); + shader_print_indexable_temp_data_type(compiler, &ins->declaration.indexable_temp); shader_print_uint_literal(compiler, ", ", ins->declaration.indexable_temp.component_count, ""); if (ins->declaration.indexable_temp.alignment) shader_print_uint_literal(compiler, ", align ", ins->declaration.indexable_temp.alignment, ""); diff --git a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c index 751e5578276..fbf9de20278 100644 --- a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c +++ b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c @@ -246,6 +246,7 @@ struct vkd3d_shader_sm1_parser bool abort; struct vkd3d_shader_parser p; + struct vsir_program *program; struct { @@ -468,7 +469,7 @@ static bool has_relative_address(uint32_t param) static const struct vkd3d_sm1_opcode_info *shader_sm1_get_opcode_info( const struct vkd3d_shader_sm1_parser *sm1, enum vkd3d_sm1_opcode opcode) { - const struct vkd3d_shader_version *version = &sm1->p.program->shader_version; + const struct vkd3d_shader_version *version = &sm1->program->shader_version; const struct vkd3d_sm1_opcode_info *info; unsigned int i = 0; @@ -542,9 +543,9 @@ static enum vkd3d_shader_register_type parse_register_type( } if (d3dbc_type == VKD3D_SM1_REG_ADDR) - return sm1->p.program->shader_version.type == VKD3D_SHADER_TYPE_PIXEL ? VKD3DSPR_TEXTURE : VKD3DSPR_ADDR; + return sm1->program->shader_version.type == VKD3D_SHADER_TYPE_PIXEL ? VKD3DSPR_TEXTURE : VKD3DSPR_ADDR; if (d3dbc_type == VKD3D_SM1_REG_TEXCRDOUT) - return vkd3d_shader_ver_ge(&sm1->p.program->shader_version, 3, 0) ? VKD3DSPR_OUTPUT : VKD3DSPR_TEXCRDOUT; + return vkd3d_shader_ver_ge(&sm1->program->shader_version, 3, 0) ? VKD3DSPR_OUTPUT : VKD3DSPR_TEXCRDOUT; for (unsigned int i = 0; i < ARRAY_SIZE(register_types); ++i) { @@ -660,7 +661,7 @@ static bool add_signature_element(struct vkd3d_shader_sm1_parser *sm1, bool outp const char *name, unsigned int index, enum vkd3d_shader_sysval_semantic sysval, unsigned int register_index, bool is_dcl, unsigned int mask) { - struct vsir_program *program = sm1->p.program; + struct vsir_program *program = sm1->program; struct shader_signature *signature; struct signature_element *element; @@ -702,7 +703,7 @@ static bool add_signature_element(struct vkd3d_shader_sm1_parser *sm1, bool outp static void add_signature_mask(struct vkd3d_shader_sm1_parser *sm1, bool output, unsigned int register_index, unsigned int mask) { - struct vsir_program *program = sm1->p.program; + struct vsir_program *program = sm1->program; struct shader_signature *signature; struct signature_element *element; @@ -749,7 +750,7 @@ static void add_signature_mask(struct vkd3d_shader_sm1_parser *sm1, bool output, static bool add_signature_element_from_register(struct vkd3d_shader_sm1_parser *sm1, const struct vkd3d_shader_register *reg, bool is_dcl, unsigned int mask) { - const struct vkd3d_shader_version *version = &sm1->p.program->shader_version; + const struct vkd3d_shader_version *version = &sm1->program->shader_version; unsigned int register_index = reg->idx_count > 0 ? reg->idx[0].offset : 0; switch (reg->type) @@ -844,7 +845,7 @@ static bool add_signature_element_from_register(struct vkd3d_shader_sm1_parser * static bool add_signature_element_from_semantic(struct vkd3d_shader_sm1_parser *sm1, const struct vkd3d_shader_semantic *semantic) { - const struct vkd3d_shader_version *version = &sm1->p.program->shader_version; + const struct vkd3d_shader_version *version = &sm1->program->shader_version; const struct vkd3d_shader_register *reg = &semantic->resource.reg.reg; enum vkd3d_shader_sysval_semantic sysval = VKD3D_SHADER_SV_NONE; unsigned int mask = semantic->resource.reg.write_mask; @@ -906,7 +907,7 @@ static void record_constant_register(struct vkd3d_shader_sm1_parser *sm1, static void shader_sm1_scan_register(struct vkd3d_shader_sm1_parser *sm1, const struct vkd3d_shader_register *reg, unsigned int mask, bool from_def) { - struct vsir_program *program = sm1->p.program; + struct vsir_program *program = sm1->program; uint32_t register_index = reg->idx[0].offset; switch (reg->type) @@ -955,7 +956,7 @@ static void shader_sm1_read_param(struct vkd3d_shader_sm1_parser *sm1, * VS >= 2.0 have relative addressing (with token) * VS >= 1.0 < 2.0 have relative addressing (without token) * The version check below should work in general. */ - if (sm1->p.program->shader_version.major < 2) + if (sm1->program->shader_version.major < 2) { *addr_token = (1u << 31) | ((VKD3DSPR_ADDR << VKD3D_SM1_REGISTER_TYPE_SHIFT2) & VKD3D_SM1_REGISTER_TYPE_MASK2) @@ -984,7 +985,7 @@ static void shader_sm1_skip_opcode(const struct vkd3d_shader_sm1_parser *sm1, co /* Version 2.0+ shaders may contain address tokens, but fortunately they * have a useful length mask - use it here. Version 1.x shaders contain no * such tokens. */ - if (sm1->p.program->shader_version.major >= 2) + if (sm1->program->shader_version.major >= 2) { length = (opcode_token & VKD3D_SM1_INSTRUCTION_LENGTH_MASK) >> VKD3D_SM1_INSTRUCTION_LENGTH_SHIFT; *ptr += length; @@ -1019,7 +1020,7 @@ static void shader_sm1_read_src_param(struct vkd3d_shader_sm1_parser *sm1, const shader_sm1_read_param(sm1, ptr, &token, &addr_token); if (has_relative_address(token)) { - if (!(src_rel_addr = vsir_program_get_src_params(sm1->p.program, 1))) + if (!(src_rel_addr = vsir_program_get_src_params(sm1->program, 1))) { vkd3d_shader_parser_error(&sm1->p, VKD3D_SHADER_ERROR_D3DBC_OUT_OF_MEMORY, "Out of memory."); @@ -1040,7 +1041,7 @@ static void shader_sm1_read_dst_param(struct vkd3d_shader_sm1_parser *sm1, const shader_sm1_read_param(sm1, ptr, &token, &addr_token); if (has_relative_address(token)) { - if (!(dst_rel_addr = vsir_program_get_src_params(sm1->p.program, 1))) + if (!(dst_rel_addr = vsir_program_get_src_params(sm1->program, 1))) { vkd3d_shader_parser_error(&sm1->p, VKD3D_SHADER_ERROR_D3DBC_OUT_OF_MEMORY, "Out of memory."); @@ -1052,9 +1053,9 @@ static void shader_sm1_read_dst_param(struct vkd3d_shader_sm1_parser *sm1, const shader_sm1_parse_dst_param(sm1, token, dst_rel_addr, dst_param); if (dst_param->reg.type == VKD3DSPR_RASTOUT && dst_param->reg.idx[0].offset == VSIR_RASTOUT_POINT_SIZE) - sm1->p.program->has_point_size = true; + sm1->program->has_point_size = true; if (dst_param->reg.type == VKD3DSPR_RASTOUT && dst_param->reg.idx[0].offset == VSIR_RASTOUT_FOG) - sm1->p.program->has_fog = true; + sm1->program->has_fog = true; } static void shader_sm1_read_semantic(struct vkd3d_shader_sm1_parser *sm1, @@ -1214,7 +1215,7 @@ static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, str { struct vkd3d_shader_src_param *src_params, *predicate; const struct vkd3d_sm1_opcode_info *opcode_info; - struct vsir_program *program = sm1->p.program; + struct vsir_program *program = sm1->program; unsigned int vsir_dst_count, vsir_src_count; struct vkd3d_shader_dst_param *dst_param; const uint32_t **ptr = &sm1->ptr; @@ -1446,7 +1447,8 @@ static enum vkd3d_result shader_sm1_init(struct vkd3d_shader_sm1_parser *sm1, st code_size != ~(size_t)0 ? token_count / 4u + 4 : 16, VSIR_CF_STRUCTURED, normalisation_level)) return VKD3D_ERROR_OUT_OF_MEMORY; - vkd3d_shader_parser_init(&sm1->p, program, message_context, compile_info->source_name); + vkd3d_shader_parser_init(&sm1->p, message_context, compile_info->source_name); + sm1->program = program; sm1->ptr = sm1->start; return VKD3D_OK; @@ -1504,8 +1506,8 @@ int d3dbc_parse(const struct vkd3d_shader_compile_info *compile_info, uint64_t c for (i = 0; i < ARRAY_SIZE(program->flat_constant_count); ++i) program->flat_constant_count[i] = get_external_constant_count(&sm1, i); - if (sm1.p.failed && ret >= 0) - ret = VKD3D_ERROR_INVALID_SHADER; + if (ret >= 0 && sm1.p.status < 0) + ret = sm1.p.status; if (ret < 0) { @@ -1952,7 +1954,7 @@ static void d3dbc_write_vsir_dcl(struct d3dbc_compiler *d3dbc, const struct vkd3 reg_id = semantic->resource.reg.reg.idx[0].offset; - if (semantic->resource.reg.reg.type != VKD3DSPR_SAMPLER) + if (semantic->resource.reg.reg.type != VKD3DSPR_COMBINED_SAMPLER) { vkd3d_shader_error(d3dbc->message_context, &ins->location, VKD3D_SHADER_ERROR_D3DBC_INVALID_REGISTER_TYPE, "dcl instruction with register type %u.", semantic->resource.reg.reg.type); @@ -2148,9 +2150,6 @@ int d3dbc_compile(struct vsir_program *program, uint64_t config_flags, if ((result = vsir_allocate_temp_registers(program, message_context))) return result; - if ((result = vsir_update_dcl_temps(program, message_context))) - return result; - d3dbc.program = program; d3dbc.message_context = message_context; switch (version->type) diff --git a/libs/vkd3d/libs/vkd3d-shader/dxil.c b/libs/vkd3d/libs/vkd3d-shader/dxil.c index c448e000cf9..f82644d4679 100644 --- a/libs/vkd3d/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d/libs/vkd3d-shader/dxil.c @@ -45,6 +45,8 @@ static const unsigned int MAX_GS_OUTPUT_STREAMS = 4; (VKD3D_SHADER_SWIZZLE_MASK << VKD3D_SHADER_SWIZZLE_SHIFT(0) \ | VKD3D_SHADER_SWIZZLE_MASK << VKD3D_SHADER_SWIZZLE_SHIFT(1)) +#define DXIL_TYPE_SIGNED 0x1u + enum bitcode_block_id { BLOCKINFO_BLOCK = 0, @@ -892,6 +894,8 @@ struct sm6_parser const uint32_t *ptr, *start, *end; unsigned int bitpos; + struct vsir_program *program; + struct dxil_block root_block; struct dxil_block *current_block; @@ -915,6 +919,7 @@ struct sm6_parser struct vkd3d_shader_dst_param *input_params; struct vkd3d_shader_dst_param *patch_constant_params; uint32_t io_regs_declared[VKD3D_BITMAP_SIZE(VKD3DSPR_COUNT)]; + struct vkd3d_shader_src_param *outpointid_param; struct sm6_function *functions; size_t function_count; @@ -976,7 +981,7 @@ static uint32_t sm6_parser_read_uint32(struct sm6_parser *sm6) { if (sm6_parser_is_end(sm6)) { - sm6->p.failed = true; + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return 0; } return *sm6->ptr++; @@ -994,7 +999,7 @@ static uint32_t sm6_parser_read_bits(struct sm6_parser *sm6, unsigned int length if (sm6_parser_is_end(sm6)) { - sm6->p.failed = true; + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return 0; } @@ -1006,7 +1011,7 @@ static uint32_t sm6_parser_read_bits(struct sm6_parser *sm6, unsigned int length ++sm6->ptr; if (sm6_parser_is_end(sm6) && l < length) { - sm6->p.failed = true; + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return bits; } sm6->bitpos = 0; @@ -1028,7 +1033,7 @@ static uint64_t sm6_parser_read_vbr(struct sm6_parser *sm6, unsigned int length) if (sm6_parser_is_end(sm6)) { - sm6->p.failed = true; + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return 0; } @@ -1039,9 +1044,10 @@ static uint64_t sm6_parser_read_vbr(struct sm6_parser *sm6, unsigned int length) bits = sm6_parser_read_bits(sm6, length); result |= (uint64_t)(bits & mask) << shift; shift += length - 1; - } while ((bits & flag) && !sm6->p.failed && shift < 64); + } while ((bits & flag) && (sm6->p.status >= 0) && shift < 64); - sm6->p.failed |= !!(bits & flag); + if (bits & flag) + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return result; } @@ -1053,7 +1059,7 @@ static void sm6_parser_align_32(struct sm6_parser *sm6) if (sm6_parser_is_end(sm6)) { - sm6->p.failed = true; + sm6->p.status = VKD3D_ERROR_INVALID_SHADER; return; } @@ -1144,8 +1150,8 @@ static enum vkd3d_result sm6_parser_read_unabbrev_record(struct sm6_parser *sm6) for (i = 0; i < count; ++i) record->operands[i] = sm6_parser_read_vbr(sm6, 6); - if (sm6->p.failed) - ret = VKD3D_ERROR_INVALID_SHADER; + if (sm6->p.status < 0) + ret = sm6->p.status; if (ret < 0 || (ret = dxil_block_add_record(block, record)) < 0) vkd3d_free(record); @@ -1156,25 +1162,25 @@ static enum vkd3d_result sm6_parser_read_unabbrev_record(struct sm6_parser *sm6) static bool sm6_parser_read_literal_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op) { *op = context; - return !sm6->p.failed; + return sm6->p.status >= 0; } static bool sm6_parser_read_fixed_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op) { *op = sm6_parser_read_bits(sm6, context); - return !sm6->p.failed; + return sm6->p.status >= 0; } static bool sm6_parser_read_vbr_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op) { *op = sm6_parser_read_vbr(sm6, context); - return !sm6->p.failed; + return sm6->p.status >= 0; } static bool sm6_parser_read_char6_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op) { *op = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789._"[sm6_parser_read_bits(sm6, 6)]; - return !sm6->p.failed; + return sm6->p.status >= 0; } static bool sm6_parser_read_blob_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op) @@ -1194,7 +1200,7 @@ static enum vkd3d_result dxil_abbrev_init(struct dxil_abbrev *abbrev, unsigned i abbrev->is_array = false; - for (i = 0, prev_type = 0; i < count && !sm6->p.failed; ++i) + for (i = 0, prev_type = 0; i < count && (sm6->p.status >= 0); ++i) { if (sm6_parser_read_bits(sm6, 1)) { @@ -1247,7 +1253,7 @@ static enum vkd3d_result dxil_abbrev_init(struct dxil_abbrev *abbrev, unsigned i abbrev->count = count; - return sm6->p.failed ? VKD3D_ERROR_INVALID_SHADER : VKD3D_OK; + return sm6->p.status; } static enum vkd3d_result sm6_parser_add_global_abbrev(struct sm6_parser *sm6) @@ -1459,7 +1465,7 @@ static enum vkd3d_result dxil_block_read(struct dxil_block *parent, struct sm6_p } break; } - } while (!sm6->p.failed); + } while (sm6->p.status >= 0); return VKD3D_ERROR_INVALID_SHADER; } @@ -1510,8 +1516,8 @@ static enum vkd3d_result dxil_block_init(struct dxil_block *block, const struct block->length = sm6_parser_read_uint32(sm6); block->start = sm6->ptr - sm6->start; - if (sm6->p.failed) - return VKD3D_ERROR_INVALID_SHADER; + if (sm6->p.status < 0) + return sm6->p.status; if ((block->abbrev_count = sm6_parser_compute_global_abbrev_count_for_block_id(sm6, block->id))) { @@ -2437,7 +2443,7 @@ static struct vkd3d_shader_src_param *instruction_src_params_alloc(struct vkd3d_ { struct vkd3d_shader_src_param *params; - if (!(params = vsir_program_get_src_params(sm6->p.program, count))) + if (!(params = vsir_program_get_src_params(sm6->program, count))) { ERR("Failed to allocate src params.\n"); vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, @@ -2454,7 +2460,7 @@ static struct vkd3d_shader_dst_param *instruction_dst_params_alloc(struct vkd3d_ { struct vkd3d_shader_dst_param *params; - if (!(params = vsir_program_get_dst_params(sm6->p.program, count))) + if (!(params = vsir_program_get_dst_params(sm6->program, count))) { ERR("Failed to allocate dst params.\n"); vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, @@ -2473,8 +2479,11 @@ static void register_init_with_id(struct vkd3d_shader_register *reg, reg->idx[0].offset = id; } -static enum vsir_data_type vsir_data_type_from_dxil(const struct sm6_type *type) +static enum vsir_data_type vsir_data_type_from_dxil(const struct sm6_type *type, + uint32_t type_flags, struct sm6_parser *dxil) { + bool is_signed = type_flags & DXIL_TYPE_SIGNED; + if (type->class == TYPE_CLASS_INTEGER) { switch (type->u.width) @@ -2482,15 +2491,16 @@ static enum vsir_data_type vsir_data_type_from_dxil(const struct sm6_type *type) case 1: return VSIR_DATA_BOOL; case 8: - return VSIR_DATA_U8; + return is_signed ? VSIR_DATA_I8 : VSIR_DATA_U8; case 16: - return VSIR_DATA_U16; + return is_signed ? VSIR_DATA_I16 : VSIR_DATA_U16; case 32: - return VSIR_DATA_U32; + return is_signed ? VSIR_DATA_I32 : VSIR_DATA_U32; case 64: - return VSIR_DATA_U64; + return is_signed ? VSIR_DATA_I64 : VSIR_DATA_U64; default: - FIXME("Unhandled width %u.\n", type->u.width); + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED, + "Unhandled integer width %u.", type->u.width); return VSIR_DATA_U32; } } @@ -2505,12 +2515,14 @@ static enum vsir_data_type vsir_data_type_from_dxil(const struct sm6_type *type) case 64: return VSIR_DATA_F64; default: - FIXME("Unhandled width %u.\n", type->u.width); + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED, + "Unhandled floating-point width %u.", type->u.width); return VSIR_DATA_F32; } } - FIXME("Unhandled type %u.\n", type->class); + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED, + "Unhandled type %#x.", type->class); return VSIR_DATA_U32; } @@ -2572,6 +2584,16 @@ static void register_convert_to_minimum_precision(struct vkd3d_shader_register * } break; + case VSIR_DATA_I16: + reg->data_type = VSIR_DATA_I32; + reg->precision = VKD3D_SHADER_REGISTER_PRECISION_MIN_INT_16; + if (reg->type == VKD3DSPR_IMMCONST) + { + for (i = 0; i < VSIR_DIMENSION_VEC4; ++i) + reg->u.immconst_u32[i] = (int16_t)reg->u.immconst_u32[i]; + } + break; + case VSIR_DATA_U16: reg->data_type = VSIR_DATA_U32; reg->precision = VKD3D_SHADER_REGISTER_PRECISION_MIN_UINT_16; @@ -2590,14 +2612,14 @@ static void register_convert_to_minimum_precision(struct vkd3d_shader_register * static void register_index_address_init(struct vkd3d_shader_register_index *idx, const struct sm6_value *address, struct sm6_parser *sm6); -static void sm6_register_from_value(struct vkd3d_shader_register *reg, const struct sm6_value *value, - struct sm6_parser *sm6) +static void vsir_register_from_dxil_value(struct vkd3d_shader_register *reg, + const struct sm6_value *value, uint32_t type_flags, struct sm6_parser *dxil) { const struct sm6_type *scalar_type; enum vsir_data_type data_type; scalar_type = sm6_type_get_scalar_type(value->type, 0); - data_type = vsir_data_type_from_dxil(scalar_type); + data_type = vsir_data_type_from_dxil(scalar_type, type_flags, dxil); switch (value->value_type) { @@ -2609,21 +2631,21 @@ static void sm6_register_from_value(struct vkd3d_shader_register *reg, const str case VALUE_TYPE_ICB: vsir_register_init(reg, VKD3DSPR_IMMCONSTBUFFER, data_type, 2); reg->idx[0].offset = value->u.icb.id; - register_index_address_init(®->idx[1], value->u.icb.index.index, sm6); + register_index_address_init(®->idx[1], value->u.icb.index.index, dxil); reg->idx[1].is_in_bounds = value->u.icb.index.is_in_bounds; break; case VALUE_TYPE_IDXTEMP: vsir_register_init(reg, VKD3DSPR_IDXTEMP, data_type, 2); reg->idx[0].offset = value->u.idxtemp.id; - register_index_address_init(®->idx[1], value->u.idxtemp.index.index, sm6); + register_index_address_init(®->idx[1], value->u.idxtemp.index.index, dxil); reg->idx[1].is_in_bounds = value->u.idxtemp.index.is_in_bounds; break; case VALUE_TYPE_GROUPSHAREDMEM: vsir_register_init(reg, VKD3DSPR_GROUPSHAREDMEM, data_type, 2); reg->idx[0].offset = value->u.groupsharedmem.id; - register_index_address_init(®->idx[1], value->u.groupsharedmem.index.index, sm6); + register_index_address_init(®->idx[1], value->u.groupsharedmem.index.index, dxil); reg->idx[1].is_in_bounds = value->u.groupsharedmem.index.is_in_bounds; break; @@ -2720,11 +2742,11 @@ static void src_param_init_vector(struct vkd3d_shader_src_param *param, unsigned param->modifiers = VKD3DSPSM_NONE; } -static void src_param_init_from_value(struct vkd3d_shader_src_param *param, const struct sm6_value *src, - struct sm6_parser *sm6) +static void src_param_init_from_value(struct vkd3d_shader_src_param *param, + const struct sm6_value *src, uint32_t type_flags, struct sm6_parser *dxil) { src_param_init(param); - sm6_register_from_value(¶m->reg, src, sm6); + vsir_register_from_dxil_value(¶m->reg, src, type_flags, dxil); } static void src_param_init_vector_from_reg(struct vkd3d_shader_src_param *param, @@ -2756,9 +2778,9 @@ static void register_index_address_init(struct vkd3d_shader_register_index *idx, } else { - struct vkd3d_shader_src_param *rel_addr = vsir_program_get_src_params(sm6->p.program, 1); + struct vkd3d_shader_src_param *rel_addr = vsir_program_get_src_params(sm6->program, 1); if (rel_addr) - src_param_init_from_value(rel_addr, address, sm6); + src_param_init_from_value(rel_addr, address, 0, sm6); idx->offset = 0; idx->rel_addr = rel_addr; } @@ -2783,17 +2805,19 @@ static void src_param_init_vector_from_handle(struct sm6_parser *sm6, src_param_init_vector_from_reg(param, ®); } -static bool instruction_dst_param_init_ssa_scalar(struct vkd3d_shader_instruction *ins, struct sm6_parser *sm6) +static bool instruction_dst_param_init_ssa_scalar(struct vkd3d_shader_instruction *ins, + uint32_t type_flags, struct sm6_parser *dxil) { - struct sm6_value *dst = sm6_parser_get_current_value(sm6); + struct sm6_value *dst = sm6_parser_get_current_value(dxil); struct vkd3d_shader_dst_param *param; - if (!(param = instruction_dst_params_alloc(ins, 1, sm6))) + if (!(param = instruction_dst_params_alloc(ins, 1, dxil))) return false; dst_param_init(param); - sm6_parser_init_ssa_value(sm6, dst); - sm6_register_from_value(¶m->reg, dst, sm6); + sm6_parser_init_ssa_value(dxil, dst); + vsir_register_from_dxil_value(¶m->reg, dst, type_flags, dxil); + return true; } @@ -2805,7 +2829,7 @@ static void instruction_dst_param_init_ssa_vector(struct vkd3d_shader_instructio dst_param_init_vector(param, component_count); sm6_parser_init_ssa_value(sm6, dst); - sm6_register_from_value(¶m->reg, dst, sm6); + vsir_register_from_dxil_value(¶m->reg, dst, 0, sm6); } static bool instruction_dst_param_init_uint_temp_vector(struct vkd3d_shader_instruction *ins, struct sm6_parser *sm6) @@ -2987,8 +3011,7 @@ static bool sm6_value_validate_is_backward_ref(const struct sm6_value *value, st { if (!value->is_back_ref) { - FIXME("Forward-referenced pointers are not supported.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Forward-referenced pointer declarations are not supported."); return false; } @@ -3200,8 +3223,7 @@ static enum vkd3d_result value_allocate_constant_array(struct sm6_value *dst, co /* Multidimensional arrays are emitted in flattened form. */ if (elem_type->class != TYPE_CLASS_INTEGER && elem_type->class != TYPE_CLASS_FLOAT) { - FIXME("Unhandled element type %u for data array.\n", elem_type->class); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "The element data type for an immediate constant buffer is not scalar integer or floating point."); return VKD3D_ERROR_INVALID_SHADER; } @@ -3224,7 +3246,7 @@ static enum vkd3d_result value_allocate_constant_array(struct sm6_value *dst, co "Out of memory allocating an immediate constant buffer of count %u.", count); return VKD3D_ERROR_OUT_OF_MEMORY; } - if (!shader_instruction_array_add_icb(&sm6->p.program->instructions, icb)) + if (!vsir_program_add_icb(sm6->program, icb)) { ERR("Failed to store icb object.\n"); vkd3d_free(icb); @@ -3237,7 +3259,7 @@ static enum vkd3d_result value_allocate_constant_array(struct sm6_value *dst, co dst->u.data = icb; icb->register_idx = sm6->icb_count++; - icb->data_type = vsir_data_type_from_dxil(elem_type); + icb->data_type = vsir_data_type_from_dxil(elem_type, 0, sm6); icb->element_count = type->u.array.count; icb->component_count = 1; icb->is_null = !operands; @@ -3254,6 +3276,12 @@ static enum vkd3d_result value_allocate_constant_array(struct sm6_value *dst, co icb->data_type = VSIR_DATA_F32; break; + case VSIR_DATA_I16: + for (i = 0; i < count; ++i) + icb->data[i] = (int16_t)operands[i]; + icb->data_type = VSIR_DATA_I32; + break; + case VSIR_DATA_U16: for (i = 0; i < count; ++i) icb->data[i] = (int16_t)operands[i]; @@ -3261,12 +3289,14 @@ static enum vkd3d_result value_allocate_constant_array(struct sm6_value *dst, co break; case VSIR_DATA_F32: + case VSIR_DATA_I32: case VSIR_DATA_U32: for (i = 0; i < count; ++i) icb->data[i] = operands[i]; break; case VSIR_DATA_F64: + case VSIR_DATA_I64: case VSIR_DATA_U64: data64 = (uint64_t *)icb->data; for (i = 0; i < count; ++i) @@ -3651,26 +3681,19 @@ static bool bitcode_parse_alignment(uint64_t encoded_alignment, unsigned int *al return true; } -static struct vkd3d_shader_instruction *sm6_parser_require_space(struct sm6_parser *sm6, size_t extra) +static struct vkd3d_shader_instruction *sm6_parser_add_instruction(struct sm6_parser *sm6, + enum vkd3d_shader_opcode op) { - struct vkd3d_shader_instruction_array *instructions = &sm6->p.program->instructions; + struct vkd3d_shader_instruction *ins; - if (!shader_instruction_array_reserve(instructions, instructions->count + extra)) + if (!(ins = vsir_program_append(sm6->program))) { - ERR("Failed to allocate instruction.\n"); + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, + "Out of memory allocating instruction."); return NULL; } - return &instructions->elements[instructions->count]; -} + vsir_instruction_init(ins, &sm6->p.location, op); -/* Space should be reserved before calling this. It is intended to require no checking of the returned pointer. */ -static struct vkd3d_shader_instruction *sm6_parser_add_instruction(struct sm6_parser *sm6, - enum vkd3d_shader_opcode handler_idx) -{ - struct vkd3d_shader_instruction *ins = sm6_parser_require_space(sm6, 1); - VKD3D_ASSERT(ins); - vsir_instruction_init(ins, &sm6->p.location, handler_idx); - ++sm6->p.program->instructions.count; return ins; } @@ -3679,7 +3702,8 @@ static void sm6_parser_declare_icb(struct sm6_parser *sm6, const struct sm6_type { struct vkd3d_shader_instruction *ins; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_IMMEDIATE_CONSTANT_BUFFER); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_IMMEDIATE_CONSTANT_BUFFER))) + return; /* The icb value index will be resolved later so forward references can be handled. */ ins->declaration.icb = (void *)(intptr_t)init; dst->value_type = VALUE_TYPE_ICB; @@ -3690,12 +3714,22 @@ static void sm6_parser_declare_indexable_temp(struct sm6_parser *sm6, const stru unsigned int count, unsigned int alignment, bool has_function_scope, unsigned int init, struct vkd3d_shader_instruction *ins, struct sm6_value *dst) { - enum vsir_data_type data_type = vsir_data_type_from_dxil(elem_type); + enum vsir_data_type data_type = vsir_data_type_from_dxil(elem_type, 0, sm6); + + if (!(sm6->program->global_flags & VKD3DSGF_FORCE_NATIVE_LOW_PRECISION)) + { + if (data_type == VSIR_DATA_F16) + data_type = VSIR_DATA_F32; + else if (data_type == VSIR_DATA_I16) + data_type = VSIR_DATA_I32; + else if (data_type == VSIR_DATA_U16) + data_type = VSIR_DATA_U32; + } if (ins) vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_DCL_INDEXABLE_TEMP); - else - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_INDEXABLE_TEMP); + else if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_INDEXABLE_TEMP))) + return; ins->declaration.indexable_temp.register_idx = sm6->indexable_temp_count++; ins->declaration.indexable_temp.register_size = count; ins->declaration.indexable_temp.alignment = alignment; @@ -3715,23 +3749,21 @@ static void sm6_parser_declare_tgsm_raw(struct sm6_parser *sm6, const struct sm6 struct vkd3d_shader_instruction *ins; unsigned int byte_count; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TGSM_RAW); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TGSM_RAW))) + return; dst_param_init(&ins->declaration.tgsm_raw.reg); dst->value_type = VALUE_TYPE_GROUPSHAREDMEM; dst->u.groupsharedmem.id = sm6->tgsm_count++; dst->structure_stride = 0; - sm6_register_from_value(&ins->declaration.tgsm_raw.reg.reg, dst, sm6); + vsir_register_from_dxil_value(&ins->declaration.tgsm_raw.reg.reg, dst, 0, sm6); ins->declaration.tgsm_raw.alignment = alignment; byte_count = elem_type->u.width / CHAR_BIT; /* Convert minimum precision types to their 32-bit equivalent. */ if (byte_count == 2) byte_count = 4; if (byte_count != 4) - { - FIXME("Unsupported byte count %u.\n", byte_count); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Raw TGSM byte count %u is not supported.", byte_count); - } ins->declaration.tgsm_raw.byte_count = byte_count; /* The initialiser value index will be resolved later when forward references can be handled. */ ins->flags = init; @@ -3742,7 +3774,8 @@ static void sm6_parser_declare_tgsm_structured(struct sm6_parser *sm6, const str { struct vkd3d_shader_instruction *ins; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TGSM_STRUCTURED); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TGSM_STRUCTURED))) + return; dst_param_init(&ins->declaration.tgsm_structured.reg); dst->value_type = VALUE_TYPE_GROUPSHAREDMEM; dst->u.groupsharedmem.id = sm6->tgsm_count++; @@ -3750,13 +3783,10 @@ static void sm6_parser_declare_tgsm_structured(struct sm6_parser *sm6, const str /* Convert minimum precision types to their 32-bit equivalent. */ if (dst->structure_stride == 2) dst->structure_stride = 4; - sm6_register_from_value(&ins->declaration.tgsm_structured.reg.reg, dst, sm6); + vsir_register_from_dxil_value(&ins->declaration.tgsm_structured.reg.reg, dst, 0, sm6); if (dst->structure_stride != 4) - { - FIXME("Unsupported structure stride %u.\n", dst->structure_stride); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Structured TGSM byte stride %u is not supported.", dst->structure_stride); - } ins->declaration.tgsm_structured.alignment = alignment; ins->declaration.tgsm_structured.byte_stride = dst->structure_stride; ins->declaration.tgsm_structured.structure_count = count; @@ -3781,8 +3811,7 @@ static bool sm6_parser_declare_global(struct sm6_parser *sm6, const struct dxil_ { if (!sm6_type_is_scalar(type->u.array.elem_type)) { - FIXME("Unsupported nested type class %u.\n", type->u.array.elem_type->class); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Global array variables with nested type class %u are not supported.", type->u.array.elem_type->class); return false; @@ -3797,8 +3826,7 @@ static bool sm6_parser_declare_global(struct sm6_parser *sm6, const struct dxil_ } else { - FIXME("Unsupported type class %u.\n", type->class); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Global variables of type class %u are not supported.", type->class); return false; } @@ -3903,14 +3931,13 @@ static bool sm6_parser_declare_global(struct sm6_parser *sm6, const struct dxil_ } else { - FIXME("Unhandled address space %"PRIu64".\n", address_space); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Global variables with address space %"PRIu64" are not supported.", address_space); return false; } ++sm6->value_count; - return true; + return (sm6->p.status >= 0); } static const struct vkd3d_shader_immediate_constant_buffer *resolve_forward_initialiser( @@ -3961,17 +3988,16 @@ static bool resolve_forward_zero_initialiser(size_t index, struct sm6_parser *sm return false; } - FIXME("Non-zero initialisers are not supported.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Non-zero TGSM initialisers are not supported."); return false; } static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6) { - struct vsir_program_iterator it = vsir_program_iterator(&sm6->p.program->instructions); - size_t i, count, base_value_idx = sm6->value_count; + struct vsir_program_iterator it = vsir_program_iterator(&sm6->program->instructions); const struct dxil_block *block = &sm6->root_block; + size_t i, base_value_idx = sm6->value_count; struct vkd3d_shader_instruction *ins; const struct dxil_record *record; enum vkd3d_result ret; @@ -3980,10 +4006,6 @@ static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6) sm6->p.location.line = block->id; sm6->p.location.column = 0; - for (i = 0, count = 0; i < block->record_count; ++i) - count += block->records[i]->code == MODULE_CODE_GLOBALVAR; - sm6_parser_require_space(sm6, count); - for (i = 0; i < block->record_count; ++i) { sm6->p.location.column = i; @@ -4001,7 +4023,10 @@ static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6) case MODULE_CODE_GLOBALVAR: if (!sm6_parser_declare_global(sm6, record)) - return VKD3D_ERROR_INVALID_SHADER; + { + VKD3D_ASSERT(sm6->p.status < 0); + return sm6->p.status; + } break; case MODULE_CODE_VERSION: @@ -4009,8 +4034,7 @@ static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6) return VKD3D_ERROR_INVALID_SHADER; if ((version = record->operands[0]) != 1) { - FIXME("Unsupported format version %#"PRIx64".\n", version); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED_BITCODE_FORMAT, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED, "Bitcode format version %#"PRIx64" is unsupported.", version); return VKD3D_ERROR_INVALID_SHADER; } @@ -4088,7 +4112,9 @@ static void src_params_init_from_operands(struct vkd3d_shader_src_param *src_par unsigned int i; for (i = 0; i < count; ++i) - src_param_init_from_value(&src_params[i], operands[i], sm6); + { + src_param_init_from_value(&src_params[i], operands[i], 0, sm6); + } } static enum vkd3d_shader_register_type register_type_from_dxil_semantic_kind( @@ -4125,7 +4151,7 @@ static enum vkd3d_shader_register_type register_type_from_dxil_semantic_kind( static void sm6_parser_init_signature(struct sm6_parser *sm6, const struct shader_signature *s, bool is_input, enum vkd3d_shader_register_type reg_type, struct vkd3d_shader_dst_param *params) { - enum vkd3d_shader_type shader_type = sm6->p.program->shader_version.type; + enum vkd3d_shader_type shader_type = sm6->program->shader_version.type; enum vkd3d_shader_register_type io_reg_type; bool is_patch_constant, is_control_point; struct vkd3d_shader_dst_param *param; @@ -4175,7 +4201,10 @@ static void sm6_parser_init_signature(struct sm6_parser *sm6, const struct shade if (is_control_point) { if (reg_type == VKD3DSPR_OUTPUT) - param->reg.idx[count].rel_addr = vsir_program_create_outpointid_param(sm6->p.program); + { + VKD3D_ASSERT(sm6->outpointid_param); + param->reg.idx[count].rel_addr = sm6->outpointid_param; + } param->reg.idx[count++].offset = 0; } @@ -4190,7 +4219,7 @@ static void sm6_parser_init_signature(struct sm6_parser *sm6, const struct shade static int sm6_parser_init_output_signature(struct sm6_parser *sm6, const struct shader_signature *output_signature) { - if (!(sm6->output_params = vsir_program_get_dst_params(sm6->p.program, output_signature->element_count))) + if (!(sm6->output_params = vsir_program_get_dst_params(sm6->program, output_signature->element_count))) { vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, "Failed to allocate output parameters."); @@ -4204,7 +4233,7 @@ static int sm6_parser_init_output_signature(struct sm6_parser *sm6, const struct static int sm6_parser_init_input_signature(struct sm6_parser *sm6, const struct shader_signature *input_signature) { - if (!(sm6->input_params = vsir_program_get_dst_params(sm6->p.program, input_signature->element_count))) + if (!(sm6->input_params = vsir_program_get_dst_params(sm6->program, input_signature->element_count))) { vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, "Failed to allocate input parameters."); @@ -4219,9 +4248,9 @@ static int sm6_parser_init_input_signature(struct sm6_parser *sm6, const struct static int sm6_parser_init_patch_constant_signature(struct sm6_parser *sm6, const struct shader_signature *patch_constant_signature) { - bool is_input = sm6->p.program->shader_version.type == VKD3D_SHADER_TYPE_DOMAIN; + bool is_input = sm6->program->shader_version.type == VKD3D_SHADER_TYPE_DOMAIN; - if (!(sm6->patch_constant_params = vsir_program_get_dst_params(sm6->p.program, + if (!(sm6->patch_constant_params = vsir_program_get_dst_params(sm6->program, patch_constant_signature->element_count))) { vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, @@ -4310,8 +4339,7 @@ static void sm6_parser_emit_alloca(struct sm6_parser *sm6, const struct dxil_rec WARN("Ignoring in_alloca flag.\n"); if (!(packed_operands & ALLOCA_FLAG_EXPLICIT_TYPE)) { - FIXME("Unhandled implicit type.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Implicit result type for ALLOCA instructions is not supported."); return; } @@ -4347,8 +4375,7 @@ static void sm6_parser_emit_alloca(struct sm6_parser *sm6, const struct dxil_rec /* A size of 1 means one instance of type[0], i.e. one array. */ if (sm6_value_get_constant_uint(size, sm6) != 1) { - FIXME("Allocation size is not 1.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "ALLOCA instruction allocation sizes other than 1 are not supported."); return; } @@ -4410,7 +4437,7 @@ static void sm6_parser_emit_atomicrmw(struct sm6_parser *sm6, const struct dxil_ || !sm6_value_validate_is_backward_ref(ptr, sm6)) return; - sm6_register_from_value(®, ptr, sm6); + vsir_register_from_dxil_value(®, ptr, 0, sm6); if (reg.type != VKD3DSPR_GROUPSHAREDMEM) { @@ -4430,8 +4457,7 @@ static void sm6_parser_emit_atomicrmw(struct sm6_parser *sm6, const struct dxil_ if ((op = map_dx_atomicrmw_op(code = record->operands[i++])) == VSIR_OP_INVALID) { - FIXME("Unhandled atomicrmw op %"PRIu64".\n", code); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Operation %"PRIu64" for an atomicrmw instruction is unhandled.", code); return; } @@ -4466,12 +4492,12 @@ static void sm6_parser_emit_atomicrmw(struct sm6_parser *sm6, const struct dxil_ src_param_init_vector_from_reg(&src_params[0], &coord); else src_param_make_constant_uint(&src_params[0], 0); - src_param_init_from_value(&src_params[1], src, sm6); + src_param_init_from_value(&src_params[1], src, 0, sm6); sm6_parser_init_ssa_value(sm6, dst); dst_params = instruction_dst_params_alloc(ins, 2, sm6); - sm6_register_from_value(&dst_params[0].reg, dst, sm6); + vsir_register_from_dxil_value(&dst_params[0].reg, dst, 0, sm6); dst_param_init(&dst_params[0]); dst_params[1].reg = reg; @@ -4559,8 +4585,7 @@ static enum vkd3d_shader_opcode map_binary_op(uint64_t code, const struct sm6_ty is_valid = is_int; break; default: - FIXME("Unhandled binary op %#"PRIx64".\n", code); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Binary operation %#"PRIx64" is unhandled.", code); return VSIR_OP_INVALID; } @@ -4581,6 +4606,7 @@ static void sm6_parser_emit_binop(struct sm6_parser *sm6, const struct dxil_reco struct vkd3d_shader_src_param *src_params; enum vkd3d_shader_opcode handler_idx; const struct sm6_value *a, *b; + uint32_t type_flags = 0; uint64_t code, flags; bool silence_warning; unsigned int i = 0; @@ -4622,6 +4648,8 @@ static void sm6_parser_emit_binop(struct sm6_parser *sm6, const struct dxil_reco silence_warning = !(flags & ~(OB_NO_UNSIGNED_WRAP | OB_NO_SIGNED_WRAP)); break; case VSIR_OP_ISHR: + type_flags |= DXIL_TYPE_SIGNED; + /* fall through */ case VSIR_OP_USHR: case VSIR_OP_IDIV: case VSIR_OP_UDIV_SIMPLE: @@ -4646,8 +4674,8 @@ static void sm6_parser_emit_binop(struct sm6_parser *sm6, const struct dxil_reco if (!(src_params = instruction_src_params_alloc(ins, 2, sm6))) return; - src_param_init_from_value(&src_params[0], a, sm6); - src_param_init_from_value(&src_params[1], b, sm6); + src_param_init_from_value(&src_params[0], a, type_flags, sm6); + src_param_init_from_value(&src_params[1], b, type_flags, sm6); if (code == BINOP_SUB) src_params[1].modifiers = VKD3DSPSM_NEG; @@ -4660,7 +4688,7 @@ static void sm6_parser_emit_binop(struct sm6_parser *sm6, const struct dxil_reco * do. */ ins->flags |= VKD3DSI_SHIFT_UNMASKED; } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, type_flags, sm6); } static const struct sm6_block *sm6_function_get_block(const struct sm6_function *function, uint64_t index, @@ -4710,7 +4738,7 @@ static void sm6_parser_emit_br(struct sm6_parser *sm6, const struct dxil_record dxil_record_validate_operand_max_count(record, i, sm6); code_block->terminator.type = TERMINATOR_COND_BR; - sm6_register_from_value(&code_block->terminator.conditional_reg, value, sm6); + vsir_register_from_dxil_value(&code_block->terminator.conditional_reg, value, 0, sm6); code_block->terminator.true_block = sm6_function_get_block(function, record->operands[0], sm6); code_block->terminator.false_block = sm6_function_get_block(function, record->operands[1], sm6); } @@ -4781,7 +4809,9 @@ static bool sm6_parser_emit_composite_construct(struct sm6_parser *sm6, const st unsigned int i; for (i = 0; i < component_count; ++i) - sm6_register_from_value(&operand_regs[i], operands[i], sm6); + { + vsir_register_from_dxil_value(&operand_regs[i], operands[i], 0, sm6); + } return sm6_parser_emit_reg_composite_construct(sm6, operand_regs, component_count, state, reg); } @@ -4797,11 +4827,11 @@ static bool sm6_parser_emit_coordinate_construct(struct sm6_parser *sm6, const s { if (!z_operand && operands[component_count]->value_type == VALUE_TYPE_UNDEFINED) break; - sm6_register_from_value(&operand_regs[component_count], operands[component_count], sm6); + vsir_register_from_dxil_value(&operand_regs[component_count], operands[component_count], 0, sm6); } if (z_operand) - sm6_register_from_value(&operand_regs[component_count++], z_operand, sm6); + vsir_register_from_dxil_value(&operand_regs[component_count++], z_operand, 0, sm6); return sm6_parser_emit_reg_composite_construct(sm6, operand_regs, component_count, state, reg); } @@ -4822,7 +4852,7 @@ static void sm6_parser_emit_dx_void(struct sm6_parser *sm6, enum dx_intrinsic_op { struct vkd3d_shader_instruction *ins = state->ins; vsir_instruction_init(ins, &sm6->p.location, sm6_dx_map_void_op(op)); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static enum vkd3d_shader_opcode map_dx_unary_op(enum dx_intrinsic_opcode op) @@ -4919,13 +4949,16 @@ static void sm6_parser_emit_dx_unary(struct sm6_parser *sm6, enum dx_intrinsic_o vsir_instruction_init(ins, &sm6->p.location, map_dx_unary_op(op)); if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } -static enum vkd3d_shader_opcode map_dx_binary_op(enum dx_intrinsic_opcode op, const struct sm6_type *type) +static enum vkd3d_shader_opcode map_dx_binary_op(enum dx_intrinsic_opcode op, + const struct sm6_type *type, uint32_t *type_flags) { + *type_flags = 0; + switch (op) { case DX_FMAX: @@ -4933,8 +4966,10 @@ static enum vkd3d_shader_opcode map_dx_binary_op(enum dx_intrinsic_opcode op, co case DX_FMIN: return type->u.width == 64 ? VSIR_OP_DMIN : VSIR_OP_MIN; case DX_IMAX: + *type_flags |= DXIL_TYPE_SIGNED; return VSIR_OP_IMAX; case DX_IMIN: + *type_flags |= DXIL_TYPE_SIGNED; return VSIR_OP_IMIN; case DX_QUAD_READ_LANE_AT: return VSIR_OP_QUAD_READ_LANE_AT; @@ -4954,14 +4989,15 @@ static void sm6_parser_emit_dx_binary(struct sm6_parser *sm6, enum dx_intrinsic_ { struct vkd3d_shader_instruction *ins = state->ins; struct vkd3d_shader_src_param *src_params; + uint32_t type_flags; - vsir_instruction_init(ins, &sm6->p.location, map_dx_binary_op(op, operands[0]->type)); + vsir_instruction_init(ins, &sm6->p.location, map_dx_binary_op(op, operands[0]->type, &type_flags)); if (!(src_params = instruction_src_params_alloc(ins, 2, sm6))) return; - src_param_init_from_value(&src_params[0], operands[0], sm6); - src_param_init_from_value(&src_params[1], operands[1], sm6); + src_param_init_from_value(&src_params[0], operands[0], type_flags, sm6); + src_param_init_from_value(&src_params[1], operands[1], type_flags, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, type_flags, sm6); } static enum vkd3d_shader_opcode map_dx_atomic_binop(const struct sm6_value *operand, struct sm6_parser *sm6) @@ -4990,8 +5026,7 @@ static enum vkd3d_shader_opcode map_dx_atomic_binop(const struct sm6_value *oper return VSIR_OP_IMM_ATOMIC_XOR; /* DXIL currently doesn't use SUB and NAND. */ default: - FIXME("Unhandled atomic binop %"PRIu64".\n", code); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "Operation %"PRIu64" for an atomic binop instruction is unhandled.", code); return VSIR_OP_INVALID; } @@ -5030,7 +5065,7 @@ static void sm6_parser_emit_dx_atomic_binop(struct sm6_parser *sm6, enum dx_intr } else { - sm6_register_from_value(®, operands[coord_idx], sm6); + vsir_register_from_dxil_value(®, operands[coord_idx], 0, sm6); } for (i = coord_idx + coord_count; i < coord_idx + 3; ++i) @@ -5051,14 +5086,14 @@ static void sm6_parser_emit_dx_atomic_binop(struct sm6_parser *sm6, enum dx_intr return; src_param_init_vector_from_reg(&src_params[0], ®); if (is_cmp_xchg) - src_param_init_from_value(&src_params[1], operands[4], sm6); - src_param_init_from_value(&src_params[1 + is_cmp_xchg], operands[5], sm6); + src_param_init_from_value(&src_params[1], operands[4], 0, sm6); + src_param_init_from_value(&src_params[1 + is_cmp_xchg], operands[5], 0, sm6); sm6_parser_init_ssa_value(sm6, dst); dst_params = instruction_dst_params_alloc(ins, 2, sm6); dst_param_init(&dst_params[0]); - sm6_register_from_value(&dst_params[0].reg, dst, sm6); + vsir_register_from_dxil_value(&dst_params[0].reg, dst, 0, sm6); dst_param_init(&dst_params[1]); sm6_register_from_handle(sm6, &resource->u.handle, &dst_params[1].reg); } @@ -5099,8 +5134,7 @@ static void sm6_parser_emit_dx_buffer_update_counter(struct sm6_parser *sm6, enu if (!sm6_value_is_constant(operands[1])) { - FIXME("Unsupported dynamic update operand.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED, "A dynamic update value for a UAV counter operation is not supported."); return; } @@ -5118,7 +5152,7 @@ static void sm6_parser_emit_dx_buffer_update_counter(struct sm6_parser *sm6, enu return; src_param_init_vector_from_handle(sm6, &src_params[0], &resource->u.handle); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_calculate_lod(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5152,7 +5186,7 @@ static void sm6_parser_emit_dx_calculate_lod(struct sm6_parser *sm6, enum dx_int src_param_init_scalar(&src_params[1], !clamp); src_param_init_vector_from_handle(sm6, &src_params[2], &sampler->u.handle); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_cbuffer_load(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5180,7 +5214,7 @@ static void sm6_parser_emit_dx_cbuffer_load(struct sm6_parser *sm6, enum dx_intr type = sm6_type_get_scalar_type(dst->type, 0); VKD3D_ASSERT(type); - src_param->reg.data_type = vsir_data_type_from_dxil(type); + src_param->reg.data_type = vsir_data_type_from_dxil(type, 0, sm6); if (data_type_is_64_bit(src_param->reg.data_type)) src_param->swizzle = vsir_swizzle_64_from_32(src_param->swizzle); else @@ -5198,7 +5232,8 @@ static void sm6_parser_dcl_register_builtin(struct sm6_parser *sm6, enum vkd3d_s if (!bitmap_is_set(sm6->io_regs_declared, reg_type)) { bitmap_set(sm6->io_regs_declared, reg_type); - ins = sm6_parser_add_instruction(sm6, handler_idx); + if (!(ins = sm6_parser_add_instruction(sm6, handler_idx))) + return; dst_param = &ins->declaration.dst; vsir_register_init(&dst_param->reg, reg_type, data_type, 0); dst_param_init_vector(dst_param, component_count); @@ -5220,7 +5255,7 @@ static void sm6_parser_emit_dx_input_register_mov(struct sm6_parser *sm6, struct src_param->reg.dimension = VSIR_DIMENSION_VEC4; src_param_init(src_param); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_coverage(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5323,7 +5358,7 @@ static void sm6_parser_emit_dx_discard(struct sm6_parser *sm6, enum dx_intrinsic vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_DISCARD); if ((src_param = instruction_src_params_alloc(ins, 1, sm6))) - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); } static void sm6_parser_emit_dx_domain_location(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5350,7 +5385,7 @@ static void sm6_parser_emit_dx_domain_location(struct sm6_parser *sm6, enum dx_i src_param->reg.dimension = VSIR_DIMENSION_VEC4; src_param_init_scalar(src_param, component_idx); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_dot(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5392,7 +5427,7 @@ static void sm6_parser_emit_dx_dot(struct sm6_parser *sm6, enum dx_intrinsic_opc src_param_init_vector_from_reg(&src_params[0], ®s[0]); src_param_init_vector_from_reg(&src_params[1], ®s[1]); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_eval_attrib(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5407,7 +5442,7 @@ static void sm6_parser_emit_dx_eval_attrib(struct sm6_parser *sm6, enum dx_intri row_index = sm6_value_get_constant_uint(operands[0], sm6); column_index = sm6_value_get_constant_uint(operands[2], sm6); - signature = &sm6->p.program->input_signature; + signature = &sm6->program->input_signature; if (row_index >= signature->element_count) { WARN("Invalid row index %u.\n", row_index); @@ -5437,9 +5472,9 @@ static void sm6_parser_emit_dx_eval_attrib(struct sm6_parser *sm6, enum dx_intri register_index_address_init(&src_params[0].reg.idx[0], operands[1], sm6); if (op == DX_EVAL_SAMPLE_INDEX) - src_param_init_from_value(&src_params[1], operands[3], sm6); + src_param_init_from_value(&src_params[1], operands[3], 0, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_fabs(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5451,10 +5486,10 @@ static void sm6_parser_emit_dx_fabs(struct sm6_parser *sm6, enum dx_intrinsic_op vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_MOV); if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); src_param->modifiers = VKD3DSPSM_ABS; - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_compute_builtin(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5494,7 +5529,7 @@ static void sm6_parser_emit_dx_compute_builtin(struct sm6_parser *sm6, enum dx_i component_idx = sm6_value_get_constant_uint(operands[0], sm6); src_param_init_scalar(src_param, component_idx); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static enum vkd3d_shader_opcode sm6_dx_map_ma_op(enum dx_intrinsic_opcode op, const struct sm6_type *type) @@ -5524,9 +5559,11 @@ static void sm6_parser_emit_dx_ma(struct sm6_parser *sm6, enum dx_intrinsic_opco if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) return; for (i = 0; i < 3; ++i) - src_param_init_from_value(&src_params[i], operands[i], sm6); + { + src_param_init_from_value(&src_params[i], operands[i], 0, sm6); + } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_get_dimensions(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5554,7 +5591,7 @@ static void sm6_parser_emit_dx_get_dimensions(struct sm6_parser *sm6, enum dx_in if (is_texture) { ins->flags = VKD3DSI_RESINFO_UINT; - src_param_init_from_value(&src_params[0], operands[1], sm6); + src_param_init_from_value(&src_params[0], operands[1], 0, sm6); component_count = VKD3D_VEC4_SIZE; if (resource_kind_is_multisampled(resource_kind)) @@ -5627,9 +5664,11 @@ static void sm6_parser_emit_dx_tertiary(struct sm6_parser *sm6, enum dx_intrinsi if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) return; for (i = 0; i < 3; ++i) - src_param_init_from_value(&src_params[i], operands[i], sm6); + { + src_param_init_from_value(&src_params[i], operands[i], 0, sm6); + } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_load_input(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5638,7 +5677,7 @@ static void sm6_parser_emit_dx_load_input(struct sm6_parser *sm6, enum dx_intrin bool is_control_point = op == DX_LOAD_OUTPUT_CONTROL_POINT; bool is_patch_constant = op == DX_LOAD_PATCH_CONSTANT; struct vkd3d_shader_instruction *ins = state->ins; - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; unsigned int count, row_index, column_index; const struct vkd3d_shader_dst_param *params; struct vkd3d_shader_src_param *src_param; @@ -5697,7 +5736,7 @@ static void sm6_parser_emit_dx_load_input(struct sm6_parser *sm6, enum dx_intrin register_index_address_init(&src_param->reg.idx[count], operands[3], sm6); } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_make_double(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5717,7 +5756,7 @@ static void sm6_parser_emit_dx_make_double(struct sm6_parser *sm6, enum dx_intri src_params[0].reg = reg; src_param_init_vector(&src_params[0], 2); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_output_control_point_id(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5768,9 +5807,9 @@ static void sm6_parser_emit_dx_quad_op(struct sm6_parser *sm6, enum dx_intrinsic if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_raw_buffer_load(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -5915,7 +5954,7 @@ static void sm6_parser_emit_dx_buffer_load(struct sm6_parser *sm6, enum dx_intri if (!(src_params = instruction_src_params_alloc(ins, 2, sm6))) return; - src_param_init_from_value(&src_params[0], operands[1], sm6); + src_param_init_from_value(&src_params[0], operands[1], 0, sm6); if (!sm6_value_is_undef(operands[2])) { /* Constant zero would be ok, but is not worth checking for unless it shows up. */ @@ -5980,7 +6019,7 @@ static void sm6_parser_emit_dx_buffer_store(struct sm6_parser *sm6, enum dx_intr if (!(src_params = instruction_src_params_alloc(ins, 2, sm6))) return; - src_param_init_from_value(&src_params[0], operands[1], sm6); + src_param_init_from_value(&src_params[0], operands[1], 0, sm6); if (!sm6_value_is_undef(operands[2])) { /* Constant zero would have no effect, but is not worth checking for unless it shows up. */ @@ -6010,7 +6049,7 @@ static void sm6_parser_emit_dx_get_sample_count(struct sm6_parser *sm6, enum dx_ src_param->reg.dimension = VSIR_DIMENSION_VEC4; src_param_init(src_param); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); ins->dst->reg.data_type = VSIR_DATA_U32; } @@ -6035,14 +6074,14 @@ static void sm6_parser_emit_dx_get_sample_pos(struct sm6_parser *sm6, enum dx_in if (op == DX_TEX2DMS_GET_SAMPLE_POS) { src_param_init_vector_from_handle(sm6, &src_params[0], &resource->u.handle); - src_param_init_from_value(&src_params[1], operands[1], sm6); + src_param_init_from_value(&src_params[1], operands[1], 0, sm6); } else { src_param_init_vector(&src_params[0], 2); vsir_register_init(&src_params[0].reg, VKD3DSPR_RASTERIZER, VSIR_DATA_F32, 0); src_params[0].reg.dimension = VSIR_DIMENSION_VEC4; - src_param_init_from_value(&src_params[1], operands[0], sm6); + src_param_init_from_value(&src_params[1], operands[0], 0, sm6); } instruction_dst_param_init_ssa_vector(ins, 2, sm6); @@ -6104,7 +6143,7 @@ static void sm6_parser_emit_dx_sample(struct sm6_parser *sm6, enum dx_intrinsic_ instruction_init_with_resource(ins, (op == DX_SAMPLE_B) ? VSIR_OP_SAMPLE_B : VSIR_OP_SAMPLE_LOD, resource, sm6); src_params = instruction_src_params_alloc(ins, 4, sm6); - src_param_init_from_value(&src_params[3], operands[9], sm6); + src_param_init_from_value(&src_params[3], operands[9], 0, sm6); break; case DX_SAMPLE_C: clamp_idx = 10; @@ -6113,7 +6152,7 @@ static void sm6_parser_emit_dx_sample(struct sm6_parser *sm6, enum dx_intrinsic_ instruction_init_with_resource(ins, (op == DX_SAMPLE_C_LZ) ? VSIR_OP_SAMPLE_C_LZ : VSIR_OP_SAMPLE_C, resource, sm6); src_params = instruction_src_params_alloc(ins, 4, sm6); - src_param_init_from_value(&src_params[3], operands[9], sm6); + src_param_init_from_value(&src_params[3], operands[9], 0, sm6); component_count = 1; break; case DX_SAMPLE_GRAD: @@ -6148,7 +6187,7 @@ static void sm6_parser_emit_dx_sample(struct sm6_parser *sm6, enum dx_intrinsic_ static void sm6_parser_emit_dx_sample_index(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, const struct sm6_value **operands, struct function_emission_state *state) { - const struct shader_signature *signature = &sm6->p.program->input_signature; + const struct shader_signature *signature = &sm6->program->input_signature; struct vkd3d_shader_instruction *ins = state->ins; struct vkd3d_shader_src_param *src_param; unsigned int element_idx; @@ -6170,7 +6209,7 @@ static void sm6_parser_emit_dx_sample_index(struct sm6_parser *sm6, enum dx_intr src_param->reg = sm6->input_params[element_idx].reg; src_param_init(src_param); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_saturate(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -6182,9 +6221,9 @@ static void sm6_parser_emit_dx_saturate(struct sm6_parser *sm6, enum dx_intrinsi vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_MOV); if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); - if (instruction_dst_param_init_ssa_scalar(ins, sm6)) + if (instruction_dst_param_init_ssa_scalar(ins, 0, sm6)) ins->dst->modifiers = VKD3DSPDM_SATURATE; } @@ -6197,7 +6236,7 @@ static void sm6_parser_emit_dx_split_double(struct sm6_parser *sm6, enum dx_intr vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_MOV); if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); instruction_dst_param_init_ssa_vector(ins, 2, sm6); } @@ -6207,7 +6246,7 @@ static void sm6_parser_emit_dx_store_output(struct sm6_parser *sm6, enum dx_intr { bool is_patch_constant = op == DX_STORE_PATCH_CONSTANT; struct vkd3d_shader_instruction *ins = state->ins; - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; struct vkd3d_shader_src_param *src_param; struct vkd3d_shader_dst_param *dst_param; const struct shader_signature *signature; @@ -6263,7 +6302,7 @@ static void sm6_parser_emit_dx_store_output(struct sm6_parser *sm6, enum dx_intr } if ((src_param = instruction_src_params_alloc(ins, 1, sm6))) - src_param_init_from_value(src_param, value, sm6); + src_param_init_from_value(src_param, value, 0, sm6); } static void sm6_parser_emit_dx_texture_gather(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -6309,7 +6348,7 @@ static void sm6_parser_emit_dx_texture_gather(struct sm6_parser *sm6, enum dx_in instruction_init_with_resource(ins, extended_offset ? VSIR_OP_GATHER4_PO_C : VSIR_OP_GATHER4_C, resource, sm6); if (!(src_params = instruction_src_params_alloc(ins, 4 + extended_offset, sm6))) return; - src_param_init_from_value(&src_params[3 + extended_offset], operands[9], sm6); + src_param_init_from_value(&src_params[3 + extended_offset], operands[9], 0, sm6); } src_param_init_vector_from_reg(&src_params[0], &coord); @@ -6372,7 +6411,7 @@ static void sm6_parser_emit_dx_texture_load(struct sm6_parser *sm6, enum dx_intr src_param_init_vector_from_reg(&src_params[0], &coord); src_param_init_vector_from_handle(sm6, &src_params[1], &resource->u.handle); if (is_multisample) - src_param_init_from_value(&src_params[2], mip_level_or_sample_count, sm6); + src_param_init_from_value(&src_params[2], mip_level_or_sample_count, 0, sm6); instruction_dst_param_init_ssa_vector(ins, VKD3D_VEC4_SIZE, sm6); } @@ -6436,7 +6475,7 @@ static void sm6_parser_emit_dx_wave_active_ballot(struct sm6_parser *sm6, enum d vsir_instruction_init(ins, &sm6->p.location, VSIR_OP_WAVE_ACTIVE_BALLOT); if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); instruction_dst_param_init_ssa_vector(ins, VKD3D_VEC4_SIZE, sm6); } @@ -6476,9 +6515,9 @@ static void sm6_parser_emit_dx_wave_active_bit(struct sm6_parser *sm6, enum dx_i if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static enum vkd3d_shader_opcode sm6_dx_map_wave_op(enum dxil_wave_op_kind op, bool is_signed, bool is_float, @@ -6527,9 +6566,9 @@ static void sm6_parser_emit_dx_wave_op(struct sm6_parser *sm6, enum dx_intrinsic if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - src_param_init_from_value(src_param, operands[0], sm6); + src_param_init_from_value(src_param, operands[0], 0, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_dx_wave_builtin(struct sm6_parser *sm6, enum dx_intrinsic_opcode op, @@ -6918,14 +6957,15 @@ static void sm6_parser_emit_call(struct sm6_parser *sm6, const struct dxil_recor fn_value->u.function.name, &operands[1], operand_count - 1, state, dst); } -static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_type *from, - const struct sm6_type *to, struct sm6_parser *sm6) +static enum vkd3d_shader_opcode dxil_map_cast_op(uint64_t code, const struct sm6_type *from, + uint32_t *src_type_flags, const struct sm6_type *to, struct sm6_parser *dxil) { enum vkd3d_shader_opcode op = VSIR_OP_INVALID; bool from_int, to_int, from_fp, to_fp; unsigned int from_width, to_width; bool is_valid = false; + *src_type_flags = 0; from_int = sm6_type_is_integer(from); to_int = sm6_type_is_integer(to); from_fp = sm6_type_is_floating_point(from); @@ -6934,15 +6974,13 @@ static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_ /* NOTE: DXIL currently doesn't use vectors here. */ if ((!from_int && !from_fp) || (!to_int && !to_fp)) { - FIXME("Unhandled cast of type class %u to type class %u.\n", from->class, to->class); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, "Cast of type class %u to type class %u is not implemented.", from->class, to->class); return VSIR_OP_INVALID; } if (to->u.width == 8 || from->u.width == 8) { - FIXME("Unhandled 8-bit value.\n"); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, "Cast to/from an 8-bit type is not implemented."); return VSIR_OP_INVALID; } @@ -6982,6 +7020,7 @@ static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_ case CAST_SITOFP: op = VSIR_OP_ITOF; is_valid = from_int && to_fp; + *src_type_flags |= DXIL_TYPE_SIGNED; break; case CAST_FPTRUNC: @@ -7000,16 +7039,14 @@ static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_ break; default: - FIXME("Unhandled cast op %"PRIu64".\n", code); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, "Cast operation %"PRIu64" is unhandled.", code); return VSIR_OP_INVALID; } if (!is_valid) { - FIXME("Invalid types %u and/or %u for op %"PRIu64".\n", from->class, to->class, code); - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, + vkd3d_shader_parser_error(&dxil->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND, "Cast operation %"PRIu64" from type class %u, width %u to type class %u, width %u is invalid.", code, from->class, from->u.width, to->class, to->u.width); return VSIR_OP_INVALID; @@ -7032,22 +7069,23 @@ static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_ return op; } -static void sm6_parser_emit_cast(struct sm6_parser *sm6, const struct dxil_record *record, +static void sm6_parser_emit_cast(struct sm6_parser *dxil, const struct dxil_record *record, struct vkd3d_shader_instruction *ins, struct sm6_value *dst) { struct vkd3d_shader_src_param *src_param; - enum vkd3d_shader_opcode handler_idx; const struct sm6_value *value; + enum vkd3d_shader_opcode op; const struct sm6_type *type; + uint32_t src_type_flags; unsigned int i = 0; - if (!(value = sm6_parser_get_value_by_ref(sm6, record, NULL, &i))) + if (!(value = sm6_parser_get_value_by_ref(dxil, record, NULL, &i))) return; - if (!dxil_record_validate_operand_count(record, i + 2, i + 2, sm6)) + if (!dxil_record_validate_operand_count(record, i + 2, i + 2, dxil)) return; - if (!(type = sm6_parser_get_type(sm6, record->operands[i++]))) + if (!(type = sm6_parser_get_type(dxil, record->operands[i++]))) return; dst->type = type; @@ -7060,28 +7098,28 @@ static void sm6_parser_emit_cast(struct sm6_parser *sm6, const struct dxil_recor return; } - if ((handler_idx = sm6_map_cast_op(record->operands[i], value->type, type, sm6)) == VSIR_OP_INVALID) + if ((op = dxil_map_cast_op(record->operands[i], value->type, &src_type_flags, type, dxil)) == VSIR_OP_INVALID) return; - vsir_instruction_init(ins, &sm6->p.location, handler_idx); + vsir_instruction_init(ins, &dxil->p.location, op); - if (handler_idx == VSIR_OP_NOP) + if (op == VSIR_OP_NOP) { *dst = *value; dst->type = type; return; } - if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) + if (!(src_param = instruction_src_params_alloc(ins, 1, dxil))) return; - src_param_init_from_value(src_param, value, sm6); + src_param_init_from_value(src_param, value, src_type_flags, dxil); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, dxil); /* VSIR bitcasts are represented by source registers with types different * from the types they were written with, rather than with different types * for the MOV source and destination. */ - if (handler_idx == VSIR_OP_MOV) + if (op == VSIR_OP_MOV) src_param->reg.data_type = ins->dst[0].reg.data_type; } @@ -7089,6 +7127,7 @@ struct sm6_cmp_info { enum vkd3d_shader_opcode handler_idx; bool src_swap; + uint32_t type_flags; }; static const struct sm6_cmp_info *sm6_map_cmp2_op(uint64_t code) @@ -7118,10 +7157,10 @@ static const struct sm6_cmp_info *sm6_map_cmp2_op(uint64_t code) [ICMP_UGE] = {VSIR_OP_UGE}, [ICMP_ULT] = {VSIR_OP_ULT}, [ICMP_ULE] = {VSIR_OP_UGE, true}, - [ICMP_SGT] = {VSIR_OP_ILT, true}, - [ICMP_SGE] = {VSIR_OP_IGE}, - [ICMP_SLT] = {VSIR_OP_ILT}, - [ICMP_SLE] = {VSIR_OP_IGE, true}, + [ICMP_SGT] = {VSIR_OP_ILT, true, DXIL_TYPE_SIGNED}, + [ICMP_SGE] = {VSIR_OP_IGE, false, DXIL_TYPE_SIGNED}, + [ICMP_SLT] = {VSIR_OP_ILT, false, DXIL_TYPE_SIGNED}, + [ICMP_SLE] = {VSIR_OP_IGE, true, DXIL_TYPE_SIGNED}, }; return (code < ARRAY_SIZE(cmp_op_table)) ? &cmp_op_table[code] : NULL; @@ -7222,10 +7261,10 @@ static void sm6_parser_emit_cmp2(struct sm6_parser *sm6, const struct dxil_recor if (!(src_params = instruction_src_params_alloc(ins, 2, sm6))) return; - src_param_init_from_value(&src_params[0 ^ cmp->src_swap], a, sm6); - src_param_init_from_value(&src_params[1 ^ cmp->src_swap], b, sm6); + src_param_init_from_value(&src_params[0 ^ cmp->src_swap], a, cmp->type_flags, sm6); + src_param_init_from_value(&src_params[1 ^ cmp->src_swap], b, cmp->type_flags, sm6); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_cmpxchg(struct sm6_parser *sm6, const struct dxil_record *record, @@ -7245,7 +7284,7 @@ static void sm6_parser_emit_cmpxchg(struct sm6_parser *sm6, const struct dxil_re || !sm6_value_validate_is_backward_ref(ptr, sm6)) return; - sm6_register_from_value(®, ptr, sm6); + vsir_register_from_dxil_value(®, ptr, 0, sm6); if (reg.type != VKD3DSPR_GROUPSHAREDMEM) { @@ -7300,14 +7339,14 @@ static void sm6_parser_emit_cmpxchg(struct sm6_parser *sm6, const struct dxil_re if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) return; src_param_make_constant_uint(&src_params[0], 0); - src_param_init_from_value(&src_params[1], cmp, sm6); - src_param_init_from_value(&src_params[2], new, sm6); + src_param_init_from_value(&src_params[1], cmp, 0, sm6); + src_param_init_from_value(&src_params[2], new, 0, sm6); sm6_parser_init_ssa_value(sm6, dst); if (!(dst_params = instruction_dst_params_alloc(ins, 2, sm6))) return; - sm6_register_from_value(&dst_params[0].reg, dst, sm6); + vsir_register_from_dxil_value(&dst_params[0].reg, dst, 0, sm6); dst_param_init(&dst_params[0]); dst_params[1].reg = reg; dst_param_init(&dst_params[1]); @@ -7366,10 +7405,10 @@ static void sm6_parser_emit_extractval(struct sm6_parser *sm6, const struct dxil if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) return; - sm6_register_from_value(&src_param->reg, src, sm6); + vsir_register_from_dxil_value(&src_param->reg, src, 0, sm6); src_param_init_scalar(src_param, elem_idx); - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static void sm6_parser_emit_gep(struct sm6_parser *sm6, const struct dxil_record *record, @@ -7515,7 +7554,7 @@ static void sm6_parser_emit_load(struct sm6_parser *sm6, const struct dxil_recor if (record->operands[i]) WARN("Ignoring volatile modifier.\n"); - sm6_register_from_value(®, ptr, sm6); + vsir_register_from_dxil_value(®, ptr, 0, sm6); if (ptr->structure_stride) { @@ -7530,7 +7569,7 @@ static void sm6_parser_emit_load(struct sm6_parser *sm6, const struct dxil_recor src_param_make_constant_uint(&src_params[0], reg.idx[1].offset); /* Struct offset is always zero as there is no struct, just an array. */ src_param_make_constant_uint(&src_params[1], 0); - src_param_init_from_value(&src_params[2], ptr, sm6); + src_param_init_from_value(&src_params[2], ptr, 0, sm6); src_params[2].reg.alignment = alignment; /* The offset is already in src_params[0]. */ src_params[2].reg.idx_count = 1; @@ -7544,11 +7583,11 @@ static void sm6_parser_emit_load(struct sm6_parser *sm6, const struct dxil_recor return; if (operand_count > 1) src_param_make_constant_uint(&src_params[0], 0); - src_param_init_from_value(&src_params[operand_count - 1], ptr, sm6); + src_param_init_from_value(&src_params[operand_count - 1], ptr, 0, sm6); src_params[operand_count - 1].reg.alignment = alignment; } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static int phi_incoming_compare(const void *a, const void *b) @@ -7702,7 +7741,7 @@ static void sm6_parser_emit_store(struct sm6_parser *sm6, const struct dxil_reco if (record->operands[i]) WARN("Ignoring volatile modifier.\n"); - sm6_register_from_value(®, ptr, sm6); + vsir_register_from_dxil_value(®, ptr, 0, sm6); if (ptr->structure_stride) { @@ -7717,7 +7756,7 @@ static void sm6_parser_emit_store(struct sm6_parser *sm6, const struct dxil_reco src_param_make_constant_uint(&src_params[0], reg.idx[1].offset); /* Struct offset is always zero as there is no struct, just an array. */ src_param_make_constant_uint(&src_params[1], 0); - src_param_init_from_value(&src_params[2], src, sm6); + src_param_init_from_value(&src_params[2], src, 0, sm6); } else { @@ -7728,7 +7767,7 @@ static void sm6_parser_emit_store(struct sm6_parser *sm6, const struct dxil_reco return; if (operand_count > 1) src_param_make_constant_uint(&src_params[0], 0); - src_param_init_from_value(&src_params[operand_count - 1], src, sm6); + src_param_init_from_value(&src_params[operand_count - 1], src, 0, sm6); } dst_param = instruction_dst_params_alloc(ins, 1, sm6); @@ -7778,7 +7817,7 @@ static void sm6_parser_emit_switch(struct sm6_parser *sm6, const struct dxil_rec return; } - sm6_register_from_value(&terminator->conditional_reg, src, sm6); + vsir_register_from_dxil_value(&terminator->conditional_reg, src, 0, sm6); terminator->type = TERMINATOR_SWITCH; terminator->case_count = record->operand_count / 2u; @@ -7854,9 +7893,11 @@ static void sm6_parser_emit_vselect(struct sm6_parser *sm6, const struct dxil_re if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) return; for (i = 0; i < 3; ++i) - src_param_init_from_value(&src_params[i], src[i], sm6); + { + src_param_init_from_value(&src_params[i], src[i], 0, sm6); + } - instruction_dst_param_init_ssa_scalar(ins, sm6); + instruction_dst_param_init_ssa_scalar(ins, 0, sm6); } static bool sm6_metadata_value_is_node(const struct sm6_metadata_value *m) @@ -8195,7 +8236,7 @@ static enum vkd3d_result sm6_function_resolve_phi_incomings(const struct sm6_fun vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_TYPE_MISMATCH, "The type of a phi incoming value does not match the result type."); } - sm6_register_from_value(&phi->incoming[j].reg, src, sm6); + vsir_register_from_dxil_value(&phi->incoming[j].reg, src, 0, sm6); } } } @@ -8206,7 +8247,7 @@ static enum vkd3d_result sm6_function_resolve_phi_incomings(const struct sm6_fun static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const struct dxil_block *block, struct sm6_function *function) { - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; struct vkd3d_shader_instruction *ins; size_t i, block_idx, block_count; const struct dxil_record *record; @@ -8355,8 +8396,8 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const return VKD3D_ERROR_INVALID_SHADER; } - if (sm6->p.failed) - return VKD3D_ERROR; + if (sm6->p.status < 0) + return sm6->p.status; if (record->attachment) metadata_attachment_record_apply(record->attachment, record->code, ins, dst, sm6); @@ -8403,18 +8444,26 @@ static void sm6_block_emit_terminator(const struct sm6_block *block, struct sm6_ case TERMINATOR_UNCOND_BR: if (!block->terminator.true_block) return; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_BRANCH); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_BRANCH))) + return; if (!(src_params = instruction_src_params_alloc(ins, 1, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } vsir_src_param_init_label(&src_params[0], block->terminator.true_block->id); break; case TERMINATOR_COND_BR: if (!block->terminator.true_block || !block->terminator.false_block) return; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_BRANCH); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_BRANCH))) + return; if (!(src_params = instruction_src_params_alloc(ins, 3, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } src_param_init(&src_params[0]); src_params[0].reg = block->terminator.conditional_reg; vsir_src_param_init_label(&src_params[1], block->terminator.true_block->id); @@ -8422,9 +8471,13 @@ static void sm6_block_emit_terminator(const struct sm6_block *block, struct sm6_ break; case TERMINATOR_SWITCH: - ins = sm6_parser_add_instruction(sm6, VSIR_OP_SWITCH_MONOLITHIC); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_SWITCH_MONOLITHIC))) + return; if (!(src_params = instruction_src_params_alloc(ins, block->terminator.case_count * 2u + 1, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } src_param_init(&src_params[0]); src_params[0].reg = block->terminator.conditional_reg; /* TODO: emit the merge block id. */ @@ -8438,7 +8491,7 @@ static void sm6_block_emit_terminator(const struct sm6_block *block, struct sm6_ switch_case = &block->terminator.cases[i]; if (!(case_block = switch_case->block)) { - VKD3D_ASSERT(sm6->p.failed); + VKD3D_ASSERT(sm6->p.status < 0); continue; } if (switch_case->is_default) @@ -8491,11 +8544,18 @@ static void sm6_block_emit_phi(const struct sm6_block *block, struct sm6_parser src_phi = &block->phi[i]; incoming_count = src_phi->incoming_count; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_PHI); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_PHI))) + return; if (!(src_params = instruction_src_params_alloc(ins, incoming_count * 2u, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } if (!(dst_param = instruction_dst_params_alloc(ins, 1, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } for (j = 0; j < incoming_count; ++j) { @@ -8507,11 +8567,11 @@ static void sm6_block_emit_phi(const struct sm6_block *block, struct sm6_parser if (incoming_block) vsir_src_param_init_label(&src_params[index + 1], incoming_block->id); else - VKD3D_ASSERT(sm6->p.failed); + VKD3D_ASSERT(sm6->p.status < 0); } dst_param_init(dst_param); - sm6_register_from_value(&dst_param->reg, &src_phi->value, sm6); + vsir_register_from_dxil_value(&dst_param->reg, &src_phi->value, 0, sm6); } } @@ -8577,42 +8637,45 @@ static void sm6_parser_emit_label(struct sm6_parser *sm6, unsigned int label_id) struct vkd3d_shader_src_param *src_param; struct vkd3d_shader_instruction *ins; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_LABEL); - + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_LABEL))) + return; if (!(src_param = instruction_src_params_alloc(ins, 1, sm6))) + { + vkd3d_shader_instruction_make_nop(ins); return; + } vsir_src_param_init_label(src_param, label_id); } static enum vkd3d_result sm6_function_emit_blocks(const struct sm6_function *function, struct sm6_parser *sm6) { - struct vsir_program *program = sm6->p.program; - unsigned int i; + struct vsir_program *program = sm6->program; + struct vkd3d_shader_instruction *ins; + unsigned int i, j; - program->block_count = function->block_count; + program->block_count = max(program->block_count, function->block_count); for (i = 0; i < function->block_count; ++i) { const struct sm6_block *block = function->blocks[i]; - /* Space for the label and terminator. */ - if (!sm6_parser_require_space(sm6, block->instruction_count + block->phi_count + 2)) - { - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, - "Out of memory emitting shader instructions."); - return VKD3D_ERROR_OUT_OF_MEMORY; - } sm6_parser_emit_label(sm6, block->id); sm6_block_emit_phi(block, sm6); - memcpy(&program->instructions.elements[program->instructions.count], block->instructions, - block->instruction_count * sizeof(*block->instructions)); - program->instructions.count += block->instruction_count; - + for (j = 0; j < block->instruction_count; ++j) + { + if (!(ins = vsir_program_append(program))) + { + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, + "Out of memory emitting block instructions."); + return sm6->p.status; + } + *ins = block->instructions[j]; + } sm6_block_emit_terminator(block, sm6); } - return VKD3D_OK; + return sm6->p.status; } static bool sm6_parser_allocate_named_metadata(struct sm6_parser *sm6) @@ -9510,8 +9573,8 @@ static enum vkd3d_result sm6_parser_descriptor_type_init(struct sm6_parser *sm6, struct vkd3d_shader_instruction *ins; const struct sm6_metadata_node *node; const struct sm6_metadata_value *m; + enum vkd3d_result ret = VKD3D_OK; struct sm6_descriptor_info *d; - enum vkd3d_result ret; unsigned int i; for (i = 0; i < descriptor_node->operand_count; ++i) @@ -9560,42 +9623,41 @@ static enum vkd3d_result sm6_parser_descriptor_type_init(struct sm6_parser *sm6, return VKD3D_ERROR_INVALID_SHADER; } - if (!(ins = sm6_parser_require_space(sm6, 1))) - { - ERR("Failed to allocate instruction.\n"); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_NOP))) return VKD3D_ERROR_OUT_OF_MEMORY; - } switch (type) { case VKD3D_SHADER_DESCRIPTOR_TYPE_CBV: if ((ret = sm6_parser_resources_load_cbv(sm6, node, d, ins)) < 0) - return ret; + goto done; break; case VKD3D_SHADER_DESCRIPTOR_TYPE_SRV: if ((ret = sm6_parser_resources_load_srv(sm6, node, d, ins)) < 0) - return ret; + goto done; break; case VKD3D_SHADER_DESCRIPTOR_TYPE_UAV: if ((ret = sm6_parser_resources_load_uav(sm6, node, d, ins)) < 0) - return ret; + goto done; break; case VKD3D_SHADER_DESCRIPTOR_TYPE_SAMPLER: if ((ret = sm6_parser_resources_load_sampler(sm6, node, d, ins)) < 0) - return ret; + goto done; break; default: - FIXME("Unsupported descriptor type %u.\n", type); vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_RESOURCES, - "Resource descriptor type %u is unsupported.", type); - return VKD3D_ERROR_INVALID_SHADER; + "Resource descriptor type %#x is unsupported.", type); + ret = VKD3D_ERROR_INVALID_SHADER; + goto done; } ++sm6->descriptor_count; - ++sm6->p.program->instructions.count; } - return VKD3D_OK; +done: + if (ret < 0) + vsir_instruction_init(ins, &ins->location, VSIR_OP_NOP); + return ret; } static enum vkd3d_result sm6_parser_resources_init(struct sm6_parser *sm6) @@ -9711,7 +9773,7 @@ static enum vkd3d_result sm6_parser_read_signature(struct sm6_parser *sm6, const { unsigned int i, j, column_count, operand_count, index; const struct sm6_metadata_node *node, *element_node; - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; struct signature_element *elements, *e; unsigned int values[10]; bool native_16bit; @@ -9930,7 +9992,7 @@ invalid: static enum vkd3d_result sm6_parser_signatures_init(struct sm6_parser *sm6, const struct sm6_metadata_value *m, enum vkd3d_tessellator_domain tessellator_domain) { - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; enum vkd3d_result ret; if (!sm6_metadata_value_is_node(m)) @@ -9953,6 +10015,14 @@ static enum vkd3d_result sm6_parser_signatures_init(struct sm6_parser *sm6, cons &program->patch_constant_signature, tessellator_domain, false)) < 0) return ret; + if (sm6->program->shader_version.type == VKD3D_SHADER_TYPE_HULL + && !(sm6->outpointid_param = vsir_program_create_outpointid_param(sm6->program))) + { + vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, + "Failed to allocate outpointid parameter."); + return VKD3D_ERROR_OUT_OF_MEMORY; + } + if ((ret = sm6_parser_init_input_signature(sm6, &program->input_signature)) < 0) return ret; @@ -9983,14 +10053,15 @@ static void sm6_parser_emit_global_flags(struct sm6_parser *sm6, const struct sm rotated_flags = (rotated_flags >> 1) | ((rotated_flags & 1) << 4); global_flags = (global_flags & ~mask) | rotated_flags; - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_GLOBAL_FLAGS); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_GLOBAL_FLAGS))) + return; ins->declaration.global_flags = global_flags; - sm6->p.program->global_flags = global_flags; + sm6->program->global_flags = global_flags; } static enum vkd3d_result sm6_parser_emit_thread_group(struct sm6_parser *sm6, const struct sm6_metadata_value *m) { - struct vkd3d_shader_version *version = &sm6->p.program->shader_version; + struct vkd3d_shader_version *version = &sm6->program->shader_version; const struct sm6_metadata_node *node; struct vkd3d_shader_instruction *ins; unsigned int group_sizes[3]; @@ -10040,11 +10111,12 @@ static enum vkd3d_result sm6_parser_emit_thread_group(struct sm6_parser *sm6, co } } - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_THREAD_GROUP); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_THREAD_GROUP))) + return VKD3D_ERROR_OUT_OF_MEMORY; ins->declaration.thread_group_size.x = group_sizes[0]; ins->declaration.thread_group_size.y = group_sizes[1]; ins->declaration.thread_group_size.z = group_sizes[2]; - sm6->p.program->thread_group_size = ins->declaration.thread_group_size; + sm6->program->thread_group_size = ins->declaration.thread_group_size; return VKD3D_OK; } @@ -10053,7 +10125,8 @@ static void sm6_parser_emit_dcl_count(struct sm6_parser *sm6, enum vkd3d_shader_ { struct vkd3d_shader_instruction *ins; - ins = sm6_parser_add_instruction(sm6, handler_idx); + if (!(ins = sm6_parser_add_instruction(sm6, handler_idx))) + return; ins->declaration.count = count; } @@ -10063,7 +10136,8 @@ static void sm6_parser_emit_dcl_primitive_topology(struct sm6_parser *sm6, { struct vkd3d_shader_instruction *ins; - ins = sm6_parser_add_instruction(sm6, handler_idx); + if (!(ins = sm6_parser_add_instruction(sm6, handler_idx))) + return; ins->declaration.primitive_type.type = primitive_type; ins->declaration.primitive_type.patch_vertex_count = patch_vertex_count; } @@ -10080,9 +10154,10 @@ static void sm6_parser_emit_dcl_tessellator_domain(struct sm6_parser *sm6, "Domain shader tessellator domain %u is unhandled.", tessellator_domain); } - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_DOMAIN); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_DOMAIN))) + return; ins->declaration.tessellator_domain = tessellator_domain; - sm6->p.program->tess_domain = tessellator_domain; + sm6->program->tess_domain = tessellator_domain; } static void sm6_parser_validate_control_point_count(struct sm6_parser *sm6, @@ -10108,10 +10183,11 @@ static void sm6_parser_emit_dcl_tessellator_partitioning(struct sm6_parser *sm6, "Hull shader tessellator partitioning %u is unhandled.", tessellator_partitioning); } - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_PARTITIONING); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_PARTITIONING))) + return; ins->declaration.tessellator_partitioning = tessellator_partitioning; - sm6->p.program->tess_partitioning = tessellator_partitioning; + sm6->program->tess_partitioning = tessellator_partitioning; } static void sm6_parser_emit_dcl_tessellator_output_primitive(struct sm6_parser *sm6, @@ -10126,10 +10202,11 @@ static void sm6_parser_emit_dcl_tessellator_output_primitive(struct sm6_parser * "Hull shader tessellator output primitive %u is unhandled.", primitive); } - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_OUTPUT_PRIMITIVE); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_TESSELLATOR_OUTPUT_PRIMITIVE))) + return; ins->declaration.tessellator_output_primitive = primitive; - sm6->p.program->tess_output_primitive = primitive; + sm6->program->tess_output_primitive = primitive; } static void sm6_parser_emit_dcl_max_tessellation_factor(struct sm6_parser *sm6, struct sm6_metadata_value *m) @@ -10153,7 +10230,8 @@ static void sm6_parser_emit_dcl_max_tessellation_factor(struct sm6_parser *sm6, "Hull shader max tessellation factor %f is invalid.", max_tessellation_factor); } - ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_HS_MAX_TESSFACTOR); + if (!(ins = sm6_parser_add_instruction(sm6, VSIR_OP_DCL_HS_MAX_TESSFACTOR))) + return; ins->declaration.max_tessellation_factor = max_tessellation_factor; } @@ -10241,8 +10319,8 @@ static void sm6_parser_gs_properties_init(struct sm6_parser *sm6, const struct s } sm6_parser_emit_dcl_primitive_topology(sm6, VSIR_OP_DCL_INPUT_PRIMITIVE, input_primitive, patch_vertex_count); - sm6->p.program->input_primitive = input_primitive; - sm6->p.program->input_control_point_count = input_control_point_count; + sm6->program->input_primitive = input_primitive; + sm6->program->input_control_point_count = input_control_point_count; i = operands[1]; /* Max total scalar count sets an upper limit. We would need to scan outputs to be more precise. */ @@ -10253,7 +10331,7 @@ static void sm6_parser_gs_properties_init(struct sm6_parser *sm6, const struct s "Geometry shader output vertex count %u is invalid.", i); } sm6_parser_emit_dcl_count(sm6, VSIR_OP_DCL_VERTICES_OUT, i); - sm6->p.program->vertices_out_count = i; + sm6->program->vertices_out_count = i; if (operands[2] > 1) { @@ -10271,7 +10349,7 @@ static void sm6_parser_gs_properties_init(struct sm6_parser *sm6, const struct s output_primitive = VKD3D_PT_TRIANGLELIST; } sm6_parser_emit_dcl_primitive_topology(sm6, VSIR_OP_DCL_OUTPUT_TOPOLOGY, output_primitive, 0); - sm6->p.program->output_topology = output_primitive; + sm6->program->output_topology = output_primitive; i = operands[4]; if (!i || i > MAX_GS_INSTANCE_COUNT) @@ -10326,7 +10404,7 @@ static enum vkd3d_tessellator_domain sm6_parser_ds_properties_init(struct sm6_pa sm6_parser_emit_dcl_tessellator_domain(sm6, operands[0]); sm6_parser_validate_control_point_count(sm6, operands[1], true, "Domain shader input"); - sm6->p.program->input_control_point_count = operands[1]; + sm6->program->input_control_point_count = operands[1]; return operands[0]; } @@ -10334,7 +10412,7 @@ static enum vkd3d_tessellator_domain sm6_parser_ds_properties_init(struct sm6_pa static enum vkd3d_tessellator_domain sm6_parser_hs_properties_init(struct sm6_parser *sm6, const struct sm6_metadata_value *m) { - struct vsir_program *program = sm6->p.program; + struct vsir_program *program = sm6->program; const struct sm6_metadata_node *node; unsigned int operands[6] = {0}; unsigned int i; @@ -10633,7 +10711,6 @@ static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, struct vsir_pro { size_t count, length, function_count, expected_function_count, byte_code_size = dxbc_desc->byte_code_size; const struct vkd3d_shader_location location = {.source_name = compile_info->source_name}; - struct shader_signature *patch_constant_signature, *output_signature, *input_signature; uint32_t version_token, dxil_version, token_count, magic; const uint32_t *byte_code = dxbc_desc->byte_code; unsigned int chunk_offset, chunk_size; @@ -10727,9 +10804,10 @@ static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, struct vsir_pro if (!vsir_program_init(program, compile_info, &version, (count + (count >> 2)) / 2u + 10, VSIR_CF_BLOCKS, VSIR_NORMALISED_SM6)) return VKD3D_ERROR_OUT_OF_MEMORY; - vkd3d_shader_parser_init(&sm6->p, program, message_context, compile_info->source_name); + vkd3d_shader_parser_init(&sm6->p, message_context, compile_info->source_name); sm6->ptr = &sm6->start[1]; sm6->bitpos = 2; + sm6->program = program; switch (program->shader_version.type) { @@ -10746,9 +10824,6 @@ static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, struct vsir_pro break; } - input_signature = &program->input_signature; - output_signature = &program->output_signature; - patch_constant_signature = &program->patch_constant_signature; program->features = dxbc_desc->features; memset(dxbc_desc, 0, sizeof(*dxbc_desc)); @@ -10890,15 +10965,6 @@ static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, struct vsir_pro goto fail; } - if (!sm6_parser_require_space(sm6, output_signature->element_count + input_signature->element_count - + patch_constant_signature->element_count)) - { - vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY, - "Out of memory emitting shader signature declarations."); - ret = VKD3D_ERROR_OUT_OF_MEMORY; - goto fail; - } - program->ssa_count = sm6->ssa_next_id; if (!(fn = sm6_parser_get_function(sm6, sm6->entry_point))) @@ -10949,15 +11015,13 @@ static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, struct vsir_pro dxil_block_destroy(&sm6->root_block); - if (sm6->p.failed) - { - ret = VKD3D_ERROR_INVALID_SHADER; + if (sm6->p.status < 0) goto fail; - } - return VKD3D_OK; fail: + if (ret >= 0 && sm6->p.status < 0) + ret = sm6->p.status; sm6_parser_cleanup(sm6); vsir_program_cleanup(program); return ret; diff --git a/libs/vkd3d/libs/vkd3d-shader/fx.c b/libs/vkd3d/libs/vkd3d-shader/fx.c index 676c501bb08..a14346a45d2 100644 --- a/libs/vkd3d/libs/vkd3d-shader/fx.c +++ b/libs/vkd3d/libs/vkd3d-shader/fx.c @@ -4205,7 +4205,7 @@ static void fx_parse_shader_blob(struct fx_parser *parser, enum vkd3d_shader_sou static const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, }; info.type = VKD3D_SHADER_STRUCTURE_TYPE_COMPILE_INFO; diff --git a/libs/vkd3d/libs/vkd3d-shader/glsl.c b/libs/vkd3d/libs/vkd3d-shader/glsl.c index dfe0a40ddf0..b2679beff9f 100644 --- a/libs/vkd3d/libs/vkd3d-shader/glsl.c +++ b/libs/vkd3d/libs/vkd3d-shader/glsl.c @@ -288,6 +288,14 @@ static void shader_glsl_print_register_name(struct vkd3d_string_buffer *buffer, shader_glsl_print_subscript(buffer, gen, reg->idx[1].rel_addr, reg->idx[1].offset); break; + case VKD3DSPR_SAMPLEMASK: + if (gen->program->shader_version.type != VKD3D_SHADER_TYPE_PIXEL) + vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, + "Internal compiler error: Unhandled sample coverage mask in shader type #%x.", + gen->program->shader_version.type); + vkd3d_string_buffer_printf(buffer, "o_mask"); + break; + default: vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, "Internal compiler error: Unhandled register type %#x.", reg->type); @@ -792,7 +800,7 @@ static void shader_glsl_print_texel_offset(struct vkd3d_string_buffer *buffer, s static void shader_glsl_ld(struct vkd3d_glsl_generator *gen, const struct vkd3d_shader_instruction *ins) { - unsigned int resource_id, resource_idx, resource_space, sample_count; + unsigned int coord_size, resource_id, resource_idx, resource_space, sample_count; const struct glsl_resource_type_info *resource_type_info; const struct vkd3d_shader_descriptor_info1 *d; enum vkd3d_shader_resource_type resource_type; @@ -800,11 +808,9 @@ static void shader_glsl_ld(struct vkd3d_glsl_generator *gen, const struct vkd3d_ enum vsir_data_type data_type; struct glsl_src coord; struct glsl_dst dst; - uint32_t coord_mask; + bool array, offset; - if (vkd3d_shader_instruction_has_texel_offset(ins)) - vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, - "Internal compiler error: Unhandled texel fetch offset."); + offset = vkd3d_shader_instruction_has_texel_offset(ins); if (ins->src[1].reg.idx[0].rel_addr || ins->src[1].reg.idx[1].rel_addr) vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_UNSUPPORTED, @@ -831,20 +837,22 @@ static void shader_glsl_ld(struct vkd3d_glsl_generator *gen, const struct vkd3d_ if ((resource_type_info = shader_glsl_get_resource_type_info(resource_type))) { - coord_mask = vkd3d_write_mask_from_component_count(resource_type_info->coord_size); + coord_size = resource_type_info->coord_size; + array = resource_type_info->array; } else { vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, "Internal compiler error: Unhandled resource type %#x.", resource_type); - coord_mask = vkd3d_write_mask_from_component_count(2); + coord_size = 2; + array = false; } glsl_dst_init(&dst, gen, ins, &ins->dst[0]); - glsl_src_init(&coord, gen, &ins->src[0], coord_mask); + glsl_src_init(&coord, gen, &ins->src[0], vkd3d_write_mask_from_component_count(coord_size)); fetch = vkd3d_string_buffer_get(&gen->string_buffers); - vkd3d_string_buffer_printf(fetch, "texelFetch("); + vkd3d_string_buffer_printf(fetch, "texelFetch%s(", offset ? "Offset" : ""); shader_glsl_print_combined_sampler_name(fetch, gen, resource_idx, resource_space, VKD3D_SHADER_DUMMY_SAMPLER_INDEX, 0); vkd3d_string_buffer_printf(fetch, ", %s", coord.str->buffer); @@ -860,6 +868,11 @@ static void shader_glsl_ld(struct vkd3d_glsl_generator *gen, const struct vkd3d_ else shader_glsl_print_src(fetch, gen, &ins->src[2], VKD3DSP_WRITEMASK_0, ins->src[2].reg.data_type); } + if (offset) + { + vkd3d_string_buffer_printf(fetch, ", "); + shader_glsl_print_texel_offset(fetch, gen, coord_size - array, &ins->texel_offset); + } vkd3d_string_buffer_printf(fetch, ")"); shader_glsl_print_swizzle(fetch, ins->src[1].swizzle, ins->dst[0].write_mask); @@ -1281,6 +1294,13 @@ static void shader_glsl_print_sysval_name(struct vkd3d_string_buffer *buffer, st vkd3d_string_buffer_printf(buffer, "intBitsToFloat(ivec4(gl_VertexID, 0, 0, 0))"); break; + case VKD3D_SHADER_SV_INSTANCE_ID: + if (version->type != VKD3D_SHADER_TYPE_VERTEX) + vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, + "Internal compiler error: Unhandled SV_INSTANCE_ID in shader type #%x.", version->type); + vkd3d_string_buffer_printf(buffer, "intBitsToFloat(ivec4(gl_InstanceID, 0, 0, 0))"); + break; + case VKD3D_SHADER_SV_IS_FRONT_FACE: if (version->type != VKD3D_SHADER_TYPE_PIXEL) vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL, @@ -1429,6 +1449,12 @@ static void shader_glsl_shader_epilogue(struct vkd3d_glsl_generator *gen) shader_glsl_print_write_mask(buffer, e->mask); vkd3d_string_buffer_printf(buffer, ";\n"); } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_SAMPLEMASK)) + { + shader_glsl_print_indent(buffer, gen->indent); + vkd3d_string_buffer_printf(gen->buffer, "gl_SampleMask[0] = floatBitsToInt(o_mask);\n"); + } } static void shader_glsl_ret(struct vkd3d_glsl_generator *gen, const struct vkd3d_shader_instruction *ins) @@ -1475,6 +1501,9 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *gen, case VSIR_OP_CONTINUE: shader_glsl_continue(gen); break; + case VSIR_OP_COS: + shader_glsl_intrinsic(gen, ins, "cos"); + break; case VSIR_OP_DCL_INDEXABLE_TEMP: shader_glsl_dcl_indexable_temp(gen, ins); break; @@ -1495,6 +1524,12 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *gen, case VSIR_OP_DP4: shader_glsl_dot(gen, ins, VKD3DSP_WRITEMASK_ALL); break; + case VSIR_OP_DSX: + shader_glsl_intrinsic(gen, ins, "dFdx"); + break; + case VSIR_OP_DSY: + shader_glsl_intrinsic(gen, ins, "dFdy"); + break; case VSIR_OP_ELSE: shader_glsl_else(gen, ins); break; @@ -1531,6 +1566,7 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *gen, break; case VSIR_OP_GEO: case VSIR_OP_IGE: + case VSIR_OP_UGE: shader_glsl_relop(gen, ins, ">=", "greaterThanEqual"); break; case VSIR_OP_IF: @@ -1620,6 +1656,9 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *gen, case VSIR_OP_RSQ: shader_glsl_intrinsic(gen, ins, "inversesqrt"); break; + case VSIR_OP_SIN: + shader_glsl_intrinsic(gen, ins, "sin"); + break; case VSIR_OP_SQRT: shader_glsl_intrinsic(gen, ins, "sqrt"); break; @@ -1629,6 +1668,12 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *gen, case VSIR_OP_SWITCH: shader_glsl_switch(gen, ins); break; + case VSIR_OP_UDIV_SIMPLE: + shader_glsl_binop(gen, ins, "/"); + break; + case VSIR_OP_UREM: + shader_glsl_binop(gen, ins, "%"); + break; case VSIR_OP_XOR: shader_glsl_binop(gen, ins, "^"); break; @@ -2318,6 +2363,8 @@ static void shader_glsl_generate_declarations(struct vkd3d_glsl_generator *gen) vkd3d_string_buffer_printf(buffer, "vec4 %s_in[%u];\n", gen->prefix, gen->limits.input_count); if (gen->limits.output_count) vkd3d_string_buffer_printf(buffer, "vec4 %s_out[%u];\n", gen->prefix, gen->limits.output_count); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_SAMPLEMASK)) + vkd3d_string_buffer_printf(gen->buffer, "float o_mask;\n"); if (program->temp_count) vkd3d_string_buffer_printf(buffer, "vec4 r[%u];\n", program->temp_count); vkd3d_string_buffer_printf(buffer, "\n"); @@ -2328,7 +2375,6 @@ static int vkd3d_glsl_generator_generate(struct vkd3d_glsl_generator *gen, struc struct vkd3d_string_buffer *buffer = gen->buffer; struct vkd3d_shader_instruction *ins; struct vsir_program_iterator it; - void *code; MESSAGE("Generating a GLSL shader. This is unsupported; you get to keep all the pieces if it breaks.\n"); @@ -2357,13 +2403,7 @@ static int vkd3d_glsl_generator_generate(struct vkd3d_glsl_generator *gen, struc if (gen->failed) return VKD3D_ERROR_INVALID_SHADER; - if ((code = vkd3d_malloc(buffer->buffer_size))) - { - memcpy(code, buffer->buffer, buffer->content_size); - out->size = buffer->content_size; - out->code = code; - } - else return VKD3D_ERROR_OUT_OF_MEMORY; + vkd3d_shader_code_from_string_buffer(out, buffer); return VKD3D_OK; } diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.c b/libs/vkd3d/libs/vkd3d-shader/hlsl.c index 62335086e20..ec1e27d9496 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl.c +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.c @@ -393,6 +393,11 @@ bool hlsl_type_is_signed_integer(const struct hlsl_type *type) vkd3d_unreachable(); } +bool hlsl_type_is_unsigned_integer(const struct hlsl_type *type) +{ + return hlsl_type_is_integer(type) && !hlsl_type_is_signed_integer(type); +} + bool hlsl_type_is_integer(const struct hlsl_type *type) { VKD3D_ASSERT(hlsl_is_numeric_type(type)); @@ -1809,6 +1814,76 @@ struct hlsl_ir_node *hlsl_new_null_constant(struct hlsl_ctx *ctx, const struct v return hlsl_new_constant(ctx, ctx->builtin_types.null, &value, loc); } +bool hlsl_constant_is_zero(struct hlsl_ir_constant *c) +{ + struct hlsl_type *data_type = c->node.data_type; + unsigned int k; + + for (k = 0; k < data_type->e.numeric.dimx; ++k) + { + switch (data_type->e.numeric.type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + if (c->value.u[k].f != 0.0f) + return false; + break; + + case HLSL_TYPE_DOUBLE: + if (c->value.u[k].d != 0.0) + return false; + break; + + case HLSL_TYPE_UINT: + case HLSL_TYPE_INT: + case HLSL_TYPE_BOOL: + case HLSL_TYPE_MIN16UINT: + if (c->value.u[k].u != 0) + return false; + break; + } + } + + return true; +} + +bool hlsl_constant_is_one(struct hlsl_ir_constant *c) +{ + struct hlsl_type *data_type = c->node.data_type; + unsigned int k; + + for (k = 0; k < data_type->e.numeric.dimx; ++k) + { + switch (data_type->e.numeric.type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + if (c->value.u[k].f != 1.0f) + return false; + break; + + case HLSL_TYPE_DOUBLE: + if (c->value.u[k].d != 1.0) + return false; + break; + + case HLSL_TYPE_UINT: + case HLSL_TYPE_INT: + case HLSL_TYPE_MIN16UINT: + if (c->value.u[k].u != 1) + return false; + break; + + case HLSL_TYPE_BOOL: + if (c->value.u[k].u != ~0) + return false; + break; + } + } + + return true; +} + static struct hlsl_ir_node *hlsl_new_expr(struct hlsl_ctx *ctx, enum hlsl_ir_expr_op op, struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS], struct hlsl_type *data_type, const struct vkd3d_shader_location *loc) @@ -3654,8 +3729,11 @@ const char *debug_hlsl_expr_op(enum hlsl_ir_expr_op op) [HLSL_OP1_BIT_NOT] = "~", [HLSL_OP1_CAST] = "cast", [HLSL_OP1_CEIL] = "ceil", + [HLSL_OP1_CLZ] = "clz", [HLSL_OP1_COS] = "cos", [HLSL_OP1_COS_REDUCED] = "cos_reduced", + [HLSL_OP1_COUNTBITS] = "countbits", + [HLSL_OP1_CTZ] = "ctz", [HLSL_OP1_DSX] = "dsx", [HLSL_OP1_DSX_COARSE] = "dsx_coarse", [HLSL_OP1_DSX_FINE] = "dsx_fine", @@ -3665,6 +3743,7 @@ const char *debug_hlsl_expr_op(enum hlsl_ir_expr_op op) [HLSL_OP1_EXP2] = "exp2", [HLSL_OP1_F16TOF32] = "f16tof32", [HLSL_OP1_F32TOF16] = "f32tof16", + [HLSL_OP1_FIND_MSB] = "find_msb", [HLSL_OP1_FLOOR] = "floor", [HLSL_OP1_FRACT] = "fract", [HLSL_OP1_ISINF] = "isinf", @@ -3718,8 +3797,9 @@ static void dump_ir_expr(struct vkd3d_string_buffer *buffer, const struct hlsl_i vkd3d_string_buffer_printf(buffer, "%s (", debug_hlsl_expr_op(expr->op)); for (i = 0; i < HLSL_MAX_OPERANDS && expr->operands[i].node; ++i) { + if (i) + vkd3d_string_buffer_printf(buffer, " "); dump_src(buffer, &expr->operands[i]); - vkd3d_string_buffer_printf(buffer, " "); } vkd3d_string_buffer_printf(buffer, ")"); } diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.h b/libs/vkd3d/libs/vkd3d-shader/hlsl.h index d67f820fe8b..8dbfd062177 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.h @@ -233,6 +233,9 @@ struct hlsl_type /* Offset where the type's description starts in the output bytecode, in bytes. */ size_t bytecode_offset; + /* Offset where the type's packed description starts in the output bytecode, in bytes. */ + size_t packed_bytecode_offset; + bool is_typedef; uint32_t is_minimum_precision : 1; @@ -715,8 +718,11 @@ enum hlsl_ir_expr_op HLSL_OP1_BIT_NOT, HLSL_OP1_CAST, HLSL_OP1_CEIL, + HLSL_OP1_CLZ, HLSL_OP1_COS, HLSL_OP1_COS_REDUCED, /* Reduced range [-pi, pi], writes to .x */ + HLSL_OP1_COUNTBITS, + HLSL_OP1_CTZ, HLSL_OP1_DSX, HLSL_OP1_DSX_COARSE, HLSL_OP1_DSX_FINE, @@ -726,6 +732,7 @@ enum hlsl_ir_expr_op HLSL_OP1_EXP2, HLSL_OP1_F16TOF32, HLSL_OP1_F32TOF16, + HLSL_OP1_FIND_MSB, HLSL_OP1_FLOOR, HLSL_OP1_FRACT, HLSL_OP1_ISINF, @@ -1184,8 +1191,8 @@ struct hlsl_ctx } constant_defs; /* 'c' registers where the constants expected by SM2 sincos are stored. */ struct hlsl_reg d3dsincosconst1, d3dsincosconst2; - /* Number of allocated SSA IDs, used in translation to vsir. */ - unsigned int ssa_count; + /* Number of allocated registers, used in translation to vsir. */ + unsigned int ssa_count, temp_count, indexable_temp_count; /* Number of threads to be executed (on the X, Y, and Z dimensions) in a single thread group in * compute shader profiles. It is set using the numthreads() attribute in the entry point. */ @@ -1541,6 +1548,12 @@ static inline bool hlsl_var_has_buffer_offset_register_reservation(struct hlsl_c return var->reg_reservation.reg_type == 'c' && var->buffer == ctx->globals_buffer; } +static inline bool hlsl_is_comparison_op(enum hlsl_ir_expr_op op) +{ + return op == HLSL_OP2_EQUAL || op == HLSL_OP2_GEQUAL + || op == HLSL_OP2_LESS || op == HLSL_OP2_NEQUAL; +} + char *hlsl_sprintf_alloc(struct hlsl_ctx *ctx, const char *fmt, ...) VKD3D_PRINTF_FUNC(2, 3); const char *debug_hlsl_expr_op(enum hlsl_ir_expr_op op); @@ -1693,6 +1706,9 @@ struct hlsl_type *hlsl_new_stream_output_type(struct hlsl_ctx *ctx, struct hlsl_ir_node *hlsl_new_ternary_expr(struct hlsl_ctx *ctx, enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2, struct hlsl_ir_node *arg3); +bool hlsl_constant_is_zero(struct hlsl_ir_constant *c); +bool hlsl_constant_is_one(struct hlsl_ir_constant *c); + void hlsl_init_simple_deref_from_var(struct hlsl_deref *deref, struct hlsl_ir_var *var); struct hlsl_ir_load *hlsl_new_var_load(struct hlsl_ctx *ctx, struct hlsl_ir_var *var, @@ -1779,10 +1795,10 @@ bool hlsl_type_is_integer(const struct hlsl_type *type); bool hlsl_type_is_floating_point(const struct hlsl_type *type); bool hlsl_type_is_row_major(const struct hlsl_type *type); bool hlsl_type_is_signed_integer(const struct hlsl_type *type); +bool hlsl_type_is_unsigned_integer(const struct hlsl_type *type); unsigned int hlsl_type_minor_size(const struct hlsl_type *type); unsigned int hlsl_type_major_size(const struct hlsl_type *type); unsigned int hlsl_type_element_count(const struct hlsl_type *type); -bool hlsl_type_is_integer(const struct hlsl_type *type); bool hlsl_type_is_minimum_precision(const struct hlsl_type *type); bool hlsl_type_is_resource(const struct hlsl_type *type); bool hlsl_type_is_shader(const struct hlsl_type *type); diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.l b/libs/vkd3d/libs/vkd3d-shader/hlsl.l index 0cdebb8a657..da9f0d39136 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl.l +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.l @@ -346,7 +346,7 @@ while {return KW_WHILE; } {ANY} {} {ANY} { - return yytext[0]; + return (unsigned char)yytext[0]; } %% diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.y b/libs/vkd3d/libs/vkd3d-shader/hlsl.y index 024d96c5663..d3004d7cc8a 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.y @@ -3181,6 +3181,20 @@ static bool elementwise_intrinsic_float_convert_args(struct hlsl_ctx *ctx, return true; } +static bool elementwise_intrinsic_int_convert_args(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_type *type; + + if (!(type = elementwise_intrinsic_get_common_type(ctx, params, loc))) + return false; + + type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_INT, type->e.numeric.dimx, type->e.numeric.dimy); + + convert_args(ctx, params, type, loc); + return true; +} + static bool elementwise_intrinsic_uint_convert_args(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -3579,6 +3593,20 @@ static bool intrinsic_cosh(struct hlsl_ctx *ctx, return write_cosh_or_sinh(ctx, params, loc, false); } +static bool intrinsic_countbits(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {0}; + struct hlsl_type *type; + + if (!elementwise_intrinsic_uint_convert_args(ctx, params, loc)) + return false; + type = convert_numeric_type(ctx, params->args[0]->data_type, HLSL_TYPE_UINT); + + operands[0] = params->args[0]; + return add_expr(ctx, params->instrs, HLSL_OP1_COUNTBITS, operands, type, loc); +} + static bool intrinsic_cross(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -3925,6 +3953,76 @@ static bool intrinsic_f32tof16(struct hlsl_ctx *ctx, return add_expr(ctx, params->instrs, HLSL_OP1_F32TOF16, operands, type, loc); } +static bool intrinsic_firstbithigh(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {0}; + struct hlsl_type *type = params->args[0]->data_type; + struct hlsl_ir_node *c, *clz, *eq, *xor; + struct hlsl_constant_value v; + + if (hlsl_version_lt(ctx, 4, 0)) + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INCOMPATIBLE_PROFILE, + "The 'firstbithigh' intrinsic requires shader model 4.0 or higher."); + + if (hlsl_type_is_unsigned_integer(type)) + { + if (!elementwise_intrinsic_uint_convert_args(ctx, params, loc)) + return false; + } + else + { + if (!elementwise_intrinsic_int_convert_args(ctx, params, loc)) + return false; + } + type = convert_numeric_type(ctx, type, HLSL_TYPE_UINT); + + operands[0] = params->args[0]; + if (hlsl_version_lt(ctx, 5, 0)) + return add_expr(ctx, params->instrs, HLSL_OP1_FIND_MSB, operands, type, loc); + + v.u[0].u = 0x1f; + if (!(c = hlsl_new_constant(ctx, hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), &v, loc))) + return false; + hlsl_block_add_instr(params->instrs, c); + + if (!(clz = add_expr(ctx, params->instrs, HLSL_OP1_CLZ, operands, type, loc))) + return false; + if (!(xor = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_BIT_XOR, c, clz, loc))) + return false; + + v.u[0].i = -1; + if (!(c = hlsl_new_constant(ctx, hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), &v, loc))) + return false; + hlsl_block_add_instr(params->instrs, c); + + if (!(eq = add_binary_comparison_expr(ctx, params->instrs, HLSL_OP2_EQUAL, clz, c, loc))) + return false; + + operands[0] = eq; + operands[1] = add_implicit_conversion(ctx, params->instrs, c, type, loc); + operands[2] = xor; + return add_expr(ctx, params->instrs, HLSL_OP3_TERNARY, operands, type, loc); +} + +static bool intrinsic_firstbitlow(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {0}; + struct hlsl_type *type; + + if (hlsl_version_lt(ctx, 4, 0)) + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INCOMPATIBLE_PROFILE, + "The 'firstbitlow' intrinsic requires shader model 4.0 or higher."); + + if (!elementwise_intrinsic_uint_convert_args(ctx, params, loc)) + return false; + type = convert_numeric_type(ctx, params->args[0]->data_type, HLSL_TYPE_UINT); + + operands[0] = params->args[0]; + return add_expr(ctx, params->instrs, HLSL_OP1_CTZ, operands, type, loc); +} + static bool intrinsic_floor(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -3983,6 +4081,53 @@ static bool intrinsic_frac(struct hlsl_ctx *ctx, return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_FRACT, arg, loc); } +static bool intrinsic_frexp(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + struct hlsl_type *type, *uint_dim_type, *int_dim_type, *bool_dim_type; + struct hlsl_ir_function_decl *func; + char *body; + + static const char template[] = + "%s frexp(%s x, out %s exp)\n" + "{\n" + /* If x is zero, always return zero for exp and mantissa. */ + " %s is_nonzero_mask = x != 0.0;\n" + " %s bits = asuint(x);\n" + /* Subtract 126, not 127, to increase the exponent */ + " %s exp_int = asint((bits & 0x7f800000u) >> 23) - 126;\n" + /* Clear the given exponent and replace it with the bit pattern + * for 2^-1 */ + " %s mantissa = asfloat((bits & 0x007fffffu) | 0x3f000000);\n" + " exp = is_nonzero_mask * %s(exp_int);\n" + " return is_nonzero_mask * mantissa;\n" + "}\n"; + + if (!elementwise_intrinsic_float_convert_args(ctx, params, loc)) + return false; + type = params->args[0]->data_type; + + if (type->e.numeric.type == HLSL_TYPE_DOUBLE) + { + hlsl_fixme(ctx, loc, "frexp() on doubles."); + return false; + } + type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_FLOAT, type->e.numeric.dimx, type->e.numeric.dimy); + uint_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_UINT, type->e.numeric.dimx, type->e.numeric.dimy); + int_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_INT, type->e.numeric.dimx, type->e.numeric.dimy); + bool_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_BOOL, type->e.numeric.dimx, type->e.numeric.dimy); + + if (!(body = hlsl_sprintf_alloc(ctx, template, type->name, type->name, type->name, + bool_dim_type->name, uint_dim_type->name, int_dim_type->name, type->name, type->name))) + return false; + func = hlsl_compile_internal_function(ctx, "frexp", body); + vkd3d_free(body); + if (!func) + return false; + + return !!add_user_call(ctx, func, params, false, loc); +} + static bool intrinsic_fwidth(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -4701,7 +4846,8 @@ static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer * } if (!strcmp(name, "tex2Dbias") - || !strcmp(name, "tex2Dlod")) + || !strcmp(name, "tex2Dlod") + || !strcmp(name, "texCUBEbias")) { struct hlsl_ir_node *lod, *c; @@ -4853,6 +4999,12 @@ static bool intrinsic_texCUBE(struct hlsl_ctx *ctx, return intrinsic_tex(ctx, params, loc, "texCUBE", HLSL_SAMPLER_DIM_CUBE); } +static bool intrinsic_texCUBEbias(struct hlsl_ctx *ctx, + const struct parse_initializer *params, const struct vkd3d_shader_location *loc) +{ + return intrinsic_tex(ctx, params, loc, "texCUBEbias", HLSL_SAMPLER_DIM_CUBE); +} + static bool intrinsic_texCUBEgrad(struct hlsl_ctx *ctx, const struct parse_initializer *params, const struct vkd3d_shader_location *loc) { @@ -5263,6 +5415,7 @@ intrinsic_functions[] = {"clip", 1, true, intrinsic_clip}, {"cos", 1, true, intrinsic_cos}, {"cosh", 1, true, intrinsic_cosh}, + {"countbits", 1, true, intrinsic_countbits}, {"cross", 2, true, intrinsic_cross}, {"ddx", 1, true, intrinsic_ddx}, {"ddx_coarse", 1, true, intrinsic_ddx_coarse}, @@ -5280,9 +5433,12 @@ intrinsic_functions[] = {"f16tof32", 1, true, intrinsic_f16tof32}, {"f32tof16", 1, true, intrinsic_f32tof16}, {"faceforward", 3, true, intrinsic_faceforward}, + {"firstbithigh", 1, true, intrinsic_firstbithigh}, + {"firstbitlow", 1, true, intrinsic_firstbitlow}, {"floor", 1, true, intrinsic_floor}, {"fmod", 2, true, intrinsic_fmod}, {"frac", 1, true, intrinsic_frac}, + {"frexp", 2, true, intrinsic_frexp}, {"fwidth", 1, true, intrinsic_fwidth}, {"isinf", 1, true, intrinsic_isinf}, {"ldexp", 2, true, intrinsic_ldexp}, @@ -5327,6 +5483,7 @@ intrinsic_functions[] = {"tex3Dgrad", 4, false, intrinsic_tex3Dgrad}, {"tex3Dproj", 2, false, intrinsic_tex3Dproj}, {"texCUBE", -1, false, intrinsic_texCUBE}, + {"texCUBEbias", 2, false, intrinsic_texCUBEbias}, {"texCUBEgrad", 4, false, intrinsic_texCUBEgrad}, {"texCUBEproj", 2, false, intrinsic_texCUBEproj}, {"transpose", 1, true, intrinsic_transpose}, @@ -8059,7 +8216,7 @@ resource_format: { uint32_t modifiers = $1; - if (!($$ = apply_type_modifiers(ctx, $2, &modifiers, false, &@1))) + if (!($$ = apply_type_modifiers(ctx, $2, &modifiers, true, &@1))) YYABORT; } diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c index 0b3dee4d2ce..2138c3776a7 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c @@ -175,12 +175,29 @@ static unsigned int hlsl_type_get_packed_size(const struct hlsl_type *type) } } +static unsigned int struct_field_get_packed_offset(const struct hlsl_type *record, unsigned int field_idx) +{ + unsigned int offset = 0; + + VKD3D_ASSERT(record->class == HLSL_CLASS_STRUCT); + VKD3D_ASSERT(field_idx < record->e.record.field_count); + + for (unsigned int i = 0; i < field_idx; ++i) + { + struct hlsl_struct_field *field = &record->e.record.fields[i]; + offset = align(offset, hlsl_type_get_packed_alignment(field->type)) + hlsl_type_get_packed_size(field->type); + } + + return align(offset, hlsl_type_get_packed_alignment(record->e.record.fields[field_idx].type)); +} + + static struct hlsl_ir_node *hlsl_block_add_packed_index_offset_append(struct hlsl_ctx *ctx, struct hlsl_block *block, struct hlsl_ir_node *prev_offset, struct hlsl_ir_node *idx, struct hlsl_type *type, const struct vkd3d_shader_location *loc) { struct hlsl_ir_node *idx_offset = NULL, *c; - unsigned int field_idx, offset, size, i; + unsigned int field_idx, offset, size; switch (type->class) { @@ -203,15 +220,7 @@ static struct hlsl_ir_node *hlsl_block_add_packed_index_offset_append(struct hls case HLSL_CLASS_STRUCT: field_idx = hlsl_ir_constant(idx)->value.u[0].u; - for (i = 0, offset = 0; i < field_idx; ++i) - { - struct hlsl_struct_field *field = &type->e.record.fields[i]; - - offset = align(offset, hlsl_type_get_packed_alignment(field->type)) - + hlsl_type_get_packed_size(field->type); - } - - offset = align(offset, hlsl_type_get_packed_alignment(type->e.record.fields[field_idx].type)); + offset = struct_field_get_packed_offset(type, field_idx); idx_offset = hlsl_block_add_uint_constant(ctx, block, offset, loc); break; @@ -5029,8 +5038,7 @@ static bool lower_comparison_operators(struct hlsl_ctx *ctx, struct hlsl_ir_node if (instr->type != HLSL_IR_EXPR) return false; expr = hlsl_ir_expr(instr); - if (expr->op != HLSL_OP2_EQUAL && expr->op != HLSL_OP2_NEQUAL && expr->op != HLSL_OP2_LESS - && expr->op != HLSL_OP2_GEQUAL) + if (!hlsl_is_comparison_op(expr->op)) return false; arg1 = expr->operands[0].node; @@ -5906,11 +5914,13 @@ static void mark_vars_usage(struct hlsl_ctx *ctx) struct register_allocator { + /* Type of registers we are allocating (not counting indexable temps). */ + enum vkd3d_shader_register_type type; + struct allocation { uint32_t reg; unsigned int writemask; - unsigned int first_write, last_read; /* Two allocations with different mode can't share the same register. */ int mode; @@ -5920,11 +5930,7 @@ struct register_allocator } *allocations; size_t count, capacity; - /* Indexable temps are allocated separately and always keep their index regardless of their - * lifetime. */ - uint32_t indexable_count; - - /* Total number of registers allocated so far. Used to declare sm4 temp count. */ + /* Total number of registers allocated so far. */ uint32_t reg_count; /* Special flag so allocations that can share registers prioritize those @@ -5936,7 +5942,7 @@ struct register_allocator }; static unsigned int get_available_writemask(const struct register_allocator *allocator, - unsigned int first_write, unsigned int last_read, uint32_t reg_idx, int mode, bool vip) + uint32_t reg_idx, int mode, bool vip) { unsigned int writemask = VKD3DSP_WRITEMASK_ALL; size_t i; @@ -5945,12 +5951,7 @@ static unsigned int get_available_writemask(const struct register_allocator *all { const struct allocation *allocation = &allocator->allocations[i]; - /* We do not overlap if first write == last read: - * this is the case where we are allocating the result of that - * expression, e.g. "add r0, r0, r1". */ - - if (allocation->reg == reg_idx - && first_write < allocation->last_read && last_read > allocation->first_write) + if (allocation->reg == reg_idx) { writemask &= ~allocation->writemask; if (allocation->mode != mode) @@ -5967,7 +5968,7 @@ static unsigned int get_available_writemask(const struct register_allocator *all } static void record_allocation(struct hlsl_ctx *ctx, struct register_allocator *allocator, uint32_t reg_idx, - unsigned int writemask, unsigned int first_write, unsigned int last_read, int mode, bool vip) + unsigned int writemask, int mode, bool vip) { struct allocation *allocation; @@ -5978,8 +5979,6 @@ static void record_allocation(struct hlsl_ctx *ctx, struct register_allocator *a allocation = &allocator->allocations[allocator->count++]; allocation->reg = reg_idx; allocation->writemask = writemask; - allocation->first_write = first_write; - allocation->last_read = last_read; allocation->mode = mode; allocation->vip = vip; @@ -5998,8 +5997,7 @@ static void record_allocation(struct hlsl_ctx *ctx, struct register_allocator *a * 'vip' can be used so that no new allocations can be made in the given register * unless they are 'vip' as well. */ static struct hlsl_reg allocate_register(struct hlsl_ctx *ctx, struct register_allocator *allocator, - unsigned int first_write, unsigned int last_read, unsigned int reg_size, - unsigned int component_count, int mode, bool force_align, bool vip) + unsigned int reg_size, unsigned int component_count, int mode, bool force_align, bool vip) { struct hlsl_reg ret = {.allocation_size = 1, .allocated = true}; unsigned int required_size = force_align ? 4 : reg_size; @@ -6012,62 +6010,34 @@ static struct hlsl_reg allocate_register(struct hlsl_ctx *ctx, struct register_a { for (uint32_t reg_idx = 0; reg_idx < allocator->reg_count; ++reg_idx) { - unsigned int available_writemask = get_available_writemask(allocator, - first_write, last_read, reg_idx, mode, vip); + unsigned int available_writemask = get_available_writemask(allocator, reg_idx, mode, vip); if (vkd3d_popcount(available_writemask) >= pref) { unsigned int writemask = hlsl_combine_writemasks(available_writemask, vkd3d_write_mask_from_component_count(reg_size)); - ret.type = VKD3DSPR_TEMP; + ret.type = allocator->type; ret.id = reg_idx; ret.writemask = hlsl_combine_writemasks(writemask, vkd3d_write_mask_from_component_count(component_count)); - record_allocation(ctx, allocator, reg_idx, writemask, first_write, last_read, mode, vip); + record_allocation(ctx, allocator, reg_idx, writemask, mode, vip); return ret; } } } - ret.type = VKD3DSPR_TEMP; + ret.type = allocator->type; ret.id = allocator->reg_count; ret.writemask = vkd3d_write_mask_from_component_count(component_count); record_allocation(ctx, allocator, allocator->reg_count, - vkd3d_write_mask_from_component_count(reg_size), first_write, last_read, mode, vip); - return ret; -} - -/* Allocate a register with writemask, while reserving reg_writemask. */ -static struct hlsl_reg allocate_register_with_masks(struct hlsl_ctx *ctx, - struct register_allocator *allocator, unsigned int first_write, unsigned int last_read, - uint32_t reg_writemask, uint32_t writemask, int mode, bool vip) -{ - struct hlsl_reg ret = {0}; - uint32_t reg_idx; - - VKD3D_ASSERT((reg_writemask & writemask) == writemask); - - for (reg_idx = 0;; ++reg_idx) - { - if ((get_available_writemask(allocator, first_write, last_read, - reg_idx, mode, vip) & reg_writemask) == reg_writemask) - break; - } - - record_allocation(ctx, allocator, reg_idx, reg_writemask, first_write, last_read, mode, vip); - - ret.type = VKD3DSPR_TEMP; - ret.id = reg_idx; - ret.allocation_size = 1; - ret.writemask = writemask; - ret.allocated = true; + vkd3d_write_mask_from_component_count(reg_size), mode, vip); return ret; } -static bool is_range_available(const struct register_allocator *allocator, unsigned int first_write, - unsigned int last_read, uint32_t reg_idx, unsigned int reg_size, int mode, bool vip) +static bool is_range_available(const struct register_allocator *allocator, + uint32_t reg_idx, unsigned int reg_size, int mode, bool vip) { unsigned int last_reg_mask = (1u << (reg_size % 4)) - 1; unsigned int writemask; @@ -6075,18 +6045,18 @@ static bool is_range_available(const struct register_allocator *allocator, unsig for (i = 0; i < (reg_size / 4); ++i) { - writemask = get_available_writemask(allocator, first_write, last_read, reg_idx + i, mode, vip); + writemask = get_available_writemask(allocator, reg_idx + i, mode, vip); if (writemask != VKD3DSP_WRITEMASK_ALL) return false; } - writemask = get_available_writemask(allocator, first_write, last_read, reg_idx + (reg_size / 4), mode, vip); + writemask = get_available_writemask(allocator, reg_idx + (reg_size / 4), mode, vip); if ((writemask & last_reg_mask) != last_reg_mask) return false; return true; } -static struct hlsl_reg allocate_range(struct hlsl_ctx *ctx, struct register_allocator *allocator, - unsigned int first_write, unsigned int last_read, unsigned int reg_size, int mode, bool vip) +static struct hlsl_reg allocate_range(struct hlsl_ctx *ctx, + struct register_allocator *allocator, unsigned int reg_size, int mode, bool vip) { struct hlsl_reg ret = {0}; uint32_t reg_idx; @@ -6094,35 +6064,33 @@ static struct hlsl_reg allocate_range(struct hlsl_ctx *ctx, struct register_allo for (reg_idx = 0;; ++reg_idx) { - if (is_range_available(allocator, first_write, last_read, reg_idx, reg_size, mode, vip)) + if (is_range_available(allocator, reg_idx, reg_size, mode, vip)) break; } for (i = 0; i < reg_size / 4; ++i) - record_allocation(ctx, allocator, reg_idx + i, VKD3DSP_WRITEMASK_ALL, first_write, last_read, mode, vip); + record_allocation(ctx, allocator, reg_idx + i, VKD3DSP_WRITEMASK_ALL, mode, vip); if (reg_size % 4) - record_allocation(ctx, allocator, reg_idx + (reg_size / 4), - (1u << (reg_size % 4)) - 1, first_write, last_read, mode, vip); + record_allocation(ctx, allocator, reg_idx + (reg_size / 4), (1u << (reg_size % 4)) - 1, mode, vip); - ret.type = VKD3DSPR_TEMP; + ret.type = allocator->type; ret.id = reg_idx; ret.allocation_size = align(reg_size, 4) / 4; ret.allocated = true; return ret; } -static struct hlsl_reg allocate_numeric_registers_for_type(struct hlsl_ctx *ctx, struct register_allocator *allocator, - unsigned int first_write, unsigned int last_read, const struct hlsl_type *type) +static struct hlsl_reg allocate_numeric_registers_for_type(struct hlsl_ctx *ctx, + struct register_allocator *allocator, const struct hlsl_type *type) { unsigned int reg_size = type->reg_size[HLSL_REGSET_NUMERIC]; /* FIXME: We could potentially pack structs or arrays more efficiently... */ if (type->class <= HLSL_CLASS_VECTOR) - return allocate_register(ctx, allocator, first_write, last_read, - type->e.numeric.dimx, type->e.numeric.dimx, 0, false, false); + return allocate_register(ctx, allocator, type->e.numeric.dimx, type->e.numeric.dimx, 0, false, false); else - return allocate_range(ctx, allocator, first_write, last_read, reg_size, 0, false); + return allocate_range(ctx, allocator, reg_size, 0, false); } static const char *debug_register(struct hlsl_reg reg, const struct hlsl_type *type) @@ -6287,10 +6255,9 @@ static void calculate_resource_register_counts(struct hlsl_ctx *ctx) } } -static void allocate_instr_temp_register(struct hlsl_ctx *ctx, - struct hlsl_ir_node *instr, struct register_allocator *allocator) +static void allocate_instr_temp_register(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr) { - unsigned int reg_writemask = 0, dst_writemask = 0; + unsigned int dst_writemask = 0; bool is_per_component = false; if (instr->reg.allocated || !instr->last_read) @@ -6302,12 +6269,10 @@ static void allocate_instr_temp_register(struct hlsl_ctx *ctx, { case HLSL_OP1_COS_REDUCED: dst_writemask = VKD3DSP_WRITEMASK_0; - reg_writemask = ctx->profile->major_version < 3 ? (1 << 3) - 1 : VKD3DSP_WRITEMASK_0; break; case HLSL_OP1_SIN_REDUCED: dst_writemask = VKD3DSP_WRITEMASK_1; - reg_writemask = ctx->profile->major_version < 3 ? (1 << 3) - 1 : VKD3DSP_WRITEMASK_1; break; case HLSL_OP1_EXP2: @@ -6329,12 +6294,22 @@ static void allocate_instr_temp_register(struct hlsl_ctx *ctx, VKD3D_ASSERT(instr->data_type->class <= HLSL_CLASS_VECTOR); - if (reg_writemask) - instr->reg = allocate_register_with_masks(ctx, allocator, - instr->index, instr->last_read, reg_writemask, dst_writemask, 0, false); + if (dst_writemask) + { + instr->reg.writemask = dst_writemask; + instr->reg.allocation_size = 1; + instr->reg.allocated = true; + instr->reg.type = VKD3DSPR_TEMP; + instr->reg.id = ctx->temp_count++; + } else if (is_per_component) - instr->reg = allocate_numeric_registers_for_type(ctx, allocator, - instr->index, instr->last_read, instr->data_type); + { + instr->reg.writemask = vkd3d_write_mask_from_component_count(instr->data_type->e.numeric.dimx); + instr->reg.allocation_size = 1; + instr->reg.allocated = true; + instr->reg.type = VKD3DSPR_TEMP; + instr->reg.id = ctx->temp_count++; + } else { instr->reg.writemask = vkd3d_write_mask_from_component_count(instr->data_type->e.numeric.dimx); @@ -6344,40 +6319,46 @@ static void allocate_instr_temp_register(struct hlsl_ctx *ctx, instr->reg.id = ctx->ssa_count++; } - TRACE("Allocated anonymous expression @%u to %s (liveness %u-%u).\n", instr->index, - debug_register(instr->reg, instr->data_type), instr->index, instr->last_read); + TRACE("Allocated anonymous expression @%u to %s.\n", instr->index, + debug_register(instr->reg, instr->data_type)); } -static void allocate_variable_temp_register(struct hlsl_ctx *ctx, - struct hlsl_ir_var *var, struct register_allocator *allocator) +static void allocate_variable_temp_register(struct hlsl_ctx *ctx, struct hlsl_ir_var *var) { + struct hlsl_reg *reg = &var->regs[HLSL_REGSET_NUMERIC]; + if (var->is_input_semantic || var->is_output_semantic || var->is_uniform) return; - if (!var->regs[HLSL_REGSET_NUMERIC].allocated && var->last_read) + if (!reg->allocated && var->last_read) { if (var->indexable) { - var->regs[HLSL_REGSET_NUMERIC].id = allocator->indexable_count++; - var->regs[HLSL_REGSET_NUMERIC].allocation_size = 1; - var->regs[HLSL_REGSET_NUMERIC].writemask = 0; - var->regs[HLSL_REGSET_NUMERIC].allocated = true; + reg->id = ctx->indexable_temp_count++; + reg->allocation_size = 1; + reg->writemask = 0; + reg->allocated = true; - TRACE("Allocated %s to x%u[].\n", var->name, var->regs[HLSL_REGSET_NUMERIC].id); + TRACE("Allocated %s to x%u[].\n", var->name, reg->id); } else { - var->regs[HLSL_REGSET_NUMERIC] = allocate_numeric_registers_for_type(ctx, allocator, - var->first_write, var->last_read, var->data_type); + reg->type = VKD3DSPR_TEMP; + reg->id = ctx->temp_count; + reg->allocation_size = align(var->data_type->reg_size[HLSL_REGSET_NUMERIC], 4) / 4; + if (var->data_type->class <= HLSL_CLASS_VECTOR) + reg->writemask = vkd3d_write_mask_from_component_count(var->data_type->e.numeric.dimx); + reg->allocated = true; + + ctx->temp_count += reg->allocation_size; - TRACE("Allocated %s to %s (liveness %u-%u).\n", var->name, - debug_register(var->regs[HLSL_REGSET_NUMERIC], var->data_type), var->first_write, var->last_read); + TRACE("Allocated %s to %s.\n", var->name, + debug_register(var->regs[HLSL_REGSET_NUMERIC], var->data_type)); } } } -static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx, - struct hlsl_block *block, struct register_allocator *allocator) +static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx, struct hlsl_block *block) { struct hlsl_ir_node *instr; @@ -6387,15 +6368,15 @@ static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx, if (ctx->profile->major_version >= 4 && instr->type == HLSL_IR_CONSTANT) continue; - allocate_instr_temp_register(ctx, instr, allocator); + allocate_instr_temp_register(ctx, instr); switch (instr->type) { case HLSL_IR_IF: { struct hlsl_ir_if *iff = hlsl_ir_if(instr); - allocate_temp_registers_recurse(ctx, &iff->then_block, allocator); - allocate_temp_registers_recurse(ctx, &iff->else_block, allocator); + allocate_temp_registers_recurse(ctx, &iff->then_block); + allocate_temp_registers_recurse(ctx, &iff->else_block); break; } @@ -6404,21 +6385,21 @@ static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx, struct hlsl_ir_load *load = hlsl_ir_load(instr); /* We need to at least allocate a variable for undefs. * FIXME: We should probably find a way to remove them instead. */ - allocate_variable_temp_register(ctx, load->src.var, allocator); + allocate_variable_temp_register(ctx, load->src.var); break; } case HLSL_IR_LOOP: { struct hlsl_ir_loop *loop = hlsl_ir_loop(instr); - allocate_temp_registers_recurse(ctx, &loop->body, allocator); + allocate_temp_registers_recurse(ctx, &loop->body); break; } case HLSL_IR_STORE: { struct hlsl_ir_store *store = hlsl_ir_store(instr); - allocate_variable_temp_register(ctx, store->lhs.var, allocator); + allocate_variable_temp_register(ctx, store->lhs.var); break; } @@ -6429,7 +6410,7 @@ static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx, LIST_FOR_EACH_ENTRY(c, &s->cases, struct hlsl_ir_switch_case, entry) { - allocate_temp_registers_recurse(ctx, &c->body, allocator); + allocate_temp_registers_recurse(ctx, &c->body); } break; } @@ -6553,8 +6534,7 @@ static void allocate_const_registers_recurse(struct hlsl_ctx *ctx, break; } - constant->reg = allocate_numeric_registers_for_type(ctx, allocator, 1, UINT_MAX, type); - constant->reg.type = VKD3DSPR_CONST; + constant->reg = allocate_numeric_registers_for_type(ctx, allocator, type); TRACE("Allocated constant @%u to %s.\n", instr->index, debug_register(constant->reg, type)); for (unsigned int x = 0, i = 0; x < 4; ++x) @@ -6651,16 +6631,14 @@ static void allocate_sincos_const_registers(struct hlsl_ctx *ctx, struct hlsl_bl { type = hlsl_get_vector_type(ctx, HLSL_TYPE_FLOAT, 4); - ctx->d3dsincosconst1 = allocate_numeric_registers_for_type(ctx, allocator, 1, UINT_MAX, type); - ctx->d3dsincosconst1.type = VKD3DSPR_CONST; + ctx->d3dsincosconst1 = allocate_numeric_registers_for_type(ctx, allocator, type); TRACE("Allocated D3DSINCOSCONST1 to %s.\n", debug_register(ctx->d3dsincosconst1, type)); record_constant(ctx, ctx->d3dsincosconst1.id * 4 + 0, -1.55009923e-06f, &instr->loc); record_constant(ctx, ctx->d3dsincosconst1.id * 4 + 1, -2.17013894e-05f, &instr->loc); record_constant(ctx, ctx->d3dsincosconst1.id * 4 + 2, 2.60416674e-03f, &instr->loc); record_constant(ctx, ctx->d3dsincosconst1.id * 4 + 3, 2.60416680e-04f, &instr->loc); - ctx->d3dsincosconst2 = allocate_numeric_registers_for_type(ctx, allocator, 1, UINT_MAX, type); - ctx->d3dsincosconst2.type = VKD3DSPR_CONST; + ctx->d3dsincosconst2 = allocate_numeric_registers_for_type(ctx, allocator, type); TRACE("Allocated D3DSINCOSCONST2 to %s.\n", debug_register(ctx->d3dsincosconst2, type)); record_constant(ctx, ctx->d3dsincosconst2.id * 4 + 0, -2.08333340e-02f, &instr->loc); record_constant(ctx, ctx->d3dsincosconst2.id * 4 + 1, -1.25000000e-01f, &instr->loc); @@ -6674,8 +6652,7 @@ static void allocate_sincos_const_registers(struct hlsl_ctx *ctx, struct hlsl_bl static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_block *body) { - struct register_allocator allocator_used = {0}; - struct register_allocator allocator = {0}; + struct register_allocator allocator = {.type = VKD3DSPR_CONST}, allocator_used = {.type = VKD3DSPR_CONST}; struct hlsl_ir_var *var; sort_uniforms_by_bind_count(ctx, HLSL_REGSET_NUMERIC); @@ -6698,15 +6675,14 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_block *bo { if (i < bind_count) { - if (get_available_writemask(&allocator_used, 1, UINT_MAX, - reg_idx + i, 0, false) != VKD3DSP_WRITEMASK_ALL) + if (get_available_writemask(&allocator_used, reg_idx + i, 0, false) != VKD3DSP_WRITEMASK_ALL) { hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, "Overlapping register() reservations on 'c%u'.", reg_idx + i); } - record_allocation(ctx, &allocator_used, reg_idx + i, VKD3DSP_WRITEMASK_ALL, 1, UINT_MAX, 0, false); + record_allocation(ctx, &allocator_used, reg_idx + i, VKD3DSP_WRITEMASK_ALL, 0, false); } - record_allocation(ctx, &allocator, reg_idx + i, VKD3DSP_WRITEMASK_ALL, 1, UINT_MAX, 0, false); + record_allocation(ctx, &allocator, reg_idx + i, VKD3DSP_WRITEMASK_ALL, 0, false); } var->regs[HLSL_REGSET_NUMERIC].type = VKD3DSPR_CONST; @@ -6730,8 +6706,7 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_block *bo if (!var->regs[HLSL_REGSET_NUMERIC].allocated) { - var->regs[HLSL_REGSET_NUMERIC] = allocate_range(ctx, &allocator, 1, UINT_MAX, alloc_size, 0, false); - var->regs[HLSL_REGSET_NUMERIC].type = VKD3DSPR_CONST; + var->regs[HLSL_REGSET_NUMERIC] = allocate_range(ctx, &allocator, alloc_size, 0, false); TRACE("Allocated %s to %s.\n", var->name, debug_register(var->regs[HLSL_REGSET_NUMERIC], var->data_type)); } @@ -6748,12 +6723,13 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_block *bo * index to all (simultaneously live) variables or intermediate values. Agnostic * as to how many registers are actually available for the current backend, and * does not handle constants. */ -static uint32_t allocate_temp_registers(struct hlsl_ctx *ctx, struct hlsl_block *body, struct list *semantic_vars) +static void allocate_temp_registers(struct hlsl_ctx *ctx, struct hlsl_block *body, struct list *semantic_vars) { - struct register_allocator allocator = {0}; struct hlsl_scope *scope; struct hlsl_ir_var *var; + ctx->indexable_temp_count = 0; + /* Reset variable temp register allocations. */ LIST_FOR_EACH_ENTRY(scope, &ctx->scopes, struct hlsl_scope, entry) { @@ -6771,25 +6747,13 @@ static uint32_t allocate_temp_registers(struct hlsl_ctx *ctx, struct hlsl_block { if (var->is_output_semantic) { - record_allocation(ctx, &allocator, 0, VKD3DSP_WRITEMASK_ALL, - var->first_write, UINT_MAX, 0, false); + ctx->temp_count = 1; break; } } } - allocate_temp_registers_recurse(ctx, body, &allocator); - vkd3d_free(allocator.allocations); - - if (allocator.indexable_count) - TRACE("Declaration of %s function required %u temp registers, and %u indexable temps.\n", - ctx->is_patch_constant_func ? "patch constant" : "main", - allocator.reg_count, allocator.indexable_count); - else - TRACE("Declaration of %s function required %u temp registers.\n", - ctx->is_patch_constant_func ? "patch constant" : "main", allocator.reg_count); - - return allocator.reg_count; + allocate_temp_registers_recurse(ctx, body); } static enum vkd3d_shader_interpolation_mode sm4_get_interpolation_mode(struct hlsl_type *type, @@ -6928,9 +6892,8 @@ static void allocate_semantic_register(struct hlsl_ctx *ctx, struct hlsl_ir_var if (special_interpolation) mode = VKD3DSIM_NONE; - var->regs[HLSL_REGSET_NUMERIC] = allocate_register(ctx, allocator, 1, UINT_MAX, + var->regs[HLSL_REGSET_NUMERIC] = allocate_register(ctx, allocator, reg_size, component_count, mode, var->force_align, vip_allocation); - var->regs[HLSL_REGSET_NUMERIC].type = output ? VKD3DSPR_OUTPUT : VKD3DSPR_INPUT; TRACE("Allocated %s to %s (mode %d).\n", var->name, debug_register(var->regs[HLSL_REGSET_NUMERIC], var->data_type), mode); @@ -6945,11 +6908,17 @@ static void allocate_semantic_registers(struct hlsl_ctx *ctx, struct list *seman bool is_pixel_shader = ctx->profile->type == VKD3D_SHADER_TYPE_PIXEL; struct hlsl_ir_var *var; + in_prim_allocator.type = VKD3DSPR_INPUT; in_prim_allocator.prioritize_smaller_writemasks = true; + patch_constant_out_patch_allocator.type = VKD3DSPR_INPUT; patch_constant_out_patch_allocator.prioritize_smaller_writemasks = true; + input_allocator.type = VKD3DSPR_INPUT; input_allocator.prioritize_smaller_writemasks = true; for (unsigned int i = 0; i < ARRAY_SIZE(output_allocators); ++i) + { + output_allocators[i].type = VKD3DSPR_OUTPUT; output_allocators[i].prioritize_smaller_writemasks = true; + } LIST_FOR_EACH_ENTRY(var, semantic_vars, struct hlsl_ir_var, extern_entry) { @@ -8266,6 +8235,282 @@ void hlsl_lower_index_loads(struct hlsl_ctx *ctx, struct hlsl_block *body) lower_ir(ctx, lower_index_loads, body); } +static enum hlsl_ir_expr_op invert_comparison_op(enum hlsl_ir_expr_op op) +{ + switch (op) + { + case HLSL_OP2_EQUAL: + return HLSL_OP2_NEQUAL; + + case HLSL_OP2_GEQUAL: + return HLSL_OP2_LESS; + + case HLSL_OP2_LESS: + return HLSL_OP2_GEQUAL; + + case HLSL_OP2_NEQUAL: + return HLSL_OP2_EQUAL; + + default: + vkd3d_unreachable(); + } +} + +static bool fold_unary_identities(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +{ + struct hlsl_ir_node *res = NULL; + struct hlsl_ir_expr *expr, *x; + + if (instr->type != HLSL_IR_EXPR) + return false; + + if (instr->data_type->class > HLSL_CLASS_VECTOR) + return false; + + expr = hlsl_ir_expr(instr); + if (!expr->operands[0].node) + return false; + + if (expr->operands[0].node->type != HLSL_IR_EXPR) + return false; + x = hlsl_ir_expr(expr->operands[0].node); + + switch (expr->op) + { + case HLSL_OP1_ABS: + if (x->op == HLSL_OP1_ABS) + { + /* ||x|| -> |x| */ + hlsl_replace_node(instr, &x->node); + return true; + } + + if (x->op == HLSL_OP1_NEG) + { + /* |-x| -> |x| */ + hlsl_src_remove(&expr->operands[0]); + hlsl_src_from_node(&expr->operands[0], x->operands[0].node); + return true; + } + break; + + case HLSL_OP1_BIT_NOT: + if (x->op == HLSL_OP1_BIT_NOT) + { + /* ~(~x) -> x */ + hlsl_replace_node(instr, x->operands[0].node); + return true; + } + break; + + case HLSL_OP1_CEIL: + case HLSL_OP1_FLOOR: + if (x->op == HLSL_OP1_CEIL || x->op == HLSL_OP1_FLOOR) + { + /* f(g(x)) -> g(x), where f(), g() are floor() or ceil() functions. */ + hlsl_replace_node(instr, &x->node); + return true; + } + break; + + case HLSL_OP1_NEG: + if (x->op == HLSL_OP1_NEG) + { + /* -(-x) -> x */ + hlsl_replace_node(instr, x->operands[0].node); + return true; + } + break; + + case HLSL_OP1_LOGIC_NOT: + if (x->op == HLSL_OP1_LOGIC_NOT) + { + /* !!x -> x */ + hlsl_replace_node(instr, x->operands[0].node); + return true; + } + + if (hlsl_is_comparison_op(x->op) + && hlsl_base_type_is_integer(x->operands[0].node->data_type->e.numeric.type) + && hlsl_base_type_is_integer(x->operands[1].node->data_type->e.numeric.type)) + { + struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {x->operands[0].node, x->operands[1].node}; + struct hlsl_block block; + + hlsl_block_init(&block); + + /* !(x == y) -> x != y, !(x < y) -> x >= y, etc. */ + res = hlsl_block_add_expr(ctx, &block, invert_comparison_op(x->op), + operands, instr->data_type, &instr->loc); + + list_move_before(&instr->entry, &block.instrs); + hlsl_replace_node(instr, res); + return true; + } + + break; + + default: + break; + } + + return false; +} + +static bool nodes_are_equivalent(const struct hlsl_ir_node *c1, const struct hlsl_ir_node *c2) +{ + if (c1 == c2) + return true; + + if (c1->type == HLSL_IR_SWIZZLE && c2->type == HLSL_IR_SWIZZLE + && hlsl_types_are_equal(c1->data_type, c2->data_type)) + { + const struct hlsl_ir_swizzle *s1 = hlsl_ir_swizzle(c1), *s2 = hlsl_ir_swizzle(c2); + + VKD3D_ASSERT(c1->data_type->class <= HLSL_CLASS_VECTOR); + + if (s1->val.node == s2->val.node && s1->u.vector == s2->u.vector) + return true; + } + + return false; +} + +/* Replaces all conditionals in an expression chain of the form (cond ? x : y) + * with x or y, assuming cond = cond_value. */ +static struct hlsl_ir_node *evaluate_conditionals_recurse(struct hlsl_ctx *ctx, + struct hlsl_block *block, const struct hlsl_ir_node *cond, bool cond_value, + struct hlsl_ir_node *instr, const struct vkd3d_shader_location *loc) +{ + struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {0}; + struct hlsl_ir_expr *expr; + struct hlsl_ir_node *res; + bool progress = false; + unsigned int i; + + if (instr->type != HLSL_IR_EXPR) + return NULL; + expr = hlsl_ir_expr(instr); + + if (expr->op == HLSL_OP3_TERNARY && nodes_are_equivalent(cond, expr->operands[0].node)) + { + struct hlsl_ir_node *x = cond_value ? expr->operands[1].node : expr->operands[2].node; + + res = evaluate_conditionals_recurse(ctx, block, cond, cond_value, x, loc); + return res ? res : x; + } + + for (i = 0; i < HLSL_MAX_OPERANDS; ++i) + { + if (!expr->operands[i].node) + break; + + operands[i] = evaluate_conditionals_recurse(ctx, block, cond, cond_value, expr->operands[i].node, loc); + + if (operands[i]) + progress = true; + else + operands[i] = expr->operands[i].node; + } + + if (progress) + return hlsl_block_add_expr(ctx, block, expr->op, operands, expr->node.data_type, loc); + + return NULL; +} + +static bool fold_conditional_identities(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +{ + struct hlsl_ir_node *c, *x, *y, *res_x, *res_y; + struct hlsl_ir_node *res = NULL; + struct hlsl_ir_expr *expr, *ec; + struct hlsl_block block; + + if (instr->type != HLSL_IR_EXPR) + return false; + + if (instr->data_type->class > HLSL_CLASS_VECTOR) + return false; + + expr = hlsl_ir_expr(instr); + if (expr->op != HLSL_OP3_TERNARY) + return false; + + c = expr->operands[0].node; + x = expr->operands[1].node; + y = expr->operands[2].node; + + VKD3D_ASSERT(c->data_type->e.numeric.type == HLSL_TYPE_BOOL); + + if (nodes_are_equivalent(x, y)) + { + /* c ? x : x -> x */ + hlsl_replace_node(instr, x); + return true; + } + + if (c->type == HLSL_IR_CONSTANT) + { + if (hlsl_constant_is_zero(hlsl_ir_constant(c))) + { + /* false ? x : y -> y */ + hlsl_replace_node(instr, y); + return true; + } + + if (hlsl_constant_is_one(hlsl_ir_constant(c))) + { + /* true ? x : y -> x */ + hlsl_replace_node(instr, x); + return true; + } + } + + hlsl_block_init(&block); + + if (x->type == HLSL_IR_CONSTANT && y->type == HLSL_IR_CONSTANT + && hlsl_types_are_equal(c->data_type, x->data_type) + && hlsl_types_are_equal(c->data_type, y->data_type)) + { + if (hlsl_constant_is_one(hlsl_ir_constant(x)) && hlsl_constant_is_zero(hlsl_ir_constant(y))) + { + /* c ? true : false -> c */ + res = c; + goto done; + } + + if (hlsl_constant_is_zero(hlsl_ir_constant(x)) && hlsl_constant_is_one(hlsl_ir_constant(y))) + { + /* c ? false : true -> !c */ + res = hlsl_block_add_unary_expr(ctx, &block, HLSL_OP1_LOGIC_NOT, c, &instr->loc); + goto done; + } + } + + ec = c->type == HLSL_IR_EXPR ? hlsl_ir_expr(c) : NULL; + if (ec && ec->op == HLSL_OP1_LOGIC_NOT) + { + /* !c ? x : y -> c ? y : x */ + res = hlsl_add_conditional(ctx, &block, ec->operands[0].node, y, x); + goto done; + } + + res_x = evaluate_conditionals_recurse(ctx, &block, c, true, x, &instr->loc); + res_y = evaluate_conditionals_recurse(ctx, &block, c, false, y, &instr->loc); + if (res_x || res_y) + res = hlsl_add_conditional(ctx, &block, c, res_x ? res_x : x, res_y ? res_y : y); + +done: + if (res) + { + list_move_before(&instr->entry, &block.instrs); + hlsl_replace_node(instr, res); + return true; + } + + hlsl_block_cleanup(&block); + return false; +} static bool simplify_exprs(struct hlsl_ctx *ctx, struct hlsl_block *block) { @@ -8275,6 +8520,8 @@ static bool simplify_exprs(struct hlsl_ctx *ctx, struct hlsl_block *block) { progress = hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, block, NULL); progress |= hlsl_transform_ir(ctx, hlsl_normalize_binary_exprs, block, NULL); + progress |= hlsl_transform_ir(ctx, fold_unary_identities, block, NULL); + progress |= hlsl_transform_ir(ctx, fold_conditional_identities, block, NULL); progress |= hlsl_transform_ir(ctx, hlsl_fold_constant_identities, block, NULL); progress |= hlsl_transform_ir(ctx, hlsl_fold_constant_swizzles, block, NULL); @@ -8448,6 +8695,7 @@ static void generate_vsir_signature_entry(struct hlsl_ctx *ctx, struct vsir_prog if (!ascii_strcasecmp(var->semantic.name, "PSIZE") && output && program->shader_version.type == VKD3D_SHADER_TYPE_VERTEX) { + program->has_point_size = true; if (var->data_type->e.numeric.dimx > 1) hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SEMANTIC, "PSIZE output must have only 1 component in this shader model."); @@ -8570,9 +8818,8 @@ static enum vsir_data_type vsir_data_type_from_hlsl_type(struct hlsl_ctx *ctx, c case HLSL_TYPE_DOUBLE: return VSIR_DATA_F64; case HLSL_TYPE_FLOAT: - return VSIR_DATA_F32; case HLSL_TYPE_HALF: - return VSIR_DATA_F16; + return VSIR_DATA_F32; case HLSL_TYPE_INT: return VSIR_DATA_I32; case HLSL_TYPE_UINT: @@ -8703,7 +8950,7 @@ static void sm1_generate_vsir_sampler_dcls(struct hlsl_ctx *ctx, semantic->resource_type = resource_type; dst_param = &semantic->resource.reg; - vsir_register_init(&dst_param->reg, VKD3DSPR_SAMPLER, VSIR_DATA_F32, 1); + vsir_register_init(&dst_param->reg, VKD3DSPR_COMBINED_SAMPLER, VSIR_DATA_F32, 1); dst_param->reg.dimension = VSIR_DIMENSION_NONE; dst_param->reg.idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].index + i; dst_param->write_mask = 0; @@ -8784,6 +9031,7 @@ static void vsir_src_from_hlsl_constant_value(struct vkd3d_shader_src_param *src } src->reg.dimension = VSIR_DIMENSION_VEC4; + src->swizzle = VKD3D_SHADER_NO_SWIZZLE; for (i = 0, j = 0; i < 4; ++i) { if ((map_writemask & (1u << i)) && (j < width)) @@ -9467,8 +9715,6 @@ static bool sm1_generate_vsir_instr_expr(struct hlsl_ctx *ctx, struct vsir_progr break; case HLSL_OP3_CMP: - if (!hlsl_type_is_floating_point(type)) - goto err; generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_CMP, 0, 0, true); break; @@ -9930,10 +10176,14 @@ static void sm1_generate_vsir(struct hlsl_ctx *ctx, const struct vkd3d_shader_co struct hlsl_ir_function_decl *func, struct list *semantic_vars, struct hlsl_block *body, uint64_t config_flags, struct vsir_program *program) { + struct hlsl_ir_var *var; struct hlsl_block block; + struct hlsl_reg *reg; + unsigned int *count; program->ssa_count = 0; - program->temp_count = allocate_temp_registers(ctx, body, semantic_vars); + program->temp_count = 0; + allocate_temp_registers(ctx, body, semantic_vars); if (ctx->result) return; @@ -9944,7 +10194,19 @@ static void sm1_generate_vsir(struct hlsl_ctx *ctx, const struct vkd3d_shader_co sm1_generate_vsir_block(ctx, body, program); + count = &program->flat_constant_count[VKD3D_SHADER_D3DBC_FLOAT_CONSTANT_REGISTER]; + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (!var->is_uniform) + continue; + + if (!(reg = &var->regs[HLSL_REGSET_NUMERIC])->allocation_size) + continue; + + *count = max(*count, reg->id + reg->allocation_size); + } program->ssa_count = ctx->ssa_count; + program->temp_count = ctx->temp_count; if (ctx->result) return; @@ -10795,11 +11057,32 @@ static bool sm4_generate_vsir_instr_expr(struct hlsl_ctx *ctx, generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_ROUND_PI, 0, 0, true); return true; + case HLSL_OP1_CLZ: + VKD3D_ASSERT(hlsl_type_is_integer(dst_type)); + VKD3D_ASSERT(hlsl_version_ge(ctx, 5, 0)); + if (hlsl_type_is_signed_integer(src_type)) + generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_FIRSTBIT_SHI, 0, 0, true); + else + generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_FIRSTBIT_HI, 0, 0, true); + return true; + case HLSL_OP1_COS: VKD3D_ASSERT(type_is_float(dst_type)); sm4_generate_vsir_expr_with_two_destinations(ctx, program, VSIR_OP_SINCOS, expr, 1); return true; + case HLSL_OP1_COUNTBITS: + VKD3D_ASSERT(hlsl_type_is_integer(dst_type)); + VKD3D_ASSERT(hlsl_version_ge(ctx, 5, 0)); + generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_COUNTBITS, 0, 0, true); + return true; + + case HLSL_OP1_CTZ: + VKD3D_ASSERT(hlsl_type_is_integer(dst_type)); + VKD3D_ASSERT(hlsl_version_ge(ctx, 5, 0)); + generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_FIRSTBIT_LO, 0, 0, true); + return true; + case HLSL_OP1_DSX: VKD3D_ASSERT(type_is_float(dst_type)); generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VSIR_OP_DSX, 0, 0, true); @@ -11437,6 +11720,7 @@ static bool sm4_generate_vsir_instr_ld(struct hlsl_ctx *ctx, const struct vkd3d_shader_version *version = &program->shader_version; const struct hlsl_ir_node *sample_index = load->sample_index.node; const struct hlsl_ir_node *texel_offset = load->texel_offset.node; + const struct hlsl_ir_node *byte_offset = load->byte_offset.node; const struct hlsl_ir_node *coords = load->coords.node; unsigned int coords_writemask = VKD3DSP_WRITEMASK_ALL; const struct hlsl_deref *resource = &load->resource; @@ -11444,20 +11728,15 @@ static bool sm4_generate_vsir_instr_ld(struct hlsl_ctx *ctx, enum hlsl_sampler_dim dim = load->sampling_dim; bool tgsm = load->resource.var->is_tgsm; struct vkd3d_shader_instruction *ins; + bool multisampled, raw, structured; enum vkd3d_shader_opcode opcode; - bool multisampled, raw; VKD3D_ASSERT(load->load_type == HLSL_RESOURCE_LOAD); - if (resource_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) - { - hlsl_fixme(ctx, &load->node.loc, "Structured buffer loads."); - return false; - } - multisampled = resource_type->class == HLSL_CLASS_TEXTURE && (resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMS || resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMSARRAY); + structured = resource_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER; if (!tgsm) { @@ -11468,15 +11747,19 @@ static bool sm4_generate_vsir_instr_ld(struct hlsl_ctx *ctx, hlsl_fixme(ctx, &load->node.loc, "Load from structured TGSM."); return false; } + VKD3D_ASSERT(!(structured && multisampled)); - if (uav) + if (structured) + opcode = VSIR_OP_LD_STRUCTURED; + else if (uav) opcode = VSIR_OP_LD_UAV_TYPED; else if (raw) opcode = VSIR_OP_LD_RAW; else opcode = multisampled ? VSIR_OP_LD2DMS : VSIR_OP_LD; - if (!(ins = generate_vsir_add_program_instruction(ctx, program, &instr->loc, opcode, 1, 2 + multisampled))) + if (!(ins = generate_vsir_add_program_instruction(ctx, program, + &instr->loc, opcode, 1, 2 + multisampled + structured))) return false; if (texel_offset && !sm4_generate_vsir_validate_texel_offset_aoffimmi(texel_offset)) @@ -11504,10 +11787,15 @@ static bool sm4_generate_vsir_instr_ld(struct hlsl_ctx *ctx, vsir_src_from_hlsl_node(&ins->src[0], ctx, coords, coords_writemask); if (!sm4_generate_vsir_init_src_param_from_deref(ctx, program, - &ins->src[1], resource, ins->dst[0].write_mask, &instr->loc)) + &ins->src[structured ? 2 : 1], resource, ins->dst[0].write_mask, &instr->loc)) return false; - if (multisampled) + if (structured) + { + VKD3D_ASSERT(byte_offset); + vsir_src_from_hlsl_node(&ins->src[1], ctx, byte_offset, VKD3DSP_WRITEMASK_ALL); + } + else if (multisampled) { if (sample_index->type == HLSL_IR_CONSTANT) vsir_src_from_hlsl_constant_value(&ins->src[2], ctx, @@ -12093,16 +12381,15 @@ static void sm4_generate_vsir_add_function(struct hlsl_ctx *ctx, struct list *se struct hlsl_block block = {0}; struct hlsl_scope *scope; struct hlsl_ir_var *var; - uint32_t temp_count; ctx->is_patch_constant_func = func == ctx->patch_constant_func; compute_liveness(ctx, body); mark_indexable_vars(ctx, body); - temp_count = allocate_temp_registers(ctx, body, semantic_vars); + allocate_temp_registers(ctx, body, semantic_vars); if (ctx->result) return; - program->temp_count = max(program->temp_count, temp_count); + program->temp_count = max(program->temp_count, ctx->temp_count); hlsl_block_init(&block); @@ -12113,8 +12400,8 @@ static void sm4_generate_vsir_add_function(struct hlsl_ctx *ctx, struct list *se sm4_generate_vsir_instr_dcl_semantic(ctx, program, var, &block, &var->loc); } - if (temp_count) - sm4_generate_vsir_instr_dcl_temps(ctx, program, temp_count, &block, &func->loc); + if (ctx->temp_count) + sm4_generate_vsir_instr_dcl_temps(ctx, program, ctx->temp_count, &block, &func->loc); LIST_FOR_EACH_ENTRY(scope, &ctx->scopes, struct hlsl_scope, entry) { @@ -12427,7 +12714,7 @@ static void sm4_generate_vsir_add_dcl_constant_buffer(struct hlsl_ctx *ctx, return; } - ins->declaration.cb.size = cbuffer->size; + ins->declaration.cb.size = align(cbuffer->size, 4) * sizeof(float); src_param = &ins->declaration.cb.src; vsir_src_param_init(src_param, VKD3DSPR_CONSTBUFFER, VSIR_DATA_F32, 0); @@ -12592,6 +12879,9 @@ static void sm4_generate_vsir_add_dcl_texture(struct hlsl_ctx *ctx, { switch (component_type->sampler_dim) { + case HLSL_SAMPLER_DIM_STRUCTURED_BUFFER: + opcode = VSIR_OP_DCL_RESOURCE_STRUCTURED; + break; case HLSL_SAMPLER_DIM_RAW_BUFFER: opcode = VSIR_OP_DCL_RESOURCE_RAW; break; @@ -12647,7 +12937,7 @@ static void sm4_generate_vsir_add_dcl_texture(struct hlsl_ctx *ctx, else if (component_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) { ins->structured = true; - ins->resource_stride = 4 * component_type->e.resource.format->reg_size[HLSL_REGSET_NUMERIC]; + ins->resource_stride = hlsl_type_get_packed_size(component_type->e.resource.format); ins->declaration.structured_resource.byte_stride = ins->resource_stride; } else @@ -12784,6 +13074,7 @@ static void sm4_generate_vsir(struct hlsl_ctx *ctx, } program->ssa_count = 0; + program->temp_count = 0; if (version->type == VKD3D_SHADER_TYPE_HULL) generate_vsir_add_program_instruction(ctx, program, @@ -12801,6 +13092,7 @@ static void sm4_generate_vsir(struct hlsl_ctx *ctx, generate_vsir_scan_global_flags(ctx, program, semantic_vars, func); program->ssa_count = ctx->ssa_count; + program->temp_count = ctx->temp_count; } /* For some reason, for matrices, values from default value initializers end @@ -12915,14 +13207,16 @@ static enum D3D_RESOURCE_RETURN_TYPE sm4_data_type(const struct hlsl_type *type) static D3D_SHADER_INPUT_TYPE sm4_resource_type(const struct hlsl_type *type) { + bool structured = type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER; + switch (type->class) { case HLSL_CLASS_SAMPLER: return D3D_SIT_SAMPLER; case HLSL_CLASS_TEXTURE: - return D3D_SIT_TEXTURE; + return structured ? D3D_SIT_STRUCTURED : D3D_SIT_TEXTURE; case HLSL_CLASS_UAV: - return D3D_SIT_UAV_RWTYPED; + return structured ? D3D_SIT_UAV_RWSTRUCTURED : D3D_SIT_UAV_RWTYPED; default: break; } @@ -12998,7 +13292,8 @@ static D3D_SHADER_VARIABLE_TYPE sm4_base_type(const struct hlsl_type *type) vkd3d_unreachable(); } -static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, struct hlsl_type *type) +static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + struct hlsl_type *type, bool structured) { const struct hlsl_type *array_type = hlsl_get_multiarray_element_type(type); const char *name = array_type->name ? array_type->name : ""; @@ -13007,7 +13302,10 @@ static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b size_t name_offset = 0; size_t i; - if (type->bytecode_offset) + if (!structured && type->bytecode_offset) + return; + + if (structured && type->packed_bytecode_offset) return; if (profile->major_version >= 5) @@ -13029,7 +13327,7 @@ static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b continue; field->name_bytecode_offset = put_string(buffer, field->name); - write_sm4_type(ctx, buffer, field->type); + write_sm4_type(ctx, buffer, field->type, structured); ++field_count; } @@ -13038,15 +13336,29 @@ static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b for (i = 0; i < array_type->e.record.field_count; ++i) { struct hlsl_struct_field *field = &array_type->e.record.fields[i]; + unsigned int field_type_offset, offset; if (!field->type->reg_size[HLSL_REGSET_NUMERIC]) continue; put_u32(buffer, field->name_bytecode_offset); - put_u32(buffer, field->type->bytecode_offset); - put_u32(buffer, field->reg_offset[HLSL_REGSET_NUMERIC] * sizeof(float)); + + if (!structured) + field_type_offset = field->type->bytecode_offset; + else + field_type_offset = field->type->packed_bytecode_offset; + put_u32(buffer, field_type_offset); + + if (!structured) + offset = field->reg_offset[HLSL_REGSET_NUMERIC] * sizeof(float); + else + offset = struct_field_get_packed_offset(array_type, i); + put_u32(buffer, offset); } - type->bytecode_offset = put_u32(buffer, vkd3d_make_u32(D3D_SVC_STRUCT, D3D_SVT_VOID)); + if (!structured) + type->bytecode_offset = put_u32(buffer, vkd3d_make_u32(D3D_SVC_STRUCT, D3D_SVT_VOID)); + else + type->packed_bytecode_offset = put_u32(buffer, vkd3d_make_u32(D3D_SVC_STRUCT, D3D_SVT_VOID)); put_u32(buffer, vkd3d_make_u32(1, hlsl_type_component_count(array_type))); put_u32(buffer, vkd3d_make_u32(array_size, field_count)); put_u32(buffer, fields_offset); @@ -13054,7 +13366,11 @@ static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b else { VKD3D_ASSERT(array_type->class <= HLSL_CLASS_LAST_NUMERIC); - type->bytecode_offset = put_u32(buffer, vkd3d_make_u32(sm4_class(array_type), sm4_base_type(array_type))); + if (!structured) + type->bytecode_offset = put_u32(buffer, vkd3d_make_u32(sm4_class(array_type), sm4_base_type(array_type))); + else + type->packed_bytecode_offset = put_u32(buffer, + vkd3d_make_u32(sm4_class(array_type), sm4_base_type(array_type))); put_u32(buffer, vkd3d_make_u32(array_type->e.numeric.dimy, array_type->e.numeric.dimx)); put_u32(buffer, vkd3d_make_u32(array_size, 0)); put_u32(buffer, 1); @@ -13073,9 +13389,9 @@ static void write_sm4_type(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rdef) { uint32_t binding_desc_size = (hlsl_version_ge(ctx, 5, 1) ? 10 : 8) * sizeof(uint32_t); - size_t cbuffers_offset, resources_offset, creator_offset, string_offset; - unsigned int cbuffer_count = 0, extern_resources_count, i, j; + size_t buffers_offset, resources_offset, creator_offset, string_offset; size_t cbuffer_position, resource_position, creator_position; + unsigned int buffer_count = 0, extern_resources_count, i, j; const struct hlsl_profile_info *profile = ctx->profile; struct vkd3d_bytecode_buffer buffer = {0}; struct extern_resource *extern_resources; @@ -13097,10 +13413,20 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry) { if (cbuffer->reg.allocated) - ++cbuffer_count; + ++buffer_count; } - put_u32(&buffer, cbuffer_count); + for (i = 0; i < extern_resources_count; ++i) + { + const struct extern_resource *resource = &extern_resources[i]; + + if (resource->buffer || resource->component_type->sampler_dim != HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) + continue; + + ++buffer_count; + } + + put_u32(&buffer, buffer_count); cbuffer_position = put_u32(&buffer, 0); put_u32(&buffer, extern_resources_count); resource_position = put_u32(&buffer, 0); @@ -13141,12 +13467,19 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd put_u32(&buffer, sm4_resource_type(resource->component_type)); if (resource->regset == HLSL_REGSET_TEXTURES || resource->regset == HLSL_REGSET_UAVS) { + bool structured = resource->component_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER; unsigned int dimx = resource->component_type->e.resource.format->e.numeric.dimx; put_u32(&buffer, sm4_data_type(resource->component_type)); put_u32(&buffer, sm4_rdef_resource_dimension(resource->component_type)); - put_u32(&buffer, ~0u); /* FIXME: multisample count */ - flags |= (dimx - 1) << VKD3D_SM4_SIF_TEXTURE_COMPONENTS_SHIFT; + + if (structured) + put_u32(&buffer, hlsl_type_get_packed_size(resource->component_type->e.resource.format)); + else + put_u32(&buffer, ~0u); /* FIXME: multisample count */ + + if (!structured) + flags |= (dimx - 1) << VKD3D_SM4_SIF_TEXTURE_COMPONENTS_SHIFT; } else { @@ -13175,8 +13508,8 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd /* Buffers. */ - cbuffers_offset = bytecode_align(&buffer); - set_u32(&buffer, cbuffer_position, cbuffers_offset); + buffers_offset = bytecode_align(&buffer); + set_u32(&buffer, cbuffer_position, buffers_offset); LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry) { unsigned int var_count = 0; @@ -13198,6 +13531,24 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd put_u32(&buffer, cbuffer->type == HLSL_BUFFER_CONSTANT ? D3D_CT_CBUFFER : D3D_CT_TBUFFER); } + for (i = 0; i < extern_resources_count; ++i) + { + const struct extern_resource *resource = &extern_resources[i]; + struct hlsl_type *resource_type; + + if (resource->buffer || resource->component_type->sampler_dim != HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) + continue; + + resource_type = resource->component_type->e.resource.format; + + put_u32(&buffer, 0); /* name */ + put_u32(&buffer, 1); /* var count */ + put_u32(&buffer, 0); /* variable offset */ + put_u32(&buffer, hlsl_type_get_packed_size(resource_type)); /* size */ + put_u32(&buffer, 0); /* FIXME: flags */ + put_u32(&buffer, D3D_CT_RESOURCE_BIND_INFO); + } + i = 0; LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry) { @@ -13205,7 +13556,18 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd continue; string_offset = put_string(&buffer, cbuffer->name); - set_u32(&buffer, cbuffers_offset + i++ * 6 * sizeof(uint32_t), string_offset); + set_u32(&buffer, buffers_offset + i++ * 6 * sizeof(uint32_t), string_offset); + } + + for (j = 0; j < extern_resources_count; ++j) + { + const struct extern_resource *resource = &extern_resources[j]; + + if (resource->buffer || resource->component_type->sampler_dim != HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) + continue; + + string_offset = put_string(&buffer, resource->name); + set_u32(&buffer, buffers_offset + i++ * 6 * sizeof(uint32_t), string_offset); } i = 0; @@ -13216,7 +13578,7 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd if (!cbuffer->reg.allocated) continue; - set_u32(&buffer, cbuffers_offset + (i++ * 6 + 2) * sizeof(uint32_t), vars_start); + set_u32(&buffer, buffers_offset + (i++ * 6 + 2) * sizeof(uint32_t), vars_start); LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) { @@ -13255,7 +13617,7 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd string_offset = put_string(&buffer, var->name); set_u32(&buffer, var_offset, string_offset); - write_sm4_type(ctx, &buffer, var->data_type); + write_sm4_type(ctx, &buffer, var->data_type, false); set_u32(&buffer, var_offset + 4 * sizeof(uint32_t), var->data_type->bytecode_offset); if (var->default_values) @@ -13298,6 +13660,42 @@ static void sm4_generate_rdef(struct hlsl_ctx *ctx, struct vkd3d_shader_code *rd } } + for (j = 0; j < extern_resources_count; ++j) + { + const struct extern_resource *resource = &extern_resources[j]; + struct hlsl_type *resource_type; + size_t vars_start; + + if (resource->buffer || resource->component_type->sampler_dim != HLSL_SAMPLER_DIM_STRUCTURED_BUFFER) + continue; + + resource_type = resource->component_type->e.resource.format; + + vars_start = bytecode_align(&buffer); + + set_u32(&buffer, buffers_offset + (i++ * 6 + 2) * sizeof(uint32_t), vars_start); + + put_u32(&buffer, 0); /* name */ + put_u32(&buffer, 0); /* offset */ + put_u32(&buffer, hlsl_type_get_packed_size(resource_type)); + put_u32(&buffer, D3D_SVF_USED); + put_u32(&buffer, 0); /* type */ + put_u32(&buffer, 0); /* default value */ + + if (profile->major_version >= 5) + { + put_u32(&buffer, ~0u); /* texture start */ + put_u32(&buffer, 0); /* texture count */ + put_u32(&buffer, ~0u); /* sampler start */ + put_u32(&buffer, 0); /* sampler count */ + } + + string_offset = put_string(&buffer, "$Element"); + set_u32(&buffer, vars_start, string_offset); + write_sm4_type(ctx, &buffer, resource_type, true); + set_u32(&buffer, vars_start + 4 * sizeof(uint32_t), resource_type->packed_bytecode_offset); + } + creator_offset = put_string(&buffer, vkd3d_shader_get_version(NULL, NULL)); set_u32(&buffer, creator_position, creator_offset); @@ -13720,6 +14118,102 @@ static void loop_unrolling_execute(struct hlsl_ctx *ctx, struct hlsl_block *bloc hlsl_transform_ir(ctx, resolve_loops, block, NULL); } +static bool lower_countbits(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, struct hlsl_block *block) +{ + struct hlsl_ir_function_decl *func; + struct hlsl_ir_node *call, *rhs; + struct hlsl_ir_expr *expr; + struct hlsl_ir_var *lhs; + char *body; + + /* Like vkd3d_popcount(). */ + static const char template[] = + "typedef uint%u uintX;\n" + "uintX countbits(uintX v)\n" + "{\n" + " v -= (v >> 1) & 0x55555555;\n" + " v = (v & 0x33333333) + ((v >> 2) & 0x33333333);\n" + " return (((v + (v >> 4)) & 0x0f0f0f0f) * 0x01010101) >> 24;\n" + "}\n"; + + if (node->type != HLSL_IR_EXPR) + return false; + + expr = hlsl_ir_expr(node); + if (expr->op != HLSL_OP1_COUNTBITS) + return false; + + rhs = expr->operands[0].node; + if (!(body = hlsl_sprintf_alloc(ctx, template, hlsl_type_component_count(rhs->data_type)))) + return false; + func = hlsl_compile_internal_function(ctx, "countbits", body); + vkd3d_free(body); + if (!func) + return false; + + lhs = func->parameters.vars[0]; + hlsl_block_add_simple_store(ctx, block, lhs, rhs); + + if (!(call = hlsl_new_call(ctx, func, &node->loc))) + return false; + hlsl_block_add_instr(block, call); + + hlsl_block_add_simple_load(ctx, block, func->return_var, &node->loc); + + return true; +} + +static bool lower_ctz(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, struct hlsl_block *block) +{ + struct hlsl_ir_function_decl *func; + struct hlsl_ir_node *call, *rhs; + struct hlsl_ir_expr *expr; + struct hlsl_ir_var *lhs; + char *body; + + /* ctz() returns the bit number of the least significant 1-bit. + * Bit numbers count from the least significant bit. */ + static const char template[] = + "typedef uint%u uintX;\n" + "uintX ctz(uintX v)\n" + "{\n" + " uintX c = 31;\n" + " v &= -v;\n" + " c = (v & 0x0000ffff) ? c - 16 : c;\n" + " c = (v & 0x00ff00ff) ? c - 8 : c;\n" + " c = (v & 0x0f0f0f0f) ? c - 4 : c;\n" + " c = (v & 0x33333333) ? c - 2 : c;\n" + " c = (v & 0x55555555) ? c - 1 : c;\n" + " return v ? c : -1;\n" + "}\n"; + + if (node->type != HLSL_IR_EXPR) + return false; + + expr = hlsl_ir_expr(node); + if (expr->op != HLSL_OP1_CTZ) + return false; + + rhs = expr->operands[0].node; + if (!(body = hlsl_sprintf_alloc(ctx, template, hlsl_type_component_count(rhs->data_type)))) + return false; + func = hlsl_compile_internal_function(ctx, "ctz", body); + vkd3d_free(body); + if (!func) + return false; + + lhs = func->parameters.vars[0]; + hlsl_block_add_simple_store(ctx, block, lhs, rhs); + + if (!(call = hlsl_new_call(ctx, func, &node->loc))) + return false; + hlsl_block_add_instr(block, call); + + hlsl_block_add_simple_load(ctx, block, func->return_var, &node->loc); + + return true; +} + static bool lower_f16tof32(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, struct hlsl_block *block) { struct hlsl_ir_function_decl *func; @@ -13862,6 +14356,69 @@ static bool lower_f32tof16(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, stru return true; } +static bool lower_find_msb(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, struct hlsl_block *block) +{ + struct hlsl_ir_function_decl *func; + struct hlsl_ir_node *call, *rhs; + struct hlsl_ir_expr *expr; + struct hlsl_ir_var *lhs; + char *body; + + /* For positive numbers, find_msb() returns the bit number of the most + * significant 1-bit. For negative numbers, it returns the bit number of + * the most significant 0-bit. Bit numbers count from the least + * significant bit. */ + static const char template[] = + "typedef %s intX;\n" + "uint%u find_msb(intX v)\n" + "{\n" + " intX c, mask;\n" + " v = v < 0 ? ~v : v;\n" + " mask = v & 0xffff0000;\n" + " v = mask ? mask : v;\n" + " c = mask ? 16 : v ? 0 : -1;\n" + " mask = v & 0xff00ff00;\n" + " v = mask ? mask : v;\n" + " c = mask ? c + 8 : c;\n" + " mask = v & 0xf0f0f0f0;\n" + " v = mask ? mask : v;\n" + " c = mask ? c + 4 : c;\n" + " mask = v & 0xcccccccc;\n" + " v = mask ? mask : v;\n" + " c = mask ? c + 2 : c;\n" + " mask = v & 0xaaaaaaaa;\n" + " v = mask ? mask : v;\n" + " c = mask ? c + 1 : c;\n" + " return c;\n" + "}\n"; + + if (node->type != HLSL_IR_EXPR) + return false; + + expr = hlsl_ir_expr(node); + if (expr->op != HLSL_OP1_FIND_MSB) + return false; + + rhs = expr->operands[0].node; + if (!(body = hlsl_sprintf_alloc(ctx, template, rhs->data_type->name, hlsl_type_component_count(rhs->data_type)))) + return false; + func = hlsl_compile_internal_function(ctx, "find_msb", body); + vkd3d_free(body); + if (!func) + return false; + + lhs = func->parameters.vars[0]; + hlsl_block_add_simple_store(ctx, block, lhs, rhs); + + if (!(call = hlsl_new_call(ctx, func, &node->loc))) + return false; + hlsl_block_add_instr(block, call); + + hlsl_block_add_simple_load(ctx, block, func->return_var, &node->loc); + + return true; +} + static bool lower_isinf(struct hlsl_ctx *ctx, struct hlsl_ir_node *node, struct hlsl_block *block) { struct hlsl_ir_function_decl *func; @@ -13978,8 +14535,11 @@ static void process_entry_function(struct hlsl_ctx *ctx, struct list *semantic_v if (hlsl_version_ge(ctx, 4, 0) && hlsl_version_lt(ctx, 5, 0)) { + lower_ir(ctx, lower_countbits, body); + lower_ir(ctx, lower_ctz, body); lower_ir(ctx, lower_f16tof32, body); lower_ir(ctx, lower_f32tof16, body); + lower_ir(ctx, lower_find_msb, body); } lower_ir(ctx, lower_isinf, body); diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c index d339a06e6c7..7b3b0470d5d 100644 --- a/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c +++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c @@ -250,6 +250,118 @@ static bool fold_ceil(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, return true; } +static bool fold_clz(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = src->node.data_type->e.numeric.type; + unsigned int k, v; + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_INT: + v = src->value.u[k].i < 0 ? ~src->value.u[k].u : src->value.u[k].u; + break; + + case HLSL_TYPE_UINT: + v = src->value.u[k].u; + break; + + default: + FIXME("Fold 'clz' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + + dst->u[k].u = v ? vkd3d_log2i(v) ^ 0x1f : ~0u; + } + + return true; +} + +static bool fold_cos(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + dst->u[k].f = cosf(src->value.u[k].f); + break; + + case HLSL_TYPE_DOUBLE: + dst->u[k].d = cos(src->value.u[k].d); + break; + + default: + FIXME("Fold 'cos' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + +static bool fold_countbits(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_UINT: + dst->u[k].u = vkd3d_popcount(src->value.u[k].u); + break; + + default: + FIXME("Fold 'countbits' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + +static bool fold_ctz(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_UINT: + if (!src->value.u[k].u) + dst->u[k].u = ~0u; + else + dst->u[k].u = vkd3d_ctz(src->value.u[k].u); + break; + + default: + FIXME("Fold 'ctz' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + static bool fold_exp2(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) { @@ -478,6 +590,48 @@ static bool fold_rcp(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons return true; } +static bool fold_reinterpret(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + unsigned int k; + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + dst->u[k] = src->value.u[k]; + } + + return true; +} + +static bool fold_round(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + /* Somewhat unfortunately, constant folded round() rounds + * halfway cases towards positive infinity, as opposed to + * nearest even like vsir/TPF round_ne. */ + dst->u[k].f = floorf(src->value.u[k].f + 0.5f); + break; + + default: + FIXME("Fold 'round' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + static bool fold_rsq(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src, const struct vkd3d_shader_location *loc) { @@ -544,6 +698,36 @@ static bool fold_sat(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons return true; } +static bool fold_sin(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, + const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + dst->u[k].f = sinf(src->value.u[k].f); + break; + + case HLSL_TYPE_DOUBLE: + dst->u[k].d = sin(src->value.u[k].d); + break; + + default: + FIXME("Fold 'sin' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + static bool fold_sqrt(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src, const struct vkd3d_shader_location *loc) { @@ -974,6 +1158,44 @@ static bool fold_lshift(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, c return true; } +static bool fold_mad(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type, + const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2, const struct hlsl_ir_constant *src3) +{ + enum hlsl_base_type type = dst_type->e.numeric.type; + unsigned int k; + + VKD3D_ASSERT(type == src1->node.data_type->e.numeric.type); + VKD3D_ASSERT(type == src2->node.data_type->e.numeric.type); + VKD3D_ASSERT(type == src3->node.data_type->e.numeric.type); + + for (k = 0; k < dst_type->e.numeric.dimx; ++k) + { + switch (type) + { + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + dst->u[k].f = fmaf(src1->value.u[k].f, src2->value.u[k].f, src3->value.u[k].f); + break; + + case HLSL_TYPE_DOUBLE: + dst->u[k].d = fma(src1->value.u[k].d, src2->value.u[k].d, src3->value.u[k].d); + break; + + case HLSL_TYPE_INT: + case HLSL_TYPE_MIN16UINT: + case HLSL_TYPE_UINT: + dst->u[k].u = src1->value.u[k].u * src2->value.u[k].u + src3->value.u[k].u; + break; + + default: + FIXME("Fold 'mad' for type %s.\n", debug_hlsl_type(ctx, dst_type)); + return false; + } + } + + return true; +} + static bool fold_max(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type, const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2) { @@ -1263,6 +1485,22 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, success = fold_ceil(ctx, &res, instr->data_type, arg1); break; + case HLSL_OP1_CLZ: + success = fold_clz(ctx, &res, instr->data_type, arg1); + break; + + case HLSL_OP1_COS: + success = fold_cos(ctx, &res, instr->data_type, arg1); + break; + + case HLSL_OP1_COUNTBITS: + success = fold_countbits(ctx, &res, instr->data_type, arg1); + break; + + case HLSL_OP1_CTZ: + success = fold_ctz(ctx, &res, instr->data_type, arg1); + break; + case HLSL_OP1_EXP2: success = fold_exp2(ctx, &res, instr->data_type, arg1); break; @@ -1291,6 +1529,14 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, success = fold_rcp(ctx, &res, instr->data_type, arg1, &instr->loc); break; + case HLSL_OP1_REINTERPRET: + success = fold_reinterpret(ctx, &res, instr->data_type, arg1); + break; + + case HLSL_OP1_ROUND: + success = fold_round(ctx, &res, instr->data_type, arg1); + break; + case HLSL_OP1_RSQ: success = fold_rsq(ctx, &res, instr->data_type, arg1, &instr->loc); break; @@ -1299,6 +1545,10 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, success = fold_sat(ctx, &res, instr->data_type, arg1); break; + case HLSL_OP1_SIN: + success = fold_sin(ctx, &res, instr->data_type, arg1); + break; + case HLSL_OP1_SQRT: success = fold_sqrt(ctx, &res, instr->data_type, arg1, &instr->loc); break; @@ -1373,6 +1623,10 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, success = fold_dp2add(ctx, &res, instr->data_type, arg1, arg2, arg3); break; + case HLSL_OP3_MAD: + success = fold_mad(ctx, &res, instr->data_type, arg1, arg2, arg3); + break; + case HLSL_OP3_TERNARY: success = fold_ternary(ctx, &res, instr->data_type, arg1, arg2, arg3); break; @@ -1393,76 +1647,9 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, return success; } -static bool constant_is_zero(struct hlsl_ir_constant *const_arg) -{ - struct hlsl_type *data_type = const_arg->node.data_type; - unsigned int k; - - for (k = 0; k < data_type->e.numeric.dimx; ++k) - { - switch (data_type->e.numeric.type) - { - case HLSL_TYPE_FLOAT: - case HLSL_TYPE_HALF: - if (const_arg->value.u[k].f != 0.0f) - return false; - break; - - case HLSL_TYPE_DOUBLE: - if (const_arg->value.u[k].d != 0.0) - return false; - break; - - case HLSL_TYPE_UINT: - case HLSL_TYPE_INT: - case HLSL_TYPE_BOOL: - case HLSL_TYPE_MIN16UINT: - if (const_arg->value.u[k].u != 0) - return false; - break; - } - } - return true; -} - -static bool constant_is_one(struct hlsl_ir_constant *const_arg) -{ - struct hlsl_type *data_type = const_arg->node.data_type; - unsigned int k; - - for (k = 0; k < data_type->e.numeric.dimx; ++k) - { - switch (data_type->e.numeric.type) - { - case HLSL_TYPE_FLOAT: - case HLSL_TYPE_HALF: - if (const_arg->value.u[k].f != 1.0f) - return false; - break; - - case HLSL_TYPE_DOUBLE: - if (const_arg->value.u[k].d != 1.0) - return false; - break; - - case HLSL_TYPE_UINT: - case HLSL_TYPE_INT: - case HLSL_TYPE_MIN16UINT: - if (const_arg->value.u[k].u != 1) - return false; - break; - - case HLSL_TYPE_BOOL: - if (const_arg->value.u[k].u != ~0) - return false; - break; - } - } - return true; -} - bool hlsl_fold_constant_identities(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) { + static const struct hlsl_constant_value zero; struct hlsl_ir_constant *const_arg = NULL; struct hlsl_ir_node *mut_arg = NULL; struct hlsl_ir_node *res_node; @@ -1502,29 +1689,40 @@ bool hlsl_fold_constant_identities(struct hlsl_ctx *ctx, struct hlsl_ir_node *in switch (expr->op) { case HLSL_OP2_ADD: - if (constant_is_zero(const_arg)) + if (hlsl_constant_is_zero(const_arg)) res_node = mut_arg; break; case HLSL_OP2_MUL: - if (constant_is_one(const_arg)) + if (hlsl_constant_is_one(const_arg)) res_node = mut_arg; break; case HLSL_OP2_LOGIC_AND: - if (constant_is_zero(const_arg)) + if (hlsl_constant_is_zero(const_arg)) res_node = &const_arg->node; - else if (constant_is_one(const_arg)) + else if (hlsl_constant_is_one(const_arg)) res_node = mut_arg; break; case HLSL_OP2_LOGIC_OR: - if (constant_is_zero(const_arg)) + if (hlsl_constant_is_zero(const_arg)) res_node = mut_arg; - else if (constant_is_one(const_arg)) + else if (hlsl_constant_is_one(const_arg)) res_node = &const_arg->node; break; + case HLSL_OP2_LESS: + /* x < 0 -> false, if x is unsigned. */ + if (!hlsl_type_is_unsigned_integer(expr->operands[0].node->data_type) + || expr->operands[1].node->type != HLSL_IR_CONSTANT + || !hlsl_constant_is_zero(hlsl_ir_constant(expr->operands[1].node))) + break; + if (!(res_node = hlsl_new_constant(ctx, instr->data_type, &zero, &instr->loc))) + break; + list_add_before(&expr->node.entry, &res_node->entry); + break; + default: break; } @@ -1649,6 +1847,9 @@ bool hlsl_normalize_binary_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *inst if (instr->data_type->class > HLSL_CLASS_VECTOR) return false; + if (expr->operands[2].node) + return false; + hlsl_block_init(&block); arg1 = expr->operands[0].node; diff --git a/libs/vkd3d/libs/vkd3d-shader/ir.c b/libs/vkd3d/libs/vkd3d-shader/ir.c index 23e059a3490..453b79c75d2 100644 --- a/libs/vkd3d/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d/libs/vkd3d-shader/ir.c @@ -27,8 +27,54 @@ struct vsir_transformation_context uint64_t config_flags; const struct vkd3d_shader_compile_info *compile_info; struct vkd3d_shader_message_context *message_context; + struct vkd3d_shader_location null_location; +}; + +static void vsir_transformation_context_init(struct vsir_transformation_context *ctx, + struct vsir_program *program, uint64_t config_flags, const struct vkd3d_shader_compile_info *compile_info, + struct vkd3d_shader_message_context *message_context) +{ + *ctx = (struct vsir_transformation_context) + { + .result = VKD3D_OK, + .program = program, + .config_flags = config_flags, + .compile_info = compile_info, + .message_context = message_context, + .null_location = {.source_name = compile_info->source_name}, + }; }; +const char *vsir_data_type_get_name(enum vsir_data_type t, const char *error) +{ + static const char * const names[] = + { + [VSIR_DATA_BOOL ] = "bool", + [VSIR_DATA_F16 ] = "f16", + [VSIR_DATA_F32 ] = "f32", + [VSIR_DATA_F64 ] = "f64", + [VSIR_DATA_I8 ] = "i8", + [VSIR_DATA_I16 ] = "i16", + [VSIR_DATA_I32 ] = "i32", + [VSIR_DATA_I64 ] = "i64", + [VSIR_DATA_U8 ] = "u8", + [VSIR_DATA_U16 ] = "u16", + [VSIR_DATA_U32 ] = "u32", + [VSIR_DATA_U64 ] = "u64", + [VSIR_DATA_SNORM ] = "snorm", + [VSIR_DATA_UNORM ] = "unorm", + [VSIR_DATA_OPAQUE ] = "opaque", + [VSIR_DATA_MIXED ] = "mixed", + [VSIR_DATA_CONTINUED] = "", + [VSIR_DATA_UNUSED ] = "", + }; + + if ((size_t)t < ARRAY_SIZE(names)) + return names[t] ? names[t] : error; + + return error; +} + const char *vsir_opcode_get_name(enum vkd3d_shader_opcode op, const char *error) { static const char * const names[] = @@ -373,6 +419,177 @@ const char *vsir_opcode_get_name(enum vkd3d_shader_opcode op, const char *error) return error; } +static struct vkd3d_shader_param_node *shader_param_allocator_node_create( + struct vkd3d_shader_param_allocator *allocator) +{ + struct vkd3d_shader_param_node *node; + + if (!(node = vkd3d_malloc(offsetof(struct vkd3d_shader_param_node, param[allocator->count * allocator->stride])))) + return NULL; + node->next = NULL; + + return node; +} + +static void shader_param_allocator_init(struct vkd3d_shader_param_allocator *allocator, size_t count, size_t stride) +{ + allocator->count = max(count, MAX_REG_OUTPUT); + allocator->stride = stride; + allocator->head = NULL; + allocator->current = NULL; + allocator->index = allocator->count; +} + +static void shader_param_allocator_destroy(struct vkd3d_shader_param_allocator *allocator) +{ + struct vkd3d_shader_param_node *current = allocator->head; + + while (current) + { + struct vkd3d_shader_param_node *next = current->next; + vkd3d_free(current); + current = next; + } +} + +void *shader_param_allocator_get(struct vkd3d_shader_param_allocator *allocator, size_t count) +{ + void *params; + + if (!allocator->current || count > allocator->count - allocator->index) + { + struct vkd3d_shader_param_node *next; + + allocator->count = max(allocator->count, count); + if (!(next = shader_param_allocator_node_create(allocator))) + return NULL; + if (allocator->current) + allocator->current->next = next; + else + allocator->head = next; + allocator->current = next; + allocator->index = 0; + } + + params = &allocator->current->param[allocator->index * allocator->stride]; + allocator->index += count; + + return params; +} + +static bool shader_instruction_array_reserve(struct vkd3d_shader_instruction_array *array, size_t reserve) +{ + if (!vkd3d_array_reserve((void **)&array->elements, &array->capacity, reserve, sizeof(*array->elements))) + { + ERR("Failed to allocate instructions.\n"); + return false; + } + + return true; +} + +bool shader_instruction_array_insert_at(struct vkd3d_shader_instruction_array *array, size_t idx, size_t count) +{ + VKD3D_ASSERT(idx <= array->count); + + if (!shader_instruction_array_reserve(array, array->count + count)) + return false; + + memmove(&array->elements[idx + count], &array->elements[idx], (array->count - idx) * sizeof(*array->elements)); + memset(&array->elements[idx], 0, count * sizeof(*array->elements)); + array->count += count; + + return true; +} + +struct vkd3d_shader_instruction *shader_instruction_array_append(struct vkd3d_shader_instruction_array *array) +{ + if (!shader_instruction_array_insert_at(array, array->count, 1)) + return NULL; + + return &array->elements[array->count - 1]; +} + +bool vsir_program_add_icb(struct vsir_program *program, struct vkd3d_shader_immediate_constant_buffer *icb) +{ + if (!vkd3d_array_reserve((void **)&program->icbs, &program->icb_capacity, + program->icb_count + 1, sizeof(*program->icbs))) + return false; + + program->icbs[program->icb_count++] = icb; + + return true; +} + +static struct vkd3d_shader_src_param *vsir_program_clone_src_params( + struct vsir_program *program, const struct vkd3d_shader_src_param *params, size_t count); + +static bool shader_register_clone_relative_addresses(struct vkd3d_shader_register *reg, struct vsir_program *program) +{ + size_t i; + + for (i = 0; i < reg->idx_count; ++i) + { + if (!reg->idx[i].rel_addr) + continue; + + if (!(reg->idx[i].rel_addr = vsir_program_clone_src_params(program, reg->idx[i].rel_addr, 1))) + return false; + } + + return true; +} + +static struct vkd3d_shader_dst_param *vsir_program_clone_dst_params( + struct vsir_program *program, const struct vkd3d_shader_dst_param *params, size_t count) +{ + struct vkd3d_shader_dst_param *dst_params; + size_t i; + + if (!(dst_params = vsir_program_get_dst_params(program, count))) + return NULL; + + memcpy(dst_params, params, count * sizeof(*params)); + for (i = 0; i < count; ++i) + { + if (!shader_register_clone_relative_addresses(&dst_params[i].reg, program)) + return NULL; + } + + return dst_params; +} + +static struct vkd3d_shader_src_param *vsir_program_clone_src_params( + struct vsir_program *program, const struct vkd3d_shader_src_param *params, size_t count) +{ + struct vkd3d_shader_src_param *src_params; + size_t i; + + if (!(src_params = vsir_program_get_src_params(program, count))) + return NULL; + + memcpy(src_params, params, count * sizeof(*params)); + for (i = 0; i < count; ++i) + { + if (!shader_register_clone_relative_addresses(&src_params[i].reg, program)) + return NULL; + } + + return src_params; +} + +static void shader_instruction_array_destroy(struct vkd3d_shader_instruction_array *array) +{ + vkd3d_free(array->elements); +} + +static bool shader_instruction_array_init(struct vkd3d_shader_instruction_array *array, size_t reserve) +{ + memset(array, 0, sizeof(*array)); + + return shader_instruction_array_reserve(array, reserve); +} + static int convert_parameter_info(const struct vkd3d_shader_compile_info *compile_info, unsigned int *ret_count, const struct vkd3d_shader_parameter1 **ret_parameters) { @@ -444,6 +661,11 @@ bool vsir_program_init(struct vsir_program *program, const struct vkd3d_shader_c program->shader_version = *version; program->cf_type = cf_type; program->normalisation_level = normalisation_level; + + /* Size the parameter initial allocations so they are large enough for most shaders. The + * code path for chained allocations will be tested if a few shaders need to use it. */ + shader_param_allocator_init(&program->dst_params, reserve - reserve / 8u, sizeof(struct vkd3d_shader_dst_param)); + shader_param_allocator_init(&program->src_params, reserve * 2u, sizeof(struct vkd3d_shader_src_param)); if (!shader_instruction_array_init(&program->instructions, reserve)) { if (program->free_parameters) @@ -471,6 +693,13 @@ void vsir_program_cleanup(struct vsir_program *program) shader_signature_cleanup(&program->output_signature); shader_signature_cleanup(&program->patch_constant_signature); vkd3d_shader_free_scan_descriptor_info1(&program->descriptors); + shader_param_allocator_destroy(&program->src_params); + shader_param_allocator_destroy(&program->dst_params); + for (i = 0; i < program->icb_count; ++i) + { + vkd3d_free(program->icbs[i]); + } + vkd3d_free(program->icbs); } const struct vkd3d_shader_parameter1 *vsir_program_get_parameter( @@ -796,16 +1025,15 @@ static void dst_param_init_output(struct vkd3d_shader_dst_param *dst, dst->write_mask = write_mask; } -void vsir_instruction_init(struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location, - enum vkd3d_shader_opcode opcode) +void vsir_instruction_init(struct vkd3d_shader_instruction *ins, + const struct vkd3d_shader_location *location, enum vkd3d_shader_opcode opcode) { - memset(ins, 0, sizeof(*ins)); - ins->location = *location; - ins->opcode = opcode; - ins->resource_data_type[0] = VSIR_DATA_F32; - ins->resource_data_type[1] = VSIR_DATA_F32; - ins->resource_data_type[2] = VSIR_DATA_F32; - ins->resource_data_type[3] = VSIR_DATA_F32; + *ins = (struct vkd3d_shader_instruction) + { + .location = *location, + .opcode = opcode, + .resource_data_type = {VSIR_DATA_F32, VSIR_DATA_F32, VSIR_DATA_F32, VSIR_DATA_F32}, + }; } bool vsir_instruction_init_with_params(struct vsir_program *program, @@ -857,11 +1085,19 @@ static bool vsir_instruction_is_dcl(const struct vkd3d_shader_instruction *instr || opcode == VSIR_OP_HS_DECLS; } -static void vkd3d_shader_instruction_make_nop(struct vkd3d_shader_instruction *ins) +/* NOTE: Immediate constant buffers are not cloned, so the source must not be destroyed while the + * destination is in use. This seems like a reasonable requirement given how this is currently used. */ +static bool vsir_program_iterator_clone_instruction(struct vsir_program *program, + struct vsir_program_iterator *dst_it, const struct vkd3d_shader_instruction *src) { - struct vkd3d_shader_location location = ins->location; + struct vkd3d_shader_instruction *dst = vsir_program_iterator_current(dst_it); + + *dst = *src; + + if (dst->dst_count && !(dst->dst = vsir_program_clone_dst_params(program, dst->dst, dst->dst_count))) + return false; - vsir_instruction_init(ins, &location, VSIR_OP_NOP); + return !dst->src_count || (dst->src = vsir_program_clone_src_params(program, dst->src, dst->src_count)); } static bool get_opcode_from_rel_op(enum vkd3d_shader_rel_op rel_op, @@ -1069,6 +1305,7 @@ static enum vkd3d_result vsir_program_lower_texkill(struct vsir_program *program ins->src[1].reg.u.immconst_f32[1] = 0.0f; ins->src[1].reg.u.immconst_f32[2] = 0.0f; ins->src[1].reg.u.immconst_f32[3] = 0.0f; + ins->src[1].swizzle = VKD3D_SHADER_NO_SWIZZLE; /* tmp.x = tmp.x || tmp.y */ /* tmp.x = tmp.x || tmp.z */ @@ -1422,7 +1659,7 @@ static enum vkd3d_result vsir_program_lower_texld_sm1(struct vsir_program *progr return VKD3D_ERROR_NOT_IMPLEMENTED; } - if (!(srcs = shader_src_param_allocator_get(&program->instructions.src_params, 3))) + if (!(srcs = vsir_program_get_src_params(program, 3))) return VKD3D_ERROR_OUT_OF_MEMORY; /* Note we run before I/O normalization. */ @@ -1498,7 +1735,7 @@ static enum vkd3d_result vsir_program_lower_texld(struct vsir_program *program, VKD3D_ASSERT(tex->src[1].reg.idx_count == 1); VKD3D_ASSERT(!tex->src[1].reg.idx[0].rel_addr); - if (!(srcs = shader_src_param_allocator_get(&program->instructions.src_params, 4))) + if (!(srcs = vsir_program_get_src_params(program, 4))) return VKD3D_ERROR_OUT_OF_MEMORY; srcs[0] = tex->src[0]; @@ -1541,7 +1778,7 @@ static enum vkd3d_result vsir_program_lower_texldd(struct vsir_program *program, VKD3D_ASSERT(texldd->src[1].reg.idx_count == 1); VKD3D_ASSERT(!texldd->src[1].reg.idx[0].rel_addr); - if (!(srcs = shader_src_param_allocator_get(&program->instructions.src_params, 5))) + if (!(srcs = vsir_program_get_src_params(program, 5))) return VKD3D_ERROR_OUT_OF_MEMORY; srcs[0] = texldd->src[0]; @@ -1567,7 +1804,7 @@ static enum vkd3d_result vsir_program_lower_texldl(struct vsir_program *program, VKD3D_ASSERT(texldl->src[1].reg.idx_count == 1); VKD3D_ASSERT(!texldl->src[1].reg.idx[0].rel_addr); - if (!(srcs = shader_src_param_allocator_get(&program->instructions.src_params, 4))) + if (!(srcs = vsir_program_get_src_params(program, 4))) return VKD3D_ERROR_OUT_OF_MEMORY; srcs[0] = texldl->src[0]; @@ -1585,6 +1822,64 @@ static enum vkd3d_result vsir_program_lower_texldl(struct vsir_program *program, return VKD3D_OK; } +static enum vkd3d_result vsir_program_lower_tex(struct vsir_program *program, struct vkd3d_shader_instruction *ins) +{ + unsigned int idx = ins->dst[0].reg.idx[0].offset; + struct vkd3d_shader_src_param *srcs; + + /* tex t# -> sample t#, t#, resource#, sampler# + * Note that the t# destination will subsequently be turned into a temp. */ + + /* We run before I/O normalization. */ + VKD3D_ASSERT(program->normalisation_level < VSIR_NORMALISED_SM6); + + if (!(srcs = vsir_program_get_src_params(program, 3))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + vsir_src_param_init(&srcs[0], VKD3DSPR_TEXTURE, VSIR_DATA_F32, 1); + srcs[0].reg.idx[0].offset = idx; + srcs[0].reg.dimension = VSIR_DIMENSION_VEC4; + srcs[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; + + vsir_src_param_init_resource(&srcs[1], idx, idx); + vsir_src_param_init_sampler(&srcs[2], idx, idx); + + ins->opcode = VSIR_OP_SAMPLE; + ins->src = srcs; + ins->src_count = 3; + + return VKD3D_OK; +} + +static enum vkd3d_result vsir_program_lower_texcoord(struct vsir_program *program, + struct vkd3d_shader_instruction *ins) +{ + unsigned int idx = ins->dst[0].reg.idx[0].offset; + struct vkd3d_shader_src_param *srcs; + + /* texcoord t# -> mov_sat t#, t# + * Note that the t# destination will subsequently be turned into a temp. */ + + /* We run before I/O normalization. */ + VKD3D_ASSERT(program->normalisation_level < VSIR_NORMALISED_SM6); + + if (!(srcs = vsir_program_get_src_params(program, 1))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + vsir_src_param_init(&srcs[0], VKD3DSPR_TEXTURE, VSIR_DATA_F32, 1); + srcs[0].reg.idx[0].offset = idx; + srcs[0].reg.dimension = VSIR_DIMENSION_VEC4; + srcs[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; + + ins->dst[0].modifiers |= VKD3DSPDM_SATURATE; + + ins->opcode = VSIR_OP_MOV; + ins->src = srcs; + ins->src_count = 1; + + return VKD3D_OK; +} + static enum vkd3d_result vsir_program_lower_dcl_input(struct vsir_program *program, struct vkd3d_shader_instruction *ins, struct vsir_transformation_context *ctx) { @@ -1669,10 +1964,31 @@ static enum vkd3d_result vsir_program_lower_d3dbc_instructions(struct vsir_progr switch (ins->opcode) { + case VSIR_OP_IFC: + ret = vsir_program_lower_ifc(program, &it, &tmp_idx, message_context); + break; + + case VSIR_OP_SINCOS: + ret = vsir_program_lower_sm1_sincos(program, &it); + break; + + case VSIR_OP_TEXCOORD: + if ((ret = vsir_program_lower_texcoord(program, ins)) < 0) + return ret; + break; + case VSIR_OP_TEXCRD: ret = vsir_program_lower_texcrd(program, ins, message_context); break; + case VSIR_OP_TEXKILL: + ret = vsir_program_lower_texkill(program, &it, &tmp_idx); + break; + + case VSIR_OP_TEX: + ret = vsir_program_lower_tex(program, ins); + break; + case VSIR_OP_TEXLD: if (program->shader_version.major == 1) ret = vsir_program_lower_texld_sm1(program, ins, message_context); @@ -1682,6 +1998,34 @@ static enum vkd3d_result vsir_program_lower_d3dbc_instructions(struct vsir_progr ret = vsir_program_lower_texld(program, ins, message_context); break; + case VSIR_OP_TEXLDD: + ret = vsir_program_lower_texldd(program, ins); + break; + + case VSIR_OP_TEXLDL: + ret = vsir_program_lower_texldl(program, ins); + break; + + case VSIR_OP_TEXBEM: + case VSIR_OP_TEXBEML: + case VSIR_OP_TEXDEPTH: + case VSIR_OP_TEXDP3: + case VSIR_OP_TEXDP3TEX: + case VSIR_OP_TEXM3x2PAD: + case VSIR_OP_TEXM3x2TEX: + case VSIR_OP_TEXM3x3DIFF: + case VSIR_OP_TEXM3x3PAD: + case VSIR_OP_TEXM3x3SPEC: + case VSIR_OP_TEXM3x3TEX: + case VSIR_OP_TEXM3x3VSPEC: + case VSIR_OP_TEXREG2AR: + case VSIR_OP_TEXREG2GB: + case VSIR_OP_TEXREG2RGB: + vkd3d_shader_error(ctx->message_context, &ins->location, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + "Aborting due to unimplemented feature: Combined sampler instruction \"%s\" (%#x).", + vsir_opcode_get_name(ins->opcode, ""), ins->opcode); + return VKD3D_ERROR_NOT_IMPLEMENTED; + default: ret = VKD3D_OK; break; @@ -1698,7 +2042,6 @@ static enum vkd3d_result vsir_program_lower_instructions(struct vsir_program *pr struct vsir_transformation_context *ctx) { struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - struct vkd3d_shader_message_context *message_context = ctx->message_context; struct vkd3d_shader_instruction *ins; unsigned int tmp_idx = ~0u; enum vkd3d_result ret; @@ -1707,16 +2050,6 @@ static enum vkd3d_result vsir_program_lower_instructions(struct vsir_program *pr { switch (ins->opcode) { - case VSIR_OP_IFC: - if ((ret = vsir_program_lower_ifc(program, &it, &tmp_idx, message_context)) < 0) - return ret; - break; - - case VSIR_OP_TEXKILL: - if ((ret = vsir_program_lower_texkill(program, &it, &tmp_idx)) < 0) - return ret; - break; - case VSIR_OP_MAD: if ((ret = vsir_program_lower_precise_mad(program, &it, &tmp_idx)) < 0) return ret; @@ -1725,6 +2058,8 @@ static enum vkd3d_result vsir_program_lower_instructions(struct vsir_program *pr case VSIR_OP_DCL: case VSIR_OP_DCL_CONSTANT_BUFFER: case VSIR_OP_DCL_GLOBAL_FLAGS: + case VSIR_OP_DCL_INPUT_PRIMITIVE: + case VSIR_OP_DCL_OUTPUT_TOPOLOGY: case VSIR_OP_DCL_SAMPLER: case VSIR_OP_DCL_TEMPS: case VSIR_OP_DCL_TESSELLATOR_DOMAIN: @@ -1765,51 +2100,67 @@ static enum vkd3d_result vsir_program_lower_instructions(struct vsir_program *pr break; case VSIR_OP_SINCOS: - if (ins->dst_count == 1) - { - if ((ret = vsir_program_lower_sm1_sincos(program, &it)) < 0) - return ret; - } - else - { - if ((ret = vsir_program_lower_sm4_sincos(program, &it, ctx)) < 0) - return ret; - } - break; - - case VSIR_OP_TEXLDD: - if ((ret = vsir_program_lower_texldd(program, ins)) < 0) + if ((ret = vsir_program_lower_sm4_sincos(program, &it, ctx)) < 0) return ret; break; - case VSIR_OP_TEXLDL: - if ((ret = vsir_program_lower_texldl(program, ins)) < 0) - return ret; + default: break; + } + } - case VSIR_OP_TEXBEM: - case VSIR_OP_TEXBEML: - case VSIR_OP_TEXCOORD: - case VSIR_OP_TEXDEPTH: - case VSIR_OP_TEXDP3: - case VSIR_OP_TEXDP3TEX: - case VSIR_OP_TEXM3x2PAD: - case VSIR_OP_TEXM3x2TEX: - case VSIR_OP_TEXM3x3DIFF: - case VSIR_OP_TEXM3x3PAD: - case VSIR_OP_TEXM3x3SPEC: - case VSIR_OP_TEXM3x3TEX: - case VSIR_OP_TEXM3x3VSPEC: - case VSIR_OP_TEXREG2AR: - case VSIR_OP_TEXREG2GB: - case VSIR_OP_TEXREG2RGB: - vkd3d_shader_error(ctx->message_context, &ins->location, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, - "Aborting due to unimplemented feature: Combined sampler instruction \"%s\" (%#x).", - vsir_opcode_get_name(ins->opcode, ""), ins->opcode); - return VKD3D_ERROR_NOT_IMPLEMENTED; + return VKD3D_OK; +} - default: - break; +/* 1.0-1.3 pixel shaders allow writing t# registers, at which point they + * effectively behave like normal r# temps. Convert them to r# registers. + * t# registers which are read before being written contain TEXCOORD varyings, + * just as in 1.4 and 2.x, and will later be lowered to v# registers. + * + * Registers which are partially written are rejected by the native validator, + * but with a "read of uninitialized component" message that suggests that once + * any component of a t# register is written, none of the components contain + * texcoord data. */ +static enum vkd3d_result vsir_program_lower_texture_writes(struct vsir_program *program, + struct vsir_transformation_context *ctx) +{ + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); + struct vkd3d_shader_instruction *ins; + unsigned int texture_temp_idx = ~0u; + uint32_t texture_written_mask = 0; + + /* We run before I/O normalization. */ + VKD3D_ASSERT(program->normalisation_level < VSIR_NORMALISED_SM6); + + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) + { + for (unsigned int i = 0; i < ins->src_count; ++i) + { + struct vkd3d_shader_src_param *src = &ins->src[i]; + + if (src->reg.type == VKD3DSPR_TEXTURE && bitmap_is_set(&texture_written_mask, src->reg.idx[0].offset)) + { + src->reg.type = VKD3DSPR_TEMP; + src->reg.idx[0].offset += texture_temp_idx; + } + } + + for (unsigned int i = 0; i < ins->dst_count; ++i) + { + struct vkd3d_shader_dst_param *dst = &ins->dst[i]; + + if (dst->reg.type == VKD3DSPR_TEXTURE) + { + bitmap_set(&texture_written_mask, dst->reg.idx[0].offset); + if (texture_temp_idx == ~0u) + { + texture_temp_idx = program->temp_count; + /* These versions have 4 texture registers. */ + program->temp_count += 4; + } + dst->reg.type = VKD3DSPR_TEMP; + dst->reg.idx[0].offset += texture_temp_idx; + } } } @@ -1823,16 +2174,19 @@ static enum vkd3d_result vsir_program_ensure_ret(struct vsir_program *program, struct vsir_transformation_context *ctx) { struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - static const struct vkd3d_shader_location no_loc; struct vkd3d_shader_instruction *ins; + struct vkd3d_shader_location loc; - ins = vsir_program_iterator_tail(&it); - if (ins && ins->opcode == VSIR_OP_RET) + if (!(ins = vsir_program_iterator_tail(&it))) + loc = ctx->null_location; + else if (ins->opcode == VSIR_OP_RET) return VKD3D_OK; + else + loc = ins->location; if (!(ins = vsir_program_append(program))) return VKD3D_ERROR_OUT_OF_MEMORY; - vsir_instruction_init(ins, &no_loc, VSIR_OP_RET); + vsir_instruction_init(ins, &loc, VSIR_OP_RET); return VKD3D_OK; } @@ -1923,7 +2277,7 @@ static enum vkd3d_result vsir_program_ensure_diffuse(struct vsir_program *progra struct vsir_transformation_context *ctx) { struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - static const struct vkd3d_shader_location no_loc; + struct vkd3d_shader_location loc = ctx->null_location; struct vkd3d_shader_instruction *ins; unsigned int i; @@ -1937,21 +2291,23 @@ static enum vkd3d_result vsir_program_ensure_diffuse(struct vsir_program *progra for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { if (!vsir_instruction_is_dcl(ins) && ins->opcode != VSIR_OP_LABEL && ins->opcode != VSIR_OP_NOP) + { + loc = ins->location; break; + } } - vsir_program_iterator_prev(&it); - if (!vsir_program_iterator_insert_after(&it, 1)) + if (!(ins = vsir_program_iterator_insert_before_and_move(&it, 1))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(&it); - vsir_instruction_init_with_params(program, ins, &no_loc, VSIR_OP_MOV, 1, 1); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); vsir_dst_param_init(&ins->dst[0], VKD3DSPR_ATTROUT, VSIR_DATA_F32, 1); ins->dst[0].reg.idx[0].offset = 0; ins->dst[0].reg.dimension = VSIR_DIMENSION_VEC4; ins->dst[0].write_mask = VKD3DSP_WRITEMASK_ALL & ~program->diffuse_written_mask; vsir_src_param_init(&ins->src[0], VKD3DSPR_IMMCONST, VSIR_DATA_F32, 0); ins->src[0].reg.dimension = VSIR_DIMENSION_VEC4; + ins->src[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; for (i = 0; i < 4; ++i) ins->src[0].reg.u.immconst_f32[i] = 1.0f; return VKD3D_OK; @@ -2162,10 +2518,8 @@ static enum vkd3d_result vsir_program_remap_output_signature(struct vsir_program continue; loc = ins->location; - vsir_program_iterator_prev(&it); - if (!vsir_program_iterator_insert_after(&it, uninit_varying_count)) + if (!(ins = vsir_program_iterator_insert_before_and_move(&it, uninit_varying_count))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(&it); for (unsigned int j = signature->element_count - uninit_varying_count; j < signature->element_count; ++j) { @@ -2175,6 +2529,7 @@ static enum vkd3d_result vsir_program_remap_output_signature(struct vsir_program dst_param_init_output(&ins->dst[0], VSIR_DATA_F32, e->register_index, e->mask); vsir_src_param_init(&ins->src[0], VKD3DSPR_IMMCONST, VSIR_DATA_F32, 0); ins->src[0].reg.dimension = VSIR_DIMENSION_VEC4; + ins->src[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; ins = vsir_program_iterator_next(&it); } } @@ -2200,8 +2555,6 @@ struct hull_flattener { struct vsir_program *program; - unsigned int instance_count; - unsigned int phase_body_idx; enum vkd3d_shader_opcode phase; struct vkd3d_shader_location last_ret_location; unsigned int *ssa_map; @@ -2213,68 +2566,6 @@ static bool flattener_is_in_fork_or_join_phase(const struct hull_flattener *flat return flattener->phase == VSIR_OP_HS_FORK_PHASE || flattener->phase == VSIR_OP_HS_JOIN_PHASE; } -struct shader_phase_location -{ - unsigned int index; - unsigned int instance_count; - unsigned int instruction_count; -}; - -struct shader_phase_location_array -{ - /* Unlikely worst case: one phase for each component of each output register. */ - struct shader_phase_location locations[MAX_REG_OUTPUT * VKD3D_VEC4_SIZE]; - unsigned int count; -}; - -static void flattener_eliminate_phase_related_dcls(struct hull_flattener *normaliser, unsigned int index, - struct vkd3d_shader_instruction *ins, struct shader_phase_location_array *locations) -{ - struct shader_phase_location *loc; - bool b; - - if (ins->opcode == VSIR_OP_HS_FORK_PHASE || ins->opcode == VSIR_OP_HS_JOIN_PHASE) - { - b = flattener_is_in_fork_or_join_phase(normaliser); - /* Reset the phase info. */ - normaliser->phase_body_idx = ~0u; - normaliser->phase = ins->opcode; - normaliser->instance_count = 1; - /* Leave the first occurrence and delete the rest. */ - if (b) - vkd3d_shader_instruction_make_nop(ins); - return; - } - else if (ins->opcode == VSIR_OP_DCL_HS_FORK_PHASE_INSTANCE_COUNT - || ins->opcode == VSIR_OP_DCL_HS_JOIN_PHASE_INSTANCE_COUNT) - { - normaliser->instance_count = ins->declaration.count + !ins->declaration.count; - vkd3d_shader_instruction_make_nop(ins); - return; - } - - if (normaliser->phase == VSIR_OP_INVALID || vsir_instruction_is_dcl(ins)) - return; - - if (normaliser->phase_body_idx == ~0u) - normaliser->phase_body_idx = index; - - if (ins->opcode == VSIR_OP_RET) - { - normaliser->last_ret_location = ins->location; - vkd3d_shader_instruction_make_nop(ins); - if (locations->count >= ARRAY_SIZE(locations->locations)) - { - FIXME("Insufficient space for phase location.\n"); - return; - } - loc = &locations->locations[locations->count++]; - loc->index = normaliser->phase_body_idx; - loc->instance_count = normaliser->instance_count; - loc->instruction_count = index - normaliser->phase_body_idx; - } -} - static void flattener_fixup_ssa_register(struct hull_flattener *normaliser, struct vkd3d_shader_register *reg, unsigned int instance_id) { @@ -2337,54 +2628,110 @@ static void flattener_fixup_registers(struct hull_flattener *normaliser, flattener_fixup_register_indices(normaliser, &ins->dst[i].reg, instance_id); } -static enum vkd3d_result flattener_flatten_phases(struct hull_flattener *normaliser, - struct shader_phase_location_array *locations) +static enum vkd3d_result flattener_replicate_location(struct hull_flattener *normaliser, + struct vsir_program_iterator *it, size_t instance_count, size_t instruction_count) { - struct vkd3d_shader_instruction_array *instructions = &normaliser->program->instructions; - struct shader_phase_location *loc; - unsigned int i, j, k, end, count; - - for (i = 0, count = 0; i < locations->count; ++i) - count += (locations->locations[i].instance_count - 1) * locations->locations[i].instruction_count; + struct vsir_program_iterator dst_it, src_it, first_it; + struct vkd3d_shader_instruction *ins, *src_ins; + unsigned int i, j; + size_t count; - if (!shader_instruction_array_reserve(instructions, instructions->count + count)) + VKD3D_ASSERT(instance_count); + count = (instance_count - 1) * instruction_count; + if (!vsir_program_iterator_insert_before(it, &first_it, count)) return VKD3D_ERROR_OUT_OF_MEMORY; - end = instructions->count; - instructions->count += count; - for (i = locations->count; i > 0; --i) + /* Make a copy of the non-dcl instructions for each instance. */ + dst_it = first_it; + for (i = 1; i < instance_count; ++i) + { + src_it = *it; + for (j = 0; j < instruction_count; ++j) + { + src_ins = vsir_program_iterator_current(&src_it); + if (!vsir_program_iterator_clone_instruction(normaliser->program, &dst_it, src_ins)) + return VKD3D_ERROR_OUT_OF_MEMORY; + + vsir_program_iterator_next(&dst_it); + vsir_program_iterator_next(&src_it); + } + } + /* Replace each reference to the instance id with a constant instance id. */ + *it = first_it; + for (i = 0; i < instance_count; ++i) { - loc = &locations->locations[i - 1]; - j = loc->index + loc->instruction_count; - memmove(&instructions->elements[j + count], &instructions->elements[j], - (end - j) * sizeof(*instructions->elements)); - end = j; - count -= (loc->instance_count - 1) * loc->instruction_count; - loc->index += count; + if (i) + memset(normaliser->ssa_map, 0xff, normaliser->orig_ssa_count * sizeof(*normaliser->ssa_map)); + + for (j = 0; j < instruction_count; ++j) + { + ins = vsir_program_iterator_current(it); + flattener_fixup_registers(normaliser, ins, i); + vsir_program_iterator_next(it); + } } - for (i = 0, count = 0; i < locations->count; ++i) + return VKD3D_OK; +} + +static enum vkd3d_result flattener_flatten_phases(struct hull_flattener *normaliser) +{ + struct vsir_program_iterator it = vsir_program_iterator(&normaliser->program->instructions); + struct vsir_program_iterator phase_body_it; + struct vkd3d_shader_instruction *ins; + bool b, phase_body_it_valid = false; + unsigned int instruction_count = 0; + unsigned int instance_count = 0; + enum vkd3d_result res; + + normaliser->phase = VSIR_OP_INVALID; + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { - loc = &locations->locations[i]; - /* Make a copy of the non-dcl instructions for each instance. */ - for (j = 1; j < loc->instance_count; ++j) + if (ins->opcode == VSIR_OP_HS_FORK_PHASE || ins->opcode == VSIR_OP_HS_JOIN_PHASE) { - for (k = 0; k < loc->instruction_count; ++k) - { - if (!shader_instruction_array_clone_instruction(instructions, - loc->index + loc->instruction_count * j + k, loc->index + k)) - return VKD3D_ERROR_OUT_OF_MEMORY; - } + b = flattener_is_in_fork_or_join_phase(normaliser); + /* Reset the phase info. */ + phase_body_it_valid = false; + normaliser->phase = ins->opcode; + instance_count = 1; + instruction_count = 0; + /* Leave the first occurrence and delete the rest. */ + if (b) + vkd3d_shader_instruction_make_nop(ins); + continue; } - /* Replace each reference to the instance id with a constant instance id. */ - for (j = 0; j < loc->instance_count; ++j) + else if (ins->opcode == VSIR_OP_DCL_HS_FORK_PHASE_INSTANCE_COUNT + || ins->opcode == VSIR_OP_DCL_HS_JOIN_PHASE_INSTANCE_COUNT) { - if (j != 0) - memset(normaliser->ssa_map, 0xff, normaliser->orig_ssa_count * sizeof(*normaliser->ssa_map)); + instance_count = ins->declaration.count + !ins->declaration.count; + vkd3d_shader_instruction_make_nop(ins); + ++instruction_count; + continue; + } + + if (normaliser->phase == VSIR_OP_INVALID) + continue; + + if (!phase_body_it_valid && !vsir_instruction_is_dcl(ins)) + { + phase_body_it_valid = true; + phase_body_it = it; + instruction_count = 0; + } - for (k = 0; k < loc->instruction_count; ++k) - flattener_fixup_registers(normaliser, - &instructions->elements[loc->index + loc->instruction_count * j + k], j); + if (ins->opcode == VSIR_OP_RET) + { + normaliser->last_ret_location = ins->location; + vkd3d_shader_instruction_make_nop(ins); + it = phase_body_it; + if ((res = flattener_replicate_location(normaliser, &it, + instance_count, instruction_count)) < 0) + return res; + phase_body_it_valid = false; + } + else + { + ++instruction_count; } } @@ -2394,19 +2741,10 @@ static enum vkd3d_result flattener_flatten_phases(struct hull_flattener *normali static enum vkd3d_result vsir_program_flatten_hull_shader_phases(struct vsir_program *program, struct vsir_transformation_context *ctx) { - struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - struct shader_phase_location_array locations; struct hull_flattener flattener = {program}; struct vkd3d_shader_instruction *ins; enum vkd3d_result result = VKD3D_OK; - unsigned int i; - flattener.phase = VSIR_OP_INVALID; - locations.count = 0; - for (ins = vsir_program_iterator_head(&it), i = 0; ins; ins = vsir_program_iterator_next(&it), ++i) - { - flattener_eliminate_phase_related_dcls(&flattener, i, ins, &locations); - } bitmap_clear(program->io_dcls, VKD3DSPR_FORKINSTID); bitmap_clear(program->io_dcls, VKD3DSPR_JOININSTID); @@ -2414,8 +2752,7 @@ static enum vkd3d_result vsir_program_flatten_hull_shader_phases(struct vsir_pro if (!(flattener.ssa_map = vkd3d_calloc(flattener.orig_ssa_count, sizeof(*flattener.ssa_map)))) return VKD3D_ERROR_OUT_OF_MEMORY; - result = flattener_flatten_phases(&flattener, &locations); - + result = flattener_flatten_phases(&flattener); vkd3d_free(flattener.ssa_map); flattener.ssa_map = NULL; @@ -2434,7 +2771,7 @@ static enum vkd3d_result vsir_program_flatten_hull_shader_phases(struct vsir_pro struct control_point_normaliser { - struct vkd3d_shader_instruction_array instructions; + struct vsir_program *program; enum vkd3d_shader_opcode phase; struct vkd3d_shader_src_param *outpointid_param; }; @@ -2446,20 +2783,15 @@ static bool control_point_normaliser_is_in_control_point_phase(const struct cont struct vkd3d_shader_src_param *vsir_program_create_outpointid_param(struct vsir_program *program) { - struct vkd3d_shader_instruction_array *instructions = &program->instructions; struct vkd3d_shader_src_param *rel_addr; - if (instructions->outpointid_param) - return instructions->outpointid_param; - - if (!(rel_addr = shader_src_param_allocator_get(&instructions->src_params, 1))) + if (!(rel_addr = vsir_program_get_src_params(program, 1))) return NULL; vsir_register_init(&rel_addr->reg, VKD3DSPR_OUTPOINTID, VSIR_DATA_U32, 0); rel_addr->swizzle = 0; rel_addr->modifiers = 0; - instructions->outpointid_param = rel_addr; return rel_addr; } @@ -2481,8 +2813,8 @@ static void shader_dst_param_normalise_outpointid(struct vkd3d_shader_dst_param } static enum vkd3d_result control_point_normaliser_emit_hs_input(struct control_point_normaliser *normaliser, - const struct shader_signature *s, unsigned int input_control_point_count, unsigned int dst, - const struct vkd3d_shader_location *location) + const struct shader_signature *s, unsigned int input_control_point_count, + struct vsir_program_iterator *dst_it, const struct vkd3d_shader_location *location) { struct vkd3d_shader_instruction *ins; const struct signature_element *e; @@ -2491,17 +2823,11 @@ static enum vkd3d_result control_point_normaliser_emit_hs_input(struct control_p for (i = 0; i < s->element_count; ++i) count += !!s->elements[i].used_mask; - if (!shader_instruction_array_reserve(&normaliser->instructions, normaliser->instructions.count + count)) + if (!(ins = vsir_program_iterator_insert_before_and_move(dst_it, count))) return VKD3D_ERROR_OUT_OF_MEMORY; - - memmove(&normaliser->instructions.elements[dst + count], &normaliser->instructions.elements[dst], - (normaliser->instructions.count - dst) * sizeof(*normaliser->instructions.elements)); - normaliser->instructions.count += count; - - ins = &normaliser->instructions.elements[dst]; vsir_instruction_init(ins, location, VSIR_OP_HS_CONTROL_POINT_PHASE); - ++ins; + ins = vsir_program_iterator_next(dst_it); for (i = 0; i < s->element_count; ++i) { @@ -2510,9 +2836,9 @@ static enum vkd3d_result control_point_normaliser_emit_hs_input(struct control_p continue; vsir_instruction_init(ins, location, VSIR_OP_MOV); - ins->dst = shader_dst_param_allocator_get(&normaliser->instructions.dst_params, 1); + ins->dst = vsir_program_get_dst_params(normaliser->program, 1); ins->dst_count = 1; - ins->src = shader_src_param_allocator_get(&normaliser->instructions.src_params, 1); + ins->src = vsir_program_get_src_params(normaliser->program, 1); ins->src_count = 1; if (!ins->dst || ! ins->src) @@ -2521,6 +2847,8 @@ static enum vkd3d_result control_point_normaliser_emit_hs_input(struct control_p return VKD3D_ERROR_OUT_OF_MEMORY; } + VKD3D_ASSERT(normaliser->outpointid_param); + vsir_dst_param_init_io(&ins->dst[0], VKD3DSPR_OUTPUT, e, 2); ins->dst[0].reg.dimension = VSIR_DIMENSION_VEC4; ins->dst[0].reg.idx[0].offset = 0; @@ -2533,7 +2861,7 @@ static enum vkd3d_result control_point_normaliser_emit_hs_input(struct control_p ins->src[0].reg.idx[0].rel_addr = normaliser->outpointid_param; ins->src[0].reg.idx[1].offset = e->register_index; - ++ins; + ins = vsir_program_iterator_next(dst_it); } vsir_instruction_init(ins, location, VSIR_OP_RET); @@ -2565,8 +2893,8 @@ static enum vkd3d_result instruction_array_normalise_hull_shader_control_point_i ERR("Failed to allocate src param.\n"); return VKD3D_ERROR_OUT_OF_MEMORY; } - normaliser.instructions = program->instructions; - it = vsir_program_iterator(&normaliser.instructions); + normaliser.program = program; + it = vsir_program_iterator(&normaliser.program->instructions); normaliser.phase = VSIR_OP_INVALID; for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) @@ -2598,16 +2926,14 @@ static enum vkd3d_result instruction_array_normalise_hull_shader_control_point_i input_control_point_count = ins->declaration.count; break; case VSIR_OP_HS_CONTROL_POINT_PHASE: - program->instructions = normaliser.instructions; program->normalisation_level = VSIR_NORMALISED_HULL_CONTROL_POINT_IO; return VKD3D_OK; case VSIR_OP_HS_FORK_PHASE: case VSIR_OP_HS_JOIN_PHASE: /* ins may be relocated if the instruction array expands. */ location = ins->location; - ret = control_point_normaliser_emit_hs_input(&normaliser, &program->input_signature, - input_control_point_count, i, &location); - program->instructions = normaliser.instructions; + ret = control_point_normaliser_emit_hs_input(&normaliser, + &program->input_signature, input_control_point_count, &it, &location); program->normalisation_level = VSIR_NORMALISED_HULL_CONTROL_POINT_IO; return ret; default: @@ -2615,7 +2941,6 @@ static enum vkd3d_result instruction_array_normalise_hull_shader_control_point_i } } - program->instructions = normaliser.instructions; program->normalisation_level = VSIR_NORMALISED_HULL_CONTROL_POINT_IO; return VKD3D_OK; } @@ -2641,12 +2966,7 @@ struct io_normaliser struct shader_signature *output_signature; struct shader_signature *patch_constant_signature; - unsigned int instance_count; - unsigned int phase_body_idx; enum vkd3d_shader_opcode phase; - unsigned int output_control_point_count; - - struct vkd3d_shader_src_param *outpointid_param; struct vkd3d_shader_dst_param *input_dcl_params[MAX_REG_OUTPUT]; struct vkd3d_shader_dst_param *output_dcl_params[MAX_REG_OUTPUT]; @@ -3268,9 +3588,6 @@ static enum vkd3d_result vsir_program_normalise_io_registers(struct vsir_program { switch (ins->opcode) { - case VSIR_OP_DCL_OUTPUT_CONTROL_POINT_COUNT: - normaliser.output_control_point_count = ins->declaration.count; - break; case VSIR_OP_DCL_INDEX_RANGE: if ((ret = io_normaliser_add_index_range(&normaliser, ins)) < 0) return ret; @@ -3355,6 +3672,7 @@ static void shader_register_normalise_flat_constants(struct vkd3d_shader_src_par { enum vkd3d_shader_d3dbc_constant_register set; struct vkd3d_shader_src_param *rel_addr; + unsigned int c; uint32_t index; size_t i, j; @@ -3369,7 +3687,11 @@ static void shader_register_normalise_flat_constants(struct vkd3d_shader_src_par param->reg.idx_count = 0; param->reg.dimension = VSIR_DIMENSION_VEC4; for (j = 0; j < 4; ++j) - param->reg.u.immconst_u32[j] = normaliser->defs[i].value[j]; + { + c = vsir_swizzle_get_component(param->swizzle, j); + param->reg.u.immconst_u32[j] = normaliser->defs[i].value[c]; + } + param->swizzle = VKD3D_SHADER_NO_SWIZZLE; return; } } @@ -3521,6 +3843,17 @@ static enum vkd3d_result vsir_program_remove_dead_code(struct vsir_program *prog return VKD3D_OK; } +/* NOTE: This invalidates all iterators, since the position of the + * vkd3d_shader_instruction_array itself changes. */ +static void vsir_program_replace_instructions(struct vsir_program *program, + struct vkd3d_shader_instruction_array *array) +{ + shader_instruction_array_destroy(&program->instructions); + + program->instructions = *array; + memset(array, 0, sizeof(*array)); +} + struct cf_flattener_if_info { struct vkd3d_shader_src_param *false_param; @@ -3544,7 +3877,7 @@ struct cf_flattener_switch_case struct cf_flattener_switch_info { - size_t ins_location; + struct vsir_program_iterator ins_it; const struct vkd3d_shader_src_param *condition; unsigned int id; unsigned int merge_block_id; @@ -3579,9 +3912,7 @@ struct cf_flattener struct vkd3d_shader_location location; enum vkd3d_result status; - struct vkd3d_shader_instruction *instructions; - size_t instruction_capacity; - size_t instruction_count; + struct vkd3d_shader_instruction_array instructions; unsigned int block_id; const char **block_names; @@ -3604,18 +3935,6 @@ static void cf_flattener_set_error(struct cf_flattener *flattener, enum vkd3d_re flattener->status = error; } -static struct vkd3d_shader_instruction *cf_flattener_require_space(struct cf_flattener *flattener, size_t count) -{ - if (!vkd3d_array_reserve((void **)&flattener->instructions, &flattener->instruction_capacity, - flattener->instruction_count + count, sizeof(*flattener->instructions))) - { - ERR("Failed to allocate instructions.\n"); - cf_flattener_set_error(flattener, VKD3D_ERROR_OUT_OF_MEMORY); - return NULL; - } - return &flattener->instructions[flattener->instruction_count]; -} - static bool cf_flattener_copy_instruction(struct cf_flattener *flattener, const struct vkd3d_shader_instruction *instruction) { @@ -3624,11 +3943,10 @@ static bool cf_flattener_copy_instruction(struct cf_flattener *flattener, if (instruction->opcode == VSIR_OP_NOP) return true; - if (!(dst_ins = cf_flattener_require_space(flattener, 1))) + if (!(dst_ins = shader_instruction_array_append(&flattener->instructions))) return false; *dst_ins = *instruction; - ++flattener->instruction_count; return true; } @@ -3656,12 +3974,13 @@ static void cf_flattener_emit_label(struct cf_flattener *flattener, unsigned int { struct vkd3d_shader_instruction *ins; - if (!(ins = cf_flattener_require_space(flattener, 1))) + if (!(ins = shader_instruction_array_append(&flattener->instructions))) return; - if (vsir_instruction_init_label(ins, &flattener->location, label_id, flattener->program)) - ++flattener->instruction_count; - else + if (!vsir_instruction_init_label(ins, &flattener->location, label_id, flattener->program)) + { + vkd3d_shader_instruction_make_nop(ins); cf_flattener_set_error(flattener, VKD3D_ERROR_OUT_OF_MEMORY); + } } /* For conditional branches, this returns the false target branch parameter. */ @@ -3673,14 +3992,17 @@ static struct vkd3d_shader_src_param *cf_flattener_emit_branch(struct cf_flatten struct vkd3d_shader_src_param *src_params, *false_branch_param; struct vkd3d_shader_instruction *ins; - if (!(ins = cf_flattener_require_space(flattener, 1))) + if (!(ins = shader_instruction_array_append(&flattener->instructions))) return NULL; vsir_instruction_init(ins, &flattener->location, VSIR_OP_BRANCH); if (condition) { if (!(src_params = instruction_src_params_alloc(ins, 4 + !!continue_block_id, flattener))) + { + vkd3d_shader_instruction_make_nop(ins); return NULL; + } src_params[0] = *condition; if (flags == VKD3D_SHADER_CONDITIONAL_OP_Z) { @@ -3701,7 +4023,10 @@ static struct vkd3d_shader_src_param *cf_flattener_emit_branch(struct cf_flatten else { if (!(src_params = instruction_src_params_alloc(ins, merge_block_id ? 3 : 1, flattener))) + { + vkd3d_shader_instruction_make_nop(ins); return NULL; + } vsir_src_param_init_label(&src_params[0], true_id); if (merge_block_id) { @@ -3713,8 +4038,6 @@ static struct vkd3d_shader_src_param *cf_flattener_emit_branch(struct cf_flatten false_branch_param = NULL; } - ++flattener->instruction_count; - return false_branch_param; } @@ -3820,8 +4143,12 @@ static enum vkd3d_result cf_flattener_iterate_instruction_array(struct cf_flatte is_hull_shader = program->shader_version.type == VKD3D_SHADER_TYPE_HULL; after_declarations_section = is_hull_shader; - if (!cf_flattener_require_space(flattener, instructions->count + 1)) + if (!shader_instruction_array_reserve(&flattener->instructions, instructions->count + 1)) + { + ERR("Failed to allocate instructions.\n"); + cf_flattener_set_error(flattener, VKD3D_ERROR_OUT_OF_MEMORY); return VKD3D_ERROR_OUT_OF_MEMORY; + } it = vsir_program_iterator(instructions); for (instruction = vsir_program_iterator_head(&it); instruction; instruction = vsir_program_iterator_next(&it)) @@ -3961,13 +4288,14 @@ static enum vkd3d_result cf_flattener_iterate_instruction_array(struct cf_flatte merge_block_id = cf_flattener_alloc_block_id(flattener); - cf_info->u.switch_.ins_location = flattener->instruction_count; cf_info->u.switch_.condition = src; - if (!(dst_ins = cf_flattener_require_space(flattener, 1))) + if (!(dst_ins = shader_instruction_array_append(&flattener->instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; vsir_instruction_init(dst_ins, &instruction->location, VSIR_OP_SWITCH_MONOLITHIC); - ++flattener->instruction_count; + + cf_info->u.switch_.ins_it = vsir_program_iterator(&flattener->instructions); + vsir_program_iterator_tail(&cf_info->u.switch_.ins_it); cf_info->u.switch_.id = flattener->switch_id; cf_info->u.switch_.merge_block_id = merge_block_id; @@ -4000,8 +4328,12 @@ static enum vkd3d_result cf_flattener_iterate_instruction_array(struct cf_flatte /* The SWITCH instruction is completed when the endswitch * instruction is processed because we do not know the number * of case statements or the default block id in advance.*/ - dst_ins = &flattener->instructions[cf_info->u.switch_.ins_location]; - if (!(src_params = instruction_src_params_alloc(dst_ins, cf_info->u.switch_.cases_count * 2 + 3, flattener))) + /* NOTE: This relies on iterators not being invalidated + * when new instructions are appended to the + * vkd3d_shader_instruction_array. */ + dst_ins = vsir_program_iterator_current(&cf_info->u.switch_.ins_it); + if (!(src_params = instruction_src_params_alloc(dst_ins, + cf_info->u.switch_.cases_count * 2 + 3, flattener))) { vkd3d_free(cf_info->u.switch_.cases); return VKD3D_ERROR_OUT_OF_MEMORY; @@ -4162,18 +4494,17 @@ static enum vkd3d_result vsir_program_flatten_control_flow_constructs(struct vsi VKD3D_ASSERT(program->cf_type == VSIR_CF_STRUCTURED); + shader_instruction_array_init(&flattener.instructions, 0); + if ((result = cf_flattener_iterate_instruction_array(&flattener, message_context)) >= 0) { - vkd3d_free(program->instructions.elements); - program->instructions.elements = flattener.instructions; - program->instructions.capacity = flattener.instruction_capacity; - program->instructions.count = flattener.instruction_count; + vsir_program_replace_instructions(program, &flattener.instructions); program->block_count = flattener.block_id; program->cf_type = VSIR_CF_BLOCKS; } else { - vkd3d_free(flattener.instructions); + shader_instruction_array_destroy(&flattener.instructions); } vkd3d_free(flattener.control_flow_info); @@ -4190,17 +4521,6 @@ static unsigned int label_from_src_param(const struct vkd3d_shader_src_param *pa return param->reg.idx[0].offset; } -static bool reserve_instructions(struct vkd3d_shader_instruction **instructions, size_t *capacity, size_t count) -{ - if (!vkd3d_array_reserve((void **)instructions, capacity, count, sizeof(**instructions))) - { - ERR("Failed to allocate instructions.\n"); - return false; - } - - return true; -} - /* A record represents replacing a jump from block `switch_label' to * block `target_label' with a jump from block `if_label' to block * `target_label'. */ @@ -4233,39 +4553,40 @@ static enum vkd3d_result vsir_program_lower_switch_to_selection_ladder(struct vs struct vsir_transformation_context *ctx) { unsigned int block_count = program->block_count, ssa_count = program->ssa_count, current_label = 0, if_label; - size_t ins_capacity = 0, ins_count = 0, i, map_capacity = 0, map_count = 0; - struct vkd3d_shader_instruction *instructions = NULL; + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct lower_switch_to_if_ladder_block_mapping *block_map = NULL; + struct vkd3d_shader_instruction_array instructions; + struct vkd3d_shader_instruction *ins, *dst_ins; + size_t map_capacity = 0, map_count = 0; VKD3D_ASSERT(program->cf_type == VSIR_CF_BLOCKS); - if (!reserve_instructions(&instructions, &ins_capacity, program->instructions.count)) - goto fail; + if (!shader_instruction_array_init(&instructions, program->instructions.count)) + return VKD3D_ERROR_OUT_OF_MEMORY; /* First subpass: convert SWITCH_MONOLITHIC instructions to * selection ladders, keeping a map between blocks before and * after the subpass. */ - for (i = 0; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; unsigned int case_count, j, default_label; switch (ins->opcode) { case VSIR_OP_LABEL: current_label = label_from_src_param(&ins->src[0]); - if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 1)) + if (!(dst_ins = shader_instruction_array_append(&instructions))) goto fail; - instructions[ins_count++] = *ins; + *dst_ins = *ins; continue; case VSIR_OP_SWITCH_MONOLITHIC: break; default: - if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 1)) + if (!(dst_ins = shader_instruction_array_append(&instructions))) goto fail; - instructions[ins_count++] = *ins; + *dst_ins = *ins; continue; } @@ -4276,17 +4597,18 @@ static enum vkd3d_result vsir_program_lower_switch_to_selection_ladder(struct vs * just have to jump to the default label. */ if (case_count == 0) { - if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 1)) + if (!(dst_ins = shader_instruction_array_append(&instructions))) goto fail; - if (!vsir_instruction_init_with_params(program, &instructions[ins_count], - &ins->location, VSIR_OP_BRANCH, 0, 1)) + if (!vsir_instruction_init_with_params(program, dst_ins, &ins->location, VSIR_OP_BRANCH, 0, 1)) + { + vkd3d_shader_instruction_make_nop(dst_ins); goto fail; - vsir_src_param_init_label(&instructions[ins_count].src[0], default_label); - ++ins_count; + } + vsir_src_param_init_label(&dst_ins->src[0], default_label); } - if (!reserve_instructions(&instructions, &ins_capacity, ins_count + 3 * case_count - 1)) + if (!shader_instruction_array_reserve(&instructions, instructions.count + 3 * case_count - 1)) goto fail; if_label = current_label; @@ -4295,13 +4617,15 @@ static enum vkd3d_result vsir_program_lower_switch_to_selection_ladder(struct vs { unsigned int fallthrough_label, case_label = label_from_src_param(&ins->src[3 + 2 * j + 1]); - if (!vsir_instruction_init_with_params(program, - &instructions[ins_count], &ins->location, VSIR_OP_IEQ, 1, 2)) + dst_ins = shader_instruction_array_append(&instructions); + if (!vsir_instruction_init_with_params(program, dst_ins, &ins->location, VSIR_OP_IEQ, 1, 2)) + { + vkd3d_shader_instruction_make_nop(dst_ins); goto fail; - dst_param_init_ssa_bool(&instructions[ins_count].dst[0], ssa_count); - instructions[ins_count].src[0] = ins->src[0]; - instructions[ins_count].src[1] = ins->src[3 + 2 * j]; - ++ins_count; + } + dst_param_init_ssa_bool(&dst_ins->dst[0], ssa_count); + dst_ins->src[0] = ins->src[0]; + dst_ins->src[1] = ins->src[3 + 2 * j]; /* For all cases except the last one we fall through to * the following case; the last one has to jump to the @@ -4311,13 +4635,15 @@ static enum vkd3d_result vsir_program_lower_switch_to_selection_ladder(struct vs else fallthrough_label = block_count + 1; - if (!vsir_instruction_init_with_params(program, &instructions[ins_count], - &ins->location, VSIR_OP_BRANCH, 0, 3)) + dst_ins = shader_instruction_array_append(&instructions); + if (!vsir_instruction_init_with_params(program, dst_ins, &ins->location, VSIR_OP_BRANCH, 0, 3)) + { + vkd3d_shader_instruction_make_nop(dst_ins); goto fail; - src_param_init_ssa_bool(&instructions[ins_count].src[0], ssa_count); - vsir_src_param_init_label(&instructions[ins_count].src[1], case_label); - vsir_src_param_init_label(&instructions[ins_count].src[2], fallthrough_label); - ++ins_count; + } + src_param_init_ssa_bool(&dst_ins->src[0], ssa_count); + vsir_src_param_init_label(&dst_ins->src[1], case_label); + vsir_src_param_init_label(&dst_ins->src[2], fallthrough_label); ++ssa_count; @@ -4333,29 +4659,28 @@ static enum vkd3d_result vsir_program_lower_switch_to_selection_ladder(struct vs } else { - if (!vsir_instruction_init_with_params(program, - &instructions[ins_count], &ins->location, VSIR_OP_LABEL, 0, 1)) + dst_ins = shader_instruction_array_append(&instructions); + if (!vsir_instruction_init_with_params(program, dst_ins, &ins->location, VSIR_OP_LABEL, 0, 1)) + { + vkd3d_shader_instruction_make_nop(dst_ins); goto fail; - vsir_src_param_init_label(&instructions[ins_count].src[0], ++block_count); - ++ins_count; + } + vsir_src_param_init_label(&dst_ins->src[0], ++block_count); if_label = block_count; } } } - vkd3d_free(program->instructions.elements); + vsir_program_replace_instructions(program, &instructions); vkd3d_free(block_map); - program->instructions.elements = instructions; - program->instructions.capacity = ins_capacity; - program->instructions.count = ins_count; program->block_count = block_count; program->ssa_count = ssa_count; return VKD3D_OK; fail: - vkd3d_free(instructions); + shader_instruction_array_destroy(&instructions); vkd3d_free(block_map); return VKD3D_ERROR_OUT_OF_MEMORY; @@ -4422,15 +4747,14 @@ static void ssas_to_temps_block_info_cleanup(struct ssas_to_temps_block_info *bl vkd3d_free(block_info); } -static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_program *program, - struct vsir_transformation_context *ctx) +static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps_in_function( + struct vsir_program *program, struct vsir_program_iterator *it) { - struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - size_t ins_capacity = 0, ins_count = 0, phi_count, incoming_count; struct ssas_to_temps_block_info *info, *block_info = NULL; - struct vkd3d_shader_instruction *instructions = NULL; + struct vsir_program_iterator it_begin = *it; struct ssas_to_temps_alloc alloc = {0}; struct vkd3d_shader_instruction *ins; + size_t phi_count, incoming_count; unsigned int current_label = 0; VKD3D_ASSERT(program->cf_type == VSIR_CF_BLOCKS); @@ -4446,14 +4770,31 @@ static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_ phi_count = 0; incoming_count = 0; - for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) + for (ins = vsir_program_iterator_current(it); ins; ins = vsir_program_iterator_next(it)) { unsigned int j, temp_idx; + bool finish = false; /* Only phi src/dst SSA values need be converted here. Structurisation may * introduce new cases of undominated SSA use, which will be handled later. */ - if (ins->opcode != VSIR_OP_PHI) - continue; + switch (ins->opcode) + { + case VSIR_OP_PHI: + break; + + case VSIR_OP_HS_CONTROL_POINT_PHASE: + case VSIR_OP_HS_FORK_PHASE: + case VSIR_OP_HS_JOIN_PHASE: + finish = true; + break; + + default: + continue; + } + + if (finish) + break; + ++phi_count; temp_idx = alloc.next_temp_idx++; @@ -4487,12 +4828,15 @@ static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_ if (!phi_count) goto done; - if (!reserve_instructions(&instructions, &ins_capacity, program->instructions.count + incoming_count - phi_count)) + if (!shader_instruction_array_reserve(&program->instructions, program->instructions.count + incoming_count)) goto fail; - for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) + *it = it_begin; + + for (ins = vsir_program_iterator_current(it); ins; ins = vsir_program_iterator_next(it)) { struct vkd3d_shader_instruction *mov_ins; + bool finish = false; size_t j; for (j = 0; j < ins->dst_count; ++j) @@ -4511,33 +4855,44 @@ static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_ case VSIR_OP_SWITCH_MONOLITHIC: info = &block_info[current_label - 1]; + mov_ins = vsir_program_iterator_insert_before_and_move(it, info->incoming_count); + VKD3D_ASSERT(mov_ins); + for (j = 0; j < info->incoming_count; ++j) { struct phi_incoming_to_temp *incoming = &info->incomings[j]; - mov_ins = &instructions[ins_count++]; if (!vsir_instruction_init_with_params(program, mov_ins, &ins->location, VSIR_OP_MOV, 1, 0)) + { + vkd3d_shader_instruction_make_nop(mov_ins); goto fail; + } *mov_ins->dst = *incoming->dst; mov_ins->src = incoming->src; mov_ins->src_count = 1; + + mov_ins = vsir_program_iterator_next(it); } break; case VSIR_OP_PHI: - continue; + vkd3d_shader_instruction_make_nop(ins); + break; + + case VSIR_OP_HS_CONTROL_POINT_PHASE: + case VSIR_OP_HS_FORK_PHASE: + case VSIR_OP_HS_JOIN_PHASE: + finish = true; + break; default: break; } - instructions[ins_count++] = *ins; + if (finish) + break; } - vkd3d_free(program->instructions.elements); - program->instructions.elements = instructions; - program->instructions.capacity = ins_capacity; - program->instructions.count = ins_count; program->temp_count = alloc.next_temp_idx; done: ssas_to_temps_block_info_cleanup(block_info, program->block_count); @@ -4546,13 +4901,51 @@ done: return VKD3D_OK; fail: - vkd3d_free(instructions); ssas_to_temps_block_info_cleanup(block_info, program->block_count); vkd3d_free(alloc.table); return VKD3D_ERROR_OUT_OF_MEMORY; } +static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_program *program, + struct vsir_transformation_context *ctx) +{ + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); + struct vkd3d_shader_instruction *ins; + enum vkd3d_result ret; + + VKD3D_ASSERT(program->cf_type == VSIR_CF_BLOCKS); + + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_current(&it)) + { + switch (ins->opcode) + { + case VSIR_OP_LABEL: + VKD3D_ASSERT(program->shader_version.type != VKD3D_SHADER_TYPE_HULL); + TRACE("Materializing undominated SSAs in a non-hull shader.\n"); + if ((ret = vsir_program_materialise_phi_ssas_to_temps_in_function(program, &it)) < 0) + return ret; + break; + + case VSIR_OP_HS_CONTROL_POINT_PHASE: + case VSIR_OP_HS_FORK_PHASE: + case VSIR_OP_HS_JOIN_PHASE: + VKD3D_ASSERT(program->shader_version.type == VKD3D_SHADER_TYPE_HULL); + TRACE("Materializing undominated SSAs in phase %u of a hull shader.\n", ins->opcode); + vsir_program_iterator_next(&it); + if ((ret = vsir_program_materialise_phi_ssas_to_temps_in_function(program, &it)) < 0) + return ret; + break; + + default: + vsir_program_iterator_next(&it); + break; + } + } + + return VKD3D_OK; +} + struct vsir_block_list { struct vsir_block **blocks; @@ -4607,7 +5000,7 @@ struct vsir_block * LABEL that introduces the block. `end' points to the terminator * instruction (either BRANCH or RET). They can coincide, meaning * that the block is empty. */ - struct vkd3d_shader_instruction *begin, *end; + struct vsir_program_iterator begin, end; struct vsir_block_list predecessors, successors; uint32_t *dominates; }; @@ -4790,8 +5183,8 @@ static void vsir_cfg_structure_cleanup(struct vsir_cfg_structure *structure) struct vsir_cfg_emit_target { - struct vkd3d_shader_instruction *instructions; - size_t ins_capacity, ins_count; + struct vkd3d_shader_instruction_array instructions; + unsigned int jump_target_temp_idx; unsigned int temp_count; }; @@ -4800,8 +5193,8 @@ struct vsir_cfg { struct vkd3d_shader_message_context *message_context; struct vsir_program *program; - size_t function_begin; - size_t function_end; + struct vsir_program_iterator function_begin; + struct vsir_program_iterator function_end; struct vsir_block *blocks; struct vsir_block *entry; size_t block_count; @@ -4926,19 +5319,21 @@ static void vsir_cfg_dump_dot(struct vsir_cfg *cfg) for (i = 0; i < cfg->block_count; ++i) { struct vsir_block *block = &cfg->blocks[i]; + struct vkd3d_shader_instruction *end; const char *shape; if (block->label == 0) continue; - switch (block->end->opcode) + end = vsir_program_iterator_current(&block->end); + switch (end->opcode) { case VSIR_OP_RET: shape = "trapezium"; break; case VSIR_OP_BRANCH: - shape = vsir_register_is_label(&block->end->src[0].reg) ? "ellipse" : "box"; + shape = vsir_register_is_label(&end->src[0].reg) ? "ellipse" : "box"; break; default: @@ -5046,19 +5441,20 @@ static void vsir_cfg_dump_structured_program(struct vsir_cfg *cfg) } static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program *program, - struct vkd3d_shader_message_context *message_context, struct vsir_cfg_emit_target *target, - size_t *pos) + struct vkd3d_shader_message_context *message_context, + struct vsir_cfg_emit_target *target, struct vsir_program_iterator *it) { struct vsir_block *current_block = NULL; + struct vkd3d_shader_instruction *ins; + size_t i, defined_block_count = 0; enum vkd3d_result ret; - size_t i; memset(cfg, 0, sizeof(*cfg)); cfg->message_context = message_context; cfg->program = program; cfg->block_count = program->block_count; cfg->target = target; - cfg->function_begin = *pos; + cfg->function_begin = *it; vsir_block_list_init(&cfg->order); @@ -5068,12 +5464,11 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program if (TRACE_ON()) vkd3d_string_buffer_init(&cfg->debug_buffer); - for (i = *pos; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_current(it); ins; ins = vsir_program_iterator_next(it)) { - struct vkd3d_shader_instruction *instruction = &program->instructions.elements[i]; bool finish = false; - switch (instruction->opcode) + switch (ins->opcode) { case VSIR_OP_PHI: case VSIR_OP_SWITCH_MONOLITHIC: @@ -5081,7 +5476,7 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program case VSIR_OP_LABEL: { - unsigned int label = label_from_src_param(&instruction->src[0]); + unsigned int label = label_from_src_param(&ins->src[0]); VKD3D_ASSERT(!current_block); VKD3D_ASSERT(label > 0); @@ -5090,16 +5485,18 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program VKD3D_ASSERT(current_block->label == 0); if ((ret = vsir_block_init(current_block, label, program->block_count)) < 0) goto fail; - current_block->begin = &program->instructions.elements[i + 1]; + current_block->begin = *it; + vsir_program_iterator_next(¤t_block->begin); if (!cfg->entry) cfg->entry = current_block; + ++defined_block_count; break; } case VSIR_OP_BRANCH: case VSIR_OP_RET: VKD3D_ASSERT(current_block); - current_block->end = instruction; + current_block->end = *it; current_block = NULL; break; @@ -5118,33 +5515,36 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program break; } - *pos = i; - cfg->function_end = *pos; + cfg->function_end = *it; + /* program->block_count is the max in any function. Set the count for this function. */ + cfg->block_count = defined_block_count; for (i = 0; i < cfg->block_count; ++i) { struct vsir_block *block = &cfg->blocks[i]; + struct vkd3d_shader_instruction *end; if (block->label == 0) continue; - switch (block->end->opcode) + end = vsir_program_iterator_current(&block->end); + switch (end->opcode) { case VSIR_OP_RET: break; case VSIR_OP_BRANCH: - if (vsir_register_is_label(&block->end->src[0].reg)) + if (vsir_register_is_label(&end->src[0].reg)) { - if ((ret = vsir_cfg_add_edge(cfg, block, &block->end->src[0])) < 0) + if ((ret = vsir_cfg_add_edge(cfg, block, &end->src[0])) < 0) goto fail; } else { - if ((ret = vsir_cfg_add_edge(cfg, block, &block->end->src[1])) < 0) + if ((ret = vsir_cfg_add_edge(cfg, block, &end->src[1])) < 0) goto fail; - if ((ret = vsir_cfg_add_edge(cfg, block, &block->end->src[2])) < 0) + if ((ret = vsir_cfg_add_edge(cfg, block, &end->src[2])) < 0) goto fail; } break; @@ -5328,8 +5728,10 @@ static enum vkd3d_result vsir_cfg_compute_loops(struct vsir_cfg *cfg) if (cfg->loops_by_header[header->label - 1] != SIZE_MAX) { + struct vkd3d_shader_instruction *begin = vsir_program_iterator_current(&header->begin); + FIXME("Block %u is header to more than one loop, this is not implemented.\n", header->label); - vkd3d_shader_error(cfg->message_context, &header->begin->location, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(cfg->message_context, &begin->location, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Block %u is header to more than one loop, this is not implemented.", header->label); return VKD3D_ERROR_NOT_IMPLEMENTED; } @@ -5411,8 +5813,10 @@ static enum vkd3d_result vsir_cfg_sort_nodes(struct vsir_cfg *cfg) if (in_degrees[i] == 0 && block != cfg->entry) { + struct vkd3d_shader_instruction *begin = vsir_program_iterator_current(&block->begin); + WARN("Unexpected entry point %u.\n", block->label); - vkd3d_shader_error(cfg->message_context, &block->begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, + vkd3d_shader_error(cfg->message_context, &begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "Block %u is unreachable from the entry point.", block->label); ret = VKD3D_ERROR_INVALID_SHADER; goto fail; @@ -5421,8 +5825,10 @@ static enum vkd3d_result vsir_cfg_sort_nodes(struct vsir_cfg *cfg) if (in_degrees[cfg->entry->label - 1] != 0) { + struct vkd3d_shader_instruction *begin = vsir_program_iterator_current(&cfg->entry->begin); + WARN("Entry point has %u incoming forward edges.\n", in_degrees[cfg->entry->label - 1]); - vkd3d_shader_error(cfg->message_context, &cfg->entry->begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, + vkd3d_shader_error(cfg->message_context, &begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "The entry point block has %u incoming forward edges.", in_degrees[cfg->entry->label - 1]); ret = VKD3D_ERROR_INVALID_SHADER; goto fail; @@ -5524,9 +5930,11 @@ static enum vkd3d_result vsir_cfg_sort_nodes(struct vsir_cfg *cfg) if (cfg->order.count != cfg->block_count) { + struct vkd3d_shader_instruction *begin = vsir_program_iterator_current(&cfg->entry->begin); + /* There is a cycle of forward edges. */ WARN("The control flow graph is not reducible.\n"); - vkd3d_shader_error(cfg->message_context, &cfg->entry->begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, + vkd3d_shader_error(cfg->message_context, &begin->location, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "The control flow graph is not reducible."); ret = VKD3D_ERROR_INVALID_SHADER; goto fail; @@ -5785,6 +6193,7 @@ static enum vkd3d_result vsir_cfg_build_structured_program(struct vsir_cfg *cfg) { struct vsir_block *block = cfg->order.blocks[i]; struct vsir_cfg_structure *structure; + struct vkd3d_shader_instruction *end; VKD3D_ASSERT(stack_depth > 0); @@ -5810,16 +6219,17 @@ static enum vkd3d_result vsir_cfg_build_structured_program(struct vsir_cfg *cfg) structure->u.block = block; /* Generate between zero and two jump instructions. */ - switch (block->end->opcode) + end = vsir_program_iterator_current(&block->end); + switch (end->opcode) { case VSIR_OP_BRANCH: { struct vsir_cfg_edge_action action_true, action_false; bool invert_condition = false; - if (vsir_register_is_label(&block->end->src[0].reg)) + if (vsir_register_is_label(&end->src[0].reg)) { - unsigned int target = label_from_src_param(&block->end->src[0]); + unsigned int target = label_from_src_param(&end->src[0]); struct vsir_block *successor = &cfg->blocks[target - 1]; vsir_cfg_compute_edge_action(cfg, block, successor, &action_true); @@ -5827,12 +6237,12 @@ static enum vkd3d_result vsir_cfg_build_structured_program(struct vsir_cfg *cfg) } else { - unsigned int target = label_from_src_param(&block->end->src[1]); + unsigned int target = label_from_src_param(&end->src[1]); struct vsir_block *successor = &cfg->blocks[target - 1]; vsir_cfg_compute_edge_action(cfg, block, successor, &action_true); - target = label_from_src_param(&block->end->src[2]); + target = label_from_src_param(&end->src[2]); successor = &cfg->blocks[target - 1]; vsir_cfg_compute_edge_action(cfg, block, successor, &action_false); @@ -5884,7 +6294,7 @@ static enum vkd3d_result vsir_cfg_build_structured_program(struct vsir_cfg *cfg) goto fail; structure->u.jump.type = action_true.jump_type; structure->u.jump.target = action_true.target; - structure->u.jump.condition = &block->end->src[0]; + structure->u.jump.condition = &vsir_program_iterator_current(&block->end)->src[0]; structure->u.jump.invert_condition = invert_condition; } @@ -6369,15 +6779,28 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_block(struct vsir_cfg *cfg struct vsir_block *block) { struct vsir_cfg_emit_target *target = cfg->target; + struct vkd3d_shader_instruction *ins, *end, *dst; + struct vsir_program_iterator it; + size_t ins_count = 0; - if (!reserve_instructions(&target->instructions, &target->ins_capacity, - target->ins_count + (block->end - block->begin))) - return VKD3D_ERROR_OUT_OF_MEMORY; + it = block->begin; + end = vsir_program_iterator_current(&block->end); - memcpy(&target->instructions[target->ins_count], block->begin, - (char *)block->end - (char *)block->begin); + for (ins = vsir_program_iterator_current(&it); ins != end; ins = vsir_program_iterator_next(&it)) + { + ++ins_count; + } + + if (!shader_instruction_array_reserve(&target->instructions, target->instructions.count + ins_count)) + return VKD3D_ERROR_OUT_OF_MEMORY; - target->ins_count += block->end - block->begin; + it = block->begin; + for (ins = vsir_program_iterator_current(&it); ins != end; ins = vsir_program_iterator_next(&it)) + { + dst = shader_instruction_array_append(&target->instructions); + VKD3D_ASSERT(dst); + *dst = *ins; + } return VKD3D_OK; } @@ -6387,20 +6810,22 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_loop(struct vsir_cfg *cfg, { struct vsir_cfg_emit_target *target = cfg->target; const struct vkd3d_shader_location no_loc = {0}; + struct vkd3d_shader_instruction *ins; enum vkd3d_result ret; - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 1)) + if (!(ins = shader_instruction_array_append(&target->instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; - vsir_instruction_init(&target->instructions[target->ins_count++], &no_loc, VSIR_OP_LOOP); + vsir_instruction_init(ins, &no_loc, VSIR_OP_LOOP); if ((ret = vsir_cfg_structure_list_emit(cfg, &loop->body, loop->idx)) < 0) return ret; - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 5)) + if (!shader_instruction_array_reserve(&target->instructions, target->instructions.count + 5)) return VKD3D_ERROR_OUT_OF_MEMORY; - vsir_instruction_init(&target->instructions[target->ins_count++], &no_loc, VSIR_OP_ENDLOOP); + ins = shader_instruction_array_append(&target->instructions); + vsir_instruction_init(ins, &no_loc, VSIR_OP_ENDLOOP); /* Add a trampoline to implement multilevel jumping depending on the stored * jump_target value. */ @@ -6414,44 +6839,50 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_loop(struct vsir_cfg *cfg, * we have to reach an outer loop, so we keep breaking. */ const unsigned int inner_break_target = loop->idx << 1; - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_IEQ, 1, 2)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_IEQ, 1, 2)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } - dst_param_init_temp_bool(&target->instructions[target->ins_count].dst[0], target->temp_count); - src_param_init_temp_uint(&target->instructions[target->ins_count].src[0], target->jump_target_temp_idx); - src_param_init_const_uint(&target->instructions[target->ins_count].src[1], outer_continue_target); + ++target->temp_count; - ++target->ins_count; + dst_param_init_temp_bool(&ins->dst[0], target->temp_count - 1); + src_param_init_temp_uint(&ins->src[0], target->jump_target_temp_idx); + src_param_init_const_uint(&ins->src[1], outer_continue_target); - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_CONTINUEP, 0, 1)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_CONTINUEP, 0, 1)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } - src_param_init_temp_bool(&target->instructions[target->ins_count].src[0], target->temp_count); - - ++target->ins_count; - ++target->temp_count; + src_param_init_temp_bool(&ins->src[0], target->temp_count - 1); - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_IEQ, 1, 2)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_IEQ, 1, 2)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } - dst_param_init_temp_bool(&target->instructions[target->ins_count].dst[0], target->temp_count); - src_param_init_temp_uint(&target->instructions[target->ins_count].src[0], target->jump_target_temp_idx); - src_param_init_const_uint(&target->instructions[target->ins_count].src[1], inner_break_target); + ++target->temp_count; - ++target->ins_count; + dst_param_init_temp_bool(&ins->dst[0], target->temp_count - 1); + src_param_init_temp_uint(&ins->src[0], target->jump_target_temp_idx); + src_param_init_const_uint(&ins->src[1], inner_break_target); - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_BREAKP, 0, 1)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_BREAKP, 0, 1)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; - target->instructions[target->ins_count].flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; - - src_param_init_temp_bool(&target->instructions[target->ins_count].src[0], target->temp_count); + } + ins->flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; - ++target->ins_count; - ++target->temp_count; + src_param_init_temp_bool(&ins->src[0], target->temp_count - 1); } return VKD3D_OK; @@ -6462,40 +6893,40 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_selection(struct vsir_cfg { struct vsir_cfg_emit_target *target = cfg->target; const struct vkd3d_shader_location no_loc = {0}; + struct vkd3d_shader_instruction *ins; enum vkd3d_result ret; - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 1)) + if (!(ins = shader_instruction_array_append(&target->instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_IF, 0, 1)) + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_IF, 0, 1)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } - target->instructions[target->ins_count].src[0] = *selection->condition; + ins->src[0] = *selection->condition; if (selection->invert_condition) - target->instructions[target->ins_count].flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; - - ++target->ins_count; + ins->flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; if ((ret = vsir_cfg_structure_list_emit(cfg, &selection->if_body, loop_idx)) < 0) return ret; if (selection->else_body.count != 0) { - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 1)) + if (!(ins = shader_instruction_array_append(&target->instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; - vsir_instruction_init(&target->instructions[target->ins_count++], &no_loc, VSIR_OP_ELSE); + vsir_instruction_init(ins, &no_loc, VSIR_OP_ELSE); if ((ret = vsir_cfg_structure_list_emit(cfg, &selection->else_body, loop_idx)) < 0) return ret; } - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 1)) + if (!(ins = shader_instruction_array_append(&target->instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; - - vsir_instruction_init(&target->instructions[target->ins_count++], &no_loc, VSIR_OP_ENDIF); + vsir_instruction_init(ins, &no_loc, VSIR_OP_ENDIF); return VKD3D_OK; } @@ -6508,6 +6939,7 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_jump(struct vsir_cfg *cfg, /* Encode the jump target as the loop index plus a bit to remember whether * we're breaking or continue-ing. */ unsigned int jump_target = jump->target << 1; + struct vkd3d_shader_instruction *ins; enum vkd3d_shader_opcode opcode; switch (jump->type) @@ -6538,32 +6970,34 @@ static enum vkd3d_result vsir_cfg_structure_list_emit_jump(struct vsir_cfg *cfg, vkd3d_unreachable(); } - if (!reserve_instructions(&target->instructions, &target->ins_capacity, target->ins_count + 2)) + if (!shader_instruction_array_reserve(&target->instructions, target->instructions.count + 2)) return VKD3D_ERROR_OUT_OF_MEMORY; if (jump->needs_launcher) { - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, VSIR_OP_MOV, 1, 1)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, VSIR_OP_MOV, 1, 1)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } - dst_param_init_temp_uint(&target->instructions[target->ins_count].dst[0], target->jump_target_temp_idx); - src_param_init_const_uint(&target->instructions[target->ins_count].src[0], jump_target); - - ++target->ins_count; + dst_param_init_temp_uint(&ins->dst[0], target->jump_target_temp_idx); + src_param_init_const_uint(&ins->src[0], jump_target); } - if (!vsir_instruction_init_with_params(cfg->program, &target->instructions[target->ins_count], - &no_loc, opcode, 0, !!jump->condition)) + ins = shader_instruction_array_append(&target->instructions); + if (!vsir_instruction_init_with_params(cfg->program, ins, &no_loc, opcode, 0, !!jump->condition)) + { + vkd3d_shader_instruction_make_nop(ins); return VKD3D_ERROR_OUT_OF_MEMORY; + } if (jump->invert_condition) - target->instructions[target->ins_count].flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; + ins->flags |= VKD3D_SHADER_CONDITIONAL_OP_Z; if (jump->condition) - target->instructions[target->ins_count].src[0] = *jump->condition; - - ++target->ins_count; + ins->src[0] = *jump->condition; return VKD3D_OK; } @@ -6616,13 +7050,13 @@ static enum vkd3d_result vsir_cfg_emit_structured_program(struct vsir_cfg *cfg) } static enum vkd3d_result vsir_program_structurize_function(struct vsir_program *program, - struct vkd3d_shader_message_context *message_context, struct vsir_cfg_emit_target *target, - size_t *pos) + struct vkd3d_shader_message_context *message_context, + struct vsir_cfg_emit_target *target, struct vsir_program_iterator *it) { enum vkd3d_result ret; struct vsir_cfg cfg; - if ((ret = vsir_cfg_init(&cfg, program, message_context, target, pos)) < 0) + if ((ret = vsir_cfg_init(&cfg, program, message_context, target, it)) < 0) return ret; vsir_cfg_compute_dominators(&cfg); @@ -6653,32 +7087,31 @@ out: static enum vkd3d_result vsir_program_structurize(struct vsir_program *program, struct vsir_transformation_context *ctx) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct vkd3d_shader_message_context *message_context = ctx->message_context; + struct vkd3d_shader_instruction *ins, *dst; struct vsir_cfg_emit_target target = {0}; enum vkd3d_result ret; - size_t i; VKD3D_ASSERT(program->cf_type == VSIR_CF_BLOCKS); target.jump_target_temp_idx = program->temp_count; target.temp_count = program->temp_count + 1; - if (!reserve_instructions(&target.instructions, &target.ins_capacity, program->instructions.count)) + if (!shader_instruction_array_init(&target.instructions, 0)) return VKD3D_ERROR_OUT_OF_MEMORY; - for (i = 0; i < program->instructions.count;) + for (ins = vsir_program_iterator_head(&it); ins;) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - switch (ins->opcode) { case VSIR_OP_LABEL: VKD3D_ASSERT(program->shader_version.type != VKD3D_SHADER_TYPE_HULL); TRACE("Structurizing a non-hull shader.\n"); - if ((ret = vsir_program_structurize_function(program, message_context, - &target, &i)) < 0) + if ((ret = vsir_program_structurize_function(program, message_context, &target, &it)) < 0) goto fail; - VKD3D_ASSERT(i == program->instructions.count); + ins = vsir_program_iterator_current(&it); + VKD3D_ASSERT(!ins); break; case VSIR_OP_HS_CONTROL_POINT_PHASE: @@ -6686,33 +7119,33 @@ static enum vkd3d_result vsir_program_structurize(struct vsir_program *program, case VSIR_OP_HS_JOIN_PHASE: VKD3D_ASSERT(program->shader_version.type == VKD3D_SHADER_TYPE_HULL); TRACE("Structurizing phase %u of a hull shader.\n", ins->opcode); - target.instructions[target.ins_count++] = *ins; - ++i; - if ((ret = vsir_program_structurize_function(program, message_context, - &target, &i)) < 0) + + dst = shader_instruction_array_append(&target.instructions); + *dst = *ins; + + vsir_program_iterator_next(&it); + if ((ret = vsir_program_structurize_function(program, message_context, &target, &it)) < 0) goto fail; + ins = vsir_program_iterator_current(&it); break; default: - if (!reserve_instructions(&target.instructions, &target.ins_capacity, target.ins_count + 1)) + if (!(dst = shader_instruction_array_append(&target.instructions))) return VKD3D_ERROR_OUT_OF_MEMORY; - target.instructions[target.ins_count++] = *ins; - ++i; + *dst = *ins; + ins = vsir_program_iterator_next(&it); break; } } - vkd3d_free(program->instructions.elements); - program->instructions.elements = target.instructions; - program->instructions.capacity = target.ins_capacity; - program->instructions.count = target.ins_count; + vsir_program_replace_instructions(program, &target.instructions); program->temp_count = target.temp_count; program->cf_type = VSIR_CF_STRUCTURED; return VKD3D_OK; fail: - vkd3d_free(target.instructions); + shader_instruction_array_destroy(&target.instructions); return ret; } @@ -6742,8 +7175,10 @@ static void register_map_undominated_use(struct vkd3d_shader_register *reg, stru static enum vkd3d_result vsir_cfg_materialize_undominated_ssas_to_temps(struct vsir_cfg *cfg) { struct vsir_program *program = cfg->program; + struct vkd3d_shader_instruction *ins, *end; struct ssas_to_temps_alloc alloc = {0}; struct vsir_block **origin_blocks; + struct vsir_program_iterator it; unsigned int j; size_t i; @@ -6761,12 +7196,13 @@ static enum vkd3d_result vsir_cfg_materialize_undominated_ssas_to_temps(struct v for (i = 0; i < cfg->block_count; ++i) { struct vsir_block *block = &cfg->blocks[i]; - struct vkd3d_shader_instruction *ins; if (block->label == 0) continue; - for (ins = block->begin; ins <= block->end; ++ins) + it = block->begin; + end = vsir_program_iterator_current(&block->end); + for (ins = vsir_program_iterator_current(&it); ins != end; ins = vsir_program_iterator_next(&it)) { for (j = 0; j < ins->dst_count; ++j) { @@ -6779,12 +7215,13 @@ static enum vkd3d_result vsir_cfg_materialize_undominated_ssas_to_temps(struct v for (i = 0; i < cfg->block_count; ++i) { struct vsir_block *block = &cfg->blocks[i]; - struct vkd3d_shader_instruction *ins; if (block->label == 0) continue; - for (ins = block->begin; ins <= block->end; ++ins) + it = block->begin; + end = vsir_program_iterator_current(&block->end); + for (ins = vsir_program_iterator_current(&it); ins != end; ins = vsir_program_iterator_next(&it)) { for (j = 0; j < ins->src_count; ++j) register_map_undominated_use(&ins->src[j].reg, &alloc, block, origin_blocks); @@ -6796,10 +7233,10 @@ static enum vkd3d_result vsir_cfg_materialize_undominated_ssas_to_temps(struct v TRACE("Emitting temps for %u values with undominated usage.\n", alloc.next_temp_idx - program->temp_count); - for (i = cfg->function_begin; i < cfg->function_end; ++i) + it = cfg->function_begin; + end = vsir_program_iterator_current(&cfg->function_end); + for (ins = vsir_program_iterator_current(&it); ins != end; ins = vsir_program_iterator_next(&it)) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - for (j = 0; j < ins->dst_count; ++j) materialize_ssas_to_temps_process_reg(program, &alloc, &ins->dst[j].reg); @@ -6815,14 +7252,13 @@ done: return VKD3D_OK; } -static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps_in_function( - struct vsir_program *program, struct vkd3d_shader_message_context *message_context, - size_t *pos) +static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps_in_function(struct vsir_program *program, + struct vkd3d_shader_message_context *message_context, struct vsir_program_iterator *it) { enum vkd3d_result ret; struct vsir_cfg cfg; - if ((ret = vsir_cfg_init(&cfg, program, message_context, NULL, pos)) < 0) + if ((ret = vsir_cfg_init(&cfg, program, message_context, NULL, it)) < 0) return ret; vsir_cfg_compute_dominators(&cfg); @@ -6837,25 +7273,23 @@ static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps_in_f static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(struct vsir_program *program, struct vsir_transformation_context *ctx) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct vkd3d_shader_message_context *message_context = ctx->message_context; + struct vkd3d_shader_instruction *ins; enum vkd3d_result ret; - size_t i; VKD3D_ASSERT(program->cf_type == VSIR_CF_BLOCKS); - for (i = 0; i < program->instructions.count;) + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_current(&it)) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - switch (ins->opcode) { case VSIR_OP_LABEL: VKD3D_ASSERT(program->shader_version.type != VKD3D_SHADER_TYPE_HULL); TRACE("Materializing undominated SSAs in a non-hull shader.\n"); if ((ret = vsir_program_materialize_undominated_ssas_to_temps_in_function( - program, message_context, &i)) < 0) + program, message_context, &it)) < 0) return ret; - VKD3D_ASSERT(i == program->instructions.count); break; case VSIR_OP_HS_CONTROL_POINT_PHASE: @@ -6863,14 +7297,14 @@ static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(stru case VSIR_OP_HS_JOIN_PHASE: VKD3D_ASSERT(program->shader_version.type == VKD3D_SHADER_TYPE_HULL); TRACE("Materializing undominated SSAs in phase %u of a hull shader.\n", ins->opcode); - ++i; + vsir_program_iterator_next(&it); if ((ret = vsir_program_materialize_undominated_ssas_to_temps_in_function( - program, message_context, &i)) < 0) + program, message_context, &it)) < 0) return ret; break; default: - ++i; + vsir_program_iterator_next(&it); break; } } @@ -6881,7 +7315,6 @@ static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(stru static bool use_flat_interpolation(const struct vsir_program *program, struct vkd3d_shader_message_context *message_context, bool *flat) { - static const struct vkd3d_shader_location no_loc; const struct vkd3d_shader_parameter1 *parameter; *flat = false; @@ -6891,13 +7324,13 @@ static bool use_flat_interpolation(const struct vsir_program *program, if (parameter->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported flat interpolation parameter type %#x.", parameter->type); return false; } if (parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid flat interpolation parameter data type %#x.", parameter->data_type); return false; } @@ -6938,7 +7371,6 @@ static enum vkd3d_result insert_alpha_test_before_ret(struct vsir_program *progr uint32_t colour_temp, struct vkd3d_shader_message_context *message_context) { struct vkd3d_shader_location loc = vsir_program_iterator_current(it)->location; - static const struct vkd3d_shader_location no_loc; struct vkd3d_shader_instruction *ins; static const struct @@ -6959,10 +7391,8 @@ static enum vkd3d_result insert_alpha_test_before_ret(struct vsir_program *progr if (compare_func == VKD3D_SHADER_COMPARISON_FUNC_NEVER) { - vsir_program_iterator_prev(it); - if (!vsir_program_iterator_insert_after(it, 1)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 1))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(it); vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_DISCARD, 0, 1); ins->flags = VKD3D_SHADER_CONDITIONAL_OP_Z; @@ -6972,10 +7402,8 @@ static enum vkd3d_result insert_alpha_test_before_ret(struct vsir_program *progr return VKD3D_OK; } - vsir_program_iterator_prev(it); - if (!vsir_program_iterator_insert_after(it, 3)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 3))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(it); switch (ref->data_type) { @@ -6994,7 +7422,7 @@ static enum vkd3d_result insert_alpha_test_before_ret(struct vsir_program *progr break; case VKD3D_SHADER_PARAMETER_DATA_TYPE_FLOAT32_VEC4: - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_PARAMETER, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_PARAMETER, "Alpha test reference data type must be a single component."); return VKD3D_ERROR_INVALID_ARGUMENT; @@ -7036,7 +7464,6 @@ static enum vkd3d_result vsir_program_insert_alpha_test(struct vsir_program *pro struct vkd3d_shader_message_context *message_context = ctx->message_context; const struct vkd3d_shader_parameter1 *func = NULL, *ref = NULL; uint32_t colour_signature_idx, colour_temp = ~0u; - static const struct vkd3d_shader_location no_loc; enum vkd3d_shader_comparison_func compare_func; struct vkd3d_shader_instruction *ins; int ret; @@ -7054,13 +7481,13 @@ static enum vkd3d_result vsir_program_insert_alpha_test(struct vsir_program *pro if (func->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported alpha test function parameter type %#x.", func->type); return VKD3D_ERROR_NOT_IMPLEMENTED; } if (func->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid alpha test function parameter data type %#x.", func->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7117,10 +7544,8 @@ static enum vkd3d_result insert_clip_planes_before_ret(struct vsir_program *prog struct vkd3d_shader_instruction *ins; unsigned int output_idx = 0; - vsir_program_iterator_prev(it); - if (!vsir_program_iterator_insert_after(it, vkd3d_popcount(mask) + 1)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, vkd3d_popcount(mask) + 1))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(it); for (unsigned int i = 0; i < 8; ++i) { @@ -7166,7 +7591,6 @@ static enum vkd3d_result vsir_program_insert_clip_planes(struct vsir_program *pr unsigned int low_signature_idx = ~0u, high_signature_idx = ~0u; const struct vkd3d_shader_parameter1 *mask_parameter = NULL; uint32_t position_signature_idx, position_temp, mask; - static const struct vkd3d_shader_location no_loc; struct signature_element *clip_element; struct vkd3d_shader_instruction *ins; unsigned int plane_count; @@ -7188,13 +7612,13 @@ static enum vkd3d_result vsir_program_insert_clip_planes(struct vsir_program *pr if (mask_parameter->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported clip plane mask parameter type %#x.", mask_parameter->type); return VKD3D_ERROR_NOT_IMPLEMENTED; } if (mask_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid clip plane mask parameter data type %#x.", mask_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7207,7 +7631,7 @@ static enum vkd3d_result vsir_program_insert_clip_planes(struct vsir_program *pr { if (signature->elements[i].sysval_semantic == VKD3D_SHADER_SV_CLIP_DISTANCE) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_PARAMETER, + vkd3d_shader_error(ctx->message_context, &ctx->null_location, VKD3D_SHADER_ERROR_VSIR_INVALID_PARAMETER, "Clip planes cannot be used if the shader writes clip distance."); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7215,7 +7639,7 @@ static enum vkd3d_result vsir_program_insert_clip_planes(struct vsir_program *pr if (!vsir_signature_find_sysval(signature, VKD3D_SHADER_SV_POSITION, 0, &position_signature_idx)) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_MISSING_SEMANTIC, + vkd3d_shader_error(ctx->message_context, &ctx->null_location, VKD3D_SHADER_ERROR_VSIR_MISSING_SEMANTIC, "Shader does not write position."); return VKD3D_ERROR_INVALID_SHADER; } @@ -7288,10 +7712,8 @@ static enum vkd3d_result insert_point_size_before_ret(struct vsir_program *progr const struct vkd3d_shader_location loc = vsir_program_iterator_current(it)->location; struct vkd3d_shader_instruction *ins; - vsir_program_iterator_prev(it); - if (!vsir_program_iterator_insert_after(it, 1)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 1))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(it); vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); vsir_dst_param_init(&ins->dst[0], VKD3DSPR_RASTOUT, VSIR_DATA_F32, 1); @@ -7307,7 +7729,6 @@ static enum vkd3d_result vsir_program_insert_point_size(struct vsir_program *pro { struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); const struct vkd3d_shader_parameter1 *size_parameter = NULL; - static const struct vkd3d_shader_location no_loc; struct vkd3d_shader_instruction *ins; if (program->has_point_size) @@ -7329,7 +7750,7 @@ static enum vkd3d_result vsir_program_insert_point_size(struct vsir_program *pro if (size_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_FLOAT32) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid point size parameter data type %#x.", size_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7356,7 +7777,6 @@ static enum vkd3d_result vsir_program_insert_point_size_clamp(struct vsir_progra { const struct vkd3d_shader_parameter1 *min_parameter = NULL, *max_parameter = NULL; struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); - static const struct vkd3d_shader_location no_loc; struct vkd3d_shader_instruction *ins; if (!program->has_point_size) @@ -7380,14 +7800,14 @@ static enum vkd3d_result vsir_program_insert_point_size_clamp(struct vsir_progra if (min_parameter && min_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_FLOAT32) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid minimum point size parameter data type %#x.", min_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } if (max_parameter && max_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_FLOAT32) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid maximum point size parameter data type %#x.", max_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7530,7 +7950,7 @@ static enum vkd3d_result vsir_program_insert_point_coord(struct vsir_program *pr { struct vsir_program_iterator it = vsir_program_iterator(&program->instructions), it2; const struct vkd3d_shader_parameter1 *sprite_parameter = NULL; - static const struct vkd3d_shader_location no_loc; + struct vkd3d_shader_location loc = ctx->null_location; struct vkd3d_shader_instruction *ins; bool used_texcoord = false; unsigned int coord_temp; @@ -7552,13 +7972,13 @@ static enum vkd3d_result vsir_program_insert_point_coord(struct vsir_program *pr if (sprite_parameter->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported point sprite parameter type %#x.", sprite_parameter->type); return VKD3D_ERROR_NOT_IMPLEMENTED; } if (sprite_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(ctx->message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(ctx->message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid point sprite parameter data type %#x.", sprite_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7578,7 +7998,10 @@ static enum vkd3d_result vsir_program_insert_point_coord(struct vsir_program *pr for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { if (!vsir_instruction_is_dcl(ins) && ins->opcode != VSIR_OP_LABEL && ins->opcode != VSIR_OP_NOP) + { + loc = ins->location; break; + } } it2 = it; @@ -7613,12 +8036,10 @@ static enum vkd3d_result vsir_program_insert_point_coord(struct vsir_program *pr if (used_texcoord) { - vsir_program_iterator_prev(&it); - if (!vsir_program_iterator_insert_after(&it, 2)) + if (!(ins = vsir_program_iterator_insert_before_and_move(&it, 2))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = vsir_program_iterator_next(&it); - vsir_instruction_init_with_params(program, ins, &no_loc, VSIR_OP_MOV, 1, 1); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); dst_param_init_temp_float4(&ins->dst[0], coord_temp); ins->dst[0].write_mask = VKD3DSP_WRITEMASK_0 | VKD3DSP_WRITEMASK_1; vsir_src_param_init(&ins->src[0], VKD3DSPR_POINT_COORD, VSIR_DATA_F32, 0); @@ -7626,11 +8047,12 @@ static enum vkd3d_result vsir_program_insert_point_coord(struct vsir_program *pr ins->src[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; ins = vsir_program_iterator_next(&it); - vsir_instruction_init_with_params(program, ins, &no_loc, VSIR_OP_MOV, 1, 1); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); dst_param_init_temp_float4(&ins->dst[0], coord_temp); ins->dst[0].write_mask = VKD3DSP_WRITEMASK_2 | VKD3DSP_WRITEMASK_3; vsir_src_param_init(&ins->src[0], VKD3DSPR_IMMCONST, VSIR_DATA_F32, 0); ins->src[0].reg.dimension = VSIR_DIMENSION_VEC4; + ins->src[0].swizzle = VKD3D_SHADER_NO_SWIZZLE; vsir_program_iterator_next(&it); program->has_point_coord = true; @@ -7668,15 +8090,12 @@ static enum vkd3d_result vsir_program_add_fog_input(struct vsir_program *program } static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *program, - const struct vkd3d_shader_instruction *ret, enum vkd3d_shader_fog_fragment_mode mode, - uint32_t fog_signature_idx, uint32_t colour_signature_idx, uint32_t colour_temp, - size_t *ret_pos, struct vkd3d_shader_message_context *message_context) + struct vsir_program_iterator *it, enum vkd3d_shader_fog_fragment_mode mode, uint32_t fog_signature_idx, + uint32_t colour_signature_idx, uint32_t colour_temp, struct vkd3d_shader_message_context *message_context) { - struct vkd3d_shader_instruction_array *instructions = &program->instructions; - struct vkd3d_shader_location loc = ret->location; + struct vkd3d_shader_instruction *ins = vsir_program_iterator_current(it); + struct vkd3d_shader_location loc = ins->location; uint32_t ssa_factor = program->ssa_count++; - size_t pos = ret - instructions->elements; - struct vkd3d_shader_instruction *ins; uint32_t ssa_temp, ssa_temp2; switch (mode) @@ -7687,16 +8106,11 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro * add sr0, FOG_END, -vFOG.x * mul_sat srFACTOR, sr0, FOG_SCALE */ - if (!shader_instruction_array_insert_at(&program->instructions, pos, 4)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 4))) return VKD3D_ERROR_OUT_OF_MEMORY; - ret = NULL; - - *ret_pos = pos + 4; ssa_temp = program->ssa_count++; - ins = &program->instructions.elements[pos]; - vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_ADD, 1, 2); dst_param_init_ssa_float(&ins->dst[0], ssa_temp); src_param_init_parameter(&ins->src[0], VKD3D_SHADER_PARAMETER_NAME_FOG_END, VSIR_DATA_F32); @@ -7705,12 +8119,15 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro ins->src[1].reg.dimension = VSIR_DIMENSION_VEC4; ins->src[1].swizzle = VKD3D_SHADER_SWIZZLE(X, X, X, X); ins->src[1].modifiers = VKD3DSPSM_NEG; + ins = vsir_program_iterator_next(it); - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_MUL, 1, 2); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MUL, 1, 2); dst_param_init_ssa_float(&ins->dst[0], ssa_factor); ins->dst[0].modifiers = VKD3DSPDM_SATURATE; src_param_init_ssa_float(&ins->src[0], ssa_temp); src_param_init_parameter(&ins->src[1], VKD3D_SHADER_PARAMETER_NAME_FOG_SCALE, VSIR_DATA_F32); + ins = vsir_program_iterator_next(it); + break; case VKD3D_SHADER_FOG_FRAGMENT_EXP: @@ -7719,16 +8136,11 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro * mul sr0, FOG_SCALE, vFOG.x * exp_sat srFACTOR, -sr0 */ - if (!shader_instruction_array_insert_at(&program->instructions, pos, 4)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 4))) return VKD3D_ERROR_OUT_OF_MEMORY; - ret = NULL; - - *ret_pos = pos + 4; ssa_temp = program->ssa_count++; - ins = &program->instructions.elements[pos]; - vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MUL, 1, 2); dst_param_init_ssa_float(&ins->dst[0], ssa_temp); src_param_init_parameter(&ins->src[0], VKD3D_SHADER_PARAMETER_NAME_FOG_SCALE, VSIR_DATA_F32); @@ -7736,12 +8148,14 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro ins->src[1].reg.idx[0].offset = fog_signature_idx; ins->src[1].reg.dimension = VSIR_DIMENSION_VEC4; ins->src[1].swizzle = VKD3D_SHADER_SWIZZLE(X, X, X, X); + ins = vsir_program_iterator_next(it); - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_EXP, 1, 1); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_EXP, 1, 1); dst_param_init_ssa_float(&ins->dst[0], ssa_factor); ins->dst[0].modifiers = VKD3DSPDM_SATURATE; src_param_init_ssa_float(&ins->src[0], ssa_temp); ins->src[0].modifiers = VKD3DSPSM_NEG; + ins = vsir_program_iterator_next(it); break; case VKD3D_SHADER_FOG_FRAGMENT_EXP2: @@ -7751,17 +8165,12 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro * mul sr1, sr0, sr0 * exp_sat srFACTOR, -sr1 */ - if (!shader_instruction_array_insert_at(&program->instructions, pos, 5)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 5))) return VKD3D_ERROR_OUT_OF_MEMORY; - ret = NULL; - - *ret_pos = pos + 5; ssa_temp = program->ssa_count++; ssa_temp2 = program->ssa_count++; - ins = &program->instructions.elements[pos]; - vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MUL, 1, 2); dst_param_init_ssa_float(&ins->dst[0], ssa_temp); src_param_init_parameter(&ins->src[0], VKD3D_SHADER_PARAMETER_NAME_FOG_SCALE, VSIR_DATA_F32); @@ -7769,17 +8178,20 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro ins->src[1].reg.idx[0].offset = fog_signature_idx; ins->src[1].reg.dimension = VSIR_DIMENSION_VEC4; ins->src[1].swizzle = VKD3D_SHADER_SWIZZLE(X, X, X, X); + ins = vsir_program_iterator_next(it); - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_MUL, 1, 2); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MUL, 1, 2); dst_param_init_ssa_float(&ins->dst[0], ssa_temp2); src_param_init_ssa_float(&ins->src[0], ssa_temp); src_param_init_ssa_float(&ins->src[1], ssa_temp); + ins = vsir_program_iterator_next(it); - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_EXP, 1, 1); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_EXP, 1, 1); dst_param_init_ssa_float(&ins->dst[0], ssa_factor); ins->dst[0].modifiers = VKD3DSPDM_SATURATE; src_param_init_ssa_float(&ins->src[0], ssa_temp2); ins->src[0].modifiers = VKD3DSPSM_NEG; + ins = vsir_program_iterator_next(it); break; default: @@ -7792,18 +8204,20 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro * mad oC0, sr0, srFACTOR, FOG_COLOUR */ - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_ADD, 1, 2); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_ADD, 1, 2); dst_param_init_ssa_float4(&ins->dst[0], program->ssa_count++); src_param_init_temp_float4(&ins->src[0], colour_temp); src_param_init_parameter_vec4(&ins->src[1], VKD3D_SHADER_PARAMETER_NAME_FOG_COLOUR, VSIR_DATA_F32); ins->src[1].modifiers = VKD3DSPSM_NEG; + ins = vsir_program_iterator_next(it); - vsir_instruction_init_with_params(program, ++ins, &loc, VSIR_OP_MAD, 1, 3); + vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MAD, 1, 3); dst_param_init_output(&ins->dst[0], VSIR_DATA_F32, colour_signature_idx, program->output_signature.elements[colour_signature_idx].mask); src_param_init_ssa_float4(&ins->src[0], program->ssa_count - 1); src_param_init_ssa_float(&ins->src[1], ssa_factor); src_param_init_parameter_vec4(&ins->src[2], VKD3D_SHADER_PARAMETER_NAME_FOG_COLOUR, VSIR_DATA_F32); + ins = vsir_program_iterator_next(it); return VKD3D_OK; } @@ -7811,14 +8225,13 @@ static enum vkd3d_result insert_fragment_fog_before_ret(struct vsir_program *pro static enum vkd3d_result vsir_program_insert_fragment_fog(struct vsir_program *program, struct vsir_transformation_context *ctx) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct vkd3d_shader_message_context *message_context = ctx->message_context; uint32_t colour_signature_idx, fog_signature_idx, colour_temp; const struct vkd3d_shader_parameter1 *mode_parameter = NULL; - static const struct vkd3d_shader_location no_loc; const struct signature_element *fog_element; enum vkd3d_shader_fog_fragment_mode mode; struct vkd3d_shader_instruction *ins; - size_t new_pos; int ret; if (program->shader_version.type != VKD3D_SHADER_TYPE_PIXEL) @@ -7832,13 +8245,13 @@ static enum vkd3d_result vsir_program_insert_fragment_fog(struct vsir_program *p if (mode_parameter->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported fog fragment mode parameter type %#x.", mode_parameter->type); return VKD3D_ERROR_NOT_IMPLEMENTED; } if (mode_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid fog fragment mode parameter data type %#x.", mode_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -7859,19 +8272,16 @@ static enum vkd3d_result vsir_program_insert_fragment_fog(struct vsir_program *p * through the whole shader and convert it to a temp. */ colour_temp = program->temp_count++; - for (size_t i = 0; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { - ins = &program->instructions.elements[i]; - if (vsir_instruction_is_dcl(ins)) continue; if (ins->opcode == VSIR_OP_RET) { - if ((ret = insert_fragment_fog_before_ret(program, ins, mode, fog_signature_idx, - colour_signature_idx, colour_temp, &new_pos, message_context)) < 0) + if ((ret = insert_fragment_fog_before_ret(program, &it, mode, fog_signature_idx, + colour_signature_idx, colour_temp, message_context)) < 0) return ret; - i = new_pos; continue; } @@ -7927,21 +8337,16 @@ static enum vkd3d_result vsir_program_add_fog_output(struct vsir_program *progra return VKD3D_OK; } -static enum vkd3d_result insert_vertex_fog_before_ret(struct vsir_program *program, - const struct vkd3d_shader_instruction *ret, enum vkd3d_shader_fog_source source, uint32_t temp, - uint32_t fog_signature_idx, uint32_t source_signature_idx, size_t *ret_pos) +static enum vkd3d_result insert_vertex_fog_before_ret(struct vsir_program *program, struct vsir_program_iterator *it, + enum vkd3d_shader_fog_source source, uint32_t temp, uint32_t fog_signature_idx, uint32_t source_signature_idx) { const struct signature_element *e = &program->output_signature.elements[source_signature_idx]; - struct vkd3d_shader_instruction_array *instructions = &program->instructions; + struct vkd3d_shader_instruction *ret = vsir_program_iterator_current(it); const struct vkd3d_shader_location loc = ret->location; - size_t pos = ret - instructions->elements; struct vkd3d_shader_instruction *ins; - if (!shader_instruction_array_insert_at(&program->instructions, pos, 2)) + if (!(ins = vsir_program_iterator_insert_before_and_move(it, 2))) return VKD3D_ERROR_OUT_OF_MEMORY; - ret = NULL; - - ins = &program->instructions.elements[pos]; /* Write the fog output. */ vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); @@ -7951,26 +8356,26 @@ static enum vkd3d_result insert_vertex_fog_before_ret(struct vsir_program *progr ins->src[0].swizzle = VKD3D_SHADER_SWIZZLE(Z, Z, Z, Z); else /* Position or specular W. */ ins->src[0].swizzle = VKD3D_SHADER_SWIZZLE(W, W, W, W); - ++ins; + ins = vsir_program_iterator_next(it); /* Write the position or specular output. */ vsir_instruction_init_with_params(program, ins, &loc, VSIR_OP_MOV, 1, 1); dst_param_init_output(&ins->dst[0], vsir_data_type_from_component_type(e->component_type), source_signature_idx, e->mask); src_param_init_temp_float4(&ins->src[0], temp); - ++ins; + ins = vsir_program_iterator_next(it); - *ret_pos = pos + 2; return VKD3D_OK; } static enum vkd3d_result vsir_program_insert_vertex_fog(struct vsir_program *program, struct vsir_transformation_context *ctx) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct vkd3d_shader_message_context *message_context = ctx->message_context; const struct vkd3d_shader_parameter1 *source_parameter = NULL; uint32_t fog_signature_idx, source_signature_idx, temp; - static const struct vkd3d_shader_location no_loc; + struct vkd3d_shader_instruction *ins; enum vkd3d_shader_fog_source source; const struct signature_element *e; @@ -7982,13 +8387,13 @@ static enum vkd3d_result vsir_program_insert_vertex_fog(struct vsir_program *pro if (source_parameter->type != VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED, "Unsupported fog source parameter type %#x.", source_parameter->type); return VKD3D_ERROR_NOT_IMPLEMENTED; } if (source_parameter->data_type != VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32) { - vkd3d_shader_error(message_context, &no_loc, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, + vkd3d_shader_error(message_context, NULL, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, "Invalid fog source parameter data type %#x.", source_parameter->data_type); return VKD3D_ERROR_INVALID_ARGUMENT; } @@ -8010,7 +8415,7 @@ static enum vkd3d_result vsir_program_insert_vertex_fog(struct vsir_program *pro if (!vsir_signature_find_sysval(&program->output_signature, VKD3D_SHADER_SV_POSITION, 0, &source_signature_idx)) { - vkd3d_shader_error(ctx->message_context, &no_loc, + vkd3d_shader_error(ctx->message_context, &ctx->null_location, VKD3D_SHADER_ERROR_VSIR_MISSING_SEMANTIC, "Shader does not write position."); return VKD3D_ERROR_INVALID_SHADER; } @@ -8027,22 +8432,18 @@ static enum vkd3d_result vsir_program_insert_vertex_fog(struct vsir_program *pro /* Insert a fog write before each ret, and convert either specular or * position output to a temp. */ - for (size_t i = 0; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - if (vsir_instruction_is_dcl(ins)) continue; if (ins->opcode == VSIR_OP_RET) { - size_t new_pos; int ret; - if ((ret = insert_vertex_fog_before_ret(program, ins, source, temp, - fog_signature_idx, source_signature_idx, &new_pos)) < 0) + if ((ret = insert_vertex_fog_before_ret(program, &it, source, temp, + fog_signature_idx, source_signature_idx)) < 0) return ret; - i = new_pos; continue; } @@ -8450,7 +8851,7 @@ struct liveness_tracker bool fixed_mask; uint8_t mask; unsigned int first_write, last_access; - } *ssa_regs; + } *ssa_regs, *temp_regs; }; static void liveness_track_src(struct liveness_tracker *tracker, @@ -8464,6 +8865,8 @@ static void liveness_track_src(struct liveness_tracker *tracker, if (src->reg.type == VKD3DSPR_SSA) tracker->ssa_regs[src->reg.idx[0].offset].last_access = index; + else if (src->reg.type == VKD3DSPR_TEMP) + tracker->temp_regs[src->reg.idx[0].offset].last_access = index; } static void liveness_track_dst(struct liveness_tracker *tracker, struct vkd3d_shader_dst_param *dst, @@ -8479,6 +8882,8 @@ static void liveness_track_dst(struct liveness_tracker *tracker, struct vkd3d_sh if (dst->reg.type == VKD3DSPR_SSA) reg = &tracker->ssa_regs[dst->reg.idx[0].offset]; + else if (dst->reg.type == VKD3DSPR_TEMP) + reg = &tracker->temp_regs[dst->reg.idx[0].offset]; else return; @@ -8565,20 +8970,22 @@ static void liveness_tracker_cleanup(struct liveness_tracker *tracker) static enum vkd3d_result track_liveness(struct vsir_program *program, struct liveness_tracker *tracker) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); + struct vkd3d_shader_instruction *ins; struct liveness_tracker_reg *regs; unsigned int loop_depth = 0; unsigned int loop_start = 0; + unsigned int i; memset(tracker, 0, sizeof(*tracker)); - if (!(regs = vkd3d_calloc(program->ssa_count, sizeof(*regs)))) + if (!(regs = vkd3d_calloc(program->ssa_count + program->temp_count, sizeof(*regs)))) return VKD3D_ERROR_OUT_OF_MEMORY; tracker->ssa_regs = regs; + tracker->temp_regs = ®s[program->ssa_count]; - for (unsigned int i = 0; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_head(&it), i = 0; ins; ins = vsir_program_iterator_next(&it), ++i) { - const struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - if (ins->opcode == VSIR_OP_LOOP || ins->opcode == VSIR_OP_REP) { if (!loop_depth++) @@ -8604,8 +9011,7 @@ static enum vkd3d_result track_liveness(struct vsir_program *program, struct liv * should be illegal for an SSA value to be read in a block * containing L.) * We don't try to perform this optimization yet, in the name of - * maximal simplicity, and also because this code is intended to - * be extended to non-SSA values. */ + * maximal simplicity. */ for (unsigned int j = 0; j < program->ssa_count; ++j) { struct liveness_tracker_reg *reg = &tracker->ssa_regs[j]; @@ -8615,6 +9021,16 @@ static enum vkd3d_result track_liveness(struct vsir_program *program, struct liv if (reg->last_access < i) reg->last_access = i; } + + for (unsigned int j = 0; j < program->temp_count; ++j) + { + struct liveness_tracker_reg *reg = &tracker->temp_regs[j]; + + if (reg->first_write > loop_start) + reg->first_write = loop_start; + if (reg->last_access < i) + reg->last_access = i; + } } } @@ -8632,74 +9048,19 @@ struct temp_allocator struct vkd3d_shader_message_context *message_context; struct temp_allocator_reg { + struct liveness_tracker_reg *liveness_reg; uint8_t allocated_mask; uint32_t temp_id; - } *ssa_regs; - size_t allocated_ssa_count; - enum vkd3d_result result; -}; - -static uint8_t get_available_writemask(const struct temp_allocator *allocator, - struct liveness_tracker *tracker, unsigned int first_write, unsigned int last_access, uint32_t temp_id) -{ - uint8_t writemask = VKD3DSP_WRITEMASK_ALL; - - for (size_t i = 0; i < allocator->allocated_ssa_count; ++i) - { - const struct temp_allocator_reg *reg = &allocator->ssa_regs[i]; - const struct liveness_tracker_reg *liveness_reg = &tracker->ssa_regs[i]; - - /* We do not overlap if first write == last read: - * this is the case where we are allocating the result of that - * expression, e.g. "add r0, r0, r1". */ - - if (reg->temp_id == temp_id - && first_write < liveness_reg->last_access - && last_access > liveness_reg->first_write) - writemask &= ~reg->allocated_mask; - - if (!writemask) - return writemask; - } - - return writemask; -} - -static bool temp_allocator_allocate(struct temp_allocator *allocator, struct liveness_tracker *tracker, - struct temp_allocator_reg *reg, const struct liveness_tracker_reg *liveness_reg, uint32_t base_id) -{ - if (!liveness_reg->written) - return false; - - for (uint32_t id = base_id;; ++id) - { - uint8_t available_mask = get_available_writemask(allocator, tracker, - liveness_reg->first_write, liveness_reg->last_access, id); - - if (liveness_reg->fixed_mask) - { - if ((available_mask & liveness_reg->mask) == liveness_reg->mask) - { - reg->temp_id = id; - reg->allocated_mask = liveness_reg->mask; - return true; - } - } - else - { - /* For SSA values the mask is always zero-based and contiguous. - * We don't correctly handle cases where it's not, currently. */ - VKD3D_ASSERT((liveness_reg->mask | (liveness_reg->mask - 1)) == liveness_reg->mask); - - if (vkd3d_popcount(available_mask) >= vkd3d_popcount(liveness_reg->mask)) - { - reg->temp_id = id; - reg->allocated_mask = vsir_combine_write_masks(available_mask, liveness_reg->mask); - return true; - } - } - } -} + enum vkd3d_shader_register_type type; + unsigned int idx; + bool force_first; + } *ssa_regs, *temp_regs; + size_t ssa_count, temp_count; + unsigned int new_temp_count; + enum vkd3d_result result; + uint8_t *current_allocation; + bool ps_1_x; +}; static void temp_allocator_set_src(struct temp_allocator *allocator, struct vkd3d_shader_src_param *src) { @@ -8713,6 +9074,8 @@ static void temp_allocator_set_src(struct temp_allocator *allocator, struct vkd3 if (src->reg.type == VKD3DSPR_SSA) reg = &allocator->ssa_regs[src->reg.idx[0].offset]; + else if (src->reg.type == VKD3DSPR_TEMP) + reg = &allocator->temp_regs[src->reg.idx[0].offset]; else return; @@ -8792,6 +9155,7 @@ static void temp_allocator_set_dst(struct temp_allocator *allocator, struct vkd3d_shader_dst_param *dst, const struct vkd3d_shader_instruction *ins) { struct temp_allocator_reg *reg; + uint32_t remapped_mask; for (unsigned int k = 0; k < dst->reg.idx_count; ++k) { @@ -8801,15 +9165,18 @@ static void temp_allocator_set_dst(struct temp_allocator *allocator, if (dst->reg.type == VKD3DSPR_SSA) reg = &allocator->ssa_regs[dst->reg.idx[0].offset]; + else if (dst->reg.type == VKD3DSPR_TEMP) + reg = &allocator->temp_regs[dst->reg.idx[0].offset]; else return; dst->reg.type = VKD3DSPR_TEMP; dst->reg.dimension = VSIR_DIMENSION_VEC4; dst->reg.idx[0].offset = reg->temp_id; - if (reg->allocated_mask != dst->write_mask) + remapped_mask = vsir_combine_write_masks(reg->allocated_mask, dst->write_mask); + if (dst->write_mask != remapped_mask) { - dst->write_mask = reg->allocated_mask; + dst->write_mask = remapped_mask; if (vsir_opcode_is_double(ins->opcode)) { @@ -8825,58 +9192,324 @@ static void temp_allocator_set_dst(struct temp_allocator *allocator, if (vsir_src_is_masked(ins->opcode, i)) { if (src->reg.type == VKD3DSPR_IMMCONST) - vsir_remap_immconst(src, dst->write_mask); + vsir_remap_immconst(src, reg->allocated_mask); else if (src->reg.type == VKD3DSPR_IMMCONST64) - vsir_remap_immconst64(src, dst->write_mask); + vsir_remap_immconst64(src, reg->allocated_mask); else - src->swizzle = vsir_map_swizzle(src->swizzle, dst->write_mask); + src->swizzle = vsir_map_swizzle(src->swizzle, reg->allocated_mask); + } + } + } +} + +static int temp_allocate_compare_open(const void *ptr1, const void *ptr2) +{ + const struct temp_allocator_reg * const *reg1 = ptr1, * const *reg2 = ptr2; + int ret; + + if ((ret = vkd3d_u32_compare((*reg1)->force_first, (*reg2)->force_first))) + return -ret; + if ((ret = vkd3d_u32_compare((*reg1)->liveness_reg->first_write, (*reg2)->liveness_reg->first_write))) + return ret; + if ((ret = vkd3d_u32_compare((*reg1)->liveness_reg->last_access, (*reg2)->liveness_reg->last_access))) + return ret; + return 0; +} + +static int temp_allocate_compare_close(const void *ptr1, const void *ptr2) +{ + const struct temp_allocator_reg * const *reg1 = ptr1, * const *reg2 = ptr2; + int ret; + + if ((ret = vkd3d_u32_compare((*reg1)->liveness_reg->last_access, (*reg2)->liveness_reg->last_access))) + return ret; + return vkd3d_u32_compare((*reg1)->liveness_reg->first_write, (*reg2)->liveness_reg->first_write); +} + +static const char *debug_temp_allocator_reg(const struct temp_allocator_reg *reg) +{ + return vkd3d_dbg_sprintf("%s%u", reg->type == VKD3DSPR_SSA ? "sr" : "r", reg->idx); +} + +static void temp_allocator_open_register(struct temp_allocator *allocator, struct temp_allocator_reg *reg) +{ + const size_t reg_count = allocator->ssa_count + allocator->temp_count; + const struct liveness_tracker_reg *liveness_reg = reg->liveness_reg; + uint8_t *current_allocation = allocator->current_allocation; + size_t i; + + if (!liveness_reg->written) + return; + + for (i = 0; i < reg_count; ++i) + { + const uint8_t available_mask = ~current_allocation[i] & 0xf; + + if (liveness_reg->fixed_mask) + { + if ((available_mask & liveness_reg->mask) == liveness_reg->mask) + { + reg->temp_id = i; + reg->allocated_mask = liveness_reg->mask; + current_allocation[i] |= reg->allocated_mask; + allocator->new_temp_count = max(allocator->new_temp_count, i + 1); + TRACE("Allocated r%u%s for %s (liveness %u-%u).\n", + reg->temp_id, debug_vsir_writemask(reg->allocated_mask), + debug_temp_allocator_reg(reg), liveness_reg->first_write, liveness_reg->last_access); + break; + } + } + else + { + /* For SSA values the mask is always zero-based and contiguous. + * For TEMP values we assume the register was allocated that way, + * but it may only be partially used. + * We currently only handle cases where the mask is zero-based and + * contiguous, so we need to fill in the missing components to + * ensure this. */ + uint8_t mask = (1u << (vkd3d_log2i(liveness_reg->mask) + 1)) - 1; + + if (vkd3d_popcount(available_mask) >= vkd3d_popcount(mask)) + { + if (mask != liveness_reg->mask) + WARN("Allocating a mask %#x with used components %#x; this is not optimized.\n", + mask, liveness_reg->mask); + + reg->temp_id = i; + reg->allocated_mask = vsir_combine_write_masks(available_mask, mask); + current_allocation[i] |= reg->allocated_mask; + allocator->new_temp_count = max(allocator->new_temp_count, i + 1); + TRACE("Allocated r%u%s for %s (liveness %u-%u).\n", + reg->temp_id, debug_vsir_writemask(reg->allocated_mask), + debug_temp_allocator_reg(reg), liveness_reg->first_write, liveness_reg->last_access); + break; + } + } + } + + VKD3D_ASSERT(i < reg_count); +} + +static void temp_allocator_close_register(struct temp_allocator *allocator, struct temp_allocator_reg *reg) +{ + const struct liveness_tracker_reg *liveness_reg = reg->liveness_reg; + + if (!liveness_reg->written) + return; + + TRACE("Register %s (liveness %u-%u) reaches end of life.\n", + debug_temp_allocator_reg(reg), liveness_reg->first_write, liveness_reg->last_access); + + allocator->current_allocation[reg->temp_id] &= ~reg->allocated_mask; +} + +/* Compute the allocation map. Each register is modeled as a time interval + * spanning from `first_write' to `last_access'. We simulate scanning through + * all the intervals in time order, keeping the set of currently allocated + * registers as a bit map: each time we open an interval (i.e., hit its + * `first_write' time) we allocate it to the first available register scanning + * the current state; each time we close an interval (i.e., hit its + * `last_access' time) we unset the corresponding bits. + * + * In general at any given time we first process all intervals to be closed and + * then all intervals to be opened at that time. This models the fact that an + * instruction can write to a register which it also reads from, and the write + * won't interfere with the read. In other words, first all reads are + * performed, then the instruction is executed, then the writes are performed. + * + * There is a corner case exception, though: the case of degenerate intervals + * that are opened and closed at the same time. This corresponds to registers + * that are written and then never read, which in principle shouldn't exist + * because they make no sense. However it's better to be robust, and we support + * them anyway. + * + * So that's what we do: + * - First all non-degenerate closes are processed. + * - Then all degenerate opens are processed, because we cannot close them + * before having opened them. + * - Then all non-degenerate opens are processed: this has to happens before + * the degenerate intervals are closed, because they need to be allocated to + * different registers. + * - Then all degenerate closes are processed. + * + * This is effected with a few different strategies: + * - In the open order, registers are primarily sorted by `first_write' and + * secondarily by `last_access'. This way degenerate registers are always + * opened before non-degenerate ones with the same `first_write' time. + * - In the close order, registers are primarily sorted by `last_access' and + * secondarily by `first_write'. This way non-degenerate registers are + * always closed before degenerate ones with the same `last_access' time. + * - There is a scheduling algorithm that decides at each iteration whether to + * open or close a register. See details below. + * + * TODO: the algorithm could be further optimized by keeping a few pointers to + * the first position in `current_allocation' that has at least one (or two, + * three and four) available components, so we don't always have to scan from + * the beginning. + */ +static enum vkd3d_result temp_allocator_compute_allocation_map(struct temp_allocator *allocator, + const struct liveness_tracker *liveness) +{ + const size_t reg_count = allocator->ssa_count + allocator->temp_count; + struct temp_allocator_reg **open_order = NULL, **close_order = NULL; + size_t i, pos_open = 0, pos_close = 0; + + /* In the worst-case scenario each of the `reg_count' registers to be + * processed requires its own allocation. We should never exceed that + * amount. */ + if (!(allocator->current_allocation = vkd3d_calloc(reg_count, sizeof(*allocator->current_allocation))) + || !(open_order = vkd3d_calloc(reg_count, sizeof(*open_order))) + || !(close_order = vkd3d_calloc(reg_count, sizeof(*close_order)))) + { + vkd3d_free(close_order); + vkd3d_free(open_order); + vkd3d_free(allocator->current_allocation); + return VKD3D_ERROR_OUT_OF_MEMORY; + } + + for (i = 0; i < reg_count; ++i) + { + struct temp_allocator_reg *reg = &allocator->ssa_regs[i]; + + if (i < allocator->ssa_count) + { + reg->type = VKD3DSPR_SSA; + reg->idx = i; + } + else + { + reg->type = VKD3DSPR_TEMP; + reg->idx = i - allocator->ssa_count; + + /* For SM 1.x ps we need to ensure that r0 is reallocated to itself, + * because it doubles as the output register. To do so we + * artificially make it alive for the whole program and make it + * compare before anything else. */ + if (reg->idx == 0 && allocator->ps_1_x) + { + reg->force_first = true; + liveness->ssa_regs[i].first_write = 0; + liveness->ssa_regs[i].last_access = UINT_MAX; } } + + reg->liveness_reg = &liveness->ssa_regs[i]; + open_order[i] = reg; + close_order[i] = reg; + } + + qsort(open_order, reg_count, sizeof(*open_order), temp_allocate_compare_open); + qsort(close_order, reg_count, sizeof(*open_order), temp_allocate_compare_close); + + for (;;) + { + struct temp_allocator_reg *reg_open = NULL, *reg_close = NULL; + bool do_open; + + if (pos_open < reg_count) + reg_open = open_order[pos_open]; + if (pos_close < reg_count) + reg_close = close_order[pos_close]; + + /* We cannot close all the registers before we finish opening them. */ + VKD3D_ASSERT(!(reg_open && !reg_close)); + + /* We finished closing registers, nothing to do any more. */ + if (!reg_close) + break; + /* There is nothing to open, so we just close. */ + else if (!reg_open) + do_open = false; + /* The next open event happens before the next close event, so we open. */ + else if (reg_open->liveness_reg->first_write < reg_close->liveness_reg->last_access) + do_open = true; + /* The other way around, we close. */ + else if (reg_close->liveness_reg->last_access < reg_open->liveness_reg->first_write) + do_open = false; + /* Ok, now we have both an open and a close happening at the same time. + * According to the strategy above, if the interval to close is + * non-degenerate, then we process it. */ + else if (reg_close->liveness_reg->first_write < reg_close->liveness_reg->last_access) + do_open = false; + /* Otherwise the interval to close is degenerate, and therefore we first + * open whatever needs to be opened. */ + else + do_open = true; + + if (do_open) + { + temp_allocator_open_register(allocator, reg_open); + ++pos_open; + } + else + { + temp_allocator_close_register(allocator, reg_close); + ++pos_close; + } } + + vkd3d_free(close_order); + vkd3d_free(open_order); + vkd3d_free(allocator->current_allocation); + return VKD3D_OK; } +/* This pass does two things: + * + * - converts SSA registers (sr#) into temp registers (r#); + * + * - contracts temp registers with non-overlapping ranges by reallocating them + * into the same register. + * + * These are done at the same time so that SSA and temp registers with + * non-overlapping liveness can share the same register. + * + * The temp contraction is not particularly sophisticated. In particular, it + * does not detect cases where a single temp register has multiple disjoint + * ranges of liveness, and it also assumes that the components used by a single + * registers is zero-based and contiguous. + * The intent for temp contraction is that HLSL will output each distinct + * variable to a unique temp ID. */ enum vkd3d_result vsir_allocate_temp_registers(struct vsir_program *program, struct vkd3d_shader_message_context *message_context) { - const unsigned int prev_temp_count = program->temp_count; + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct temp_allocator allocator = {0}; + struct vkd3d_shader_instruction *ins; struct temp_allocator_reg *regs; struct liveness_tracker tracker; enum vkd3d_result ret; - if (!program->ssa_count) + if (!program->ssa_count && !program->temp_count) return VKD3D_OK; if ((ret = track_liveness(program, &tracker))) return ret; - if (!(regs = vkd3d_calloc(program->ssa_count, sizeof(*regs)))) + if (!(regs = vkd3d_calloc(program->ssa_count + program->temp_count, sizeof(*regs)))) { liveness_tracker_cleanup(&tracker); return VKD3D_ERROR_OUT_OF_MEMORY; } allocator.message_context = message_context; + allocator.ssa_count = program->ssa_count; + allocator.temp_count = program->temp_count; allocator.ssa_regs = regs; + allocator.temp_regs = regs + program->ssa_count; + allocator.new_temp_count = 0; - for (unsigned int i = 0; i < program->ssa_count; ++i) - { - const struct liveness_tracker_reg *liveness_reg = &tracker.ssa_regs[i]; - struct temp_allocator_reg *reg = &allocator.ssa_regs[i]; + if (program->shader_version.type == VKD3D_SHADER_TYPE_PIXEL && program->shader_version.major < 2) + allocator.ps_1_x = true; - if (temp_allocator_allocate(&allocator, &tracker, reg, liveness_reg, prev_temp_count)) - { - TRACE("Allocated r%u%s to sr%u (liveness %u-%u).\n", - reg->temp_id, debug_vsir_writemask(reg->allocated_mask), i, - liveness_reg->first_write, liveness_reg->last_access); - program->temp_count = max(program->temp_count, reg->temp_id + 1); - } - ++allocator.allocated_ssa_count; + if ((ret = temp_allocator_compute_allocation_map(&allocator, &tracker)) < 0) + { + liveness_tracker_cleanup(&tracker); + vkd3d_free(regs); + return ret; } - for (unsigned int i = 0; i < program->instructions.count; ++i) + for (ins = vsir_program_iterator_head(&it); ins; ins = vsir_program_iterator_next(&it)) { - const struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; - /* Make sure we do the srcs first; setting the dst writemask may need * to remap their swizzles. */ for (unsigned int j = 0; j < ins->src_count; ++j) @@ -8886,9 +9519,11 @@ enum vkd3d_result vsir_allocate_temp_registers(struct vsir_program *program, } program->ssa_count = 0; + program->temp_count = allocator.new_temp_count; vkd3d_free(regs); liveness_tracker_cleanup(&tracker); + return allocator.result; } @@ -8902,11 +9537,17 @@ enum vkd3d_result vsir_allocate_temp_registers(struct vsir_program *program, enum vkd3d_result vsir_update_dcl_temps(struct vsir_program *program, struct vkd3d_shader_message_context *message_context) { + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); + struct vkd3d_shader_location location; + struct vkd3d_shader_instruction *ins; unsigned int temp_count = 0; - for (int i = program->instructions.count - 1; i >= 0; --i) + if (program->shader_version.major < 4) + return VKD3D_OK; + + for (ins = vsir_program_iterator_tail(&it); ins; ins = vsir_program_iterator_prev(&it)) { - struct vkd3d_shader_instruction *ins = &program->instructions.elements[i]; + location = ins->location; if (ins->opcode == VSIR_OP_DCL_TEMPS) { @@ -8915,18 +9556,17 @@ enum vkd3d_result vsir_update_dcl_temps(struct vsir_program *program, continue; } - if (temp_count && program->shader_version.major >= 4 - && (ins->opcode == VSIR_OP_HS_CONTROL_POINT_PHASE - || ins->opcode == VSIR_OP_HS_FORK_PHASE - || ins->opcode == VSIR_OP_HS_JOIN_PHASE)) + if (temp_count && (ins->opcode == VSIR_OP_HS_CONTROL_POINT_PHASE + || ins->opcode == VSIR_OP_HS_FORK_PHASE + || ins->opcode == VSIR_OP_HS_JOIN_PHASE)) { /* The phase didn't have a dcl_temps instruction, but we added * temps here, so we need to insert one. */ - if (!shader_instruction_array_insert_at(&program->instructions, i + 1, 1)) + if (!vsir_program_iterator_insert_after(&it, 1)) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = &program->instructions.elements[i + 1]; - vsir_instruction_init(ins, &program->instructions.elements[i].location, VSIR_OP_DCL_TEMPS); + ins = vsir_program_iterator_next(&it); + vsir_instruction_init(ins, &location, VSIR_OP_DCL_TEMPS); ins->declaration.count = temp_count; temp_count = 0; continue; @@ -8945,15 +9585,15 @@ enum vkd3d_result vsir_update_dcl_temps(struct vsir_program *program, } } - if (temp_count && program->shader_version.major >= 4) + if (temp_count) { - struct vkd3d_shader_instruction *ins; + ins = vsir_program_iterator_head(&it); + location = ins->location; - if (!shader_instruction_array_insert_at(&program->instructions, 0, 1)) + if (!(ins = vsir_program_iterator_insert_before_and_move(&it, 1))) return VKD3D_ERROR_OUT_OF_MEMORY; - ins = &program->instructions.elements[0]; - vsir_instruction_init(ins, &program->instructions.elements[1].location, VSIR_OP_DCL_TEMPS); + vsir_instruction_init(ins, &location, VSIR_OP_DCL_TEMPS); ins->declaration.count = temp_count; } @@ -8965,7 +9605,7 @@ struct validation_context struct vkd3d_shader_message_context *message_context; const struct vsir_program *program; size_t instruction_idx; - struct vkd3d_shader_location null_location; + struct vkd3d_shader_location location; bool invalid_instruction_idx; enum vkd3d_result status; bool dcl_temps_found; @@ -9024,13 +9664,12 @@ static void VKD3D_PRINTF_FUNC(3, 4) validator_error(struct validation_context *c if (ctx->invalid_instruction_idx) { - vkd3d_shader_error(ctx->message_context, &ctx->null_location, error, "%s", buf.buffer); + vkd3d_shader_error(ctx->message_context, &ctx->location, error, "%s", buf.buffer); WARN("VSIR validation error: %s\n", buf.buffer); } else { - const struct vkd3d_shader_instruction *ins = &ctx->program->instructions.elements[ctx->instruction_idx]; - vkd3d_shader_error(ctx->message_context, &ins->location, error, + vkd3d_shader_error(ctx->message_context, &ctx->location, error, "instruction %zu: %s", ctx->instruction_idx + 1, buf.buffer); WARN("VSIR validation error: instruction %zu: %s\n", ctx->instruction_idx + 1, buf.buffer); } @@ -9359,6 +9998,37 @@ static void vsir_validate_io_register(struct validation_context *ctx, const stru reg->idx[control_point_index].offset, control_point_count, reg->type); } +static void vsir_validate_texture_register(struct validation_context *ctx, const struct vkd3d_shader_register *reg) +{ + const struct vkd3d_shader_version *version = &ctx->program->shader_version; + uint32_t idx; + + if (version->type != VKD3D_SHADER_TYPE_PIXEL) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_REGISTER_TYPE, + "TEXTURE registers cannot be used in shader type %#x.", version->type); + + if (reg->idx_count != 1) + { + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_INDEX_COUNT, + "Invalid index count %u for a TEXTURE register.", reg->idx_count); + return; + } + + if (reg->idx[0].rel_addr) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_INDEX, + "Non-NULL relative address for a TEXTURE register."); + + if (version->major >= 3) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_REGISTER_TYPE, + "TEXTURE registers cannot be used in version %u.%u.", version->major, version->minor); + + idx = reg->idx[0].offset; + if (idx >= 8 || (vkd3d_shader_ver_le(version, 1, 4) && idx >= 6) + || (vkd3d_shader_ver_le(version, 1, 3) && idx >= 4)) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_INDEX, + "Register t%u exceeds limits for version %u.%u.", idx, version->major, version->minor); +} + static void vsir_validate_temp_register(struct validation_context *ctx, const struct vkd3d_shader_register *reg) { @@ -9459,7 +10129,8 @@ static void vsir_validate_label_register(struct validation_context *ctx, if (reg->data_type != VSIR_DATA_UNUSED) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for a LABEL register.", reg->data_type); + "Invalid data type \"%s\" (%#x) for a LABEL register.", + vsir_data_type_get_name(reg->data_type, ""), reg->data_type); if (reg->dimension != VSIR_DIMENSION_NONE) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DIMENSION, @@ -9538,7 +10209,8 @@ static void vsir_validate_sampler_register(struct validation_context *ctx, if (reg->data_type != VSIR_DATA_UNUSED) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for a SAMPLER register.", reg->data_type); + "Invalid data type \"%s\" (%#x) for a SAMPLER register.", + vsir_data_type_get_name(reg->data_type, ""), reg->data_type); /* VEC4 is allowed in gather operations. */ if (reg->dimension == VSIR_DIMENSION_SCALAR) @@ -9564,7 +10236,8 @@ static void vsir_validate_resource_register(struct validation_context *ctx, if (reg->data_type != VSIR_DATA_UNUSED) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for a RESOURCE register.", reg->data_type); + "Invalid data type \"%s\" (%#x) for a RESOURCE register.", + vsir_data_type_get_name(reg->data_type, ""), reg->data_type); if (reg->dimension != VSIR_DIMENSION_VEC4) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DIMENSION, @@ -9590,8 +10263,8 @@ static void vsir_validate_uav_register(struct validation_context *ctx, if (reg->data_type != VSIR_DATA_UNUSED) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for a UAV register.", - reg->data_type); + "Invalid data type \"%s\" (%#x) for a UAV register.", + vsir_data_type_get_name(reg->data_type, ""), reg->data_type); /* NONE is allowed in counter operations. */ if (reg->dimension == VSIR_DIMENSION_SCALAR) @@ -9663,9 +10336,10 @@ static void vsir_validate_ssa_register(struct validation_context *ctx, if (data_type_is_64_bit(data->data_type) != data_type_is_64_bit(reg->data_type)) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for a SSA register: " - "it has already been seen with data type %#x at instruction %zu.", - reg->data_type, data->data_type, data->first_seen); + "Invalid data type \"%s\" (%#x) for SSA register %u: " + "it has already been seen with data type \"%s\" (%#x) at instruction %zu.", + vsir_data_type_get_name(reg->data_type, ""), reg->data_type, reg->idx[0].offset, + vsir_data_type_get_name(data->data_type, ""), data->data_type, data->first_seen); } } @@ -9824,6 +10498,10 @@ static void vsir_validate_register(struct validation_context *ctx, vsir_validate_register_without_indices(ctx, reg); break; + case VKD3DSPR_TEXTURE: + vsir_validate_texture_register(ctx, reg); + break; + case VKD3DSPR_SSA: vsir_validate_ssa_register(ctx, reg); break; @@ -9872,6 +10550,8 @@ static void vsir_validate_io_dst_param(struct validation_context *ctx, static void vsir_validate_dst_param(struct validation_context *ctx, const struct vkd3d_shader_dst_param *dst) { + const struct vkd3d_shader_version *version = &ctx->program->shader_version; + vsir_validate_register(ctx, &dst->reg); if (dst->write_mask & ~VKD3DSP_WRITEMASK_ALL) @@ -9913,7 +10593,8 @@ static void vsir_validate_dst_param(struct validation_context *ctx, default: validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for destination with saturate modifier.", dst->reg.data_type); + "Invalid data type \"%s\" (%#x) for destination with saturate modifier.", + vsir_data_type_get_name(dst->reg.data_type, ""), dst->reg.data_type); break; } @@ -9932,7 +10613,8 @@ static void vsir_validate_dst_param(struct validation_context *ctx, case 15: if (dst->reg.data_type != VSIR_DATA_F32) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for destination with shift.", dst->reg.data_type); + "Invalid data type \"%s\" (%#x) for destination with shift.", + vsir_data_type_get_name(dst->reg.data_type, ""), dst->reg.data_type); break; default: @@ -9999,6 +10681,11 @@ static void vsir_validate_dst_param(struct validation_context *ctx, vsir_validate_io_dst_param(ctx, dst); break; + case VKD3DSPR_TEXTURE: + if (vkd3d_shader_ver_ge(version, 1, 4)) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_REGISTER_TYPE, + "Texture registers cannot be written in version %u.%u.", version->major, version->minor); + default: break; } @@ -10062,6 +10749,16 @@ static void vsir_validate_src_param(struct validation_context *ctx, validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SWIZZLE, "Source of dimension %u has invalid swizzle %#x.", src->reg.dimension, src->swizzle); + if (src->reg.dimension == VSIR_DIMENSION_VEC4 && src->reg.type == VKD3DSPR_IMMCONST + && src->swizzle != VKD3D_SHADER_NO_SWIZZLE) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SWIZZLE, + "Immediate constant source has invalid swizzle %#x.", src->swizzle); + + if (src->reg.dimension == VSIR_DIMENSION_VEC4 && src->reg.type == VKD3DSPR_IMMCONST64 + && src->swizzle != VKD3D_SHADER_SWIZZLE(X, Y, X, X)) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SWIZZLE, + "Immediate constant source has invalid swizzle %#x.", src->swizzle); + if (src->modifiers >= VKD3DSPSM_COUNT) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_MODIFIERS, "Source has invalid modifiers %#x.", src->modifiers); @@ -10070,7 +10767,8 @@ static void vsir_validate_src_param(struct validation_context *ctx, { if (!(src_modifier_data[src->modifiers].data_type_mask & (1u << src->reg.data_type))) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_MODIFIERS, - "Source has invalid modifier %#x for data type %u.", src->modifiers, src->reg.data_type); + "Source has invalid modifier %#x for data type \"%s\" (%#x).", + src->modifiers, vsir_data_type_get_name(src->reg.data_type, ""), src->reg.data_type); } switch (src->reg.type) @@ -10722,7 +11420,7 @@ static void vsir_validate_hull_shader_phase(struct validation_context *ctx, static void vsir_validate_elementwise_operation(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction, const bool types[VSIR_DATA_TYPE_COUNT]) { - enum vsir_data_type dst_data_type; + enum vsir_data_type dst_data_type, src_data_type; unsigned int i; if (instruction->dst_count < 1) @@ -10735,16 +11433,18 @@ static void vsir_validate_elementwise_operation(struct validation_context *ctx, if (!types[dst_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for elementwise operation \"%s\" (%#x).", - dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid data type \"%s\" (%#x) for elementwise operation \"%s\" (%#x).", + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); for (i = 0; i < instruction->src_count; ++i) { - if (instruction->src[i].reg.data_type != dst_data_type) + if ((src_data_type = instruction->src[i].reg.data_type) != dst_data_type) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Data type %#x for operand %u doesn't match the destination data type %#x " + "Data type \"%s\" (%#x) for operand %u doesn't match the destination data type \"%s\" (%#x) " "for elementwise operation \"%s\" (%#x).", - instruction->src[i].reg.data_type, i, dst_data_type, + vsir_data_type_get_name(src_data_type, ""), src_data_type, i, + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); } } @@ -10777,6 +11477,7 @@ static void vsir_validate_integer_elementwise_operation(struct validation_contex static const bool types[VSIR_DATA_TYPE_COUNT] = { [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, [VSIR_DATA_U32] = true, [VSIR_DATA_U64] = true, }; @@ -10784,6 +11485,18 @@ static void vsir_validate_integer_elementwise_operation(struct validation_contex vsir_validate_elementwise_operation(ctx, instruction, types); } +static void vsir_validate_signed_integer_elementwise_operation(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) +{ + static const bool types[VSIR_DATA_TYPE_COUNT] = + { + [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, + }; + + vsir_validate_elementwise_operation(ctx, instruction, types); +} + static void vsir_validate_logic_elementwise_operation(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction) { @@ -10801,7 +11514,7 @@ static void vsir_validate_logic_elementwise_operation(struct validation_context static void vsir_validate_comparison_operation(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction, const bool types[VSIR_DATA_TYPE_COUNT]) { - enum vsir_data_type dst_data_type, src_data_type; + enum vsir_data_type dst_data_type, src_data_type, data_type; unsigned int i; if (instruction->dst_count < 1) @@ -10811,8 +11524,9 @@ static void vsir_validate_comparison_operation(struct validation_context *ctx, if (dst_data_type != VSIR_DATA_U32 && dst_data_type != VSIR_DATA_BOOL) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for result of comparison operation \"%s\" (%#x).", - dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid data type \"%s\" (%#x) for result of comparison operation \"%s\" (%#x).", + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); if (instruction->src_count == 0) return; @@ -10824,16 +11538,18 @@ static void vsir_validate_comparison_operation(struct validation_context *ctx, if (!types[src_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid data type %#x for comparison operation \"%s\" (%#x).", - src_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid data type \"%s\" (%#x) for comparison operation \"%s\" (%#x).", + vsir_data_type_get_name(src_data_type, ""), src_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); for (i = 1; i < instruction->src_count; ++i) { - if (instruction->src[i].reg.data_type != src_data_type) + if ((data_type = instruction->src[i].reg.data_type) != src_data_type) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Data type %#x for operand %u doesn't match the first operands data type %#x " - "for comparison operation \"%s\" (%#x).", - instruction->src[i].reg.data_type, i, src_data_type, + "Data type \"%s\" (%#x) for operand %u doesn't match the first " + "operands data type \"%s\" (%#x) for comparison operation \"%s\" (%#x).", + vsir_data_type_get_name(data_type, ""), data_type, i, + vsir_data_type_get_name(src_data_type, ""), src_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); } } @@ -10874,6 +11590,18 @@ static void vsir_validate_integer_comparison_operation(struct validation_context vsir_validate_comparison_operation(ctx, instruction, types); } +static void vsir_validate_signed_integer_comparison_operation(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) +{ + static const bool types[VSIR_DATA_TYPE_COUNT] = + { + [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, + }; + + vsir_validate_comparison_operation(ctx, instruction, types); +} + static void vsir_validate_cast_operation(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction, const bool src_types[VSIR_DATA_TYPE_COUNT], const bool dst_types[VSIR_DATA_TYPE_COUNT]) @@ -10891,45 +11619,72 @@ static void vsir_validate_cast_operation(struct validation_context *ctx, if (!src_types[src_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid source data type %#x for cast operation \"%s\" (%#x).", - src_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid source data type \"%s\" (%#x) for cast operation \"%s\" (%#x).", + vsir_data_type_get_name(src_data_type, ""), src_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); if (!dst_types[dst_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid destination data type %#x for cast operation \"%s\" (%#x).", - dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid destination data type \"%s\" (%#x) for cast operation \"%s\" (%#x).", + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); } static void vsir_validate_shift_operation(struct validation_context *ctx, - const struct vkd3d_shader_instruction *instruction) + const struct vkd3d_shader_instruction *instruction, const bool types[VSIR_DATA_TYPE_COUNT]) { - enum vsir_data_type data_type; + enum vsir_data_type dst_data_type, src_data_type; - static const bool types[] = + static const bool shift_types[] = { [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, [VSIR_DATA_U32] = true, [VSIR_DATA_U64] = true, }; - data_type = instruction->dst[0].reg.data_type; - if ((size_t)data_type >= ARRAY_SIZE(types) || !types[data_type]) + dst_data_type = instruction->dst[0].reg.data_type; + if ((size_t)dst_data_type >= VSIR_DATA_TYPE_COUNT || !types[dst_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid destination data type %#x for shift operation \"%s\" (%#x).", - data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid destination data type \"%s\" (%#x) for shift operation \"%s\" (%#x).", + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); - if (instruction->src[0].reg.data_type != data_type) + if ((src_data_type = instruction->src[0].reg.data_type) != dst_data_type) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Data type %#x for source operand 0 doesn't match destination data type %#x " + "Data type \"%s\" (%#x) for source operand 0 doesn't match destination data type \"%s\" (%#x) " "for shift operation \"%s\" (%#x).", - instruction->src[0].reg.data_type, data_type, + vsir_data_type_get_name(src_data_type, ""), src_data_type, + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); - data_type = instruction->src[1].reg.data_type; - if ((size_t)data_type >= ARRAY_SIZE(types) || !types[data_type]) + src_data_type = instruction->src[1].reg.data_type; + if ((size_t)src_data_type >= ARRAY_SIZE(shift_types) || !shift_types[src_data_type]) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid source operand 1 data type %#x for shift operation \"%s\" (%#x).", - data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); + "Invalid source operand 1 data type \"%s\" (%#x) for shift operation \"%s\" (%#x).", + vsir_data_type_get_name(src_data_type, ""), src_data_type, + vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode); +} + +static void vsir_validate_bem(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction) +{ + const struct vkd3d_shader_version *version = &ctx->program->shader_version; + + if (version->type != VKD3D_SHADER_TYPE_PIXEL) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_REGISTER_TYPE, + "BEM cannot be used in shader type %#x.", version->type); + + if (version->major != 1 || version->minor != 4) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_OPCODE, + "BEM cannot be used in version %u.%u.", version->major, version->minor); + + if (instruction->dst[0].write_mask != 0x3) + validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_WRITE_MASK, + "Invalid BEM write mask %#x.", instruction->dst[0].write_mask); + + /* Strictly not an elementwise operation, but we expect all the arguments + * to be float. */ + vsir_validate_float_elementwise_operation(ctx, instruction); } static void vsir_validate_branch(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction) @@ -11490,14 +12245,39 @@ static void vsir_validate_ifc(struct validation_context *ctx, const struct vkd3d vsir_validator_push_block(ctx, VSIR_OP_IF); } +static void vsir_validate_ishl(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) +{ + static const bool types[VSIR_DATA_TYPE_COUNT] = + { + [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, + [VSIR_DATA_U32] = true, + [VSIR_DATA_U64] = true, + }; + + vsir_validate_shift_operation(ctx, instruction, types); +} + +static void vsir_validate_ishr(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) +{ + static const bool types[VSIR_DATA_TYPE_COUNT] = + { + [VSIR_DATA_I32] = true, + [VSIR_DATA_I64] = true, + }; + + vsir_validate_shift_operation(ctx, instruction, types); +} + static void vsir_validate_itof(struct validation_context *ctx, const struct vkd3d_shader_instruction *instruction) { static const bool src_types[VSIR_DATA_TYPE_COUNT] = { [VSIR_DATA_BOOL] = true, [VSIR_DATA_I32] = true, - [VSIR_DATA_U32] = true, - [VSIR_DATA_U64] = true, + [VSIR_DATA_I64] = true, }; static const bool dst_types[VSIR_DATA_TYPE_COUNT] = { @@ -11624,7 +12404,8 @@ static void vsir_validate_throw_invalid_dst_type_error_with_flags(struct validat enum vsir_data_type dst_data_type = instruction->dst[0].reg.data_type; validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DATA_TYPE, - "Invalid destination data type %#x for operation \"%s\" (%#x) with flags %#x.", dst_data_type, + "Invalid destination data type \"%s\" (%#x) for operation \"%s\" (%#x) with flags %#x.", + vsir_data_type_get_name(dst_data_type, ""), dst_data_type, vsir_opcode_get_name(instruction->opcode, ""), instruction->opcode, instruction->flags); } @@ -11721,6 +12502,18 @@ static void vsir_validate_switch_monolithic(struct validation_context *ctx, ctx->inside_block = false; } +static void vsir_validate_ushr(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) +{ + static const bool types[VSIR_DATA_TYPE_COUNT] = + { + [VSIR_DATA_U32] = true, + [VSIR_DATA_U64] = true, + }; + + vsir_validate_shift_operation(ctx, instruction, types); +} + struct vsir_validator_instruction_desc { unsigned int dst_param_count; @@ -11736,6 +12529,7 @@ static const struct vsir_validator_instruction_desc vsir_validator_instructions[ [VSIR_OP_AND] = {1, 2, vsir_validate_logic_elementwise_operation}, [VSIR_OP_ASIN] = {1, 1, vsir_validate_float_elementwise_operation}, [VSIR_OP_ATAN] = {1, 1, vsir_validate_float_elementwise_operation}, + [VSIR_OP_BEM] = {1, 2, vsir_validate_bem}, [VSIR_OP_BRANCH] = {0, ~0u, vsir_validate_branch}, [VSIR_OP_DADD] = {1, 2, vsir_validate_double_elementwise_operation}, [VSIR_OP_DCL_GS_INSTANCES] = {0, 0, vsir_validate_dcl_gs_instances}, @@ -11800,17 +12594,17 @@ static const struct vsir_validator_instruction_desc vsir_validator_instructions[ [VSIR_OP_IEQ] = {1, 2, vsir_validate_integer_comparison_operation}, [VSIR_OP_IF] = {0, 1, vsir_validate_if}, [VSIR_OP_IFC] = {0, 2, vsir_validate_ifc}, - [VSIR_OP_IGE] = {1, 2, vsir_validate_integer_comparison_operation}, - [VSIR_OP_ILT] = {1, 2, vsir_validate_integer_comparison_operation}, + [VSIR_OP_IGE] = {1, 2, vsir_validate_signed_integer_comparison_operation}, + [VSIR_OP_ILT] = {1, 2, vsir_validate_signed_integer_comparison_operation}, [VSIR_OP_IMAD] = {1, 3, vsir_validate_integer_elementwise_operation}, - [VSIR_OP_IMAX] = {1, 2, vsir_validate_integer_elementwise_operation}, - [VSIR_OP_IMIN] = {1, 2, vsir_validate_integer_elementwise_operation}, + [VSIR_OP_IMAX] = {1, 2, vsir_validate_signed_integer_elementwise_operation}, + [VSIR_OP_IMIN] = {1, 2, vsir_validate_signed_integer_elementwise_operation}, [VSIR_OP_INE] = {1, 2, vsir_validate_integer_comparison_operation}, [VSIR_OP_INEG] = {1, 1, vsir_validate_integer_elementwise_operation}, [VSIR_OP_IREM] = {1, 2, vsir_validate_integer_elementwise_operation}, [VSIR_OP_ISFINITE] = {1, 1, vsir_validate_float_comparison_operation}, - [VSIR_OP_ISHL] = {1, 2, vsir_validate_shift_operation}, - [VSIR_OP_ISHR] = {1, 2, vsir_validate_shift_operation}, + [VSIR_OP_ISHL] = {1, 2, vsir_validate_ishl}, + [VSIR_OP_ISHR] = {1, 2, vsir_validate_ishr}, [VSIR_OP_ISINF] = {1, 1, vsir_validate_float_comparison_operation}, [VSIR_OP_ISNAN] = {1, 1, vsir_validate_float_comparison_operation}, [VSIR_OP_ITOF] = {1, 1, vsir_validate_itof}, @@ -11840,16 +12634,15 @@ static const struct vsir_validator_instruction_desc vsir_validator_instructions[ [VSIR_OP_SAMPLE_INFO] = {1, 1, vsir_validate_sample_info}, [VSIR_OP_SWITCH] = {0, 1, vsir_validate_switch}, [VSIR_OP_SWITCH_MONOLITHIC] = {0, ~0u, vsir_validate_switch_monolithic}, + [VSIR_OP_USHR] = {1, 2, vsir_validate_ushr}, }; -static void vsir_validate_instruction(struct validation_context *ctx) +static void vsir_validate_instruction(struct validation_context *ctx, + const struct vkd3d_shader_instruction *instruction) { const struct vkd3d_shader_version *version = &ctx->program->shader_version; - const struct vkd3d_shader_instruction *instruction; size_t i; - instruction = &ctx->program->instructions.elements[ctx->instruction_idx]; - for (i = 0; i < instruction->dst_count; ++i) vsir_validate_dst_param(ctx, &instruction->dst[i]); @@ -11927,7 +12720,7 @@ enum vkd3d_result vsir_program_validate(struct vsir_program *program, uint64_t c { .message_context = message_context, .program = program, - .null_location = {.source_name = source_name}, + .location = {.source_name = source_name}, .status = VKD3D_OK, .phase = VSIR_OP_INVALID, .invalid_instruction_idx = true, @@ -11938,6 +12731,8 @@ enum vkd3d_result vsir_program_validate(struct vsir_program *program, uint64_t c .inner_tess_idxs[0] = ~0u, .inner_tess_idxs[1] = ~0u, }; + struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); + struct vkd3d_shader_instruction *ins; unsigned int i; if (!(config_flags & VKD3D_SHADER_CONFIG_FLAG_FORCE_VALIDATION)) @@ -12046,11 +12841,17 @@ enum vkd3d_result vsir_program_validate(struct vsir_program *program, uint64_t c ctx.invalid_instruction_idx = false; - for (ctx.instruction_idx = 0; ctx.instruction_idx < program->instructions.count - && ctx.status != VKD3D_ERROR_OUT_OF_MEMORY; ++ctx.instruction_idx) - vsir_validate_instruction(&ctx); + ctx.instruction_idx = 0; + for (ins = vsir_program_iterator_head(&it); ins && ctx.status != VKD3D_ERROR_OUT_OF_MEMORY; + ins = vsir_program_iterator_next(&it)) + { + ctx.location = ins->location; + vsir_validate_instruction(&ctx, ins); + ++ctx.instruction_idx; + } ctx.invalid_instruction_idx = true; + ctx.location = (struct vkd3d_shader_location){.source_name = source_name}; if (ctx.depth != 0) validator_error(&ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "%zu nested blocks were not closed.", ctx.depth); @@ -12113,14 +12914,9 @@ static void vsir_transform_( enum vkd3d_result vsir_program_transform_early(struct vsir_program *program, uint64_t config_flags, const struct vkd3d_shader_compile_info *compile_info, struct vkd3d_shader_message_context *message_context) { - struct vsir_transformation_context ctx = - { - .result = VKD3D_OK, - .program = program, - .config_flags = config_flags, - .compile_info = compile_info, - .message_context = message_context, - }; + struct vsir_transformation_context ctx; + + vsir_transformation_context_init(&ctx, program, config_flags, compile_info, message_context); /* For vsir_program_ensure_diffuse(). */ if (program->shader_version.major <= 2) @@ -12138,18 +12934,18 @@ enum vkd3d_result vsir_program_transform_early(struct vsir_program *program, uin enum vkd3d_result vsir_program_lower_d3dbc(struct vsir_program *program, uint64_t config_flags, const struct vkd3d_shader_compile_info *compile_info, struct vkd3d_shader_message_context *message_context) { - struct vsir_transformation_context ctx = - { - .result = VKD3D_OK, - .program = program, - .config_flags = config_flags, - .compile_info = compile_info, - .message_context = message_context, - }; + struct vsir_transformation_context ctx; + vsir_transformation_context_init(&ctx, program, config_flags, compile_info, message_context); vsir_transform(&ctx, vsir_program_lower_d3dbc_instructions); + if (program->shader_version.major == 1 && program->shader_version.type == VKD3D_SHADER_TYPE_PIXEL) + { + if (program->shader_version.minor < 4) + vsir_transform(&ctx, vsir_program_lower_texture_writes); + vsir_transform(&ctx, vsir_program_normalise_ps1_output); + } if (TRACE_ON()) vsir_program_trace(program); @@ -12160,15 +12956,9 @@ enum vkd3d_result vsir_program_lower_d3dbc(struct vsir_program *program, uint64_ enum vkd3d_result vsir_program_transform(struct vsir_program *program, uint64_t config_flags, const struct vkd3d_shader_compile_info *compile_info, struct vkd3d_shader_message_context *message_context) { - struct vsir_transformation_context ctx = - { - .result = VKD3D_OK, - .program = program, - .config_flags = config_flags, - .compile_info = compile_info, - .message_context = message_context, - }; + struct vsir_transformation_context ctx; + vsir_transformation_context_init(&ctx, program, config_flags, compile_info, message_context); vsir_transform(&ctx, vsir_program_lower_instructions); if (program->shader_version.major >= 6) diff --git a/libs/vkd3d/libs/vkd3d-shader/msl.c b/libs/vkd3d/libs/vkd3d-shader/msl.c index c6e048adb20..8a89dcabd5e 100644 --- a/libs/vkd3d/libs/vkd3d-shader/msl.c +++ b/libs/vkd3d/libs/vkd3d-shader/msl.c @@ -167,6 +167,7 @@ static void msl_print_register_datatype(struct vkd3d_string_buffer *buffer, case VSIR_DATA_F32: vkd3d_string_buffer_printf(buffer, "f"); break; + case VSIR_DATA_BOOL: case VSIR_DATA_I32: vkd3d_string_buffer_printf(buffer, "i"); break; @@ -485,6 +486,22 @@ static enum msl_data_type msl_print_register_name(struct vkd3d_string_buffer *bu vkd3d_string_buffer_printf(buffer, "o_mask"); return MSL_DATA_UNION; + case VKD3DSPR_THREADID: + vkd3d_string_buffer_printf(buffer, "v_thread_id"); + return MSL_DATA_UNION; + + case VKD3DSPR_THREADGROUPID: + vkd3d_string_buffer_printf(buffer, "v_thread_group_id"); + return MSL_DATA_UNION; + + case VKD3DSPR_LOCALTHREADID: + vkd3d_string_buffer_printf(buffer, "v_local_thread_id"); + return MSL_DATA_UNION; + + case VKD3DSPR_LOCALTHREADINDEX: + vkd3d_string_buffer_printf(buffer, "v_local_thread_index"); + return MSL_DATA_UNION; + default: msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_INTERNAL, "Internal compiler error: Unhandled register type %#x.", reg->type); @@ -799,6 +816,7 @@ static void msl_relop(struct msl_generator *gen, const struct vkd3d_shader_instr static void msl_cast(struct msl_generator *gen, const struct vkd3d_shader_instruction *ins, const char *constructor) { unsigned int component_count; + const char *negate; struct msl_src src; struct msl_dst dst; uint32_t mask; @@ -806,10 +824,11 @@ static void msl_cast(struct msl_generator *gen, const struct vkd3d_shader_instru mask = msl_dst_init(&dst, gen, ins, &ins->dst[0]); msl_src_init(&src, gen, &ins->src[0], mask); + negate = ins->opcode == VSIR_OP_UTOF && data_type_is_bool(ins->src[0].reg.data_type) ? "-" : ""; if ((component_count = vsir_write_mask_component_count(mask)) > 1) - msl_print_assignment(gen, &dst, "%s%u(%s)", constructor, component_count, src.str->buffer); + msl_print_assignment(gen, &dst, "%s%u(%s%s)", constructor, component_count, negate, src.str->buffer); else - msl_print_assignment(gen, &dst, "%s(%s)", constructor, src.str->buffer); + msl_print_assignment(gen, &dst, "%s(%s%s)", constructor, negate, src.str->buffer); msl_src_cleanup(&src, &gen->string_buffers); msl_dst_cleanup(&dst, &gen->string_buffers); @@ -943,8 +962,8 @@ static void msl_print_texel_offset(struct vkd3d_string_buffer *buffer, struct ms static void msl_ld(struct msl_generator *gen, const struct vkd3d_shader_instruction *ins) { + unsigned int resource_id, resource_idx, resource_space, sample_count; const struct msl_resource_type_info *resource_type_info; - unsigned int resource_id, resource_idx, resource_space; const struct vkd3d_shader_descriptor_info1 *descriptor; const struct vkd3d_shader_descriptor_binding *binding; enum vkd3d_shader_resource_type resource_type; @@ -969,6 +988,7 @@ static void msl_ld(struct msl_generator *gen, const struct vkd3d_shader_instruct { resource_type = descriptor->resource_type; resource_space = descriptor->register_space; + sample_count = descriptor->sample_count; data_type = descriptor->resource_data_type; } else @@ -977,6 +997,7 @@ static void msl_ld(struct msl_generator *gen, const struct vkd3d_shader_instruct "Internal compiler error: Undeclared resource descriptor %u.", resource_id); resource_space = 0; resource_type = VKD3D_SHADER_RESOURCE_TEXTURE_2D; + sample_count = 1; data_type = VSIR_DATA_F32; } @@ -988,6 +1009,16 @@ static void msl_ld(struct msl_generator *gen, const struct vkd3d_shader_instruct msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_UNSUPPORTED, "Texel fetches from resource type %#x are not supported.", resource_type); + if (sample_count == 1) + { + /* Similar to the SPIR-V and GLSL targets, we map multi-sample + * textures with sample count 1 to their single-sample equivalents. */ + if (resource_type == VKD3D_SHADER_RESOURCE_TEXTURE_2DMS) + resource_type = VKD3D_SHADER_RESOURCE_TEXTURE_2D; + else if (resource_type == VKD3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY) + resource_type = VKD3D_SHADER_RESOURCE_TEXTURE_2DARRAY; + } + if (!(resource_type_info = msl_get_resource_type_info(resource_type))) { msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_INTERNAL, @@ -1030,6 +1061,10 @@ static void msl_ld(struct msl_generator *gen, const struct vkd3d_shader_instruct vkd3d_string_buffer_printf(read, ", "); if (ins->opcode != VSIR_OP_LD2DMS) msl_print_src_with_type(read, gen, &ins->src[0], VKD3DSP_WRITEMASK_3, VSIR_DATA_U32); + else if (sample_count == 1) + /* If the resource isn't a true multisample resource, this is the + * "lod" parameter instead of the "sample" parameter. */ + vkd3d_string_buffer_printf(read, "0"); else msl_print_src_with_type(read, gen, &ins->src[2], VKD3DSP_WRITEMASK_0, VSIR_DATA_U32); } @@ -1294,6 +1329,11 @@ static void msl_store_uav_typed(struct msl_generator *gen, const struct vkd3d_sh data_type = VSIR_DATA_F32; } + if (resource_type == VKD3D_SHADER_RESOURCE_TEXTURE_2DMS + || resource_type == VKD3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY) + msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_UNSUPPORTED, + "Storing to resource type %#x is not supported.", resource_type); + if (!(resource_type_info = msl_get_resource_type_info(resource_type))) { msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_INTERNAL, @@ -1414,6 +1454,12 @@ static void msl_ret(struct msl_generator *gen, const struct vkd3d_shader_instruc static void msl_dcl_indexable_temp(struct msl_generator *gen, const struct vkd3d_shader_instruction *ins) { const char *type = ins->declaration.indexable_temp.component_count == 4 ? "vkd3d_vec4" : "vkd3d_scalar"; + + if (ins->declaration.indexable_temp.initialiser) + msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_INTERNAL, + "Internal compiler error: Unhandled initialiser for indexable temporary %u.", + ins->declaration.indexable_temp.register_idx); + msl_print_indent(gen->buffer, gen->indent); vkd3d_string_buffer_printf(gen->buffer, "%s x%u[%u];\n", type, ins->declaration.indexable_temp.register_idx, @@ -1426,6 +1472,9 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d switch (ins->opcode) { + case VSIR_OP_ACOS: + msl_intrinsic(gen, ins, "acos"); + break; case VSIR_OP_ADD: case VSIR_OP_IADD: msl_binop(gen, ins, "+"); @@ -1433,6 +1482,12 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d case VSIR_OP_AND: msl_binop(gen, ins, "&"); break; + case VSIR_OP_ASIN: + msl_intrinsic(gen, ins, "asin"); + break; + case VSIR_OP_ATAN: + msl_intrinsic(gen, ins, "atan"); + break; case VSIR_OP_BREAK: msl_break(gen); break; @@ -1442,6 +1497,9 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d case VSIR_OP_CONTINUE: msl_continue(gen); break; + case VSIR_OP_COS: + msl_intrinsic(gen, ins, "cos"); + break; case VSIR_OP_DCL_INDEXABLE_TEMP: msl_dcl_indexable_temp(gen, ins); break; @@ -1497,6 +1555,9 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d case VSIR_OP_FRC: msl_intrinsic(gen, ins, "fract"); break; + case VSIR_OP_FREM: + msl_intrinsic(gen, ins, "fmod"); + break; case VSIR_OP_FTOI: msl_cast(gen, ins, "int"); break; @@ -1516,8 +1577,18 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d break; case VSIR_OP_GEO: case VSIR_OP_IGE: + case VSIR_OP_UGE: msl_relop(gen, ins, ">="); break; + case VSIR_OP_HCOS: + msl_intrinsic(gen, ins, "cosh"); + break; + case VSIR_OP_HSIN: + msl_intrinsic(gen, ins, "sinh"); + break; + case VSIR_OP_HTAN: + msl_intrinsic(gen, ins, "tanh"); + break; case VSIR_OP_IF: msl_if(gen, ins); break; @@ -1602,6 +1673,9 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d case VSIR_OP_RSQ: msl_intrinsic(gen, ins, "rsqrt"); break; + case VSIR_OP_SIN: + msl_intrinsic(gen, ins, "sin"); + break; case VSIR_OP_SQRT: msl_intrinsic(gen, ins, "sqrt"); break; @@ -1611,6 +1685,9 @@ static void msl_handle_instruction(struct msl_generator *gen, const struct vkd3d case VSIR_OP_SWITCH: msl_switch(gen, ins); break; + case VSIR_OP_TAN: + msl_intrinsic(gen, ins, "tan"); + break; case VSIR_OP_XOR: msl_binop(gen, ins, "^"); break; @@ -1951,6 +2028,30 @@ static void msl_generate_entrypoint_prologue(struct msl_generator *gen) msl_print_write_mask(buffer, e->mask); vkd3d_string_buffer_printf(buffer, ";\n"); } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADID)) + { + msl_print_indent(gen->buffer, 1); + vkd3d_string_buffer_printf(buffer, "v_thread_id.u = uint4(thread_id, 0u);\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADGROUPID)) + { + msl_print_indent(gen->buffer, 1); + vkd3d_string_buffer_printf(buffer, "v_thread_group_id.u = uint4(thread_group_id, 0u);\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADID)) + { + msl_print_indent(gen->buffer, 1); + vkd3d_string_buffer_printf(buffer, "v_local_thread_id.u = uint4(local_thread_id, 0u);\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADINDEX)) + { + msl_print_indent(gen->buffer, 1); + vkd3d_string_buffer_printf(buffer, "v_local_thread_index.u = uint4(local_thread_index, 0u, 0u, 0u);\n"); + } } static void msl_generate_entrypoint_epilogue(struct msl_generator *gen) @@ -1995,6 +2096,7 @@ static void msl_generate_entrypoint_epilogue(struct msl_generator *gen) static void msl_generate_entrypoint(struct msl_generator *gen) { enum vkd3d_shader_type type = gen->program->shader_version.type; + bool output = true; switch (type) { @@ -2006,13 +2108,21 @@ static void msl_generate_entrypoint(struct msl_generator *gen) vkd3d_string_buffer_printf(gen->buffer, "[[early_fragment_tests]]\n"); vkd3d_string_buffer_printf(gen->buffer, "fragment "); break; + case VKD3D_SHADER_TYPE_COMPUTE: + vkd3d_string_buffer_printf(gen->buffer, "kernel "); + output = false; + break; default: msl_compiler_error(gen, VKD3D_SHADER_ERROR_MSL_INTERNAL, "Internal compiler error: Unhandled shader type %#x.", type); return; } - vkd3d_string_buffer_printf(gen->buffer, "vkd3d_%s_out shader_entry(\n", gen->prefix); + if (output) + vkd3d_string_buffer_printf(gen->buffer, "vkd3d_%s_out ", gen->prefix); + else + vkd3d_string_buffer_printf(gen->buffer, "void "); + vkd3d_string_buffer_printf(gen->buffer, "shader_entry(\n"); if (gen->program->descriptors.descriptor_count) { @@ -2028,6 +2138,30 @@ static void msl_generate_entrypoint(struct msl_generator *gen) vkd3d_string_buffer_printf(gen->buffer, "uint vertex_id [[vertex_id]],\n"); } + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADID)) + { + msl_print_indent(gen->buffer, 2); + vkd3d_string_buffer_printf(gen->buffer, "uint3 thread_id [[thread_position_in_grid]],\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADGROUPID)) + { + msl_print_indent(gen->buffer, 2); + vkd3d_string_buffer_printf(gen->buffer, "uint3 thread_group_id [[threadgroup_position_in_grid]],\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADID)) + { + msl_print_indent(gen->buffer, 2); + vkd3d_string_buffer_printf(gen->buffer, "uint3 local_thread_id [[thread_position_in_threadgroup]],\n"); + } + + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADINDEX)) + { + msl_print_indent(gen->buffer, 2); + vkd3d_string_buffer_printf(gen->buffer, "uint local_thread_index [[thread_index_in_threadgroup]],\n"); + } + msl_print_indent(gen->buffer, 2); vkd3d_string_buffer_printf(gen->buffer, "vkd3d_%s_in input [[stage_in]])\n{\n", gen->prefix); @@ -2037,6 +2171,14 @@ static void msl_generate_entrypoint(struct msl_generator *gen) vkd3d_string_buffer_printf(gen->buffer, " vkd3d_%s_out output;\n", gen->prefix); if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_SAMPLEMASK)) vkd3d_string_buffer_printf(gen->buffer, " vkd3d_scalar o_mask;\n"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADID)) + vkd3d_string_buffer_printf(gen->buffer, " vkd3d_vec4 v_thread_id;\n"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADGROUPID)) + vkd3d_string_buffer_printf(gen->buffer, " vkd3d_vec4 v_thread_group_id;\n"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADID)) + vkd3d_string_buffer_printf(gen->buffer, " vkd3d_vec4 v_local_thread_id;\n"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADINDEX)) + vkd3d_string_buffer_printf(gen->buffer, " vkd3d_vec4 v_local_thread_index;\n"); vkd3d_string_buffer_printf(gen->buffer, "\n"); msl_generate_entrypoint_prologue(gen); @@ -2048,13 +2190,23 @@ static void msl_generate_entrypoint(struct msl_generator *gen) vkd3d_string_buffer_printf(gen->buffer, ", output.shader_out_depth"); if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_SAMPLEMASK)) vkd3d_string_buffer_printf(gen->buffer, ", o_mask"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADID)) + vkd3d_string_buffer_printf(gen->buffer, ", v_thread_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADGROUPID)) + vkd3d_string_buffer_printf(gen->buffer, ", v_thread_group_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADID)) + vkd3d_string_buffer_printf(gen->buffer, ", v_local_thread_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADINDEX)) + vkd3d_string_buffer_printf(gen->buffer, ", v_local_thread_index"); if (gen->program->descriptors.descriptor_count) vkd3d_string_buffer_printf(gen->buffer, ", descriptors"); vkd3d_string_buffer_printf(gen->buffer, ");\n\n"); msl_generate_entrypoint_epilogue(gen); - vkd3d_string_buffer_printf(gen->buffer, " return output;\n}\n"); + if (output) + vkd3d_string_buffer_printf(gen->buffer, " return output;\n"); + vkd3d_string_buffer_printf(gen->buffer, "}\n"); } static int msl_generator_generate(struct msl_generator *gen, struct vkd3d_shader_code *out) @@ -2119,6 +2271,14 @@ static int msl_generator_generate(struct msl_generator *gen, struct vkd3d_shader vkd3d_string_buffer_printf(gen->buffer, ", thread float &o_depth"); if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_SAMPLEMASK)) vkd3d_string_buffer_printf(gen->buffer, ", thread vkd3d_scalar &o_mask"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADID)) + vkd3d_string_buffer_printf(gen->buffer, ", thread vkd3d_vec4 &v_thread_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_THREADGROUPID)) + vkd3d_string_buffer_printf(gen->buffer, ", thread vkd3d_vec4 &v_thread_group_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADID)) + vkd3d_string_buffer_printf(gen->buffer, ", thread vkd3d_vec4 &v_local_thread_id"); + if (bitmap_is_set(gen->program->io_dcls, VKD3DSPR_LOCALTHREADINDEX)) + vkd3d_string_buffer_printf(gen->buffer, ", thread vkd3d_vec4 &v_local_thread_index"); if (gen->program->descriptors.descriptor_count) vkd3d_string_buffer_printf(gen->buffer, ", constant descriptor *descriptors"); vkd3d_string_buffer_printf(gen->buffer, ")\n{\n"); diff --git a/libs/vkd3d/libs/vkd3d-shader/preproc.l b/libs/vkd3d/libs/vkd3d-shader/preproc.l index a8c0db358bc..f9b1d67ac36 100644 --- a/libs/vkd3d/libs/vkd3d-shader/preproc.l +++ b/libs/vkd3d/libs/vkd3d-shader/preproc.l @@ -20,6 +20,7 @@ %{ +#include "preproc.h" #include "preproc.tab.h" #undef ERROR /* defined in wingdi.h */ @@ -823,7 +824,6 @@ int preproc_lexer_parse(const struct vkd3d_shader_compile_info *compile_info, static const struct vkd3d_shader_preprocess_info default_preprocess_info = {0}; struct preproc_ctx ctx = {0}; char *source_name = NULL; - void *output_code; unsigned int i; vkd3d_string_buffer_init(&ctx.buffer); @@ -900,16 +900,9 @@ int preproc_lexer_parse(const struct vkd3d_shader_compile_info *compile_info, return VKD3D_ERROR_INVALID_SHADER; } - if (!(output_code = vkd3d_malloc(ctx.buffer.content_size))) - { - vkd3d_string_buffer_cleanup(&ctx.buffer); - return VKD3D_ERROR_OUT_OF_MEMORY; - } - memcpy(output_code, ctx.buffer.buffer, ctx.buffer.content_size); - out->size = ctx.buffer.content_size; - out->code = output_code; vkd3d_string_buffer_trace(&ctx.buffer); - vkd3d_string_buffer_cleanup(&ctx.buffer); + + vkd3d_shader_code_from_string_buffer(out, &ctx.buffer); return VKD3D_OK; fail: diff --git a/libs/vkd3d/libs/vkd3d-shader/spirv.c b/libs/vkd3d/libs/vkd3d-shader/spirv.c index 97c0d0e73a8..9c5f71ee11f 100644 --- a/libs/vkd3d/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d/libs/vkd3d-shader/spirv.c @@ -2518,7 +2518,7 @@ static uint32_t vkd3d_spirv_build_op_glsl_std450_nclamp(struct vkd3d_spirv_build GLSLstd450NClamp, operands, ARRAY_SIZE(operands)); } -static uint32_t vkd3d_spirv_get_type_id(struct vkd3d_spirv_builder *builder, +static uint32_t spirv_get_type_id_for_component_type(struct vkd3d_spirv_builder *builder, enum vkd3d_shader_component_type component_type, unsigned int component_count) { uint32_t scalar_id, type_id; @@ -2553,8 +2553,9 @@ static uint32_t vkd3d_spirv_get_type_id(struct vkd3d_spirv_builder *builder, case VKD3D_SHADER_COMPONENT_DOUBLE: type_id = vkd3d_spirv_get_op_type_float(builder, 64); break; + case VKD3D_SHADER_COMPONENT_INT64: case VKD3D_SHADER_COMPONENT_UINT64: - type_id = vkd3d_spirv_get_op_type_int(builder, 64, 0); + type_id = vkd3d_spirv_get_op_type_int(builder, 64, component_type == VKD3D_SHADER_COMPONENT_INT64); break; default: FIXME("Unhandled component type %#x.\n", component_type); @@ -2564,7 +2565,7 @@ static uint32_t vkd3d_spirv_get_type_id(struct vkd3d_spirv_builder *builder, else { VKD3D_ASSERT(component_type != VKD3D_SHADER_COMPONENT_VOID); - scalar_id = vkd3d_spirv_get_type_id(builder, component_type, 1); + scalar_id = spirv_get_type_id_for_component_type(builder, component_type, 1); type_id = vkd3d_spirv_get_op_type_vector(builder, scalar_id, component_count); } @@ -2573,13 +2574,13 @@ static uint32_t vkd3d_spirv_get_type_id(struct vkd3d_spirv_builder *builder, return type_id; } -static uint32_t vkd3d_spirv_get_type_id_for_data_type(struct vkd3d_spirv_builder *builder, +static uint32_t spirv_get_type_id(struct vkd3d_spirv_builder *builder, enum vsir_data_type data_type, unsigned int component_count) { enum vkd3d_shader_component_type component_type; component_type = vkd3d_component_type_from_data_type(data_type); - return vkd3d_spirv_get_type_id(builder, component_type, component_count); + return spirv_get_type_id_for_component_type(builder, component_type, component_count); } static void vkd3d_spirv_builder_init(struct vkd3d_spirv_builder *builder, @@ -2794,7 +2795,7 @@ struct vkd3d_symbol_register_data { SpvStorageClass storage_class; uint32_t member_idx; - enum vkd3d_shader_component_type component_type; + enum vsir_data_type data_type; unsigned int write_mask; unsigned int structure_stride; unsigned int binding_base_idx; @@ -2804,7 +2805,7 @@ struct vkd3d_symbol_register_data struct vkd3d_symbol_resource_data { struct vkd3d_shader_register_range range; - enum vkd3d_shader_component_type sampled_type; + enum vsir_data_type sampled_type; uint32_t type_id; const struct vkd3d_spirv_resource_type *resource_type_info; unsigned int structure_stride; @@ -2919,15 +2920,14 @@ static void vkd3d_symbol_make_io(struct vkd3d_symbol *symbol, symbol->key.reg.idx = index; } -static void vkd3d_symbol_set_register_info(struct vkd3d_symbol *symbol, - uint32_t val_id, SpvStorageClass storage_class, - enum vkd3d_shader_component_type component_type, uint32_t write_mask) +static void vkd3d_symbol_set_register_info(struct vkd3d_symbol *symbol, uint32_t val_id, + SpvStorageClass storage_class, enum vsir_data_type data_type, uint32_t write_mask) { symbol->id = val_id; symbol->descriptor_array = NULL; symbol->info.reg.storage_class = storage_class; symbol->info.reg.member_idx = 0; - symbol->info.reg.component_type = component_type; + symbol->info.reg.data_type = data_type; symbol->info.reg.write_mask = write_mask; symbol->info.reg.structure_stride = 0; symbol->info.reg.binding_base_idx = 0; @@ -3253,6 +3253,14 @@ static struct spirv_compiler *spirv_compiler_create(struct vsir_program *program compiler->features = option->value; break; + case VKD3D_SHADER_COMPILE_OPTION_PACK_MATRIX_ORDER: + case VKD3D_SHADER_COMPILE_OPTION_BACKWARD_COMPATIBILITY: + case VKD3D_SHADER_COMPILE_OPTION_CHILD_EFFECT: + case VKD3D_SHADER_COMPILE_OPTION_WARN_IMPLICIT_TRUNCATION: + case VKD3D_SHADER_COMPILE_OPTION_INCLUDE_EMPTY_BUFFERS_IN_EFFECTS: + /* Explicitly ignored for this target. */ + break; + default: WARN("Ignoring unrecognised option %#x with value %#x.\n", option->name, option->value); break; @@ -3649,122 +3657,128 @@ static const struct vkd3d_symbol *spirv_compiler_put_symbol(struct spirv_compile } static uint32_t spirv_compiler_get_constant(struct spirv_compiler *compiler, - enum vkd3d_shader_component_type component_type, unsigned int component_count, const uint32_t *values) + enum vsir_data_type data_type, unsigned int component_count, const uint32_t *values) { uint32_t type_id, scalar_type_id, component_ids[VKD3D_VEC4_SIZE]; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; unsigned int i; VKD3D_ASSERT(0 < component_count && component_count <= VKD3D_VEC4_SIZE); - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id(builder, data_type, component_count); - switch (component_type) + switch (data_type) { - case VKD3D_SHADER_COMPONENT_UINT: - case VKD3D_SHADER_COMPONENT_INT: - case VKD3D_SHADER_COMPONENT_FLOAT: - break; - case VKD3D_SHADER_COMPONENT_BOOL: + case VSIR_DATA_BOOL: if (component_count == 1) return vkd3d_spirv_get_op_constant_bool(builder, type_id, *values); - FIXME("Unsupported vector of bool.\n"); spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_TYPE, "Vectors of bool type are not supported."); return vkd3d_spirv_get_op_undef(builder, type_id); + + case VSIR_DATA_F16: + case VSIR_DATA_F32: + case VSIR_DATA_I32: + case VSIR_DATA_U16: + case VSIR_DATA_U32: + case VSIR_DATA_SNORM: + case VSIR_DATA_UNORM: + break; + default: spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_TYPE, - "Unhandled component_type %#x.", component_type); + "Unhandled data type \"%s\" (%#x).", + vsir_data_type_get_name(data_type, ""), data_type); return vkd3d_spirv_get_op_undef(builder, type_id); } if (component_count == 1) - { return vkd3d_spirv_get_op_constant(builder, type_id, *values); - } - else + + scalar_type_id = spirv_get_type_id(builder, data_type, 1); + for (i = 0; i < component_count; ++i) { - scalar_type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); - for (i = 0; i < component_count; ++i) - component_ids[i] = vkd3d_spirv_get_op_constant(builder, scalar_type_id, values[i]); - return vkd3d_spirv_get_op_constant_composite(builder, type_id, component_ids, component_count); + component_ids[i] = vkd3d_spirv_get_op_constant(builder, scalar_type_id, values[i]); } + + return vkd3d_spirv_get_op_constant_composite(builder, type_id, component_ids, component_count); } static uint32_t spirv_compiler_get_constant64(struct spirv_compiler *compiler, - enum vkd3d_shader_component_type component_type, unsigned int component_count, const uint64_t *values) + enum vsir_data_type data_type, unsigned int component_count, const uint64_t *values) { uint32_t type_id, scalar_type_id, component_ids[VKD3D_DVEC2_SIZE]; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; unsigned int i; VKD3D_ASSERT(0 < component_count && component_count <= VKD3D_DVEC2_SIZE); - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id(builder, data_type, component_count); - if (component_type != VKD3D_SHADER_COMPONENT_DOUBLE && component_type != VKD3D_SHADER_COMPONENT_UINT64) + if (data_type != VSIR_DATA_F64 && data_type != VSIR_DATA_I64 && data_type != VSIR_DATA_U64) { - FIXME("Unhandled component_type %#x.\n", component_type); + spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_TYPE, + "Unhandled data type \"%s\" (%#x).", + vsir_data_type_get_name(data_type, ""), data_type); return vkd3d_spirv_get_op_undef(builder, type_id); } if (component_count == 1) - { return vkd3d_spirv_get_op_constant64(builder, type_id, *values); - } - else + + scalar_type_id = spirv_get_type_id(builder, data_type, 1); + for (i = 0; i < component_count; ++i) { - scalar_type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); - for (i = 0; i < component_count; ++i) - component_ids[i] = vkd3d_spirv_get_op_constant64(builder, scalar_type_id, values[i]); - return vkd3d_spirv_get_op_constant_composite(builder, type_id, component_ids, component_count); + component_ids[i] = vkd3d_spirv_get_op_constant64(builder, scalar_type_id, values[i]); } + + return vkd3d_spirv_get_op_constant_composite(builder, type_id, component_ids, component_count); } -static uint32_t spirv_compiler_get_constant_uint(struct spirv_compiler *compiler, - uint32_t value) +static uint32_t spirv_compiler_get_constant_uint(struct spirv_compiler *compiler, uint32_t value) { - return spirv_compiler_get_constant(compiler, VKD3D_SHADER_COMPONENT_UINT, 1, &value); + return spirv_compiler_get_constant(compiler, VSIR_DATA_U32, 1, &value); } -static uint32_t spirv_compiler_get_constant_float(struct spirv_compiler *compiler, - float value) +static uint32_t spirv_compiler_get_constant_float(struct spirv_compiler *compiler, float value) { - return spirv_compiler_get_constant(compiler, VKD3D_SHADER_COMPONENT_FLOAT, 1, (uint32_t *)&value); + return spirv_compiler_get_constant(compiler, VSIR_DATA_F32, 1, (uint32_t *)&value); } static uint32_t spirv_compiler_get_constant_vector(struct spirv_compiler *compiler, - enum vkd3d_shader_component_type component_type, unsigned int component_count, uint32_t value) + enum vsir_data_type data_type, unsigned int component_count, uint32_t value) { const uint32_t values[] = {value, value, value, value}; - return spirv_compiler_get_constant(compiler, component_type, component_count, values); + + return spirv_compiler_get_constant(compiler, data_type, component_count, values); } static uint32_t spirv_compiler_get_constant_uint_vector(struct spirv_compiler *compiler, uint32_t value, unsigned int component_count) { - return spirv_compiler_get_constant_vector(compiler, VKD3D_SHADER_COMPONENT_UINT, component_count, value); + return spirv_compiler_get_constant_vector(compiler, VSIR_DATA_U32, component_count, value); } static uint32_t spirv_compiler_get_constant_float_vector(struct spirv_compiler *compiler, float value, unsigned int component_count) { const float values[] = {value, value, value, value}; - return spirv_compiler_get_constant(compiler, VKD3D_SHADER_COMPONENT_FLOAT, - component_count, (const uint32_t *)values); + + return spirv_compiler_get_constant(compiler, VSIR_DATA_F32, component_count, (const uint32_t *)values); } static uint32_t spirv_compiler_get_constant_double_vector(struct spirv_compiler *compiler, double value, unsigned int component_count) { const double values[] = {value, value}; - return spirv_compiler_get_constant64(compiler, VKD3D_SHADER_COMPONENT_DOUBLE, - component_count, (const uint64_t *)values); + + return spirv_compiler_get_constant64(compiler, VSIR_DATA_F64, component_count, (const uint64_t *)values); } static uint32_t spirv_compiler_get_constant_uint64_vector(struct spirv_compiler *compiler, uint64_t value, unsigned int component_count) { const uint64_t values[] = {value, value}; - return spirv_compiler_get_constant64(compiler, VKD3D_SHADER_COMPONENT_UINT64, component_count, values); + + return spirv_compiler_get_constant64(compiler, VSIR_DATA_U64, component_count, values); } static uint32_t spirv_compiler_get_type_id_for_reg(struct spirv_compiler *compiler, @@ -3772,9 +3786,7 @@ static uint32_t spirv_compiler_get_type_id_for_reg(struct spirv_compiler *compil { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - return vkd3d_spirv_get_type_id(builder, - vkd3d_component_type_from_data_type(reg->data_type), - vsir_write_mask_component_count(write_mask)); + return spirv_get_type_id(builder, reg->data_type, vsir_write_mask_component_count(write_mask)); } static uint32_t spirv_compiler_get_type_id_for_dst(struct spirv_compiler *compiler, @@ -3901,7 +3913,7 @@ static uint32_t spirv_compiler_emit_array_variable(struct spirv_compiler *compil uint32_t type_id, length_id, ptr_type_id; unsigned int i; - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id_for_component_type(builder, component_type, component_count); for (i = 0; i < length_count; ++i) { if (!array_lengths[i]) @@ -4000,8 +4012,8 @@ static uint32_t spirv_compiler_emit_spec_constant(struct spirv_compiler *compile info = get_spec_constant_info(name); default_value = info ? info->default_value.u : 0; - scalar_type_id = vkd3d_spirv_get_type_id(builder, vkd3d_component_type_from_data_type(type), 1); - vector_type_id = vkd3d_spirv_get_type_id(builder, vkd3d_component_type_from_data_type(type), component_count); + scalar_type_id = spirv_get_type_id(builder, type, 1); + vector_type_id = spirv_get_type_id(builder, type, component_count); for (unsigned int i = 0; i < component_count; ++i) { @@ -4050,7 +4062,7 @@ static uint32_t spirv_compiler_get_buffer_parameter(struct spirv_compiler *compi unsigned int index = parameter - compiler->program->parameters; uint32_t type_id, ptr_id, ptr_type_id; - type_id = vkd3d_spirv_get_type_id(builder, vkd3d_component_type_from_data_type(type), component_count); + type_id = spirv_get_type_id(builder, type, component_count); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, type_id); ptr_id = vkd3d_spirv_build_op_access_chain1(builder, ptr_type_id, compiler->spirv_parameter_info[index].buffer_id, @@ -4087,8 +4099,8 @@ static uint32_t spirv_compiler_emit_shader_parameter(struct spirv_compiler *comp type, component_count, name, parameter->data_type); if (parameter->type == VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) - return spirv_compiler_get_constant(compiler, vkd3d_component_type_from_data_type(type), - component_count, (const uint32_t *)¶meter->u.immediate_constant); + return spirv_compiler_get_constant(compiler, type, component_count, + (const uint32_t *)¶meter->u.immediate_constant); if (parameter->type == VKD3D_SHADER_PARAMETER_TYPE_SPECIALIZATION_CONSTANT) return spirv_compiler_get_spec_constant(compiler, name, @@ -4104,32 +4116,32 @@ default_parameter: } static uint32_t spirv_compiler_emit_construct_vector(struct spirv_compiler *compiler, - enum vkd3d_shader_component_type component_type, unsigned int component_count, - uint32_t val_id, unsigned int val_component_idx, unsigned int val_component_count) + enum vsir_data_type data_type, unsigned int component_count, uint32_t val_id, + unsigned int val_component_idx, unsigned int val_component_count) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t components[VKD3D_VEC4_SIZE]; - uint32_t type_id, result_id; + uint32_t type_id; unsigned int i; VKD3D_ASSERT(val_component_idx < val_component_count); - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id(builder, data_type, component_count); if (val_component_count == 1) { for (i = 0; i < component_count; ++i) + { components[i] = val_id; - result_id = vkd3d_spirv_build_op_composite_construct(builder, - type_id, components, component_count); + } + return vkd3d_spirv_build_op_composite_construct(builder, type_id, components, component_count); } - else + + for (i = 0; i < component_count; ++i) { - for (i = 0; i < component_count; ++i) - components[i] = val_component_idx; - result_id = vkd3d_spirv_build_op_vector_shuffle(builder, - type_id, val_id, val_id, components, component_count); + components[i] = val_component_idx; } - return result_id; + + return vkd3d_spirv_build_op_vector_shuffle(builder, type_id, val_id, val_id, components, component_count); } static uint32_t spirv_compiler_emit_load_src(struct spirv_compiler *compiler, @@ -4147,10 +4159,11 @@ static uint32_t spirv_compiler_emit_register_addressing(struct spirv_compiler *c addr_id = spirv_compiler_emit_load_src(compiler, reg_index->rel_addr, VKD3DSP_WRITEMASK_0); if (reg_index->offset) { - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); - addr_id = vkd3d_spirv_build_op_iadd(builder, type_id, - addr_id, spirv_compiler_get_constant_uint(compiler, reg_index->offset)); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); + addr_id = vkd3d_spirv_build_op_iadd(builder, type_id, addr_id, + spirv_compiler_get_constant_uint(compiler, reg_index->offset)); } + return addr_id; } @@ -4159,7 +4172,7 @@ struct vkd3d_shader_register_info uint32_t id; const struct vkd3d_symbol *descriptor_array; SpvStorageClass storage_class; - enum vkd3d_shader_component_type component_type; + enum vsir_data_type data_type; unsigned int write_mask; uint32_t member_idx; unsigned int structure_stride; @@ -4182,7 +4195,7 @@ static bool spirv_compiler_get_register_info(struct spirv_compiler *compiler, register_info->storage_class = SpvStorageClassPrivate; register_info->descriptor_array = NULL; register_info->member_idx = 0; - register_info->component_type = VKD3D_SHADER_COMPONENT_FLOAT; + register_info->data_type = VSIR_DATA_F32; register_info->write_mask = VKD3DSP_WRITEMASK_ALL; register_info->structure_stride = 0; register_info->binding_base_idx = 0; @@ -4203,7 +4216,7 @@ static bool spirv_compiler_get_register_info(struct spirv_compiler *compiler, memset(register_info, 0, sizeof(*register_info)); register_info->id = ssa->id; register_info->storage_class = SpvStorageClassMax; - register_info->component_type = vkd3d_component_type_from_data_type(ssa->data_type); + register_info->data_type = ssa->data_type; register_info->write_mask = ssa->write_mask; return true; } @@ -4222,7 +4235,7 @@ static bool spirv_compiler_get_register_info(struct spirv_compiler *compiler, register_info->descriptor_array = symbol->descriptor_array; register_info->storage_class = symbol->info.reg.storage_class; register_info->member_idx = symbol->info.reg.member_idx; - register_info->component_type = symbol->info.reg.component_type; + register_info->data_type = symbol->info.reg.data_type; register_info->write_mask = symbol->info.reg.write_mask; register_info->structure_stride = symbol->info.reg.structure_stride; register_info->binding_base_idx = symbol->info.reg.binding_base_idx; @@ -4296,7 +4309,7 @@ static uint32_t spirv_compiler_get_descriptor_index(struct spirv_compiler *compi struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t type_id, ptr_type_id, ptr_id, offset_id, index_ids[2]; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); if (!(offset_id = compiler->descriptor_offset_ids[push_constant_index])) { index_ids[0] = compiler->descriptor_offsets_member_id; @@ -4369,7 +4382,7 @@ static void spirv_compiler_emit_dereference_register(struct spirv_compiler *comp if (index_count) { component_count = vsir_write_mask_component_count(register_info->write_mask); - type_id = vkd3d_spirv_get_type_id(builder, register_info->component_type, component_count); + type_id = spirv_get_type_id(builder, register_info->data_type, component_count); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, register_info->storage_class, type_id); register_info->id = vkd3d_spirv_build_op_access_chain(builder, ptr_type_id, register_info->id, indexes, index_count); @@ -4428,7 +4441,7 @@ static uint32_t spirv_compiler_emit_swizzle(struct spirv_compiler *compiler, && (component_count == 1 || vkd3d_swizzle_is_equal(val_write_mask, swizzle, write_mask))) return val_id; - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id_for_component_type(builder, component_type, component_count); if (component_count == 1) { @@ -4479,7 +4492,7 @@ static uint32_t spirv_compiler_emit_vector_shuffle(struct spirv_compiler *compil components[i] = VKD3D_VEC4_SIZE + vsir_swizzle_get_component(swizzle, i); } - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id_for_component_type(builder, component_type, component_count); return vkd3d_spirv_build_op_vector_shuffle(builder, type_id, vector1_id, vector2_id, components, component_count); } @@ -4494,10 +4507,11 @@ static uint32_t spirv_compiler_emit_int_to_bool(struct spirv_compiler *compiler, VKD3D_ASSERT(!(condition & ~(VKD3D_SHADER_CONDITIONAL_OP_NZ | VKD3D_SHADER_CONDITIONAL_OP_Z))); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count); op = condition & VKD3D_SHADER_CONDITIONAL_OP_Z ? SpvOpIEqual : SpvOpINotEqual; + return vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, op, type_id, val_id, - data_type == VSIR_DATA_U64 + data_type_is_64_bit(data_type) ? spirv_compiler_get_constant_uint64_vector(compiler, 0, component_count) : spirv_compiler_get_constant_uint_vector(compiler, 0, component_count)); } @@ -4510,7 +4524,8 @@ static uint32_t spirv_compiler_emit_bool_to_int(struct spirv_compiler *compiler, true_id = spirv_compiler_get_constant_uint_vector(compiler, signedness ? 0xffffffff : 1, component_count); false_id = spirv_compiler_get_constant_uint_vector(compiler, 0, component_count); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, component_count); + return vkd3d_spirv_build_op_select(builder, type_id, val_id, true_id, false_id); } @@ -4523,7 +4538,8 @@ static uint32_t spirv_compiler_emit_bool_to_int64(struct spirv_compiler *compile true_id = spirv_compiler_get_constant_uint64_vector(compiler, signedness ? UINT64_MAX : 1, component_count); false_id = spirv_compiler_get_constant_uint64_vector(compiler, 0, component_count); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT64, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_U64, component_count); + return vkd3d_spirv_build_op_select(builder, type_id, val_id, true_id, false_id); } @@ -4535,7 +4551,8 @@ static uint32_t spirv_compiler_emit_bool_to_float(struct spirv_compiler *compile true_id = spirv_compiler_get_constant_float_vector(compiler, signedness ? -1.0f : 1.0f, component_count); false_id = spirv_compiler_get_constant_float_vector(compiler, 0.0f, component_count); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, component_count); + return vkd3d_spirv_build_op_select(builder, type_id, val_id, true_id, false_id); } @@ -4547,7 +4564,8 @@ static uint32_t spirv_compiler_emit_bool_to_double(struct spirv_compiler *compil true_id = spirv_compiler_get_constant_double_vector(compiler, signedness ? -1.0 : 1.0, component_count); false_id = spirv_compiler_get_constant_double_vector(compiler, 0.0, component_count); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_DOUBLE, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_F64, component_count); + return vkd3d_spirv_build_op_select(builder, type_id, val_id, true_id, false_id); } @@ -4574,8 +4592,7 @@ static uint32_t spirv_compiler_emit_load_constant(struct spirv_compiler *compile } } - return spirv_compiler_get_constant(compiler, - vkd3d_component_type_from_data_type(reg->data_type), component_count, values); + return spirv_compiler_get_constant(compiler, reg->data_type, component_count, values); } static uint32_t spirv_compiler_emit_load_constant64(struct spirv_compiler *compiler, @@ -4601,8 +4618,7 @@ static uint32_t spirv_compiler_emit_load_constant64(struct spirv_compiler *compi } } - return spirv_compiler_get_constant64(compiler, - vkd3d_component_type_from_data_type(reg->data_type), component_count, values); + return spirv_compiler_get_constant64(compiler, reg->data_type, component_count, values); } static uint32_t spirv_compiler_emit_load_undef(struct spirv_compiler *compiler, @@ -4614,18 +4630,18 @@ static uint32_t spirv_compiler_emit_load_undef(struct spirv_compiler *compiler, VKD3D_ASSERT(reg->type == VKD3DSPR_UNDEF); - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, reg->data_type, component_count); + type_id = spirv_get_type_id(builder, reg->data_type, component_count); + return vkd3d_spirv_get_op_undef(builder, type_id); } static uint32_t spirv_compiler_emit_load_scalar(struct spirv_compiler *compiler, - const struct vkd3d_shader_register *reg, uint32_t swizzle, uint32_t write_mask, - const struct vkd3d_shader_register_info *reg_info) + const struct vkd3d_shader_register *reg, uint32_t swizzle, + uint32_t write_mask, const struct vkd3d_shader_register_info *reg_info) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t type_id, ptr_type_id, index, reg_id, val_id; unsigned int component_idx, reg_component_count; - enum vkd3d_shader_component_type component_type; uint32_t skipped_component_mask; VKD3D_ASSERT(!register_is_constant_or_undef(reg)); @@ -4636,7 +4652,6 @@ static uint32_t spirv_compiler_emit_load_scalar(struct spirv_compiler *compiler, skipped_component_mask = ~reg_info->write_mask & ((VKD3DSP_WRITEMASK_0 << component_idx) - 1); if (skipped_component_mask) component_idx -= vsir_write_mask_component_count(skipped_component_mask); - component_type = vkd3d_component_type_from_data_type(reg->data_type); reg_component_count = vsir_write_mask_component_count(reg_info->write_mask); @@ -4646,7 +4661,7 @@ static uint32_t spirv_compiler_emit_load_scalar(struct spirv_compiler *compiler, component_idx, reg->type, reg->idx[0].offset, reg_info->write_mask); } - type_id = vkd3d_spirv_get_type_id(builder, reg_info->component_type, 1); + type_id = spirv_get_type_id(builder, reg_info->data_type, 1); reg_id = reg_info->id; if (reg_component_count != 1) { @@ -4656,27 +4671,22 @@ static uint32_t spirv_compiler_emit_load_scalar(struct spirv_compiler *compiler, } val_id = vkd3d_spirv_build_op_load(builder, type_id, reg_id, SpvMemoryAccessMaskNone); + if (reg->data_type == reg_info->data_type) + return val_id; - if (component_type != reg_info->component_type) + if (reg->data_type != VSIR_DATA_BOOL) { - if (component_type == VKD3D_SHADER_COMPONENT_BOOL) - { - if (reg_info->component_type != VKD3D_SHADER_COMPONENT_UINT) - { - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); - val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - } - val_id = spirv_compiler_emit_int_to_bool(compiler, - VKD3D_SHADER_CONDITIONAL_OP_NZ, VSIR_DATA_U32, 1, val_id); - } - else - { - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); - val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - } + type_id = spirv_get_type_id(builder, reg->data_type, 1); + return vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); } - return val_id; + if (reg_info->data_type != VSIR_DATA_U32) + { + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); + val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); + } + + return spirv_compiler_emit_int_to_bool(compiler, VKD3D_SHADER_CONDITIONAL_OP_NZ, VSIR_DATA_U32, 1, val_id); } static uint32_t spirv_compiler_emit_constant_array(struct spirv_compiler *compiler, @@ -4684,14 +4694,12 @@ static uint32_t spirv_compiler_emit_constant_array(struct spirv_compiler *compil { uint32_t *elements, elem_type_id, length_id, type_id, const_id; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - enum vkd3d_shader_component_type component_type; unsigned int i, element_count, component_count; element_count = icb->element_count; - component_type = vkd3d_component_type_from_data_type(icb->data_type); component_count = icb->component_count; - elem_type_id = vkd3d_spirv_get_type_id_for_data_type(builder, icb->data_type, component_count); + elem_type_id = spirv_get_type_id(builder, icb->data_type, component_count); length_id = spirv_compiler_get_constant_uint(compiler, element_count); type_id = vkd3d_spirv_get_op_type_array(builder, elem_type_id, length_id); @@ -4706,7 +4714,6 @@ static uint32_t spirv_compiler_emit_constant_array(struct spirv_compiler *compil if (!(elements = vkd3d_calloc(element_count, sizeof(*elements)))) { - ERR("Failed to allocate %u elements.", element_count); spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_OUT_OF_MEMORY, "Failed to allocate %u constant array elements.", element_count); return 0; @@ -4718,22 +4725,27 @@ static uint32_t spirv_compiler_emit_constant_array(struct spirv_compiler *compil case VSIR_DATA_I32: case VSIR_DATA_U32: for (i = 0; i < element_count; ++i) - elements[i] = spirv_compiler_get_constant(compiler, component_type, component_count, - &icb->data[component_count * i]); + { + elements[i] = spirv_compiler_get_constant(compiler, icb->data_type, + component_count, &icb->data[component_count * i]); + } break; + case VSIR_DATA_F64: + case VSIR_DATA_I64: case VSIR_DATA_U64: { uint64_t *data = (uint64_t *)icb->data; for (i = 0; i < element_count; ++i) - elements[i] = spirv_compiler_get_constant64(compiler, component_type, component_count, - &data[component_count * i]); + elements[i] = spirv_compiler_get_constant64(compiler, icb->data_type, + component_count, &data[component_count * i]); break; } + default: - FIXME("Unhandled data type %u.\n", icb->data_type); spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_TYPE, - "Immediate constant buffer data type %u is unhandled.", icb->data_type); + "Immediate constant buffer data type \"%s\" (%#x) is unhandled.", + vsir_data_type_get_name(icb->data_type, ""), icb->data_type); break; } @@ -4756,7 +4768,6 @@ static uint32_t spirv_compiler_emit_load_reg(struct spirv_compiler *compiler, const struct vkd3d_shader_register *reg, uint32_t swizzle, uint32_t write_mask) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - enum vkd3d_shader_component_type component_type; struct vkd3d_shader_register_info reg_info; unsigned int component_count; uint32_t type_id, val_id; @@ -4773,16 +4784,15 @@ static uint32_t spirv_compiler_emit_load_reg(struct spirv_compiler *compiler, reg->data_type, reg->dimension == VSIR_DIMENSION_VEC4 ? 4 : 1); component_count = vsir_write_mask_component_count(write_mask); - component_type = vkd3d_component_type_from_data_type(reg->data_type); if (!spirv_compiler_get_register_info(compiler, reg, ®_info)) { - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id(builder, reg->data_type, component_count); return vkd3d_spirv_get_op_undef(builder, type_id); } spirv_compiler_emit_dereference_register(compiler, reg, ®_info); - val_write_mask = (data_type_is_64_bit(reg->data_type) && !component_type_is_64_bit(reg_info.component_type)) + val_write_mask = (data_type_is_64_bit(reg->data_type) && !data_type_is_64_bit(reg_info.data_type)) ? vsir_write_mask_32_from_64(write_mask) : write_mask; /* Intermediate value (no storage class). */ @@ -4796,35 +4806,31 @@ static uint32_t spirv_compiler_emit_load_reg(struct spirv_compiler *compiler, } else { - type_id = vkd3d_spirv_get_type_id(builder, - reg_info.component_type, vsir_write_mask_component_count(reg_info.write_mask)); + type_id = spirv_get_type_id(builder, reg_info.data_type, + vsir_write_mask_component_count(reg_info.write_mask)); val_id = vkd3d_spirv_build_op_load(builder, type_id, reg_info.id, SpvMemoryAccessMaskNone); swizzle = data_type_is_64_bit(reg->data_type) ? vsir_swizzle_32_from_64(swizzle) : swizzle; } - val_id = spirv_compiler_emit_swizzle(compiler, - val_id, reg_info.write_mask, reg_info.component_type, swizzle, val_write_mask); + val_id = spirv_compiler_emit_swizzle(compiler, val_id, reg_info.write_mask, + vkd3d_component_type_from_data_type(reg_info.data_type), swizzle, val_write_mask); + if (reg->data_type == reg_info.data_type) + return val_id; - if (component_type != reg_info.component_type) + if (reg->data_type != VSIR_DATA_BOOL) { - if (component_type == VKD3D_SHADER_COMPONENT_BOOL) - { - if (reg_info.component_type != VKD3D_SHADER_COMPONENT_UINT) - { - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, component_count); - val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - } - val_id = spirv_compiler_emit_int_to_bool(compiler, - VKD3D_SHADER_CONDITIONAL_OP_NZ, VSIR_DATA_U32, component_count, val_id); - } - else - { - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); - val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - } + type_id = spirv_get_type_id(builder, reg->data_type, component_count); + return vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); } - return val_id; + if (reg_info.data_type != VSIR_DATA_U32) + { + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, component_count); + val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); + } + + return spirv_compiler_emit_int_to_bool(compiler, VKD3D_SHADER_CONDITIONAL_OP_NZ, + VSIR_DATA_U32, component_count, val_id); } static void spirv_compiler_emit_execution_mode(struct spirv_compiler *compiler, @@ -4923,7 +4929,7 @@ static void spirv_compiler_emit_store_scalar(struct spirv_compiler *compiler, if (vsir_write_mask_component_count(dst_write_mask) > 1) { - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); + type_id = spirv_get_type_id_for_component_type(builder, component_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, storage_class, type_id); component_idx = vsir_write_mask_get_component_idx(write_mask); component_idx -= vsir_write_mask_get_component_idx(dst_write_mask); @@ -4951,7 +4957,7 @@ static void spirv_compiler_emit_store(struct spirv_compiler *compiler, if (dst_component_count == 1 && component_count != 1) { - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); + type_id = spirv_get_type_id_for_component_type(builder, component_type, 1); val_id = vkd3d_spirv_build_op_composite_extract1(builder, type_id, val_id, vsir_write_mask_get_component_idx(dst_write_mask)); write_mask &= dst_write_mask; @@ -4966,7 +4972,7 @@ static void spirv_compiler_emit_store(struct spirv_compiler *compiler, if (dst_component_count != component_count) { - type_id = vkd3d_spirv_get_type_id(builder, component_type, dst_component_count); + type_id = spirv_get_type_id_for_component_type(builder, component_type, dst_component_count); dst_val_id = vkd3d_spirv_build_op_load(builder, type_id, dst_id, SpvMemoryAccessMaskNone); VKD3D_ASSERT(component_count <= ARRAY_SIZE(components)); @@ -4993,9 +4999,9 @@ static void spirv_compiler_emit_store_reg(struct spirv_compiler *compiler, const struct vkd3d_shader_register *reg, uint32_t write_mask, uint32_t val_id) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - enum vkd3d_shader_component_type component_type; struct vkd3d_shader_register_info reg_info; uint32_t src_write_mask = write_mask; + enum vsir_data_type data_type; uint32_t type_id; VKD3D_ASSERT(!register_is_constant_or_undef(reg)); @@ -5010,22 +5016,22 @@ static void spirv_compiler_emit_store_reg(struct spirv_compiler *compiler, return; spirv_compiler_emit_dereference_register(compiler, reg, ®_info); - component_type = vkd3d_component_type_from_data_type(reg->data_type); - if (component_type != reg_info.component_type) + data_type = reg->data_type; + if (data_type != reg_info.data_type) { - if (data_type_is_64_bit(reg->data_type)) + if (data_type_is_64_bit(data_type)) src_write_mask = vsir_write_mask_32_from_64(write_mask); - if (component_type == VKD3D_SHADER_COMPONENT_BOOL) + if (data_type == VSIR_DATA_BOOL) val_id = spirv_compiler_emit_bool_to_int(compiler, vsir_write_mask_component_count(src_write_mask), val_id, false); - type_id = vkd3d_spirv_get_type_id(builder, reg_info.component_type, + type_id = spirv_get_type_id(builder, reg_info.data_type, vsir_write_mask_component_count(src_write_mask)); val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - component_type = reg_info.component_type; + data_type = reg_info.data_type; } - spirv_compiler_emit_store(compiler, - reg_info.id, reg_info.write_mask, component_type, reg_info.storage_class, src_write_mask, val_id); + spirv_compiler_emit_store(compiler, reg_info.id, reg_info.write_mask, + vkd3d_component_type_from_data_type(data_type), reg_info.storage_class, src_write_mask, val_id); } static uint32_t spirv_compiler_emit_sat(struct spirv_compiler *compiler, @@ -5101,7 +5107,7 @@ static void spirv_compiler_emit_store_dst_components(struct spirv_compiler *comp struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t type_id, dst_type_id, val_id; - type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + type_id = spirv_get_type_id_for_component_type(builder, component_type, component_count); if (component_count > 1) { val_id = vkd3d_spirv_build_op_composite_construct(builder, @@ -5112,7 +5118,7 @@ static void spirv_compiler_emit_store_dst_components(struct spirv_compiler *comp val_id = *component_ids; } - dst_type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, component_count); + dst_type_id = spirv_get_type_id(builder, dst->reg.data_type, component_count); if (dst_type_id != type_id) val_id = vkd3d_spirv_build_op_bitcast(builder, dst_type_id, val_id); @@ -5281,9 +5287,8 @@ static uint32_t spirv_compiler_emit_draw_parameter_fixup(struct spirv_compiler * vkd3d_spirv_add_iface_variable(builder, base_var_id); spirv_compiler_decorate_builtin(compiler, base_var_id, base); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_INT, 1); - base_id = vkd3d_spirv_build_op_load(builder, - type_id, base_var_id, SpvMemoryAccessMaskNone); + type_id = spirv_get_type_id(builder, VSIR_DATA_I32, 1); + base_id = vkd3d_spirv_build_op_load(builder, type_id, base_var_id, SpvMemoryAccessMaskNone); return vkd3d_spirv_build_op_isub(builder, type_id, index_id, base_id); } @@ -5311,23 +5316,22 @@ static uint32_t sv_front_face_fixup(struct spirv_compiler *compiler, } /* frag_coord.w = 1.0f / frag_coord.w */ -static uint32_t frag_coord_fixup(struct spirv_compiler *compiler, - uint32_t frag_coord_id) +static uint32_t frag_coord_fixup(struct spirv_compiler *compiler, uint32_t frag_coord_id) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t type_id, w_id; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 1); w_id = vkd3d_spirv_build_op_composite_extract1(builder, type_id, frag_coord_id, 3); - w_id = vkd3d_spirv_build_op_fdiv(builder, type_id, - spirv_compiler_get_constant_float(compiler, 1.0f), w_id); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, VKD3D_VEC4_SIZE); + w_id = vkd3d_spirv_build_op_fdiv(builder, type_id, spirv_compiler_get_constant_float(compiler, 1.0f), w_id); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, VKD3D_VEC4_SIZE); + return vkd3d_spirv_build_op_composite_insert1(builder, type_id, w_id, frag_coord_id, 3); } struct vkd3d_spirv_builtin { - enum vkd3d_shader_component_type component_type; + enum vsir_data_type data_type; unsigned int component_count; SpvBuiltIn spirv_builtin; vkd3d_spirv_builtin_fixup_pfn fixup_pfn; @@ -5347,43 +5351,43 @@ static const struct } vkd3d_system_value_builtins[] = { - {VKD3D_SHADER_SV_VERTEX_ID, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInVertexId}, + {VKD3D_SHADER_SV_VERTEX_ID, {VSIR_DATA_I32, 1, SpvBuiltInVertexId}, VKD3D_SHADER_SPIRV_ENVIRONMENT_OPENGL_4_5}, - {VKD3D_SHADER_SV_INSTANCE_ID, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInInstanceId}, + {VKD3D_SHADER_SV_INSTANCE_ID, {VSIR_DATA_I32, 1, SpvBuiltInInstanceId}, VKD3D_SHADER_SPIRV_ENVIRONMENT_OPENGL_4_5}, - {VKD3D_SHADER_SV_POSITION, {VKD3D_SHADER_COMPONENT_FLOAT, 4, SpvBuiltInPosition}}, - {VKD3D_SHADER_SV_VERTEX_ID, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInVertexIndex, sv_vertex_id_fixup}}, - {VKD3D_SHADER_SV_INSTANCE_ID, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInInstanceIndex, sv_instance_id_fixup}}, + {VKD3D_SHADER_SV_POSITION, {VSIR_DATA_F32, 4, SpvBuiltInPosition}}, + {VKD3D_SHADER_SV_VERTEX_ID, {VSIR_DATA_I32, 1, SpvBuiltInVertexIndex, sv_vertex_id_fixup}}, + {VKD3D_SHADER_SV_INSTANCE_ID, {VSIR_DATA_I32, 1, SpvBuiltInInstanceIndex, sv_instance_id_fixup}}, - {VKD3D_SHADER_SV_PRIMITIVE_ID, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInPrimitiveId}}, + {VKD3D_SHADER_SV_PRIMITIVE_ID, {VSIR_DATA_I32, 1, SpvBuiltInPrimitiveId}}, - {VKD3D_SHADER_SV_RENDER_TARGET_ARRAY_INDEX, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInLayer}}, - {VKD3D_SHADER_SV_VIEWPORT_ARRAY_INDEX, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInViewportIndex}}, + {VKD3D_SHADER_SV_RENDER_TARGET_ARRAY_INDEX, {VSIR_DATA_I32, 1, SpvBuiltInLayer}}, + {VKD3D_SHADER_SV_VIEWPORT_ARRAY_INDEX, {VSIR_DATA_I32, 1, SpvBuiltInViewportIndex}}, - {VKD3D_SHADER_SV_IS_FRONT_FACE, {VKD3D_SHADER_COMPONENT_BOOL, 1, SpvBuiltInFrontFacing, sv_front_face_fixup}}, + {VKD3D_SHADER_SV_IS_FRONT_FACE, {VSIR_DATA_BOOL, 1, SpvBuiltInFrontFacing, sv_front_face_fixup}}, - {VKD3D_SHADER_SV_SAMPLE_INDEX, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInSampleId}}, + {VKD3D_SHADER_SV_SAMPLE_INDEX, {VSIR_DATA_U32, 1, SpvBuiltInSampleId}}, - {VKD3D_SHADER_SV_CLIP_DISTANCE, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInClipDistance, NULL, 1}}, - {VKD3D_SHADER_SV_CULL_DISTANCE, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInCullDistance, NULL, 1}}, + {VKD3D_SHADER_SV_CLIP_DISTANCE, {VSIR_DATA_F32, 1, SpvBuiltInClipDistance, NULL, 1}}, + {VKD3D_SHADER_SV_CULL_DISTANCE, {VSIR_DATA_F32, 1, SpvBuiltInCullDistance, NULL, 1}}, - {VKD3D_SHADER_SV_TESS_FACTOR_QUADEDGE, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelOuter, NULL, 4}}, - {VKD3D_SHADER_SV_TESS_FACTOR_QUADINT, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelInner, NULL, 2}}, + {VKD3D_SHADER_SV_TESS_FACTOR_QUADEDGE, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelOuter, NULL, 4}}, + {VKD3D_SHADER_SV_TESS_FACTOR_QUADINT, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelInner, NULL, 2}}, - {VKD3D_SHADER_SV_TESS_FACTOR_TRIEDGE, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelOuter, NULL, 4}}, - {VKD3D_SHADER_SV_TESS_FACTOR_TRIINT, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelInner, NULL, 2}}, + {VKD3D_SHADER_SV_TESS_FACTOR_TRIEDGE, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelOuter, NULL, 4}}, + {VKD3D_SHADER_SV_TESS_FACTOR_TRIINT, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelInner, NULL, 2}}, - {VKD3D_SHADER_SV_TESS_FACTOR_LINEDEN, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelOuter, NULL, 4, 0}}, - {VKD3D_SHADER_SV_TESS_FACTOR_LINEDET, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInTessLevelOuter, NULL, 4, 1}}, + {VKD3D_SHADER_SV_TESS_FACTOR_LINEDEN, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelOuter, NULL, 4, 0}}, + {VKD3D_SHADER_SV_TESS_FACTOR_LINEDET, {VSIR_DATA_F32, 1, SpvBuiltInTessLevelOuter, NULL, 4, 1}}, }; static const struct vkd3d_spirv_builtin vkd3d_pixel_shader_position_builtin = { - VKD3D_SHADER_COMPONENT_FLOAT, 4, SpvBuiltInFragCoord, frag_coord_fixup, + VSIR_DATA_F32, 4, SpvBuiltInFragCoord, frag_coord_fixup, }; static const struct vkd3d_spirv_builtin vkd3d_output_point_size_builtin = { - VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInPointSize, + VSIR_DATA_F32, 1, SpvBuiltInPointSize, }; static const struct { @@ -5393,31 +5397,31 @@ static const struct } vkd3d_register_builtins[] = { - {VKD3DSPR_THREADID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 3, SpvBuiltInGlobalInvocationId}}, - {VKD3DSPR_LOCALTHREADID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 3, SpvBuiltInLocalInvocationId}}, - {VKD3DSPR_LOCALTHREADINDEX, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInLocalInvocationIndex}}, - {VKD3DSPR_THREADGROUPID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 3, SpvBuiltInWorkgroupId}}, + {VKD3DSPR_THREADID, SpvStorageClassInput, {VSIR_DATA_I32, 3, SpvBuiltInGlobalInvocationId}}, + {VKD3DSPR_LOCALTHREADID, SpvStorageClassInput, {VSIR_DATA_I32, 3, SpvBuiltInLocalInvocationId}}, + {VKD3DSPR_LOCALTHREADINDEX, SpvStorageClassInput, {VSIR_DATA_I32, 1, SpvBuiltInLocalInvocationIndex}}, + {VKD3DSPR_THREADGROUPID, SpvStorageClassInput, {VSIR_DATA_I32, 3, SpvBuiltInWorkgroupId}}, - {VKD3DSPR_GSINSTID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInInvocationId}}, - {VKD3DSPR_OUTPOINTID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInInvocationId}}, + {VKD3DSPR_GSINSTID, SpvStorageClassInput, {VSIR_DATA_I32, 1, SpvBuiltInInvocationId}}, + {VKD3DSPR_OUTPOINTID, SpvStorageClassInput, {VSIR_DATA_I32, 1, SpvBuiltInInvocationId}}, - {VKD3DSPR_PRIMID, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_INT, 1, SpvBuiltInPrimitiveId}}, + {VKD3DSPR_PRIMID, SpvStorageClassInput, {VSIR_DATA_I32, 1, SpvBuiltInPrimitiveId}}, - {VKD3DSPR_TESSCOORD, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_FLOAT, 3, SpvBuiltInTessCoord}}, + {VKD3DSPR_TESSCOORD, SpvStorageClassInput, {VSIR_DATA_F32, 3, SpvBuiltInTessCoord}}, - {VKD3DSPR_POINT_COORD, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_FLOAT, 2, SpvBuiltInPointCoord}}, + {VKD3DSPR_POINT_COORD, SpvStorageClassInput, {VSIR_DATA_F32, 2, SpvBuiltInPointCoord}}, - {VKD3DSPR_COVERAGE, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInSampleMask, NULL, 1}}, - {VKD3DSPR_SAMPLEMASK, SpvStorageClassOutput, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInSampleMask, NULL, 1}}, + {VKD3DSPR_COVERAGE, SpvStorageClassInput, {VSIR_DATA_U32, 1, SpvBuiltInSampleMask, NULL, 1}}, + {VKD3DSPR_SAMPLEMASK, SpvStorageClassOutput, {VSIR_DATA_U32, 1, SpvBuiltInSampleMask, NULL, 1}}, - {VKD3DSPR_DEPTHOUT, SpvStorageClassOutput, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInFragDepth}}, - {VKD3DSPR_DEPTHOUTGE, SpvStorageClassOutput, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInFragDepth}}, - {VKD3DSPR_DEPTHOUTLE, SpvStorageClassOutput, {VKD3D_SHADER_COMPONENT_FLOAT, 1, SpvBuiltInFragDepth}}, + {VKD3DSPR_DEPTHOUT, SpvStorageClassOutput, {VSIR_DATA_F32, 1, SpvBuiltInFragDepth}}, + {VKD3DSPR_DEPTHOUTGE, SpvStorageClassOutput, {VSIR_DATA_F32, 1, SpvBuiltInFragDepth}}, + {VKD3DSPR_DEPTHOUTLE, SpvStorageClassOutput, {VSIR_DATA_F32, 1, SpvBuiltInFragDepth}}, - {VKD3DSPR_OUTSTENCILREF, SpvStorageClassOutput, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInFragStencilRefEXT}}, + {VKD3DSPR_OUTSTENCILREF, SpvStorageClassOutput, {VSIR_DATA_U32, 1, SpvBuiltInFragStencilRefEXT}}, - {VKD3DSPR_WAVELANECOUNT, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInSubgroupSize}}, - {VKD3DSPR_WAVELANEINDEX, SpvStorageClassInput, {VKD3D_SHADER_COMPONENT_UINT, 1, SpvBuiltInSubgroupLocalInvocationId}}, + {VKD3DSPR_WAVELANECOUNT, SpvStorageClassInput, {VSIR_DATA_U32, 1, SpvBuiltInSubgroupSize}}, + {VKD3DSPR_WAVELANEINDEX, SpvStorageClassInput, {VSIR_DATA_U32, 1, SpvBuiltInSubgroupLocalInvocationId}}, }; static void spirv_compiler_emit_register_execution_mode(struct spirv_compiler *compiler, @@ -5527,7 +5531,8 @@ static uint32_t spirv_compiler_emit_load_invocation_id(struct spirv_compiler *co uint32_t type_id, id; id = spirv_compiler_get_invocation_id(compiler); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_INT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_I32, 1); + return vkd3d_spirv_build_op_load(builder, type_id, id, SpvMemoryAccessMaskNone); } @@ -5641,13 +5646,12 @@ static uint32_t spirv_compiler_emit_builtin_variable_v(struct spirv_compiler *co sizes[0] = max(sizes[0], builtin->spirv_array_size); id = spirv_compiler_emit_array_variable(compiler, &builder->global_stream, storage_class, - builtin->component_type, builtin->component_count, array_sizes, size_count); + vkd3d_component_type_from_data_type(builtin->data_type), builtin->component_count, array_sizes, size_count); vkd3d_spirv_add_iface_variable(builder, id); spirv_compiler_decorate_builtin(compiler, id, builtin->spirv_builtin); if (compiler->shader_type == VKD3D_SHADER_TYPE_PIXEL && storage_class == SpvStorageClassInput - && builtin->component_type != VKD3D_SHADER_COMPONENT_FLOAT - && builtin->component_type != VKD3D_SHADER_COMPONENT_BOOL) + && builtin->data_type != VSIR_DATA_F32 && builtin->data_type != VSIR_DATA_BOOL) vkd3d_spirv_build_op_decorate(builder, id, SpvDecorationFlat, NULL, 0); return id; @@ -5711,8 +5715,7 @@ static const struct vkd3d_symbol *spirv_compiler_emit_io_register(struct spirv_c spirv_compiler_emit_register_debug_name(builder, id, reg); write_mask = vkd3d_write_mask_from_component_count(builtin->component_count); - vkd3d_symbol_set_register_info(®_symbol, id, - storage_class, builtin->component_type, write_mask); + vkd3d_symbol_set_register_info(®_symbol, id, storage_class, builtin->data_type, write_mask); reg_symbol.info.reg.is_aggregate = builtin->spirv_array_size; return spirv_compiler_put_symbol(compiler, ®_symbol); @@ -5733,6 +5736,7 @@ static void spirv_compiler_emit_input(struct spirv_compiler *compiler, uint32_t val_id, input_id, var_id; uint32_t type_id, float_type_id; struct vkd3d_symbol reg_symbol; + enum vsir_data_type data_type; SpvStorageClass storage_class; bool use_private_var = false; unsigned int array_sizes[2]; @@ -5781,12 +5785,14 @@ static void spirv_compiler_emit_input(struct spirv_compiler *compiler, if (builtin) { - component_type = builtin->component_type; + data_type = builtin->data_type; + component_type = vkd3d_component_type_from_data_type(data_type); input_component_count = builtin->component_count; } else { component_type = signature_element->component_type; + data_type = vsir_data_type_from_component_type(component_type); input_component_count = vsir_write_mask_component_count(signature_element->mask); } @@ -5852,7 +5858,7 @@ static void spirv_compiler_emit_input(struct spirv_compiler *compiler, } vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - use_private_var ? VKD3D_SHADER_COMPONENT_FLOAT : component_type, + use_private_var ? VSIR_DATA_F32 : data_type, use_private_var ? VKD3DSP_WRITEMASK_ALL : reg_write_mask); reg_symbol.info.reg.is_aggregate = array_sizes[0] || array_sizes[1]; VKD3D_ASSERT(!builtin || !builtin->spirv_array_size || use_private_var || array_sizes[0] || array_sizes[1]); @@ -5867,16 +5873,16 @@ static void spirv_compiler_emit_input(struct spirv_compiler *compiler, vsir_register_init(&dst_reg, reg_type, VSIR_DATA_F32, 1); dst_reg.idx[0].offset = element_idx; - type_id = vkd3d_spirv_get_type_id(builder, component_type, input_component_count); + type_id = spirv_get_type_id(builder, data_type, input_component_count); val_id = vkd3d_spirv_build_op_load(builder, type_id, input_id, SpvMemoryAccessMaskNone); if (builtin && builtin->fixup_pfn) val_id = builtin->fixup_pfn(compiler, val_id); - if (component_type != VKD3D_SHADER_COMPONENT_FLOAT) + if (data_type != VSIR_DATA_F32) { - float_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, input_component_count); + float_type_id = spirv_get_type_id(builder, VSIR_DATA_F32, input_component_count); val_id = vkd3d_spirv_build_op_bitcast(builder, float_type_id, val_id); } @@ -6024,13 +6030,14 @@ static void spirv_compiler_emit_output(struct spirv_compiler *compiler, const struct shader_signature *shader_signature; const struct vkd3d_spirv_builtin *builtin; enum vkd3d_shader_sysval_semantic sysval; - uint32_t write_mask; bool use_private_variable = false; struct vkd3d_symbol reg_symbol; + enum vsir_data_type data_type; SpvStorageClass storage_class; unsigned int array_sizes[2]; bool is_patch_constant; uint32_t id, var_id; + uint32_t write_mask; is_patch_constant = (reg_type == VKD3DSPR_PATCHCONST); @@ -6055,13 +6062,15 @@ static void spirv_compiler_emit_output(struct spirv_compiler *compiler, output_component_count = vsir_write_mask_component_count(write_mask); if (builtin) { - component_type = builtin->component_type; + data_type = builtin->data_type; + component_type = vkd3d_component_type_from_data_type(data_type); if (!builtin->spirv_array_size) output_component_count = builtin->component_count; } else { component_type = signature_element->component_type; + data_type = vsir_data_type_from_component_type(component_type); } storage_class = SpvStorageClassOutput; @@ -6152,7 +6161,7 @@ static void spirv_compiler_emit_output(struct spirv_compiler *compiler, } vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - use_private_variable ? VKD3D_SHADER_COMPONENT_FLOAT : component_type, + use_private_variable ? VSIR_DATA_F32 : data_type, use_private_variable ? VKD3DSP_WRITEMASK_ALL : write_mask); reg_symbol.info.reg.is_aggregate = array_sizes[0] || array_sizes[1]; VKD3D_ASSERT(!builtin || !builtin->spirv_array_size || use_private_variable || array_sizes[0] || array_sizes[1]); @@ -6225,7 +6234,7 @@ static void spirv_compiler_emit_store_shader_output(struct spirv_compiler *compi if (output_info->component_type != VKD3D_SHADER_COMPONENT_FLOAT) { - type_id = vkd3d_spirv_get_type_id(builder, output_info->component_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id_for_component_type(builder, output_info->component_type, VKD3D_VEC4_SIZE); val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); } @@ -6236,7 +6245,7 @@ static void spirv_compiler_emit_store_shader_output(struct spirv_compiler *compi /* Set values to 0 for not initialized shader output components. */ write_mask |= uninit_mask; zero_id = spirv_compiler_get_constant_vector(compiler, - output_info->component_type, VKD3D_VEC4_SIZE, 0); + vsir_data_type_from_component_type(output_info->component_type), VKD3D_VEC4_SIZE, 0); val_id = spirv_compiler_emit_vector_shuffle(compiler, zero_id, val_id, swizzle, uninit_mask, output_info->component_type, vsir_write_mask_component_count(write_mask)); @@ -6250,7 +6259,7 @@ static void spirv_compiler_emit_store_shader_output(struct spirv_compiler *compi output_id = output_info->id; if (output_index_id) { - type_id = vkd3d_spirv_get_type_id(builder, + type_id = spirv_get_type_id_for_component_type(builder, output_info->component_type, vsir_write_mask_component_count(dst_write_mask)); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassOutput, type_id); output_id = vkd3d_spirv_build_op_access_chain1(builder, ptr_type_id, output_id, output_index_id); @@ -6263,7 +6272,7 @@ static void spirv_compiler_emit_store_shader_output(struct spirv_compiler *compi return; } - type_id = vkd3d_spirv_get_type_id(builder, output_info->component_type, 1); + type_id = spirv_get_type_id_for_component_type(builder, output_info->component_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassOutput, type_id); mask = output_info->array_element_mask; array_idx = spirv_compiler_get_output_array_index(compiler, output); @@ -6305,7 +6314,7 @@ static void spirv_compiler_emit_shader_epilogue_function(struct spirv_compiler * function_id = compiler->epilogue_function_id; void_id = vkd3d_spirv_get_op_type_void(builder); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 4); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 4); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassPrivate, type_id); for (i = 0, count = 0; i < ARRAY_SIZE(compiler->private_output_variable); ++i) { @@ -6514,7 +6523,6 @@ static void spirv_compiler_emit_dcl_indexable_temp(struct spirv_compiler *compil const struct vkd3d_shader_indexable_temp *temp = &instruction->declaration.indexable_temp; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t id, type_id, length_id, ptr_type_id, init_id = 0; - enum vkd3d_shader_component_type component_type; struct vkd3d_shader_register reg; struct vkd3d_symbol reg_symbol; SpvStorageClass storage_class; @@ -6535,8 +6543,7 @@ static void spirv_compiler_emit_dcl_indexable_temp(struct spirv_compiler *compil function_location = spirv_compiler_get_current_function_location(compiler); vkd3d_spirv_begin_function_stream_insertion(builder, function_location); - component_type = vkd3d_component_type_from_data_type(temp->data_type); - type_id = vkd3d_spirv_get_type_id(builder, component_type, temp->component_count); + type_id = spirv_get_type_id(builder, temp->data_type, temp->component_count); length_id = spirv_compiler_get_constant_uint(compiler, temp->register_size); type_id = vkd3d_spirv_get_op_type_array(builder, type_id, length_id); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, storage_class, type_id); @@ -6549,8 +6556,8 @@ static void spirv_compiler_emit_dcl_indexable_temp(struct spirv_compiler *compil vkd3d_spirv_end_function_stream_insertion(builder); vkd3d_symbol_make_register(®_symbol, ®); - vkd3d_symbol_set_register_info(®_symbol, id, storage_class, - component_type, vkd3d_write_mask_from_component_count(temp->component_count)); + vkd3d_symbol_set_register_info(®_symbol, id, storage_class, temp->data_type, + vkd3d_write_mask_from_component_count(temp->component_count)); spirv_compiler_put_symbol(compiler, ®_symbol); } @@ -6577,7 +6584,7 @@ static void spirv_compiler_emit_push_constant_buffers(struct spirv_compiler *com if (!(member_ids = vkd3d_calloc(count, sizeof(*member_ids)))) return; - vec4_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, VKD3D_VEC4_SIZE); + vec4_id = spirv_get_type_id(builder, VSIR_DATA_F32, VKD3D_VEC4_SIZE); for (i = 0, j = 0; i < compiler->shader_interface.push_constant_buffer_count; ++i) { @@ -6594,7 +6601,7 @@ static void spirv_compiler_emit_push_constant_buffers(struct spirv_compiler *com if (compiler->offset_info.descriptor_table_count) { - uint32_t type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + uint32_t type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); length_id = spirv_compiler_get_constant_uint(compiler, compiler->offset_info.descriptor_table_count); member_ids[j] = vkd3d_spirv_build_op_type_array(builder, type_id, length_id); vkd3d_spirv_build_op_decorate1(builder, member_ids[j], SpvDecorationArrayStride, 4); @@ -6626,8 +6633,7 @@ static void spirv_compiler_emit_push_constant_buffers(struct spirv_compiler *com vkd3d_spirv_build_op_member_name(builder, struct_id, j, "cb%u", reg_idx); vkd3d_symbol_make_register(®_symbol, &cb->reg); - vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - VKD3D_SHADER_COMPONENT_FLOAT, VKD3DSP_WRITEMASK_ALL); + vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, VSIR_DATA_F32, VKD3DSP_WRITEMASK_ALL); reg_symbol.info.reg.member_idx = j; spirv_compiler_put_symbol(compiler, ®_symbol); @@ -6788,7 +6794,7 @@ static void spirv_compiler_emit_cbv_declaration(struct spirv_compiler *compiler, return; } - vec4_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, VKD3D_VEC4_SIZE); + vec4_id = spirv_get_type_id(builder, VSIR_DATA_F32, VKD3D_VEC4_SIZE); length_id = spirv_compiler_get_constant_uint(compiler, size); array_type_id = vkd3d_spirv_build_op_type_array(builder, vec4_id, length_id); vkd3d_spirv_build_op_decorate1(builder, array_type_id, SpvDecorationArrayStride, 16); @@ -6802,8 +6808,7 @@ static void spirv_compiler_emit_cbv_declaration(struct spirv_compiler *compiler, ®, range, VKD3D_SHADER_RESOURCE_BUFFER, descriptor, false, &var_info); vkd3d_symbol_make_register(®_symbol, ®); - vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - VKD3D_SHADER_COMPONENT_FLOAT, VKD3DSP_WRITEMASK_ALL); + vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, VSIR_DATA_F32, VKD3DSP_WRITEMASK_ALL); reg_symbol.descriptor_array = var_info.array_symbol; reg_symbol.info.reg.binding_base_idx = var_info.binding_base_idx; spirv_compiler_put_symbol(compiler, ®_symbol); @@ -6829,8 +6834,7 @@ static void spirv_compiler_emit_dcl_immediate_constant_buffer(struct spirv_compi reg.idx[0].offset = icb->register_idx; vkd3d_symbol_make_register(®_symbol, ®); vkd3d_symbol_set_register_info(®_symbol, icb_id, SpvStorageClassPrivate, - vkd3d_component_type_from_data_type(icb->data_type), - vkd3d_write_mask_from_component_count(icb->component_count)); + icb->data_type, vkd3d_write_mask_from_component_count(icb->component_count)); spirv_compiler_put_symbol(compiler, ®_symbol); } @@ -6859,8 +6863,7 @@ static void spirv_compiler_emit_sampler_declaration(struct spirv_compiler *compi ®, range, VKD3D_SHADER_RESOURCE_NONE, descriptor, false, &var_info); vkd3d_symbol_make_register(®_symbol, ®); - vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - VKD3D_SHADER_COMPONENT_FLOAT, VKD3DSP_WRITEMASK_ALL); + vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, VSIR_DATA_F32, VKD3DSP_WRITEMASK_ALL); reg_symbol.descriptor_array = var_info.array_symbol; reg_symbol.info.reg.binding_base_idx = var_info.binding_base_idx; spirv_compiler_put_symbol(compiler, ®_symbol); @@ -6927,7 +6930,7 @@ static uint32_t spirv_compiler_get_image_type_id(struct spirv_compiler *compiler vkd3d_spirv_enable_capability(builder, SpvCapabilityStorageImageReadWithoutFormat); } - sampled_type_id = vkd3d_spirv_get_type_id(builder, data_type, 1); + sampled_type_id = spirv_get_type_id_for_component_type(builder, data_type, 1); return vkd3d_spirv_get_op_type_image(builder, sampled_type_id, resource_type_info->dim, 2, resource_type_info->arrayed, resource_type_info->ms, reg->type == VKD3DSPR_UAV ? 2 : 1, format); @@ -6996,7 +6999,7 @@ static void spirv_compiler_emit_combined_sampler_declarations(struct spirv_compi symbol.id = var_id; symbol.descriptor_array = NULL; symbol.info.resource.range = *resource_range; - symbol.info.resource.sampled_type = sampled_type; + symbol.info.resource.sampled_type = vsir_data_type_from_component_type(sampled_type); symbol.info.resource.type_id = image_type_id; symbol.info.resource.resource_type_info = resource_type_info; symbol.info.resource.structure_stride = structure_stride; @@ -7053,7 +7056,7 @@ static void spirv_compiler_emit_resource_declaration(struct spirv_compiler *comp { uint32_t array_type_id, struct_id; - type_id = vkd3d_spirv_get_type_id(builder, sampled_type, 1); + type_id = spirv_get_type_id(builder, descriptor->resource_data_type, 1); array_type_id = vkd3d_spirv_get_op_type_runtime_array(builder, type_id); vkd3d_spirv_build_op_decorate1(builder, array_type_id, SpvDecorationArrayStride, 4); @@ -7094,7 +7097,7 @@ static void spirv_compiler_emit_resource_declaration(struct spirv_compiler *comp { VKD3D_ASSERT(structure_stride); /* counters are valid only for structured buffers */ - counter_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + counter_type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); if (spirv_compiler_is_opengl_target(compiler)) { vkd3d_spirv_enable_capability(builder, SpvCapabilityAtomicStorage); @@ -7126,7 +7129,7 @@ static void spirv_compiler_emit_resource_declaration(struct spirv_compiler *comp resource_symbol.id = var_id; resource_symbol.descriptor_array = var_info.array_symbol; resource_symbol.info.resource.range = *range; - resource_symbol.info.resource.sampled_type = sampled_type; + resource_symbol.info.resource.sampled_type = descriptor->resource_data_type; resource_symbol.info.resource.type_id = type_id; resource_symbol.info.resource.resource_type_info = resource_type_info; resource_symbol.info.resource.structure_stride = structure_stride; @@ -7158,7 +7161,7 @@ static void spirv_compiler_emit_workgroup_memory(struct spirv_compiler *compiler if (alignment) TRACE("Ignoring alignment %u.\n", alignment); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); length_id = spirv_compiler_get_constant_uint(compiler, size); array_type_id = vkd3d_spirv_get_op_type_array(builder, type_id, length_id); @@ -7170,8 +7173,7 @@ static void spirv_compiler_emit_workgroup_memory(struct spirv_compiler *compiler spirv_compiler_emit_register_debug_name(builder, var_id, reg); vkd3d_symbol_make_register(®_symbol, reg); - vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, - VKD3D_SHADER_COMPONENT_UINT, VKD3DSP_WRITEMASK_0); + vkd3d_symbol_set_register_info(®_symbol, var_id, storage_class, VSIR_DATA_U32, VKD3DSP_WRITEMASK_0); reg_symbol.info.reg.structure_stride = structure_stride; spirv_compiler_put_symbol(compiler, ®_symbol); } @@ -7209,10 +7211,9 @@ static void spirv_compiler_emit_output_vertex_count(struct spirv_compiler *compi SpvExecutionModeOutputVertices, instruction->declaration.count); } -static void spirv_compiler_emit_dcl_input_primitive(struct spirv_compiler *compiler, - const struct vkd3d_shader_instruction *instruction) +static void spirv_compiler_emit_input_primitive(struct spirv_compiler *compiler) { - enum vkd3d_primitive_type primitive_type = instruction->declaration.primitive_type.type; + enum vkd3d_primitive_type primitive_type = compiler->program->input_primitive; SpvExecutionMode mode; switch (primitive_type) @@ -7233,7 +7234,8 @@ static void spirv_compiler_emit_dcl_input_primitive(struct spirv_compiler *compi mode = SpvExecutionModeInputTrianglesAdjacency; break; default: - FIXME("Unhandled primitive type %#x.\n", primitive_type); + spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED, + "Unhandled input primitive type %#x.", primitive_type); return; } @@ -7263,10 +7265,9 @@ static void spirv_compiler_emit_point_size(struct spirv_compiler *compiler) } } -static void spirv_compiler_emit_dcl_output_topology(struct spirv_compiler *compiler, - const struct vkd3d_shader_instruction *instruction) +static void spirv_compiler_emit_output_topology(struct spirv_compiler *compiler) { - enum vkd3d_primitive_type primitive_type = instruction->declaration.primitive_type.type; + enum vkd3d_primitive_type primitive_type = compiler->program->output_topology; SpvExecutionMode mode; switch (primitive_type) @@ -7282,7 +7283,8 @@ static void spirv_compiler_emit_dcl_output_topology(struct spirv_compiler *compi mode = SpvExecutionModeOutputTriangleStrip; break; default: - ERR("Unexpected primitive type %#x.\n", primitive_type); + spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED, + "Unhandled output topology %#x.", primitive_type); return; } @@ -7624,11 +7626,12 @@ static void spirv_compiler_emit_bool_cast(struct spirv_compiler *compiler, /* ITOD is not supported. Frontends which emit bool casts must use ITOF for double. */ val_id = spirv_compiler_emit_bool_to_double(compiler, 1, val_id, instruction->opcode == VSIR_OP_ITOF); } - else if (dst->reg.data_type == VSIR_DATA_U16 || dst->reg.data_type == VSIR_DATA_U32) + else if (dst->reg.data_type == VSIR_DATA_I16 || dst->reg.data_type == VSIR_DATA_I32 + || dst->reg.data_type == VSIR_DATA_U16 || dst->reg.data_type == VSIR_DATA_U32) { val_id = spirv_compiler_emit_bool_to_int(compiler, 1, val_id, instruction->opcode == VSIR_OP_ITOI); } - else if (dst->reg.data_type == VSIR_DATA_U64) + else if (dst->reg.data_type == VSIR_DATA_I64 || dst->reg.data_type == VSIR_DATA_U64) { val_id = spirv_compiler_emit_bool_to_int64(compiler, 1, val_id, instruction->opcode == VSIR_OP_ITOI); } @@ -7724,7 +7727,7 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil condition_id = spirv_compiler_emit_int_to_bool(compiler, VKD3D_SHADER_CONDITIONAL_OP_NZ, src[1].reg.data_type, component_count, src_ids[1]); - if (dst[0].reg.data_type == VSIR_DATA_U64) + if (data_type_is_64_bit(dst[0].reg.data_type)) uint_max_id = spirv_compiler_get_constant_uint64_vector(compiler, UINT64_MAX, component_count); else uint_max_id = spirv_compiler_get_constant_uint_vector(compiler, UINT_MAX, component_count); @@ -7741,7 +7744,7 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil || instruction->opcode == VSIR_OP_ISHR || instruction->opcode == VSIR_OP_USHR)) { uint32_t mask_id = spirv_compiler_get_constant_vector(compiler, - VKD3D_SHADER_COMPONENT_UINT, vsir_write_mask_component_count(dst->write_mask), 0x1f); + VSIR_DATA_U32, vsir_write_mask_component_count(dst->write_mask), 0x1f); src_ids[1] = vkd3d_spirv_build_op_and(builder, type_id, src_ids[1], mask_id); } @@ -7839,7 +7842,7 @@ static void spirv_compiler_emit_ext_glsl_instruction(struct spirv_compiler *comp unsigned int i, component_count; enum GLSLstd450 glsl_inst; - if (src[0].reg.data_type == VSIR_DATA_U64 && (instruction->opcode == VSIR_OP_FIRSTBIT_HI + if (data_type_is_64_bit(src[0].reg.data_type) && (instruction->opcode == VSIR_OP_FIRSTBIT_HI || instruction->opcode == VSIR_OP_FIRSTBIT_LO || instruction->opcode == VSIR_OP_FIRSTBIT_SHI)) { /* At least some drivers support this anyway, but if validation is enabled it will fail. */ @@ -7878,7 +7881,7 @@ static void spirv_compiler_emit_ext_glsl_instruction(struct spirv_compiler *comp component_count = vsir_write_mask_component_count(dst->write_mask); uint_max_id = spirv_compiler_get_constant_uint_vector(compiler, UINT32_MAX, component_count); condition_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpIEqual, - vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count), val_id, uint_max_id); + spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count), val_id, uint_max_id); rev_val_id = vkd3d_spirv_build_op_isub(builder, type_id, spirv_compiler_get_constant_uint_vector(compiler, 31, component_count), val_id); val_id = vkd3d_spirv_build_op_select(builder, type_id, condition_id, val_id, rev_val_id); @@ -7905,7 +7908,7 @@ static void spirv_compiler_emit_mov(struct spirv_compiler *compiler, spirv_compiler_get_register_info(compiler, &dst->reg, &dst_reg_info); spirv_compiler_get_register_info(compiler, &src->reg, &src_reg_info); - if (dst_reg_info.component_type != src_reg_info.component_type + if (dst_reg_info.data_type != src_reg_info.data_type || dst_reg_info.write_mask != src_reg_info.write_mask) goto general_implementation; @@ -7928,7 +7931,7 @@ static void spirv_compiler_emit_mov(struct spirv_compiler *compiler, dst_id = spirv_compiler_get_register_id(compiler, &dst->reg); src_id = spirv_compiler_get_register_id(compiler, &src->reg); - type_id = vkd3d_spirv_get_type_id(builder, dst_reg_info.component_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, dst_reg_info.data_type, VKD3D_VEC4_SIZE); val_id = vkd3d_spirv_build_op_load(builder, type_id, src_id, SpvMemoryAccessMaskNone); dst_val_id = vkd3d_spirv_build_op_load(builder, type_id, dst_id, SpvMemoryAccessMaskNone); @@ -7957,7 +7960,7 @@ general_implementation: val_id = spirv_compiler_emit_load_src(compiler, src, write_mask); if (dst->reg.data_type != src->reg.data_type) { - val_id = vkd3d_spirv_build_op_bitcast(builder, vkd3d_spirv_get_type_id_for_data_type(builder, + val_id = vkd3d_spirv_build_op_bitcast(builder, spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)), val_id); } spirv_compiler_emit_store_dst(compiler, dst, val_id); @@ -7983,7 +7986,7 @@ static void spirv_compiler_emit_movc(struct spirv_compiler *compiler, { if (instruction->opcode == VSIR_OP_CMP) condition_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpFOrdGreaterThanEqual, - vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count), condition_id, + spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count), condition_id, spirv_compiler_get_constant_float_vector(compiler, 0.0f, component_count)); else condition_id = spirv_compiler_emit_int_to_bool(compiler, @@ -8010,7 +8013,7 @@ static void spirv_compiler_emit_swapc(struct spirv_compiler *compiler, src2_id = spirv_compiler_emit_load_src(compiler, &src[2], dst->write_mask); component_count = vsir_write_mask_component_count(dst->write_mask); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, component_count); condition_id = spirv_compiler_emit_int_to_bool(compiler, VKD3D_SHADER_CONDITIONAL_OP_NZ, src[0].reg.data_type, component_count, condition_id); @@ -8027,13 +8030,13 @@ static void spirv_compiler_emit_dot(struct spirv_compiler *compiler, struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; const struct vkd3d_shader_dst_param *dst = instruction->dst; const struct vkd3d_shader_src_param *src = instruction->src; - enum vkd3d_shader_component_type component_type; uint32_t type_id, val_id, src_ids[2]; unsigned int component_count, i; + enum vsir_data_type data_type; uint32_t write_mask; component_count = vsir_write_mask_component_count(dst->write_mask); - component_type = vkd3d_component_type_from_data_type(dst->reg.data_type); + data_type = dst->reg.data_type; if (instruction->opcode == VSIR_OP_DP4) write_mask = VKD3DSP_WRITEMASK_ALL; @@ -8046,15 +8049,12 @@ static void spirv_compiler_emit_dot(struct spirv_compiler *compiler, for (i = 0; i < ARRAY_SIZE(src_ids); ++i) src_ids[i] = spirv_compiler_emit_load_src(compiler, &src[i], write_mask); - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); + type_id = spirv_get_type_id(builder, data_type, 1); val_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpDot, type_id, src_ids[0], src_ids[1]); if (component_count > 1) - { - val_id = spirv_compiler_emit_construct_vector(compiler, - component_type, component_count, val_id, 0, 1); - } + val_id = spirv_compiler_emit_construct_vector(compiler, data_type, component_count, val_id, 0, 1); if (instruction->flags & VKD3DSI_PRECISE_XYZW) vkd3d_spirv_build_op_decorate(builder, val_id, SpvDecorationNoContraction, NULL, 0); @@ -8092,7 +8092,7 @@ static void spirv_compiler_emit_imad(struct spirv_compiler *compiler, unsigned int i, component_count; component_count = vsir_write_mask_component_count(dst->write_mask); - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, component_count); + type_id = spirv_get_type_id(builder, dst->reg.data_type, component_count); for (i = 0; i < ARRAY_SIZE(src_ids); ++i) src_ids[i] = spirv_compiler_emit_load_src(compiler, &src[i], dst->write_mask); @@ -8111,7 +8111,6 @@ static void spirv_compiler_emit_ftoi(struct spirv_compiler *compiler, const struct vkd3d_shader_dst_param *dst = instruction->dst; const struct vkd3d_shader_src_param *src = instruction->src; uint32_t src_type_id, dst_type_id, condition_type_id; - enum vkd3d_shader_component_type component_type; unsigned int component_count; uint32_t write_mask; @@ -8143,17 +8142,16 @@ static void spirv_compiler_emit_ftoi(struct spirv_compiler *compiler, val_id = vkd3d_spirv_build_op_glsl_std450_max(builder, src_type_id, src_id, int_min_id); /* VSIR allows the destination of a signed conversion to be unsigned. */ - component_type = vkd3d_component_type_from_data_type(dst->reg.data_type); - int_max_id = spirv_compiler_get_constant_vector(compiler, component_type, component_count, INT_MAX); - condition_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count); + int_max_id = spirv_compiler_get_constant_vector(compiler, dst->reg.data_type, component_count, INT_MAX); + condition_type_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count); condition_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpFOrdGreaterThanEqual, condition_type_id, val_id, float_max_id); val_id = vkd3d_spirv_build_op_tr1(builder, &builder->function_stream, SpvOpConvertFToS, dst_type_id, val_id); val_id = vkd3d_spirv_build_op_select(builder, dst_type_id, condition_id, int_max_id, val_id); - zero_id = spirv_compiler_get_constant_vector(compiler, component_type, component_count, 0); + zero_id = spirv_compiler_get_constant_vector(compiler, dst->reg.data_type, component_count, 0); condition_id = vkd3d_spirv_build_op_tr1(builder, &builder->function_stream, SpvOpIsNan, condition_type_id, src_id); val_id = vkd3d_spirv_build_op_select(builder, dst_type_id, condition_id, zero_id, val_id); @@ -8199,7 +8197,7 @@ static void spirv_compiler_emit_ftou(struct spirv_compiler *compiler, val_id = vkd3d_spirv_build_op_glsl_std450_max(builder, src_type_id, src_id, zero_id); uint_max_id = spirv_compiler_get_constant_uint_vector(compiler, UINT_MAX, component_count); - condition_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count); + condition_type_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count); condition_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpFOrdGreaterThanEqual, condition_type_id, val_id, float_max_id); @@ -8224,7 +8222,7 @@ static void spirv_compiler_emit_dtof(struct spirv_compiler *compiler, src_id = spirv_compiler_emit_load_src(compiler, src, write_mask); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, component_count); val_id = vkd3d_spirv_build_op_tr1(builder, &builder->function_stream, SpvOpFConvert, type_id, src_id); if (instruction->flags & VKD3DSI_PRECISE_XYZW) vkd3d_spirv_build_op_decorate(builder, val_id, SpvDecorationNoContraction, NULL, 0); @@ -8248,8 +8246,8 @@ static void spirv_compiler_emit_bitfield_instruction(struct spirv_compiler *comp VKD3D_ASSERT(2 <= src_count && src_count <= ARRAY_SIZE(src_ids)); component_type = vkd3d_component_type_from_data_type(dst->reg.data_type); - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); - size = (src[src_count - 1].reg.data_type == VSIR_DATA_U64) ? 0x40 : 0x20; + type_id = spirv_get_type_id(builder, dst->reg.data_type, 1); + size = data_type_is_64_bit(src[src_count - 1].reg.data_type) ? 0x40 : 0x20; mask_id = spirv_compiler_get_constant_uint(compiler, size - 1); size_id = spirv_compiler_get_constant_uint(compiler, size); @@ -8305,8 +8303,8 @@ static void spirv_compiler_emit_f16tof32(struct spirv_compiler *compiler, unsigned int i, j; instr_set_id = vkd3d_spirv_get_glsl_std450_instr_set(builder); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 2); - scalar_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 2); + scalar_type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 1); /* FIXME: Consider a single UnpackHalf2x16 instruction per 2 components. */ VKD3D_ASSERT(dst->write_mask & VKD3DSP_WRITEMASK_ALL); @@ -8338,8 +8336,8 @@ static void spirv_compiler_emit_f32tof16(struct spirv_compiler *compiler, unsigned int i, j; instr_set_id = vkd3d_spirv_get_glsl_std450_instr_set(builder); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 2); - scalar_type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 2); + scalar_type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); zero_id = spirv_compiler_get_constant_float(compiler, 0.0f); /* FIXME: Consider a single PackHalf2x16 instruction per 2 components. */ @@ -8418,7 +8416,7 @@ static void spirv_compiler_emit_comparison_instruction(struct spirv_compiler *co src0_id = spirv_compiler_emit_load_src(compiler, &src[0], write_mask); src1_id = spirv_compiler_emit_load_src(compiler, &src[1], write_mask); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count); result_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, op, type_id, src0_id, src1_id); @@ -8470,7 +8468,7 @@ static void spirv_compiler_emit_float_comparison_instruction(struct spirv_compil src0_id = spirv_compiler_emit_load_src(compiler, &src[0], dst->write_mask); src1_id = spirv_compiler_emit_load_src(compiler, &src[1], dst->write_mask); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, component_count); result_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, op, type_id, src0_id, src1_id); result_id = spirv_compiler_emit_bool_to_float(compiler, component_count, result_id, false); @@ -8806,7 +8804,7 @@ struct vkd3d_shader_image uint32_t image_id; uint32_t sampled_image_id; - enum vkd3d_shader_component_type sampled_type; + enum vsir_data_type sampled_type; uint32_t image_type_id; const struct vkd3d_spirv_resource_type *resource_type_info; unsigned int structure_stride; @@ -8926,9 +8924,9 @@ static void spirv_compiler_prepare_image(struct spirv_compiler *compiler, image->image_id = 0; } - image->image_type_id = spirv_compiler_get_image_type_id(compiler, resource_reg, - &symbol->info.resource.range, image->resource_type_info, - image->sampled_type, image->structure_stride || image->raw); + image->image_type_id = spirv_compiler_get_image_type_id(compiler, resource_reg, &symbol->info.resource.range, + image->resource_type_info, vkd3d_component_type_from_data_type(image->sampled_type), + image->structure_stride || image->raw); if (sampled) { @@ -8977,8 +8975,7 @@ static uint32_t spirv_compiler_emit_texel_offset(struct spirv_compiler *compiler int32_t data[4] = {offset->u, offset->v, offset->w, 0}; VKD3D_ASSERT(resource_type_info->dim != SpvDimCube); - return spirv_compiler_get_constant(compiler, - VKD3D_SHADER_COMPONENT_INT, component_count, (const uint32_t *)data); + return spirv_compiler_get_constant(compiler, VSIR_DATA_I32, component_count, (const uint32_t *)data); } static void spirv_compiler_emit_ld(struct spirv_compiler *compiler, @@ -8999,7 +8996,7 @@ static void spirv_compiler_emit_ld(struct spirv_compiler *compiler, spirv_compiler_prepare_image(compiler, &image, &src[1].reg, NULL, VKD3D_IMAGE_FLAG_NONE); - type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); coordinate_mask = (1u << image.resource_type_info->coordinate_component_count) - 1; coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], coordinate_mask); if (image.resource_type_info->resource_type != VKD3D_SHADER_RESOURCE_BUFFER && !multisample) @@ -9024,8 +9021,8 @@ static void spirv_compiler_emit_ld(struct spirv_compiler *compiler, val_id = vkd3d_spirv_build_op_image_fetch(builder, type_id, image.image_id, coordinate_id, operands_mask, image_operands, image_operand_count); - spirv_compiler_emit_store_dst_swizzled(compiler, - dst, val_id, image.sampled_type, src[1].swizzle); + spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), src[1].swizzle); } static void spirv_compiler_emit_lod(struct spirv_compiler *compiler, @@ -9045,13 +9042,13 @@ static void spirv_compiler_emit_lod(struct spirv_compiler *compiler, spirv_compiler_prepare_image(compiler, &image, &resource->reg, &sampler->reg, VKD3D_IMAGE_FLAG_SAMPLED); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 2); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 2); coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_ALL); val_id = vkd3d_spirv_build_op_image_query_lod(builder, type_id, image.sampled_image_id, coordinate_id); - spirv_compiler_emit_store_dst_swizzled(compiler, - dst, val_id, image.sampled_type, resource->swizzle); + spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), resource->swizzle); } static void spirv_compiler_emit_sample(struct spirv_compiler *compiler, @@ -9115,14 +9112,14 @@ static void spirv_compiler_emit_sample(struct spirv_compiler *compiler, instruction, image.resource_type_info); } - sampled_type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); + sampled_type_id = spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_ALL); VKD3D_ASSERT(image_operand_count <= ARRAY_SIZE(image_operands)); val_id = vkd3d_spirv_build_op_image_sample(builder, op, sampled_type_id, image.sampled_image_id, coordinate_id, operands_mask, image_operands, image_operand_count); - spirv_compiler_emit_store_dst_swizzled(compiler, - dst, val_id, image.sampled_type, resource->swizzle); + spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), resource->swizzle); } static void spirv_compiler_emit_sample_c(struct spirv_compiler *compiler, @@ -9160,15 +9157,15 @@ static void spirv_compiler_emit_sample_c(struct spirv_compiler *compiler, instruction, image.resource_type_info); } - sampled_type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, 1); + sampled_type_id = spirv_get_type_id(builder, image.sampled_type, 1); coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_ALL); dref_id = spirv_compiler_emit_load_src(compiler, &src[3], VKD3DSP_WRITEMASK_0); val_id = vkd3d_spirv_build_op_image_sample_dref(builder, op, sampled_type_id, image.sampled_image_id, coordinate_id, dref_id, operands_mask, image_operands, image_operand_count); - spirv_compiler_emit_store_dst_scalar(compiler, - dst, val_id, image.sampled_type, src[1].swizzle); + spirv_compiler_emit_store_dst_scalar(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), src[1].swizzle); } static void spirv_compiler_emit_gather4(struct spirv_compiler *compiler, @@ -9219,7 +9216,7 @@ static void spirv_compiler_emit_gather4(struct spirv_compiler *compiler, instruction, image.resource_type_info); } - sampled_type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); + sampled_type_id = spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); coordinate_mask = (1u << image.resource_type_info->coordinate_component_count) - 1; coordinate_id = spirv_compiler_emit_load_src(compiler, addr, coordinate_mask); if (image_flags & VKD3D_IMAGE_FLAG_DEPTH) @@ -9234,15 +9231,14 @@ static void spirv_compiler_emit_gather4(struct spirv_compiler *compiler, { component_idx = vsir_swizzle_get_component(sampler->swizzle, 0); /* Nvidia driver requires signed integer type. */ - component_id = spirv_compiler_get_constant(compiler, - VKD3D_SHADER_COMPONENT_INT, 1, &component_idx); + component_id = spirv_compiler_get_constant(compiler, VSIR_DATA_I32, 1, &component_idx); val_id = vkd3d_spirv_build_op_image_gather(builder, sampled_type_id, image.sampled_image_id, coordinate_id, component_id, operands_mask, image_operands, image_operand_count); } - spirv_compiler_emit_store_dst_swizzled(compiler, - dst, val_id, image.sampled_type, resource->swizzle); + spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), resource->swizzle); } static uint32_t spirv_compiler_emit_raw_structured_addressing( @@ -9301,10 +9297,10 @@ static void spirv_compiler_emit_ld_raw_structured_srv_uav(struct spirv_compiler if (storage_buffer_uav) { - texel_type_id = vkd3d_spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); + texel_type_id = spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, texel_type_id); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id_for_component_type(builder, VKD3D_SHADER_COMPONENT_UINT, 1); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, resource_symbol->info.resource.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); @@ -9336,11 +9332,11 @@ static void spirv_compiler_emit_ld_raw_structured_srv_uav(struct spirv_compiler spirv_compiler_prepare_image(compiler, &image, &resource->reg, NULL, VKD3D_IMAGE_FLAG_NONE); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id_for_component_type(builder, VKD3D_SHADER_COMPONENT_UINT, 1); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, image.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); - texel_type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); + texel_type_id = spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); VKD3D_ASSERT(dst->write_mask & VKD3DSP_WRITEMASK_ALL); for (i = 0, j = 0; i < VKD3D_VEC4_SIZE; ++i) { @@ -9379,7 +9375,7 @@ static void spirv_compiler_emit_ld_tgsm(struct spirv_compiler *compiler, if (!spirv_compiler_get_register_info(compiler, &resource->reg, ®_info)) return; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, reg_info.storage_class, type_id); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, reg_info.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); @@ -9438,10 +9434,10 @@ static void spirv_compiler_emit_store_uav_raw_structured(struct spirv_compiler * if (spirv_compiler_use_storage_buffer(compiler, &resource_symbol->info.resource)) { - type_id = vkd3d_spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); + type_id = spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, type_id); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id_for_component_type(builder, VKD3D_SHADER_COMPONENT_UINT, 1); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, resource_symbol->info.resource.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); @@ -9469,7 +9465,7 @@ static void spirv_compiler_emit_store_uav_raw_structured(struct spirv_compiler * } else { - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id_for_component_type(builder, VKD3D_SHADER_COMPONENT_UINT, 1); spirv_compiler_prepare_image(compiler, &image, &dst->reg, NULL, VKD3D_IMAGE_FLAG_NONE); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, image.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); @@ -9482,7 +9478,7 @@ static void spirv_compiler_emit_store_uav_raw_structured(struct spirv_compiler * for (component_idx = 0; component_idx < component_count; ++component_idx) { /* Mesa Vulkan drivers require the texel parameter to be a vector. */ - data_id = spirv_compiler_emit_construct_vector(compiler, VKD3D_SHADER_COMPONENT_UINT, + data_id = spirv_compiler_emit_construct_vector(compiler, VSIR_DATA_U32, VKD3D_VEC4_SIZE, val_id, component_idx, component_count); coordinate_id = base_coordinate_id; @@ -9512,7 +9508,7 @@ static void spirv_compiler_emit_store_tgsm(struct spirv_compiler *compiler, if (!spirv_compiler_get_register_info(compiler, &dst->reg, ®_info)) return; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, reg_info.storage_class, type_id); base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler, type_id, reg_info.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0); @@ -9570,7 +9566,7 @@ static void spirv_compiler_emit_ld_uav_typed(struct spirv_compiler *compiler, if (spirv_compiler_use_storage_buffer(compiler, &resource_symbol->info.resource)) { - type_id = vkd3d_spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); + type_id = spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, type_id); coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_0); indices[0] = spirv_compiler_get_constant_uint(compiler, 0); @@ -9580,20 +9576,20 @@ static void spirv_compiler_emit_ld_uav_typed(struct spirv_compiler *compiler, val_id = vkd3d_spirv_build_op_load(builder, type_id, ptr_id, SpvMemoryAccessMaskNone); spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, - resource_symbol->info.resource.sampled_type, src[1].swizzle); + vkd3d_component_type_from_data_type(resource_symbol->info.resource.sampled_type), src[1].swizzle); } else { spirv_compiler_prepare_image(compiler, &image, &src[1].reg, NULL, VKD3D_IMAGE_FLAG_NONE); - type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, image.sampled_type, VKD3D_VEC4_SIZE); coordinate_mask = (1u << image.resource_type_info->coordinate_component_count) - 1; coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], coordinate_mask); val_id = vkd3d_spirv_build_op_image_read(builder, type_id, image.image_id, coordinate_id, SpvImageOperandsMaskNone, NULL, 0); - spirv_compiler_emit_store_dst_swizzled(compiler, - dst, val_id, image.sampled_type, src[1].swizzle); + spirv_compiler_emit_store_dst_swizzled(compiler, dst, val_id, + vkd3d_component_type_from_data_type(image.sampled_type), src[1].swizzle); } } @@ -9613,14 +9609,14 @@ static void spirv_compiler_emit_store_uav_typed(struct spirv_compiler *compiler, if (spirv_compiler_use_storage_buffer(compiler, &resource_symbol->info.resource)) { - type_id = vkd3d_spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); + type_id = spirv_get_type_id(builder, resource_symbol->info.resource.sampled_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, type_id); coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_0); indices[0] = spirv_compiler_get_constant_uint(compiler, 0); indices[1] = coordinate_id; - val_id = spirv_compiler_emit_load_src_with_type(compiler, &src[1], - VKD3DSP_WRITEMASK_0, resource_symbol->info.resource.sampled_type); + val_id = spirv_compiler_emit_load_src_with_type(compiler, &src[1], VKD3DSP_WRITEMASK_0, + vkd3d_component_type_from_data_type(resource_symbol->info.resource.sampled_type)); ptr_id = vkd3d_spirv_build_op_access_chain(builder, ptr_type_id, resource_symbol->id, indices, 2); vkd3d_spirv_build_op_store(builder, ptr_id, val_id, SpvMemoryAccessMaskNone); } @@ -9631,7 +9627,8 @@ static void spirv_compiler_emit_store_uav_typed(struct spirv_compiler *compiler, spirv_compiler_prepare_image(compiler, &image, &dst->reg, NULL, VKD3D_IMAGE_FLAG_NONE); coordinate_mask = (1u << image.resource_type_info->coordinate_component_count) - 1; coordinate_id = spirv_compiler_emit_load_src(compiler, &src[0], coordinate_mask); - texel_id = spirv_compiler_emit_load_src_with_type(compiler, &src[1], dst->write_mask, image.sampled_type); + texel_id = spirv_compiler_emit_load_src_with_type(compiler, &src[1], + dst->write_mask, vkd3d_component_type_from_data_type(image.sampled_type)); vkd3d_spirv_build_op_image_write(builder, image.image_id, coordinate_id, texel_id, SpvImageOperandsMaskNone, NULL, 0); @@ -9658,7 +9655,7 @@ static void spirv_compiler_emit_uav_counter_instruction(struct spirv_compiler *c counter_id = resource_symbol->info.resource.uav_counter_id; VKD3D_ASSERT(counter_id); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); if (resource_symbol->info.resource.uav_counter_array) { @@ -9772,6 +9769,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil struct vkd3d_shader_register_info reg_info; SpvMemorySemanticsMask memory_semantic; struct vkd3d_shader_image image; + enum vsir_data_type data_type; unsigned int structure_stride; uint32_t coordinate_mask; uint32_t operands[6]; @@ -9820,7 +9818,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil } } - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id_for_component_type(builder, VKD3D_SHADER_COMPONENT_UINT, 1); if (structure_stride || raw) { VKD3D_ASSERT(!raw != !structure_stride); @@ -9836,7 +9834,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil if (resource->reg.type == VKD3DSPR_GROUPSHAREDMEM) { - component_type = VKD3D_SHADER_COMPONENT_UINT; + data_type = VSIR_DATA_U32; ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, reg_info.storage_class, type_id); pointer_id = vkd3d_spirv_build_op_access_chain1(builder, ptr_type_id, reg_info.id, coordinate_id); } @@ -9844,8 +9842,8 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil { if (spirv_compiler_use_storage_buffer(compiler, &resource_symbol->info.resource)) { - component_type = resource_symbol->info.resource.sampled_type; - type_id = vkd3d_spirv_get_type_id(builder, component_type, 1); + data_type = resource_symbol->info.resource.sampled_type; + type_id = spirv_get_type_id(builder, data_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassUniform, type_id); operands[0] = spirv_compiler_get_constant_uint(compiler, 0); operands[1] = coordinate_id; @@ -9853,8 +9851,8 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil } else { - component_type = image.sampled_type; - type_id = vkd3d_spirv_get_type_id(builder, image.sampled_type, 1); + data_type = image.sampled_type; + type_id = spirv_get_type_id(builder, data_type, 1); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassImage, type_id); sample_id = spirv_compiler_get_constant_uint(compiler, 0); pointer_id = vkd3d_spirv_build_op_image_texel_pointer(builder, @@ -9862,6 +9860,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil } } + component_type = vkd3d_component_type_from_data_type(data_type); val_id = spirv_compiler_emit_load_src_with_type(compiler, &src[1], VKD3DSP_WRITEMASK_0, component_type); if (instruction->flags & VKD3DARF_VOLATILE) @@ -9907,7 +9906,7 @@ static void spirv_compiler_emit_bufinfo(struct spirv_compiler *compiler, { resource_symbol = spirv_compiler_find_resource(compiler, &src->reg); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); val_id = vkd3d_spirv_build_op_array_length(builder, type_id, resource_symbol->id, 0); write_mask = VKD3DSP_WRITEMASK_0; } @@ -9917,7 +9916,7 @@ static void spirv_compiler_emit_bufinfo(struct spirv_compiler *compiler, spirv_compiler_prepare_image(compiler, &image, &src->reg, NULL, VKD3D_IMAGE_FLAG_NONE); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); val_id = vkd3d_spirv_build_op_image_query_size(builder, type_id, image.image_id); write_mask = VKD3DSP_WRITEMASK_0; } @@ -9927,7 +9926,7 @@ static void spirv_compiler_emit_bufinfo(struct spirv_compiler *compiler, stride_id = spirv_compiler_get_constant_uint(compiler, image.structure_stride); constituents[0] = vkd3d_spirv_build_op_udiv(builder, type_id, val_id, stride_id); constituents[1] = stride_id; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, ARRAY_SIZE(constituents)); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, ARRAY_SIZE(constituents)); val_id = vkd3d_spirv_build_op_composite_construct(builder, type_id, constituents, ARRAY_SIZE(constituents)); write_mask |= VKD3DSP_WRITEMASK_1; @@ -9966,14 +9965,14 @@ static void spirv_compiler_emit_resinfo(struct spirv_compiler *compiler, size_component_count = image.resource_type_info->coordinate_component_count; if (image.resource_type_info->dim == SpvDimCube) --size_component_count; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, size_component_count); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, size_component_count); supports_mipmaps = src[1].reg.type != VKD3DSPR_UAV && !image.resource_type_info->ms; if (supports_mipmaps) { lod_id = spirv_compiler_emit_load_src(compiler, &src[0], VKD3DSP_WRITEMASK_0); val_id = vkd3d_spirv_build_op_image_query_size_lod(builder, type_id, image.image_id, lod_id); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); miplevel_count_id = vkd3d_spirv_build_op_image_query_levels(builder, type_id, image.image_id); } else @@ -9987,14 +9986,13 @@ static void spirv_compiler_emit_resinfo(struct spirv_compiler *compiler, for (i = 0; i < 3 - size_component_count; ++i) constituents[i + 1] = spirv_compiler_get_constant_uint(compiler, 0); constituents[i + 1] = miplevel_count_id; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, VKD3D_VEC4_SIZE); - val_id = vkd3d_spirv_build_op_composite_construct(builder, - type_id, constituents, i + 2); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, VKD3D_VEC4_SIZE); + val_id = vkd3d_spirv_build_op_composite_construct(builder, type_id, constituents, i + 2); if (!(instruction->flags & VKD3DSI_RESINFO_UINT)) { component_type = VKD3D_SHADER_COMPONENT_FLOAT; - type_id = vkd3d_spirv_get_type_id(builder, component_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, VKD3D_VEC4_SIZE); val_id = vkd3d_spirv_build_op_convert_utof(builder, type_id, val_id); if (instruction->flags & VKD3DSI_PRECISE_XYZW) vkd3d_spirv_build_op_decorate(builder, val_id, SpvDecorationNoContraction, NULL, 0); @@ -10022,7 +10020,7 @@ static uint32_t spirv_compiler_emit_query_sample_count(struct spirv_compiler *co vkd3d_spirv_enable_capability(builder, SpvCapabilityImageQuery); spirv_compiler_prepare_image(compiler, &image, &src->reg, NULL, VKD3D_IMAGE_FLAG_NONE); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); val_id = vkd3d_spirv_build_op_image_query_samples(builder, type_id, image.image_id); } @@ -10048,14 +10046,16 @@ static void spirv_compiler_emit_sample_info(struct spirv_compiler *compiler, constituents[0] = val_id; for (i = 1; i < VKD3D_VEC4_SIZE; ++i) + { constituents[i] = spirv_compiler_get_constant_uint(compiler, 0); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, VKD3D_VEC4_SIZE); + } + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, VKD3D_VEC4_SIZE); val_id = vkd3d_spirv_build_op_composite_construct(builder, type_id, constituents, VKD3D_VEC4_SIZE); if (!(instruction->flags & VKD3DSI_SAMPLE_INFO_UINT)) { component_type = VKD3D_SHADER_COMPONENT_FLOAT; - type_id = vkd3d_spirv_get_type_id(builder, component_type, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, VKD3D_VEC4_SIZE); val_id = vkd3d_spirv_build_op_convert_utof(builder, type_id, val_id); if (instruction->flags & VKD3DSI_PRECISE_XYZW) vkd3d_spirv_build_op_decorate(builder, val_id, SpvDecorationNoContraction, NULL, 0); @@ -10121,13 +10121,12 @@ static void spirv_compiler_emit_sample_position(struct spirv_compiler *compiler, sample_count_id = spirv_compiler_emit_query_sample_count(compiler, &instruction->src[0]); sample_index_id = spirv_compiler_emit_load_src(compiler, &instruction->src[1], VKD3DSP_WRITEMASK_0); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); index_id = vkd3d_spirv_build_op_iadd(builder, type_id, sample_count_id, sample_index_id); - index_id = vkd3d_spirv_build_op_isub(builder, - type_id, index_id, spirv_compiler_get_constant_uint(compiler, 1)); + index_id = vkd3d_spirv_build_op_isub(builder, type_id, index_id, spirv_compiler_get_constant_uint(compiler, 1)); /* Validate sample index. */ - bool_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, 1); + bool_id = spirv_get_type_id(builder, VSIR_DATA_BOOL, 1); id = vkd3d_spirv_build_op_logical_and(builder, bool_id, vkd3d_spirv_build_op_uless_than(builder, bool_id, sample_index_id, sample_count_id), vkd3d_spirv_build_op_uless_than_equal(builder, @@ -10135,16 +10134,16 @@ static void spirv_compiler_emit_sample_position(struct spirv_compiler *compiler, index_id = vkd3d_spirv_build_op_select(builder, type_id, id, index_id, spirv_compiler_get_constant_uint(compiler, 0)); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, 2); + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, 2); if (!(id = compiler->sample_positions_id)) { length_id = spirv_compiler_get_constant_uint(compiler, ARRAY_SIZE(standard_sample_positions)); array_type_id = vkd3d_spirv_get_op_type_array(builder, type_id, length_id); - for (i = 0; i < ARRAY_SIZE(standard_sample_positions); ++ i) + for (i = 0; i < ARRAY_SIZE(standard_sample_positions); ++i) { - constituents[i] = spirv_compiler_get_constant(compiler, - VKD3D_SHADER_COMPONENT_FLOAT, 2, (const uint32_t *)standard_sample_positions[i]); + constituents[i] = spirv_compiler_get_constant(compiler, VSIR_DATA_F32, + 2, (const uint32_t *)standard_sample_positions[i]); } id = vkd3d_spirv_build_op_constant_composite(builder, array_type_id, constituents, ARRAY_SIZE(constituents)); @@ -10199,7 +10198,7 @@ static void spirv_compiler_emit_eval_attrib(struct spirv_compiler *compiler, src_ids[src_count++] = spirv_compiler_emit_load_src(compiler, &src[1], VKD3DSP_WRITEMASK_0); } - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_FLOAT, + type_id = spirv_get_type_id(builder, VSIR_DATA_F32, vsir_write_mask_component_count(register_info.write_mask)); instr_set_id = vkd3d_spirv_get_glsl_std450_instr_set(builder); @@ -10334,9 +10333,8 @@ static void spirv_compiler_emit_quad_read_across(struct spirv_compiler *compiler const struct vkd3d_shader_src_param *src = instruction->src; uint32_t type_id, direction_type_id, direction_id, val_id; - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, - vsir_write_mask_component_count(dst->write_mask)); - direction_type_id = vkd3d_spirv_get_type_id_for_data_type(builder, VSIR_DATA_U32, 1); + type_id = spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)); + direction_type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask); direction_id = map_quad_read_across_direction(instruction->opcode); direction_id = vkd3d_spirv_get_op_constant(builder, direction_type_id, direction_id); @@ -10355,14 +10353,12 @@ static void spirv_compiler_emit_quad_read_lane_at(struct spirv_compiler *compile if (!register_is_constant_or_undef(&src[1].reg)) { - FIXME("Unsupported non-constant quad read lane index.\n"); spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED, "Non-constant quad read lane indices are not supported."); return; } - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, - vsir_write_mask_component_count(dst->write_mask)); + type_id = spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)); val_id = spirv_compiler_emit_load_src(compiler, &src[0], dst->write_mask); lane_id = spirv_compiler_emit_load_src(compiler, &src[1], VKD3DSP_WRITEMASK_0); val_id = vkd3d_spirv_build_op_group_nonuniform_quad_broadcast(builder, type_id, val_id, lane_id); @@ -10411,7 +10407,7 @@ static uint32_t spirv_compiler_emit_group_nonuniform_ballot(struct spirv_compile struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t type_id, val_id; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, VKD3D_VEC4_SIZE); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, VKD3D_VEC4_SIZE); val_id = spirv_compiler_emit_load_src(compiler, src, VKD3DSP_WRITEMASK_0); val_id = vkd3d_spirv_build_op_group_nonuniform_ballot(builder, type_id, val_id); @@ -10470,8 +10466,7 @@ static void spirv_compiler_emit_wave_alu_op(struct spirv_compiler *compiler, op = map_wave_alu_op(instruction->opcode, data_type_is_floating_point(src->reg.data_type)); - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, - vsir_write_mask_component_count(dst->write_mask)); + type_id = spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)); val_id = spirv_compiler_emit_load_src(compiler, &src[0], dst->write_mask); vkd3d_spirv_enable_capability(builder, SpvCapabilityGroupNonUniformArithmetic); @@ -10495,7 +10490,7 @@ static void spirv_compiler_emit_wave_bit_count(struct spirv_compiler *compiler, : SpvGroupOperationReduce; val_id = spirv_compiler_emit_group_nonuniform_ballot(compiler, instruction->src); - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1); + type_id = spirv_get_type_id(builder, VSIR_DATA_U32, 1); val_id = vkd3d_spirv_build_op_group_nonuniform_ballot_bit_count(builder, type_id, group_op, val_id); spirv_compiler_emit_store_dst(compiler, dst, val_id); @@ -10520,8 +10515,7 @@ static void spirv_compiler_emit_wave_read_lane_at(struct spirv_compiler *compile const struct vkd3d_shader_src_param *src = instruction->src; uint32_t type_id, lane_id, val_id; - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, - vsir_write_mask_component_count(dst->write_mask)); + type_id = spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)); val_id = spirv_compiler_emit_load_src(compiler, &src[0], dst->write_mask); lane_id = spirv_compiler_emit_load_src(compiler, &src[1], VKD3DSP_WRITEMASK_0); @@ -10548,8 +10542,7 @@ static void spirv_compiler_emit_wave_read_lane_first(struct spirv_compiler *comp const struct vkd3d_shader_src_param *src = instruction->src; uint32_t type_id, val_id; - type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type, - vsir_write_mask_component_count(dst->write_mask)); + type_id = spirv_get_type_id(builder, dst->reg.data_type, vsir_write_mask_component_count(dst->write_mask)); val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask); val_id = vkd3d_spirv_build_op_group_nonuniform_broadcast_first(builder, type_id, val_id); @@ -10608,12 +10601,6 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler, case VSIR_OP_DCL_VERTICES_OUT: spirv_compiler_emit_output_vertex_count(compiler, instruction); break; - case VSIR_OP_DCL_INPUT_PRIMITIVE: - spirv_compiler_emit_dcl_input_primitive(compiler, instruction); - break; - case VSIR_OP_DCL_OUTPUT_TOPOLOGY: - spirv_compiler_emit_dcl_output_topology(compiler, instruction); - break; case VSIR_OP_DCL_GS_INSTANCES: spirv_compiler_emit_dcl_gs_instances(compiler, instruction); break; @@ -11066,7 +11053,14 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler, if (program->ssa_count) spirv_compiler_allocate_ssa_register_ids(compiler, program->ssa_count); if (compiler->shader_type == VKD3D_SHADER_TYPE_COMPUTE) + { spirv_compiler_emit_thread_group_size(compiler, &program->thread_group_size); + } + else if (compiler->shader_type == VKD3D_SHADER_TYPE_GEOMETRY) + { + spirv_compiler_emit_input_primitive(compiler); + spirv_compiler_emit_output_topology(compiler); + } spirv_compiler_emit_global_flags(compiler, program->global_flags); spirv_compiler_emit_descriptor_declarations(compiler); @@ -11080,8 +11074,7 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler, { uint32_t type_id, struct_id, ptr_type_id, var_id; - type_id = vkd3d_spirv_get_type_id(builder, - vkd3d_component_type_from_data_type(parameter_data_type_map[parameter->data_type].type), + type_id = spirv_get_type_id(builder, parameter_data_type_map[parameter->data_type].type, parameter_data_type_map[parameter->data_type].component_count); struct_id = vkd3d_spirv_build_op_type_struct(builder, &type_id, 1); diff --git a/libs/vkd3d/libs/vkd3d-shader/tpf.c b/libs/vkd3d/libs/vkd3d-shader/tpf.c index ea15c1a9ad5..c6d239847c8 100644 --- a/libs/vkd3d/libs/vkd3d-shader/tpf.c +++ b/libs/vkd3d/libs/vkd3d-shader/tpf.c @@ -665,6 +665,8 @@ struct vkd3d_shader_sm4_parser { const uint32_t *start, *end, *ptr; + struct vsir_program *program; + enum vkd3d_shader_opcode phase; bool has_control_point_phase; unsigned int input_register_masks[MAX_REG_OUTPUT]; @@ -764,7 +766,7 @@ static const enum vsir_data_type data_type_table[] = static bool shader_is_sm_5_1(const struct vkd3d_shader_sm4_parser *sm4) { - const struct vkd3d_shader_version *version = &sm4->p.program->shader_version; + const struct vkd3d_shader_version *version = &sm4->program->shader_version; return version->major >= 5 && version->minor >= 1; } @@ -849,7 +851,7 @@ static void shader_sm4_read_shader_data(struct vkd3d_shader_instruction *ins, ui icb->element_count = icb_size / VKD3D_VEC4_SIZE; icb->is_null = false; memcpy(icb->data, tokens, sizeof(*tokens) * icb_size); - shader_instruction_array_add_icb(&priv->p.program->instructions, icb); + vsir_program_add_icb(priv->program, icb); ins->declaration.icb = icb; } @@ -971,7 +973,7 @@ static void shader_sm4_read_dcl_index_range(struct vkd3d_shader_instruction *ins uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *priv) { struct vkd3d_shader_index_range *index_range = &ins->declaration.index_range; - struct vsir_program *program = priv->p.program; + struct vsir_program *program = priv->program; unsigned int i, register_idx, register_count; const struct shader_signature *signature; enum vkd3d_shader_register_type type; @@ -1094,14 +1096,14 @@ static void shader_sm4_read_dcl_output_topology(struct vkd3d_shader_instruction if (ins->declaration.primitive_type.type == VKD3D_PT_UNDEFINED) FIXME("Unhandled output primitive type %#x.\n", primitive_type); - priv->p.program->output_topology = ins->declaration.primitive_type.type; + priv->program->output_topology = ins->declaration.primitive_type.type; } static void shader_sm4_read_dcl_input_primitive(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *sm4) { enum vkd3d_sm4_input_primitive_type primitive_type; - struct vsir_program *program = sm4->p.program; + struct vsir_program *program = sm4->program; primitive_type = (opcode_token & VKD3D_SM4_PRIMITIVE_TYPE_MASK) >> VKD3D_SM4_PRIMITIVE_TYPE_SHIFT; if (VKD3D_SM5_INPUT_PT_PATCH1 <= primitive_type && primitive_type <= VKD3D_SM5_INPUT_PT_PATCH32) @@ -1129,7 +1131,7 @@ static void shader_sm4_read_dcl_input_primitive(struct vkd3d_shader_instruction static void shader_sm4_read_declaration_count(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *sm4) { - struct vsir_program *program = sm4->p.program; + struct vsir_program *program = sm4->program; ins->declaration.count = *tokens; if (opcode == VKD3D_SM4_OP_DCL_TEMPS) @@ -1161,7 +1163,7 @@ static void shader_sm4_read_dcl_input_ps(struct vkd3d_shader_instruction *ins, u if (shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VSIR_DATA_F32, dst)) { struct signature_element *e = vsir_signature_find_element_for_reg( - &priv->p.program->input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); + &priv->program->input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); if (!e) { @@ -1187,7 +1189,7 @@ static void shader_sm4_read_dcl_input_ps_siv(struct vkd3d_shader_instruction *in if (shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VSIR_DATA_F32, dst)) { struct signature_element *e = vsir_signature_find_element_for_reg( - &priv->p.program->input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); + &priv->program->input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); if (!e) { @@ -1220,7 +1222,7 @@ static void shader_sm4_read_dcl_global_flags(struct vkd3d_shader_instruction *in uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *sm4) { ins->declaration.global_flags = (opcode_token & VKD3D_SM4_GLOBAL_FLAGS_MASK) >> VKD3D_SM4_GLOBAL_FLAGS_SHIFT; - sm4->p.program->global_flags = ins->declaration.global_flags; + sm4->program->global_flags = ins->declaration.global_flags; } static void shader_sm5_read_fcall(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, @@ -1256,7 +1258,7 @@ static void shader_sm5_read_dcl_interface(struct vkd3d_shader_instruction *ins, static void shader_sm5_read_control_point_count(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *sm4) { - struct vsir_program *program = sm4->p.program; + struct vsir_program *program = sm4->program; ins->declaration.count = (opcode_token & VKD3D_SM5_CONTROL_POINT_COUNT_MASK) >> VKD3D_SM5_CONTROL_POINT_COUNT_SHIFT; @@ -1272,7 +1274,7 @@ static void shader_sm5_read_dcl_tessellator_domain(struct vkd3d_shader_instructi { ins->declaration.tessellator_domain = (opcode_token & VKD3D_SM5_TESSELLATOR_MASK) >> VKD3D_SM5_TESSELLATOR_SHIFT; - priv->p.program->tess_domain = ins->declaration.tessellator_domain; + priv->program->tess_domain = ins->declaration.tessellator_domain; } static void shader_sm5_read_dcl_tessellator_partitioning(struct vkd3d_shader_instruction *ins, uint32_t opcode, @@ -1280,7 +1282,7 @@ static void shader_sm5_read_dcl_tessellator_partitioning(struct vkd3d_shader_ins { ins->declaration.tessellator_partitioning = (opcode_token & VKD3D_SM5_TESSELLATOR_MASK) >> VKD3D_SM5_TESSELLATOR_SHIFT; - priv->p.program->tess_partitioning = ins->declaration.tessellator_partitioning; + priv->program->tess_partitioning = ins->declaration.tessellator_partitioning; } static void shader_sm5_read_dcl_tessellator_output_primitive(struct vkd3d_shader_instruction *ins, uint32_t opcode, @@ -1288,7 +1290,7 @@ static void shader_sm5_read_dcl_tessellator_output_primitive(struct vkd3d_shader { ins->declaration.tessellator_output_primitive = (opcode_token & VKD3D_SM5_TESSELLATOR_MASK) >> VKD3D_SM5_TESSELLATOR_SHIFT; - priv->p.program->tess_output_primitive = ins->declaration.tessellator_output_primitive; + priv->program->tess_output_primitive = ins->declaration.tessellator_output_primitive; } static void shader_sm5_read_dcl_hs_max_tessfactor(struct vkd3d_shader_instruction *ins, uint32_t opcode, @@ -1300,7 +1302,7 @@ static void shader_sm5_read_dcl_hs_max_tessfactor(struct vkd3d_shader_instructio static void shader_sm5_read_dcl_thread_group(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *sm4) { - struct vsir_program *program = sm4->p.program; + struct vsir_program *program = sm4->program; ins->declaration.thread_group_size.x = *tokens++; ins->declaration.thread_group_size.y = *tokens++; @@ -2009,7 +2011,7 @@ static bool shader_sm4_read_reg_idx(struct vkd3d_shader_sm4_parser *priv, const { if (addressing & VKD3D_SM4_ADDRESSING_RELATIVE) { - struct vkd3d_shader_src_param *rel_addr = vsir_program_get_src_params(priv->p.program, 1); + struct vkd3d_shader_src_param *rel_addr = vsir_program_get_src_params(priv->program, 1); if (!(reg_idx->rel_addr = rel_addr)) { @@ -2284,7 +2286,7 @@ static bool register_is_control_point_input(const struct vkd3d_shader_register * { return reg->type == VKD3DSPR_INCONTROLPOINT || reg->type == VKD3DSPR_OUTCONTROLPOINT || (reg->type == VKD3DSPR_INPUT && (priv->phase == VSIR_OP_HS_CONTROL_POINT_PHASE - || priv->p.program->shader_version.type == VKD3D_SHADER_TYPE_GEOMETRY)); + || priv->program->shader_version.type == VKD3D_SHADER_TYPE_GEOMETRY)); } static uint32_t mask_from_swizzle(uint32_t swizzle) @@ -2608,8 +2610,8 @@ static void shader_sm4_read_instruction_modifier(uint32_t modifier, struct vkd3d static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, struct vkd3d_shader_instruction *ins) { const struct vkd3d_sm4_opcode_info *opcode_info; - struct vsir_program *program = sm4->p.program; uint32_t opcode_token, opcode, previous_token; + struct vsir_program *program = sm4->program; struct vkd3d_shader_dst_param *dst_params; struct vkd3d_shader_src_param *src_params; const uint32_t **ptr = &sm4->ptr; @@ -2814,8 +2816,9 @@ static bool shader_sm4_init(struct vkd3d_shader_sm4_parser *sm4, struct vsir_pro if (!vsir_program_init(program, compile_info, &version, token_count / 7u + 20, VSIR_CF_STRUCTURED, VSIR_NORMALISED_SM4)) return false; - vkd3d_shader_parser_init(&sm4->p, program, message_context, compile_info->source_name); + vkd3d_shader_parser_init(&sm4->p, message_context, compile_info->source_name); sm4->ptr = sm4->start; + sm4->program = program; init_sm4_lookup_tables(&sm4->lookup); @@ -2973,13 +2976,13 @@ int tpf_parse(const struct vkd3d_shader_compile_info *compile_info, uint64_t con } } if (program->shader_version.type == VKD3D_SHADER_TYPE_HULL - && !sm4.has_control_point_phase && !sm4.p.failed) + && !sm4.has_control_point_phase && sm4.p.status >= 0) shader_sm4_validate_default_phase_index_ranges(&sm4); - if (sm4.p.failed) + if (sm4.p.status < 0) { vsir_program_cleanup(program); - return VKD3D_ERROR_INVALID_SHADER; + return sm4.p.status; } return VKD3D_OK; @@ -3480,7 +3483,9 @@ static uint32_t sm4_encode_register(const struct tpf_compiler *tpf, const struct switch (sm4_swizzle_type) { case VKD3D_SM4_SWIZZLE_NONE: - VKD3D_ASSERT(sm4_swizzle || register_is_constant(reg)); + if (register_is_constant(reg)) + break; + VKD3D_ASSERT(sm4_swizzle); token |= (sm4_swizzle << VKD3D_SM4_WRITEMASK_SHIFT) & VKD3D_SM4_WRITEMASK_MASK; break; @@ -3692,7 +3697,7 @@ static void write_sm4_instruction(const struct tpf_compiler *tpf, const struct s static void tpf_dcl_constant_buffer(const struct tpf_compiler *tpf, const struct vkd3d_shader_instruction *ins) { const struct vkd3d_shader_constant_buffer *cb = &ins->declaration.cb; - size_t size = (cb->size + 3) / 4; + size_t size = cb->size / VKD3D_VEC4_SIZE / sizeof(float); struct sm4_instruction instr = { @@ -4222,6 +4227,7 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_ case VSIR_OP_DCL: case VSIR_OP_DCL_RESOURCE_RAW: + case VSIR_OP_DCL_RESOURCE_STRUCTURED: case VSIR_OP_DCL_UAV_RAW: case VSIR_OP_DCL_UAV_STRUCTURED: case VSIR_OP_DCL_UAV_TYPED: @@ -4242,6 +4248,7 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_ case VSIR_OP_BREAK: case VSIR_OP_CASE: case VSIR_OP_CONTINUE: + case VSIR_OP_COUNTBITS: case VSIR_OP_CUT: case VSIR_OP_CUT_STREAM: case VSIR_OP_DCL_STREAM: @@ -4267,6 +4274,9 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_ case VSIR_OP_EXP: case VSIR_OP_F16TOF32: case VSIR_OP_F32TOF16: + case VSIR_OP_FIRSTBIT_HI: + case VSIR_OP_FIRSTBIT_LO: + case VSIR_OP_FIRSTBIT_SHI: case VSIR_OP_FRC: case VSIR_OP_FTOI: case VSIR_OP_FTOU: @@ -4305,6 +4315,7 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_ case VSIR_OP_LD: case VSIR_OP_LD2DMS: case VSIR_OP_LD_RAW: + case VSIR_OP_LD_STRUCTURED: case VSIR_OP_LD_UAV_TYPED: case VSIR_OP_LOG: case VSIR_OP_LOOP: diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c index 5fcc836aae1..6087df80970 100644 --- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c +++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c @@ -23,6 +23,8 @@ #include #include +/* VKD3D_DEBUG_ENV_NAME("VKD3D_SHADER_DEBUG"); */ + static inline int char_to_int(char c) { if ('0' <= c && c <= '9') @@ -722,14 +724,13 @@ uint64_t vkd3d_shader_init_config_flags(void) return config_flags; } -void vkd3d_shader_parser_init(struct vkd3d_shader_parser *parser, struct vsir_program *program, +void vkd3d_shader_parser_init(struct vkd3d_shader_parser *parser, struct vkd3d_shader_message_context *message_context, const char *source_name) { parser->message_context = message_context; parser->location.source_name = source_name; parser->location.line = 1; parser->location.column = 0; - parser->program = program; } void VKD3D_PRINTF_FUNC(3, 4) vkd3d_shader_parser_error(struct vkd3d_shader_parser *parser, @@ -741,7 +742,32 @@ void VKD3D_PRINTF_FUNC(3, 4) vkd3d_shader_parser_error(struct vkd3d_shader_parse vkd3d_shader_verror(parser->message_context, &parser->location, error, format, args); va_end(args); - parser->failed = true; + if (parser->status >= 0) + { + switch (error) + { + case VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED: + case VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED: + case VKD3D_SHADER_ERROR_D3DBC_NOT_IMPLEMENTED: + case VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED: + case VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED: + case VKD3D_SHADER_ERROR_FX_NOT_IMPLEMENTED: + parser->status = VKD3D_ERROR_NOT_IMPLEMENTED; + break; + case VKD3D_SHADER_ERROR_DXBC_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_TPF_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_SPV_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_RS_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_D3DBC_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY: + case VKD3D_SHADER_ERROR_FX_OUT_OF_MEMORY: + parser->status = VKD3D_ERROR_OUT_OF_MEMORY; + break; + default: + parser->status = VKD3D_ERROR_INVALID_SHADER; + break; + } + } } void VKD3D_PRINTF_FUNC(3, 4) vkd3d_shader_parser_warning(struct vkd3d_shader_parser *parser, @@ -1523,7 +1549,6 @@ static int vkd3d_shader_scan_instruction(struct vkd3d_shader_scan_context *conte if (context->cf_info_count) context->cf_info[context->cf_info_count - 1].inside_block = false; break; - case VSIR_OP_TEX: case VSIR_OP_TEXBEM: case VSIR_OP_TEXBEML: case VSIR_OP_TEXDP3TEX: @@ -1683,6 +1708,7 @@ static int vsir_program_scan(struct vsir_program *program, const struct vkd3d_sh struct vsir_program_iterator it = vsir_program_iterator(&program->instructions); struct vkd3d_shader_scan_combined_resource_sampler_info *combined_sampler_info; struct vkd3d_shader_scan_hull_shader_tessellation_info *tessellation_info; + struct vkd3d_shader_scan_thread_group_size_info *thread_group_size_info; struct vkd3d_shader_scan_descriptor_info *descriptor_info; struct vkd3d_shader_scan_signature_info *signature_info; struct vkd3d_shader_scan_context context; @@ -1704,6 +1730,7 @@ static int vsir_program_scan(struct vsir_program *program, const struct vkd3d_sh } tessellation_info = vkd3d_find_struct(compile_info->next, SCAN_HULL_SHADER_TESSELLATION_INFO); + thread_group_size_info = vkd3d_find_struct(compile_info->next, SCAN_THREAD_GROUP_SIZE_INFO); vkd3d_shader_scan_context_init(&context, &program->shader_version, compile_info, add_descriptor_info ? &program->descriptors : NULL, combined_sampler_info, message_context); @@ -1756,6 +1783,13 @@ static int vsir_program_scan(struct vsir_program *program, const struct vkd3d_sh tessellation_info->partitioning = context.partitioning; } + if (!ret && thread_group_size_info) + { + thread_group_size_info->x = program->thread_group_size.x; + thread_group_size_info->y = program->thread_group_size.y; + thread_group_size_info->z = program->thread_group_size.z; + } + if (ret < 0) { if (combined_sampler_info) @@ -2182,6 +2216,9 @@ const enum vkd3d_shader_target_type *vkd3d_shader_get_supported_target_types( VKD3D_SHADER_TARGET_D3D_BYTECODE, VKD3D_SHADER_TARGET_DXBC_TPF, VKD3D_SHADER_TARGET_FX, +#ifdef VKD3D_SHADER_UNSUPPORTED_MSL + VKD3D_SHADER_TARGET_MSL, +#endif }; static const enum vkd3d_shader_target_type d3dbc_types[] = @@ -2253,6 +2290,7 @@ int vkd3d_shader_preprocess(const struct vkd3d_shader_compile_info *compile_info struct vkd3d_shader_code *out, char **messages) { struct vkd3d_shader_message_context message_context; + struct shader_dump_data dump_data; int ret; TRACE("compile_info %p, out %p, messages %p.\n", compile_info, out, messages); @@ -2265,7 +2303,11 @@ int vkd3d_shader_preprocess(const struct vkd3d_shader_compile_info *compile_info vkd3d_shader_message_context_init(&message_context, compile_info->log_level); - ret = preproc_lexer_parse(compile_info, out, &message_context); + fill_shader_dump_data(compile_info, &dump_data); + vkd3d_shader_dump_shader(&dump_data, compile_info->source.code, compile_info->source.size, SHADER_DUMP_TYPE_SOURCE); + + if ((ret = preproc_lexer_parse(compile_info, out, &message_context)) >= 0) + vkd3d_shader_dump_shader(&dump_data, out->code, out->size, SHADER_DUMP_TYPE_PREPROC); vkd3d_shader_message_context_trace_messages(&message_context); if (!vkd3d_shader_message_context_copy_messages(&message_context, messages)) @@ -2279,204 +2321,6 @@ void vkd3d_shader_set_log_callback(PFN_vkd3d_log callback) vkd3d_dbg_set_log_callback(callback); } -static struct vkd3d_shader_param_node *shader_param_allocator_node_create( - struct vkd3d_shader_param_allocator *allocator) -{ - struct vkd3d_shader_param_node *node; - - if (!(node = vkd3d_malloc(offsetof(struct vkd3d_shader_param_node, param[allocator->count * allocator->stride])))) - return NULL; - node->next = NULL; - return node; -} - -static void shader_param_allocator_init(struct vkd3d_shader_param_allocator *allocator, - size_t count, size_t stride) -{ - allocator->count = max(count, MAX_REG_OUTPUT); - allocator->stride = stride; - allocator->head = NULL; - allocator->current = NULL; - allocator->index = allocator->count; -} - -static void shader_param_allocator_destroy(struct vkd3d_shader_param_allocator *allocator) -{ - struct vkd3d_shader_param_node *current = allocator->head; - - while (current) - { - struct vkd3d_shader_param_node *next = current->next; - vkd3d_free(current); - current = next; - } -} - -void *shader_param_allocator_get(struct vkd3d_shader_param_allocator *allocator, size_t count) -{ - void *params; - - if (!allocator->current || count > allocator->count - allocator->index) - { - struct vkd3d_shader_param_node *next; - - /* Monolithic switch has no definite parameter count limit. */ - allocator->count = max(allocator->count, count); - - if (!(next = shader_param_allocator_node_create(allocator))) - return NULL; - if (allocator->current) - allocator->current->next = next; - else - allocator->head = next; - allocator->current = next; - allocator->index = 0; - } - - params = &allocator->current->param[allocator->index * allocator->stride]; - allocator->index += count; - return params; -} - -bool shader_instruction_array_init(struct vkd3d_shader_instruction_array *instructions, size_t reserve) -{ - memset(instructions, 0, sizeof(*instructions)); - /* Size the parameter initial allocations so they are large enough for most shaders. The - * code path for chained allocations will be tested if a few shaders need to use it. */ - shader_param_allocator_init(&instructions->dst_params, reserve - reserve / 8u, - sizeof(struct vkd3d_shader_dst_param)); - shader_param_allocator_init(&instructions->src_params, reserve * 2u, sizeof(struct vkd3d_shader_src_param)); - return shader_instruction_array_reserve(instructions, reserve); -} - -bool shader_instruction_array_reserve(struct vkd3d_shader_instruction_array *instructions, size_t reserve) -{ - if (!vkd3d_array_reserve((void **)&instructions->elements, &instructions->capacity, reserve, - sizeof(*instructions->elements))) - { - ERR("Failed to allocate instructions.\n"); - return false; - } - return true; -} - -bool shader_instruction_array_insert_at(struct vkd3d_shader_instruction_array *instructions, - size_t idx, size_t count) -{ - VKD3D_ASSERT(idx <= instructions->count); - - if (!shader_instruction_array_reserve(instructions, instructions->count + count)) - return false; - - memmove(&instructions->elements[idx + count], &instructions->elements[idx], - (instructions->count - idx) * sizeof(*instructions->elements)); - memset(&instructions->elements[idx], 0, count * sizeof(*instructions->elements)); - - instructions->count += count; - - return true; -} - -bool shader_instruction_array_add_icb(struct vkd3d_shader_instruction_array *instructions, - struct vkd3d_shader_immediate_constant_buffer *icb) -{ - if (!vkd3d_array_reserve((void **)&instructions->icbs, &instructions->icb_capacity, instructions->icb_count + 1, - sizeof(*instructions->icbs))) - return false; - instructions->icbs[instructions->icb_count++] = icb; - return true; -} - -static struct vkd3d_shader_src_param *shader_instruction_array_clone_src_params( - struct vkd3d_shader_instruction_array *instructions, const struct vkd3d_shader_src_param *params, - size_t count); - -static bool shader_register_clone_relative_addresses(struct vkd3d_shader_register *reg, - struct vkd3d_shader_instruction_array *instructions) -{ - unsigned int i; - - for (i = 0; i < reg->idx_count; ++i) - { - if (!reg->idx[i].rel_addr) - continue; - - if (!(reg->idx[i].rel_addr = shader_instruction_array_clone_src_params(instructions, reg->idx[i].rel_addr, 1))) - return false; - } - - return true; -} - -static struct vkd3d_shader_dst_param *shader_instruction_array_clone_dst_params( - struct vkd3d_shader_instruction_array *instructions, const struct vkd3d_shader_dst_param *params, - size_t count) -{ - struct vkd3d_shader_dst_param *dst_params; - size_t i; - - if (!(dst_params = shader_dst_param_allocator_get(&instructions->dst_params, count))) - return NULL; - - memcpy(dst_params, params, count * sizeof(*params)); - for (i = 0; i < count; ++i) - { - if (!shader_register_clone_relative_addresses(&dst_params[i].reg, instructions)) - return NULL; - } - - return dst_params; -} - -static struct vkd3d_shader_src_param *shader_instruction_array_clone_src_params( - struct vkd3d_shader_instruction_array *instructions, const struct vkd3d_shader_src_param *params, - size_t count) -{ - struct vkd3d_shader_src_param *src_params; - size_t i; - - if (!(src_params = shader_src_param_allocator_get(&instructions->src_params, count))) - return NULL; - - memcpy(src_params, params, count * sizeof(*params)); - for (i = 0; i < count; ++i) - { - if (!shader_register_clone_relative_addresses(&src_params[i].reg, instructions)) - return NULL; - } - - return src_params; -} - -/* NOTE: Immediate constant buffers are not cloned, so the source must not be destroyed while the - * destination is in use. This seems like a reasonable requirement given how this is currently used. */ -bool shader_instruction_array_clone_instruction(struct vkd3d_shader_instruction_array *instructions, - size_t dst, size_t src) -{ - struct vkd3d_shader_instruction *ins = &instructions->elements[dst]; - - *ins = instructions->elements[src]; - - if (ins->dst_count && ins->dst && !(ins->dst = shader_instruction_array_clone_dst_params(instructions, - ins->dst, ins->dst_count))) - return false; - - return !ins->src_count || !!(ins->src = shader_instruction_array_clone_src_params(instructions, - ins->src, ins->src_count)); -} - -void shader_instruction_array_destroy(struct vkd3d_shader_instruction_array *instructions) -{ - unsigned int i; - - vkd3d_free(instructions->elements); - shader_param_allocator_destroy(&instructions->dst_params); - shader_param_allocator_destroy(&instructions->src_params); - for (i = 0; i < instructions->icb_count; ++i) - vkd3d_free(instructions->icbs[i]); - vkd3d_free(instructions->icbs); -} - void vkd3d_shader_build_varying_map(const struct vkd3d_shader_signature *output_signature, const struct vkd3d_shader_signature *input_signature, unsigned int *ret_count, struct vkd3d_shader_varying_map *varyings) diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h index e758c16b3d4..3637aa4fdb9 100644 --- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h @@ -213,7 +213,7 @@ enum vkd3d_shader_error VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND_COUNT = 8005, VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_TABLE = 8006, VKD3D_SHADER_ERROR_DXIL_INVALID_VALUE_SYMTAB = 8007, - VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED_BITCODE_FORMAT = 8008, + VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED = 8008, VKD3D_SHADER_ERROR_DXIL_INVALID_FUNCTION_DCL = 8009, VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_ID = 8010, VKD3D_SHADER_ERROR_DXIL_INVALID_MODULE = 8011, @@ -226,6 +226,7 @@ enum vkd3d_shader_error VKD3D_SHADER_ERROR_DXIL_INVALID_RESOURCES = 8018, VKD3D_SHADER_ERROR_DXIL_INVALID_RESOURCE_HANDLE = 8019, VKD3D_SHADER_ERROR_DXIL_INVALID_CONSTANT = 8020, + VKD3D_SHADER_ERROR_DXIL_NOT_IMPLEMENTED = 8021, VKD3D_SHADER_WARNING_DXIL_UNKNOWN_MAGIC_NUMBER = 8300, VKD3D_SHADER_WARNING_DXIL_UNKNOWN_SHADER_TYPE = 8301, @@ -716,7 +717,10 @@ enum vsir_data_type VSIR_DATA_F32, VSIR_DATA_F64, + VSIR_DATA_I8, + VSIR_DATA_I16, VSIR_DATA_I32, + VSIR_DATA_I64, VSIR_DATA_U8, VSIR_DATA_U16, @@ -734,10 +738,18 @@ enum vsir_data_type VSIR_DATA_TYPE_COUNT, }; +const char *vsir_data_type_get_name(enum vsir_data_type t, const char *error); + static inline bool data_type_is_integer(enum vsir_data_type data_type) { - return data_type == VSIR_DATA_I32 || data_type == VSIR_DATA_U8 || data_type == VSIR_DATA_U16 - || data_type == VSIR_DATA_U32 || data_type == VSIR_DATA_U64; + return data_type == VSIR_DATA_I8 + || data_type == VSIR_DATA_I16 + || data_type == VSIR_DATA_I32 + || data_type == VSIR_DATA_I64 + || data_type == VSIR_DATA_U8 + || data_type == VSIR_DATA_U16 + || data_type == VSIR_DATA_U32 + || data_type == VSIR_DATA_U64; } static inline bool data_type_is_bool(enum vsir_data_type data_type) @@ -752,7 +764,7 @@ static inline bool data_type_is_floating_point(enum vsir_data_type data_type) static inline bool data_type_is_64_bit(enum vsir_data_type data_type) { - return data_type == VSIR_DATA_F64 || data_type == VSIR_DATA_U64; + return data_type == VSIR_DATA_F64 || data_type == VSIR_DATA_I64 || data_type == VSIR_DATA_U64; } enum vsir_dimension @@ -1357,6 +1369,13 @@ static inline bool vkd3d_shader_ver_le(const struct vkd3d_shader_version *v, uns void vsir_instruction_init(struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location, enum vkd3d_shader_opcode opcode); +static inline void vkd3d_shader_instruction_make_nop(struct vkd3d_shader_instruction *ins) +{ + struct vkd3d_shader_location location = ins->location; + + vsir_instruction_init(ins, &location, VSIR_OP_NOP); +} + static inline bool vkd3d_shader_instruction_has_texel_offset(const struct vkd3d_shader_instruction *ins) { return ins->texel_offset.u || ins->texel_offset.v || ins->texel_offset.w; @@ -1410,44 +1429,15 @@ struct vkd3d_shader_param_allocator void *shader_param_allocator_get(struct vkd3d_shader_param_allocator *allocator, size_t count); -static inline struct vkd3d_shader_src_param *shader_src_param_allocator_get( - struct vkd3d_shader_param_allocator *allocator, size_t count) -{ - VKD3D_ASSERT(allocator->stride == sizeof(struct vkd3d_shader_src_param)); - return shader_param_allocator_get(allocator, count); -} - -static inline struct vkd3d_shader_dst_param *shader_dst_param_allocator_get( - struct vkd3d_shader_param_allocator *allocator, size_t count) -{ - VKD3D_ASSERT(allocator->stride == sizeof(struct vkd3d_shader_dst_param)); - return shader_param_allocator_get(allocator, count); -} - struct vkd3d_shader_instruction_array { struct vkd3d_shader_instruction *elements; size_t capacity; size_t count; - - struct vkd3d_shader_param_allocator src_params; - struct vkd3d_shader_param_allocator dst_params; - struct vkd3d_shader_immediate_constant_buffer **icbs; - size_t icb_capacity; - size_t icb_count; - - struct vkd3d_shader_src_param *outpointid_param; }; -bool shader_instruction_array_init(struct vkd3d_shader_instruction_array *instructions, size_t reserve); -bool shader_instruction_array_reserve(struct vkd3d_shader_instruction_array *instructions, size_t reserve); -bool shader_instruction_array_insert_at(struct vkd3d_shader_instruction_array *instructions, - size_t idx, size_t count); -bool shader_instruction_array_add_icb(struct vkd3d_shader_instruction_array *instructions, - struct vkd3d_shader_immediate_constant_buffer *icb); -bool shader_instruction_array_clone_instruction(struct vkd3d_shader_instruction_array *instructions, - size_t dst, size_t src); -void shader_instruction_array_destroy(struct vkd3d_shader_instruction_array *instructions); +struct vkd3d_shader_instruction *shader_instruction_array_append(struct vkd3d_shader_instruction_array *array); +bool shader_instruction_array_insert_at(struct vkd3d_shader_instruction_array *instructions, size_t idx, size_t count); struct vsir_program_iterator { @@ -1503,13 +1493,46 @@ static inline struct vkd3d_shader_instruction *vsir_program_iterator_prev( } /* When insertion takes place, argument `it' is updated to point to the same - * instruction as before the insertion, but all other iterators and pointers - * to the same container are invalidated and cannot be used any more. */ + * instruction as before the insertion, but all existing pointers to the same + * container, as well as any iterators pointing to instructions after the + * insertion point should be considered invalid. */ static inline bool vsir_program_iterator_insert_after(struct vsir_program_iterator *it, size_t count) { return shader_instruction_array_insert_at(it->array, it->idx + 1, count); } +/* When insertion takes place, argument `it' is updated to point to the same + * instruction as before the insertion, and the optional argument `ins_it' is + * initialized to point to the first inserted instruction. + * A pointer to the first inserted instruction is returned. */ +static inline struct vkd3d_shader_instruction *vsir_program_iterator_insert_before( + struct vsir_program_iterator *it, struct vsir_program_iterator *ins_it, size_t count) +{ + VKD3D_ASSERT(it != ins_it); + VKD3D_ASSERT(it->idx != SIZE_MAX); + + if (!shader_instruction_array_insert_at(it->array, it->idx, count)) + return NULL; + + *ins_it = *it; + it->idx += count; + + return vsir_program_iterator_current(ins_it); +} + +/* When insertion takes place, argument `it' is updated to point to the first + * inserted instruction. A pointer to this first inserted instruction is + * returned. */ +static inline struct vkd3d_shader_instruction *vsir_program_iterator_insert_before_and_move( + struct vsir_program_iterator *it, size_t count) +{ + VKD3D_ASSERT(it->idx != SIZE_MAX); + + if (!shader_instruction_array_insert_at(it->array, it->idx, count)) + return NULL; + return vsir_program_iterator_current(it); +} + enum vkd3d_shader_config_flags { VKD3D_SHADER_CONFIG_FLAG_FORCE_VALIDATION = 0x00000001, @@ -1575,7 +1598,7 @@ struct vsir_program unsigned int input_control_point_count, output_control_point_count; struct vsir_thread_group_size thread_group_size; unsigned int flat_constant_count[3]; - unsigned int block_count; + unsigned int block_count; /* maximum block count in any function */ unsigned int temp_count; unsigned int ssa_count; enum vsir_global_flags global_flags; @@ -1599,12 +1622,21 @@ struct vsir_program struct vkd3d_shader_source_list source_files; const char **block_names; size_t block_name_count; + + struct vkd3d_shader_immediate_constant_buffer **icbs; + size_t icb_capacity; + size_t icb_count; + + struct vkd3d_shader_param_allocator src_params; + struct vkd3d_shader_param_allocator dst_params; }; enum vkd3d_result vsir_allocate_temp_registers(struct vsir_program *program, struct vkd3d_shader_message_context *message_context); enum vkd3d_result vsir_update_dcl_temps(struct vsir_program *program, struct vkd3d_shader_message_context *message_context); + +bool vsir_program_add_icb(struct vsir_program *program, struct vkd3d_shader_immediate_constant_buffer *icb); void vsir_program_cleanup(struct vsir_program *program); const struct vkd3d_shader_parameter1 *vsir_program_get_parameter( const struct vsir_program *program, enum vkd3d_shader_parameter_name name); @@ -1627,37 +1659,39 @@ bool vsir_instruction_init_with_params(struct vsir_program *program, static inline struct vkd3d_shader_instruction *vsir_program_append(struct vsir_program *program) { - struct vkd3d_shader_instruction_array *array = &program->instructions; - - if (!shader_instruction_array_insert_at(array, array->count, 1)) - return NULL; - - return &array->elements[array->count - 1]; + return shader_instruction_array_append(&program->instructions); } static inline struct vkd3d_shader_dst_param *vsir_program_get_dst_params( struct vsir_program *program, unsigned int count) { - return shader_dst_param_allocator_get(&program->instructions.dst_params, count); + struct vkd3d_shader_param_allocator *allocator = &program->dst_params; + + VKD3D_ASSERT(allocator->stride == sizeof(struct vkd3d_shader_dst_param)); + + return shader_param_allocator_get(allocator, count); } static inline struct vkd3d_shader_src_param *vsir_program_get_src_params( struct vsir_program *program, unsigned int count) { - return shader_src_param_allocator_get(&program->instructions.src_params, count); + struct vkd3d_shader_param_allocator *allocator = &program->src_params; + + VKD3D_ASSERT(allocator->stride == sizeof(struct vkd3d_shader_src_param)); + + return shader_param_allocator_get(allocator, count); } struct vkd3d_shader_parser { struct vkd3d_shader_message_context *message_context; struct vkd3d_shader_location location; - struct vsir_program *program; - bool failed; + enum vkd3d_result status; }; void vkd3d_shader_parser_error(struct vkd3d_shader_parser *parser, enum vkd3d_shader_error error, const char *format, ...) VKD3D_PRINTF_FUNC(3, 4); -void vkd3d_shader_parser_init(struct vkd3d_shader_parser *parser, struct vsir_program *program, +void vkd3d_shader_parser_init(struct vkd3d_shader_parser *parser, struct vkd3d_shader_message_context *message_context, const char *source_name); void vkd3d_shader_parser_warning(struct vkd3d_shader_parser *parser, enum vkd3d_shader_error error, const char *format, ...) VKD3D_PRINTF_FUNC(3, 4); @@ -1852,20 +1886,28 @@ static inline enum vkd3d_shader_component_type vkd3d_component_type_from_data_ty { case VSIR_DATA_BOOL: return VKD3D_SHADER_COMPONENT_BOOL; - case VSIR_DATA_F16: /* Minimum precision. TODO: native 16-bit */ + case VSIR_DATA_F16: + return VKD3D_SHADER_COMPONENT_FLOAT16; case VSIR_DATA_F32: case VSIR_DATA_SNORM: case VSIR_DATA_UNORM: return VKD3D_SHADER_COMPONENT_FLOAT; case VSIR_DATA_F64: return VKD3D_SHADER_COMPONENT_DOUBLE; + case VSIR_DATA_I16: + return VKD3D_SHADER_COMPONENT_INT16; case VSIR_DATA_I32: return VKD3D_SHADER_COMPONENT_INT; - case VSIR_DATA_U16: /* Minimum precision. TODO: native 16-bit */ + case VSIR_DATA_I64: + return VKD3D_SHADER_COMPONENT_INT64; + case VSIR_DATA_U16: + return VKD3D_SHADER_COMPONENT_UINT16; case VSIR_DATA_U32: return VKD3D_SHADER_COMPONENT_UINT; case VSIR_DATA_U64: return VKD3D_SHADER_COMPONENT_UINT64; + case VSIR_DATA_UNUSED: + return VKD3D_SHADER_COMPONENT_VOID; default: FIXME("Unhandled data type %#x.\n", data_type); /* fall-through */ @@ -1878,23 +1920,41 @@ static inline enum vsir_data_type vsir_data_type_from_component_type(enum vkd3d_ { switch (component_type) { - case VKD3D_SHADER_COMPONENT_FLOAT: - return VSIR_DATA_F32; + case VKD3D_SHADER_COMPONENT_VOID: + return VSIR_DATA_UNUSED; case VKD3D_SHADER_COMPONENT_UINT: return VSIR_DATA_U32; case VKD3D_SHADER_COMPONENT_INT: return VSIR_DATA_I32; + case VKD3D_SHADER_COMPONENT_FLOAT: + return VSIR_DATA_F32; + case VKD3D_SHADER_COMPONENT_BOOL: + return VSIR_DATA_BOOL; case VKD3D_SHADER_COMPONENT_DOUBLE: return VSIR_DATA_F64; - default: - FIXME("Unhandled component type %#x.\n", component_type); - return VSIR_DATA_F32; + case VKD3D_SHADER_COMPONENT_UINT64: + return VSIR_DATA_U64; + case VKD3D_SHADER_COMPONENT_INT64: + return VSIR_DATA_I64; + case VKD3D_SHADER_COMPONENT_FLOAT16: + return VSIR_DATA_F16; + case VKD3D_SHADER_COMPONENT_UINT16: + return VSIR_DATA_U16; + case VKD3D_SHADER_COMPONENT_INT16: + return VSIR_DATA_I16; + case VKD3D_SHADER_COMPONENT_TYPE_FORCE_32BIT: + break; } + + FIXME("Unhandled component type %#x.\n", component_type); + + return VSIR_DATA_UNUSED; } static inline bool component_type_is_64_bit(enum vkd3d_shader_component_type component_type) { - return component_type == VKD3D_SHADER_COMPONENT_DOUBLE || component_type == VKD3D_SHADER_COMPONENT_UINT64; + return component_type == VKD3D_SHADER_COMPONENT_DOUBLE || component_type == VKD3D_SHADER_COMPONENT_INT64 + || component_type == VKD3D_SHADER_COMPONENT_UINT64; } static inline unsigned int vsir_write_mask_get_component_idx(uint32_t write_mask) diff --git a/libs/vkd3d/libs/vkd3d-utils/vkd3d_utils_main.c b/libs/vkd3d/libs/vkd3d-utils/vkd3d_utils_main.c index f2967835b62..fea8c2440d1 100644 --- a/libs/vkd3d/libs/vkd3d-utils/vkd3d_utils_main.c +++ b/libs/vkd3d/libs/vkd3d-utils/vkd3d_utils_main.c @@ -19,6 +19,8 @@ #include "vkd3d_utils_private.h" #undef D3D12CreateDevice +/* VKD3D_DEBUG_ENV_NAME("VKD3D_DEBUG"); */ + static const char *debug_d3d_blob_part(D3D_BLOB_PART part) { switch (part) @@ -269,7 +271,7 @@ HRESULT WINAPI D3DCompile2VKD3D(const void *data, SIZE_T data_size, const char * option = &options[0]; option->name = VKD3D_SHADER_COMPILE_OPTION_API_VERSION; - option->value = VKD3D_SHADER_API_VERSION_1_17; + option->value = VKD3D_SHADER_API_VERSION_CURRENT; compile_info.type = VKD3D_SHADER_STRUCTURE_TYPE_COMPILE_INFO; compile_info.next = &preprocess_info; @@ -433,7 +435,7 @@ HRESULT WINAPI D3DPreprocess(const void *data, SIZE_T size, const char *filename static const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, }; TRACE("data %p, size %"PRIuPTR", filename %s, macros %p, include %p, preprocessed_blob %p, messages_blob %p.\n", @@ -482,7 +484,10 @@ HRESULT WINAPI D3DPreprocess(const void *data, SIZE_T size, const char *filename if (!ret) { - if (FAILED(hr = vkd3d_blob_create((void *)preprocessed_code.code, preprocessed_code.size, preprocessed_blob))) + /* vkd3d-shader output is null-terminated, but the null terminator isn't + * included in the size. Increase the size to account for that. */ + if (FAILED(hr = vkd3d_blob_create((void *)preprocessed_code.code, + preprocessed_code.size + 1, preprocessed_blob))) { vkd3d_shader_free_shader_code(&preprocessed_code); return hr; @@ -979,7 +984,7 @@ HRESULT WINAPI D3DDisassemble(const void *data, SIZE_T data_size, static const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, }; TRACE("data %p, data_size %"PRIuPTR", flags %#x, comments %p, blob %p.\n", @@ -1032,7 +1037,9 @@ HRESULT WINAPI D3DDisassemble(const void *data, SIZE_T data_size, return hresult_from_vkd3d_result(ret); } - if (FAILED(hr = vkd3d_blob_create((void *)output.code, output.size, blob))) + /* vkd3d-shader output is null-terminated, but the null terminator isn't + * included in the size. Increase the size to account for that. */ + if (FAILED(hr = vkd3d_blob_create((void *)output.code, output.size + 1, blob))) vkd3d_shader_free_shader_code(&output); return hr; diff --git a/libs/vkd3d/libs/vkd3d/command.c b/libs/vkd3d/libs/vkd3d/command.c index 074d8430585..69f42280e8a 100644 --- a/libs/vkd3d/libs/vkd3d/command.c +++ b/libs/vkd3d/libs/vkd3d/command.c @@ -2688,32 +2688,12 @@ static void d3d12_command_list_reset_state(struct d3d12_command_list *list, static HRESULT STDMETHODCALLTYPE d3d12_command_list_Reset(ID3D12GraphicsCommandList6 *iface, ID3D12CommandAllocator *allocator, ID3D12PipelineState *initial_pipeline_state) { - struct d3d12_command_allocator *allocator_impl = unsafe_impl_from_ID3D12CommandAllocator(allocator); struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList6(iface); - HRESULT hr; TRACE("iface %p, allocator %p, initial_pipeline_state %p.\n", iface, allocator, initial_pipeline_state); - if (!allocator_impl) - { - WARN("Command allocator is NULL.\n"); - return E_INVALIDARG; - } - - if (list->is_recording) - { - WARN("Command list is in the recording state.\n"); - return E_FAIL; - } - - if (SUCCEEDED(hr = d3d12_command_allocator_allocate_command_buffer(allocator_impl, list))) - { - list->allocator = allocator_impl; - d3d12_command_list_reset_state(list, initial_pipeline_state); - } - - return hr; + return d3d12_command_list_reset(list, allocator, initial_pipeline_state); } static void STDMETHODCALLTYPE d3d12_command_list_ClearState(ID3D12GraphicsCommandList6 *iface, @@ -3049,7 +3029,7 @@ static void d3d12_command_list_update_descriptor_table(struct d3d12_command_list if (descriptor_count > range->vk_binding_count) { - ERR("Heap descriptor count %u exceeds maximum Vulkan count %u. Reducing to the Vulkan maximum.\n", + MESSAGE("Heap descriptor count %u exceeds maximum Vulkan count %u. Reducing to the Vulkan maximum.\n", descriptor_count, range->vk_binding_count); descriptor_count = range->vk_binding_count; } @@ -6405,9 +6385,8 @@ static struct d3d12_command_list *unsafe_impl_from_ID3D12CommandList(ID3D12Comma return CONTAINING_RECORD(iface, struct d3d12_command_list, ID3D12GraphicsCommandList6_iface); } -static HRESULT d3d12_command_list_init(struct d3d12_command_list *list, struct d3d12_device *device, - D3D12_COMMAND_LIST_TYPE type, struct d3d12_command_allocator *allocator, - ID3D12PipelineState *initial_pipeline_state) +static HRESULT d3d12_command_list_init(struct d3d12_command_list *list, + struct d3d12_device *device, D3D12_COMMAND_LIST_TYPE type) { HRESULT hr; @@ -6421,31 +6400,37 @@ static HRESULT d3d12_command_list_init(struct d3d12_command_list *list, struct d d3d12_device_add_ref(list->device = device); - list->allocator = allocator; + return hr; +} - list->descriptor_heap_count = 0; +HRESULT d3d12_command_list_create(struct d3d12_device *device, UINT node_mask, + D3D12_COMMAND_LIST_TYPE type, struct d3d12_command_list **list) +{ + struct d3d12_command_list *object; + HRESULT hr; - if (SUCCEEDED(hr = d3d12_command_allocator_allocate_command_buffer(allocator, list))) - { - list->pipeline_bindings[VKD3D_PIPELINE_BIND_POINT_GRAPHICS].vk_uav_counter_views = NULL; - list->pipeline_bindings[VKD3D_PIPELINE_BIND_POINT_COMPUTE].vk_uav_counter_views = NULL; - d3d12_command_list_reset_state(list, initial_pipeline_state); - } - else + debug_ignored_node_mask(node_mask); + + if (!(object = vkd3d_calloc(1, sizeof(*object)))) + return E_OUTOFMEMORY; + + if (FAILED(hr = d3d12_command_list_init(object, device, type))) { - vkd3d_private_store_destroy(&list->private_store); - d3d12_device_release(device); + vkd3d_free(object); + return hr; } - return hr; + TRACE("Created command list %p.\n", object); + + *list = object; + + return S_OK; } -HRESULT d3d12_command_list_create(struct d3d12_device *device, - UINT node_mask, D3D12_COMMAND_LIST_TYPE type, ID3D12CommandAllocator *allocator_iface, - ID3D12PipelineState *initial_pipeline_state, struct d3d12_command_list **list) +HRESULT d3d12_command_list_reset(struct d3d12_command_list *list, ID3D12CommandAllocator *allocator_iface, + ID3D12PipelineState *initial_pipeline_state) { struct d3d12_command_allocator *allocator; - struct d3d12_command_list *object; HRESULT hr; if (!(allocator = unsafe_impl_from_ID3D12CommandAllocator(allocator_iface))) @@ -6454,29 +6439,26 @@ HRESULT d3d12_command_list_create(struct d3d12_device *device, return E_INVALIDARG; } - if (allocator->type != type) + if (allocator->type != list->type) { WARN("Command list types do not match (allocator %#x, list %#x).\n", - allocator->type, type); + allocator->type, list->type); return E_INVALIDARG; } - debug_ignored_node_mask(node_mask); - - if (!(object = vkd3d_malloc(sizeof(*object)))) - return E_OUTOFMEMORY; - - if (FAILED(hr = d3d12_command_list_init(object, device, type, allocator, initial_pipeline_state))) + if (list->is_recording) { - vkd3d_free(object); - return hr; + WARN("Command list is in the recording state.\n"); + return E_FAIL; } - TRACE("Created command list %p.\n", object); - - *list = object; + if (SUCCEEDED(hr = d3d12_command_allocator_allocate_command_buffer(allocator, list))) + { + list->allocator = allocator; + d3d12_command_list_reset_state(list, initial_pipeline_state); + } - return S_OK; + return hr; } /* ID3D12CommandQueue */ diff --git a/libs/vkd3d/libs/vkd3d/device.c b/libs/vkd3d/libs/vkd3d/device.c index 67f84aafa29..6af5e2a5c7d 100644 --- a/libs/vkd3d/libs/vkd3d/device.c +++ b/libs/vkd3d/libs/vkd3d/device.c @@ -3271,10 +3271,15 @@ static HRESULT STDMETHODCALLTYPE d3d12_device_CreateCommandList(ID3D12Device9 *i iface, node_mask, type, command_allocator, initial_pipeline_state, debugstr_guid(riid), command_list); - if (FAILED(hr = d3d12_command_list_create(device, node_mask, type, command_allocator, - initial_pipeline_state, &object))) + if (FAILED(hr = d3d12_command_list_create(device, node_mask, type, &object))) return hr; + if (FAILED(hr = d3d12_command_list_reset(object, command_allocator, initial_pipeline_state))) + { + ID3D12GraphicsCommandList6_Release(&object->ID3D12GraphicsCommandList6_iface); + return hr; + } + return return_interface(&object->ID3D12GraphicsCommandList6_iface, &IID_ID3D12GraphicsCommandList6, riid, command_list); } @@ -5082,10 +5087,21 @@ static HRESULT STDMETHODCALLTYPE d3d12_device_CreateCommandList1(ID3D12Device9 * UINT node_mask, D3D12_COMMAND_LIST_TYPE type, D3D12_COMMAND_LIST_FLAGS flags, REFIID iid, void **command_list) { - FIXME("iface %p, node_mask 0x%08x, type %#x, flags %#x, iid %s, command_list %p stub!\n", + struct d3d12_device *device = impl_from_ID3D12Device9(iface); + struct d3d12_command_list *object; + HRESULT hr; + + TRACE("iface %p, node_mask 0x%08x, type %#x, flags %#x, iid %s, command_list %p.\n", iface, node_mask, type, flags, debugstr_guid(iid), command_list); - return E_NOTIMPL; + if (flags) + FIXME("Ignoring flags %#x.\n", flags); + + if (FAILED(hr = d3d12_command_list_create(device, node_mask, type, &object))) + return hr; + + return return_interface(&object->ID3D12GraphicsCommandList6_iface, + &IID_ID3D12GraphicsCommandList6, iid, command_list); } static HRESULT STDMETHODCALLTYPE d3d12_device_CreateProtectedResourceSession(ID3D12Device9 *iface, diff --git a/libs/vkd3d/libs/vkd3d/state.c b/libs/vkd3d/libs/vkd3d/state.c index 6bbd6533b74..b6055a50a99 100644 --- a/libs/vkd3d/libs/vkd3d/state.c +++ b/libs/vkd3d/libs/vkd3d/state.c @@ -2391,7 +2391,7 @@ static HRESULT create_shader_stage(struct d3d12_device *device, const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, {VKD3D_SHADER_COMPILE_OPTION_TYPED_UAV, typed_uav_compile_option(device)}, {VKD3D_SHADER_COMPILE_OPTION_WRITE_TESS_GEOM_POINT_SIZE, 0}, {VKD3D_SHADER_COMPILE_OPTION_FEATURE, feature_flags_compile_option(device)}, @@ -2456,7 +2456,7 @@ static int vkd3d_scan_dxbc(const struct d3d12_device *device, const D3D12_SHADER const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, {VKD3D_SHADER_COMPILE_OPTION_TYPED_UAV, typed_uav_compile_option(device)}, }; @@ -4135,7 +4135,7 @@ static int compile_hlsl_cs(const struct vkd3d_shader_code *hlsl, struct vkd3d_sh static const struct vkd3d_shader_compile_option options[] = { - {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_1_17}, + {VKD3D_SHADER_COMPILE_OPTION_API_VERSION, VKD3D_SHADER_API_VERSION_CURRENT}, }; info.type = VKD3D_SHADER_STRUCTURE_TYPE_COMPILE_INFO; diff --git a/libs/vkd3d/libs/vkd3d/utils.c b/libs/vkd3d/libs/vkd3d/utils.c index c2832a61f67..2d0510e5456 100644 --- a/libs/vkd3d/libs/vkd3d/utils.c +++ b/libs/vkd3d/libs/vkd3d/utils.c @@ -703,7 +703,7 @@ const char *debug_vk_extent_3d(VkExtent3D extent) const char *debug_vk_queue_flags(VkQueueFlags flags) { - char buffer[191]; + char buffer[222]; buffer[0] = '\0'; #define FLAG_TO_STR(f) if (flags & f) { strcat(buffer, " | "#f); flags &= ~f; } @@ -716,6 +716,7 @@ const char *debug_vk_queue_flags(VkQueueFlags flags) #define FLAG_TO_STR(f, n) if (flags & f) { strcat(buffer, " | "#n); flags &= ~f; } FLAG_TO_STR(0x20, VK_QUEUE_VIDEO_DECODE_BIT_KHR) FLAG_TO_STR(0x40, VK_QUEUE_VIDEO_ENCODE_BIT_KHR) + FLAG_TO_STR(0x100, VK_QUEUE_OPTICAL_FLOW_BIT_NV) #undef FLAG_TO_STR if (flags) FIXME("Unrecognized flag(s) %#x.\n", flags); diff --git a/libs/vkd3d/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/libs/vkd3d/vkd3d_private.h index 9fb6834158f..0a8c5aef674 100644 --- a/libs/vkd3d/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/libs/vkd3d/vkd3d_private.h @@ -1327,8 +1327,9 @@ struct d3d12_command_list }; HRESULT d3d12_command_list_create(struct d3d12_device *device, - UINT node_mask, D3D12_COMMAND_LIST_TYPE type, ID3D12CommandAllocator *allocator_iface, - ID3D12PipelineState *initial_pipeline_state, struct d3d12_command_list **list); + UINT node_mask, D3D12_COMMAND_LIST_TYPE type, struct d3d12_command_list **list); +HRESULT d3d12_command_list_reset(struct d3d12_command_list *list, + ID3D12CommandAllocator *allocator_iface, ID3D12PipelineState *initial_pipeline_state); struct vkd3d_queue { -- 2.51.0