Skip to content

G
glslang
Commit 06a0ba3c by John Kessenich
Options

Merge remote-tracking branch 'GitHub/master'

parents c4f5ec87 ed834895
Showing with 270 additions and 49 deletions
CMakeLists.txt
......@@ -94,9 +94,10 @@ endif()
if(ENABLE_OPT)
message(STATUS "optimizer enabled")
add_definitions(-DENABLE_OPT)
add_definitions(-DENABLE_OPT=1)
elseif(ENABLE_HLSL)
message(STATUS "spirv-tools not linked - illegal SPIRV may be generated for HLSL")
add_definitions(-DENABLE_OPT=0)
endif()
add_subdirectory(glslang)
......
README.md
......@@ -197,8 +197,11 @@ bool InitializeProcess();
void FinalizeProcess();
class TShader
setStrings(...);
setEnvInput(EShSourceHlsl or EShSourceGlsl, stage, EShClientVulkan or EShClientOpenGL, 100);
setEnvClient(EShClientVulkan or EShClientOpenGL, EShTargetVulkan_1_0 or EShTargetVulkan_1_1 or EShTargetOpenGL_450);
setEnvTarget(EShTargetSpv, EShTargetSpv_1_0 or EShTargetSpv_1_3);
bool parse(...);
void setStrings(...);
const char* getInfoLog();
class TProgram
......
SPIRV/CMakeLists.txt
......@@ -53,7 +53,7 @@ if(ENABLE_OPT)
PRIVATE ${spirv-tools_SOURCE_DIR}/include
PRIVATE ${spirv-tools_SOURCE_DIR}/source
)
target_link_libraries(SPIRV glslang SPIRV-Tools-opt SPVRemapper)
target_link_libraries(SPIRV glslang SPIRV-Tools-opt)
else()
target_link_libraries(SPIRV glslang)
endif(ENABLE_OPT)
......
SPIRV/GLSL.ext.NV.h
......@@ -51,4 +51,7 @@ const char* const E_SPV_NV_stereo_view_rendering = "SPV_NV_stereo_view_rendering
//SPV_NVX_multiview_per_view_attributes
const char* const E_SPV_NVX_multiview_per_view_attributes = "SPV_NVX_multiview_per_view_attributes";
//SPV_NV_shader_subgroup_partitioned
const char* const E_SPV_NV_shader_subgroup_partitioned = "SPV_NV_shader_subgroup_partitioned";
#endif // #ifndef GLSLextNV_H
\ No newline at end of file
SPIRV/GlslangToSpv.cpp 100644 → 100755
This diff is collapsed. Click to expand it.
SPIRV/SpvBuilder.cpp
......@@ -854,9 +854,9 @@ Id Builder::makeFpConstant(Id type, double d, bool specConstant)
switch (getScalarTypeWidth(type)) {
case 16:
return makeFloat16Constant(d, specConstant);
return makeFloat16Constant((float)d, specConstant);
case 32:
return makeFloatConstant(d, specConstant);
return makeFloatConstant((float)d, specConstant);
case 64:
return makeDoubleConstant(d, specConstant);
default:
......
SPIRV/doc.cpp 100644 → 100755
This diff is collapsed. Click to expand it.
SPIRV/doc.h
......@@ -190,7 +190,6 @@ protected:
class EnumParameters {
public:
EnumParameters() : desc(0) { }
EnumCaps caps;
const char* desc;
};
......@@ -235,7 +234,6 @@ public:
bool hasType() const { return typePresent != 0; }
const char* opDesc;
EnumCaps capabilities;
OpcodeClass opClass;
OperandParameters operands;
......@@ -244,8 +242,6 @@ protected:
int resultPresent : 1;
};
const int OpcodeCeiling = 321;
// The set of objects that hold all the instruction/operand
// parameterization information.
extern InstructionParameters InstructionDesc[];
......
SPIRV/spirv.hpp
......@@ -573,6 +573,9 @@ enum GroupOperation {
GroupOperationInclusiveScan = 1,
GroupOperationExclusiveScan = 2,
GroupOperationClusteredReduce = 3,
GroupOperationPartitionedReduceNV = 6,
GroupOperationPartitionedInclusiveScanNV = 7,
GroupOperationPartitionedExclusiveScanNV = 8,
GroupOperationMax = 0x7fffffff,
};
......@@ -701,6 +704,7 @@ enum Capability {
CapabilityInputAttachmentArrayNonUniformIndexingEXT = 5310,
CapabilityUniformTexelBufferArrayNonUniformIndexingEXT = 5311,
CapabilityStorageTexelBufferArrayNonUniformIndexingEXT = 5312,
CapabilityGroupNonUniformPartitionedNV = 5297,
CapabilitySubgroupShuffleINTEL = 5568,
CapabilitySubgroupBufferBlockIOINTEL = 5569,
CapabilitySubgroupImageBlockIOINTEL = 5570,
......@@ -1064,6 +1068,7 @@ enum Op {
OpGroupSMaxNonUniformAMD = 5007,
OpFragmentMaskFetchAMD = 5011,
OpFragmentFetchAMD = 5012,
OpGroupNonUniformPartitionNV = 5296,
OpSubgroupShuffleINTEL = 5571,
OpSubgroupShuffleDownINTEL = 5572,
OpSubgroupShuffleUpINTEL = 5573,
......
StandAlone/StandAlone.cpp
......@@ -574,7 +574,7 @@ void ProcessArguments(std::vector<std::unique_ptr<glslang::TWorkItem>>& workItem
if (argv[0][2] == 'd')
Options |= EOptionOptimizeDisable;
else if (argv[0][2] == 's')
#ifdef ENABLE_OPT
#if ENABLE_OPT
Options |= EOptionOptimizeSize;
#else
Error("-Os not available; optimizer not linked");
......@@ -720,6 +720,8 @@ void SetMessageOptions(EShMessages& messages)
messages = (EShMessages)(messages | EShMsgDebugInfo);
if (HlslEnable16BitTypes)
messages = (EShMessages)(messages | EShMsgHlslEnable16BitTypes);
if ((Options & EOptionOptimizeDisable) || !ENABLE_OPT)
messages = (EShMessages)(messages | EShMsgHlslLegalization);
}
//
......
Test/baseLegalResults/hlsl.aliasOpaque.frag.out
hlsl.aliasOpaque.frag
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 87
......
Test/baseLegalResults/hlsl.flattenOpaque.frag.out
hlsl.flattenOpaque.frag
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 185
......
Test/baseLegalResults/hlsl.flattenOpaqueInit.vert.out
hlsl.flattenOpaqueInit.vert
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 134
......
Test/baseLegalResults/hlsl.flattenOpaqueInitMix.vert.out
hlsl.flattenOpaqueInitMix.vert
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 97
......
Test/baseLegalResults/hlsl.flattenSubset.frag.out
hlsl.flattenSubset.frag
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 66
......
Test/baseLegalResults/hlsl.flattenSubset2.frag.out
hlsl.flattenSubset2.frag
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 53
......
Test/baseLegalResults/hlsl.partialFlattenLocal.vert.out
hlsl.partialFlattenLocal.vert
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 169
......
Test/baseLegalResults/hlsl.partialFlattenMixed.vert.out
hlsl.partialFlattenMixed.vert
WARNING: AST will form illegal SPIR-V; need to transform to legalize
// Module Version 10000
// Generated by (magic number): 80006
// Id's are bound by 36
......
Test/baseResults/310AofA.vert.out
......@@ -11,6 +11,7 @@ ERROR: 0:79: 'assign' : cannot convert from ' global 4-element array of 7-eleme
ERROR: 0:81: 'foo' : no matching overloaded function found
ERROR: 0:86: '==' : wrong operand types: no operation '==' exists that takes a left-hand operand of type ' global 4-element array of 7-element array of highp float' and a right operand of type ' global 5-element array of 7-element array of highp float' (or there is no acceptable conversion)
ERROR: 0:90: '[' : array index out of range '5'
ERROR: 0:94: '[' : index out of range '-1'
ERROR: 0:96: 'assign' : cannot convert from ' temp 3-element array of highp 4-component vector of float' to 'layout( column_major shared) buffer unsized 2-element array of highp 4-component vector of float'
ERROR: 0:103: '' : array size required
ERROR: 0:104: '' : array size required
......@@ -20,7 +21,7 @@ ERROR: 0:107: '' : array size required
ERROR: 0:110: 'vertex input arrays' : not supported with this profile: es
ERROR: 0:111: 'vertex-shader array-of-array output' : not supported with this profile: es
ERROR: 0:113: 'array-of-array of block' : not supported with this profile: es
ERROR: 21 compilation errors. No code generated.
ERROR: 22 compilation errors. No code generated.
Shader version: 310
......
Test/baseResults/spv.shaderBallotAMD.comp.out
This source diff could not be displayed because it is too large. You can view the blob instead.
Test/baseResults/spv.specConstant.vert.out
......@@ -11,7 +11,7 @@ spv.specConstant.vert
Source GLSL 400
Name 4 "main"
Name 9 "arraySize"
Name 14 "foo(vf4[s2148];"
Name 14 "foo(vf4[s2468];"
Name 13 "p"
Name 17 "builtin_spec_constant("
Name 20 "color"
......@@ -102,10 +102,10 @@ spv.specConstant.vert
Store 20(color) 46
48: 10 Load 22(ucol)
Store 47(param) 48
49: 2 FunctionCall 14(foo(vf4[s2148];) 47(param)
49: 2 FunctionCall 14(foo(vf4[s2468];) 47(param)
Return
FunctionEnd
14(foo(vf4[s2148];): 2 Function None 12
14(foo(vf4[s2468];): 2 Function None 12
13(p): 11(ptr) FunctionParameter
15: Label
54: 24(ptr) AccessChain 53(dupUcol) 23
......
Test/baseResults/spv.subgroupPartitioned.comp.out 0 → 100755
This source diff could not be displayed because it is too large. You can view the blob instead.
Test/spv.shaderBallotAMD.comp
......@@ -2,6 +2,7 @@
#extension GL_ARB_gpu_shader_int64: enable
#extension GL_AMD_gpu_shader_half_float: enable
#extension GL_AMD_gpu_shader_int16: enable
#extension GL_AMD_shader_ballot: enable
layout (local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
......@@ -15,6 +16,8 @@ layout(binding = 0) buffer Buffers
int64_t i64;
u64vec2 u64v;
f16vec3 f16v;
i16vec4 i16v;
uint16_t u16;
};
void main()
......@@ -26,6 +29,8 @@ void main()
i64 = minInvocationsAMD(i64);
u64v = minInvocationsAMD(u64v);
f16v = minInvocationsAMD(f16v);
i16v = minInvocationsAMD(i16v);
u16 = minInvocationsAMD(u16);
i = maxInvocationsAMD(i);
uv = maxInvocationsAMD(uv);
......@@ -34,6 +39,8 @@ void main()
i64 = maxInvocationsAMD(i64);
u64v = maxInvocationsAMD(u64v);
f16v = maxInvocationsAMD(f16v);
i16v = maxInvocationsAMD(i16v);
u16 = maxInvocationsAMD(u16);
i = addInvocationsAMD(i);
uv = addInvocationsAMD(uv);
......@@ -42,6 +49,8 @@ void main()
i64 = addInvocationsAMD(i64);
u64v = addInvocationsAMD(u64v);
f16v = addInvocationsAMD(f16v);
i16v = addInvocationsAMD(i16v);
u16 = addInvocationsAMD(u16);
i = minInvocationsNonUniformAMD(i);
uv = minInvocationsNonUniformAMD(uv);
......@@ -50,6 +59,8 @@ void main()
i64 = minInvocationsNonUniformAMD(i64);
u64v = minInvocationsNonUniformAMD(u64v);
f16v = minInvocationsNonUniformAMD(f16v);
i16v = minInvocationsNonUniformAMD(i16v);
u16 = minInvocationsNonUniformAMD(u16);
i = maxInvocationsNonUniformAMD(i);
uv = maxInvocationsNonUniformAMD(uv);
......@@ -58,6 +69,8 @@ void main()
i64 = maxInvocationsNonUniformAMD(i64);
u64v = maxInvocationsNonUniformAMD(u64v);
f16v = maxInvocationsNonUniformAMD(f16v);
i16v = maxInvocationsNonUniformAMD(i16v);
u16 = maxInvocationsNonUniformAMD(u16);
i = addInvocationsNonUniformAMD(i);
uv = addInvocationsNonUniformAMD(uv);
......@@ -66,6 +79,8 @@ void main()
i64 = addInvocationsNonUniformAMD(i64);
u64v = addInvocationsNonUniformAMD(u64v);
f16v = addInvocationsNonUniformAMD(f16v);
i16v = addInvocationsNonUniformAMD(i16v);
u16 = addInvocationsNonUniformAMD(u16);
i = minInvocationsInclusiveScanAMD(i);
uv = minInvocationsInclusiveScanAMD(uv);
......@@ -74,6 +89,8 @@ void main()
i64 = minInvocationsInclusiveScanAMD(i64);
u64v = minInvocationsInclusiveScanAMD(u64v);
f16v = minInvocationsInclusiveScanAMD(f16v);
i16v = minInvocationsInclusiveScanAMD(i16v);
u16 = minInvocationsInclusiveScanAMD(u16);
i = maxInvocationsInclusiveScanAMD(i);
uv = maxInvocationsInclusiveScanAMD(uv);
......@@ -82,6 +99,8 @@ void main()
i64 = maxInvocationsInclusiveScanAMD(i64);
u64v = maxInvocationsInclusiveScanAMD(u64v);
f16v = maxInvocationsInclusiveScanAMD(f16v);
i16v = maxInvocationsInclusiveScanAMD(i16v);
u16 = maxInvocationsInclusiveScanAMD(u16);
i = addInvocationsInclusiveScanAMD(i);
uv = addInvocationsInclusiveScanAMD(uv);
......@@ -90,6 +109,8 @@ void main()
i64 = addInvocationsInclusiveScanAMD(i64);
u64v = addInvocationsInclusiveScanAMD(u64v);
f16v = addInvocationsInclusiveScanAMD(f16v);
i16v = addInvocationsInclusiveScanAMD(i16v);
u16 = addInvocationsInclusiveScanAMD(u16);
i = minInvocationsExclusiveScanAMD(i);
uv = minInvocationsExclusiveScanAMD(uv);
......@@ -98,6 +119,8 @@ void main()
i64 = minInvocationsExclusiveScanAMD(i64);
u64v = minInvocationsExclusiveScanAMD(u64v);
f16v = minInvocationsExclusiveScanAMD(f16v);
i16v = minInvocationsExclusiveScanAMD(i16v);
u16 = minInvocationsExclusiveScanAMD(u16);
i = maxInvocationsExclusiveScanAMD(i);
uv = maxInvocationsExclusiveScanAMD(uv);
......@@ -106,6 +129,8 @@ void main()
i64 = maxInvocationsExclusiveScanAMD(i64);
u64v = maxInvocationsExclusiveScanAMD(u64v);
f16v = maxInvocationsExclusiveScanAMD(f16v);
i16v = maxInvocationsExclusiveScanAMD(i16v);
u16 = maxInvocationsExclusiveScanAMD(u16);
i = addInvocationsExclusiveScanAMD(i);
uv = addInvocationsExclusiveScanAMD(uv);
......@@ -114,6 +139,8 @@ void main()
i64 = addInvocationsExclusiveScanAMD(i64);
u64v = addInvocationsExclusiveScanAMD(u64v);
f16v = addInvocationsExclusiveScanAMD(f16v);
i16v = addInvocationsExclusiveScanAMD(i16v);
u16 = addInvocationsExclusiveScanAMD(u16);
i = minInvocationsInclusiveScanNonUniformAMD(i);
uv = minInvocationsInclusiveScanNonUniformAMD(uv);
......@@ -122,6 +149,8 @@ void main()
i64 = minInvocationsInclusiveScanNonUniformAMD(i64);
u64v = minInvocationsInclusiveScanNonUniformAMD(u64v);
f16v = minInvocationsInclusiveScanNonUniformAMD(f16v);
i16v = minInvocationsInclusiveScanNonUniformAMD(i16v);
u16 = minInvocationsInclusiveScanNonUniformAMD(u16);
i = maxInvocationsInclusiveScanNonUniformAMD(i);
uv = maxInvocationsInclusiveScanNonUniformAMD(uv);
......@@ -130,6 +159,8 @@ void main()
i64 = maxInvocationsInclusiveScanNonUniformAMD(i64);
u64v = maxInvocationsInclusiveScanNonUniformAMD(u64v);
f16v = maxInvocationsInclusiveScanNonUniformAMD(f16v);
i16v = maxInvocationsInclusiveScanNonUniformAMD(i16v);
u16 = maxInvocationsInclusiveScanNonUniformAMD(u16);
i = addInvocationsInclusiveScanNonUniformAMD(i);
uv = addInvocationsInclusiveScanNonUniformAMD(uv);
......@@ -138,6 +169,8 @@ void main()
i64 = addInvocationsInclusiveScanNonUniformAMD(i64);
u64v = addInvocationsInclusiveScanNonUniformAMD(u64v);
f16v = addInvocationsInclusiveScanNonUniformAMD(f16v);
i16v = addInvocationsInclusiveScanNonUniformAMD(i16v);
u16 = addInvocationsInclusiveScanNonUniformAMD(u16);
i = minInvocationsExclusiveScanNonUniformAMD(i);
uv = minInvocationsExclusiveScanNonUniformAMD(uv);
......@@ -146,6 +179,8 @@ void main()
i64 = minInvocationsExclusiveScanNonUniformAMD(i64);
u64v = minInvocationsExclusiveScanNonUniformAMD(u64v);
f16v = minInvocationsExclusiveScanNonUniformAMD(f16v);
i16v = minInvocationsExclusiveScanNonUniformAMD(i16v);
u16 = minInvocationsExclusiveScanNonUniformAMD(u16);
i = maxInvocationsExclusiveScanNonUniformAMD(i);
uv = maxInvocationsExclusiveScanNonUniformAMD(uv);
......@@ -154,6 +189,8 @@ void main()
i64 = maxInvocationsExclusiveScanNonUniformAMD(i64);
u64v = maxInvocationsExclusiveScanNonUniformAMD(u64v);
f16v = maxInvocationsExclusiveScanNonUniformAMD(f16v);
i16v = maxInvocationsExclusiveScanNonUniformAMD(i16v);
u16 = maxInvocationsExclusiveScanNonUniformAMD(u16);
i = addInvocationsExclusiveScanNonUniformAMD(i);
uv = addInvocationsExclusiveScanNonUniformAMD(uv);
......@@ -162,4 +199,6 @@ void main()
i64 = addInvocationsExclusiveScanNonUniformAMD(i64);
u64v = addInvocationsExclusiveScanNonUniformAMD(u64v);
f16v = addInvocationsExclusiveScanNonUniformAMD(f16v);
i16v = addInvocationsExclusiveScanNonUniformAMD(i16v);
u16 = addInvocationsExclusiveScanNonUniformAMD(u16);
}
glslang/Include/intermediate.h 100644 → 100755
......@@ -527,6 +527,32 @@ enum TOperator {
EOpSubgroupQuadSwapHorizontal,
EOpSubgroupQuadSwapVertical,
EOpSubgroupQuadSwapDiagonal,
#ifdef NV_EXTENSIONS
EOpSubgroupPartition,
EOpSubgroupPartitionedAdd,
EOpSubgroupPartitionedMul,
EOpSubgroupPartitionedMin,
EOpSubgroupPartitionedMax,
EOpSubgroupPartitionedAnd,
EOpSubgroupPartitionedOr,
EOpSubgroupPartitionedXor,
EOpSubgroupPartitionedInclusiveAdd,
EOpSubgroupPartitionedInclusiveMul,
EOpSubgroupPartitionedInclusiveMin,
EOpSubgroupPartitionedInclusiveMax,
EOpSubgroupPartitionedInclusiveAnd,
EOpSubgroupPartitionedInclusiveOr,
EOpSubgroupPartitionedInclusiveXor,
EOpSubgroupPartitionedExclusiveAdd,
EOpSubgroupPartitionedExclusiveMul,
EOpSubgroupPartitionedExclusiveMin,
EOpSubgroupPartitionedExclusiveMax,
EOpSubgroupPartitionedExclusiveAnd,
EOpSubgroupPartitionedExclusiveOr,
EOpSubgroupPartitionedExclusiveXor,
#endif
EOpSubgroupGuardStop,
#ifdef AMD_EXTENSIONS
......
glslang/MachineIndependent/ParseHelper.cpp
......@@ -372,6 +372,9 @@ TIntermTyped* TParseContext::handleBracketDereference(const TSourceLoc& loc, TIn
} else {
// at least one of base and index is not a front-end constant variable...
if (index->getQualifier().isFrontEndConstant())
checkIndex(loc, base->getType(), indexValue);
if (base->getAsSymbolNode() && isIoResizeArray(base->getType()))
handleIoResizeArrayAccess(loc, base);
......
glslang/MachineIndependent/ShaderLang.cpp
......@@ -1925,6 +1925,7 @@ const char* TProgram::getUniformBlockName(int index) const { return reflection
int TProgram::getUniformBlockSize(int index) const { return reflection->getUniformBlock(index).size; }
int TProgram::getUniformIndex(const char* name) const { return reflection->getIndex(name); }
int TProgram::getUniformBinding(int index) const { return reflection->getUniform(index).getBinding(); }
int TProgram::getUniformBlockBinding(int index) const { return reflection->getUniformBlock(index).getBinding(); }
int TProgram::getUniformBlockIndex(int index) const { return reflection->getUniform(index).index; }
int TProgram::getUniformBlockCounterIndex(int index) const { return reflection->getUniformBlock(index).counterIndex; }
int TProgram::getUniformType(int index) const { return reflection->getUniform(index).glDefineType; }
......
glslang/MachineIndependent/Versions.cpp 100644 → 100755
......@@ -227,6 +227,7 @@ void TParseVersions::initializeExtensionBehavior()
extensionBehavior[E_GL_NV_shader_atomic_int64] = EBhDisable;
extensionBehavior[E_GL_NV_conservative_raster_underestimation] = EBhDisable;
extensionBehavior[E_GL_NV_shader_noperspective_interpolation] = EBhDisable;
extensionBehavior[E_GL_NV_shader_subgroup_partitioned] = EBhDisable;
#endif
// AEP
......@@ -401,6 +402,7 @@ void TParseVersions::getPreamble(std::string& preamble)
"#define GL_NV_viewport_array2 1\n"
"#define GL_NV_shader_atomic_int64 1\n"
"#define GL_NV_conservative_raster_underestimation 1\n"
"#define GL_NV_shader_subgroup_partitioned 1\n"
#endif
"#define GL_KHX_shader_explicit_arithmetic_types 1\n"
"#define GL_KHX_shader_explicit_arithmetic_types_int8 1\n"
......@@ -762,6 +764,10 @@ void TParseVersions::updateExtensionBehavior(int line, const char* extension, co
updateExtensionBehavior(line, "GL_KHR_shader_subgroup_basic", behaviorString);
else if (strcmp(extension, "GL_KHR_shader_subgroup_quad") == 0)
updateExtensionBehavior(line, "GL_KHR_shader_subgroup_basic", behaviorString);
#ifdef NV_EXTENSIONS
else if (strcmp(extension, "GL_NV_shader_subgroup_partitioned") == 0)
updateExtensionBehavior(line, "GL_KHR_shader_subgroup_basic", behaviorString);
#endif
}
void TParseVersions::updateExtensionBehavior(const char* extension, TExtensionBehavior behavior)
......
glslang/MachineIndependent/Versions.h 100644 → 100755
......@@ -198,6 +198,7 @@ const char* const E_GL_NVX_multiview_per_view_attributes = "GL_NVX_multiv
const char* const E_GL_NV_shader_atomic_int64 = "GL_NV_shader_atomic_int64";
const char* const E_GL_NV_conservative_raster_underestimation = "GL_NV_conservative_raster_underestimation";
const char* const E_GL_NV_shader_noperspective_interpolation = "GL_NV_shader_noperspective_interpolation";
const char* const E_GL_NV_shader_subgroup_partitioned = "GL_NV_shader_subgroup_partitioned";
// Arrays of extensions for the above viewportEXTs duplications
......
glslang/MachineIndependent/intermOut.cpp
......@@ -599,6 +599,31 @@ bool TOutputTraverser::visitUnary(TVisit /* visit */, TIntermUnary* node)
case EOpSubgroupQuadSwapVertical: out.debug << "subgroupQuadSwapVertical"; break;
case EOpSubgroupQuadSwapDiagonal: out.debug << "subgroupQuadSwapDiagonal"; break;
#ifdef NV_EXTENSIONS
case EOpSubgroupPartition: out.debug << "subgroupPartitionNV"; break;
case EOpSubgroupPartitionedAdd: out.debug << "subgroupPartitionedAddNV"; break;
case EOpSubgroupPartitionedMul: out.debug << "subgroupPartitionedMulNV"; break;
case EOpSubgroupPartitionedMin: out.debug << "subgroupPartitionedMinNV"; break;
case EOpSubgroupPartitionedMax: out.debug << "subgroupPartitionedMaxNV"; break;
case EOpSubgroupPartitionedAnd: out.debug << "subgroupPartitionedAndNV"; break;
case EOpSubgroupPartitionedOr: out.debug << "subgroupPartitionedOrNV"; break;
case EOpSubgroupPartitionedXor: out.debug << "subgroupPartitionedXorNV"; break;
case EOpSubgroupPartitionedInclusiveAdd: out.debug << "subgroupPartitionedInclusiveAddNV"; break;
case EOpSubgroupPartitionedInclusiveMul: out.debug << "subgroupPartitionedInclusiveMulNV"; break;
case EOpSubgroupPartitionedInclusiveMin: out.debug << "subgroupPartitionedInclusiveMinNV"; break;
case EOpSubgroupPartitionedInclusiveMax: out.debug << "subgroupPartitionedInclusiveMaxNV"; break;
case EOpSubgroupPartitionedInclusiveAnd: out.debug << "subgroupPartitionedInclusiveAndNV"; break;
case EOpSubgroupPartitionedInclusiveOr: out.debug << "subgroupPartitionedInclusiveOrNV"; break;
case EOpSubgroupPartitionedInclusiveXor: out.debug << "subgroupPartitionedInclusiveXorNV"; break;
case EOpSubgroupPartitionedExclusiveAdd: out.debug << "subgroupPartitionedExclusiveAddNV"; break;
case EOpSubgroupPartitionedExclusiveMul: out.debug << "subgroupPartitionedExclusiveMulNV"; break;
case EOpSubgroupPartitionedExclusiveMin: out.debug << "subgroupPartitionedExclusiveMinNV"; break;
case EOpSubgroupPartitionedExclusiveMax: out.debug << "subgroupPartitionedExclusiveMaxNV"; break;
case EOpSubgroupPartitionedExclusiveAnd: out.debug << "subgroupPartitionedExclusiveAndNV"; break;
case EOpSubgroupPartitionedExclusiveOr: out.debug << "subgroupPartitionedExclusiveOrNV"; break;
case EOpSubgroupPartitionedExclusiveXor: out.debug << "subgroupPartitionedExclusiveXorNV"; break;
#endif
case EOpClip: out.debug << "clip"; break;
case EOpIsFinite: out.debug << "isfinite"; break;
case EOpLog10: out.debug << "log10"; break;
......
glslang/Public/ShaderLang.h
......@@ -215,6 +215,7 @@ enum EShMessages {
EShMsgHlslOffsets = (1 << 9), // allow block offsets to follow HLSL rules instead of GLSL rules
EShMsgDebugInfo = (1 << 10), // save debug information
EShMsgHlslEnable16BitTypes = (1 << 11), // enable use of 16-bit types in SPIR-V for HLSL
EShMsgHlslLegalization = (1 << 12), // enable HLSL Legalization messages
};
//
......@@ -670,6 +671,7 @@ public:
int getUniformBlockSize(int blockIndex) const; // can be used for glGetActiveUniformBlockiv(UNIFORM_BLOCK_DATA_SIZE)
int getUniformIndex(const char* name) const; // can be used for glGetUniformIndices()
int getUniformBinding(int index) const; // returns the binding number
int getUniformBlockBinding(int index) const; // returns the block binding number
int getUniformBlockIndex(int index) const; // can be used for glGetActiveUniformsiv(GL_UNIFORM_BLOCK_INDEX)
int getUniformBlockCounterIndex(int index) const; // returns block index of associated counter.
int getUniformType(int index) const; // can be used for glGetActiveUniformsiv(GL_UNIFORM_TYPE)
......
gtests/Hlsl.FromFile.cpp
......@@ -92,7 +92,7 @@ TEST_P(HlslLegalizeTest, FromFile)
loadFileCompileAndCheck(GlobalTestSettings.testRoot, GetParam().fileName,
Source::HLSL, Semantics::Vulkan, glslang::EShTargetVulkan_1_0,
Target::Spv, true, GetParam().entryPoint,
"/baseLegalResults/", false);
"/baseLegalResults/", true);
}
// clang-format off
......@@ -410,7 +410,7 @@ INSTANTIATE_TEST_CASE_P(
);
// clang-format on
#ifdef ENABLE_OPT
#if ENABLE_OPT
// clang-format off
INSTANTIATE_TEST_CASE_P(
ToSpirv, HlslLegalizeTest,
......
gtests/Spv.FromFile.cpp 100644 → 100755
......@@ -339,6 +339,7 @@ INSTANTIATE_TEST_CASE_P(
"spv.precise.tese",
"spv.precise.tesc",
"spv.vulkan100.subgroupArithmetic.comp",
"spv.vulkan100.subgroupPartitioned.comp",
"spv.xfb.vert",
"spv.xfb2.vert",
"spv.xfb3.vert",
......@@ -369,6 +370,7 @@ INSTANTIATE_TEST_CASE_P(
"spv.subgroupBallot.comp",
"spv.subgroupClustered.comp",
"spv.subgroupClusteredNeg.comp",
"spv.subgroupPartitioned.comp",
"spv.subgroupShuffle.comp",
"spv.subgroupShuffleRelative.comp",
"spv.subgroupQuad.comp",
......
gtests/TestFixture.cpp
......@@ -100,6 +100,8 @@ EShMessages DeriveOptions(Source source, Semantics semantics, Target target)
break;
}
result = static_cast<EShMessages>(result | EShMsgHlslLegalization);
return result;
}
......
gtests/TestFixture.h
......@@ -200,7 +200,7 @@ public:
glslang::EShTargetClientVersion clientTargetVersion,
bool flattenUniformArrays = false,
EShTextureSamplerTransformMode texSampTransMode = EShTexSampTransKeep,
bool disableOptimizer = true,
bool enableOptimizer = false,
bool automap = true)
{
const EShLanguage stage = GetShaderStage(GetSuffix(shaderName));
......@@ -242,7 +242,7 @@ public:
if (success && (controls & EShMsgSpvRules)) {
std::vector<uint32_t> spirv_binary;
glslang::SpvOptions options;
options.disableOptimizer = disableOptimizer;
options.disableOptimizer = !enableOptimizer;
glslang::GlslangToSpv(*program.getIntermediate(stage),
spirv_binary, &logger, &options);
......@@ -412,7 +412,7 @@ public:
bool automap = true,
const std::string& entryPointName="",
const std::string& baseDir="/baseResults/",
const bool disableOptimizer = true)
const bool enableOptimizer = false)
{
const std::string inputFname = testDir + "/" + testName;
const std::string expectedOutputFname =
......@@ -422,9 +422,11 @@ public:
tryLoadFile(inputFname, "input", &input);
tryLoadFile(expectedOutputFname, "expected output", &expectedOutput);
const EShMessages controls = DeriveOptions(source, semantics, target);
EShMessages controls = DeriveOptions(source, semantics, target);
if (enableOptimizer)
controls = static_cast<EShMessages>(controls & ~EShMsgHlslLegalization);
GlslangResult result = compileAndLink(testName, input, entryPointName, controls, clientTargetVersion, false,
EShTexSampTransKeep, disableOptimizer, automap);
EShTexSampTransKeep, enableOptimizer, automap);
// Generate the hybrid output in the way of glslangValidator.
std::ostringstream stream;
......
hlsl/hlslParseHelper.cpp
......@@ -829,6 +829,9 @@ TIntermTyped* HlslParseContext::handleBracketDereference(const TSourceLoc& loc,
} else {
// at least one of base and index is variable...
if (index->getQualifier().isFrontEndConstant())
checkIndex(loc, base->getType(), indexValue);
if (base->getType().isScalarOrVec1())
result = base;
else if (base->getAsSymbolNode() && wasFlattened(base)) {
......@@ -844,11 +847,10 @@ TIntermTyped* HlslParseContext::handleBracketDereference(const TSourceLoc& loc,
else
checkIndex(loc, base->getType(), indexValue);
result = intermediate.addIndex(EOpIndexDirect, base, index, loc);
} else {
} else
result = intermediate.addIndex(EOpIndexIndirect, base, index, loc);
}
}
}
if (result == nullptr) {
// Insert dummy error-recovery result
......@@ -9925,7 +9927,7 @@ void HlslParseContext::finish()
// Communicate out (esp. for command line) that we formed AST that will make
// illegal AST SPIR-V and it needs transforms to legalize it.
if (intermediate.needsLegalization())
if (intermediate.needsLegalization() && (messages & EShMsgHlslLegalization))
infoSink.info << "WARNING: AST will form illegal SPIR-V; need to transform to legalize";
TParseContextBase::finish();
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment