Updated glslang.

3rdparty/glslang/SPIRV/GLSL.ext.KHR.h

@@ -50,5 +50,6 @@ static const char* const E_SPV_KHR_ray_tracing                  = "SPV_KHR_ray_t
 static const char* const E_SPV_KHR_ray_query                    = "SPV_KHR_ray_query";
 static const char* const E_SPV_KHR_fragment_shading_rate        = "SPV_KHR_fragment_shading_rate";
 static const char* const E_SPV_KHR_terminate_invocation         = "SPV_KHR_terminate_invocation";
+static const char* const E_SPV_KHR_workgroup_memory_explicit_layout = "SPV_KHR_workgroup_memory_explicit_layout";
 
 #endif  // #ifndef GLSLextKHR_H

3rdparty/glslang/SPIRV/GlslangToSpv.cpp

@@ -380,6 +380,7 @@ spv::Decoration TranslateBlockDecoration(const glslang::TType& type, bool useSto
         case glslang::EvqBuffer:       return useStorageBuffer ? spv::DecorationBlock : spv::DecorationBufferBlock;
         case glslang::EvqVaryingIn:    return spv::DecorationBlock;
         case glslang::EvqVaryingOut:   return spv::DecorationBlock;
+        case glslang::EvqShared:       return spv::DecorationBlock;
 #ifndef GLSLANG_WEB
         case glslang::EvqPayload:      return spv::DecorationBlock;
         case glslang::EvqPayloadIn:    return spv::DecorationBlock;
@@ -436,6 +437,7 @@ spv::Decoration TranslateLayoutDecoration(const glslang::TType& type, glslang::T
             break;
         case glslang::EbtBlock:
             switch (type.getQualifier().storage) {
+            case glslang::EvqShared:
             case glslang::EvqUniform:
             case glslang::EvqBuffer:
                 switch (type.getQualifier().layoutPacking) {
@@ -1278,6 +1280,12 @@ spv::StorageClass TGlslangToSpvTraverser::TranslateStorageClass(const glslang::T
         return spv::StorageClassUniformConstant;
     }
 
+    if (type.getQualifier().storage == glslang::EvqShared && type.getBasicType() == glslang::EbtBlock) {
+        builder.addExtension(spv::E_SPV_KHR_workgroup_memory_explicit_layout);
+        builder.addCapability(spv::CapabilityWorkgroupMemoryExplicitLayoutKHR);
+        return spv::StorageClassWorkgroup;
+    }
+
     switch (type.getQualifier().storage) {
     case glslang::EvqGlobal:        return spv::StorageClassPrivate;
     case glslang::EvqConstReadOnly: return spv::StorageClassFunction;
@@ -3623,6 +3631,11 @@ spv::Id TGlslangToSpvTraverser::createSpvVariable(const glslang::TIntermSymbol*
             break;
 #endif
         default:
+            if (storageClass == spv::StorageClassWorkgroup &&
+                node->getType().getBasicType() == glslang::EbtBlock) {
+                builder.addCapability(spv::CapabilityWorkgroupMemoryExplicitLayout16BitAccessKHR);
+                break;
+            }
             if (node->getType().contains16BitFloat())
                 builder.addCapability(spv::CapabilityFloat16);
             if (node->getType().contains16BitInt())
@@ -3641,6 +3654,9 @@ spv::Id TGlslangToSpvTraverser::createSpvVariable(const glslang::TIntermSymbol*
         } else if (storageClass == spv::StorageClassStorageBuffer) {
             builder.addIncorporatedExtension(spv::E_SPV_KHR_8bit_storage, spv::Spv_1_5);
             builder.addCapability(spv::CapabilityStorageBuffer8BitAccess);
+        } else if (storageClass == spv::StorageClassWorkgroup &&
+                   node->getType().getBasicType() == glslang::EbtBlock) {
+            builder.addCapability(spv::CapabilityWorkgroupMemoryExplicitLayout8BitAccessKHR);
         } else {
             builder.addCapability(spv::CapabilityInt8);
         }
@@ -4407,6 +4423,7 @@ glslang::TLayoutPacking TGlslangToSpvTraverser::getExplicitLayout(const glslang:
     // has to be a uniform or buffer block or task in/out blocks
     if (type.getQualifier().storage != glslang::EvqUniform &&
         type.getQualifier().storage != glslang::EvqBuffer &&
+        type.getQualifier().storage != glslang::EvqShared &&
         !type.getQualifier().isTaskMemory())
         return glslang::ElpNone;
 
@@ -6996,6 +7013,10 @@ spv::Id TGlslangToSpvTraverser::createAtomicOperation(glslang::TOperator op, spv
         builder.addCapability(spv::CapabilityVulkanMemoryModelKHR);
     }
 
+    if (builder.getConstantScalar(scopeId) == spv::ScopeQueueFamily) {
+        builder.addCapability(spv::CapabilityVulkanMemoryModelKHR);
+    }
+
     if (glslangIntermediate->usingVulkanMemoryModel() && builder.getConstantScalar(scopeId) == spv::ScopeDevice) {
         builder.addCapability(spv::CapabilityVulkanMemoryModelDeviceScopeKHR);
     }

3rdparty/glslang/SPIRV/SpvPostProcess.cpp

@@ -436,6 +436,38 @@ void Builder::postProcessFeatures() {
             }
         }
     }
+
+    // If any Vulkan memory model-specific functionality is used, update the
+    // OpMemoryModel to match.
+    if (capabilities.find(spv::CapabilityVulkanMemoryModelKHR) != capabilities.end()) {
+        memoryModel = spv::MemoryModelVulkanKHR;
+        addIncorporatedExtension(spv::E_SPV_KHR_vulkan_memory_model, spv::Spv_1_5);
+    }
+
+    // Add Aliased decoration if there's more than one Workgroup Block variable.
+    if (capabilities.find(spv::CapabilityWorkgroupMemoryExplicitLayoutKHR) != capabilities.end()) {
+        assert(entryPoints.size() == 1);
+        auto &ep = entryPoints[0];
+
+        std::vector<Id> workgroup_variables;
+        for (int i = 0; i < (int)ep->getNumOperands(); i++) {
+            if (!ep->isIdOperand(i))
+                continue;
+
+            const Id id = ep->getIdOperand(i);
+            const Instruction *instr = module.getInstruction(id);
+            if (instr->getOpCode() != spv::OpVariable)
+                continue;
+
+            if (instr->getImmediateOperand(0) == spv::StorageClassWorkgroup)
+                workgroup_variables.push_back(id);
+        }
+
+        if (workgroup_variables.size() > 1) {
+            for (size_t i = 0; i < workgroup_variables.size(); i++)
+                addDecoration(workgroup_variables[i], spv::DecorationAliased);
+        }
+    }
 }
 #endif
 

3rdparty/glslang/SPIRV/SpvTools.cpp

@@ -153,6 +153,8 @@ void SpirvToolsValidate(const glslang::TIntermediate& intermediate, std::vector<
     spv_validator_options options = spvValidatorOptionsCreate();
     spvValidatorOptionsSetRelaxBlockLayout(options, intermediate.usingHlslOffsets());
     spvValidatorOptionsSetBeforeHlslLegalization(options, prelegalization);
+    spvValidatorOptionsSetScalarBlockLayout(options, intermediate.usingScalarBlockLayout());
+    spvValidatorOptionsSetWorkgroupScalarBlockLayout(options, intermediate.usingScalarBlockLayout());
     spvValidateWithOptions(context, options, &binary, &diagnostic);
 
     // report

3rdparty/glslang/SPIRV/doc.cpp

@@ -967,6 +967,10 @@ const char* CapabilityString(int info)
     case CapabilityAtomicFloat32AddEXT:                     return "AtomicFloat32AddEXT";
     case CapabilityAtomicFloat64AddEXT:                     return "AtomicFloat64AddEXT";
 
+    case CapabilityWorkgroupMemoryExplicitLayoutKHR:            return "CapabilityWorkgroupMemoryExplicitLayoutKHR";
+    case CapabilityWorkgroupMemoryExplicitLayout8BitAccessKHR:  return "CapabilityWorkgroupMemoryExplicitLayout8BitAccessKHR";
+    case CapabilityWorkgroupMemoryExplicitLayout16BitAccessKHR: return "CapabilityWorkgroupMemoryExplicitLayout16BitAccessKHR";
+
     default: return "Bad";
     }
 }

3rdparty/glslang/SPIRV/spirv.hpp

@@ -168,10 +168,16 @@ enum ExecutionMode {
     ExecutionModeSampleInterlockUnorderedEXT = 5369,
     ExecutionModeShadingRateInterlockOrderedEXT = 5370,
     ExecutionModeShadingRateInterlockUnorderedEXT = 5371,
+    ExecutionModeSharedLocalMemorySizeINTEL = 5618,
+    ExecutionModeRoundingModeRTPINTEL = 5620,
+    ExecutionModeRoundingModeRTNINTEL = 5621,
+    ExecutionModeFloatingPointModeALTINTEL = 5622,
+    ExecutionModeFloatingPointModeIEEEINTEL = 5623,
     ExecutionModeMaxWorkgroupSizeINTEL = 5893,
     ExecutionModeMaxWorkDimINTEL = 5894,
     ExecutionModeNoGlobalOffsetINTEL = 5895,
     ExecutionModeNumSIMDWorkitemsINTEL = 5896,
+    ExecutionModeSchedulerTargetFmaxMhzINTEL = 5903,
     ExecutionModeMax = 0x7fffffff,
 };
 
@@ -204,6 +210,8 @@ enum StorageClass {
     StorageClassPhysicalStorageBuffer = 5349,
     StorageClassPhysicalStorageBufferEXT = 5349,
     StorageClassCodeSectionINTEL = 5605,
+    StorageClassDeviceOnlyINTEL = 5936,
+    StorageClassHostOnlyINTEL = 5937,
     StorageClassMax = 0x7fffffff,
 };
 
@@ -374,6 +382,8 @@ enum FPFastMathModeShift {
     FPFastMathModeNSZShift = 2,
     FPFastMathModeAllowRecipShift = 3,
     FPFastMathModeFastShift = 4,
+    FPFastMathModeAllowContractFastINTELShift = 16,
+    FPFastMathModeAllowReassocINTELShift = 17,
     FPFastMathModeMax = 0x7fffffff,
 };
 
@@ -384,6 +394,8 @@ enum FPFastMathModeMask {
     FPFastMathModeNSZMask = 0x00000004,
     FPFastMathModeAllowRecipMask = 0x00000008,
     FPFastMathModeFastMask = 0x00000010,
+    FPFastMathModeAllowContractFastINTELMask = 0x00010000,
+    FPFastMathModeAllowReassocINTELMask = 0x00020000,
 };
 
 enum FPRoundingMode {
@@ -484,12 +496,22 @@ enum Decoration {
     DecorationRestrictPointerEXT = 5355,
     DecorationAliasedPointer = 5356,
     DecorationAliasedPointerEXT = 5356,
+    DecorationSIMTCallINTEL = 5599,
     DecorationReferencedIndirectlyINTEL = 5602,
+    DecorationClobberINTEL = 5607,
+    DecorationSideEffectsINTEL = 5608,
+    DecorationVectorComputeVariableINTEL = 5624,
+    DecorationFuncParamIOKindINTEL = 5625,
+    DecorationVectorComputeFunctionINTEL = 5626,
+    DecorationStackCallINTEL = 5627,
+    DecorationGlobalVariableOffsetINTEL = 5628,
     DecorationCounterBuffer = 5634,
     DecorationHlslCounterBufferGOOGLE = 5634,
     DecorationHlslSemanticGOOGLE = 5635,
     DecorationUserSemantic = 5635,
     DecorationUserTypeGOOGLE = 5636,
+    DecorationFunctionRoundingModeINTEL = 5822,
+    DecorationFunctionDenormModeINTEL = 5823,
     DecorationRegisterINTEL = 5825,
     DecorationMemoryINTEL = 5826,
     DecorationNumbanksINTEL = 5827,
@@ -502,6 +524,17 @@ enum Decoration {
     DecorationMergeINTEL = 5834,
     DecorationBankBitsINTEL = 5835,
     DecorationForcePow2DepthINTEL = 5836,
+    DecorationBurstCoalesceINTEL = 5899,
+    DecorationCacheSizeINTEL = 5900,
+    DecorationDontStaticallyCoalesceINTEL = 5901,
+    DecorationPrefetchINTEL = 5902,
+    DecorationStallEnableINTEL = 5905,
+    DecorationFuseLoopsInFunctionINTEL = 5907,
+    DecorationBufferLocationINTEL = 5921,
+    DecorationIOPipeStorageINTEL = 5944,
+    DecorationFunctionFloatingPointModeINTEL = 6080,
+    DecorationSingleElementVectorINTEL = 6085,
+    DecorationVectorComputeCallableFunctionINTEL = 6087,
     DecorationMax = 0x7fffffff,
 };
 
@@ -656,6 +689,7 @@ enum LoopControlShift {
     LoopControlLoopCoalesceINTELShift = 20,
     LoopControlMaxInterleavingINTELShift = 21,
     LoopControlSpeculatedIterationsINTELShift = 22,
+    LoopControlNoFusionINTELShift = 23,
     LoopControlMax = 0x7fffffff,
 };
 
@@ -677,6 +711,7 @@ enum LoopControlMask {
     LoopControlLoopCoalesceINTELMask = 0x00100000,
     LoopControlMaxInterleavingINTELMask = 0x00200000,
     LoopControlSpeculatedIterationsINTELMask = 0x00400000,
+    LoopControlNoFusionINTELMask = 0x00800000,
 };
 
 enum FunctionControlShift {
@@ -876,6 +911,9 @@ enum Capability {
     CapabilityFragmentShadingRateKHR = 4422,
     CapabilitySubgroupBallotKHR = 4423,
     CapabilityDrawParameters = 4427,
+    CapabilityWorkgroupMemoryExplicitLayoutKHR = 4428,
+    CapabilityWorkgroupMemoryExplicitLayout8BitAccessKHR = 4429,
+    CapabilityWorkgroupMemoryExplicitLayout16BitAccessKHR = 4430,
     CapabilitySubgroupVoteKHR = 4431,
     CapabilityStorageBuffer16BitAccess = 4433,
     CapabilityStorageUniformBufferBlock16 = 4433,
@@ -966,21 +1004,37 @@ enum Capability {
     CapabilitySubgroupBufferBlockIOINTEL = 5569,
     CapabilitySubgroupImageBlockIOINTEL = 5570,
     CapabilitySubgroupImageMediaBlockIOINTEL = 5579,
+    CapabilityRoundToInfinityINTEL = 5582,
+    CapabilityFloatingPointModeINTEL = 5583,
     CapabilityIntegerFunctions2INTEL = 5584,
     CapabilityFunctionPointersINTEL = 5603,
     CapabilityIndirectReferencesINTEL = 5604,
+    CapabilityAsmINTEL = 5606,
+    CapabilityVectorComputeINTEL = 5617,
+    CapabilityVectorAnyINTEL = 5619,
     CapabilitySubgroupAvcMotionEstimationINTEL = 5696,
     CapabilitySubgroupAvcMotionEstimationIntraINTEL = 5697,
     CapabilitySubgroupAvcMotionEstimationChromaINTEL = 5698,
+    CapabilityVariableLengthArrayINTEL = 5817,
+    CapabilityFunctionFloatControlINTEL = 5821,
     CapabilityFPGAMemoryAttributesINTEL = 5824,
+    CapabilityFPFastMathModeINTEL = 5837,
+    CapabilityArbitraryPrecisionIntegersINTEL = 5844,
     CapabilityUnstructuredLoopControlsINTEL = 5886,
     CapabilityFPGALoopControlsINTEL = 5888,
     CapabilityKernelAttributesINTEL = 5892,
     CapabilityFPGAKernelAttributesINTEL = 5897,
+    CapabilityFPGAMemoryAccessesINTEL = 5898,
+    CapabilityFPGAClusterAttributesINTEL = 5904,
+    CapabilityLoopFuseINTEL = 5906,
+    CapabilityFPGABufferLocationINTEL = 5920,
+    CapabilityUSMStorageClassesINTEL = 5935,
+    CapabilityIOPipesINTEL = 5943,
     CapabilityBlockingPipesINTEL = 5945,
     CapabilityFPGARegINTEL = 5948,
     CapabilityAtomicFloat32AddEXT = 6033,
     CapabilityAtomicFloat64AddEXT = 6034,
+    CapabilityLongConstantCompositeINTEL = 6089,
     CapabilityMax = 0x7fffffff,
 };
 
@@ -1047,6 +1101,18 @@ enum FragmentShadingRateMask {
     FragmentShadingRateHorizontal4PixelsMask = 0x00000008,
 };
 
+enum FPDenormMode {
+  FPDenormModePreserve = 0,
+  FPDenormModeFlushToZero = 1,
+  FPDenormModeMax = 0x7fffffff,
+};
+
+enum FPOperationMode {
+  FPOperationModeIEEE = 0,
+  FPOperationModeALT = 1,
+  FPOperationModeMax = 0x7fffffff,
+};
+
 enum Op {
     OpNop = 0,
     OpUndef = 1,
@@ -1466,8 +1532,11 @@ enum Op {
     OpUSubSatINTEL = 5596,
     OpIMul32x16INTEL = 5597,
     OpUMul32x16INTEL = 5598,
-    OpFunctionPointerINTEL = 5600,
+    OpConstFunctionPointerINTEL = 5600,
     OpFunctionPointerCallINTEL = 5601,
+    OpAsmTargetINTEL = 5609,
+    OpAsmINTEL = 5610,
+    OpAsmCallINTEL = 5611,
     OpDecorateString = 5632,
     OpDecorateStringGOOGLE = 5632,
     OpMemberDecorateString = 5633,
@@ -1590,7 +1659,12 @@ enum Op {
     OpSubgroupAvcSicGetPackedSkcLumaCountThresholdINTEL = 5814,
     OpSubgroupAvcSicGetPackedSkcLumaSumThresholdINTEL = 5815,
     OpSubgroupAvcSicGetInterRawSadsINTEL = 5816,
+    OpVariableLengthArrayINTEL = 5818,
+    OpSaveMemoryINTEL = 5819,
+    OpRestoreMemoryINTEL = 5820,
     OpLoopControlINTEL = 5887,
+    OpPtrCastToCrossWorkgroupINTEL = 5934,
+    OpCrossWorkgroupCastToPtrINTEL = 5938,
     OpReadPipeBlockingINTEL = 5946,
     OpWritePipeBlockingINTEL = 5947,
     OpFPGARegINTEL = 5949,
@@ -1612,6 +1686,10 @@ enum Op {
     OpRayQueryGetIntersectionObjectToWorldKHR = 6031,
     OpRayQueryGetIntersectionWorldToObjectKHR = 6032,
     OpAtomicFAddEXT = 6035,
+    OpTypeBufferSurfaceINTEL = 6086,
+    OpTypeStructContinuedINTEL = 6090,
+    OpConstantCompositeContinuedINTEL = 6091,
+    OpSpecConstantCompositeContinuedINTEL = 6092,
     OpMax = 0x7fffffff,
 };
 
@@ -2036,8 +2114,11 @@ inline void HasResultAndType(Op opcode, bool *hasResult, bool *hasResultType) {
     case OpUSubSatINTEL: *hasResult = true; *hasResultType = true; break;
     case OpIMul32x16INTEL: *hasResult = true; *hasResultType = true; break;
     case OpUMul32x16INTEL: *hasResult = true; *hasResultType = true; break;
-    case OpFunctionPointerINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpConstFunctionPointerINTEL: *hasResult = true; *hasResultType = true; break;
     case OpFunctionPointerCallINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpAsmTargetINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpAsmINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpAsmCallINTEL: *hasResult = true; *hasResultType = true; break;
     case OpDecorateString: *hasResult = false; *hasResultType = false; break;
     case OpMemberDecorateString: *hasResult = false; *hasResultType = false; break;
     case OpVmeImageINTEL: *hasResult = true; *hasResultType = true; break;
@@ -2158,7 +2239,12 @@ inline void HasResultAndType(Op opcode, bool *hasResult, bool *hasResultType) {
     case OpSubgroupAvcSicGetPackedSkcLumaCountThresholdINTEL: *hasResult = true; *hasResultType = true; break;
     case OpSubgroupAvcSicGetPackedSkcLumaSumThresholdINTEL: *hasResult = true; *hasResultType = true; break;
     case OpSubgroupAvcSicGetInterRawSadsINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpVariableLengthArrayINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpSaveMemoryINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpRestoreMemoryINTEL: *hasResult = false; *hasResultType = false; break;
     case OpLoopControlINTEL: *hasResult = false; *hasResultType = false; break;
+    case OpPtrCastToCrossWorkgroupINTEL: *hasResult = true; *hasResultType = true; break;
+    case OpCrossWorkgroupCastToPtrINTEL: *hasResult = true; *hasResultType = true; break;
     case OpReadPipeBlockingINTEL: *hasResult = true; *hasResultType = true; break;
     case OpWritePipeBlockingINTEL: *hasResult = true; *hasResultType = true; break;
     case OpFPGARegINTEL: *hasResult = true; *hasResultType = true; break;
@@ -2180,6 +2266,10 @@ inline void HasResultAndType(Op opcode, bool *hasResult, bool *hasResultType) {
     case OpRayQueryGetIntersectionObjectToWorldKHR: *hasResult = true; *hasResultType = true; break;
     case OpRayQueryGetIntersectionWorldToObjectKHR: *hasResult = true; *hasResultType = true; break;
     case OpAtomicFAddEXT: *hasResult = true; *hasResultType = true; break;
+    case OpTypeBufferSurfaceINTEL: *hasResult = true; *hasResultType = false; break;
+    case OpTypeStructContinuedINTEL: *hasResult = false; *hasResultType = false; break;
+    case OpConstantCompositeContinuedINTEL: *hasResult = false; *hasResultType = false; break;
+    case OpSpecConstantCompositeContinuedINTEL: *hasResult = false; *hasResultType = false; break;
     }
 }
 #endif /* SPV_ENABLE_UTILITY_CODE */

3rdparty/glslang/glslang/MachineIndependent/Intermediate.cpp

@@ -2870,7 +2870,7 @@ void TIntermediate::addToCallGraph(TInfoSink& /*infoSink*/, const TString& calle
             return;
     }
 
-    callGraph.push_front(TCall(caller, callee));
+    callGraph.emplace_front(caller, callee);
 }
 
 //

3rdparty/glslang/glslang/MachineIndependent/ParseHelper.cpp

@@ -87,6 +87,10 @@ TParseContext::TParseContext(TSymbolTable& symbolTable, TIntermediate& interm, b
     globalInputDefaults.clear();
     globalOutputDefaults.clear();
 
+    globalSharedDefaults.clear();
+    globalSharedDefaults.layoutMatrix = ElmColumnMajor;
+    globalSharedDefaults.layoutPacking = ElpStd430;
+
 #ifndef GLSLANG_WEB
     // "Shaders in the transform
     // feedback capturing mode have an initial global default of
@@ -6033,12 +6037,28 @@ void TParseContext::layoutTypeCheck(const TSourceLoc& loc, const TType& type)
     }
 }
 
+static bool storageCanHaveLayoutInBlock(const enum TStorageQualifier storage)
+{
+    switch (storage) {
+    case EvqUniform:
+    case EvqBuffer:
+    case EvqShared:
+        return true;
+    default:
+        return false;
+    }
+}
+
 // Do layout error checking that can be done within a layout qualifier proper, not needing to know
 // if there are blocks, atomic counters, variables, etc.
 void TParseContext::layoutQualifierCheck(const TSourceLoc& loc, const TQualifier& qualifier)
 {
-    if (qualifier.storage == EvqShared && qualifier.hasLayout())
-        error(loc, "cannot apply layout qualifiers to a shared variable", "shared", "");
+    if (qualifier.storage == EvqShared && qualifier.hasLayout()) {
+        if (spvVersion.spv > 0 && spvVersion.spv < EShTargetSpv_1_4) {
+            error(loc, "shared block requires at least SPIR-V 1.4", "shared block", "");
+        }
+        profileRequires(loc, EEsProfile | ECoreProfile | ECompatibilityProfile, 0, E_GL_EXT_shared_memory_block, "shared block");
+    }
 
     // "It is a compile-time error to use *component* without also specifying the location qualifier (order does not matter)."
     if (qualifier.hasComponent() && ! qualifier.hasLocation())
@@ -6121,7 +6141,7 @@ void TParseContext::layoutQualifierCheck(const TSourceLoc& loc, const TQualifier
             error(loc, "can only be used on an output", "xfb layout qualifier", "");
     }
     if (qualifier.hasUniformLayout()) {
-        if (! qualifier.isUniformOrBuffer() && !qualifier.isTaskMemory()) {
+        if (!storageCanHaveLayoutInBlock(qualifier.storage) && !qualifier.isTaskMemory()) {
             if (qualifier.hasMatrix() || qualifier.hasPacking())
                 error(loc, "matrix or packing qualifiers can only be used on a uniform or buffer", "layout", "");
             if (qualifier.hasOffset() || qualifier.hasAlign())
@@ -7667,6 +7687,7 @@ void TParseContext::declareBlock(const TSourceLoc& loc, TTypeList& typeList, con
     case EvqBuffer:     defaultQualification = globalBufferDefaults;     break;
     case EvqVaryingIn:  defaultQualification = globalInputDefaults;      break;
     case EvqVaryingOut: defaultQualification = globalOutputDefaults;     break;
+    case EvqShared:     defaultQualification = globalSharedDefaults;     break;
     default:            defaultQualification.clear();                    break;
     }
 
@@ -7930,6 +7951,12 @@ void TParseContext::blockStageIoCheck(const TSourceLoc& loc, const TQualifier& q
             error(loc, "output blocks cannot be used in a task shader", "out", "");
         }
         break;
+    case EvqShared:
+        if (spvVersion.spv > 0 && spvVersion.spv < EShTargetSpv_1_4) {
+            error(loc, "shared block requires at least SPIR-V 1.4", "shared block", "");
+        }
+        profileRequires(loc, EEsProfile | ECoreProfile | ECompatibilityProfile, 0, E_GL_EXT_shared_memory_block, "shared block");
+        break;
 #ifndef GLSLANG_WEB
     case EvqPayload:
         profileRequires(loc, ~EEsProfile, 460, 2, extsrt, "rayPayloadNV block");
@@ -8088,7 +8115,7 @@ void TParseContext::fixXfbOffsets(TQualifier& qualifier, TTypeList& typeList)
 //
 void TParseContext::fixBlockUniformOffsets(TQualifier& qualifier, TTypeList& typeList)
 {
-    if (!qualifier.isUniformOrBuffer() && !qualifier.isTaskMemory())
+    if (!storageCanHaveLayoutInBlock(qualifier.storage) && !qualifier.isTaskMemory())
         return;
     if (qualifier.layoutPacking != ElpStd140 && qualifier.layoutPacking != ElpStd430 && qualifier.layoutPacking != ElpScalar)
         return;
@@ -8602,8 +8629,14 @@ void TParseContext::updateStandaloneQualifierDefaults(const TSourceLoc& loc, con
         }
 #endif
         break;
+    case EvqShared:
+        if (qualifier.hasMatrix())
+            globalSharedDefaults.layoutMatrix = qualifier.layoutMatrix;
+        if (qualifier.hasPacking())
+            globalSharedDefaults.layoutPacking = qualifier.layoutPacking;
+        break;
     default:
-        error(loc, "default qualifier requires 'uniform', 'buffer', 'in', or 'out' storage qualification", "", "");
+        error(loc, "default qualifier requires 'uniform', 'buffer', 'in', 'out' or 'shared' storage qualification", "", "");
         return;
     }
 

3rdparty/glslang/glslang/MachineIndependent/ParseHelper.h

@@ -485,6 +485,7 @@ protected:
     TQualifier globalUniformDefaults;
     TQualifier globalInputDefaults;
     TQualifier globalOutputDefaults;
+    TQualifier globalSharedDefaults;
     TString currentCaller;        // name of last function body entered (not valid when at global scope)
 #ifndef GLSLANG_WEB
     int* atomicUintOffsets;       // to become an array of the right size to hold an offset per binding point

3rdparty/glslang/glslang/MachineIndependent/Versions.cpp

@@ -331,6 +331,7 @@ void TParseVersions::initializeExtensionBehavior()
     extensionBehavior[E_GL_EXT_fragment_shading_rate]       = EBhDisable;
     extensionBehavior[E_GL_EXT_shader_image_int64]   = EBhDisable;
     extensionBehavior[E_GL_EXT_terminate_invocation]        = EBhDisable;
+    extensionBehavior[E_GL_EXT_shared_memory_block]         = EBhDisable;
 
     // OVR extensions
     extensionBehavior[E_GL_OVR_multiview]                = EBhDisable;
@@ -472,6 +473,7 @@ void TParseVersions::getPreamble(std::string& preamble)
             "#define GL_EXT_demote_to_helper_invocation 1\n"
             "#define GL_EXT_debug_printf 1\n"
             "#define GL_EXT_fragment_shading_rate 1\n"
+            "#define GL_EXT_shared_memory_block 1\n"
 
             // GL_KHR_shader_subgroup
             "#define GL_KHR_shader_subgroup_basic 1\n"

3rdparty/glslang/glslang/MachineIndependent/Versions.h

@@ -202,6 +202,7 @@ const char* const E_GL_EXT_shader_implicit_conversions      = "GL_EXT_shader_imp
 const char* const E_GL_EXT_fragment_shading_rate            = "GL_EXT_fragment_shading_rate";
 const char* const E_GL_EXT_shader_image_int64               = "GL_EXT_shader_image_int64";
 const char* const E_GL_EXT_null_initializer                 = "GL_EXT_null_initializer";
+const char* const E_GL_EXT_shared_memory_block              = "GL_EXT_shared_memory_block";
 
 // Arrays of extensions for the above viewportEXTs duplications
 

3rdparty/glslang/glslang/MachineIndependent/linkValidate.cpp

@@ -653,6 +653,25 @@ void TIntermediate::mergeErrorCheck(TInfoSink& infoSink, const TIntermSymbol& sy
 #endif
 }
 
+void TIntermediate::sharedBlockCheck(TInfoSink& infoSink)
+{
+    bool has_shared_block = false;
+    bool has_shared_non_block = false;
+    TIntermSequence& linkObjects = findLinkerObjects()->getSequence();
+    for (size_t i = 0; i < linkObjects.size(); ++i) {
+        const TType& type = linkObjects[i]->getAsTyped()->getType();
+        const TQualifier& qualifier = type.getQualifier();
+        if (qualifier.storage == glslang::EvqShared) {
+            if (type.getBasicType() == glslang::EbtBlock)
+                has_shared_block = true;
+            else
+                has_shared_non_block = true;
+        }
+    }
+    if (has_shared_block && has_shared_non_block)
+        error(infoSink, "cannot mix use of shared variables inside and outside blocks");
+}
+
 //
 // Do final link-time error checking of a complete (merged) intermediate representation.
 // (Much error checking was done during merging).
@@ -778,6 +797,7 @@ void TIntermediate::finalCheck(TInfoSink& infoSink, bool keepUncalled)
             error(infoSink, "post_depth_coverage requires early_fragment_tests");
         break;
     case EShLangCompute:
+        sharedBlockCheck(infoSink);
         break;
     case EShLangRayGen:
     case EShLangIntersect:
@@ -810,6 +830,7 @@ void TIntermediate::finalCheck(TInfoSink& infoSink, bool keepUncalled)
     case EShLangTaskNV:
         if (numTaskNVBlocks > 1)
             error(infoSink, "Only one taskNV interface block is allowed per shader");
+        sharedBlockCheck(infoSink);
         break;
     default:
         error(infoSink, "Unknown Stage.");

3rdparty/glslang/glslang/MachineIndependent/localintermediate.h

@@ -858,6 +858,14 @@ public:
     bool usingHlslIoMapping() { return false; }
 #endif
 
+    bool usingScalarBlockLayout() const {
+        for (auto extIt = requestedExtensions.begin(); extIt != requestedExtensions.end(); ++extIt) {
+            if (*extIt == E_GL_EXT_scalar_block_layout)
+                return true;
+        }
+        return false;
+    }
+
     void addToCallGraph(TInfoSink&, const TString& caller, const TString& callee);
     void merge(TInfoSink&, TIntermediate&);
     void finalCheck(TInfoSink&, bool keepUncalled);
@@ -946,6 +954,7 @@ protected:
     void checkCallGraphCycles(TInfoSink&);
     void checkCallGraphBodies(TInfoSink&, bool keepUncalled);
     void inOutLocationCheck(TInfoSink&);
+    void sharedBlockCheck(TInfoSink&);
     bool userOutputUsed() const;
     bool isSpecializationOperation(const TIntermOperator&) const;
     bool isNonuniformPropagating(TOperator) const;