AuroraMiddleware/glslang
1
0
Fork 0
mirror of https://github.com/KhronosGroup/glslang synced 2024-11-10 20:30:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #398 from amdrexu/feature

Browse Source 
AMD: Implement 4 AMD-specific extensions.
This commit is contained in:
John Kessenich 2016-07-21 14:39:27 -06:00 committed by GitHub
commit df018fa994
20 changed files with 5425 additions and 4120 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
CMakeLists.txt
View File

@ -1,12 +1,18 @@
cmake_minimum_required(VERSION 2.8.12)
set_property(GLOBAL PROPERTY USE_FOLDERS ON)
option(ENABLE_AMD_EXTENSIONS "Enables support of AMD-specific extensions" OFF)
enable_testing()
set(CMAKE_INSTALL_PREFIX "install" CACHE STRING "prefix")
project(glslang)
if(ENABLE_AMD_EXTENSIONS)
add_definitions(-DAMD_EXTENSIONS)
endif(ENABLE_AMD_EXTENSIONS)
if(WIN32)
set(CMAKE_DEBUG_POSTFIX "d")
include(ChooseMSVCCRT.cmake)

5
SPIRV/CMakeLists.txt
View File

@ -18,6 +18,11 @@ set(HEADERS
doc.h
disassemble.h)
if(ENABLE_AMD_EXTENSIONS)
set(HEADERS
GLSL.ext.AMD.h)
endif(ENABLE_AMD_EXTENSIONS)
add_library(SPIRV STATIC ${SOURCES} ${HEADERS})
set_property(TARGET SPIRV PROPERTY FOLDER glslang)

113
SPIRV/GLSL.ext.AMD.h Normal file
View File

@ -0,0 +1,113 @@
/*
** Copyright (c) 2014-2016 The Khronos Group Inc.
**
** Permission is hereby granted, free of charge, to any person obtaining a copy
** of this software and/or associated documentation files (the "Materials"),
** to deal in the Materials without restriction, including without limitation
** the rights to use, copy, modify, merge, publish, distribute, sublicense,
** and/or sell copies of the Materials, and to permit persons to whom the
** Materials are furnished to do so, subject to the following conditions:
**
** The above copyright notice and this permission notice shall be included in
** all copies or substantial portions of the Materials.
**
** MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS KHRONOS
** STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS SPECIFICATIONS AND
** HEADER INFORMATION ARE LOCATED AT https://www.khronos.org/registry/
**
** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
** OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
** FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
** THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
** LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
** FROM,OUT OF OR IN CONNECTION WITH THE MATERIALS OR THE USE OR OTHER DEALINGS
** IN THE MATERIALS.
*/
#ifndef GLSLextAMD_H
#define GLSLextAMD_H
enum BuiltIn;
enum Decoration;
enum Op;
static const int GLSLextAMDVersion = 100;
static const int GLSLextAMDRevision = 1;
// SPV_AMD_shader_ballot
static const char* const E_SPV_AMD_shader_ballot = "SPV_AMD_shader_ballot";
static const Op OpGroupIAddNonUniformAMD = static_cast<Op>(5000);
static const Op OpGroupFAddNonUniformAMD = static_cast<Op>(5001);
static const Op OpGroupFMinNonUniformAMD = static_cast<Op>(5002);
static const Op OpGroupUMinNonUniformAMD = static_cast<Op>(5003);
static const Op OpGroupSMinNonUniformAMD = static_cast<Op>(5004);
static const Op OpGroupFMaxNonUniformAMD = static_cast<Op>(5005);
static const Op OpGroupUMaxNonUniformAMD = static_cast<Op>(5006);
static const Op OpGroupSMaxNonUniformAMD = static_cast<Op>(5007);
enum ShaderBallotAMD {
ShaderBallotBadAMD = 0, // Don't use
SwizzleInvocationsAMD = 1,
SwizzleInvocationsMaskedAMD = 2,
WriteInvocationAMD = 3,
MbcntAMD = 4,
ShaderBallotCountAMD
};
// SPV_AMD_shader_trinary_minmax
static const char* const E_SPV_AMD_shader_trinary_minmax = "SPV_AMD_shader_trinary_minmax";
enum ShaderTrinaryMinMaxAMD {
ShaderTrinaryMinMaxBadAMD = 0, // Don't use
FMin3AMD = 1,
UMin3AMD = 2,
SMin3AMD = 3,
FMax3AMD = 4,
UMax3AMD = 5,
SMax3AMD = 6,
FMid3AMD = 7,
UMid3AMD = 8,
SMid3AMD = 9,
ShaderTrinaryMinMaxCountAMD
};
// SPV_AMD_shader_explicit_vertex_parameter
static const char* const E_SPV_AMD_shader_explicit_vertex_parameter = "SPV_AMD_shader_explicit_vertex_parameter";
static const BuiltIn BuiltInBaryCoordNoPerspAMD = static_cast<BuiltIn>(4992);
static const BuiltIn BuiltInBaryCoordNoPerspCentroidAMD = static_cast<BuiltIn>(4993);
static const BuiltIn BuiltInBaryCoordNoPerspSampleAMD = static_cast<BuiltIn>(4994);
static const BuiltIn BuiltInBaryCoordSmoothAMD = static_cast<BuiltIn>(4995);
static const BuiltIn BuiltInBaryCoordSmoothCentroidAMD = static_cast<BuiltIn>(4996);
static const BuiltIn BuiltInBaryCoordSmoothSampleAMD = static_cast<BuiltIn>(4997);
static const BuiltIn BuiltInBaryCoordPullModelAMD = static_cast<BuiltIn>(4998);
static const Decoration DecorationExplicitInterpAMD = static_cast<Decoration>(4999);
enum ShaderExplicitVertexParameterAMD {
ShaderExplicitVertexParameterBadAMD = 0, // Don't use
InterpolateAtVertexAMD = 1,
ShaderExplicitVertexParameterCountAMD
};
// SPV_AMD_gcn_shader
static const char* const E_SPV_AMD_gcn_shader = "SPV_AMD_gcn_shader";
enum GcnShaderAMD {
GcnShaderBadAMD = 0, // Don't use
CubeFaceIndexAMD = 1,
CubeFaceCoordAMD = 2,
TimeAMD = 3,
GcnShaderCountAMD
};
#endif // #ifndef GLSLextAMD_H

222
SPIRV/GlslangToSpv.cpp
View File

@ -43,6 +43,9 @@
#include "SpvBuilder.h"
namespace spv {
#include "GLSL.std.450.h"
#ifdef AMD_EXTENSIONS
#include "GLSL.ext.AMD.h"
#endif
}
// Glslang includes
@ -147,9 +150,9 @@ protected:
spv::Id createConversion(glslang::TOperator op, spv::Decoration precision, spv::Decoration noContraction, spv::Id destTypeId, spv::Id operand, glslang::TBasicType typeProxy);
spv::Id makeSmearedConstant(spv::Id constant, int vectorSize);
spv::Id createAtomicOperation(glslang::TOperator op, spv::Decoration precision, spv::Id typeId, std::vector<spv::Id>& operands, glslang::TBasicType typeProxy);
spv::Id createInvocationsOperation(glslang::TOperator, spv::Id typeId, spv::Id operand);
spv::Id createInvocationsOperation(glslang::TOperator, spv::Id typeId, spv::Id operand, glslang::TBasicType typeProxy);
spv::Id createMiscOperation(glslang::TOperator op, spv::Decoration precision, spv::Id typeId, std::vector<spv::Id>& operands, glslang::TBasicType typeProxy);
spv::Id createNoArgOperation(glslang::TOperator op);
spv::Id createNoArgOperation(glslang::TOperator op, spv::Decoration precision, spv::Id typeId);
spv::Id getSymbolId(const glslang::TIntermSymbol* node);
void addDecoration(spv::Id id, spv::Decoration dec);
void addDecoration(spv::Id id, spv::Decoration dec, unsigned value);
@ -160,6 +163,7 @@ protected:
bool isTrivialLeaf(const glslang::TIntermTyped* node);
bool isTrivial(const glslang::TIntermTyped* node);
spv::Id createShortCircuit(glslang::TOperator, glslang::TIntermTyped& left, glslang::TIntermTyped& right);
spv::Id getExtBuiltins(const char* name);
spv::Function* shaderEntry;
spv::Instruction* entryPoint;
@ -175,6 +179,7 @@ protected:
std::set<spv::Id> iOSet; // all input/output variables from either static use or declaration of interface
const glslang::TIntermediate* glslangIntermediate;
spv::Id stdBuiltins;
std::unordered_map<const char*, spv::Id> extBuiltinMap;
std::unordered_map<int, spv::Id> symbolValues;
std::unordered_set<int> constReadOnlyParameters; // set of formal function parameters that have glslang qualifier constReadOnly, so we know they are not local function "const" that are write-once
@ -371,6 +376,10 @@ spv::Decoration TranslateInterpolationDecoration(const glslang::TQualifier& qual
return spv::DecorationNoPerspective;
else if (qualifier.flat)
return spv::DecorationFlat;
#ifdef AMD_EXTENSIONS
else if (qualifier.explicitInterp)
return spv::DecorationExplicitInterpAMD;
#endif
else
return spv::DecorationMax;
}
@ -508,6 +517,15 @@ spv::BuiltIn TGlslangToSpvTraverser::TranslateBuiltInDecoration(glslang::TBuiltI
// TODO: Add SPIR-V builtin ID.
logger->missingFunctionality("shader ballot");
return spv::BuiltInMax;
#ifdef AMD_EXTENSIONS
case glslang::EbvBaryCoordNoPersp: return spv::BuiltInBaryCoordNoPerspAMD;
case glslang::EbvBaryCoordNoPerspCentroid: return spv::BuiltInBaryCoordNoPerspCentroidAMD;
case glslang::EbvBaryCoordNoPerspSample: return spv::BuiltInBaryCoordNoPerspSampleAMD;
case glslang::EbvBaryCoordSmooth: return spv::BuiltInBaryCoordSmoothAMD;
case glslang::EbvBaryCoordSmoothCentroid: return spv::BuiltInBaryCoordSmoothCentroidAMD;
case glslang::EbvBaryCoordSmoothSample: return spv::BuiltInBaryCoordSmoothSampleAMD;
case glslang::EbvBaryCoordPullModel: return spv::BuiltInBaryCoordPullModelAMD;
#endif
default: return spv::BuiltInMax;
}
}
@ -628,6 +646,10 @@ void InheritQualifiers(glslang::TQualifier& child, const glslang::TQualifier& pa
child.invariant = true;
if (parent.nopersp)
child.nopersp = true;
#ifdef AMD_EXTENSIONS
if (parent.explicitInterp)
child.explicitInterp = true;
#endif
if (parent.flat)
child.flat = true;
if (parent.centroid)
@ -1484,6 +1506,9 @@ bool TGlslangToSpvTraverser::visitAggregate(glslang::TVisit visit, glslang::TInt
break;
case glslang::EOpInterpolateAtSample:
case glslang::EOpInterpolateAtOffset:
#ifdef AMD_EXTENSIONS
case glslang::EOpInterpolateAtVertex:
#endif
if (arg == 0)
lvalue = true;
break;
@ -1524,7 +1549,7 @@ bool TGlslangToSpvTraverser::visitAggregate(glslang::TVisit visit, glslang::TInt
// Pass through to generic operations.
switch (glslangOperands.size()) {
case 0:
result = createNoArgOperation(node->getOp());
result = createNoArgOperation(node->getOp(), precision, convertGlslangToSpvType(node->getType()));
break;
case 1:
result = createUnaryOperation(
@ -3169,6 +3194,7 @@ spv::Id TGlslangToSpvTraverser::createBinaryMatrixOperation(spv::Op op, spv::Dec
spv::Id TGlslangToSpvTraverser::createUnaryOperation(glslang::TOperator op, spv::Decoration precision, spv::Decoration noContraction, spv::Id typeId, spv::Id operand, glslang::TBasicType typeProxy)
{
spv::Op unaryOp = spv::OpNop;
int extBuiltins = -1;
int libCall = -1;
bool isUnsigned = typeProxy == glslang::EbtUint || typeProxy == glslang::EbtUint64;
bool isFloat = typeProxy == glslang::EbtFloat || typeProxy == glslang::EbtDouble;
@ -3448,7 +3474,32 @@ spv::Id TGlslangToSpvTraverser::createUnaryOperation(glslang::TOperator op, spv:
case glslang::EOpAnyInvocation:
case glslang::EOpAllInvocations:
case glslang::EOpAllInvocationsEqual:
return createInvocationsOperation(op, typeId, operand);
#ifdef AMD_EXTENSIONS
case glslang::EOpMinInvocations:
case glslang::EOpMaxInvocations:
case glslang::EOpAddInvocations:
case glslang::EOpMinInvocationsNonUniform:
case glslang::EOpMaxInvocationsNonUniform:
case glslang::EOpAddInvocationsNonUniform:
#endif
return createInvocationsOperation(op, typeId, operand, typeProxy);
#ifdef AMD_EXTENSIONS
case glslang::EOpMbcnt:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_ballot);
libCall = spv::MbcntAMD;
break;
case glslang::EOpCubeFaceIndex:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_gcn_shader);
libCall = spv::CubeFaceIndexAMD;
break;
case glslang::EOpCubeFaceCoord:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_gcn_shader);
libCall = spv::CubeFaceCoordAMD;
break;
#endif
default:
return 0;
@ -3458,7 +3509,7 @@ spv::Id TGlslangToSpvTraverser::createUnaryOperation(glslang::TOperator op, spv:
if (libCall >= 0) {
std::vector<spv::Id> args;
args.push_back(operand);
id = builder.createBuiltinCall(typeId, stdBuiltins, libCall, args);
id = builder.createBuiltinCall(typeId, extBuiltins >= 0 ? extBuiltins : stdBuiltins, libCall, args);
} else {
id = builder.createUnaryOp(unaryOp, typeId, operand);
}
@ -3765,12 +3816,20 @@ spv::Id TGlslangToSpvTraverser::createAtomicOperation(glslang::TOperator op, spv
}
// Create group invocation operations.
spv::Id TGlslangToSpvTraverser::createInvocationsOperation(glslang::TOperator op, spv::Id typeId, spv::Id operand)
spv::Id TGlslangToSpvTraverser::createInvocationsOperation(glslang::TOperator op, spv::Id typeId, spv::Id operand, glslang::TBasicType typeProxy)
{
bool isUnsigned = typeProxy == glslang::EbtUint || typeProxy == glslang::EbtUint64;
bool isFloat = typeProxy == glslang::EbtFloat || typeProxy == glslang::EbtDouble;
builder.addCapability(spv::CapabilityGroups);
std::vector<spv::Id> operands;
operands.push_back(builder.makeUintConstant(spv::ScopeSubgroup));
#ifdef AMD_EXTENSIONS
if (op == glslang::EOpMinInvocations || op == glslang::EOpMaxInvocations || op == glslang::EOpAddInvocations ||
op == glslang::EOpMinInvocationsNonUniform || op == glslang::EOpMaxInvocationsNonUniform || op == glslang::EOpAddInvocationsNonUniform)
operands.push_back(spv::GroupOperationReduce);
#endif
operands.push_back(operand);
switch (op) {
@ -3786,6 +3845,74 @@ spv::Id TGlslangToSpvTraverser::createInvocationsOperation(glslang::TOperator op
return builder.createBinOp(spv::OpLogicalOr, typeId, groupAll,
builder.createUnaryOp(spv::OpLogicalNot, typeId, groupAny));
}
#ifdef AMD_EXTENSIONS
case glslang::EOpMinInvocations:
case glslang::EOpMaxInvocations:
case glslang::EOpAddInvocations:
{
spv::Op spvOp = spv::OpNop;
if (op == glslang::EOpMinInvocations) {
if (isFloat)
spvOp = spv::OpGroupFMin;
else {
if (isUnsigned)
spvOp = spv::OpGroupUMin;
else
spvOp = spv::OpGroupSMin;
}
} else if (op == glslang::EOpMaxInvocations) {
if (isFloat)
spvOp = spv::OpGroupFMax;
else {
if (isUnsigned)
spvOp = spv::OpGroupUMax;
else
spvOp = spv::OpGroupSMax;
}
} else {
if (isFloat)
spvOp = spv::OpGroupFAdd;
else
spvOp = spv::OpGroupIAdd;
}
return builder.createOp(spvOp, typeId, operands);
}
case glslang::EOpMinInvocationsNonUniform:
case glslang::EOpMaxInvocationsNonUniform:
case glslang::EOpAddInvocationsNonUniform:
{
spv::Op spvOp = spv::OpNop;
if (op == glslang::EOpMinInvocationsNonUniform) {
if (isFloat)
spvOp = spv::OpGroupFMinNonUniformAMD;
else {
if (isUnsigned)
spvOp = spv::OpGroupUMinNonUniformAMD;
else
spvOp = spv::OpGroupSMinNonUniformAMD;
}
}
else if (op == glslang::EOpMaxInvocationsNonUniform) {
if (isFloat)
spvOp = spv::OpGroupFMaxNonUniformAMD;
else {
if (isUnsigned)
spvOp = spv::OpGroupUMaxNonUniformAMD;
else
spvOp = spv::OpGroupSMaxNonUniformAMD;
}
}
else {
if (isFloat)
spvOp = spv::OpGroupFAddNonUniformAMD;
else
spvOp = spv::OpGroupIAddNonUniformAMD;
}
return builder.createOp(spvOp, typeId, operands);
}
#endif
default:
logger->missingFunctionality("invocation operation");
return spv::NoResult;
@ -3798,6 +3925,7 @@ spv::Id TGlslangToSpvTraverser::createMiscOperation(glslang::TOperator op, spv::
bool isFloat = typeProxy == glslang::EbtFloat || typeProxy == glslang::EbtDouble;
spv::Op opCode = spv::OpNop;
int extBuiltins = -1;
int libCall = -1;
size_t consumedOperands = operands.size();
spv::Id typeId0 = 0;
@ -3941,6 +4069,60 @@ spv::Id TGlslangToSpvTraverser::createMiscOperation(glslang::TOperator op, spv::
libCall = spv::GLSLstd450Bad;
break;
#ifdef AMD_EXTENSIONS
case glslang::EOpSwizzleInvocations:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_ballot);
libCall = spv::SwizzleInvocationsAMD;
break;
case glslang::EOpSwizzleInvocationsMasked:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_ballot);
libCall = spv::SwizzleInvocationsMaskedAMD;
break;
case glslang::EOpWriteInvocation:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_ballot);
libCall = spv::WriteInvocationAMD;
break;
case glslang::EOpMin3:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_trinary_minmax);
if (isFloat)
libCall = spv::FMin3AMD;
else {
if (isUnsigned)
libCall = spv::UMin3AMD;
else
libCall = spv::SMin3AMD;
}
break;
case glslang::EOpMax3:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_trinary_minmax);
if (isFloat)
libCall = spv::FMax3AMD;
else {
if (isUnsigned)
libCall = spv::UMax3AMD;
else
libCall = spv::SMax3AMD;
}
break;
case glslang::EOpMid3:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_trinary_minmax);
if (isFloat)
libCall = spv::FMid3AMD;
else {
if (isUnsigned)
libCall = spv::UMid3AMD;
else
libCall = spv::SMid3AMD;
}
break;
case glslang::EOpInterpolateAtVertex:
extBuiltins = getExtBuiltins(spv::E_SPV_AMD_shader_explicit_vertex_parameter);
libCall = spv::InterpolateAtVertexAMD;
break;
#endif
default:
return 0;
}
@ -3951,7 +4133,7 @@ spv::Id TGlslangToSpvTraverser::createMiscOperation(glslang::TOperator op, spv::
// Construct the call arguments, without modifying the original operands vector.
// We might need the remaining arguments, e.g. in the EOpFrexp case.
std::vector<spv::Id> callArguments(operands.begin(), operands.begin() + consumedOperands);
id = builder.createBuiltinCall(typeId, stdBuiltins, libCall, callArguments);
id = builder.createBuiltinCall(typeId, extBuiltins >= 0 ? extBuiltins : stdBuiltins, libCall, callArguments);
} else {
switch (consumedOperands) {
case 0:
@ -3997,8 +4179,8 @@ spv::Id TGlslangToSpvTraverser::createMiscOperation(glslang::TOperator op, spv::
return builder.setPrecision(id, precision);
}
// Intrinsics with no arguments, no return value, and no precision.
spv::Id TGlslangToSpvTraverser::createNoArgOperation(glslang::TOperator op)
// Intrinsics with no arguments (or no return value, and no precision).
spv::Id TGlslangToSpvTraverser::createNoArgOperation(glslang::TOperator op, spv::Decoration precision, spv::Id typeId)
{
// TODO: get the barrier operands correct
@ -4045,6 +4227,14 @@ spv::Id TGlslangToSpvTraverser::createNoArgOperation(glslang::TOperator op)
// Control barrier with non-"None" semantic is also a memory barrier.
builder.createControlBarrier(spv::ScopeWorkgroup, spv::ScopeWorkgroup, spv::MemorySemanticsWorkgroupMemoryMask);
return 0;
#ifdef AMD_EXTENSIONS
case glslang::EOpTime:
{
std::vector<spv::Id> args; // Dummy arguments
spv::Id id = builder.createBuiltinCall(typeId, getExtBuiltins(spv::E_SPV_AMD_gcn_shader), spv::TimeAMD, args);
return builder.setPrecision(id, precision);
}
#endif
default:
logger->missingFunctionality("unknown operation with no arguments");
return 0;
@ -4437,6 +4627,20 @@ spv::Id TGlslangToSpvTraverser::createShortCircuit(glslang::TOperator op, glslan
return builder.createOp(spv::OpPhi, boolTypeId, phiOperands);
}
// Return type Id of the imported set of extended instructions corresponds to the name.
// Import this set if it has not been imported yet.
spv::Id TGlslangToSpvTraverser::getExtBuiltins(const char* name)
{
if (extBuiltinMap.find(name) != extBuiltinMap.end())
return extBuiltinMap[name];
else {
builder.addExtensions(name);
spv::Id extBuiltins = builder.import(name);
extBuiltinMap[name] = extBuiltins;
return extBuiltins;
}
}
}; // end anonymous namespace
namespace glslang {

14
SPIRV/SpvBuilder.cpp
View File

@ -2312,7 +2312,11 @@ void Builder::dump(std::vector<unsigned int>& out) const
capInst.dump(out);
}
// TBD: OpExtension ...
for (int e = 0; e < (int)extensions.size(); ++e) {
Instruction extInst(0, 0, OpExtension);
extInst.addStringOperand(extensions[e]);
extInst.dump(out);
}
dumpInstructions(out, imports);
Instruction memInst(0, 0, OpMemoryModel);
@ -2331,10 +2335,10 @@ void Builder::dump(std::vector<unsigned int>& out) const
sourceInst.addImmediateOperand(sourceVersion);
sourceInst.dump(out);
}
for (int e = 0; e < (int)extensions.size(); ++e) {
Instruction extInst(0, 0, OpSourceExtension);
extInst.addStringOperand(extensions[e]);
extInst.dump(out);
for (int e = 0; e < (int)sourceExtensions.size(); ++e) {
Instruction sourceExtInst(0, 0, OpSourceExtension);
sourceExtInst.addStringOperand(sourceExtensions[e]);
sourceExtInst.dump(out);
}
dumpInstructions(out, names);
dumpInstructions(out, lines);

4
SPIRV/SpvBuilder.h
View File

@ -70,7 +70,8 @@ public:
source = lang;
sourceVersion = version;
}
void addSourceExtension(const char* ext) { extensions.push_back(ext); }
void addSourceExtension(const char* ext) { sourceExtensions.push_back(ext); }
void addExtensions(const char* ext) { extensions.push_back(ext); }
Id import(const char*);
void setMemoryModel(spv::AddressingModel addr, spv::MemoryModel mem)
{
@ -552,6 +553,7 @@ public:
SourceLanguage source;
int sourceVersion;
std::vector<const char*> extensions;
std::vector<const char*> sourceExtensions;
AddressingModel addressModel;
MemoryModel memoryModel;
std::set<spv::Capability> capabilities;

82
SPIRV/disassemble.cpp
View File

@ -44,16 +44,25 @@
#include <sstream>
#include <cstring>
namespace spv {
// Include C-based headers that don't have a namespace
#include "GLSL.std.450.h"
}
const char* GlslStd450DebugNames[spv::GLSLstd450Count];
#include "disassemble.h"
#include "doc.h"
namespace spv {
extern "C" {
// Include C-based headers that don't have a namespace
#include "GLSL.std.450.h"
#ifdef AMD_EXTENSIONS
#include "GLSL.ext.AMD.h"
#endif
}
}
const char* GlslStd450DebugNames[spv::GLSLstd450Count];
namespace spv {
#ifdef AMD_EXTENSIONS
static const char* GLSLextAMDGetDebugNames(const char*, unsigned);
#endif
static void Kill(std::ostream& out, const char* message)
{
@ -64,6 +73,9 @@ static void Kill(std::ostream& out, const char* message)
// used to identify the extended instruction library imported when printing
enum ExtInstSet {
GLSL450Inst,
#ifdef AMD_EXTENSIONS
GLSLextAMDInst,
#endif
OpenCLExtInst,
};
@ -446,14 +458,26 @@ void SpirvStream::disassembleInstruction(Id resultId, Id /*typeId*/, Op opCode,
--numOperands;
if (opCode == OpExtInst) {
ExtInstSet extInstSet = GLSL450Inst;
if (0 == memcmp("OpenCL", (const char*)(idDescriptor[stream[word-2]].c_str()), 6)) {
const char* name = idDescriptor[stream[word - 2]].c_str();
if (0 == memcmp("OpenCL", name, 6)) {
extInstSet = OpenCLExtInst;
#ifdef AMD_EXTENSIONS
} else if (strcmp(spv::E_SPV_AMD_shader_ballot, name) == 0 ||
strcmp(spv::E_SPV_AMD_shader_trinary_minmax, name) == 0 ||
strcmp(spv::E_SPV_AMD_shader_explicit_vertex_parameter, name) == 0 ||
strcmp(spv::E_SPV_AMD_gcn_shader, name) == 0) {
extInstSet = GLSLextAMDInst;
#endif
}
unsigned entrypoint = stream[word - 1];
if (extInstSet == GLSL450Inst) {
if (entrypoint < GLSLstd450Count) {
out << "(" << GlslStd450DebugNames[entrypoint] << ")";
}
#ifdef AMD_EXTENSIONS
} else if (extInstSet == GLSLextAMDInst) {
out << "(" << GLSLextAMDGetDebugNames(name, entrypoint) << ")";
#endif
}
}
break;
@ -561,6 +585,50 @@ static void GLSLstd450GetDebugNames(const char** names)
names[GLSLstd450InterpolateAtOffset] = "InterpolateAtOffset";
}
#ifdef AMD_EXTENSIONS
static const char* GLSLextAMDGetDebugNames(const char* name, unsigned entrypoint)
{
if (strcmp(name, spv::E_SPV_AMD_shader_ballot) == 0) {
switch (entrypoint) {
case SwizzleInvocationsAMD: return "SwizzleInvocationsAMD";
case SwizzleInvocationsMaskedAMD: return "SwizzleInvocationsMaskedAMD";
case WriteInvocationAMD: return "WriteInvocationAMD";
case MbcntAMD: return "MbcntAMD";
default: return "Bad";
}
} else if (strcmp(name, spv::E_SPV_AMD_shader_trinary_minmax) == 0) {
switch (entrypoint) {
case FMin3AMD: return "FMin3AMD";
case UMin3AMD: return "UMin3AMD";
case SMin3AMD: return "SMin3AMD";
case FMax3AMD: return "FMax3AMD";
case UMax3AMD: return "UMax3AMD";
case SMax3AMD: return "SMax3AMD";
case FMid3AMD: return "FMid3AMD";
case UMid3AMD: return "UMid3AMD";
case SMid3AMD: return "SMid3AMD";
default: return "Bad";
}
} else if (strcmp(name, spv::E_SPV_AMD_shader_explicit_vertex_parameter) == 0) {
switch (entrypoint) {
case InterpolateAtVertexAMD: return "InterpolateAtVertexAMD";
default: return "Bad";
}
}
else if (strcmp(name, spv::E_SPV_AMD_gcn_shader) == 0) {
switch (entrypoint) {
case CubeFaceIndexAMD: return "CubeFaceIndexAMD";
case CubeFaceCoordAMD: return "CubeFaceCoordAMD";
case TimeAMD: return "TimeAMD";
default:
break;
}
}
return "Bad";
}
#endif
void Disassemble(std::ostream& out, const std::vector<unsigned int>& stream)
{
SpirvStream SpirvStream(out, stream);

80
SPIRV/doc.cpp
View File

@ -45,6 +45,15 @@
#include <string.h>
#include <algorithm>
#ifdef AMD_EXTENSIONS
namespace spv {
extern "C" {
// Include C-based headers that don't have a namespace
#include "GLSL.ext.AMD.h"
}
}
#endif
namespace spv {
//
@ -243,6 +252,10 @@ const char* DecorationString(int decoration)
case DecorationCeiling:
default: return "Bad";
#ifdef AMD_EXTENSIONS
case 4999: return "ExplicitInterpAMD";
#endif
}
}
@ -298,6 +311,16 @@ const char* BuiltInString(int builtIn)
case BuiltInCeiling:
default: return "Bad";
#ifdef AMD_EXTENSIONS
case 4992: return "BaryCoordNoPerspAMD";
case 4993: return "BaryCoordNoPerspCentroidAMD";
case 4994: return "BaryCoordNoPerspSampleAMD";
case 4995: return "BaryCoordSmoothAMD";
case 4996: return "BaryCoordSmoothCentroidAMD";
case 4997: return "BaryCoordSmoothSampleAMD";
case 4998: return "BaryCoordPullModelAMD";
#endif
}
}
@ -1107,12 +1130,27 @@ const char* OpcodeString(int op)
case OpcodeCeiling:
default:
return "Bad";
#ifdef AMD_EXTENSIONS
case 5000: return "OpGroupIAddNonUniformAMD";
case 5001: return "OpGroupFAddNonUniformAMD";
case 5002: return "OpGroupFMinNonUniformAMD";
case 5003: return "OpGroupUMinNonUniformAMD";
case 5004: return "OpGroupSMinNonUniformAMD";
case 5005: return "OpGroupFMaxNonUniformAMD";
case 5006: return "OpGroupUMaxNonUniformAMD";
case 5007: return "OpGroupSMaxNonUniformAMD";
#endif
}
}
// The set of objects that hold all the instruction/operand
// parameterization information.
#ifdef AMD_EXTENSIONS
InstructionParameters InstructionDesc[OpCodeMask + 1];
#else
InstructionParameters InstructionDesc[OpcodeCeiling];
#endif
OperandParameters ExecutionModeOperands[ExecutionModeCeiling];
OperandParameters DecorationOperands[DecorationCeiling];
@ -2703,6 +2741,48 @@ void Parameterize()
InstructionDesc[OpEnqueueMarker].operands.push(OperandId, "'Num Events'");
InstructionDesc[OpEnqueueMarker].operands.push(OperandId, "'Wait Events'");
InstructionDesc[OpEnqueueMarker].operands.push(OperandId, "'Ret Event'");
#ifdef AMD_EXTENSIONS
InstructionDesc[OpGroupIAddNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupIAddNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupIAddNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupIAddNonUniformAMD].operands.push(OperandId, "'X'");
InstructionDesc[OpGroupFAddNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupFAddNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupFAddNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupFAddNonUniformAMD].operands.push(OperandId, "'X'");
InstructionDesc[OpGroupUMinNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupUMinNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupUMinNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupUMinNonUniformAMD].operands.push(OperandId, "'X'");
InstructionDesc[OpGroupSMinNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupSMinNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupSMinNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupSMinNonUniformAMD].operands.push(OperandId, "X");
InstructionDesc[OpGroupFMinNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupFMinNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupFMinNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupFMinNonUniformAMD].operands.push(OperandId, "X");
InstructionDesc[OpGroupUMaxNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupUMaxNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupUMaxNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupUMaxNonUniformAMD].operands.push(OperandId, "X");
InstructionDesc[OpGroupSMaxNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupSMaxNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupSMaxNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupSMaxNonUniformAMD].operands.push(OperandId, "X");
InstructionDesc[OpGroupFMaxNonUniformAMD].capabilities.push_back(CapabilityGroups);
InstructionDesc[OpGroupFMaxNonUniformAMD].operands.push(OperandScope, "'Execution'");
InstructionDesc[OpGroupFMaxNonUniformAMD].operands.push(OperandGroupOperation, "'Operation'");
InstructionDesc[OpGroupFMaxNonUniformAMD].operands.push(OperandId, "X");
#endif
}
}; // end spv namespace

18
glslang/Include/BaseTypes.h
View File

@ -186,6 +186,15 @@ enum TBuiltInVariable {
EbvSamplePosition,
EbvSampleMask,
EbvHelperInvocation,
#ifdef AMD_EXTENSIONS
EbvBaryCoordNoPersp,
EbvBaryCoordNoPerspCentroid,
EbvBaryCoordNoPerspSample,
EbvBaryCoordSmooth,
EbvBaryCoordSmoothCentroid,
EbvBaryCoordSmoothSample,
EbvBaryCoordPullModel,
#endif
EbvLast
};
@ -286,6 +295,15 @@ __inline const char* GetBuiltInVariableString(TBuiltInVariable v)
case EbvSamplePosition: return "SamplePosition";
case EbvSampleMask: return "SampleMaskIn";
case EbvHelperInvocation: return "HelperInvocation";
#ifdef AMD_EXTENSIONS
case EbvBaryCoordNoPersp: return "BaryCoordNoPersp";
case EbvBaryCoordNoPerspCentroid: return "BaryCoordNoPerspCentroid";
case EbvBaryCoordNoPerspSample: return "BaryCoordNoPerspSample";
case EbvBaryCoordSmooth: return "BaryCoordSmooth";
case EbvBaryCoordSmoothCentroid: return "BaryCoordSmoothCentroid";
case EbvBaryCoordSmoothSample: return "BaryCoordSmoothSample";
case EbvBaryCoordPullModel: return "BaryCoordPullModel";
#endif
default: return "unknown built-in variable";
}
}

14
glslang/Include/Types.h
View File

@ -405,6 +405,9 @@ public:
smooth = false;
flat = false;
nopersp = false;
#ifdef AMD_EXTENSIONS
explicitInterp = false;
#endif
patch = false;
sample = false;
coherent = false;
@ -438,6 +441,9 @@ public:
bool smooth : 1;
bool flat : 1;
bool nopersp : 1;
#ifdef AMD_EXTENSIONS
bool explicitInterp : 1;
#endif
bool patch : 1;
bool sample : 1;
bool coherent : 1;
@ -453,7 +459,11 @@ public:
}
bool isInterpolation() const
{
#ifdef AMD_EXTENSIONS
return flat || smooth || nopersp || explicitInterp;
#else
return flat || smooth || nopersp;
#endif
}
bool isAuxiliary() const
{
@ -1518,6 +1528,10 @@ public:
p += snprintf(p, end - p, "flat ");
if (qualifier.nopersp)
p += snprintf(p, end - p, "noperspective ");
#ifdef AMD_EXTENSIONS
if (qualifier.explicitInterp)
p += snprintf(p, end - p, "__explicitInterpAMD ");
#endif
if (qualifier.patch)
p += snprintf(p, end - p, "patch ");
if (qualifier.sample)

27
glslang/Include/intermediate.h
View File

@ -246,6 +246,12 @@ enum TOperator {
EOpReflect,
EOpRefract,
#ifdef AMD_EXTENSIONS
EOpMin3,
EOpMax3,
EOpMid3,
#endif
EOpDPdx, // Fragment only
EOpDPdy, // Fragment only
EOpFwidth, // Fragment only
@ -260,6 +266,10 @@ enum TOperator {
EOpInterpolateAtSample, // Fragment only
EOpInterpolateAtOffset, // Fragment only
#ifdef AMD_EXTENSIONS
EOpInterpolateAtVertex,
#endif
EOpMatrixTimesMatrix,
EOpOuterProduct,
EOpDeterminant,
@ -291,6 +301,23 @@ enum TOperator {
EOpAllInvocations,
EOpAllInvocationsEqual,
#ifdef AMD_EXTENSIONS
EOpMinInvocations,
EOpMaxInvocations,
EOpAddInvocations,
EOpMinInvocationsNonUniform,
EOpMaxInvocationsNonUniform,
EOpAddInvocationsNonUniform,
EOpSwizzleInvocations,
EOpSwizzleInvocationsMasked,
EOpWriteInvocation,
EOpMbcnt,
EOpCubeFaceIndex,
EOpCubeFaceCoord,
EOpTime,
#endif
EOpAtomicAdd,
EOpAtomicMin,
EOpAtomicMax,

350
glslang/MachineIndependent/Initialize.cpp
View File

@ -801,10 +801,85 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
"bvec3 notEqual(u64vec3, u64vec3);"
"bvec4 notEqual(u64vec4, u64vec4);"
#ifdef AMD_EXTENSIONS
"int findLSB(int64_t);"
"ivec2 findLSB(i64vec2);"
"ivec3 findLSB(i64vec3);"
"ivec4 findLSB(i64vec4);"
"int findLSB(uint64_t);"
"ivec2 findLSB(u64vec2);"
"ivec3 findLSB(u64vec3);"
"ivec4 findLSB(u64vec4);"
"int findMSB(int64_t);"
"ivec2 findMSB(i64vec2);"
"ivec3 findMSB(i64vec3);"
"ivec4 findMSB(i64vec4);"
"int findMSB(uint64_t);"
"ivec2 findMSB(u64vec2);"
"ivec3 findMSB(u64vec3);"
"ivec4 findMSB(u64vec4);"
#endif
"\n"
);
}
#ifdef AMD_EXTENSIONS
// GL_AMD_shader_trinary_minmax
if (profile != EEsProfile && version >= 430) {
commonBuiltins.append(
"float min3(float, float, float);"
"vec2 min3(vec2, vec2, vec2);"
"vec3 min3(vec3, vec3, vec3);"
"vec4 min3(vec4, vec4, vec4);"
"int min3(int, int, int);"
"ivec2 min3(ivec2, ivec2, ivec2);"
"ivec3 min3(ivec3, ivec3, ivec3);"
"ivec4 min3(ivec4, ivec4, ivec4);"
"uint min3(uint, uint, uint);"
"uvec2 min3(uvec2, uvec2, uvec2);"
"uvec3 min3(uvec3, uvec3, uvec3);"
"uvec4 min3(uvec4, uvec4, uvec4);"
"float max3(float, float, float);"
"vec2 max3(vec2, vec2, vec2);"
"vec3 max3(vec3, vec3, vec3);"
"vec4 max3(vec4, vec4, vec4);"
"int max3(int, int, int);"
"ivec2 max3(ivec2, ivec2, ivec2);"
"ivec3 max3(ivec3, ivec3, ivec3);"
"ivec4 max3(ivec4, ivec4, ivec4);"
"uint max3(uint, uint, uint);"
"uvec2 max3(uvec2, uvec2, uvec2);"
"uvec3 max3(uvec3, uvec3, uvec3);"
"uvec4 max3(uvec4, uvec4, uvec4);"
"float mid3(float, float, float);"
"vec2 mid3(vec2, vec2, vec2);"
"vec3 mid3(vec3, vec3, vec3);"
"vec4 mid3(vec4, vec4, vec4);"
"int mid3(int, int, int);"
"ivec2 mid3(ivec2, ivec2, ivec2);"
"ivec3 mid3(ivec3, ivec3, ivec3);"
"ivec4 mid3(ivec4, ivec4, ivec4);"
"uint mid3(uint, uint, uint);"
"uvec2 mid3(uvec2, uvec2, uvec2);"
"uvec3 mid3(uvec3, uvec3, uvec3);"
"uvec4 mid3(uvec4, uvec4, uvec4);"
"\n"
);
}
#endif
if ((profile == EEsProfile && version >= 310) ||
(profile != EEsProfile && version >= 430)) {
commonBuiltins.append(
@ -1468,6 +1543,161 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
"\n");
}
#ifdef AMD_EXTENSIONS
// GL_AMD_shader_ballot
if (profile != EEsProfile && version >= 450) {
commonBuiltins.append(
"float minInvocationsAMD(float);"
"vec2 minInvocationsAMD(vec2);"
"vec3 minInvocationsAMD(vec3);"
"vec4 minInvocationsAMD(vec4);"
"int minInvocationsAMD(int);"
"ivec2 minInvocationsAMD(ivec2);"
"ivec3 minInvocationsAMD(ivec3);"
"ivec4 minInvocationsAMD(ivec4);"
"uint minInvocationsAMD(uint);"
"uvec2 minInvocationsAMD(uvec2);"
"uvec3 minInvocationsAMD(uvec3);"
"uvec4 minInvocationsAMD(uvec4);"
"float maxInvocationsAMD(float);"
"vec2 maxInvocationsAMD(vec2);"
"vec3 maxInvocationsAMD(vec3);"
"vec4 maxInvocationsAMD(vec4);"
"int maxInvocationsAMD(int);"
"ivec2 maxInvocationsAMD(ivec2);"
"ivec3 maxInvocationsAMD(ivec3);"
"ivec4 maxInvocationsAMD(ivec4);"
"uint maxInvocationsAMD(uint);"
"uvec2 maxInvocationsAMD(uvec2);"
"uvec3 maxInvocationsAMD(uvec3);"
"uvec4 maxInvocationsAMD(uvec4);"
"float addInvocationsAMD(float);"
"vec2 addInvocationsAMD(vec2);"
"vec3 addInvocationsAMD(vec3);"
"vec4 addInvocationsAMD(vec4);"
"int addInvocationsAMD(int);"
"ivec2 addInvocationsAMD(ivec2);"
"ivec3 addInvocationsAMD(ivec3);"
"ivec4 addInvocationsAMD(ivec4);"
"uint addInvocationsAMD(uint);"
"uvec2 addInvocationsAMD(uvec2);"
"uvec3 addInvocationsAMD(uvec3);"
"uvec4 addInvocationsAMD(uvec4);"
"float minInvocationsNonUniformAMD(float);"
"vec2 minInvocationsNonUniformAMD(vec2);"
"vec3 minInvocationsNonUniformAMD(vec3);"
"vec4 minInvocationsNonUniformAMD(vec4);"
"int minInvocationsNonUniformAMD(int);"
"ivec2 minInvocationsNonUniformAMD(ivec2);"
"ivec3 minInvocationsNonUniformAMD(ivec3);"
"ivec4 minInvocationsNonUniformAMD(ivec4);"
"uint minInvocationsNonUniformAMD(uint);"
"uvec2 minInvocationsNonUniformAMD(uvec2);"
"uvec3 minInvocationsNonUniformAMD(uvec3);"
"uvec4 minInvocationsNonUniformAMD(uvec4);"
"float maxInvocationsNonUniformAMD(float);"
"vec2 maxInvocationsNonUniformAMD(vec2);"
"vec3 maxInvocationsNonUniformAMD(vec3);"
"vec4 maxInvocationsNonUniformAMD(vec4);"
"int maxInvocationsNonUniformAMD(int);"
"ivec2 maxInvocationsNonUniformAMD(ivec2);"
"ivec3 maxInvocationsNonUniformAMD(ivec3);"
"ivec4 maxInvocationsNonUniformAMD(ivec4);"
"uint maxInvocationsNonUniformAMD(uint);"
"uvec2 maxInvocationsNonUniformAMD(uvec2);"
"uvec3 maxInvocationsNonUniformAMD(uvec3);"
"uvec4 maxInvocationsNonUniformAMD(uvec4);"
"float addInvocationsNonUniformAMD(float);"
"vec2 addInvocationsNonUniformAMD(vec2);"
"vec3 addInvocationsNonUniformAMD(vec3);"
"vec4 addInvocationsNonUniformAMD(vec4);"
"int addInvocationsNonUniformAMD(int);"
"ivec2 addInvocationsNonUniformAMD(ivec2);"
"ivec3 addInvocationsNonUniformAMD(ivec3);"
"ivec4 addInvocationsNonUniformAMD(ivec4);"
"uint addInvocationsNonUniformAMD(uint);"
"uvec2 addInvocationsNonUniformAMD(uvec2);"
"uvec3 addInvocationsNonUniformAMD(uvec3);"
"uvec4 addInvocationsNonUniformAMD(uvec4);"
"float swizzleInvocationsAMD(float, uvec4);"
"vec2 swizzleInvocationsAMD(vec2, uvec4);"
"vec3 swizzleInvocationsAMD(vec3, uvec4);"
"vec4 swizzleInvocationsAMD(vec4, uvec4);"
"int swizzleInvocationsAMD(int, uvec4);"
"ivec2 swizzleInvocationsAMD(ivec2, uvec4);"
"ivec3 swizzleInvocationsAMD(ivec3, uvec4);"
"ivec4 swizzleInvocationsAMD(ivec4, uvec4);"
"uint swizzleInvocationsAMD(uint, uvec4);"
"uvec2 swizzleInvocationsAMD(uvec2, uvec4);"
"uvec3 swizzleInvocationsAMD(uvec3, uvec4);"
"uvec4 swizzleInvocationsAMD(uvec4, uvec4);"
"float swizzleInvocationsMaskedAMD(float, uvec3);"
"vec2 swizzleInvocationsMaskedAMD(vec2, uvec3);"
"vec3 swizzleInvocationsMaskedAMD(vec3, uvec3);"
"vec4 swizzleInvocationsMaskedAMD(vec4, uvec3);"
"int swizzleInvocationsMaskedAMD(int, uvec3);"
"ivec2 swizzleInvocationsMaskedAMD(ivec2, uvec3);"
"ivec3 swizzleInvocationsMaskedAMD(ivec3, uvec3);"
"ivec4 swizzleInvocationsMaskedAMD(ivec4, uvec3);"
"uint swizzleInvocationsMaskedAMD(uint, uvec3);"
"uvec2 swizzleInvocationsMaskedAMD(uvec2, uvec3);"
"uvec3 swizzleInvocationsMaskedAMD(uvec3, uvec3);"
"uvec4 swizzleInvocationsMaskedAMD(uvec4, uvec3);"
"float writeInvocationAMD(float, float, uint);"
"vec2 writeInvocationAMD(vec2, vec2, uint);"
"vec3 writeInvocationAMD(vec3, vec3, uint);"
"vec4 writeInvocationAMD(vec4, vec4, uint);"
"int writeInvocationAMD(int, int, uint);"
"ivec2 writeInvocationAMD(ivec2, ivec2, uint);"
"ivec3 writeInvocationAMD(ivec3, ivec3, uint);"
"ivec4 writeInvocationAMD(ivec4, ivec4, uint);"
"uint writeInvocationAMD(uint, uint, uint);"
"uvec2 writeInvocationAMD(uvec2, uvec2, uint);"
"uvec3 writeInvocationAMD(uvec3, uvec3, uint);"
"uvec4 writeInvocationAMD(uvec4, uvec4, uint);"
"uint mbcntAMD(uint64_t);"
"\n");
}
// GL_AMD_gcn_shader
if (profile != EEsProfile && version >= 450) {
commonBuiltins.append(
"float cubeFaceIndexAMD(vec3);"
"vec2 cubeFaceCoordAMD(vec3);"
"uint64_t timeAMD();"
"\n");
}
#endif
//============================================================================
//
// Prototypes for built-in functions seen by vertex shaders only.
@ -1713,6 +1943,29 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
"\n");
}
#ifdef AMD_EXTENSIONS
// GL_AMD_shader_explicit_vertex_parameter
if (profile != EEsProfile && version >= 450) {
stageBuiltins[EShLangFragment].append(
"float interpolateAtVertexAMD(float, uint);"
"vec2 interpolateAtVertexAMD(vec2, uint);"
"vec3 interpolateAtVertexAMD(vec3, uint);"
"vec4 interpolateAtVertexAMD(vec4, uint);"
"int interpolateAtVertexAMD(int, uint);"
"ivec2 interpolateAtVertexAMD(ivec2, uint);"
"ivec3 interpolateAtVertexAMD(ivec3, uint);"
"ivec4 interpolateAtVertexAMD(ivec4, uint);"
"uint interpolateAtVertexAMD(uint, uint);"
"uvec2 interpolateAtVertexAMD(uvec2, uint);"
"uvec3 interpolateAtVertexAMD(uvec3, uint);"
"uvec4 interpolateAtVertexAMD(uvec4, uint);"
"\n");
}
#endif
//============================================================================
//
// Standard Uniforms
@ -2332,6 +2585,19 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
"in float gl_CullDistance[];"
"bool gl_HelperInvocation;" // needs qualifier fixed later
);
#ifdef AMD_EXTENSIONS
if (version >= 450)
stageBuiltins[EShLangFragment].append(
"in vec2 gl_BaryCoordNoPerspAMD;"
"in vec2 gl_BaryCoordNoPerspCentroidAMD;"
"in vec2 gl_BaryCoordNoPerspSampleAMD;"
"in vec2 gl_BaryCoordSmoothAMD;"
"in vec2 gl_BaryCoordSmoothCentroidAMD;"
"in vec2 gl_BaryCoordSmoothSampleAMD;"
"in vec3 gl_BaryCoordPullModelAMD;"
);
#endif
} else {
// ES profile
@ -3460,6 +3726,14 @@ void TBuiltIns::initialize(const TBuiltInResource &resources, int version, EProf
s.append(builtInConstant);
}
#ifdef AMD_EXTENSIONS
// GL_AMD_gcn_shader
if (profile != EEsProfile && version >= 450) {
snprintf(builtInConstant, maxSize, "const int gl_SIMDGroupSizeAMD = 64;");
s.append(builtInConstant);
}
#endif
s.append("\n");
}
@ -3588,6 +3862,33 @@ void TBuiltIns::identifyBuiltIns(int version, EProfile profile, const SpvVersion
symbolTable.setFunctionExtensions("allInvocationsEqualARB", 1, &E_GL_ARB_shader_group_vote);
}
#ifdef AMD_EXTENSIONS
if (profile != EEsProfile) {
symbolTable.setFunctionExtensions("minInvocationsAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("maxInvocationsAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("addInvocationsAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("minInvocationsNonUniformAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("maxInvocationsNonUniformAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("addInvocationsNonUniformAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("swizzleInvocationsAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("swizzleInvocationsWithPatternAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("writeInvocationAMD", 1, &E_GL_AMD_shader_ballot);
symbolTable.setFunctionExtensions("mbcntAMD", 1, &E_GL_AMD_shader_ballot);
}
if (profile != EEsProfile) {
symbolTable.setFunctionExtensions("min3", 1, &E_GL_AMD_shader_trinary_minmax);
symbolTable.setFunctionExtensions("max3", 1, &E_GL_AMD_shader_trinary_minmax);
symbolTable.setFunctionExtensions("mid3", 1, &E_GL_AMD_shader_trinary_minmax);
}
if (profile != EEsProfile) {
symbolTable.setFunctionExtensions("cubeFaceIndexAMD", 1, &E_GL_AMD_gcn_shader);
symbolTable.setFunctionExtensions("cubeFaceCoordAMD", 1, &E_GL_AMD_gcn_shader);
symbolTable.setFunctionExtensions("timeAMD", 1, &E_GL_AMD_gcn_shader);
}
#endif
// Compatibility variables, vertex only
if (spvVersion.spv == 0) {
BuiltInVariable("gl_Color", EbvColor, symbolTable);
@ -3872,6 +4173,29 @@ void TBuiltIns::identifyBuiltIns(int version, EProfile profile, const SpvVersion
symbolTable.setFunctionExtensions("textureGradOffsetClampARB", 1, &E_GL_ARB_sparse_texture_clamp);
}
#ifdef AMD_EXTENSIONS
// E_GL_AMD_shader_explicit_vertex_parameter
if (profile != EEsProfile) {
symbolTable.setVariableExtensions("gl_BaryCoordNoPerspAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordNoPerspCentroidAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordNoPerspSampleAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordSmoothAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordSmoothCentroidAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordSmoothSampleAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setVariableExtensions("gl_BaryCoordPullModelAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
symbolTable.setFunctionExtensions("interpolateAtVertexAMD", 1, &E_GL_AMD_shader_explicit_vertex_parameter);
BuiltInVariable("gl_BaryCoordNoPerspAMD", EbvBaryCoordNoPersp, symbolTable);
BuiltInVariable("gl_BaryCoordNoPerspCentroidAMD", EbvBaryCoordNoPerspCentroid, symbolTable);
BuiltInVariable("gl_BaryCoordNoPerspSampleAMD", EbvBaryCoordNoPerspSample, symbolTable);
BuiltInVariable("gl_BaryCoordSmoothAMD", EbvBaryCoordSmooth, symbolTable);
BuiltInVariable("gl_BaryCoordSmoothCentroidAMD", EbvBaryCoordSmoothCentroid, symbolTable);
BuiltInVariable("gl_BaryCoordSmoothSampleAMD", EbvBaryCoordSmoothSample, symbolTable);
BuiltInVariable("gl_BaryCoordPullModelAMD", EbvBaryCoordPullModel, symbolTable);
}
#endif
symbolTable.setVariableExtensions("gl_FragDepthEXT", 1, &E_GL_EXT_frag_depth);
if (profile == EEsProfile) {
@ -4205,6 +4529,27 @@ void TBuiltIns::identifyBuiltIns(int version, EProfile profile, const SpvVersion
symbolTable.relateToOperator("anyInvocationARB", EOpAnyInvocation);
symbolTable.relateToOperator("allInvocationsARB", EOpAllInvocations);
symbolTable.relateToOperator("allInvocationsEqualARB", EOpAllInvocationsEqual);
#ifdef AMD_EXTENSIONS
symbolTable.relateToOperator("minInvocationsAMD", EOpMinInvocations);
symbolTable.relateToOperator("maxInvocationsAMD", EOpMaxInvocations);
symbolTable.relateToOperator("addInvocationsAMD", EOpAddInvocations);
symbolTable.relateToOperator("minInvocationsNonUniformAMD", EOpMinInvocationsNonUniform);
symbolTable.relateToOperator("maxInvocationsNonUniformAMD", EOpMaxInvocationsNonUniform);
symbolTable.relateToOperator("addInvocationsNonUniformAMD", EOpAddInvocationsNonUniform);
symbolTable.relateToOperator("swizzleInvocationsAMD", EOpSwizzleInvocations);
symbolTable.relateToOperator("swizzleInvocationsMaskedAMD", EOpSwizzleInvocationsMasked);
symbolTable.relateToOperator("writeInvocationAMD", EOpWriteInvocation);
symbolTable.relateToOperator("mbcntAMD", EOpMbcnt);
symbolTable.relateToOperator("min3", EOpMin3);
symbolTable.relateToOperator("max3", EOpMax3);
symbolTable.relateToOperator("mid3", EOpMid3);
symbolTable.relateToOperator("cubeFaceIndexAMD", EOpCubeFaceIndex);
symbolTable.relateToOperator("cubeFaceCoordAMD", EOpCubeFaceCoord);
symbolTable.relateToOperator("timeAMD", EOpTime);
#endif
}
}
@ -4238,7 +4583,12 @@ void TBuiltIns::identifyBuiltIns(int version, EProfile profile, const SpvVersion
symbolTable.relateToOperator("interpolateAtCentroid", EOpInterpolateAtCentroid);
symbolTable.relateToOperator("interpolateAtSample", EOpInterpolateAtSample);
symbolTable.relateToOperator("interpolateAtOffset", EOpInterpolateAtOffset);
#ifdef AMD_EXTENSIONS
if (profile != EEsProfile)
symbolTable.relateToOperator("interpolateAtVertexAMD", EOpInterpolateAtVertex);
break;
#endif
case EShLangCompute:
symbolTable.relateToOperator("memoryBarrierShared", EOpMemoryBarrierShared);

17
glslang/MachineIndependent/ParseHelper.cpp
View File

@ -2119,6 +2119,10 @@ void TParseContext::rValueErrorCheck(const TSourceLoc& loc, const char* op, TInt
TIntermSymbol* symNode = node->getAsSymbolNode();
if (symNode && symNode->getQualifier().writeonly)
error(loc, "can't read from writeonly object: ", op, symNode->getName().c_str());
#ifdef AMD_EXTENSIONS
else if (symNode && symNode->getQualifier().explicitInterp)
error(loc, "can't read from explicitly-interpolated object: ", op, symNode->getName().c_str());
#endif
}
//
@ -2665,7 +2669,11 @@ void TParseContext::globalQualifierTypeCheck(const TSourceLoc& loc, const TQuali
publicType.basicType == EbtDouble)
profileRequires(loc, EEsProfile, 300, nullptr, "shader input/output");
if (! qualifier.flat) {
#ifdef AMD_EXTENSIONS
if (! qualifier.flat && ! qualifier.explicitInterp) {
#else
if (!qualifier.flat) {
#endif
if (publicType.basicType == EbtInt || publicType.basicType == EbtUint ||
publicType.basicType == EbtInt64 || publicType.basicType == EbtUint64 ||
publicType.basicType == EbtDouble ||
@ -2802,7 +2810,11 @@ void TParseContext::mergeQualifiers(const TSourceLoc& loc, TQualifier& dst, cons
// Multiple interpolation qualifiers (mostly done later by 'individual qualifiers')
if (src.isInterpolation() && dst.isInterpolation())
#ifdef AMD_EXTENSIONS
error(loc, "can only have one interpolation qualifier (flat, smooth, noperspective, __explicitInterpAMD)", "", "");
#else
error(loc, "can only have one interpolation qualifier (flat, smooth, noperspective)", "", "");
#endif
// Ordering
if (! force && ((profile != EEsProfile && version < 420) ||
@ -2858,6 +2870,9 @@ void TParseContext::mergeQualifiers(const TSourceLoc& loc, TQualifier& dst, cons
MERGE_SINGLETON(smooth);
MERGE_SINGLETON(flat);
MERGE_SINGLETON(nopersp);
#ifdef AMD_EXTENSIONS
MERGE_SINGLETON(explicitInterp);
#endif
MERGE_SINGLETON(patch);
MERGE_SINGLETON(sample);
MERGE_SINGLETON(coherent);

11
glslang/MachineIndependent/Scan.cpp
View File

@ -553,6 +553,9 @@ void TScanContext::fillInKeywordMap()
(*KeywordMap)["noperspective"] = NOPERSPECTIVE;
(*KeywordMap)["smooth"] = SMOOTH;
(*KeywordMap)["flat"] = FLAT;
#ifdef AMD_EXTENSIONS
(*KeywordMap)["__explicitInterpAMD"] = __EXPLICITINTERPAMD;
#endif
(*KeywordMap)["centroid"] = CENTROID;
(*KeywordMap)["precise"] = PRECISE;
(*KeywordMap)["invariant"] = INVARIANT;
@ -1117,6 +1120,14 @@ int TScanContext::tokenizeIdentifier()
return identifierOrType();
return keyword;
#ifdef AMD_EXTENSIONS
case __EXPLICITINTERPAMD:
if (parseContext.profile != EEsProfile && parseContext.version >= 450 &&
parseContext.extensionTurnedOn(E_GL_AMD_shader_explicit_vertex_parameter))
return keyword;
return identifierOrType();
#endif
case FLAT:
if (parseContext.profile == EEsProfile && parseContext.version < 300)
reservedWord();

14
glslang/MachineIndependent/Versions.cpp
View File

@ -187,6 +187,13 @@ void TParseVersions::initializeExtensionBehavior()
extensionBehavior[E_GL_GOOGLE_cpp_style_line_directive] = EBhDisable;
extensionBehavior[E_GL_GOOGLE_include_directive] = EBhDisable;
#ifdef AMD_EXTENSIONS
extensionBehavior[E_GL_AMD_shader_ballot] = EBhDisable;
extensionBehavior[E_GL_AMD_shader_trinary_minmax] = EBhDisable;
extensionBehavior[E_GL_AMD_shader_explicit_vertex_parameter] = EBhDisable;
extensionBehavior[E_GL_AMD_gcn_shader] = EBhDisable;
#endif
// AEP
extensionBehavior[E_GL_ANDROID_extension_pack_es31a] = EBhDisable;
extensionBehavior[E_GL_KHR_blend_equation_advanced] = EBhDisable;
@ -286,6 +293,13 @@ void TParseVersions::getPreamble(std::string& preamble)
"#define GL_ARB_sparse_texture_clamp 1\n"
// "#define GL_ARB_cull_distance 1\n" // present for 4.5, but need extension control over block members
"#define GL_EXT_shader_non_constant_global_initializers 1\n"
#ifdef AMD_EXTENSIONS
"#define GL_AMD_shader_ballot 1\n"
"#define GL_AMD_shader_trinary_minmax 1\n"
"#define GL_AMD_shader_explicit_vertex_parameter 1\n"
"#define GL_AMD_gcn_shader 1\n"
#endif
;
}

7
glslang/MachineIndependent/Versions.h
View File

@ -135,6 +135,13 @@ const char* const E_GL_EXT_shader_non_constant_global_initializers = "GL_EXT_sha
const char* const E_GL_GOOGLE_cpp_style_line_directive = "GL_GOOGLE_cpp_style_line_directive";
const char* const E_GL_GOOGLE_include_directive = "GL_GOOGLE_include_directive";
#ifdef AMD_EXTENSIONS
const char* const E_GL_AMD_shader_ballot = "GL_AMD_shader_ballot";
const char* const E_GL_AMD_shader_trinary_minmax = "GL_AMD_shader_trinary_minmax";
const char* const E_GL_AMD_shader_explicit_vertex_parameter = "GL_AMD_shader_explicit_vertex_parameter";
const char* const E_GL_AMD_gcn_shader = "GL_AMD_gcn_shader";
#endif
// AEP
const char* const E_GL_ANDROID_extension_pack_es31a = "GL_ANDROID_extension_pack_es31a";
const char* const E_GL_KHR_blend_equation_advanced = "GL_KHR_blend_equation_advanced";

11
glslang/MachineIndependent/glslang.y
View File

@ -126,7 +126,7 @@ extern int yylex(YYSTYPE*, TParseContext&);
%token <lex> UNIFORM PATCH SAMPLE BUFFER SHARED
%token <lex> COHERENT VOLATILE RESTRICT READONLY WRITEONLY
%token <lex> DVEC2 DVEC3 DVEC4 DMAT2 DMAT3 DMAT4
%token <lex> NOPERSPECTIVE FLAT SMOOTH LAYOUT
%token <lex> NOPERSPECTIVE FLAT SMOOTH LAYOUT __EXPLICITINTERPAMD
%token <lex> MAT2X2 MAT2X3 MAT2X4
%token <lex> MAT3X2 MAT3X3 MAT3X4
@ -1050,6 +1050,15 @@ interpolation_qualifier
$$.init($1.loc);
$$.qualifier.nopersp = true;
}
| __EXPLICITINTERPAMD {
#ifdef AMD_EXTENSIONS
parseContext.globalCheck($1.loc, "__explicitInterpAMD");
parseContext.profileRequires($1.loc, ECoreProfile, 450, E_GL_AMD_shader_explicit_vertex_parameter, "explicit interpolation");
parseContext.profileRequires($1.loc, ECompatibilityProfile, 450, E_GL_AMD_shader_explicit_vertex_parameter, "explicit interpolation");
$$.init($1.loc);
$$.qualifier.explicitInterp = true;
#endif
}
;
layout_qualifier

7927
glslang/MachineIndependent/glslang_tab.cpp
View File

File diff suppressed because it is too large Load Diff

594
glslang/MachineIndependent/glslang_tab.cpp.h
View File

@ -1,19 +1,19 @@
/* A Bison parser, made by GNU Bison 3.0.4. */
/* A Bison parser, made by GNU Bison 2.7. */
/* Bison interface for Yacc-like parsers in C
Copyright (C) 1984, 1989-1990, 2000-2015 Free Software Foundation, Inc.
Copyright (C) 1984, 1989-1990, 2000-2012 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
@ -26,13 +26,13 @@
special exception, which will cause the skeleton and the resulting
Bison output files to be licensed under the GNU General Public
License without this special exception.
This special exception was added by the Free Software Foundation in
version 2.2 of Bison. */
#ifndef YY_YY_MACHINEINDEPENDENT_GLSLANG_TAB_CPP_H_INCLUDED
# define YY_YY_MACHINEINDEPENDENT_GLSLANG_TAB_CPP_H_INCLUDED
/* Debug traces. */
#ifndef YY_YY_GLSLANG_TAB_CPP_H_INCLUDED
# define YY_YY_GLSLANG_TAB_CPP_H_INCLUDED
/* Enabling traces. */
#ifndef YYDEBUG
# define YYDEBUG 1
#endif
@ -40,287 +40,289 @@
extern int yydebug;
#endif
/* Token type. */
/* Tokens. */
#ifndef YYTOKENTYPE
# define YYTOKENTYPE
enum yytokentype
{
ATTRIBUTE = 258,
VARYING = 259,
CONST = 260,
BOOL = 261,
FLOAT = 262,
DOUBLE = 263,
INT = 264,
UINT = 265,
INT64_T = 266,
UINT64_T = 267,
BREAK = 268,
CONTINUE = 269,
DO = 270,
ELSE = 271,
FOR = 272,
IF = 273,
DISCARD = 274,
RETURN = 275,
SWITCH = 276,
CASE = 277,
DEFAULT = 278,
SUBROUTINE = 279,
BVEC2 = 280,
BVEC3 = 281,
BVEC4 = 282,
IVEC2 = 283,
IVEC3 = 284,
IVEC4 = 285,
I64VEC2 = 286,
I64VEC3 = 287,
I64VEC4 = 288,
UVEC2 = 289,
UVEC3 = 290,
UVEC4 = 291,
U64VEC2 = 292,
U64VEC3 = 293,
U64VEC4 = 294,
VEC2 = 295,
VEC3 = 296,
VEC4 = 297,
MAT2 = 298,
MAT3 = 299,
MAT4 = 300,
CENTROID = 301,
IN = 302,
OUT = 303,
INOUT = 304,
UNIFORM = 305,
PATCH = 306,
SAMPLE = 307,
BUFFER = 308,
SHARED = 309,
COHERENT = 310,
VOLATILE = 311,
RESTRICT = 312,
READONLY = 313,
WRITEONLY = 314,
DVEC2 = 315,
DVEC3 = 316,
DVEC4 = 317,
DMAT2 = 318,
DMAT3 = 319,
DMAT4 = 320,
NOPERSPECTIVE = 321,
FLAT = 322,
SMOOTH = 323,
LAYOUT = 324,
MAT2X2 = 325,
MAT2X3 = 326,
MAT2X4 = 327,
MAT3X2 = 328,
MAT3X3 = 329,
MAT3X4 = 330,
MAT4X2 = 331,
MAT4X3 = 332,
MAT4X4 = 333,
DMAT2X2 = 334,
DMAT2X3 = 335,
DMAT2X4 = 336,
DMAT3X2 = 337,
DMAT3X3 = 338,
DMAT3X4 = 339,
DMAT4X2 = 340,
DMAT4X3 = 341,
DMAT4X4 = 342,
ATOMIC_UINT = 343,
SAMPLER1D = 344,
SAMPLER2D = 345,
SAMPLER3D = 346,
SAMPLERCUBE = 347,
SAMPLER1DSHADOW = 348,
SAMPLER2DSHADOW = 349,
SAMPLERCUBESHADOW = 350,
SAMPLER1DARRAY = 351,
SAMPLER2DARRAY = 352,
SAMPLER1DARRAYSHADOW = 353,
SAMPLER2DARRAYSHADOW = 354,
ISAMPLER1D = 355,
ISAMPLER2D = 356,
ISAMPLER3D = 357,
ISAMPLERCUBE = 358,
ISAMPLER1DARRAY = 359,
ISAMPLER2DARRAY = 360,
USAMPLER1D = 361,
USAMPLER2D = 362,
USAMPLER3D = 363,
USAMPLERCUBE = 364,
USAMPLER1DARRAY = 365,
USAMPLER2DARRAY = 366,
SAMPLER2DRECT = 367,
SAMPLER2DRECTSHADOW = 368,
ISAMPLER2DRECT = 369,
USAMPLER2DRECT = 370,
SAMPLERBUFFER = 371,
ISAMPLERBUFFER = 372,
USAMPLERBUFFER = 373,
SAMPLERCUBEARRAY = 374,
SAMPLERCUBEARRAYSHADOW = 375,
ISAMPLERCUBEARRAY = 376,
USAMPLERCUBEARRAY = 377,
SAMPLER2DMS = 378,
ISAMPLER2DMS = 379,
USAMPLER2DMS = 380,
SAMPLER2DMSARRAY = 381,
ISAMPLER2DMSARRAY = 382,
USAMPLER2DMSARRAY = 383,
SAMPLEREXTERNALOES = 384,
SAMPLER = 385,
SAMPLERSHADOW = 386,
TEXTURE1D = 387,
TEXTURE2D = 388,
TEXTURE3D = 389,
TEXTURECUBE = 390,
TEXTURE1DARRAY = 391,
TEXTURE2DARRAY = 392,
ITEXTURE1D = 393,
ITEXTURE2D = 394,
ITEXTURE3D = 395,
ITEXTURECUBE = 396,
ITEXTURE1DARRAY = 397,
ITEXTURE2DARRAY = 398,
UTEXTURE1D = 399,
UTEXTURE2D = 400,
UTEXTURE3D = 401,
UTEXTURECUBE = 402,
UTEXTURE1DARRAY = 403,
UTEXTURE2DARRAY = 404,
TEXTURE2DRECT = 405,
ITEXTURE2DRECT = 406,
UTEXTURE2DRECT = 407,
TEXTUREBUFFER = 408,
ITEXTUREBUFFER = 409,
UTEXTUREBUFFER = 410,
TEXTURECUBEARRAY = 411,
ITEXTURECUBEARRAY = 412,
UTEXTURECUBEARRAY = 413,
TEXTURE2DMS = 414,
ITEXTURE2DMS = 415,
UTEXTURE2DMS = 416,
TEXTURE2DMSARRAY = 417,
ITEXTURE2DMSARRAY = 418,
UTEXTURE2DMSARRAY = 419,
SUBPASSINPUT = 420,
SUBPASSINPUTMS = 421,
ISUBPASSINPUT = 422,
ISUBPASSINPUTMS = 423,
USUBPASSINPUT = 424,
USUBPASSINPUTMS = 425,
IMAGE1D = 426,
IIMAGE1D = 427,
UIMAGE1D = 428,
IMAGE2D = 429,
IIMAGE2D = 430,
UIMAGE2D = 431,
IMAGE3D = 432,
IIMAGE3D = 433,
UIMAGE3D = 434,
IMAGE2DRECT = 435,
IIMAGE2DRECT = 436,
UIMAGE2DRECT = 437,
IMAGECUBE = 438,
IIMAGECUBE = 439,
UIMAGECUBE = 440,
IMAGEBUFFER = 441,
IIMAGEBUFFER = 442,
UIMAGEBUFFER = 443,
IMAGE1DARRAY = 444,
IIMAGE1DARRAY = 445,
UIMAGE1DARRAY = 446,
IMAGE2DARRAY = 447,
IIMAGE2DARRAY = 448,
UIMAGE2DARRAY = 449,
IMAGECUBEARRAY = 450,
IIMAGECUBEARRAY = 451,
UIMAGECUBEARRAY = 452,
IMAGE2DMS = 453,
IIMAGE2DMS = 454,
UIMAGE2DMS = 455,
IMAGE2DMSARRAY = 456,
IIMAGE2DMSARRAY = 457,
UIMAGE2DMSARRAY = 458,
STRUCT = 459,
VOID = 460,
WHILE = 461,
IDENTIFIER = 462,
TYPE_NAME = 463,
FLOATCONSTANT = 464,
DOUBLECONSTANT = 465,
INTCONSTANT = 466,
UINTCONSTANT = 467,
INT64CONSTANT = 468,
UINT64CONSTANT = 469,
BOOLCONSTANT = 470,
LEFT_OP = 471,
RIGHT_OP = 472,
INC_OP = 473,
DEC_OP = 474,
LE_OP = 475,
GE_OP = 476,
EQ_OP = 477,
NE_OP = 478,
AND_OP = 479,
OR_OP = 480,
XOR_OP = 481,
MUL_ASSIGN = 482,
DIV_ASSIGN = 483,
ADD_ASSIGN = 484,
MOD_ASSIGN = 485,
LEFT_ASSIGN = 486,
RIGHT_ASSIGN = 487,
AND_ASSIGN = 488,
XOR_ASSIGN = 489,
OR_ASSIGN = 490,
SUB_ASSIGN = 491,
LEFT_PAREN = 492,
RIGHT_PAREN = 493,
LEFT_BRACKET = 494,
RIGHT_BRACKET = 495,
LEFT_BRACE = 496,
RIGHT_BRACE = 497,
DOT = 498,
COMMA = 499,
COLON = 500,
EQUAL = 501,
SEMICOLON = 502,
BANG = 503,
DASH = 504,
TILDE = 505,
PLUS = 506,
STAR = 507,
SLASH = 508,
PERCENT = 509,
LEFT_ANGLE = 510,
RIGHT_ANGLE = 511,
VERTICAL_BAR = 512,
CARET = 513,
AMPERSAND = 514,
QUESTION = 515,
INVARIANT = 516,
PRECISE = 517,
HIGH_PRECISION = 518,
MEDIUM_PRECISION = 519,
LOW_PRECISION = 520,
PRECISION = 521,
PACKED = 522,
RESOURCE = 523,
SUPERP = 524
};
/* Put the tokens into the symbol table, so that GDB and other debuggers
know about them. */
enum yytokentype {
ATTRIBUTE = 258,
VARYING = 259,
CONST = 260,
BOOL = 261,
FLOAT = 262,
DOUBLE = 263,
INT = 264,
UINT = 265,
INT64_T = 266,
UINT64_T = 267,
BREAK = 268,
CONTINUE = 269,
DO = 270,
ELSE = 271,
FOR = 272,
IF = 273,
DISCARD = 274,
RETURN = 275,
SWITCH = 276,
CASE = 277,
DEFAULT = 278,
SUBROUTINE = 279,
BVEC2 = 280,
BVEC3 = 281,
BVEC4 = 282,
IVEC2 = 283,
IVEC3 = 284,
IVEC4 = 285,
I64VEC2 = 286,
I64VEC3 = 287,
I64VEC4 = 288,
UVEC2 = 289,
UVEC3 = 290,
UVEC4 = 291,
U64VEC2 = 292,
U64VEC3 = 293,
U64VEC4 = 294,
VEC2 = 295,
VEC3 = 296,
VEC4 = 297,
MAT2 = 298,
MAT3 = 299,
MAT4 = 300,
CENTROID = 301,
IN = 302,
OUT = 303,
INOUT = 304,
UNIFORM = 305,
PATCH = 306,
SAMPLE = 307,
BUFFER = 308,
SHARED = 309,
COHERENT = 310,
VOLATILE = 311,
RESTRICT = 312,
READONLY = 313,
WRITEONLY = 314,
DVEC2 = 315,
DVEC3 = 316,
DVEC4 = 317,
DMAT2 = 318,
DMAT3 = 319,
DMAT4 = 320,
NOPERSPECTIVE = 321,
FLAT = 322,
SMOOTH = 323,
LAYOUT = 324,
__EXPLICITINTERPAMD = 325,
MAT2X2 = 326,
MAT2X3 = 327,
MAT2X4 = 328,
MAT3X2 = 329,
MAT3X3 = 330,
MAT3X4 = 331,
MAT4X2 = 332,
MAT4X3 = 333,
MAT4X4 = 334,
DMAT2X2 = 335,
DMAT2X3 = 336,
DMAT2X4 = 337,
DMAT3X2 = 338,
DMAT3X3 = 339,
DMAT3X4 = 340,
DMAT4X2 = 341,
DMAT4X3 = 342,
DMAT4X4 = 343,
ATOMIC_UINT = 344,
SAMPLER1D = 345,
SAMPLER2D = 346,
SAMPLER3D = 347,
SAMPLERCUBE = 348,
SAMPLER1DSHADOW = 349,
SAMPLER2DSHADOW = 350,
SAMPLERCUBESHADOW = 351,
SAMPLER1DARRAY = 352,
SAMPLER2DARRAY = 353,
SAMPLER1DARRAYSHADOW = 354,
SAMPLER2DARRAYSHADOW = 355,
ISAMPLER1D = 356,
ISAMPLER2D = 357,
ISAMPLER3D = 358,
ISAMPLERCUBE = 359,
ISAMPLER1DARRAY = 360,
ISAMPLER2DARRAY = 361,
USAMPLER1D = 362,
USAMPLER2D = 363,
USAMPLER3D = 364,
USAMPLERCUBE = 365,
USAMPLER1DARRAY = 366,
USAMPLER2DARRAY = 367,
SAMPLER2DRECT = 368,
SAMPLER2DRECTSHADOW = 369,
ISAMPLER2DRECT = 370,
USAMPLER2DRECT = 371,
SAMPLERBUFFER = 372,
ISAMPLERBUFFER = 373,
USAMPLERBUFFER = 374,
SAMPLERCUBEARRAY = 375,
SAMPLERCUBEARRAYSHADOW = 376,
ISAMPLERCUBEARRAY = 377,
USAMPLERCUBEARRAY = 378,
SAMPLER2DMS = 379,
ISAMPLER2DMS = 380,
USAMPLER2DMS = 381,
SAMPLER2DMSARRAY = 382,
ISAMPLER2DMSARRAY = 383,
USAMPLER2DMSARRAY = 384,
SAMPLEREXTERNALOES = 385,
SAMPLER = 386,
SAMPLERSHADOW = 387,
TEXTURE1D = 388,
TEXTURE2D = 389,
TEXTURE3D = 390,
TEXTURECUBE = 391,
TEXTURE1DARRAY = 392,
TEXTURE2DARRAY = 393,
ITEXTURE1D = 394,
ITEXTURE2D = 395,
ITEXTURE3D = 396,
ITEXTURECUBE = 397,
ITEXTURE1DARRAY = 398,
ITEXTURE2DARRAY = 399,
UTEXTURE1D = 400,
UTEXTURE2D = 401,
UTEXTURE3D = 402,
UTEXTURECUBE = 403,
UTEXTURE1DARRAY = 404,
UTEXTURE2DARRAY = 405,
TEXTURE2DRECT = 406,
ITEXTURE2DRECT = 407,
UTEXTURE2DRECT = 408,
TEXTUREBUFFER = 409,
ITEXTUREBUFFER = 410,
UTEXTUREBUFFER = 411,
TEXTURECUBEARRAY = 412,
ITEXTURECUBEARRAY = 413,
UTEXTURECUBEARRAY = 414,
TEXTURE2DMS = 415,
ITEXTURE2DMS = 416,
UTEXTURE2DMS = 417,
TEXTURE2DMSARRAY = 418,
ITEXTURE2DMSARRAY = 419,
UTEXTURE2DMSARRAY = 420,
SUBPASSINPUT = 421,
SUBPASSINPUTMS = 422,
ISUBPASSINPUT = 423,
ISUBPASSINPUTMS = 424,
USUBPASSINPUT = 425,
USUBPASSINPUTMS = 426,
IMAGE1D = 427,
IIMAGE1D = 428,
UIMAGE1D = 429,
IMAGE2D = 430,
IIMAGE2D = 431,
UIMAGE2D = 432,
IMAGE3D = 433,
IIMAGE3D = 434,
UIMAGE3D = 435,
IMAGE2DRECT = 436,
IIMAGE2DRECT = 437,
UIMAGE2DRECT = 438,
IMAGECUBE = 439,
IIMAGECUBE = 440,
UIMAGECUBE = 441,
IMAGEBUFFER = 442,
IIMAGEBUFFER = 443,
UIMAGEBUFFER = 444,
IMAGE1DARRAY = 445,
IIMAGE1DARRAY = 446,
UIMAGE1DARRAY = 447,
IMAGE2DARRAY = 448,
IIMAGE2DARRAY = 449,
UIMAGE2DARRAY = 450,
IMAGECUBEARRAY = 451,
IIMAGECUBEARRAY = 452,
UIMAGECUBEARRAY = 453,
IMAGE2DMS = 454,
IIMAGE2DMS = 455,
UIMAGE2DMS = 456,
IMAGE2DMSARRAY = 457,
IIMAGE2DMSARRAY = 458,
UIMAGE2DMSARRAY = 459,
STRUCT = 460,
VOID = 461,
WHILE = 462,
IDENTIFIER = 463,
TYPE_NAME = 464,
FLOATCONSTANT = 465,
DOUBLECONSTANT = 466,
INTCONSTANT = 467,
UINTCONSTANT = 468,
INT64CONSTANT = 469,
UINT64CONSTANT = 470,
BOOLCONSTANT = 471,
LEFT_OP = 472,
RIGHT_OP = 473,
INC_OP = 474,
DEC_OP = 475,
LE_OP = 476,
GE_OP = 477,
EQ_OP = 478,
NE_OP = 479,
AND_OP = 480,
OR_OP = 481,
XOR_OP = 482,
MUL_ASSIGN = 483,
DIV_ASSIGN = 484,
ADD_ASSIGN = 485,
MOD_ASSIGN = 486,
LEFT_ASSIGN = 487,
RIGHT_ASSIGN = 488,
AND_ASSIGN = 489,
XOR_ASSIGN = 490,
OR_ASSIGN = 491,
SUB_ASSIGN = 492,
LEFT_PAREN = 493,
RIGHT_PAREN = 494,
LEFT_BRACKET = 495,
RIGHT_BRACKET = 496,
LEFT_BRACE = 497,
RIGHT_BRACE = 498,
DOT = 499,
COMMA = 500,
COLON = 501,
EQUAL = 502,
SEMICOLON = 503,
BANG = 504,
DASH = 505,
TILDE = 506,
PLUS = 507,
STAR = 508,
SLASH = 509,
PERCENT = 510,
LEFT_ANGLE = 511,
RIGHT_ANGLE = 512,
VERTICAL_BAR = 513,
CARET = 514,
AMPERSAND = 515,
QUESTION = 516,
INVARIANT = 517,
PRECISE = 518,
HIGH_PRECISION = 519,
MEDIUM_PRECISION = 520,
LOW_PRECISION = 521,
PRECISION = 522,
PACKED = 523,
RESOURCE = 524,
SUPERP = 525
};
#endif
/* Value type. */
#if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED
union YYSTYPE
#if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED
typedef union YYSTYPE
{
#line 66 "MachineIndependent/glslang.y" /* yacc.c:1909 */
/* Line 2058 of yacc.c */
#line 66 "glslang.y"
struct {
glslang::TSourceLoc loc;
@ -354,16 +356,28 @@ union YYSTYPE
};
} interm;
#line 358 "MachineIndependent/glslang_tab.cpp.h" /* yacc.c:1909 */
};
typedef union YYSTYPE YYSTYPE;
/* Line 2058 of yacc.c */
#line 362 "glslang_tab.cpp.h"
} YYSTYPE;
# define YYSTYPE_IS_TRIVIAL 1
# define yystype YYSTYPE /* obsolescent; will be withdrawn */
# define YYSTYPE_IS_DECLARED 1
#endif
#ifdef YYPARSE_PARAM
#if defined __STDC__ || defined __cplusplus
int yyparse (void *YYPARSE_PARAM);
#else
int yyparse ();
#endif
#else /* ! YYPARSE_PARAM */
#if defined __STDC__ || defined __cplusplus
int yyparse (glslang::TParseContext* pParseContext);
#else
int yyparse ();
#endif
#endif /* ! YYPARSE_PARAM */
#endif /* !YY_YY_MACHINEINDEPENDENT_GLSLANG_TAB_CPP_H_INCLUDED */
#endif /* !YY_YY_GLSLANG_TAB_CPP_H_INCLUDED */

29
glslang/MachineIndependent/intermOut.cpp
View File

@ -355,6 +355,7 @@ bool TOutputTraverser::visitUnary(TVisit /* visit */, TIntermUnary* node)
case EOpBallot: out.debug << "ballot"; break;
case EOpReadFirstInvocation: out.debug << "readFirstInvocation"; break;
case EOpAnyInvocation: out.debug << "anyInvocation"; break;
case EOpAllInvocations: out.debug << "allInvocations"; break;
case EOpAllInvocationsEqual: out.debug << "allInvocationsEqual"; break;
@ -365,6 +366,19 @@ bool TOutputTraverser::visitUnary(TVisit /* visit */, TIntermUnary* node)
case EOpRcp: out.debug << "rcp"; break;
case EOpSaturate: out.debug << "saturate"; break;
#ifdef AMD_EXTENSIONS
case EOpMinInvocations: out.debug << "minInvocations"; break;
case EOpMaxInvocations: out.debug << "maxInvocations"; break;
case EOpAddInvocations: out.debug << "addInvocations"; break;
case EOpMinInvocationsNonUniform: out.debug << "minInvocationsNonUniform"; break;
case EOpMaxInvocationsNonUniform: out.debug << "maxInvocationsNonUniform"; break;
case EOpAddInvocationsNonUniform: out.debug << "addInvocationsNonUniform"; break;
case EOpMbcnt: out.debug << "mbcnt"; break;
case EOpCubeFaceIndex: out.debug << "cubeFaceIndex"; break;
case EOpCubeFaceCoord: out.debug << "cubeFaceCoord"; break;
#endif
default: out.debug.message(EPrefixError, "Bad unary op");
}
@ -482,6 +496,18 @@ bool TOutputTraverser::visitAggregate(TVisit /* visit */, TIntermAggregate* node
case EOpReadInvocation: out.debug << "readInvocation"; break;
#ifdef AMD_EXTENSIONS
case EOpSwizzleInvocations: out.debug << "swizzleInvocations"; break;
case EOpSwizzleInvocationsMasked: out.debug << "swizzleInvocationsMasked"; break;
case EOpWriteInvocation: out.debug << "writeInvocation"; break;
case EOpMin3: out.debug << "min3"; break;
case EOpMax3: out.debug << "max3"; break;
case EOpMid3: out.debug << "mid3"; break;
case EOpTime: out.debug << "time"; break;
#endif
case EOpAtomicAdd: out.debug << "AtomicAdd"; break;
case EOpAtomicMin: out.debug << "AtomicMin"; break;
case EOpAtomicMax: out.debug << "AtomicMax"; break;
@ -539,6 +565,9 @@ bool TOutputTraverser::visitAggregate(TVisit /* visit */, TIntermAggregate* node
case EOpInterpolateAtSample: out.debug << "interpolateAtSample"; break;
case EOpInterpolateAtOffset: out.debug << "interpolateAtOffset"; break;
#ifdef AMD_EXTENSIONS
case EOpInterpolateAtVertex: out.debug << "interpolateAtVertex"; break;
#endif
case EOpSinCos: out.debug << "sincos"; break;
case EOpGenMul: out.debug << "mul"; break;