Merge pull request #2668 from cmarcelo/GLSL_EXT_shader_atomic_float2
Implement GLSL_EXT_shader_atomic_float2
This commit is contained in:
Greg Fischer
GitHub
commit
b9d06bdb01
9 changed files with 319 additions and 24 deletions
|
|
@ -1436,11 +1436,23 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
|
|||
" int64_t atomicMin(coherent volatile inout int64_t, int64_t);"
|
||||
"uint64_t atomicMin(coherent volatile inout uint64_t, uint64_t, int, int, int);"
|
||||
" int64_t atomicMin(coherent volatile inout int64_t, int64_t, int, int, int);"
|
||||
"float16_t atomicMin(coherent volatile inout float16_t, float16_t);"
|
||||
"float16_t atomicMin(coherent volatile inout float16_t, float16_t, int, int, int);"
|
||||
" float atomicMin(coherent volatile inout float, float);"
|
||||
" float atomicMin(coherent volatile inout float, float, int, int, int);"
|
||||
" double atomicMin(coherent volatile inout double, double);"
|
||||
" double atomicMin(coherent volatile inout double, double, int, int, int);"
|
||||
|
||||
"uint64_t atomicMax(coherent volatile inout uint64_t, uint64_t);"
|
||||
" int64_t atomicMax(coherent volatile inout int64_t, int64_t);"
|
||||
"uint64_t atomicMax(coherent volatile inout uint64_t, uint64_t, int, int, int);"
|
||||
" int64_t atomicMax(coherent volatile inout int64_t, int64_t, int, int, int);"
|
||||
"float16_t atomicMax(coherent volatile inout float16_t, float16_t);"
|
||||
"float16_t atomicMax(coherent volatile inout float16_t, float16_t, int, int, int);"
|
||||
" float atomicMax(coherent volatile inout float, float);"
|
||||
" float atomicMax(coherent volatile inout float, float, int, int, int);"
|
||||
" double atomicMax(coherent volatile inout double, double);"
|
||||
" double atomicMax(coherent volatile inout double, double, int, int, int);"
|
||||
|
||||
"uint64_t atomicAnd(coherent volatile inout uint64_t, uint64_t);"
|
||||
" int64_t atomicAnd(coherent volatile inout int64_t, int64_t);"
|
||||
|
|
@ -1461,6 +1473,8 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
|
|||
" int64_t atomicAdd(coherent volatile inout int64_t, int64_t);"
|
||||
"uint64_t atomicAdd(coherent volatile inout uint64_t, uint64_t, int, int, int);"
|
||||
" int64_t atomicAdd(coherent volatile inout int64_t, int64_t, int, int, int);"
|
||||
"float16_t atomicAdd(coherent volatile inout float16_t, float16_t);"
|
||||
"float16_t atomicAdd(coherent volatile inout float16_t, float16_t, int, int, int);"
|
||||
" float atomicAdd(coherent volatile inout float, float);"
|
||||
" float atomicAdd(coherent volatile inout float, float, int, int, int);"
|
||||
" double atomicAdd(coherent volatile inout double, double);"
|
||||
|
|
@ -1470,6 +1484,8 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
|
|||
" int64_t atomicExchange(coherent volatile inout int64_t, int64_t);"
|
||||
"uint64_t atomicExchange(coherent volatile inout uint64_t, uint64_t, int, int, int);"
|
||||
" int64_t atomicExchange(coherent volatile inout int64_t, int64_t, int, int, int);"
|
||||
"float16_t atomicExchange(coherent volatile inout float16_t, float16_t);"
|
||||
"float16_t atomicExchange(coherent volatile inout float16_t, float16_t, int, int, int);"
|
||||
" float atomicExchange(coherent volatile inout float, float);"
|
||||
" float atomicExchange(coherent volatile inout float, float, int, int, int);"
|
||||
" double atomicExchange(coherent volatile inout double, double);"
|
||||
|
|
@ -1482,11 +1498,13 @@ void TBuiltIns::initialize(int version, EProfile profile, const SpvVersion& spvV
|
|||
|
||||
"uint64_t atomicLoad(coherent volatile in uint64_t, int, int, int);"
|
||||
" int64_t atomicLoad(coherent volatile in int64_t, int, int, int);"
|
||||
"float16_t atomicLoad(coherent volatile in float16_t, int, int, int);"
|
||||
" float atomicLoad(coherent volatile in float, int, int, int);"
|
||||
" double atomicLoad(coherent volatile in double, int, int, int);"
|
||||
|
||||
"void atomicStore(coherent volatile out uint64_t, uint64_t, int, int, int);"
|
||||
"void atomicStore(coherent volatile out int64_t, int64_t, int, int, int);"
|
||||
"void atomicStore(coherent volatile out float16_t, float16_t, int, int, int);"
|
||||
"void atomicStore(coherent volatile out float, float, int, int, int);"
|
||||
"void atomicStore(coherent volatile out double, double, int, int, int);"
|
||||
"\n");
|
||||
|
|
@ -6478,6 +6496,24 @@ void TBuiltIns::addImageFunctions(TSampler sampler, const TString& typeName, int
|
|||
commonBuiltins.append(imageParams);
|
||||
commonBuiltins.append(", float");
|
||||
commonBuiltins.append(", int, int, int);\n");
|
||||
|
||||
commonBuiltins.append("float imageAtomicMin(volatile coherent ");
|
||||
commonBuiltins.append(imageParams);
|
||||
commonBuiltins.append(", float);\n");
|
||||
|
||||
commonBuiltins.append("float imageAtomicMin(volatile coherent ");
|
||||
commonBuiltins.append(imageParams);
|
||||
commonBuiltins.append(", float");
|
||||
commonBuiltins.append(", int, int, int);\n");
|
||||
|
||||
commonBuiltins.append("float imageAtomicMax(volatile coherent ");
|
||||
commonBuiltins.append(imageParams);
|
||||
commonBuiltins.append(", float);\n");
|
||||
|
||||
commonBuiltins.append("float imageAtomicMax(volatile coherent ");
|
||||
commonBuiltins.append(imageParams);
|
||||
commonBuiltins.append(", float");
|
||||
commonBuiltins.append(", int, int, int);\n");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -2341,18 +2341,23 @@ void TParseContext::builtInOpCheck(const TSourceLoc& loc, const TFunction& fnCan
|
|||
error(loc, "only supported on image with format r64i", fnCandidate.getName().c_str(), "");
|
||||
else if (callNode.getType().getBasicType() == EbtUint64 && imageType.getQualifier().getFormat() != ElfR64ui)
|
||||
error(loc, "only supported on image with format r64ui", fnCandidate.getName().c_str(), "");
|
||||
} else {
|
||||
bool isImageAtomicOnFloatAllowed = ((fnCandidate.getName().compare(0, 14, "imageAtomicAdd") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 15, "imageAtomicLoad") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 16, "imageAtomicStore") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 19, "imageAtomicExchange") == 0));
|
||||
if (imageType.getSampler().type == EbtFloat && isImageAtomicOnFloatAllowed &&
|
||||
(fnCandidate.getName().compare(0, 19, "imageAtomicExchange") != 0)) // imageAtomicExchange doesn't require GL_EXT_shader_atomic_float
|
||||
} else if (imageType.getSampler().type == EbtFloat) {
|
||||
if (fnCandidate.getName().compare(0, 19, "imageAtomicExchange") == 0) {
|
||||
// imageAtomicExchange doesn't require an extension
|
||||
} else if ((fnCandidate.getName().compare(0, 14, "imageAtomicAdd") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 15, "imageAtomicLoad") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 16, "imageAtomicStore") == 0)) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float, fnCandidate.getName().c_str());
|
||||
if (!isImageAtomicOnFloatAllowed)
|
||||
} else if ((fnCandidate.getName().compare(0, 14, "imageAtomicMin") == 0) ||
|
||||
(fnCandidate.getName().compare(0, 14, "imageAtomicMax") == 0)) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float2, fnCandidate.getName().c_str());
|
||||
} else {
|
||||
error(loc, "only supported on integer images", fnCandidate.getName().c_str(), "");
|
||||
else if (imageType.getQualifier().getFormat() != ElfR32f && isEsProfile())
|
||||
}
|
||||
if (imageType.getQualifier().getFormat() != ElfR32f && isEsProfile())
|
||||
error(loc, "only supported on image with format r32f", fnCandidate.getName().c_str(), "");
|
||||
} else {
|
||||
error(loc, "not supported on this image type", fnCandidate.getName().c_str(), "");
|
||||
}
|
||||
|
||||
const size_t maxArgs = imageType.getSampler().isMultiSample() ? 5 : 4;
|
||||
|
|
@ -2381,16 +2386,28 @@ void TParseContext::builtInOpCheck(const TSourceLoc& loc, const TFunction& fnCan
|
|||
memorySemanticsCheck(loc, fnCandidate, callNode);
|
||||
if ((callNode.getOp() == EOpAtomicAdd || callNode.getOp() == EOpAtomicExchange ||
|
||||
callNode.getOp() == EOpAtomicLoad || callNode.getOp() == EOpAtomicStore) &&
|
||||
(arg0->getType().isFloatingDomain())) {
|
||||
(arg0->getType().getBasicType() == EbtFloat ||
|
||||
arg0->getType().getBasicType() == EbtDouble)) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float, fnCandidate.getName().c_str());
|
||||
} else if ((callNode.getOp() == EOpAtomicAdd || callNode.getOp() == EOpAtomicExchange ||
|
||||
callNode.getOp() == EOpAtomicLoad || callNode.getOp() == EOpAtomicStore ||
|
||||
callNode.getOp() == EOpAtomicMin || callNode.getOp() == EOpAtomicMax) &&
|
||||
arg0->getType().isFloatingDomain()) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float2, fnCandidate.getName().c_str());
|
||||
}
|
||||
} else if (arg0->getType().getBasicType() == EbtInt64 || arg0->getType().getBasicType() == EbtUint64) {
|
||||
const char* const extensions[2] = { E_GL_NV_shader_atomic_int64,
|
||||
E_GL_EXT_shader_atomic_int64 };
|
||||
requireExtensions(loc, 2, extensions, fnCandidate.getName().c_str());
|
||||
} else if ((callNode.getOp() == EOpAtomicAdd || callNode.getOp() == EOpAtomicExchange) &&
|
||||
(arg0->getType().isFloatingDomain())) {
|
||||
(arg0->getType().getBasicType() == EbtFloat ||
|
||||
arg0->getType().getBasicType() == EbtDouble)) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float, fnCandidate.getName().c_str());
|
||||
} else if ((callNode.getOp() == EOpAtomicAdd || callNode.getOp() == EOpAtomicExchange ||
|
||||
callNode.getOp() == EOpAtomicLoad || callNode.getOp() == EOpAtomicStore ||
|
||||
callNode.getOp() == EOpAtomicMin || callNode.getOp() == EOpAtomicMax) &&
|
||||
arg0->getType().isFloatingDomain()) {
|
||||
requireExtensions(loc, 1, &E_GL_EXT_shader_atomic_float2, fnCandidate.getName().c_str());
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
|
|
|||
|
|
@ -355,6 +355,7 @@ void TParseVersions::initializeExtensionBehavior()
|
|||
extensionBehavior[E_GL_EXT_shader_subgroup_extended_types_int64] = EBhDisable;
|
||||
extensionBehavior[E_GL_EXT_shader_subgroup_extended_types_float16] = EBhDisable;
|
||||
extensionBehavior[E_GL_EXT_shader_atomic_float] = EBhDisable;
|
||||
extensionBehavior[E_GL_EXT_shader_atomic_float2] = EBhDisable;
|
||||
}
|
||||
|
||||
#endif // GLSLANG_WEB
|
||||
|
|
@ -539,6 +540,7 @@ void TParseVersions::getPreamble(std::string& preamble)
|
|||
"#define GL_EXT_shader_subgroup_extended_types_float16 1\n"
|
||||
|
||||
"#define GL_EXT_shader_atomic_float 1\n"
|
||||
"#define GL_EXT_shader_atomic_float2 1\n"
|
||||
;
|
||||
|
||||
if (version >= 150) {
|
||||
|
|
|
|||
|
|
@ -307,6 +307,7 @@ const char* const E_GL_EXT_shader_subgroup_extended_types_float16 = "GL_EXT_shad
|
|||
const char* const E_GL_EXT_terminate_invocation = "GL_EXT_terminate_invocation";
|
||||
|
||||
const char* const E_GL_EXT_shader_atomic_float = "GL_EXT_shader_atomic_float";
|
||||
const char* const E_GL_EXT_shader_atomic_float2 = "GL_EXT_shader_atomic_float2";
|
||||
|
||||
// Arrays of extensions for the above AEP duplications
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue