diff options
| -rw-r--r-- | source/slang/core.meta.slang | 3 | ||||
| -rw-r--r-- | source/slang/diff.meta.slang | 73 | ||||
| -rw-r--r-- | source/slang/hlsl.meta.slang | 2 | ||||
| -rw-r--r-- | source/slang/hlsl.meta.slang.temp.h | 23204 | ||||
| -rw-r--r-- | source/slang/slang-check-expr.cpp | 20 | ||||
| -rw-r--r-- | source/slang/slang-check-overload.cpp | 27 | ||||
| -rw-r--r-- | tests/bugs/gh-4971-2.slang | 23 | ||||
| -rw-r--r-- | tests/bugs/gh-4971.slang | 25 |
8 files changed, 23346 insertions, 31 deletions
diff --git a/source/slang/core.meta.slang b/source/slang/core.meta.slang index e4e24dddb..62c9a2800 100644 --- a/source/slang/core.meta.slang +++ b/source/slang/core.meta.slang @@ -747,6 +747,7 @@ struct Ptr { // If a 'Ptr[index]' is referred to by a '__ref', call 'kIROp_GetOffsetPtr(index)' __intrinsic_op($(kIROp_GetOffsetPtr)) + [nonmutating] ref; } }; @@ -2996,7 +2997,7 @@ attribute_syntax [__vulkanHitAttributes] : VulkanHitAttributesAttribute; __attributeTarget(FunctionDeclBase) attribute_syntax [mutating] : MutatingAttribute; -__attributeTarget(SetterDecl) +__attributeTarget(AccessorDecl) attribute_syntax [nonmutating] : NonmutatingAttribute; __attributeTarget(FunctionDeclBase) diff --git a/source/slang/diff.meta.slang b/source/slang/diff.meta.slang index b9cc0b103..b39d91494 100644 --- a/source/slang/diff.meta.slang +++ b/source/slang/diff.meta.slang @@ -275,9 +275,10 @@ struct TensorView { [ForceInline] [__NoSideEffect] get { return load(index); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(index, newValue); } @@ -295,9 +296,10 @@ struct TensorView { [ForceInline] [__NoSideEffect] get { return load(i1, i2); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i1, i2, newValue); } @@ -315,13 +317,15 @@ struct TensorView { [ForceInline] [__NoSideEffect] get { return load(i.x, i.y); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i.x, i.y, newValue); } [__NoSideEffect] + [nonmutating] ref { __target_switch @@ -334,14 +338,16 @@ struct TensorView __subscript(uint i1, uint i2, uint i3) -> T { [ForceInline] [__NoSideEffect] get { return load(i1, i2, i3); } - - [ForceInline] set + + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i1, i2, i3, newValue); } [__NoSideEffect] + [nonmutating] ref { __target_switch @@ -355,13 +361,15 @@ struct TensorView { [ForceInline] [__NoSideEffect] get { return load(i.x, i.y, i.z); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i.x, i.y, i.z, newValue); } [__NoSideEffect] + [nonmutating] ref { __target_switch @@ -373,11 +381,14 @@ struct TensorView } __subscript(uint i1, uint i2, uint i3, uint i4) -> T { - [ForceInline] [__NoSideEffect] get { return load(i1, i2, i3, i4); } + [ForceInline] + [__NoSideEffect] + get { return load(i1, i2, i3, i4); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i1, i2, i3, i4, newValue); } @@ -395,9 +406,10 @@ struct TensorView { [__NoSideEffect][ForceInline] get { return load(i.x, i.y, i.z, i.w); } - [ForceInline] set + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i.x, i.y, i.z, i.w, newValue); } @@ -414,14 +426,16 @@ struct TensorView __subscript(uint i1, uint i2, uint i3, uint i4, uint i5) -> T { [ForceInline] [__NoSideEffect] get { return load(i1, i2, i3, i4, i5); } - - [ForceInline] set + + [ForceInline] + [nonmutating] + set { - unmodified(this); store(i1, i2, i3, i4, i5, newValue); } [__NoSideEffect] + [nonmutating] ref { __target_switch @@ -894,63 +908,70 @@ struct DiffTensorView __subscript(uint index)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(index); } - [__unsafeForceInlineEarly] [Differentiable] set { store(index, newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(index, newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint2 index)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(index); } - [__unsafeForceInlineEarly] [Differentiable] set { store(index, newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(index, newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint x, uint y)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(uint2(x, y)); } - [__unsafeForceInlineEarly] [Differentiable] set { store(uint2(x, y), newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(uint2(x, y), newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint3 index)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(index); } - [__unsafeForceInlineEarly] [Differentiable] set { store(index, newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(index, newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint x, uint y, uint z)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(uint3(x, y, z)); } - [__unsafeForceInlineEarly] [Differentiable] set { store(uint3(x, y, z), newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(uint3(x, y, z), newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint4 index)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(index); } - [__unsafeForceInlineEarly] [Differentiable] set { store(index, newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(index, newValue); } [__NoSideEffect] + [nonmutating] ref; } __subscript(uint x, uint y, uint z, uint w)->T { [__unsafeForceInlineEarly] [Differentiable] [__NoSideEffect] get { return load(uint4(x, y, z, w)); } - [__unsafeForceInlineEarly] [Differentiable] set { store(uint4(x, y, z, w), newValue); } + [__unsafeForceInlineEarly] [Differentiable] [nonmutating] set { store(uint4(x, y, z, w), newValue); } [__NoSideEffect] + [nonmutating] ref; } diff --git a/source/slang/hlsl.meta.slang b/source/slang/hlsl.meta.slang index fd5c2ebb7..b0553aa19 100644 --- a/source/slang/hlsl.meta.slang +++ b/source/slang/hlsl.meta.slang @@ -5778,6 +5778,7 @@ struct $(item.name) // // This allows call's to stay aware that the input is from a 'Buffer'. [__NoSideEffect] + [nonmutating] __intrinsic_op($(kIROp_RWStructuredBufferGetElementPtr)) ref; } @@ -16113,6 +16114,7 @@ ${{{{ // If a 'Texture[index]' is referred to by a '__ref', call 'kIROp_ImageSubscript(index)'. // This allows call's to stay aware that the input is from a 'Texture'. __intrinsic_op($(kIROp_ImageSubscript)) + [nonmutating] ref; ${{{{ } // access != SLANG_RESOURCE_ACCESS_READ diff --git a/source/slang/hlsl.meta.slang.temp.h b/source/slang/hlsl.meta.slang.temp.h new file mode 100644 index 000000000..8b2548338 --- /dev/null +++ b/source/slang/hlsl.meta.slang.temp.h @@ -0,0 +1,23204 @@ +SLANG_RAW("// Slang HLSL compatibility library\n") +SLANG_RAW("\n") +SLANG_RAW("typedef uint UINT;\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_RequireGLSLExtension +) +SLANG_RAW(")\n") +SLANG_RAW("void __requireGLSLExtension(String extensionName);\n") +SLANG_RAW("\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("interface IBufferDataLayout\n") +SLANG_RAW("{\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_DefaultBufferLayoutType +) +SLANG_RAW(")\n") +SLANG_RAW("struct DefaultDataLayout : IBufferDataLayout\n") +SLANG_RAW("{};\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_Std140BufferLayoutType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(spirv)]\n") +SLANG_RAW("[require(glsl)]\n") +SLANG_RAW("struct Std140DataLayout : IBufferDataLayout\n") +SLANG_RAW("{};\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_Std430BufferLayoutType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(spirv)]\n") +SLANG_RAW("[require(glsl)]\n") +SLANG_RAW("struct Std430DataLayout : IBufferDataLayout\n") +SLANG_RAW("{};\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_ScalarBufferLayoutType +) +SLANG_RAW(")\n") +SLANG_RAW("struct ScalarDataLayout : IBufferDataLayout\n") +SLANG_RAW("{};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, L : IBufferDataLayout = DefaultDataLayout>\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_GLSLShaderStorageBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("__magic_type(GLSLShaderStorageBufferType)\n") +SLANG_RAW("struct GLSLShaderStorageBuffer {}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T,L:IBufferDataLayout>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferGetDimensions +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, appendstructuredbuffer)]\n") +SLANG_RAW("uint2 __structuredBufferGetDimensions(AppendStructuredBuffer<T,L> buffer);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T,L:IBufferDataLayout>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferGetDimensions +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, consumestructuredbuffer)]\n") +SLANG_RAW("uint2 __structuredBufferGetDimensions(ConsumeStructuredBuffer<T,L> buffer);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferGetDimensions +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer)]\n") +SLANG_RAW("uint2 __structuredBufferGetDimensions<T,L:IBufferDataLayout>(StructuredBuffer<T,L> buffer);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferGetDimensions +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer_rw)]\n") +SLANG_RAW("uint2 __structuredBufferGetDimensions<T,L:IBufferDataLayout>(RWStructuredBuffer<T,L> buffer);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferGetDimensions +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer_rw)]\n") +SLANG_RAW("uint2 __structuredBufferGetDimensions<T,L:IBufferDataLayout>(RasterizerOrderedStructuredBuffer<T,L> buffer);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, L:IBufferDataLayout=DefaultDataLayout>\n") +SLANG_RAW("__magic_type(HLSLAppendStructuredBufferType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLAppendStructuredBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, appendstructuredbuffer)]\n") +SLANG_RAW("struct AppendStructuredBuffer\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferAppend +) +SLANG_RAW(")\n") +SLANG_RAW(" void Append(T value);\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void GetDimensions(\n") +SLANG_RAW(" out uint numStructs,\n") +SLANG_RAW(" out uint stride)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let result = __structuredBufferGetDimensions(this);\n") +SLANG_RAW(" numStructs = result.x;\n") +SLANG_RAW(" stride = result.y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__magic_type(HLSLByteAddressBufferType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLByteAddressBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW("struct ByteAddressBuffer\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer)]\n") +SLANG_RAW(" void GetDimensions(out uint dim)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" dim = __structuredBufferGetDimensions(__getEquivalentStructuredBuffer<uint>(this)).x*4;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint Load(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer)]\n") +SLANG_RAW(" uint Load(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint2 Load2(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint2 Load2(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint2 Load2Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, __naturalStrideOf<uint2>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer)]\n") +SLANG_RAW(" uint2 Load2(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint3 Load3(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint3 Load3(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint3 Load3Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, __naturalStrideOf<uint3>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer)]\n") +SLANG_RAW(" uint3 Load3(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint4 Load4(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint4 Load4(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW(" uint4 Load4Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, __naturalStrideOf<uint4>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer)]\n") +SLANG_RAW(" uint4 Load4(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Load<T>(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Load<T>(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T LoadAligned<T>(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, __naturalStrideOf<T>());\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("[builtin]\n") +SLANG_RAW("interface __ITextureShape\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor;\n") +SLANG_RAW(" static const int dimensions;\n") +SLANG_RAW(" static const int planeDimensions;\n") +SLANG_RAW("}\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("[builtin]\n") +SLANG_RAW("interface __ITextureShape1D2D3D : __ITextureShape\n") +SLANG_RAW("{\n") +SLANG_RAW("}\n") +SLANG_RAW("__magic_type(TextureShape1DType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureShape1DType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __Shape1D : __ITextureShape1D2D3D\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor = ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int dimensions = 1;\n") +SLANG_RAW(" static const int planeDimensions = 1;\n") +SLANG_RAW("}\n") +SLANG_RAW("__magic_type(TextureShape2DType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureShape2DType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __Shape2D : __ITextureShape1D2D3D\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor = ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int dimensions = 2;\n") +SLANG_RAW(" static const int planeDimensions = 2;\n") +SLANG_RAW("}\n") +SLANG_RAW("__magic_type(TextureShape3DType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureShape3DType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __Shape3D : __ITextureShape1D2D3D\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor = ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int dimensions = 3;\n") +SLANG_RAW(" static const int planeDimensions = 3;\n") +SLANG_RAW("}\n") +SLANG_RAW("__magic_type(TextureShapeCubeType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureShapeCubeType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __ShapeCube : __ITextureShape\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor = ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int dimensions = 3;\n") +SLANG_RAW(" static const int planeDimensions = 2;\n") +SLANG_RAW("}\n") +SLANG_RAW("__magic_type(TextureShapeBufferType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureShapeBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __ShapeBuffer : __ITextureShape\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int flavor = ") +SLANG_SPLICE(SLANG_TEXTURE_BUFFER +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int dimensions = 1;\n") +SLANG_RAW(" static const int planeDimensions = 1;\n") +SLANG_RAW("}\n") +SLANG_RAW("__intrinsic_op(vectorReshape)\n") +SLANG_RAW("vector<T,N> __vectorReshape<let N : int, T, let M : int>(vector<T,M> vin);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(makeVector)\n") +SLANG_RAW("__generic<T, let N:int>\n") +SLANG_RAW("vector<T,N+1> __makeVector(vector<T,N> vec, T scalar);\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__magic_type(TextureType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __TextureImpl<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("{\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Combined texture sampler specific functions\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture<TSampler, TCoord>(TSampler s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_1d_shadow<TSampler, TCoord>(TSampler s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_3d_array_shadow<TSampler, TCoord>(TSampler s, TCoord value, float compare)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture_offset<TSampler, TCoord, TOffset>( TSampler s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureOffset($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_1d_shadow<TSampler, TCoord, TOffset>(TSampler s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureOffset($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture_level_zero<TSampler, TCoord>(TSampler s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLod($0, $1, 0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_level_zero_1d_shadow<TSampler, TCoord>(TSampler s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLod($0, $1, 0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_level_zero<TSampler, TCoord, TOffset>(TSampler s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLodOffset($0, $1, 0, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_level_zero_1d_shadow<TSampler, TCoord, TOffset>(TSampler s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLodOffset($0, $1, 0, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture<TTexture, TCoord>(TTexture t, SamplerComparisonState s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($p, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_1d_shadow<TTexture, TCoord>(TTexture t, SamplerComparisonState s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($p, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_3d_array_shadow<TTexture, TCoord>(TTexture t, SamplerComparisonState s, TCoord value, float compare)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"texture($p, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture_offset<TTexture, TCoord, TOffset>(TTexture t,SamplerComparisonState s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureOffset($p, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_1d_shadow<TTexture, TCoord, TOffset>(TTexture t,SamplerComparisonState s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureOffset($p, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl, texture_sm_4_1)]\n") +SLANG_RAW("float __glsl_texture_level_zero<TTexture, TCoord>(TTexture t,SamplerComparisonState s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLod($p, $2, 0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_level_zero_1d_shadow<TTexture, TCoord>(TTexture t,SamplerComparisonState s, TCoord value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLod($p, $2, 0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_level_zero<TTexture, TCoord, TOffset>(TTexture t,SamplerComparisonState s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLodOffset($p, $2, 0, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_texture_shadow_lod)\n") +SLANG_RAW("[require(glsl, texture_shadowlod)]\n") +SLANG_RAW("float __glsl_texture_offset_level_zero_1d_shadow<TTexture, TCoord, TOffset>(TTexture t,SamplerComparisonState s, TCoord value, constexpr TOffset offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"textureLodOffset($p, $2, 0, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,isMS,sampleCount,0,isShadow,1,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int access = 0;\n") +SLANG_RAW("\n") +SLANG_RAW(" typealias TextureCoord = vector<float, Shape.dimensions>;\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CombinedTextureSamplerGetTexture +) +SLANG_RAW(")\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, isMS, sampleCount, 0, isShadow, 0, format> __getTexture();\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CombinedTextureSamplerGetSampler +) +SLANG_RAW(")\n") +SLANG_RAW(" SamplerState __getSampler();\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CombinedTextureSamplerGetSampler +) +SLANG_RAW(")\n") +SLANG_RAW(" SamplerComparisonState __getComparisonSampler();\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv, texture_querylod)]\n") +SLANG_RAW(" float CalculateLevelOfDetail(TextureCoord location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().CalculateLevelOfDetail(__getSampler(), location);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().CalculateLevelOfDetail(__getSampler(), location);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureQueryLod($0, $1).x\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float2 = OpImageQueryLod $this $location\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv, texture_querylod)]\n") +SLANG_RAW(" float CalculateLevelOfDetailUnclamped(TextureCoord location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().CalculateLevelOfDetailUnclamped(__getSampler(), location);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().CalculateLevelOfDetailUnclamped(__getSampler(), location);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureQueryLod($0, $1).y\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float2 = OpImageQueryLod $this $location\n") +SLANG_RAW(" }).y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(vector<float, Shape.dimensions+isArray> location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexture($0, $1)$z\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLayered<$T0>($0, ($1).x, int(($1).y))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLayered<$T0>($0, ($1).x, ($1).y, int(($1).z))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLayered<$T0>($0, ($1).x, ($1).y, ($1).z, int(($1).w))\";\n") +SLANG_RAW(" default: __intrinsic_asm \"invalid texture shape\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1D<$T0>($0, ($1))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2D<$T0>($0, ($1).x, ($1).y)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex3D<$T0>($0, ($1).x, ($1).y, ($1).z)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemap<$T0>($0, ($1).x, ($1).y, ($1).z)\";\n") +SLANG_RAW(" default: __intrinsic_asm \"invalid texture shape\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod $this $location None;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(vector<float, Shape.dimensions+isArray> location, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location, offset);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location, offset);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffsetClampARB($0, $1, $2)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod $this $location None|ConstOffset $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(vector<float, Shape.dimensions+isArray> location, vector<int, Shape.planeDimensions> offset, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location, offset, clamp);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().Sample(__getSampler(), location, offset, clamp);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffsetClampARB($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod $this $location None|ConstOffset|MinLod $offset $clamp;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Sample(vector<float, Shape.dimensions+isArray> location, vector<int, Shape.planeDimensions> offset, float clamp, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Sample(location, offset, clamp);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T SampleBias(vector<float, Shape.dimensions+isArray> location, float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexture($0, $1, $2)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod $this $location None|Bias $bias;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW("\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T SampleBias(vector<float, Shape.dimensions+isArray> location, float bias, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias, offset);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias, offset);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffset($0, $1, $3, $2)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod $this $location None|Bias|ConstOffset $bias $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleBias(__getSampler(), location, bias, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmp(vector<float, Shape.dimensions+isArray> location, float compareValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_1d_shadow(this, __makeVector(__makeVector(location, 0.0), compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (Shape.dimensions == 3 && isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_3d_array_shadow(this, location, compareValue);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture(this, __makeVector(location, compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefImplicitLod $this $location $compareValue;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmpLevelZero(vector<float, Shape.dimensions+isArray> location, float compareValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_level_zero_1d_shadow(this, __makeVector(__makeVector(location, 0.0), compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_level_zero(this, __makeVector(location, compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const float zeroFloat = 0.0f;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefExplicitLod $this $location $compareValue Lod $zeroFloat;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone] \n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmp(vector<float, Shape.dimensions+isArray> location, float compareValue, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_1d_shadow(this, __makeVector(__makeVector(location, 0.0), compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset(this, __makeVector(location, compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefImplicitLod $this $location $compareValue ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleCmp(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmpLevelZero(vector<float, Shape.dimensions+isArray> location, float compareValue, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_level_zero_1d_shadow(this, __makeVector(__makeVector(location,0.0), compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_level_zero(this, __makeVector(location, compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const float zeroFloat = 0.0f;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefExplicitLod $this $location $compareValue Lod|ConstOffset $zeroFloat $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleCmpLevelZero(__getComparisonSampler(), location, compareValue, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGrad($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod $this $location None|Grad $gradX $gradY;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY, constexpr vector<int, Shape.dimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY, offset);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY, offset);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGradOffset($0, $1, $2, $3, $4)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod $this $location None|Grad|ConstOffset $gradX $gradY $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY, constexpr vector<int, Shape.dimensions> offset, float lodClamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY, offset, lodClamp);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleGrad(__getSampler(), location, gradX, gradY, offset, lodClamp);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGradOffsetClampARB($0, $1, $2, $3, $4, $5)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod $this $location None|Grad|ConstOffset|MinLod $gradX $gradY $offset $lodClamp;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleLevel(vector<float, Shape.dimensions+isArray> location, float level)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureLod($0, $1, $2)$z\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLayeredLod<$T0>($0, ($1).x, int(($1).y), ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLayeredLod<$T0>($0, ($1).x, ($1).y, int(($1).z), ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLayeredLod<$T0>($0, ($1).x, ($1).y, ($1).z, int(($1).w), ($2))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLod<$T0>($0, ($1), ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLod<$T0>($0, ($1).x, ($1).y, ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex3DLod<$T0>($0, ($1).x, ($1).y, ($1).z, ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLod<$T0>($0, ($1).x, ($1).y, ($1).z, ($2))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod $this $location None|Lod $level;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleLevel(vector<float, Shape.dimensions+isArray> location, float level, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level, offset);\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level, offset);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureLodOffset($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod $this $location None|Lod|ConstOffset $level $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __getTexture().SampleLevel(__getSampler(), location, level, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Non-combined texture types specific functions\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,isMS,sampleCount,access,isShadow,0,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" typealias TextureCoord = vector<float, Shape.dimensions>;\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv, texture_querylod)]\n") +SLANG_RAW(" float CalculateLevelOfDetail(SamplerState s, TextureCoord location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".CalculateLevelOfDetail\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \".calculate_clamped_lod\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureQueryLod($p, $2).x\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float2 = OpImageQueryLod %sampledImage $location;\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv, texture_querylod)]\n") +SLANG_RAW(" float CalculateLevelOfDetailUnclamped(SamplerState s, TextureCoord location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".CalculateLevelOfDetailUnclamped\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \".calculate_unclamped_lod\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureQueryLod($p, $2).y\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float2 = OpImageQueryLod %sampledImage $location;\n") +SLANG_RAW(" }).y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,isMS,sampleCount,0,isShadow,0,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(SamplerState s, vector<float, Shape.dimensions+isArray> location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).x, uint(($2).y))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xyz, uint(($2).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexture($p, $2)$z\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLayered<$T0>($0, ($2).x, int(($2).y))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLayered<$T0>($0, ($2).x, ($2).y, int(($2).z))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLayered<$T0>($0, ($2).x, ($2).y, ($2).z, int(($2).w))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1D<$T0>($0, ($2))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2D<$T0>($0, ($2).x, ($2).y)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex3D<$T0>($0, ($2).x, ($2).y, ($2).z)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemap<$T0>($0, ($2).x, ($2).y, ($2).z)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod %sampledImage $location None;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).x, i32(($2).y))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).xy, i32(($2).z))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).xyz, i32(($2).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(SamplerState s, vector<float, Shape.dimensions+isArray> location, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffset($p, $2, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod %sampledImage $location None|ConstOffset $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).x, i32(($2).y), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).xy, i32(($2).z), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, ($2).xyz, i32(($2).w), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSample($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T Sample(SamplerState s, vector<float, Shape.dimensions+isArray> location, constexpr vector<int, Shape.planeDimensions> offset, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), min_lod_clamp($4), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, min_lod_clamp($4), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffsetClampARB($p, $2, $3, $4)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod %sampledImage $location None|ConstOffset|MinLod $offset $clamp;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Sample(SamplerState s, vector<float, Shape.dimensions+isArray> location, constexpr vector<int, Shape.planeDimensions> offset, float clamp, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".Sample\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Sample(s, location, offset, clamp);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T SampleBias(SamplerState s, vector<float, Shape.dimensions+isArray> location, float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleBias\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleBias\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), bias($3))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xyz, uint(($2).w), bias($3))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, bias($3))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexture($p, $2, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod %sampledImage $location None|Bias $bias;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).x, i32(($2).y), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).xy, i32(($2).z), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).xyz, i32(($2).w), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0_fragment)]\n") +SLANG_RAW(" T SampleBias(SamplerState s, vector<float, Shape.dimensions+isArray> location, float bias, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleBias\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleBias\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), bias($3), $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, bias($3), $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureOffset($p, $2, $4, $3)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleImplicitLod %sampledImage $location None|Bias|ConstOffset $bias $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).x, i32(($2).y), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).xy, i32(($2).z), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, ($2).xyz, i32(($2).w), $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleBias($0, $1, $2, $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmp(SamplerComparisonState s, vector<float, Shape.dimensions+isArray> location, float compareValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_1d_shadow(this, s, __makeVector(__makeVector(location, 0.0), compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (Shape.dimensions == 3 && isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_3d_array_shadow(this, s, location, compareValue);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture(this, s, __makeVector(location,compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleCmp\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xy, uint(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xyz, uint(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \".sample_compare\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefImplicitLod %sampledImage $location $compareValue;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).x, i32(($2).y), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).xy, i32(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).xyz, i32(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmpLevelZero(SamplerComparisonState s, vector<float, Shape.dimensions+isArray> location, float compareValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_level_zero_1d_shadow(this, s, __makeVector(__makeVector(location, 0.0), compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_level_zero(this, s, __makeVector(location,compareValue));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleCmpLevelZero\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xy, uint(($2).z), $3, level(0))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xyz, uint(($2).w), $3, level(0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, $2, $3, level(0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const float zeroFloat = 0.0f;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefExplicitLod %sampledImage $location $compareValue Lod $zeroFloat;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).x, i32(($2).y), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).xy, i32(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).xyz, i32(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmp(SamplerComparisonState s, vector<float, Shape.dimensions+isArray> location, float compareValue, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_1d_shadow(this, s, __makeVector(__makeVector(location, 0.0), compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset(this, s, __makeVector(location,compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleCmp\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xy, uint(($2).z), $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \".sample_compare\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefImplicitLod %sampledImage $location $compareValue ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).x, i32(($2).y), $3, $4)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).xy, i32(($2).z), $3, $4)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, ($2).xyz, i32(($2).w), $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompare($0, $1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_shadowlod)]\n") +SLANG_RAW(" float SampleCmpLevelZero(SamplerComparisonState s, vector<float, Shape.dimensions+isArray> location, float compareValue, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (Shape.dimensions == 1 && isArray == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_level_zero_1d_shadow(this, s, __makeVector(__makeVector(location,0.0),compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __glsl_texture_offset_level_zero(this, s, __makeVector(location,compareValue), offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleCmpLevelZero\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // T sample_compare(sampler s, float2 coord, uint array, float compare_value, lod_options options, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, ($2).xy, uint(($2).z), $3, level(0), $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // T sample_compare(sampler s, float2 coord, float compare_value, lod_options options, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample_compare($1, $2, $3, level(0), $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const float zeroFloat = 0.0f;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" result:$$float = OpImageSampleDrefExplicitLod %sampledImage $location $compareValue Lod|ConstOffset $zeroFloat $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).x, i32(($2).y), $3, $4)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).xy, i32(($2).z), $3, $4)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, ($2).xyz, i32(($2).w), $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleCompareLevel($0, $1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(SamplerState s, vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), gradient2d($3, $4))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xyz, uint(($2).w), gradientcube($3, $4))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient2d($3, $4))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient3d($3, $4))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradientcube($3, $4))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGrad($p, $2, $3, $4)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod %sampledImage $location None|Grad $gradX $gradY;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).x, i32(($2).y), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).xy, i32(($2).z), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).xyz, i32(($2).w), $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, $2, $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(SamplerState s, vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY, constexpr vector<int, Shape.dimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), gradient2d($3, $4), $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient2d($3, $4), $5)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient3d($3, $4), $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGradOffset($p, $2, $3, $4, $5)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod %sampledImage $location None|Grad|ConstOffset $gradX $gradY $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW("\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).x, i32(($2).y), $3, $4, $5)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).xy, i32(($2).z), $3, $4, $5)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, ($2).xyz, i32(($2).w), $3, $4, $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleGrad($0, $1, $2, $3, $4, $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleGrad(SamplerState s, vector<float, Shape.dimensions+isArray> location, vector<float, Shape.dimensions> gradX, vector<float, Shape.dimensions> gradY, constexpr vector<int, Shape.dimensions> offset, float lodClamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleGrad\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), gradient2d($3, $4), min_lod_clamp($6), $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient2d($3, $4), min_lod_clamp($6), $5)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, gradient3d($3, $4), min_lod_clamp($6), $5)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureGradOffsetClampARB($p, $2, $3, $4, $5, $6)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod %sampledImage $location None|Grad|ConstOffset|MinLod $gradX $gradY $offset $lodClamp;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleLevel(SamplerState s, vector<float, Shape.dimensions+isArray> location, float level)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleLevel\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleLevel\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).x, uint(($2).y))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), level($3))$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xyz, uint(($2).w), level($3))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, level($3))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureLod($p, $2, $3)$z\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLayeredLod<$T0>($0, ($2).x, int(($2).y), ($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLayeredLod<$T0>($0, ($2).x, ($2).y, int(($2).z), ($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLayeredLod<$T0>($0, ($2).x, ($2).y, ($2).z, int(($2).w), ($3))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex1DLod<$T0>($0, ($2), ($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex2DLod<$T0>($0, ($2).x, ($2).y, ($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"tex3DLod<$T0>($0, ($2).x, ($2).y, ($2).z, ($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"texCubemapLod<$T0>($0, ($2).x, ($2).y, ($2).z, ($3))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod %sampledImage $location None|Lod $level;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).x, i32(($2).y), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).xy, i32(($2).z), $3)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).xyz, i32(($2).w), $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, $2, $3)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_0)]\n") +SLANG_RAW(" T SampleLevel(SamplerState s, vector<float, Shape.dimensions+isArray> location, float level, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".SampleLevel\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" || T is half || T is vector<half,2> || T is vector<half,3> || T is vector<half,4>\n") +SLANG_RAW(" , \"HLSL supports only float and half type textures\");\n") +SLANG_RAW(" __intrinsic_asm \".SampleLevel\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xy, uint(($2).z), level($3), $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, ($2).xyz, uint(($2).w), level($3), $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.sample($1, $2, level($3), $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctextureLodOffset($p, $2, $3, $4)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(this) = OpSampledImage $this $s;\n") +SLANG_RAW(" %sampled : __sampledType(T) = OpImageSampleExplicitLod %sampledImage $location None|Lod|ConstOffset $level $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(T is float || T is vector<float,2> || T is vector<float,3> || T is vector<float,4>\n") +SLANG_RAW(" , \"WGSL supports only f32 type textures\");\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).x, i32(($2).y), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).xy, i32(($2).z), $3, $4)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, ($2).xyz, i32(($2).w), $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureSampleLevel($0, $1, $2, $3, $4)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture.GetDimensions and Sampler.GetDimensions\n") + +const char* kTextureShapeTypeNames[] = { + "__Shape1D", "__Shape2D", "__Shape3D", "__ShapeCube"}; +for (int shapeIndex = 0; shapeIndex < 4; shapeIndex++) +for (int isArray = 0; isArray <= 1; isArray++) +for (int isMS = 0; isMS <= 1; isMS++) { + if (isMS) + { + if (shapeIndex != kStdlibShapeIndex2D) + continue; + } + if (isArray) + { + if (shapeIndex == kStdlibShapeIndex3D) + continue; + } + auto shapeTypeName = kTextureShapeTypeNames[shapeIndex]; + TextureTypeInfo textureTypeInfo(kBaseTextureShapes[shapeIndex], isArray, isMS, 0, sb, path); +SLANG_RAW("#line 2166 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,") +SLANG_SPLICE(shapeTypeName +) +SLANG_RAW(",") +SLANG_SPLICE(isArray +) +SLANG_RAW(",") +SLANG_SPLICE(isMS +) +SLANG_RAW(",sampleCount,access,isShadow,isCombined,format>\n") +SLANG_RAW("{\n") + + textureTypeInfo.writeGetDimensionFunctions(); + +SLANG_RAW("#line 2173 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} +SLANG_RAW("#line 2178 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture.GetSamplePosition(int s);\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,1,sampleCount,access,isShadow,isCombined,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, texture_sm_4_1_vertex_fragment_geometry)]\n") +SLANG_RAW(" float2 GetSamplePosition(int s);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_MakeArray +) +SLANG_RAW(")\n") +SLANG_RAW("Array<T,4> __makeArray<T>(T v0, T v1, T v2, T v3);\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Beginning of Texture Gather\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerState s,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($p, $2, $3)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" // Tv gather(sampler s, float2 coord, uint array, int2 offset = int2(0), component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, ($2).xy, uint(($2).z), int2(0), metal::component($3))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" // Tv gather(sampler s, float3 coord, uint array, component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, ($2).xyz, uint(($2).w), metal::component($3))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" if (Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(")\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Tv gather(sampler s, float3 coord, component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, $2, metal::component($3))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // Tv gather(sampler s, float2 coord, int2 offset = int2(0), component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, $2, int2(0), metal::component($3))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather %sampledImage $location $component;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // If depth texture, `textureGather` doesn't take channel value, `$3`.\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, ($2).xy, u32(($2).z))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, ($2).xyz, u32(($2).w))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($3, $0, $1, ($2).xy, u32(($2).z))\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($3, $0, $1, ($2).xyz, u32(($2).w))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($3, $0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather $sampler $location $component;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather_offset(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerState s,\n") +SLANG_RAW(" constexpr vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffset($p, $2, $3, $4)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Metal supports offset variant of Gather only for 2D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Tv gather(sampler s, float2 coord, uint array, int2 offset = int2(0), component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, ($2).xy, uint(($2).z), $3, metal::component($4))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // Tv gather(sampler s, float2 coord, int2 offset = int2(0), component c = component::x) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather($1, $2, $3, metal::component($4))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather %sampledImage $location $component ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // If depth texture, `textureGather` doesn't take channel value, `$4`.\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, ($2).xy, u32(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, ($2).xyz, u32(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($4, $0, $1, ($2).xy, u32(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($4, $0, $1, ($2).xyz, u32(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($4, $0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather_offset(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffset($0, $1, $2, $3)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather $sampler $location $component ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather_offsets(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerState s,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset4,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffsets($p, $2, $T3[]($3, $4, $5, $6)), $7\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let offsets = __makeArray(offset1,offset2,offset3,offset4);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageGatherExtended;\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather %sampledImage $location $component ConstOffsets $offsets;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gather_offsets(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset4,\n") +SLANG_RAW(" int component)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffsets($0, $1, $T2[]($2, $3, $4, $5), $6)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let offsets = __makeArray(offset1,offset2,offset3,offset4);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageGatherExtended;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageGather $sampler $location $component ConstOffsets $offsets;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerComparisonState s,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($p, $2, $3)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" // Tv gather_compare(sampler s, float2 coord, uint array, float compare_value, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather_compare($1, ($2).xy, uint(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" // Tv gather_compare(sampler s, float3 coord, uint array, float compare_value) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather_compare($1, ($2).xyz, uint(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // Tv gather_compare(sampler s, float2 coord, float compare_value, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather_compare($1, $2, $3)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather %sampledImage $location $compareValue;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(isShadow == 1, \"WGSL supports textureGatherCompare only for depth textures.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, ($2).xy, u32(($2).z), $3)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, ($2).xyz, u32(($2).w), $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGather($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather $sampler $location $compareValue;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp_offset(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerComparisonState s,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffset($p, $2, $3, $4)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Metal supports depth compare Gather only for 2D texture\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Tv gather_compare(sampler s, float2 coord, uint array, float compare_value, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather_compare($1, ($2).xy, uint(($2).z), $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // Tv gather_compare(sampler s, float2 coord, float compare_value, int2 offset = int2(0)) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.gather_compare($1, $2, $3, $4)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather %sampledImage $location $compareValue ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(isShadow == 1, \"WGSL supports textureGatherCompare only for depth textures.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, ($2).xy, u32(($2).z), $3, $4)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, ($2).xyz, u32(($2).w), $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherCompare($0, $1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp_offset(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffset($0, $1, $2, $3)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather $sampler $location $compareValue ConstOffset $offset;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp_offsets(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 0, format> texture,\n") +SLANG_RAW(" SamplerComparisonState s,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset4)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffsets($p, $2, $3, $T4[]($4, $5, $6, $7))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let offsets = __makeArray(offset1,offset2,offset3,offset4);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageGatherExtended;\n") +SLANG_RAW(" %sampledImage : __sampledImageType(texture) = OpSampledImage $texture $s;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather %sampledImage $location $compareValue ConstOffsets $offsets;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<TElement, T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let access:int, let isShadow:int, let format:int>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, texture_gather)]\n") +SLANG_RAW("vector<TElement,4> __texture_gatherCmp_offsets(\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0, sampleCount, access, isShadow, 1, format> sampler,\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location,\n") +SLANG_RAW(" TElement compareValue,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" vector<int, Shape.planeDimensions> offset4)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureGatherOffsets($0, $1, $2, $T3[]($3, $4, $5, $6))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let offsets = __makeArray(offset1,offset2,offset3,offset4);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageGatherExtended;\n") +SLANG_RAW(" result:$$vector<TElement,4> = OpImageDrefGather $sampler $location $compareValue ConstOffsets $offsets;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +for (int isCombined = 0; isCombined < 2; isCombined++) +for (int isScalarTexture = 0; isScalarTexture < 2; isScalarTexture++) +{ + const char* extSizeParam = isScalarTexture ? "" : ", let N:int"; + const char* extTexType = isScalarTexture ? "T" : "vector<T,N>"; + +SLANG_RAW("#line 2609 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("// Gather for [TextureType = ") +SLANG_SPLICE(extTexType +) +SLANG_RAW(", isCombined = ") +SLANG_SPLICE(isCombined +) +SLANG_RAW("]\n") +SLANG_RAW("__generic<T:__BuiltinArithmeticType ") +SLANG_SPLICE(extSizeParam +) +SLANG_RAW(", Shape: __ITextureShape, let isArray:int, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<") +SLANG_SPLICE(extTexType +) +SLANG_RAW(",Shape,isArray,0,sampleCount,0,isShadow,") +SLANG_SPLICE(isCombined +) +SLANG_RAW(",format>\n") +SLANG_RAW("{\n") + + for (int isShadow = 0; isShadow < 2; isShadow++) + for (auto componentId = 0; componentId < 5; componentId++) + { + const char* compareFunc = isShadow ? "Cmp" : ""; + const char* compareParam = isShadow ? ", T compareValue" : ""; + const char* compareArg = isShadow ? ", compareValue" : ""; + + // Some targets support the combined texture natively + const char* samplerParam = isCombined ? "" : (isShadow ? "SamplerComparisonState s," : "SamplerState s,"); + const char* samplerArg = isCombined ? "" : ", s"; + const char* getTexture = isCombined ? "__getTexture()" : "this"; + const char* getSampler = isCombined ? (isShadow ? ", __getComparisonSampler()" : ", __getSampler()") : samplerArg; + + const char* componentFuncString[] = { "", "Red", "Green", "Blue", "Alpha"}; + const char* componentArgString[] = { ", 0", ", 0", ", 1", ", 2", ", 3" }; + const char* componentFunc = componentFuncString[componentId]; + const char* componentArg = (isShadow ? "" : componentArgString[componentId]); +SLANG_RAW("#line 2632 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(")\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __texture_gather") +SLANG_SPLICE(compareFunc +) +SLANG_RAW("<T>(") +SLANG_SPLICE(getTexture +) +SLANG_RAW(" ") +SLANG_SPLICE(getSampler +) +SLANG_RAW(", location ") +SLANG_SPLICE(compareArg +) +SLANG_RAW(" ") +SLANG_SPLICE(componentArg +) +SLANG_RAW(");\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return __texture_gather") +SLANG_SPLICE(compareFunc +) +SLANG_RAW("<T>(this ") +SLANG_SPLICE(samplerArg +) +SLANG_RAW(", location ") +SLANG_SPLICE(compareArg +) +SLANG_RAW(" ") +SLANG_SPLICE(componentArg +) +SLANG_RAW(");\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(",\n") +SLANG_RAW(" out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_metal_spirv_wgsl, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(",\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return __texture_gather") +SLANG_SPLICE(compareFunc +) +SLANG_RAW("_offset<T>(") +SLANG_SPLICE(getTexture +) +SLANG_RAW(" ") +SLANG_SPLICE(getSampler +) +SLANG_RAW(", location ") +SLANG_SPLICE(compareArg +) +SLANG_RAW(", offset ") +SLANG_SPLICE(componentArg +) +SLANG_RAW(");\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return __texture_gather") +SLANG_SPLICE(compareFunc +) +SLANG_RAW("_offset<T>(this ") +SLANG_SPLICE(samplerArg +) +SLANG_RAW(", location ") +SLANG_SPLICE(compareArg +) +SLANG_RAW(", offset ") +SLANG_SPLICE(componentArg +) +SLANG_RAW(");\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(",\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset,\n") +SLANG_RAW(" out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(",\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset4)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return __texture_gather") +SLANG_SPLICE(compareFunc +) +SLANG_RAW("_offsets<T>(this ") +SLANG_SPLICE(samplerArg +) +SLANG_RAW(", location ") +SLANG_SPLICE(compareArg +) +SLANG_RAW(", offset1,offset2,offset3,offset4 ") +SLANG_SPLICE(componentArg +) +SLANG_RAW(");\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, texture_gather)]\n") +SLANG_RAW(" vector<T,4> Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("(\n") +SLANG_RAW(" ") +SLANG_SPLICE(samplerParam +) +SLANG_RAW("\n") +SLANG_RAW(" vector<float, Shape.dimensions+isArray> location\n") +SLANG_RAW(" ") +SLANG_SPLICE(compareParam +) +SLANG_RAW(",\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset1,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset2,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset3,\n") +SLANG_RAW(" constexpr vector<int, Shape.planeDimensions> offset4,\n") +SLANG_RAW(" out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(",\n") +SLANG_RAW(" \"Gather is supported only for 2D and 3D textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Gather") +SLANG_SPLICE(compareFunc +) +SLANG_SPLICE(componentFunc +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + } // for (componentId) +SLANG_RAW("#line 2759 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("} // End of: Gather for [TextureType = ") +SLANG_SPLICE(extTexType +) +SLANG_RAW(", isCombined = ") +SLANG_SPLICE(isCombined +) +SLANG_RAW("]\n") +SLANG_RAW("\n") + +} // for (isScalarTexture) +SLANG_RAW("#line 2764 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("// End of all Texture Gather\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Load/Subscript for readonly, no MS textures\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,0,sampleCount,0,isShadow,isCombined,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int isMS = 0;\n") +SLANG_RAW(" static const int access = ") +SLANG_SPLICE(kStdlibResourceAccessReadOnly +) +SLANG_RAW(";\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [require(glsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T __glsl_load(vector<int, Shape.dimensions+isArray> location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __intrinsic_asm \"$ctexelFetch($0, ($1), 0)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray+1> location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" // lod is not supported for 1D texture\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(uint(($1).x), uint(($1).y))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(uint(($1).x))$z\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // T read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 0 && isArray == 0)\n") +SLANG_RAW(" // Tv read(uint3 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,3>(($1).xyz), uint(($1).w))$z\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" static_assert(isArray == 0, \"Unsupported 'Load' of 'texture cube array' for 'metal' target\");\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // T read(uint2 coord, uint face, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsics>\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint2 coord, uint face, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsics>\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" static_assert(false, \"Unsupported 'Load' of 'texture' for 'metal' target\");\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsics>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexelFetch($0, ($1).$w1b, ($1).$w1e)$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const int lodLoc = Shape.dimensions+isArray;\n") +SLANG_RAW(" let coord = __vectorReshape<Shape.dimensions+isArray>(location);\n") +SLANG_RAW(" let lod = location[lodLoc];\n") +SLANG_RAW(" if (isCombined != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %image:__imageType(this) = OpImage $this;\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch %image $coord Lod $lod;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch $this $coord Lod $lod;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL doesn't supports textureLoad for Cube texture.\");\n") +SLANG_RAW(" static_assert(isArray == 0 || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL supports textureLoad for texture_2d_array but not for array of 1D, 3D or Cube.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, ($1).xy, i32(($1).z), ($1).w)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, ($1).x, ($1).y)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, ($1).xy, ($1).z)$z\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, ($1).xyz, ($1).w)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return T();\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray+1> location, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexelFetchOffset($0, ($1).$w1b, ($1).$w1e, ($2))$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const int lodLoc = Shape.dimensions+isArray;\n") +SLANG_RAW(" let coord = __vectorReshape<Shape.dimensions+isArray>(location);\n") +SLANG_RAW(" let lod = location[lodLoc];\n") +SLANG_RAW(" if (isCombined != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %image:__imageType(this) = OpImage $this;\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch %image $coord Lod|ConstOffset $lod $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch $this $coord Lod|ConstOffset $lod $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray+1> location, constexpr vector<int, Shape.planeDimensions> offset, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Load(location, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(vector<uint, Shape.dimensions+isArray> location) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".operator[]\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" return Load(__makeVector(location, 0));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" return __glsl_load(location);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (isCombined != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %image:__imageType(this) = OpImage $this;\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch %image $location;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch $this $location;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return Load(__makeVector(location, 0));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture Load/Subscript for readonly, MS textures\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,1,sampleCount,0,isShadow,isCombined,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" static const int access = ") +SLANG_SPLICE(kStdlibResourceAccessReadOnly +) +SLANG_RAW(";\n") +SLANG_RAW(" static const int isMS = 1;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Document seems to have a typo. `lod` must be `sample`.\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(($1).xy, ($1).z, uint($2))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint sample) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read($1, uint($2))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Document seems to have a typo. `lod` must be `sample`.\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(($1).xy, ($1).z, uint($2))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint sample) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read($1, uint($2))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<Not supported>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexelFetch($0, $1, ($2))$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (isCombined != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %image:__imageType(this) = OpImage $this;\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch %image $location Sample $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch $this $location Sample $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL supports textureLoad for texture_multisampled_2d but not for multisampled of 1D, 3D or Cube.\");\n") +SLANG_RAW(" static_assert(isArray == 0\n") +SLANG_RAW(" , \"WGSL doesn't support array variants of multisampled textures for textureLoad.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, $1, $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions + isArray + 1> locationAndSampleIndex)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return Load(__vectorReshape<Shape.dimensions + isArray>(locationAndSampleIndex), locationAndSampleIndex[Shape.dimensions + isArray]);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex, constexpr vector<int, Shape.planeDimensions> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$ctexelFetchOffset($0, $1, ($2), ($3))$z\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (isCombined != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %image:__imageType(this) = OpImage $this;\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch %image $location ConstOffset|Sample $offset $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageFetch $this $location ConstOffset|Sample $offset $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex, constexpr vector<int, Shape.planeDimensions> offset, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Load(location, sampleIndex, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(vector<uint, Shape.dimensions+isArray> location) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"($0).sample[$1]\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return Load(location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __subscript(vector<uint, Shape.dimensions+isArray> location, int sampleIndex) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_samplerless)]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"($0).sample[$2][$1]\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return Load(location, sampleIndex);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Load/Subscript for readwrite textures\n") + + for (int access = kStdlibResourceAccessReadWrite; access<=kStdlibResourceAccessRasterizerOrdered; access++) { + const char* glslIntrinsic = "$cimageLoad($0, $1)$z"; + const char* glslIntrinsicOffset = "$cimageLoad($0, ($1)+($2))$z"; + const char* glslIntrinsicMS = "$cimageLoad($0, $1, $2)$z"; + const char* glslIntrinsicMSOffset = "$cimageLoad($0, ($1)+($3), $2)$z"; +SLANG_RAW("#line 3185 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,0,sampleCount,") +SLANG_SPLICE(access +) +SLANG_RAW(",isShadow, 0,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"") +SLANG_SPLICE(glslIntrinsic +) +SLANG_RAW("\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf1DLayeredread$C<$T0>($0, ($1).x * $E, ($1).y, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf2DLayeredread$C<$T0>($0, ($1).x * $E, ($1).y, ($1).z, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf3DLayeredread$C<$T0>($0, ($1).x * $E, ($1).y, ($1).z, ($1).w, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf1Dread$C<$T0>($0, ($1) * $E, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf2Dread$C<$T0>($0, ($1).x * $E, ($1).y, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf3Dread$C<$T0>($0, ($1).x * $E, ($1).y, ($1).z, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageRead $this $location;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" // lod is not supported for 1D texture\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(uint(($1).x), uint(($1).y))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(uint($1))$z\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // T read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z))$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 0 && isArray == 0)\n") +SLANG_RAW(" // Tv read(uint3 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,3>(($1).xyz))$z\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" static_assert(isArray == 0, \"Unsupported 'Load' of 'texture cube array' for 'metal' target\");\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // T read(uint2 coord, uint face, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.read(vec<uint,2>(($1).xy), uint(($1).z)%6, uint(($1).z)/6, uint(($1).w))\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // Tv read(uint2 coord, uint face, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.read(vec<uint,2>(($1).xy), uint(($1).z)%6, uint(($1).z)/6, uint(($1).w))\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" static_assert(false, \"Unsupported 'Load' of 'texture' for 'metal' target\");\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsics>\";\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL doesn't supports textureLoad for Cube texture.\");\n") +SLANG_RAW(" static_assert(isArray == 0 || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL supports textureLoad for 2d_array but not for array of 1D, 3D or Cube.\");\n") +SLANG_RAW(" static_assert(isShadow == 0 || T is float\n") +SLANG_RAW(" , \"WGSL supports only f32 depth textures\");\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, ($1).xy, i32(($1).z))$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, $1)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv, texture_sm_4_1)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, vector<int, Shape.dimensions+isArray> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"") +SLANG_SPLICE(glslIntrinsicOffset +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageRead $this $location ConstOffset $offset;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, vector<int, Shape.dimensions+isArray> offset, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Load(location, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(glsl, texture_sm_4_1)]\n") +SLANG_RAW(" void __glslImageStore(vector<int, Shape.dimensions+isArray> location, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __intrinsic_asm \"imageStore($0, $1, $V2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(vector<uint, Shape.dimensions+isArray> location) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1)]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".operator[]\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return Load(location);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [nonmutating]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1)]\n") +SLANG_RAW(" set(T newValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".operator[]\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslImageStore(location, newValue);\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" if (isArray != 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf1DLayeredwrite$C<$T0>($2, $0, ($1).x * $E, ($1).y, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf2DLayeredwrite$C<$T0>($2, $0, ($1).x * $E, ($1).y, ($1).z, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf3DLayeredwrite$C<$T0>($2, $0, ($1).x * $E, ($1).y, ($1).z, ($1).w, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" switch(Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf1Dwrite$C<$T0>($2, $0, ($1) * $E, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf2Dwrite$C<$T0>($2, $0, ($1).x * $E, ($1).y, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" __intrinsic_asm \"surf3Dwrite$C<$T0>($2, $0, ($1).x * $E, ($1).y, ($1).z, SLANG_CUDA_BOUNDARY_MODE)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"<invalid intrinsic>\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpImageWrite $this $location __convertTexel(newValue);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW(":\n") +SLANG_RAW(" // lod is not supported for 1D texture\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // void write(Tv color, uint coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, uint(($1).x), uint(($1).y))\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // void write(Tv color, uint coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, uint($1))\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy), uint(($1).z))\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy), uint(($1).z))\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 0 && isArray == 0)\n") +SLANG_RAW(" // void write(Tv color, uint3 coord, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,3>(($1).xyz))\";\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_CUBE +) +SLANG_RAW(":\n") +SLANG_RAW(" static_assert(isArray == 0, \"Unsupported 'Store' of 'texture cube array' for 'metal' target\");\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // void write(Tv color, uint2 coord, uint face, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$0.write($2, vec<uint,2>(($1).xy), uint(($1).z), uint(($1).w))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_1D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_3D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL doesn't supports textureStore for Cube texture.\");\n") +SLANG_RAW(" static_assert(isArray == 0 || Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL supports textureStore for texture_store_2d_array but not for array of 1D, 3D or Cube.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" // WGSL requires the value type to be always `vec4`\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (T is int32_t || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<i32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is int32_t2 || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<i32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is int32_t3 || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<i32>($2, 1))\";\n") +SLANG_RAW(" if (T is uint32_t || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<u32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is uint32_t2 || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<u32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is uint32_t3 || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<u32>($2, 1))\";\n") +SLANG_RAW(" if (T is half) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f16>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is half2) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f16>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is half3) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f16>($2, 1))\";\n") +SLANG_RAW(" if (T is float) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is float2) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is float3) __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), vec4<f32>($2, 1))\";\n") +SLANG_RAW(" __intrinsic_asm \"textureStore($0, ($1).xy, i32(($1).z), $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" if (T is int32_t || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, $1, vec4<i32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is int32_t2 || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, $1, vec4<i32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is int32_t3 || T is int16_t || T is int8_t) __intrinsic_asm \"textureStore($0, $1, vec4<i32>($2, 1))\";\n") +SLANG_RAW(" if (T is uint32_t || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, $1, vec4<u32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is uint32_t2 || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, $1, vec4<u32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is uint32_t3 || T is uint16_t || T is uint8_t) __intrinsic_asm \"textureStore($0, $1, vec4<u32>($2, 1))\";\n") +SLANG_RAW(" if (T is half) __intrinsic_asm \"textureStore($0, $1, vec4<f16>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is half2) __intrinsic_asm \"textureStore($0, $1, vec4<f16>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is half3) __intrinsic_asm \"textureStore($0, $1, vec4<f16>($2, 1))\";\n") +SLANG_RAW(" if (T is float) __intrinsic_asm \"textureStore($0, $1, vec4<f32>($2, 0, 0, 1))\";\n") +SLANG_RAW(" if (T is float2) __intrinsic_asm \"textureStore($0, $1, vec4<f32>($2, 0, 1))\";\n") +SLANG_RAW(" if (T is float3) __intrinsic_asm \"textureStore($0, $1, vec4<f32>($2, 1))\";\n") +SLANG_RAW(" __intrinsic_asm \"textureStore($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // If a 'Texture[location]' is referred to by a '__ref', call 'kIROp_ImageSubscript(location)'.\n") +SLANG_RAW(" // This allows call's to stay aware that the input is from a 'Texture'.\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_ImageSubscript +) +SLANG_RAW(") \n") +SLANG_RAW(" ref;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +if (access == kStdlibResourceAccessReadWrite) { +SLANG_RAW("#line 3529 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("// RW MS textures.\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let isArray:int, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,isArray,1,sampleCount,") +SLANG_SPLICE(access +) +SLANG_RAW(",isShadow, 0,format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_metal_spirv_wgsl, texture_sm_4_1_compute_fragment)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" switch (Shape.flavor)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW(":\n") +SLANG_RAW(" if (isShadow == 1)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // The document seems to have a typo. `lod` must mean `sample`.\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), $2)$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // T read(uint2 coord, uint sample) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isArray == 1)\n") +SLANG_RAW(" // The document seems to have a typo. `lod` must mean `sample`.\n") +SLANG_RAW(" // Tv read(uint2 coord, uint array, uint lod = 0) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), uint(($1).z), $2)$z\";\n") +SLANG_RAW(" else\n") +SLANG_RAW(" // Tv read(uint2 coord, uint sample) const\n") +SLANG_RAW(" __intrinsic_asm \"$c$0.read(vec<uint,2>(($1).xy), $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" break;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" // TODO: This needs to be handled by the capability system\n") +SLANG_RAW(" __intrinsic_asm \"<Not supported>\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"") +SLANG_SPLICE(glslIntrinsicMS +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageRead $this $location Sample $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" static_assert(Shape.flavor == ") +SLANG_SPLICE(SLANG_TEXTURE_2D +) +SLANG_RAW("\n") +SLANG_RAW(" , \"WGSL supports textureLoad for texture_multisampled_2d but not for multisampled of 1D, 3D or Cube.\");\n") +SLANG_RAW(" static_assert(isArray == 0\n") +SLANG_RAW(" , \"WGSL doesn't support array variants of multisampled textures for textureLoad.\");\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_asm \"textureLoad($0, $1, $2)$z\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv, texture_sm_4_1_compute_fragment)]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex, vector<int, Shape.dimensions+isArray> offset)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"") +SLANG_SPLICE(glslIntrinsicMSOffset +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %sampled:__sampledType(T) = OpImageRead $this $location ConstOffset|Sample $offset $sampleIndex;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" T Load(vector<int, Shape.dimensions+isArray> location, int sampleIndex, vector<int, Shape.dimensions+isArray> offset, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" status = 0;\n") +SLANG_RAW(" return Load(location, sampleIndex, offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(glsl, texture_sm_4_1_compute_fragment)]\n") +SLANG_RAW(" void __glslImageStore(vector<int, Shape.dimensions+isArray> location, int sampleIndex, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __intrinsic_asm \"imageStore($0, $1, $2, $V3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(vector<uint, Shape.dimensions+isArray> location, int sampleIndex) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv_wgsl, texture_sm_4_1_compute_fragment)]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample[$2][$1]\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return Load(location, sampleIndex);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [nonmutating]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_glsl_hlsl_spirv, texture_sm_4_1_compute_fragment)]\n") +SLANG_RAW(" set(T newValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.sample[$2][$1]\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslImageStore(location, sampleIndex, newValue);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpImageWrite $this $location __convertTexel(newValue) Sample $sampleIndex;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // If a 'Texture[location, sampleIndex]' is referred to by a '__ref', call 'kIROp_ImageSubscript(location, sampleIndex)'.\n") +SLANG_RAW(" // This allows call's to stay aware that the input is from a 'Texture'.\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_ImageSubscript +) +SLANG_RAW(")\n") +SLANG_RAW(" ref;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} // if (access == kStdlibResourceAccessReadWrite) // for RW MS textures. +} // for (access). +SLANG_RAW("#line 3681 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Definitions to support the legacy texture .mips[][] operator.\n") +SLANG_RAW("struct __TextureMip<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format> tex;\n") +SLANG_RAW(" int mip;\n") +SLANG_RAW(" __subscript(vector<int, isArray + Shape.dimensions> pos)->T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return tex.Load(__makeVector(pos, mip)); }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("struct __TextureMips<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 0 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format> tex;\n") +SLANG_RAW(" __subscript(int mip)->__TextureMip<T, Shape, isArray, isCombined, format>\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return { tex, mip }; }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("extension __TextureImpl<T, Shape, isArray, 0 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" property __TextureMips<T, Shape, isArray, isCombined, format> mips\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return { this }; }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Definitions to support the .sample[][] operator.\n") +SLANG_RAW("struct __TextureSample<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 1 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format> tex;\n") +SLANG_RAW(" int sample;\n") +SLANG_RAW(" __subscript(vector<int, isArray + Shape.dimensions> pos)->T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return tex[pos, sample]; }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("struct __TextureSampleMS<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __TextureImpl<T, Shape, isArray, 1 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format> tex;\n") +SLANG_RAW(" __subscript(int sample)->__TextureSample<T, Shape, isArray, isCombined, format>\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return { tex, sample }; }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape : __ITextureShape, let isArray : int, let isCombined : int, let format : int>\n") +SLANG_RAW("extension __TextureImpl<T, Shape, isArray, 1 /*isMS*/, 0 /*sampleCount*/, 0 /*access*/, 0 /*isShadow*/, isCombined, format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" property __TextureSampleMS<T, Shape, isArray, isCombined, format> sample\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get { return { this }; }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture type aliases.\n") +SLANG_RAW("// T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int\n") + + const char* shapeTypeNames[] = {"1D", "2D", "3D", "Cube"}; + const char* accessPrefix[] = {"", "RW", "RasterizerOrdered", "Feedback"}; + const char* arrayPostFix[] = {"", "Array"}; + const char* msPostFix[] = {"", "MS"}; + for (int shape = 0; shape < 4; shape++) + for (int isArray = 0; isArray<=1; isArray++) + for (int isMS = 0; isMS<=1; isMS++) + for (int isCombined = 0; isCombined<=1; isCombined++) + for (int access = kStdlibResourceAccessReadOnly; access<=kStdlibResourceAccessFeedback; access++) { + if (access != kStdlibResourceAccessReadOnly) + { + // No RW Cube. + if (shape == kStdlibShapeIndexCube) continue; + } + if (access == kStdlibResourceAccessFeedback) + { + // Feedback only defined for Texture2D and Texture2DArray. + if (shape != 1) continue; + if (isMS) continue; + if (isCombined) continue; + } + if (isMS) + { + // Only Texture2DMS. + if (shape != kStdlibShapeIndex2D) + continue; + // Only Texture2DMS or RWTexture2DMS. + if (access >= kStdlibShapeIndex3D) + continue; + } + // No 3D Array. + if (shape == kStdlibShapeIndex3D && isArray == 1) + continue; + const char* textureTypeName = isCombined ? "Sampler" : "Texture"; +SLANG_RAW("#line 3784 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("typealias ") +SLANG_SPLICE(accessPrefix[access] +) +SLANG_SPLICE(textureTypeName +) +SLANG_SPLICE(shapeTypeNames[shape] +) +SLANG_SPLICE(msPostFix[isMS] +) +SLANG_SPLICE(arrayPostFix[isArray] +) +SLANG_RAW("<T=float4, let sampleCount:int=0, let format:int=0> = __TextureImpl<T, __Shape") +SLANG_SPLICE(shapeTypeNames[shape] +) +SLANG_RAW(", ") +SLANG_SPLICE(isArray +) +SLANG_RAW(", ") +SLANG_SPLICE(isMS +) +SLANG_RAW(", sampleCount, ") +SLANG_SPLICE(access +) +SLANG_RAW(", 0, ") +SLANG_SPLICE(isCombined +) +SLANG_RAW(", format>;\n") + +} +SLANG_RAW("#line 3788 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// AtomicAdd\n") +SLANG_RAW("\n") +SLANG_RAW("// Make the GLSL atomicAdd available.\n") +SLANG_RAW("// We have separate int/float implementations, as the float version requires some specific extensions\n") +SLANG_RAW("// https://www.khronos.org/registry/OpenGL/extensions/NV/NV_shader_atomic_float.txt\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_float)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_float1)]\n") +SLANG_RAW("float __atomicAdd(__ref float value, float amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_add\";\n") +SLANG_RAW(" OpCapability AtomicFloat32AddEXT;\n") +SLANG_RAW(" result:$$float = OpAtomicFAddEXT &value Device None $amount\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_atomic_fp16_vector)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_halfvec)]\n") +SLANG_RAW("half2 __atomicAdd(__ref half2 value, half2 amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_add\";\n") +SLANG_RAW(" OpCapability AtomicFloat32AddEXT;\n") +SLANG_RAW(" result:$$half2 = OpAtomicFAddEXT &value Device None $amount\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Helper for hlsl, using NVAPI\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicAdd(RWByteAddressBuffer buf, uint offset, uint2)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedAddUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// atomic add for hlsl using SM6.6\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW("void __atomicAdd(RWByteAddressBuffer buf, uint offset, int64_t value, out int64_t originalValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.InterlockedAdd64($1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW("void __atomicAdd(RWByteAddressBuffer buf, uint offset, uint64_t value, out uint64_t originalValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.InterlockedAdd64($1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Int versions require glsl 4.30\n") +SLANG_RAW("// https://www.khronos.org/registry/OpenGL-Refpages/gl4/html/atomicAdd.xhtml\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl)]\n") +SLANG_RAW("int __atomicAdd(__ref int value, int amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpAtomicIAdd &value Device None $amount;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl)]\n") +SLANG_RAW("uint __atomicAdd(__ref uint value, uint amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpAtomicIAdd &value Device None $amount;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("int64_t __atomicAdd(__ref int64_t value, int64_t amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$int64_t = OpAtomicIAdd &value Device None $amount\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicAdd(__ref uint64_t value, uint64_t amount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicIAdd &value Device None $amount\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Cas - Compare and swap\n") +SLANG_RAW("\n") +SLANG_RAW("// Helper for HLSL, using NVAPI\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __cas(RWByteAddressBuffer buf, uint offset, uint2 compareValue, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedCompareExchangeUint64($0, $1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// CAS using SM6.6\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW("void __cas(RWByteAddressBuffer buf, uint offset, in int64_t compare_value, in int64_t value, out int64_t original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.InterlockedCompareExchange64($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW("void __cas(RWByteAddressBuffer buf, uint offset, in uint64_t compare_value, in uint64_t value, out uint64_t original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.InterlockedCompareExchange64($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("int64_t __cas(__ref int64_t ioValue, int64_t compareValue, int64_t newValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicCompSwap($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$int64_t = OpAtomicCompareExchange &ioValue Device None None $newValue $compareValue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __cas(__ref uint64_t ioValue, uint64_t compareValue, uint64_t newValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicCompSwap($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicCompareExchange &ioValue Device None None $newValue $compareValue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Max\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicMax(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedMaxUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicMax(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMax($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicUMax &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_float2)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_float2)]\n") +SLANG_RAW("float __atomicMax(__ref float ioValue, float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMax($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_min_max\";\n") +SLANG_RAW(" OpCapability AtomicFloat32MinMaxEXT;\n") +SLANG_RAW(" result:$$float = OpAtomicFMaxEXT &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_float2)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_float2)]\n") +SLANG_RAW("half __atomicMax(__ref half ioValue, half value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMax($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_min_max\";\n") +SLANG_RAW(" OpCapability AtomicFloat16MinMaxEXT;\n") +SLANG_RAW(" result:$$half = OpAtomicFMaxEXT &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Min\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicMin(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedMinUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicMin(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMin($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicUMin &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_float2)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_float2)]\n") +SLANG_RAW("float __atomicMin(__ref float ioValue, float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMin($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_min_max\";\n") +SLANG_RAW(" OpCapability AtomicFloat32MinMaxEXT;\n") +SLANG_RAW(" result:$$float = OpAtomicFMinEXT &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_float2)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_float2)]\n") +SLANG_RAW("half __atomicMin(__ref half ioValue, half value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMin($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_shader_atomic_float_min_max\";\n") +SLANG_RAW(" OpCapability AtomicFloat16MinMaxEXT;\n") +SLANG_RAW(" result:$$half = OpAtomicFMinEXT &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// And\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicAnd(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedAndUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicAnd(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAnd($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicAnd &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Or\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicOr(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedOrUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicOr(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicOr($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicOr &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Xor\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicXor(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedXorUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicXor(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicXor($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicXor &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Exchange\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW("uint2 __atomicExchange(RWByteAddressBuffer buf, uint offset, uint2 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInterlockedExchangeUint64($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_atomic_int64)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, atomic_glsl_int64)]\n") +SLANG_RAW("uint64_t __atomicExchange(__ref uint64_t ioValue, uint64_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicExchange($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$uint64_t = OpAtomicExchange &ioValue Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Conversion between uint64_t and uint2\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, sm_4_0_version)]\n") +SLANG_RAW("uint2 __asuint2(uint64_t i)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return uint2(uint(i), uint(uint64_t(i) >> 32));\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, sm_4_0_version)]\n") +SLANG_RAW("uint64_t __asuint64(uint2 i)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return (uint64_t(i.y) << 32) | i.x;\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ByteAddressBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer)]\n") +SLANG_RAW("T __byteAddressBufferLoad<T>(ByteAddressBuffer buffer, int offset, int alignment);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ByteAddressBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW("T __byteAddressBufferLoad<T>(RWByteAddressBuffer buffer, int offset, int alignment);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ByteAddressBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW("T __byteAddressBufferLoad<T>(RasterizerOrderedByteAddressBuffer buffer, int offset, int alignment);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ByteAddressBufferStore +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW("void __byteAddressBufferStore<T>(RWByteAddressBuffer buffer, int offset, int alignment, T value);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ByteAddressBufferStore +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW("void __byteAddressBufferStore<T>(RasterizerOrderedByteAddressBuffer buffer, int offset, int alignment, T value);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, L:IBufferDataLayout=DefaultDataLayout>\n") +SLANG_RAW("__magic_type(HLSLStructuredBufferType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLStructuredBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("struct StructuredBuffer\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void GetDimensions(\n") +SLANG_RAW(" out uint numStructs,\n") +SLANG_RAW(" out uint stride)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let rs = __structuredBufferGetDimensions(this);\n") +SLANG_RAW(" numStructs = rs.x;\n") +SLANG_RAW(" stride = rs.y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, structuredbuffer)]\n") +SLANG_RAW(" T Load<TIndex : __BuiltinIntegerType>(TIndex location);\n") +SLANG_RAW("\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferLoadStatus +) +SLANG_RAW(")\n") +SLANG_RAW(" [require(hlsl, structuredbuffer)]\n") +SLANG_RAW(" T Load<TIndex : __BuiltinIntegerType>(TIndex location, out uint status);\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TIndex : __BuiltinIntegerType>\n") +SLANG_RAW(" __subscript(TIndex index) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, structuredbuffer)]\n") +SLANG_RAW(" get;\n") +SLANG_RAW(" };\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, L:IBufferDataLayout=DefaultDataLayout>\n") +SLANG_RAW("__magic_type(HLSLConsumeStructuredBufferType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLConsumeStructuredBufferType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, consumestructuredbuffer)]\n") +SLANG_RAW("struct ConsumeStructuredBuffer\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_StructuredBufferConsume +) +SLANG_RAW(")\n") +SLANG_RAW(" T Consume();\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void GetDimensions(\n") +SLANG_RAW(" out uint numStructs,\n") +SLANG_RAW(" out uint stride)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let result = __structuredBufferGetDimensions(this);\n") +SLANG_RAW(" numStructs = result.x;\n") +SLANG_RAW(" stride = result.y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetElement +) +SLANG_RAW(")\n") +SLANG_RAW("T __getElement<T, U, I>(U collection, I index);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, let N : int>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, hull)]\n") +SLANG_RAW("__magic_type(HLSLInputPatchType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLInputPatchType +) +SLANG_RAW(")\n") +SLANG_RAW("struct InputPatch\n") +SLANG_RAW("{\n") +SLANG_RAW(" __generic<TIndex : __BuiltinIntegerType>\n") +SLANG_RAW(" __subscript(TIndex index)->T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".operator[]\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __getElement<T>(this, index);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, let N : int>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, domain_hull)]\n") +SLANG_RAW("__magic_type(HLSLOutputPatchType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLOutputPatchType +) +SLANG_RAW(")\n") +SLANG_RAW("struct OutputPatch\n") +SLANG_RAW("{\n") +SLANG_RAW(" __generic<TIndex : __BuiltinIntegerType>\n") +SLANG_RAW(" __subscript(TIndex index)->T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".operator[]\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __getElement<T>(this, index);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") + +static const struct { + IROp op; + char const* name; +} kMutableByteAddressBufferCases[] = +{ + { kIROp_HLSLRWByteAddressBufferType, "RWByteAddressBuffer" }, + { kIROp_HLSLRasterizerOrderedByteAddressBufferType, "RasterizerOrderedByteAddressBuffer" }, +}; +for(auto item : kMutableByteAddressBufferCases) { +SLANG_RAW("#line 4410 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__magic_type(HLSL") +SLANG_SPLICE(item.name +) +SLANG_RAW("Type)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(item.op +) +SLANG_RAW(")\n") +SLANG_RAW("struct ") +SLANG_SPLICE(item.name +) +SLANG_RAW("\n") +SLANG_RAW("{\n") +SLANG_RAW(" // Note(tfoley): supports all operations from `ByteAddressBuffer`\n") +SLANG_RAW(" // TODO(tfoley): can this be made a sub-type?\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, structuredbuffer_rw)]\n") +SLANG_RAW(" void GetDimensions(out uint dim)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" dim = __structuredBufferGetDimensions(__getEquivalentStructuredBuffer<uint>(this)).x*4;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint Load(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint Load(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint2 Load2(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint2 Load2(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint2 Load2Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint2>(this, location, __naturalStrideOf<uint2>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint2 Load2(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load2\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint3 Load3(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint3 Load3(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint3 Load3Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint3>(this, location, __naturalStrideOf<uint3>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint3 Load3(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load3\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint4 Load4(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint4 Load4(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint4 Load4Aligned(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return __byteAddressBufferLoad<uint4>(this, location, __naturalStrideOf<uint4>());\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, byteaddressbuffer_rw)]\n") +SLANG_RAW(" uint4 Load4(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load4\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" T Load<T>(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, 0);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" T Load<T>(int location, int alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, alignment);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" T LoadAligned<T>(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return __byteAddressBufferLoad<T>(this, location, __naturalStrideOf<T>());\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + if (item.op == kIROp_HLSLRWByteAddressBufferType) + { +SLANG_RAW("#line 4635 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" // float32 and int64 atomic support. This is a Slang specific extension, it uses\n") +SLANG_RAW(" // GL_EXT_shader_atomic_float on Vulkan\n") +SLANG_RAW(" // NvAPI support on DX\n") +SLANG_RAW(" // NOTE! To use this feature on HLSL based targets the path to 'nvHLSLExtns.h' from the NvAPI SDK must\n") +SLANG_RAW(" // be set. That this include will be added to the *output* that is passed to a downstram compiler.\n") +SLANG_RAW(" // Also note that you *can* include NVAPI headers in your Slang source, and directly use NVAPI functions\n") +SLANG_RAW(" // Directly using NVAPI functions does *not* add the #include on the output\n") +SLANG_RAW(" // Finally note you can *mix* NVAPI direct calls, and use of NVAPI intrinsics below. This doesn't cause\n") +SLANG_RAW(" // any clashes, as Slang will emit any NVAPI function it parsed (say via a include in Slang source) with\n") +SLANG_RAW(" // unique functions.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/vkspec.html#VK_EXT_shader_atomic_float\n") +SLANG_RAW(" // https://htmlpreview.github.io/?https://github.com/KhronosGroup/SPIRV-Registry/blob/master/extensions/EXT/SPV_EXT_shader_atomic_float_add.html\n") +SLANG_RAW("\n") +SLANG_RAW(" // F32 Add\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(2.0)\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_nvapi_cuda_metal_float1)]\n") +SLANG_RAW(" void InterlockedAddF32(uint byteAddress, float valueToAdd, out float originalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($3 = NvInterlockedAddFp32($0, $1, $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicAdd($0._getPtrAt<float>($1), $2))\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<float>(this);\n") +SLANG_RAW(" __metalInterlocked_add(__getMetalAtomicRef(buf[byteAddress / 4]), valueToAdd, originalValue);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<float>(this);\n") +SLANG_RAW(" originalValue = __atomicAdd(buf[byteAddress / 4], valueToAdd);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // FP16x2\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_nvapi)]\n") +SLANG_RAW(" uint _NvInterlockedAddFp16x2(uint byteAddress, uint fp16x2Value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvInterlockedAddFp16x2($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void InterlockedAddF16(uint byteAddress, half value, out half originalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" if ((byteAddress & 2) == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" uint packedInput = asuint16(value);\n") +SLANG_RAW(" originalValue = asfloat16((uint16_t)_NvInterlockedAddFp16x2(byteAddress, packedInput));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" byteAddress = byteAddress & ~3;\n") +SLANG_RAW(" uint packedInput = ((uint)asuint16(value)) << 16;\n") +SLANG_RAW(" originalValue = asfloat16((uint16_t)(_NvInterlockedAddFp16x2(byteAddress, packedInput) >> 16));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<half2>(this);\n") +SLANG_RAW(" if ((byteAddress & 2) == 0)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" originalValue = __atomicAdd(buf[byteAddress/4], half2(value, half(0.0))).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" originalValue = __atomicAdd(buf[byteAddress/4], half2(half(0.0), value)).y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Without returning original value\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __cuda_sm_version(2.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_nvapi_cuda_metal_float1)]\n") +SLANG_RAW(" void InterlockedAddF32(uint byteAddress, float valueToAdd)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"(NvInterlockedAddFp32($0, $1, $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicAdd($0._getPtrAt<float>($1), $2)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<float>(this);\n") +SLANG_RAW(" __metalInterlocked_add(__getMetalAtomicRef(buf[byteAddress / 4]), valueToAdd);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<float>(this);\n") +SLANG_RAW(" __atomicAdd(buf[byteAddress / 4], valueToAdd);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Int64 Add\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __cuda_sm_version(6.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda6_int64)]\n") +SLANG_RAW(" void InterlockedAddI64(uint byteAddress, int64_t valueToAdd, out int64_t originalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicAdd($0._getPtrAt<uint64_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" originalValue = __asuint64(__atomicAdd(this, byteAddress, __asuint2(valueToAdd)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<int64_t>(this);\n") +SLANG_RAW(" originalValue = __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Without returning original value\n") +SLANG_RAW(" __cuda_sm_version(6.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda6_int64)]\n") +SLANG_RAW(" void InterlockedAddI64(uint byteAddress, int64_t valueToAdd)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicAdd($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __atomicAdd(this, byteAddress, __asuint2(valueToAdd));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<int64_t>(this);\n") +SLANG_RAW(" __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Cas uint64_t\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedCompareExchangeU64(uint byteAddress, uint64_t compareValue, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$4 = atomicCAS($0._getPtrAt<uint64_t>($1), $2, $3))\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" outOriginalValue = __asuint64(__cas(this, byteAddress, __asuint2(compareValue), __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" outOriginalValue = __cas(buf[byteAddress / 8], compareValue, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Max\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(5.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda5_int64)]\n") +SLANG_RAW(" uint64_t InterlockedMaxU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicMax($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicMax(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicMax(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMax64(uint byteAddress, int64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMax64(uint byteAddress, int64_t value, out int64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMax64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMax64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Min\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(5.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda5_int64)]\n") +SLANG_RAW(" uint64_t InterlockedMinU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicMin($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicMin(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicMin(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMin64(uint byteAddress, int64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMin64(uint byteAddress, int64_t value, out int64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMin64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedMin64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // And\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(5.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda5_int64)]\n") +SLANG_RAW(" uint64_t InterlockedAndU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicAnd($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicAnd(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicAnd(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedAnd64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAnd64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedAnd64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAnd64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Or\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(5.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda5_int64)]\n") +SLANG_RAW(" uint64_t InterlockedOrU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicOr($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicOr(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicOr(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedOr64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedOr64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedOr64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedOr64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Xor\n") +SLANG_RAW("\n") +SLANG_RAW(" __cuda_sm_version(5.0)\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda5_int64)]\n") +SLANG_RAW(" uint64_t InterlockedXorU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicXor($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicXor(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicXor(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedXor64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedXor64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedXor64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedXor64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Exchange\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda9_int64)]\n") +SLANG_RAW(" uint64_t InterlockedExchangeU64(uint byteAddress, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicExch($0._getPtrAt<uint64_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" return __asuint64(__atomicExchange(this, byteAddress, __asuint2(value)));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" return __atomicExchange(buf[byteAddress / 8], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedExchangeFloat(uint byteAddress, float value, out float outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedExchangeFloat\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedExchange64(uint byteAddress, int64_t value, out int64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedExchange64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedExchange64(uint byteAddress, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedExchange64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // SM6.6 6 64bit atomics.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedAdd64(uint byteAddress, int64_t valueToAdd)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd64\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<int64_t>(this);\n") +SLANG_RAW(" __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedAdd64(uint byteAddress, int64_t valueToAdd, out int64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd64\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<int64_t>(this);\n") +SLANG_RAW(" outOriginalValue = __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedAdd64(uint byteAddress, uint64_t valueToAdd)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd64\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedAdd64(uint byteAddress, uint64_t valueToAdd, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd64\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" outOriginalValue = __atomicAdd(buf[byteAddress / 8], valueToAdd);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedCompareExchange64(uint byteAddress, int64_t compareValue, int64_t value, out int64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __cas(this, byteAddress, compareValue, value, outOriginalValue);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<int64_t>(this);\n") +SLANG_RAW(" outOriginalValue = __cas(buf[byteAddress / 8], compareValue, value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_cuda9_int64)]\n") +SLANG_RAW(" void InterlockedCompareExchange64(uint byteAddress, uint64_t compareValue, uint64_t value, out uint64_t outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __cas(this, byteAddress, compareValue, value, outOriginalValue);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint64_t>(this);\n") +SLANG_RAW(" outOriginalValue = __cas(buf[byteAddress / 8], compareValue, value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedCompareStoreFloatBitwise(uint byteAddress, float compareValue, float value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareStoreFloatBitwise\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedCompareExchangeFloatBitwise(uint byteAddress, float compareValue, float value, out float outOriginalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareExchangeFloatBitwise\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedCompareStore64(uint byteAddress, int64_t compareValue, int64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareStore64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl, atomic_hlsl_sm_6_6)]\n") +SLANG_RAW(" void InterlockedCompareStore64(uint byteAddress, uint64_t compareValue, uint64_t value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareStore64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + } // endif (type == RWByteAddressBuffer) +SLANG_RAW("#line 5222 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" // Added operations:\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedAdd(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicAdd($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicAdd($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_add(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedAdd(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedAdd(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAdd($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicAdd($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAdd\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_add(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedAdd(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedAnd(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"$3 = atomicAnd($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicAnd($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAnd\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_and(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedAnd(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedAnd(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicAnd($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicAnd($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedAnd\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_and(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedAnd(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedCompareExchange(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT compare_value,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($4 = atomicCompSwap($0._data[$1/4], $2, $3))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$4 = atomicCAS($0._getPtrAt<uint32_t>($1), $2, $3))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareExchange\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(buf[dest / 4]), compare_value, value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedCompareExchange(buf[dest / 4], compare_value, value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedCompareStore(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT compare_value,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicCompSwap($0._data[$1/4], $2, $3)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicCAS($0._getPtrAt<uint32_t>($1), $2, $3)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedCompareStore\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(buf[dest / 4]), compare_value, value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedCompareStore(buf[dest / 4], compare_value, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedExchange(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicExchange($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicExch($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedExchange\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_exchange(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedExchange(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedMax(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicMax($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicMax($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_max(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedMax(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedMax(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMax($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicMax($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMax\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_max(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedMax(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedMin(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicMin($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicMin($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_min(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedMin(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedMin(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicMin($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicMin($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedMin\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_min(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedMin(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedOr(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicOr($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicOr($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedOr\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_or(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedOr(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedOr(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicOr($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicOr($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedOr\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_or(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedOr(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedXor(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value,\n") +SLANG_RAW(" out UINT original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($3 = atomicXor($0._data[$1/4], $2))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(*$3 = atomicXor($0._getPtrAt<uint32_t>($1), $2))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedXor\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_xor(__getMetalAtomicRef(buf[dest / 4]), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedXor(buf[dest / 4], value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void InterlockedXor(\n") +SLANG_RAW(" UINT dest,\n") +SLANG_RAW(" UINT value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atomicXor($0._data[$1/4], $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicXor($0._getPtrAt<uint32_t>($1), $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".InterlockedXor\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" __metalInterlocked_xor(__getMetalAtomicRef(buf[dest / 4]), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let buf = __getEquivalentStructuredBuffer<uint>(this);\n") +SLANG_RAW(" ::InterlockedXor(buf[dest / 4], value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store(uint address, uint value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, 0, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store2(uint address, uint2 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, 0, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store2(uint address, uint2 value, uint alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, alignment, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store2Aligned(uint address, uint2 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, __naturalStrideOf<uint2>(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store3(uint address, uint3 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, 0, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store3(uint address, uint3 value, uint alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, alignment, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store3Aligned(uint address, uint3 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, __naturalStrideOf<uint3>(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store4(uint address, uint4 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, 0, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store4(uint address, uint4 value, uint alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, alignment, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, byteaddressbuffer_rw)]\n") +SLANG_RAW(" void Store4Aligned(uint address, uint4 value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Store4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __byteAddressBufferStore(this, address, __naturalStrideOf<uint4>(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void Store<T>(int offset, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __byteAddressBufferStore(this, offset, 0, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void Store<T>(int offset, T value, uint alignment)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __byteAddressBufferStore(this, offset, alignment, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" void StoreAligned<T>(int offset, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __byteAddressBufferStore(this, offset, __naturalStrideOf<T>(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") + +} +SLANG_RAW("#line 5725 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") + +static const struct { + IROp op; + char const* name; +} kMutableStructuredBufferCases[] = +{ + { kIROp_HLSLRWStructuredBufferType, "RWStructuredBuffer" }, + { kIROp_HLSLRasterizerOrderedStructuredBufferType, "RasterizerOrderedStructuredBuffer" }, +}; +for(auto item : kMutableStructuredBufferCases) { +SLANG_RAW("#line 5737 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, L:IBufferDataLayout=DefaultDataLayout>\n") +SLANG_RAW("__magic_type(HLSL") +SLANG_SPLICE(item.name +) +SLANG_RAW("Type)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(item.op +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer_rw)]\n") +SLANG_RAW("struct ") +SLANG_SPLICE(item.name +) +SLANG_RAW("\n") +SLANG_RAW("{\n") +SLANG_RAW(" uint DecrementCounter();\n") +SLANG_RAW("\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cpp_cuda_glsl_hlsl_metal_spirv, structuredbuffer_rw)]\n") +SLANG_RAW(" void GetDimensions(\n") +SLANG_RAW(" out uint numStructs,\n") +SLANG_RAW(" out uint stride)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let rs = __structuredBufferGetDimensions(this);\n") +SLANG_RAW(" numStructs = rs.x;\n") +SLANG_RAW(" stride = rs.y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" uint IncrementCounter();\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_RWStructuredBufferLoad +) +SLANG_RAW(")\n") +SLANG_RAW(" T Load<TIndex : __BuiltinIntegerType>(TIndex location);\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_RWStructuredBufferLoadStatus +) +SLANG_RAW(")\n") +SLANG_RAW(" T Load<TIndex : __BuiltinIntegerType>(TIndex location, out uint status);\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TIndex : __BuiltinIntegerType>\n") +SLANG_RAW(" __subscript(TIndex index) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // If a 'Buffer[index]' is referred to by a '__ref', call 'kIROp_RWStructuredBufferGetElementPtr(index)'.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // This allows call's to stay aware that the input is from a 'Buffer'.\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [nonmutating]\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_RWStructuredBufferGetElementPtr +) +SLANG_RAW(")\n") +SLANG_RAW(" ref;\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") + +} +SLANG_RAW("#line 5789 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, geometry)]\n") +SLANG_RAW("__magic_type(HLSLPointStreamType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLPointStreamType +) +SLANG_RAW(")\n") +SLANG_RAW("struct PointStream\n") +SLANG_RAW("{\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamAppend\")]\n") +SLANG_RAW(" void Append(T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EmitVertex()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Append\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEmitVertex; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamRestart\")]\n") +SLANG_RAW(" void RestartStrip()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EndPrimitive()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".RestartStrip\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEndPrimitive; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, geometry)]\n") +SLANG_RAW("__magic_type(HLSLLineStreamType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLLineStreamType +) +SLANG_RAW(")\n") +SLANG_RAW("struct LineStream\n") +SLANG_RAW("{\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamAppend\")]\n") +SLANG_RAW(" void Append(T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EmitVertex()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Append\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEmitVertex; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamRestart\")]\n") +SLANG_RAW(" void RestartStrip()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EndPrimitive()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".RestartStrip\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEndPrimitive; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, geometry)]\n") +SLANG_RAW("__magic_type(HLSLTriangleStreamType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLTriangleStreamType +) +SLANG_RAW(")\n") +SLANG_RAW("struct TriangleStream\n") +SLANG_RAW("{\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamAppend\")]\n") +SLANG_RAW(" void Append(T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EmitVertex()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Append\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEmitVertex; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [KnownBuiltin(\"GeometryStreamRestart\")]\n") +SLANG_RAW(" void RestartStrip()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"EndPrimitive()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".RestartStrip\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpEndPrimitive; };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("#define VECTOR_MAP_UNARY(TYPE, COUNT, FUNC, VALUE) \\\n") +SLANG_RAW(" vector<TYPE,COUNT> result; for(int i = 0; i < COUNT; ++i) { result[i] = FUNC(VALUE[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("#define MATRIX_MAP_UNARY(TYPE, ROWS, COLS, FUNC, VALUE) \\\n") +SLANG_RAW(" matrix<TYPE,ROWS,COLS> result; for(int i = 0; i < ROWS; ++i) { result[i] = FUNC(VALUE[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("#define VECTOR_MAP_BINARY(TYPE, COUNT, FUNC, LEFT, RIGHT) \\\n") +SLANG_RAW(" vector<TYPE,COUNT> result; for(int i = 0; i < COUNT; ++i) { result[i] = FUNC(LEFT[i], RIGHT[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("#define MATRIX_MAP_BINARY(TYPE, ROWS, COLS, FUNC, LEFT, RIGHT) \\\n") +SLANG_RAW(" matrix<TYPE,ROWS,COLS> result; for(int i = 0; i < ROWS; ++i) { result[i] = FUNC(LEFT[i], RIGHT[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("#define VECTOR_MAP_TRINARY(TYPE, COUNT, FUNC, A, B, C) \\\n") +SLANG_RAW(" vector<TYPE,COUNT> result; for(int i = 0; i < COUNT; ++i) { result[i] = FUNC(A[i], B[i], C[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("#define MATRIX_MAP_TRINARY(TYPE, ROWS, COLS, FUNC, A, B, C) \\\n") +SLANG_RAW(" matrix<TYPE,ROWS,COLS> result; for(int i = 0; i < ROWS; ++i) { result[i] = FUNC(A[i], B[i], C[i]); } return result\n") +SLANG_RAW("\n") +SLANG_RAW("// Try to terminate the current draw or dispatch call (HLSL SM 4.0)\n") +SLANG_RAW("void abort();\n") +SLANG_RAW("\n") +SLANG_RAW("// Absolute value (HLSL SM 1.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T abs(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_abs($0)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_abs($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 SAbs $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" //default:\n") +SLANG_RAW(" // Note: this simple definition may not be appropriate for floating-point inputs\n") +SLANG_RAW(" // return x < 0 ? -x : x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> abs(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 SAbs $x;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, abs, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> abs(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, abs, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T abs(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_abs($0)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_abs($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 FAbs $x;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> abs(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 FAbs $x;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, abs, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> abs(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"abs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, abs, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T fabs(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fabs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return abs(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> fabs(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fabs\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return abs(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Inverse cosine (HLSL SM 1.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T acos(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_acos($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_acos($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Acos $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> acos(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Acos $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, acos, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> acos(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"acos\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, acos, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Inverse hyperbolic cosine\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T acosh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_acosh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_acosh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Acosh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return log(x + sqrt( x * x - T(1)));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N:int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> acosh(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Acosh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"acosh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, acosh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Test if all components are non-zero (HLSL SM 1.0)\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("bool all(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"bool($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let zero = __default<T>();\n") +SLANG_RAW(" if (__isInt<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpINotEqual $$bool result $x $zero\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFUnordNotEqual $$bool result $x $zero\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isBool<T>())\n") +SLANG_RAW(" return __slang_noop_cast<bool>(x);\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return false;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("bool all(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" if(N == 1)\n") +SLANG_RAW(" return all(x[0]);\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"all(bvec$N0($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isBool<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpAll $$bool result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zero = __default<vector<T,N>>();\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpINotEqual $$vector<bool,N> %castResult $x $zero;\n") +SLANG_RAW(" OpAll $$bool result %castResult\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zero = __default<vector<T,N>>();\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFUnordNotEqual $$vector<bool,N> %castResult $x $zero;\n") +SLANG_RAW(" OpAll $$bool result %castResult\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"all\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" bool result = true;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result = result && all(x[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("bool all(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"all\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" bool result = true;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result = result && all(x[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Barrier for writes to all memory spaces (HLSL SM 5.0)\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void AllMemoryBarrier()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"AllMemoryBarrier\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"memoryBarrier(gl_ScopeDevice, (gl_StorageSemanticsShared|gl_StorageSemanticsImage|gl_StorageSemanticsBuffer), gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__threadfence()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_device | mem_flags::mem_threadgroup | mem_flags::mem_texture | mem_flags::mem_threadgroup_imageblock)\";\n") +SLANG_RAW(" case spirv: spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpMemoryBarrier Device AcquireRelease|UniformMemory|WorkgroupMemory|ImageMemory;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"storageBarrier(); textureBarrier(); workgroupBarrier();\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Thread-group sync and barrier for writes to all memory spaces (HLSL SM 5.0)\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void AllMemoryBarrierWithGroupSync()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"AllMemoryBarrierWithGroupSync\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"controlBarrier(gl_ScopeWorkgroup, gl_ScopeDevice, (gl_StorageSemanticsShared|gl_StorageSemanticsImage|gl_StorageSemanticsBuffer), gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__syncthreads()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_device | mem_flags::mem_threadgroup | mem_flags::mem_texture | mem_flags::mem_threadgroup_imageblock)\";\n") +SLANG_RAW(" case spirv: spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpControlBarrier Workgroup Device AcquireRelease|UniformMemory|WorkgroupMemory|ImageMemory;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"storageBarrier(); textureBarrier(); workgroupBarrier();\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Returns the workgroup size of the calling entry point.\n") +SLANG_RAW("[require(compute)]\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetWorkGroupSize +) +SLANG_RAW(")\n") +SLANG_RAW("int3 WorkgroupSize();\n") +SLANG_RAW("\n") +SLANG_RAW("// Test if any components is non-zero (HLSL SM 1.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("bool any(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" __intrinsic_asm \"bool($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let zero = __default<T>();\n") +SLANG_RAW(" if (__isInt<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpINotEqual $$bool result $x $zero\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFUnordNotEqual $$bool result $x $zero\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isBool<T>())\n") +SLANG_RAW(" return __slang_noop_cast<bool>(x);\n") +SLANG_RAW(" return false;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("bool any(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" if(N == 1)\n") +SLANG_RAW(" return any(x[0]);\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"any(bvec$N0($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isBool<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpAny $$bool result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zero = __default<vector<T,N>>();\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpINotEqual $$vector<bool,N> %castResult $x $zero;\n") +SLANG_RAW(" OpAny $$bool result %castResult\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zero = __default<vector<T,N>>();\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFUnordNotEqual $$vector<bool,N> %castResult $x $zero;\n") +SLANG_RAW(" OpAny $$bool result %castResult\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"any\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" bool result = false;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result = result || any(x[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv)]\n") +SLANG_RAW("bool any(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"any\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" bool result = false;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result = result || any(x[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Reinterpret bits as a double (HLSL SM 5.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_ARB_gpu_shader5)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("double asdouble(uint lowbits, uint highbits)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asdouble\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"packDouble2x32(uvec2($0, $1))\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asdouble($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asdouble($0, $1)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" %v:$$uint2 = OpCompositeConstruct $lowbits $highbits;\n") +SLANG_RAW(" result:$$double = OpExtInst glsl450 59 %v\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_ARB_gpu_shader5)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("double2 asdouble(uint2 lowbits, uint2 highbits)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"asdouble($0, $1)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return double2(asdouble(lowbits.x, highbits.x), asdouble(lowbits.y, highbits.y));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reinterpret bits as a float (HLSL SM 4.0)\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("float asfloat(int x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asfloat($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asfloat($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"intBitsToFloat\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$float result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("float asfloat(uint x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asfloat($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asfloat($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"uintBitsToFloat\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$float result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<float, N> asfloat(vector< int, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"intBitsToFloat\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<float, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float, N, asfloat, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<float,N> asfloat(vector<uint,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"uintBitsToFloat\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<float,N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float, N, asfloat, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<float,N,M> asfloat(matrix< int,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(float, N, M, asfloat, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<float,N,M> asfloat(matrix<uint,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(float, N, M, asfloat, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// No op\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("float asfloat(float x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("vector<float,N> asfloat(vector<float,N> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("matrix<float,N,M> asfloat(matrix<float,N,M> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("// Inverse sine (HLSL SM 1.0)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T asin(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asin($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asin($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Asin $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> asin(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Asin $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T,N,asin,x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> asin(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T,N,M,asin,x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Inverse hyperbolic sine\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T asinh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asinh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asinh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Asinh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return log(x + sqrt(x * x + T(1)));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N:int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> asinh(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Asinh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"asinh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, asinh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reinterpret bits as an int (HLSL SM 4.0)\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("int asint(float x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asint($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asint($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floatBitsToInt\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$int result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("int asint(uint x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asint($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asint($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"int($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$int result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<int, N> asint(vector<float, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floatBitsToInt\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<int, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(int, N, asint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<int, N> asint(vector<uint, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" if(N == 1)\n") +SLANG_RAW(" return vector<int, N>(asint(x[0]));\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ivec$N0($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<int, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(int, N, asint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("matrix<int, N, M> asint(matrix<float, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(int, N, M, asint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("matrix<int, N, M> asint(matrix<uint, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(int, N, M, asint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// No op\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("int asint(int x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("vector<int,N> asint(vector<int,N> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("matrix<int,N,M> asint(matrix<int,N,M> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("// Reinterpret bits of double as a uint (HLSL SM 5.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_ARB_gpu_shader5)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("void asuint(double value, out uint lowbits, out uint highbits)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"{ uvec2 v = unpackDouble2x32($0); $1 = v.x; $2 = v.y; }\";\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"$P_asuint($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let uv = spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result : $$uint2 = OpBitcast $value;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" lowbits = uv.x;\n") +SLANG_RAW(" highbits = uv.y;\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reinterpret bits as a uint (HLSL SM 4.0)\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("uint asuint(float x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asuint($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asuint($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floatBitsToUint\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$uint result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("uint asuint(int x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_asuint($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_asuint($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"uint($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$uint result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<uint,N> asuint(vector<float,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floatBitsToUint\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<uint,N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, asuint, x);\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_4_0)]\n") +SLANG_RAW("vector<uint, N> asuint(vector<int, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" if(N == 1)\n") +SLANG_RAW(" return vector<uint, N>(asuint(x[0]));\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"uvec$N0($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<uint, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<$TR>($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, asuint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("matrix<uint,N,M> asuint(matrix<float,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(uint, N, M, asuint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("matrix<uint, N, M> asuint(matrix<int, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(uint, N, M, asuint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("uint asuint(uint x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("vector<uint,N> asuint(vector<uint,N> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int, let M : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("matrix<uint,N,M> asuint(matrix<uint,N,M> x)\n") +SLANG_RAW("{ return x; }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// 16-bit bitcast ops (HLSL SM 6.2)\n") +SLANG_RAW("//\n") +SLANG_RAW("// TODO: We need to map these to GLSL/SPIR-V\n") +SLANG_RAW("// operations that don't require an intermediate\n") +SLANG_RAW("// conversion to fp32.\n") +SLANG_RAW("\n") +SLANG_RAW("// Identity cases:\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] float16_t asfloat16(float16_t value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] vector<float16_t,N> asfloat16<let N : int>(vector<float16_t,N> value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] matrix<float16_t,R,C> asfloat16<let R : int, let C : int>(matrix<float16_t,R,C> value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] int16_t asint16(int16_t value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] vector<int16_t,N> asint16<let N : int>(vector<int16_t,N> value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] matrix<int16_t,R,C> asint16<let R : int, let C : int>(matrix<int16_t,R,C> value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] uint16_t asuint16(uint16_t value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] vector<uint16_t,N> asuint16<let N : int>(vector<uint16_t,N> value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] matrix<uint16_t,R,C> asuint16<let R : int, let C : int>(matrix<uint16_t,R,C> value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("// Signed<->unsigned cases:\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] int16_t asint16(uint16_t value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] vector<int16_t,N> asint16<let N : int>(vector<uint16_t,N> value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] matrix<int16_t,R,C> asint16<let R : int, let C : int>(matrix<uint16_t,R,C> value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] uint16_t asuint16(int16_t value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] vector<uint16_t,N> asuint16<let N : int>(vector<int16_t,N> value) { return value; }\n") +SLANG_RAW("[__unsafeForceInlineEarly][__readNone] matrix<uint16_t,R,C> asuint16<let R : int, let C : int>(matrix<int16_t,R,C> value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("// Float->unsigned cases:\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("uint16_t asuint16(float16_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half_as_ushort\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"uint16_t(packHalf2x16(vec2($0, 0.0)))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint16\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$uint16_t result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint16_t,N> asuint16<let N : int>(vector<float16_t,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asuint16\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<uint16_t,N> = OpBitcast $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint16_t, N, asuint16, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<uint16_t,R,C> asuint16<let R : int, let C : int>(matrix<float16_t,R,C> value)\n") +SLANG_RAW("{ MATRIX_MAP_UNARY(uint16_t, R, C, asuint16, value); }\n") +SLANG_RAW("\n") +SLANG_RAW("// Unsigned->float cases:\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("float16_t asfloat16(uint16_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__ushort_as_half\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"float16_t(unpackHalf2x16($0).x)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat16\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$float16_t result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<float16_t,N> asfloat16<let N : int>(vector<uint16_t,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat16\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<float16_t,N> = OpBitcast $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float16_t, N, asfloat16, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("matrix<float16_t,R,C> asfloat16<let R : int, let C : int>(matrix<uint16_t,R,C> value)\n") +SLANG_RAW("{ MATRIX_MAP_UNARY(float16_t, R, C, asfloat16, value); }\n") +SLANG_RAW("\n") +SLANG_RAW("// Float<->signed cases:\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("int16_t asint16(float16_t value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half_as_short\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint16\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$int16_t result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default: return asuint16(value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone] \n") +SLANG_RAW("[require(cuda_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<int16_t,N> asint16<let N : int>(vector<float16_t,N> value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint16\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" default: return asuint16(value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<int16_t,R,C> asint16<let R : int, let C : int>(matrix<float16_t,R,C> value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asint16\";\n") +SLANG_RAW(" default: return asuint16(value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[__unsafeForceInlineEarly] \n") +SLANG_RAW("[require(cuda_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("float16_t asfloat16(int16_t value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__short_as_half\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat16\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$float16_t result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default: return asfloat16(asuint16(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<float16_t,N> asfloat16<let N : int>(vector<int16_t,N> value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat16\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<$TR>($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpBitcast $$vector<float16_t,N> result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default: return asfloat16(asuint16(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<float16_t,R,C> asfloat16<let R : int, let C : int>(matrix<int16_t,R,C> value) \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"asfloat16\";\n") +SLANG_RAW(" default: return asfloat16(asuint16(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Inverse tangent (HLSL SM 1.0)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T atan(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_atan($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_atan($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Atan $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> atan(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Atan $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, atan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> atan(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, atan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T atan2(T y, T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_atan2($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_atan2($0, $1)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atan($0,$1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Atan2 $y $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> atan2(vector<T, N> y, vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atan($0,$1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Atan2 $y $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, atan2, y, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> atan2(matrix<T,N,M> y, matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"atan2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, atan2, y, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Hyperbolic inverse tangent\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T atanh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_atanh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_atanh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Atanh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return T(0.5) * log((T(1) + x) / (T(1) - x));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N:int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> atanh(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Atanh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"atanh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, atanh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Ceiling (HLSL SM 1.0)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T ceil(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_ceil($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_ceil($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Ceil $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> ceil(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Ceil $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, ceil, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> ceil(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ceil\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, ceil, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Copy-sign\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("vector<half,N> copysign_half(vector<half,N> x, vector<half,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" let ux = reinterpret<vector<uint16_t,N>>(x);\n") +SLANG_RAW(" let uy = reinterpret<vector<uint16_t,N>>(y);\n") +SLANG_RAW(" vector<uint16_t,N> signY = (uy & (uint16_t(1) << uint16_t(15)));\n") +SLANG_RAW(" vector<uint16_t,N> newX = (ux & ((uint16_t(1) << uint16_t(15)) - uint16_t(1))) + signY;\n") +SLANG_RAW(" return reinterpret<vector<half,N>>(newX);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("vector<float,N> copysign_float(vector<float,N> x, vector<float,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" let ux = reinterpret<vector<uint32_t,N>>(x);\n") +SLANG_RAW(" let uy = reinterpret<vector<uint32_t,N>>(y);\n") +SLANG_RAW(" vector<uint32_t,N> signY = (uy & (uint32_t(1) << uint32_t(31)));\n") +SLANG_RAW(" vector<uint32_t,N> newX = (ux & ((uint32_t(1) << uint32_t(31)) - uint32_t(1))) + signY;\n") +SLANG_RAW(" return reinterpret<vector<float,N>>(newX);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("vector<double,N> copysign_double(vector<double,N> x, vector<double,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" let ux = reinterpret<vector<uint64_t,N>>(x);\n") +SLANG_RAW(" let uy = reinterpret<vector<uint64_t,N>>(y);\n") +SLANG_RAW(" vector<uint64_t,N> signY = (uy & (uint64_t(1) << uint64_t(63)));\n") +SLANG_RAW(" vector<uint64_t,N> newX = (ux & ((uint64_t(1) << uint64_t(63)) - uint64_t(1))) + signY;\n") +SLANG_RAW(" return reinterpret<vector<double,N>>(newX);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T:__BuiltinFloatingPointType, U:__BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_FloatCast +) +SLANG_RAW(")\n") +SLANG_RAW("vector<T,N> __real_cast(vector<U,N> val);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("vector<T,N> copysign(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"copysign\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // sign of -0.0 needs to be respected.\n") +SLANG_RAW(" if (T is half)\n") +SLANG_RAW(" return __real_cast<T>(copysign_half(\n") +SLANG_RAW(" __real_cast<half>(x),\n") +SLANG_RAW(" __real_cast<half>(y)));\n") +SLANG_RAW(" if (T is float)\n") +SLANG_RAW(" return __real_cast<T>(copysign_float(\n") +SLANG_RAW(" __real_cast<float>(x),\n") +SLANG_RAW(" __real_cast<float>(y)));\n") +SLANG_RAW(" return __real_cast<T>(copysign_double(\n") +SLANG_RAW(" __real_cast<double>(x),\n") +SLANG_RAW(" __real_cast<double>(y)));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("T copysign(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"copysign\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return copysign(vector<T,1>(x), vector<T,1>(y))[0];\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Check access status to tiled resource\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("bool CheckAccessFullyMapped(out uint status)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"CheckAccessFullyMapped\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Clamp (HLSL SM 1.0)\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T clamp(T x, T minBound, T maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 SClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 UClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> clamp(vector<T, N> x, vector<T, N> minBound, vector<T, N> maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T, N> = OpExtInst glsl450 SClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T, N> = OpExtInst glsl450 UClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> clamp(matrix<T,N,M> x, matrix<T,N,M> minBound, matrix<T,N,M> maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T clamp(T x, T minBound, T maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 FClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> clamp(vector<T, N> x, vector<T, N> minBound, vector<T, N> maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 FClamp $x $minBound $maxBound\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> clamp(matrix<T,N,M> x, matrix<T,N,M> minBound, matrix<T,N,M> maxBound)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clamp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(max(x, minBound), maxBound);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Clip (discard) fragment conditionally\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, fragment)]\n") +SLANG_RAW("void clip(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clip\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if(x < T(0)) discard;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, fragment)]\n") +SLANG_RAW("void clip(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clip\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if(any(x < T(0))) discard;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, fragment)]\n") +SLANG_RAW("void clip(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"clip\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if(any(x < T(0))) discard;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Cosine\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T cos(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_cos($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_cos($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Cos $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> cos(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Cos $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T,N, cos, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> cos(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cos\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, cos, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Hyperbolic cosine\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T cosh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_cosh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_cosh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Cosh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> cosh(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Cosh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T,N, cosh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> cosh(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cosh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, cosh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Cosine degree\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T cospi(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cospi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return cos(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> cospi(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cospi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return cos(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Population count\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("uint countbits(uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"countbits\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"bitCount\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"popcount\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_countbits($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpBitCount $$uint result $value};\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"countOneBits\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic <let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint, N> countbits(vector<uint, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"countbits\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"bitCount\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"popcount\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpBitCount $$vector<uint, N> result $value};\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"countOneBits\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, countbits, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Cross product\n") +SLANG_RAW("// TODO: SPIRV does not support integer vectors.\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,3> cross(vector<T,3> left, vector<T,3> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,3> result glsl450 Cross $left $right\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return vector<T,3>(\n") +SLANG_RAW(" left.y * right.z - left.z * right.y,\n") +SLANG_RAW(" left.z * right.x - left.x * right.z,\n") +SLANG_RAW(" left.x * right.y - left.y * right.x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, 3> cross(vector<T, 3> left, vector<T, 3> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, 3> result glsl450 Cross $left $right\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"cross\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return vector<T, 3>(\n") +SLANG_RAW(" left.y * right.z - left.z * right.y,\n") +SLANG_RAW(" left.z * right.x - left.x * right.z,\n") +SLANG_RAW(" left.x * right.y - left.y * right.x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Convert encoded color\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("int4 D3DCOLORtoUBYTE4(float4 color)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"D3DCOLORtoUBYTE4\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"bitcast<vec4i>(pack4x8unorm($0)).zyxw\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let scaled = color.zyxw * 255.001999f;\n") +SLANG_RAW(" return int4(scaled);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Partial-difference derivatives\n") + +const char* diffDimensions[2] = {"x", "y"}; +for (auto xOrY : diffDimensions) { +SLANG_RAW("#line 7770 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]\n") +SLANG_RAW("T dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"dfd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW(" $$T result $x};\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"dpd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]\n") +SLANG_RAW("vector<T, N> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"dfd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW(" $$vector<T, N> result $x};\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"dpd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]\n") +SLANG_RAW("matrix<T, N, M> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative(); \n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW(", x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("__glsl_extension(GL_ARB_derivative_control)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("T dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Coarse\";\n") +SLANG_RAW(" case spirv: return spirv_asm {OpCapability DerivativeControl; result:$$T = OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Coarse $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_ARB_derivative_control)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("vector<T, N> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Coarse\";\n") +SLANG_RAW(" case spirv: return spirv_asm {OpCapability DerivativeControl; result:$$vector<T,N> = OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Coarse $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("matrix<T, N, M> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative(); \n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_coarse, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("__glsl_extension(GL_ARB_derivative_control)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("T dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Fine\";\n") +SLANG_RAW(" case spirv: return spirv_asm {OpCapability DerivativeControl; result:$$T = OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Fine $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_ARB_derivative_control)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("vector<T, N> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dFd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Fine\";\n") +SLANG_RAW(" case spirv: return spirv_asm {OpCapability DerivativeControl; result:$$vector<T,N> = OpDPd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("Fine $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing_derivativecontrol)]\n") +SLANG_RAW("matrix<T, N, M> dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative(); \n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, dd") +SLANG_SPLICE(xOrY +) +SLANG_RAW("_fine, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} // for (xOrY) +SLANG_RAW("#line 7924 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Radians to degrees\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("T degrees(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Degrees $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * (T(180) / T.getPi());\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("vector<T, N> degrees(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Degrees $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, degrees, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("matrix<T, N, M> degrees(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"degrees\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, degrees, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Matrix determinant\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[PreferCheckpoint]\n") +SLANG_RAW("[require(glsl_hlsl_metal_spirv_wgsl)]\n") +SLANG_RAW("T determinant(matrix<T,N,N> m)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"determinant\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"determinant\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"determinant\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Determinant $m\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"determinant\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Barrier for device memory\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void DeviceMemoryBarrier()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"DeviceMemoryBarrier\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"memoryBarrier(gl_ScopeDevice, (gl_StorageSemanticsImage|gl_StorageSemanticsBuffer), gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__threadfence()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_device | mem_flags::mem_threadgroup | mem_flags::mem_texture | mem_flags::mem_threadgroup_imageblock)\";\n") +SLANG_RAW(" case spirv: spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpMemoryBarrier Device AcquireRelease|UniformMemory|ImageMemory;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"storageBarrier(); textureBarrier(); workgroupBarrier();\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void DeviceMemoryBarrierWithGroupSync()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"DeviceMemoryBarrierWithGroupSync\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"controlBarrier(gl_ScopeWorkgroup, gl_ScopeDevice, (gl_StorageSemanticsImage|gl_StorageSemanticsBuffer), gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__syncthreads()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_device | mem_flags::mem_threadgroup | mem_flags::mem_texture | mem_flags::mem_threadgroup_imageblock)\";\n") +SLANG_RAW(" case spirv: spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpControlBarrier Workgroup Device AcquireRelease|UniformMemory|ImageMemory;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"storageBarrier(); textureBarrier(); workgroupBarrier();\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Vector distance\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T distance(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Distance $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return length(x - y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T distance(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Distance $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"distance\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return length(x - y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// fdim\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fdim(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fdim\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return max(T(0), x - y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fdim(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fdim\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return max(T(0), x - y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// divide\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("T divide(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"divide\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x / y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv)]\n") +SLANG_RAW("vector<T,N> divide(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"divide\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x / y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Vector dot product\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T dot(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * y;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T dot(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpDot $$T result $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" T result = T(0);\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result += x[i] * y[i];\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T dot(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" T result = T(0);\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" result += x[i] * y[i];\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Helper for computing distance terms for lighting (obsolete)\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType> vector<T,4> dst(vector<T,4> x, vector<T,4> y);\n") +SLANG_RAW("\n") +SLANG_RAW("// Given a RWByteAddressBuffer allow it to be interpreted as a RWStructuredBuffer\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetEquivalentStructuredBuffer +) +SLANG_RAW(")\n") +SLANG_RAW("RWStructuredBuffer<T> __getEquivalentStructuredBuffer<T>(RWByteAddressBuffer b);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetEquivalentStructuredBuffer +) +SLANG_RAW(")\n") +SLANG_RAW("StructuredBuffer<T> __getEquivalentStructuredBuffer<T>(ByteAddressBuffer b);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetEquivalentStructuredBuffer +) +SLANG_RAW(")\n") +SLANG_RAW("RasterizerOrderedStructuredBuffer<T> __getEquivalentStructuredBuffer<T>(RasterizerOrderedByteAddressBuffer b);\n") +SLANG_RAW("\n") +SLANG_RAW("// Error message\n") +SLANG_RAW("\n") +SLANG_RAW("// void errorf( string format, ... );\n") +SLANG_RAW("\n") +SLANG_RAW("// Attribute evaluation\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: The matrix cases of these functions won't actuall work\n") +SLANG_RAW("// when compiled to GLSL, since they only support scalar/vector\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: Should these be constrains to `__BuiltinFloatingPointType`?\n") +SLANG_RAW("// TODO: SPIRV-direct does not support non-floating-point types.\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("T EvaluateAttributeAtCentroid(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtCentroid\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 InterpolateAtCentroid $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("vector<T,N> EvaluateAttributeAtCentroid(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtCentroid\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 InterpolateAtCentroid $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("matrix<T,N,M> EvaluateAttributeAtCentroid(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtCentroid\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, EvaluateAttributeAtCentroid, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("T EvaluateAttributeAtSample(T x, uint sampleindex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtSample($0, int($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 InterpolateAtSample $x $sampleindex\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("vector<T,N> EvaluateAttributeAtSample(vector<T,N> x, uint sampleindex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtSample($0, int($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 InterpolateAtSample $x $sampleindex\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("matrix<T,N,M> EvaluateAttributeAtSample(matrix<T,N,M> x, uint sampleindex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtSample($0, int($1))\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,N,M> result;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result[i] = EvaluateAttributeAtSample(x[i], sampleindex);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("T EvaluateAttributeSnapped(T x, int2 offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtOffset($0, vec2($1) / 16.0f)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" const float2 tmp = float2(16.f, 16.f);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %foffset:$$float2 = OpConvertSToF $offset;\n") +SLANG_RAW(" %offsetdiv16:$$float2 = OpFDiv %foffset $tmp;\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 InterpolateAtOffset $x %offsetdiv16\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("vector<T,N> EvaluateAttributeSnapped(vector<T,N> x, int2 offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtOffset($0, vec2($1) / 16.0f)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" const float2 tmp = float2(16.f, 16.f);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %foffset:$$float2 = OpConvertSToF $offset;\n") +SLANG_RAW(" %offsetdiv16:$$float2 = OpFDiv %foffset $tmp;\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 InterpolateAtOffset $x %offsetdiv16\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("matrix<T,N,M> EvaluateAttributeSnapped(matrix<T,N,M> x, int2 offset)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"interpolateAtOffset($0, vec2($1) / 16.0f)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,N,M> result;\n") +SLANG_RAW(" for(int i = 0; i < N; ++i)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result[i] = EvaluateAttributeSnapped(x[i], offset);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-e exponent\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T exp(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_exp($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_exp($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Exp $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> exp(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Exp $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, exp, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> exp(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"exp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, exp, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-2 exponent\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T exp2(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"exp2($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isHalf<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm { OpExtInst $$T result glsl450 Exp2 $x };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" float xf = __realCast<float>(x);\n") +SLANG_RAW(" return T(spirv_asm {\n") +SLANG_RAW(" result:$$float = OpExtInst glsl450 Exp2 $xf\n") +SLANG_RAW(" });\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"exp2($0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" case cpp: \n") +SLANG_RAW(" __intrinsic_asm \"$P_exp2($0)\";\n") +SLANG_RAW(" case cuda: \n") +SLANG_RAW(" __intrinsic_asm \"$P_exp2($0)\";\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" \n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> exp2(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"exp2($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Exp2 $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, exp2, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> exp2(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"exp2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, exp2, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-10 exponent\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T exp10(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp10\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" const T ln10 = T(2.302585092994045901); // ln(10)\n") +SLANG_RAW(" return exp(x * ln10);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> exp10(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"exp10\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" const T ln10 = T(2.30258509299); // ln(10)\n") +SLANG_RAW(" return exp(x * ln10);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Convert 16-bit float stored in low bits of integer\n") +SLANG_RAW("__glsl_version(420)\n") +SLANG_RAW("__cuda_sm_version(6.0)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("float f16tof32(uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"unpackHalf2x16($0).x\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f16tof32($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half2float(__ushort_as_half($0))\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"f16tof32($0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<half>((ushort)($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %lowBits = OpUConvert $$uint16_t $value;\n") +SLANG_RAW(" %half = OpBitcast $$half %lowBits;\n") +SLANG_RAW(" result:$$float = OpFConvert %half\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"unpack2x16float($0).x\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<float, N> f16tof32(vector<uint, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f16tof32\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %lowBits = OpUConvert $$vector<uint16_t,N> $value;\n") +SLANG_RAW(" %half = OpBitcast $$vector<half,N> %lowBits;\n") +SLANG_RAW(" result:$$vector<float,N> = OpFConvert %half\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float, N, f16tof32, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Convert to 16-bit float stored in low bits of integer\n") +SLANG_RAW("__glsl_version(420)\n") +SLANG_RAW("__cuda_sm_version(6.0)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("uint f32tof16(float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"packHalf2x16(vec2($0,0.0))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f32tof16($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half_as_ushort(__float2half($0))\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"f32tof16($0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"as_type<ushort>((half)($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %half = OpFConvert $$half $value;\n") +SLANG_RAW(" %lowBits = OpBitcast $$uint16_t %half;\n") +SLANG_RAW(" result:$$uint = OpUConvert %lowBits\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"pack2x16float(vec2f($0,0.0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint, N> f32tof16(vector<float, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f32tof16\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %half = OpFConvert $$vector<half,N> $value;\n") +SLANG_RAW(" %lowBits = OpBitcast $$vector<uint16_t,N> %half;\n") +SLANG_RAW(" result:$$vector<uint,N> = OpUConvert %lowBits\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, f32tof16, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n") +SLANG_RAW("// The following is Slang specific and NOT part of standard HLSL\n") +SLANG_RAW("// It's not clear what happens with float16 time in HLSL -> can the float16 coerce to uint for example? If so that would\n") +SLANG_RAW("// give the wrong result\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(420)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("float f16tof32(float16_t value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"unpackHalf2x16($0).x\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f16tof32($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half2float($0)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"f16tof32($0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"float($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$float = OpFConvert $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"f32($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<float, N> f16tof32(vector<float16_t, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__half2float\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"f16tof32\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"$TR($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpFConvert $$vector<float, N> result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float, N, f16tof32, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Convert to float16_t\n") +SLANG_RAW("__glsl_version(420)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("float16_t f32tof16_(float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__float2half\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"packHalf2x16(vec2($0,0.0))\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"half($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpFConvert $$float16_t result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"f16($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cuda_glsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<float16_t, N> f32tof16_(vector<float, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__float2half\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"$TR($0)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpFConvert $$vector<float16_t, N> result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(float16_t, N, f32tof16_, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n") +SLANG_RAW("\n") +SLANG_RAW("// Flip surface normal to face forward, if needed\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> faceforward(vector<T,N> n, vector<T,N> i, vector<T,N> ng)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"faceforward\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"faceforward\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"faceforward\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 FaceForward $n $i $ng\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"faceForward\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return dot(ng, i) < T(0.0f) ? n : -n;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Find first set bit starting at high bit and working down\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("int firstbithigh(int value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_firstbithigh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_firstbithigh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findMSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbithigh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$int result glsl450 FindSMsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstLeadingBit\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<int, N> firstbithigh(vector<int, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findMSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbithigh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<int, N> result glsl450 FindSMsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstLeadingBit\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(int, N, firstbithigh, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("uint firstbithigh(uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_firstbithigh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_firstbithigh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findMSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbithigh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$uint result glsl450 FindUMsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstLeadingBit\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint,N> firstbithigh(vector<uint,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findMSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbithigh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"clz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<uint,N> result glsl450 FindUMsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstLeadingBit\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, firstbithigh, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Find first set bit starting at low bit and working up\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("int firstbitlow(int value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_firstbitlow($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_firstbitlow($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findLSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbitlow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ctz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$int result glsl450 FindILsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstTrailingBit\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<int,N> firstbitlow(vector<int,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findLSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbitlow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ctz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<int,N> result glsl450 FindILsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstTrailingBit\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(int, N, firstbitlow, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("uint firstbitlow(uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_firstbitlow($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_firstbitlow($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findLSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbitlow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ctz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$uint result glsl450 FindILsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstTrailingBit\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint,N> firstbitlow(vector<uint,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"findLSB\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"firstbitlow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ctz\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<uint,N> result glsl450 FindILsb $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"firstTrailingBit\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, firstbitlow, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Floor (HLSL SM 1.0)\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T floor(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_floor($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_floor($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Floor $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> floor(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Floor $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, floor, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> floor(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"floor\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, floor, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Fused multiply-add\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("T fma(T a, T b, T c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" if (__isFloat<T>() || __isHalf<T>())\n") +SLANG_RAW(" return mad(a, b, c);\n") +SLANG_RAW(" else\n") +SLANG_RAW(" __intrinsic_asm \"fma($0, $1, $2)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Fma $a $b $c\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return a*b + c;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<T, N> fma(vector<T, N> a, vector<T, N> b, vector<T, N> c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Fma $a $b $c\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, fma, a, b, c);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<T, N, M> fma(matrix<T, N, M> a, matrix<T, N, M> b, matrix<T, N, M> c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_TRINARY(T, N, M, fma, a, b, c);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Floating point remainder of x/y\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T fmod(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" // In HLSL, `fmod` returns a remainder.\n") +SLANG_RAW(" // Definition of `fmod` in HLSL is,\n") +SLANG_RAW(" // \"The floating-point remainder is calculated such that x = i * y + f,\n") +SLANG_RAW(" // where i is an integer, f has the same sign as x, and the absolute value\n") +SLANG_RAW(" // of f is less than the absolute value of y.\"\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // In GLSL, `mod` is a Modulus function.\n") +SLANG_RAW(" // OpenGL document defines \"Modulus\" as \"Returns x - y * floor(x / y)\".\n") +SLANG_RAW(" // The use of \"Floor()\" makes the difference.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // In Metal, `fmod` is Modulus function.\n") +SLANG_RAW(" // Metal document defines it as \"Returns x - y * trunc(x/y)\".\n") +SLANG_RAW(" // Note that the function name is same to HLSL but it behaves differently.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // The tricky ones are when x or y is a negative value.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // | Remainder | Modulus\n") +SLANG_RAW(" // x y | x= i*y +f | x-y*floor(x/y)\n") +SLANG_RAW(" // ------+-----------+------------------------------\n") +SLANG_RAW(" // 4 3 | 4= 1*3 +1 | 4-3*floor( 4/3) = 4-3* 1 = 1\n") +SLANG_RAW(" // 3 3 | 3= 1*3 +0 | 3-3*floor( 3/3) = 3-3* 1 = 0\n") +SLANG_RAW(" // 2 3 | 2= 0*3 +2 | 2-3*floor( 2/3) = 2-3* 0 = 2\n") +SLANG_RAW(" // 1 3 | 1= 0*3 +1 | 1-3*floor( 1/3) = 1-3* 0 = 1\n") +SLANG_RAW(" // 0 3 | 0= 0*3 +0 | 0-3*floor( 0/3) = 0-3* 0 = 0\n") +SLANG_RAW(" // -1 3 |-1= 0*3 -1 |-1-3*floor(-1/3) =-1-3*-1 = 2\n") +SLANG_RAW(" // -2 3 |-2= 0*3 -2 |-2-3*floor(-2/3) =-2-3*-1 = 1\n") +SLANG_RAW(" // -3 3 |-3=-1*3 0 |-3-3*floor(-3/3) =-3-3*-1 = 0\n") +SLANG_RAW(" // -4 3 |-4=-1*3 -1 |-4-3*floor(-4/3) =-4-3*-2 = 2\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // When y is a negative value,\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // | Remainder | Modulus\n") +SLANG_RAW(" // x y | x= i*y +f | x-y*floor(x/y)\n") +SLANG_RAW(" // ------+-----------+------------------------------\n") +SLANG_RAW(" // 4 -3 | 4=-1*-3+1 | 4+3*floor( 4/-3) = 4+3*-2 =-2\n") +SLANG_RAW(" // 3 -3 | 3=-1*-3+0 | 3+3*floor( 3/-3) = 3+3*-1 = 0\n") +SLANG_RAW(" // 2 -3 | 2= 0*-3+2 | 2+3*floor( 2/-3) = 2+3*-1 =-1\n") +SLANG_RAW(" // 1 -3 | 1= 0*-3+1 | 1+3*floor( 1/-3) = 1+3*-1 =-2\n") +SLANG_RAW(" // 0 -3 | 0= 0*-3+0 | 0+3*floor( 0/-3) = 0+3* 0 = 0\n") +SLANG_RAW(" // -1 -3 |-1= 0*-3-1 |-1+3*floor(-1/-3) =-1+3* 0 =-1\n") +SLANG_RAW(" // -2 -3 |-2= 0*-3-2 |-2+3*floor(-2/-3) =-2+3* 0 =-2\n") +SLANG_RAW(" // -3 -3 |-3= 1*-3 0 |-3+3*floor(-3/-3) =-3+3* 1 = 0\n") +SLANG_RAW(" // -4 -3 |-4= 1*-3-1 |-4+3*floor(-4/-3) =-4+3* 1 =-1\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_fmod($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_fmod($0, $1)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" // GLSL doesn't have a function for remainder.\n") +SLANG_RAW(" __intrinsic_asm \"(($0 < 0.0) ? -mod(-$0,abs($1)) : mod($0,abs($1)))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"fmod\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" // Metal doesn't have a function for remainder.\n") +SLANG_RAW(" __intrinsic_asm \"(($0 < 0.0) ? -fmod(-$0,abs($1)) : fmod($0,abs($1)))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" // OpFRem return \"The floating-point remainder whose sign\n") +SLANG_RAW(" // matches the sign of Operand 1\", where Operand 1 is \"x\".\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$T = OpFRem $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"(($0) % ($1))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> fmod(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"fmod\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpFRem $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, fmod, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> fmod(matrix<T, N, M> x, matrix<T, N, M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"fmod\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, fmod, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Fractional part\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T frac(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_frac($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_frac($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"frac\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Fract $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> frac(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"frac\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Fract $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"fract\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, frac, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("matrix<T, N, M> frac(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, frac, x);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T fract(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return frac(x);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N:int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> fract(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return frac(x);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Split float into mantissa and exponent\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T frexp(T x, out int exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_frexp($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_frexp($0, $1)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"frexp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"frexp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"frexp($0, *($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 Frexp $x &exp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" T fract;\n") +SLANG_RAW(" __wgsl_frexp<T>(x, fract, exp);\n") +SLANG_RAW(" return fract;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(wgsl)]\n") +SLANG_RAW("void __wgsl_frexp(T x, out T fract, out int exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"{ var s = frexp($0); ($1) = s.fract; ($2) = s.exp; }\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> frexp(vector<T, N> x, out vector<int, N> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"frexp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"frexp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"frexp($0, *($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T, N> = OpExtInst glsl450 Frexp $x &exp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" vector<T,N> fract;\n") +SLANG_RAW(" __wgsl_frexp<T>(x, fract, exp);\n") +SLANG_RAW(" return fract;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, frexp, x, exp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(wgsl)]\n") +SLANG_RAW("void __wgsl_frexp(vector<T, N> x, out vector<T, N> fract, out vector<int, N> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"{ var s = frexp($0); ($1) = s.fract; ($2) = s.exp; }\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int, let L : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> frexp(matrix<T, N, M> x, out matrix<int, N, M, L> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"frexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, frexp, x, exp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Texture filter width\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_metal_spirv_wgsl, fragmentprocessing)]\n") +SLANG_RAW("T fwidth(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFwidth $$T result $x;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv_wgsl, fragmentprocessing)]\n") +SLANG_RAW("vector<T, N> fwidth(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __requireComputeDerivative();\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpFwidth $$vector<T, N> result $x;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, fragmentprocessing)]\n") +SLANG_RAW("matrix<T, N, M> fwidth(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"fwidth($0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, fwidth, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetPerVertexInputArray +) +SLANG_RAW(")\n") +SLANG_RAW("Array<T, 3> __GetPerVertexInputArray<T>(T attribute);\n") +SLANG_RAW("\n") +SLANG_RAW("/// Get the value of a vertex attribute at a specific vertex.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// The `GetAttributeAtVertex()` function can be used in a fragment shader\n") +SLANG_RAW("/// to get the value of the given `attribute` at the vertex of the primitive\n") +SLANG_RAW("/// that corresponds to the given `vertexIndex`.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Note that the `attribute` must have been a declared varying input to\n") +SLANG_RAW("/// the fragment shader with the `nointerpolation` modifier.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// This function can be applied to scalars, vectors, and matrices of\n") +SLANG_RAW("/// built-in scalar types.\n") +SLANG_RAW("///\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("__glsl_extension(GL_EXT_fragment_shader_barycentric)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, getattributeatvertex)]\n") +SLANG_RAW("[KnownBuiltin(\"GetAttributeAtVertex\")]\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("T GetAttributeAtVertex(T attribute, uint vertexIndex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"GetAttributeAtVertex\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return __GetPerVertexInputArray(attribute)[vertexIndex];\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("/// Get the value of a vertex attribute at a specific vertex.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// The `GetAttributeAtVertex()` function can be used in a fragment shader\n") +SLANG_RAW("/// to get the value of the given `attribute` at the vertex of the primitive\n") +SLANG_RAW("/// that corresponds to the given `vertexIndex`.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Note that the `attribute` must have been a declared varying input to\n") +SLANG_RAW("/// the fragment shader with the `nointerpolation` modifier.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// This function can be applied to scalars, vectors, and matrices of\n") +SLANG_RAW("/// built-in scalar types.\n") +SLANG_RAW("///\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("__glsl_extension(GL_EXT_fragment_shader_barycentric)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, getattributeatvertex)]\n") +SLANG_RAW("vector<T,N> GetAttributeAtVertex(vector<T,N> attribute, uint vertexIndex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"GetAttributeAtVertex\";\n") +SLANG_RAW(" case glsl: \n") +SLANG_RAW(" __intrinsic_asm \"$0[$1]\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %_ptr_Input_vectorT = OpTypePointer Input $$vector<T,N>;\n") +SLANG_RAW(" %addr = OpAccessChain %_ptr_Input_vectorT $attribute $vertexIndex;\n") +SLANG_RAW(" result:$$vector<T,N> = OpLoad %addr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("/// Get the value of a vertex attribute at a specific vertex.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// The `GetAttributeAtVertex()` function can be used in a fragment shader\n") +SLANG_RAW("/// to get the value of the given `attribute` at the vertex of the primitive\n") +SLANG_RAW("/// that corresponds to the given `vertexIndex`.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Note that the `attribute` must have been a declared varying input to\n") +SLANG_RAW("/// the fragment shader with the `nointerpolation` modifier.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// This function can be applied to scalars, vectors, and matrices of\n") +SLANG_RAW("/// built-in scalar types.\n") +SLANG_RAW("///\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("__glsl_extension(GL_EXT_fragment_shader_barycentric)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, getattributeatvertex)]\n") +SLANG_RAW("matrix<T,N,M> GetAttributeAtVertex(matrix<T,N,M> attribute, uint vertexIndex)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"GetAttributeAtVertex\";\n") +SLANG_RAW(" case glsl: \n") +SLANG_RAW(" __intrinsic_asm \"$0[$1]\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %_ptr_Input_matrixT = OpTypePointer Input $$matrix<T,N,M>;\n") +SLANG_RAW(" %addr = OpAccessChain %_ptr_Input_matrixT $attribute $vertexIndex;\n") +SLANG_RAW(" result:$$matrix<T,N,M> = OpLoad %addr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Get number of samples in render target\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(hlsl, sm_4_0)]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("uint GetRenderTargetSampleCount()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GetRenderTargetSampleCount\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"get_num_samples\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Get position of given sample\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(hlsl, sm_4_0)]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("float2 GetRenderTargetSamplePosition(int Index)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GetRenderTargetSamplePosition\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"get_sample_position\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Group memory barrier\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void GroupMemoryBarrier()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"memoryBarrier(gl_ScopeWorkgroup, gl_StorageSemanticsShared, gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GroupMemoryBarrier\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__threadfence_block\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_threadgroup)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpMemoryBarrier Workgroup AcquireRelease|WorkgroupMemory\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"workgroupBarrier\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, memorybarrier)]\n") +SLANG_RAW("void __subgroupBarrier()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBarrier\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GroupMemoryBarrierWithGroupSync\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__syncthreads()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_threadgroup)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpControlBarrier Subgroup Subgroup AcquireRelease|WorkgroupMemory|ImageMemory|UniformMemory\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_memory_scope_semantics)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv_wgsl, memorybarrier)]\n") +SLANG_RAW("void GroupMemoryBarrierWithGroupSync()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"controlBarrier(gl_ScopeWorkgroup, gl_ScopeWorkgroup, gl_StorageSemanticsShared, gl_SemanticsAcquireRelease)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GroupMemoryBarrierWithGroupSync\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__syncthreads()\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"threadgroup_barrier(mem_flags::mem_threadgroup)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpControlBarrier Workgroup Workgroup AcquireRelease|WorkgroupMemory\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"workgroupBarrier\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Atomics\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_MetalAtomicCast +) +SLANG_RAW(")\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("T* __getMetalAtomicRef(__ref T x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Checks if input is a ImageSubscript\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_IsTextureAccess +) +SLANG_RAW(")\n") +SLANG_RAW("bool __isTextureAccess(__ref T x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Checks if input is a texture of T type scalar\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_IsTextureScalarAccess +) +SLANG_RAW(")\n") +SLANG_RAW("bool __isTextureScalarAccess(__ref T x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Checks if input is a texture array\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_IsTextureArrayAccess +) +SLANG_RAW(")\n") +SLANG_RAW("bool __isTextureArrayAccess(__ref T x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Accepts an ImageSubscript\n") +SLANG_RAW("// Gets Texture used with ImageSubscript.\n") +SLANG_RAW("__generic<TextureAccess>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ExtractTextureFromTextureAccess +) +SLANG_RAW(")\n") +SLANG_RAW("TextureAccess* __extractTextureFromTextureAccess(__ref TextureAccess x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Accepts an ImageSubscript\n") +SLANG_RAW("// Gets Coord from ImageSubscript. Swizzles out ArrayCoord if applicable\n") +SLANG_RAW("__generic<TextureAccess>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ExtractCoordFromTextureAccess +) +SLANG_RAW(")\n") +SLANG_RAW("uint __extractCoordFromTextureAccess(__ref TextureAccess x);\n") +SLANG_RAW("\n") +SLANG_RAW("// Accepts an ImageSubscript\n") +SLANG_RAW("// Gets ArrayCoord from ImageSubscript\n") +SLANG_RAW("__generic<TextureAccess>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ExtractArrayCoordFromTextureAccess +) +SLANG_RAW(")\n") +SLANG_RAW("uint __extractArrayCoordFromTextureAccess(__ref TextureAccess x);\n") +SLANG_RAW("\n") + +for (bool isArray : {false, true}) +{ + StringBuilder coordBuilder; + StringBuilder coordFetchBuilder; + + StringBuilder threeParamsASMBuilder; + StringBuilder threeParamsOutputParamASMBuilder; + + StringBuilder fourParamsASMBuilder; + + coordBuilder << "Coord coord"; + coordFetchBuilder << "coord"; + + threeParamsASMBuilder << "$1, $2"; + + fourParamsASMBuilder << "$1, $2, $3"; + if(isArray) + { + coordBuilder << ", uint arrayCoord"; + coordFetchBuilder << ", arrayCoord"; + threeParamsASMBuilder << ", $3"; + fourParamsASMBuilder << ", $4"; + threeParamsOutputParamASMBuilder << "$4"; + } + else + { + threeParamsOutputParamASMBuilder << "$3"; + } + auto coordString = coordBuilder.toString(); + auto coordFetchString = coordFetchBuilder.toString(); + + auto threeParamsASMString = threeParamsASMBuilder.toString(); + auto threeParamsOutputParamASMString = threeParamsOutputParamASMBuilder.toString(); + + auto fourParamsASMString = fourParamsASMBuilder.toString(); +SLANG_RAW("#line 9541 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") + + for (const char* atomicOperation : {"add", "and", "max", "min", "or", "sub", "xor"}) + { +SLANG_RAW("#line 9546 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" vector<T, 4> __metalImageInterlocked_") +SLANG_SPLICE(atomicOperation +) +SLANG_RAW("(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", vector<T, 4> value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"__metalImageInterlocked only allows 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" __intrinsic_asm \"$0.atomic_fetch_") +SLANG_SPLICE(atomicOperation +) +SLANG_RAW("(") +SLANG_SPLICE(threeParamsASMString +) +SLANG_RAW(")\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalImageInterlocked_") +SLANG_SPLICE(atomicOperation +) +SLANG_RAW("(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", vector<T, 4> value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"__metalImageInterlocked only allows 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" original_value = __metalImageInterlocked_") +SLANG_SPLICE(atomicOperation +) +SLANG_RAW("(tex, ") +SLANG_SPLICE(coordFetchString +) +SLANG_RAW(", value)[0];\n") +SLANG_RAW(" }\n") + + } // atomicOperation +SLANG_RAW("#line 9570 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" vector<T, 4> __metalImageInterlocked_exchange(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", vector<T, 4> value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"__metalImageInterlocked only allows 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" __intrinsic_asm \"($0.atomic_exchange(") +SLANG_SPLICE(threeParamsASMString +) +SLANG_RAW("))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalImageInterlocked_exchange(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", vector<T, 4> value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"Metal atomic texture operations only allow 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" original_value = __metalImageInterlocked_exchange(tex, ") +SLANG_SPLICE(coordFetchString +) +SLANG_RAW(", value)[0];\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalImageInterlocked_compare_exchange(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", __ref vector<T, 4> compare_value, vector<T, 4> value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"__metalImageInterlocked only allows 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" __intrinsic_asm \"($0.atomic_compare_exchange_weak(") +SLANG_SPLICE(fourParamsASMString +) +SLANG_RAW("))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __generic<TextureType, T, Coord>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalImageInterlocked_compare_exchange(TextureType tex, ") +SLANG_SPLICE(coordString +) +SLANG_RAW(", vector<T, 4> compare_value, vector<T, 4> value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(T is int || T is uint, \"__metalImageInterlocked only allows 'int'/'uint' textures\");\n") +SLANG_RAW(" static_assert(Coord is uint || Coord is vector<uint,2> || Coord is vector<uint,3> || Coord is vector<uint,4>,\n") +SLANG_RAW(" \"__metalImageInterlocked implementation only allows 'uint' coordinates\");\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(tex, ") +SLANG_SPLICE(coordFetchString +) +SLANG_RAW(", compare_value, value);\n") +SLANG_RAW(" original_value = compare_value[0];\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + +} // isArray +SLANG_RAW("#line 9617 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") + + +// Generated functions: + +// atomicAdd, InterlockedAdd, atomic_fetch_add_explicit, OpAtomicIAdd, OpAtomicFAddEXT +// __cudaInterlocked_add, __glslInterlocked_add, __hlslInterlocked_add, __metalInterlocked_add, __spirvInterlocked_add + +// atomicAnd, InterlockedAnd, atomic_fetch_and_explicit, OpAtomicAnd +// __cudaInterlocked_and, __glslInterlocked_and, __hlslInterlocked_and, __metalInterlocked_and, __spirvInterlocked_and + +// atomicMax, InterlockedMax, atomic_fetch_max_explicit, OpAtomicUMax, OpAtomicSMax, OpAtomicFMaxEXT +// __cudaInterlocked_max, __glslInterlocked_max, __hlslInterlocked_max, __metalInterlocked_max, __spirvInterlocked_max + +// atomicMin, InterlockedMin, atomic_fetch_min_explicit, OpAtomicUMin, OpAtomicSMin, OpAtomicFMinEXT +// __cudaInterlocked_min, __glslInterlocked_min, __hlslInterlocked_min, __metalInterlocked_min, __spirvInterlocked_min + +// atomicOr, InterlockedOr, atomic_fetch_or_explicit, OpAtomicOr +// __cudaInterlocked_or, __glslInterlocked_or, __hlslInterlocked_or, __metalInterlocked_or, __spirvInterlocked_or + +// atomicXor, InterlockedXor, atomic_fetch_xor_explicit, OpAtomicXor +// __cudaInterlocked_xor, __glslInterlocked_xor, __hlslInterlocked_xor, __metalInterlocked_xor, __spirvInterlocked_xor + +// atomicExchange, atomicExch, InterlockedExchange, atomic_exchange_explicit, OpAtomicExchange +// __cudaInterlocked_exchange, __glslInterlocked_exchange, __hlslInterlocked_exchange, __metalInterlocked_exchange, __spirvInterlocked_exchange + +struct InternalAtomicOperationInfo +{ + const char* slangSuffix; + const char* cudaSuffix; + const char* glslSuffix; + const char* hlslSuffix; + const char* metalSuffix; + const char* spirvFloatSuffix; + const char* spirvUIntSuffix; + const char* spirvIntSuffix; + + const char* assertExpr; +}; + +InternalAtomicOperationInfo internalAtomicOperationInfo[7] = { + { "add", "Add", "Add", "Add", "fetch_add", "FAddEXT", "IAdd", "IAdd", "true" }, + { "and", "And", "And", "And", "fetch_and", "And", "And", "And", "!__isFloat<T>()" }, + { "max", "Max", "Max", "Max", "fetch_max", "FMaxEXT", "UMax", "SMax", "true" }, + { "min", "Min", "Min", "Min", "fetch_min", "FMinEXT", "UMin", "SMin", "true" }, + { "or", "Or", "Or", "Or", "fetch_or", "Or", "Or", "Or", "!__isFloat<T>()" }, + { "xor", "Xor", "Xor", "Xor", "fetch_xor", "Xor", "Xor", "Xor", "!__isFloat<T>()" }, + { "exchange", "Exch", "Exchange", "Exchange", "exchange", "Exchange", "Exchange", "Exchange", "true" }, +}; + +for (InternalAtomicOperationInfo atomicOp : internalAtomicOperationInfo) +{ +SLANG_RAW("#line 9670 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" __generic<AtomicType, T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(AtomicType dest, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"atomic_") +SLANG_SPLICE(atomicOp.metalSuffix +) +SLANG_RAW("_explicit($0, $1, memory_order_relaxed)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<AtomicType, T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(metal)]\n") +SLANG_RAW(" void __metalInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(AtomicType dest, T value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"((*($2)) = (atomic_") +SLANG_SPLICE(atomicOp.metalSuffix +) +SLANG_RAW("_explicit($0, $1, memory_order_relaxed)))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda)]\n") +SLANG_RAW(" void __cudaInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"atomic") +SLANG_SPLICE(atomicOp.cudaSuffix +) +SLANG_RAW("((int*)$0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(cuda)]\n") +SLANG_RAW(" void __cudaInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"(*$2 = atomic") +SLANG_SPLICE(atomicOp.cudaSuffix +) +SLANG_RAW("((int*)$0, $1))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl)]\n") +SLANG_RAW(" void __glslInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"$atomic") +SLANG_SPLICE(atomicOp.glslSuffix +) +SLANG_RAW("($A, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl)]\n") +SLANG_RAW(" void __glslInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"($2 = $atomic") +SLANG_SPLICE(atomicOp.glslSuffix +) +SLANG_RAW("($A, $1))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl)]\n") +SLANG_RAW(" void __hlslInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"Interlocked") +SLANG_SPLICE(atomicOp.hlslSuffix +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(hlsl)]\n") +SLANG_RAW(" void __hlslInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" __intrinsic_asm \"Interlocked") +SLANG_SPLICE(atomicOp.hlslSuffix +) +SLANG_RAW("\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(spirv)]\n") +SLANG_RAW(" void __spirvInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvFloatSuffix +) +SLANG_RAW(" &dest Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvUIntSuffix +) +SLANG_RAW(" &dest Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvIntSuffix +) +SLANG_RAW(" &dest Device None $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __generic<T>\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(spirv)]\n") +SLANG_RAW(" void __spirvInterlocked_") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("(__ref T dest, T value, out T original_value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" static_assert(") +SLANG_SPLICE(atomicOp.assertExpr +) +SLANG_RAW(", \"Unable to use float with Atomic") +SLANG_SPLICE(atomicOp.slangSuffix +) +SLANG_RAW("\"); \n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %original:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvFloatSuffix +) +SLANG_RAW(" &dest Device None $value;\n") +SLANG_RAW(" OpStore &original_value %original\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %original:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvUIntSuffix +) +SLANG_RAW(" &dest Device None $value;\n") +SLANG_RAW(" OpStore &original_value %original\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %original:$$T = OpAtomic") +SLANG_SPLICE(atomicOp.spirvIntSuffix +) +SLANG_RAW(" &dest Device None $value;\n") +SLANG_RAW(" OpStore &original_value %original\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + +} // fetchAndModify +SLANG_RAW("#line 9806 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<AtomicType, T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("void __metalInterlocked_compare_exchange(AtomicType dest, __ref T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"atomic_compare_exchange_weak_explicit($0, $1, $2, memory_order_relaxed, memory_order_relaxed)\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<AtomicType, T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("void __metalInterlocked_compare_exchange(AtomicType dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(dest, compare_value, value);\n") +SLANG_RAW(" original_value = compare_value;\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cuda)]\n") +SLANG_RAW("void __cudaInterlocked_compare_exchange(__ref T dest, __ref T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"atomicCAS($0, $1, $2)\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cuda)]\n") +SLANG_RAW("void __cudaInterlocked_compare_exchange(__ref T dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"*$3 = atomicCAS($0, $1, $2)\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl)]\n") +SLANG_RAW("void __glslInterlocked_compare_exchange(__ref T dest, __ref T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"$atomicCompSwap($A, $1, $2)\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl)]\n") +SLANG_RAW("void __glslInterlocked_compare_exchange(__ref T dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"($3 = $atomicCompSwap($A, $1, $2))\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(hlsl)]\n") +SLANG_RAW("void __hlslInterlocked_compare_exchange(__ref T dest, __ref T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"InterlockedCompareExchange\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(hlsl)]\n") +SLANG_RAW("void __hlslInterlocked_compare_exchange(__ref T dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"InterlockedCompareExchange\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(spirv)]\n") +SLANG_RAW("void __spirvInterlocked_compare_exchange(__ref T dest, __ref T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %result:$$T = OpAtomicCompareExchange &dest Device None None $value $compare_value;\n") +SLANG_RAW(" };\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(spirv)]\n") +SLANG_RAW("void __spirvInterlocked_compare_exchange(__ref T dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %original:$$T = OpAtomicCompareExchange &dest Device None None $value $compare_value;\n") +SLANG_RAW(" OpStore &original_value %original\n") +SLANG_RAW(" };\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(hlsl)]\n") +SLANG_RAW("void __hlslInterlocked_compare_exchange_float_bitwise(__ref T dest, T compare_value, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"InterlockedCompareExchangeFloatBitwise\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(hlsl)]\n") +SLANG_RAW("void __hlslInterlocked_compare_exchange_float_bitwise(__ref T dest, T compare_value, T value, out T original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"InterlockedCompareExchangeFloatBitwise\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +// Generates code for: +// InterlockedAdd, InterlockedAnd, InterlockedOr, InterlockedXor, +// InterlockedMax, InterlockedMin, InterlockedExchange +struct SlangAtomicOperationInfo +{ + const char* slangCallSuffix; + const char* internalCallSuffix; +}; + +SlangAtomicOperationInfo slangAtomicOperationInfo[7] = { + { "Add", "add" }, + { "And", "and" }, + { "Or", "or" }, + { "Xor", "xor" }, + { "Max", "max" }, + { "Min", "min" }, + { "Exchange", "exchange" }, +}; + +for (SlangAtomicOperationInfo atomicOp : slangAtomicOperationInfo) +{ + for(const char* T : {"int", "uint"}) + { +SLANG_RAW("#line 9937 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void Interlocked") +SLANG_SPLICE(atomicOp.slangCallSuffix +) +SLANG_RAW("(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value);\n") +SLANG_RAW(" case glsl: __glslInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value);\n") +SLANG_RAW(" case spirv: __spirvInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vector<") +SLANG_SPLICE(T +) +SLANG_RAW(", 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vector<") +SLANG_SPLICE(T +) +SLANG_RAW(", 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__getMetalAtomicRef(dest), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void Interlocked") +SLANG_SPLICE(atomicOp.slangCallSuffix +) +SLANG_RAW("(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to a scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value, original_value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value, original_value);\n") +SLANG_RAW(" case glsl: __glslInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value, original_value);\n") +SLANG_RAW(" case spirv: __spirvInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(dest, value, original_value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vector<") +SLANG_SPLICE(T +) +SLANG_RAW(",4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vector<") +SLANG_SPLICE(T +) +SLANG_RAW(",4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" __metalInterlocked_") +SLANG_SPLICE(atomicOp.internalCallSuffix +) +SLANG_RAW("(__getMetalAtomicRef(dest), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + + } // for(const char* T : {"int64_t", "uint64_t"}) +SLANG_RAW("#line 10005 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void Interlocked") +SLANG_SPLICE(atomicOp.slangCallSuffix +) +SLANG_RAW("(__ref uint dest, int value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" Interlocked") +SLANG_SPLICE(atomicOp.slangCallSuffix +) +SLANG_RAW("(dest, (uint)value);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} // for (SlangAtomicOperationInfo atomicOp : slangAtomicOperationInfo) +SLANG_RAW("#line 10015 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") + +for(const char* T : {"int64_t", "uint64_t"}) +{ +SLANG_RAW("#line 10020 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void InterlockedAdd(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{ \n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_add(dest, value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_add(dest, value);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __requireGLSLExtension(\"GL_EXT_shader_atomic_int64\");\n") +SLANG_RAW(" __glslInterlocked_add(dest, value);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" result:$$") +SLANG_SPLICE(T +) +SLANG_RAW(" = OpAtomicIAdd &dest Device None $value;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedAdd(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_add(dest, value, original_value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_add(dest, value, original_value);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __requireGLSLExtension(\"GL_EXT_shader_atomic_int64\");\n") +SLANG_RAW(" __glslInterlocked_add(dest, value, original_value);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability Int64Atomics;\n") +SLANG_RAW(" %origin:$$") +SLANG_SPLICE(T +) +SLANG_RAW(" = OpAtomicIAdd &dest Device None $value;\n") +SLANG_RAW(" OpStore &original_value %origin\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedAnd(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_and(dest, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedAnd(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_and(dest, value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareExchange(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" compare_value, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange(dest, compare_value, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareExchange(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" compare_value, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareStore(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" compare_value, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value);\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedCompareStore\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedExchange(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedExchange\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedExchange(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedExchange\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedMax(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedMax\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedMax(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedMax\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedMin(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedMin\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedMin(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedMin\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedOr(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedOr\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedOr(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedOr\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedXor(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedXor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedXor(__ref ") +SLANG_SPLICE(T +) +SLANG_RAW(" dest, ") +SLANG_SPLICE(T +) +SLANG_RAW(" value, out ") +SLANG_SPLICE(T +) +SLANG_RAW(" original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedXor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} // for(const char* T : {"int64_t", "uint64_t"}) +SLANG_RAW("#line 10198 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void InterlockedCompareExchange(__ref int dest, int compare_value, int value, out int original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case glsl: __glslInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case spirv: __spirvInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" vector<int, 4> vec_compare_value = vector<int, 4>(compare_value);\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vec_compare_value, vector<int, 4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vec_compare_value, vector<int, 4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void InterlockedCompareExchange(__ref uint dest, uint compare_value, uint value, out uint original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case cuda: __cudaInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case glsl: __glslInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case spirv: __spirvInterlocked_compare_exchange(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" vector<uint, 4> vec_compare_value = vector<uint, 4>(compare_value);\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vec_compare_value, vector<uint, 4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vec_compare_value, vector<uint, 4>(value), original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value, original_value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareExchangeFloatBitwise(__ref float dest, float compare_value, float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange_float_bitwise(dest, compare_value, value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(!__isTextureAccess(dest), \"float atomic texture operations are disallowed with Metal target's\");\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareExchangeFloatBitwise(__ref float dest, float compare_value, float value, out float original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_compare_exchange_float_bitwise(dest, compare_value, value, original_value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(!__isTextureAccess(dest), \"float atomic texture operations are disallowed with Metal target's\");\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void InterlockedCompareStore(__ref int dest, int compare_value, int value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedCompareStore\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"$atomicCompSwap($A, $1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicCAS($0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpAtomicCompareExchange &dest Device None None $value $compare_value;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" vector<int, 4> vec_compare_value = vector<int, 4>(compare_value);\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vec_compare_value, vector<int, 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vec_compare_value, vector<int, 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__glsl_version(430)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_metal_spirv, atomic_glsl_hlsl_cuda_metal)]\n") +SLANG_RAW("void InterlockedCompareStore(__ref uint dest, uint compare_value, uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedCompareStore\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"$atomicCompSwap($A, $1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"atomicCAS((int*)$0, $1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpAtomicCompareExchange &dest Device None None $value $compare_value;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" if (__isTextureAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" vector<uint, 4> vec_compare_value = vector<uint, 4>(compare_value);\n") +SLANG_RAW(" if(__isTextureArrayAccess(dest))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), __extractArrayCoordFromTextureAccess(dest), vec_compare_value, vector<uint, 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalImageInterlocked_compare_exchange(__extractTextureFromTextureAccess(dest), \n") +SLANG_RAW(" __extractCoordFromTextureAccess(dest), vec_compare_value, vector<uint, 4>(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __metalInterlocked_compare_exchange(__getMetalAtomicRef(dest), compare_value, value);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedCompareStoreFloatBitwise(__ref float dest, float compare_value, float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InterlockedCompareStoreFloatBitwise\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedExchange(__ref float dest, float value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_exchange(dest, value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(!__isTextureAccess(dest), \"'float' atomic texture operations are disallowed with Metal target's\");\n") +SLANG_RAW(" __metalInterlocked_exchange(__getMetalAtomicRef(dest), value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("void InterlockedExchange(__ref float dest, float value, out float original_value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" static_assert(__isTextureScalarAccess(dest) || !__isTextureAccess(dest), \"Atomic must be applied to scalar texture or non-texture\");\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __hlslInterlocked_exchange(dest, value, original_value);\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" static_assert(!__isTextureAccess(dest), \"'float' atomic texture operations are disallowed with Metal target's\");\n") +SLANG_RAW(" __metalInterlocked_exchange(__getMetalAtomicRef(dest), value, original_value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Is floating-point value finite?\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("bool isfinite(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"isfinite\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_isfinite($0)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isfinite\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return !(isinf(x) || isnan(x));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<bool, N> isfinite(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"isfinite\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return !(isinf(x) || isnan(x));\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isfinite\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(bool, N, isfinite, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<bool, N, M> isfinite(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"isfinite\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(bool, N, M, isfinite, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Is floating-point value infinite?\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("bool isinf(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isinf\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_isinf($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm { result:$$bool = OpIsInf $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<bool, N> isinf(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isinf\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm { result:$$vector<bool,N> = OpIsInf $x};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(bool, N, isinf, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<bool, N, M> isinf(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"isinf\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(bool, N, M, isinf, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Is floating-point value not-a-number?\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("bool isnan(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isnan\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_isnan($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm { result:$$bool = OpIsNan $x};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<bool, N> isnan(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"isnan\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm { result:$$vector<bool, N> = OpIsNan $x};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(bool, N, isnan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<bool, N, M> isnan(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"isnan\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(bool, N, M, isnan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Construct float from mantissa and exponent\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T ldexp(T x, T exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * exp2(exp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> ldexp(vector<T, N> x, vector<T, N> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * exp2(exp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> ldexp(matrix<T, N, M> x, matrix<T, N, M> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, ldexp, x, exp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, E : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T ldexp(T x, E exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Ldexp $x $exp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return ldexp(x, __realCast<T>(exp));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, E : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> ldexp(vector<T, N> x, vector<E, N> exp)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Ldexp $x $exp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"ldexp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,N> temp;\n") +SLANG_RAW(" [ForceUnroll]\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" temp[i] = __realCast<T>(exp[i]);\n") +SLANG_RAW(" return ldexp(x, temp);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Vector length\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T length(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"length\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"length\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"length\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Length $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"length\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return sqrt(dot(x, x));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Scalar float length\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T length(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"length\"; \n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Length $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"length\"; \n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return abs(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Linear interpolation\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T lerp(T x, T y, T s)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"mix\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"mix\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"lerp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 FMix $x $y $s\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x + (y - x) * s;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> lerp(vector<T, N> x, vector<T, N> y, vector<T, N> s)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"mix\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"mix\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"lerp\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 FMix $x $y $s\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x + (y - x) * s;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> lerp(matrix<T,N,M> x, matrix<T,N,M> y, matrix<T,N,M> s)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"lerp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_TRINARY(T, N, M, lerp, x, y, s);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Legacy lighting function (obsolete)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("float4 lit(float n_dot_l, float n_dot_h, float m)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"lit\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let ambient = 1.0f;\n") +SLANG_RAW(" let diffuse = max(n_dot_l, 0.0f);\n") +SLANG_RAW(" let specular = step(0.0f, n_dot_l) * max(pow(n_dot_h, m), 0.0f);\n") +SLANG_RAW(" return float4(ambient, diffuse, specular, 1.0f);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-e logarithm\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T log(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_log($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_log($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Log $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> log(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Log $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, log, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> log(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, log, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-10 logarithm\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T log10(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log10\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log10\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(log( $0 ) * $S0( 0.43429448190325182765112891891661) )\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_log10($0)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_log10($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" const T tmp = T(0.43429448190325182765112891891661);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %baseElog:$$T = OpExtInst glsl450 Log $x;\n") +SLANG_RAW(" result:$$T = OpFMul %baseElog $tmp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> log10(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log10\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log10\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(log( $0 ) * $S0(0.43429448190325182765112891891661) )\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" const T tmp = T(0.43429448190325182765112891891661);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %baseElog:$$vector<T,N> = OpExtInst glsl450 Log $x;\n") +SLANG_RAW(" result:$$vector<T,N> = OpVectorTimesScalar %baseElog $tmp\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, log10, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> log10(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log10\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, log10, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Base-2 logarithm\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T log2(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_log2($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_log2($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Log2 $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> log2(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Log2 $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, log2, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> log2(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"log2\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, log2, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// multiply-add\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("T mad(T mvalue, T avalue, T bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Fma $mvalue $avalue $bvalue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<T, N> mad(vector<T, N> mvalue, vector<T, N> avalue, vector<T, N> bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Fma $mvalue $avalue $bvalue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, mad, mvalue, avalue, bvalue);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<T, N, M> mad(matrix<T, N, M> mvalue, matrix<T, N, M> avalue, matrix<T, N, M> bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_TRINARY(T, N, M, mad, mvalue, avalue, bvalue);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("T mad(T mvalue, T avalue, T bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_fma($0, $1, $2)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Fma $mvalue $avalue $bvalue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("vector<T, N> mad(vector<T, N> mvalue, vector<T, N> avalue, vector<T, N> bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"fma\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Fma $mvalue $avalue $bvalue\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, mad, mvalue, avalue, bvalue);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shader5_sm_5_0)]\n") +SLANG_RAW("matrix<T, N, M> mad(matrix<T, N, M> mvalue, matrix<T, N, M> avalue, matrix<T, N, M> bvalue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mad\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_TRINARY(T, N, M, mad, mvalue, avalue, bvalue);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// maximum\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T max(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" // Note: a stdlib implementation of `max` (or `min`) will require splitting\n") +SLANG_RAW(" // floating-point and integer cases apart, because the floating-point\n") +SLANG_RAW(" // version needs to correctly handle the case where one of the inputs\n") +SLANG_RAW(" // is not-a-number.\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_max($0, $1)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_max($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 SMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 UMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> max(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 SMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 UMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, max, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> max(matrix<T, N, M> x, matrix<T, N, M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, max, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T max(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_max($0, $1)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_max($0, $1)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 FMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> max(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T, N> = OpExtInst glsl450 FMax $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, max, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> max(matrix<T, N, M> x, matrix<T, N, M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"max\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, max, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T max3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return max(x, max(y, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> max3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"max3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return max(x, max(y, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fmax(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmax\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if (isnan(x)) return y;\n") +SLANG_RAW(" return max(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fmax(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmax\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, fmax, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fmax3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmax3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" bool isnanX = isnan(x);\n") +SLANG_RAW(" bool isnanY = isnan(y);\n") +SLANG_RAW(" bool isnanZ = isnan(z);\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isnanX)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return isnanY ? z : y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (isnanY)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isnanZ)\n") +SLANG_RAW(" return x;\n") +SLANG_RAW(" return max(x, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (isnanZ)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return max(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" return max(y, max(x, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fmax3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmax3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, fmax3, x, y, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// minimum\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T min(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"min\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_min($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 SMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 UMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> min(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 SMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 UMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, min, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> min(matrix<T,N,M> x, matrix<T,N,M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, min, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T min(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_min($0, $1)\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_min($0, $1)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 FMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> min(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 FMin $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, min, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> min(matrix<T,N,M> x, matrix<T,N,M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"min\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, min, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T min3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"min3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(x, min(y, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> min3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"min3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return min(x, min(y, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fmin(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if (isnan(x)) return y;\n") +SLANG_RAW(" return min(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fmin(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, fmin, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fmin3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmin3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" bool isnanX = isnan(x);\n") +SLANG_RAW(" bool isnanY = isnan(y);\n") +SLANG_RAW(" bool isnanZ = isnan(z);\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isnan(x))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return isnanY ? z : y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (isnanY)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" if (isnanZ)\n") +SLANG_RAW(" return x;\n") +SLANG_RAW(" return min(x, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (isnanZ)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return min(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" return min(x, min(y, z));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fmin3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmin3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, fmin3, x, y, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Median\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T median3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"median3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // | a | b | c | m |\n") +SLANG_RAW(" // ----------+---+---+---+---+\n") +SLANG_RAW(" // x > y > z | z | y | x | y |\n") +SLANG_RAW(" // x > z > y | y | z | x | z |\n") +SLANG_RAW(" // y > x > z | z | y | x | x |\n") +SLANG_RAW(" // y > z > x | z | y | z | z |\n") +SLANG_RAW(" // z > x > y | y | z | x | x |\n") +SLANG_RAW(" // z > y > x | y | z | y | y |\n") +SLANG_RAW("\n") +SLANG_RAW(" T a = min(y, z);\n") +SLANG_RAW(" T b = max(y, z);\n") +SLANG_RAW(" T c = max(x, a);\n") +SLANG_RAW(" T m = min(b, c);\n") +SLANG_RAW(" return m;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> median3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"median3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" vector<T,N> a = min(y, z);\n") +SLANG_RAW(" vector<T,N> b = max(y, z);\n") +SLANG_RAW(" vector<T,N> c = max(x, a);\n") +SLANG_RAW(" vector<T,N> m = min(b, c);\n") +SLANG_RAW(" return m;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T fmedian3(T x, T y, T z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmedian3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" bool isnanX = isnan(x);\n") +SLANG_RAW(" bool isnanY = isnan(y);\n") +SLANG_RAW(" bool isnanZ = isnan(z);\n") +SLANG_RAW("\n") +SLANG_RAW(" if (isnanX)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return isnanY ? z : y;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else if (isnanY || isnanZ)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // \"the function can return either non-NaN value\"\n") +SLANG_RAW(" return x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" return median3(x, y, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N: int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> fmedian3(vector<T,N> x, vector<T,N> y, vector<T,N> z)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"fmedian3\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, fmedian3, x, y, z);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// split into integer and fractional parts (both with same sign)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T modf(T x, out T ip)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_modf($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_modf($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"modf\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"modf\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"modf($0, *($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpExtInst glsl450 Modf $x &ip\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" T fract;\n") +SLANG_RAW(" __wgsl_modf<T>(x, fract, ip);\n") +SLANG_RAW(" return fract;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(wgsl)]\n") +SLANG_RAW("void __wgsl_modf(T x, out T fract, out T whole)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"{ var s = modf($0); ($1) = s.fract; ($2) = s.whole; }\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> modf(vector<T,N> x, out vector<T,N> ip)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"modf\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"modf\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"modf($0, *($1))\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$vector<T,N> = OpExtInst glsl450 Modf $x &ip\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" vector<T,N> fract;\n") +SLANG_RAW(" __wgsl_modf<T>(x, fract, ip);\n") +SLANG_RAW(" return fract;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, modf, x, ip);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(wgsl)]\n") +SLANG_RAW("void __wgsl_modf(vector<T,N> x, out vector<T,N> fract, out vector<T,N> whole)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_asm \"{ var s = modf($0); ($1) = s.fract; ($2) = s.whole; }\";\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int, let L : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> modf(matrix<T,N,M> x, out matrix<T,N,M,L> ip)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"modf\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, modf, x, ip);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// msad4 (whatever that is)\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, sm_4_0_version)]\n") +SLANG_RAW("uint4 msad4(uint reference, uint2 source, uint4 accum)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"msad4\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" int4 bytesRef = (reference >> uint4(24, 16, 8, 0)) & 0xFF;\n") +SLANG_RAW(" int4 bytesX = (source.x >> uint4(24, 16, 8, 0)) & 0xFF;\n") +SLANG_RAW(" int4 bytesY = (source.y >> uint4(24, 16, 8, 0)) & 0xFF;\n") +SLANG_RAW(" \n") +SLANG_RAW(" uint4 mask = select(bytesRef == 0, 0, 0xFFFFFFFFu);\n") +SLANG_RAW(" \n") +SLANG_RAW(" uint4 result = accum;\n") +SLANG_RAW(" result += mask.x & abs(bytesRef - int4(bytesX.x, bytesY.y, bytesY.z, bytesY.w));\n") +SLANG_RAW(" result += mask.y & abs(bytesRef - int4(bytesX.x, bytesX.y, bytesY.z, bytesY.w));\n") +SLANG_RAW(" result += mask.z & abs(bytesRef - int4(bytesX.x, bytesX.y, bytesX.z, bytesY.w));\n") +SLANG_RAW(" result += mask.w & abs(bytesRef - int4(bytesX.x, bytesX.y, bytesX.z, bytesX.w));\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// General inner products\n") +SLANG_RAW("\n") +SLANG_RAW("// scalar-scalar\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_Mul +) +SLANG_RAW(")\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T mul(T x, T y);\n") +SLANG_RAW("\n") +SLANG_RAW("// scalar-vector and vector-scalar\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_Mul +) +SLANG_RAW(")\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> mul(vector<T, N> x, T y);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_Mul +) +SLANG_RAW(")\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> mul(T x, vector<T, N> y);\n") +SLANG_RAW("\n") +SLANG_RAW("// scalar-matrix and matrix-scalar\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M :int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_Mul +) +SLANG_RAW(")\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> mul(matrix<T, N, M> x, T y);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M :int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_Mul +) +SLANG_RAW(")\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> mul(T x, matrix<T, N, M> y);\n") +SLANG_RAW("\n") +SLANG_RAW("// vector-vector (dot product)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T mul(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"dot\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return dot(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T mul(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return dot(x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// vector-matrix\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, M> mul(vector<T, N> left, matrix<T, N, M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpMatrixTimesVector $$vector<T, M> result $right $left\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,M> result;\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[i] * right[i][j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[j] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, M> mul(vector<T, N> left, matrix<T, N, M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,M> result;\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[i] * right[i][j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[j] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinLogicalType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, M> mul(vector<T, N> left, matrix<T, N, M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,M> result;\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum |= left[i] & right[i][j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[j] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// matrix-vector\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> mul(matrix<T,N,M> left, vector<T,M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpVectorTimesMatrix $$vector<T,N> result $right $left\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,N> result;\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[i][j] * right[j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[i] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> mul(matrix<T,N,M> left, vector<T,M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,N> result;\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[i][j] * right[j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[i] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinLogicalType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[OverloadRank(-1)]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> mul(matrix<T,N,M> left, vector<T,M> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" vector<T,N> result;\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int j = 0; j < M; ++j )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum |= left[i][j] & right[j];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[i] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// matrix-matrix\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let R : int, let N : int, let C : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,R,C> mul(matrix<T,R,N> left, matrix<T,N,C> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpMatrixTimesMatrix $$matrix<T,R,C> result $right $left\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,R,C> result;\n") +SLANG_RAW(" for( int r = 0; r < R; ++r)\n") +SLANG_RAW(" for( int c = 0; c < C; ++c)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[r][i] * right[i][c];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[r][c] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let R : int, let N : int, let C : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,R,C> mul(matrix<T,R,N> left, matrix<T,N,C> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,R,C> result;\n") +SLANG_RAW(" for( int r = 0; r < R; ++r)\n") +SLANG_RAW(" for( int c = 0; c < C; ++c)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum += left[r][i] * right[i][c];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[r][c] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinLogicalType, let R : int, let N : int, let C : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,R,C> mul(matrix<T,R,N> left, matrix<T,N,C> right)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"mul\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"($1 * $0)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,R,C> result;\n") +SLANG_RAW(" for( int r = 0; r < R; ++r)\n") +SLANG_RAW(" for( int c = 0; c < C; ++c)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T sum = T(0);\n") +SLANG_RAW(" for( int i = 0; i < N; ++i )\n") +SLANG_RAW(" {\n") +SLANG_RAW(" sum |= left[r][i] & right[i][c];\n") +SLANG_RAW(" }\n") +SLANG_RAW(" result[r][c] = sum;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// next-after: next representable floating-point value\n") +SLANG_RAW("// after x in the direction of y\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("T nextafter(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"nextafter\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" if (isnan(x)) return x;\n") +SLANG_RAW(" if (isnan(y)) return y;\n") +SLANG_RAW(" if (x == y) return y;\n") +SLANG_RAW(" if (T is half)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T delta = __realCast<T>(bit_cast<half>(uint16_t(1)));\n") +SLANG_RAW(" return x + ((x < y) ? delta : -delta);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" if (T is float)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T delta = __realCast<T>(bit_cast<float>(uint32_t(1)));\n") +SLANG_RAW(" return x + ((x < y) ? delta : -delta);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" T delta = __realCast<T>(bit_cast<double>(uint64_t(1)));\n") +SLANG_RAW(" return x + ((x < y) ? delta : -delta);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, shader5_sm_4_0)]\n") +SLANG_RAW("vector<T,N> nextafter(vector<T,N> x, vector<T,N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"nextafter\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, nextafter, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// noise (deprecated)\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[deprecated(\"Always returns 0\")]\n") +SLANG_RAW("float noise(float x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return 0;\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[deprecated(\"Always returns 0\")]\n") +SLANG_RAW("__generic<let N : int> float noise(vector<float, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return 0;\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("/// Indicate that an index may be non-uniform at execution time.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Shader Model 5.1 and 6.x introduce support for dynamic indexing\n") +SLANG_RAW("/// of arrays of resources, but place the restriction that *by default*\n") +SLANG_RAW("/// the implementation can assume that any value used as an index into\n") +SLANG_RAW("/// such arrays will be dynamically uniform across an entire `Draw` or `Dispatch`\n") +SLANG_RAW("/// (when using instancing, the value must be uniform across all instances;\n") +SLANG_RAW("/// it does not seem that the restriction extends to draws within a multi-draw).\n") +SLANG_RAW("///\n") +SLANG_RAW("/// In order to indicate to the implementation that it cannot make the\n") +SLANG_RAW("/// uniformity assumption, a shader programmer is required to pass the index\n") +SLANG_RAW("/// to the `NonUniformResourceIndex` function before using it as an index.\n") +SLANG_RAW("/// The function superficially acts like an identity function.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Note: a future version of Slang may take responsibility for inserting calls\n") +SLANG_RAW("/// to this function as necessary in output code, rather than make this\n") +SLANG_RAW("/// the user's responsibility, so that the default behavior of the language\n") +SLANG_RAW("/// is more semantically \"correct.\"\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(spirv)]\n") +SLANG_RAW("T __copyObject<T>(T v)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch {\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpCopyObject $v;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("/// `NonUniformResourceIndex` function is used to indicate if the resource index is\n") +SLANG_RAW("/// divergent, and ensure scalarization happens correctly for each divergent lane.\n") +SLANG_RAW("__generic<T:__BuiltinArithmeticType>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_NonUniformResourceIndex +) +SLANG_RAW(")\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, nonuniformqualifier)]\n") +SLANG_RAW("T NonUniformResourceIndex(T index);\n") +SLANG_RAW("\n") +SLANG_RAW("/// HLSL allows NonUniformResourceIndex around non int/uint types.\n") +SLANG_RAW("/// It's effect is presumably to ignore it, which the following implementation does.\n") +SLANG_RAW("/// We should also look to add a warning for this scenario.\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("[deprecated(\"NonUniformResourceIndex on a type other than uint/int is deprecated and has no effect\")]\n") +SLANG_RAW("T NonUniformResourceIndex<T>(T value) { return value; }\n") +SLANG_RAW("\n") +SLANG_RAW("// Normalize a vector\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> normalize(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Normalize $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x / length(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T normalize(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Normalize $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"normalize\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x / length(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Raise to a power\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T pow(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_pow($0, $1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_pow($0, $1)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Pow $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> pow(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Pow $x $y\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, pow, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> pow(matrix<T,N,M> x, matrix<T,N,M> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"pow\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, pow, x, y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T powr(T x, T y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"powr\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return pow(abs(x), y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> powr(vector<T, N> x, vector<T, N> y)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"powr\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return pow(abs(x), y);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Output message\n") +SLANG_RAW("// TODO: add check to ensure format is const literal.\n") +SLANG_RAW("\n") + +for (int argCount = 0; argCount < 12; argCount++) +{ + StringBuilder paramList; + StringBuilder argList; + StringBuilder spirvArgList; + StringBuilder genericParamList; + if (argCount > 0) + genericParamList << "<"; + for (int i = 0; i < argCount; i++) + { + if (i > 0) + genericParamList << ", "; + genericParamList << "T" << i; + + paramList << ", T" << i << " v" << i; + argList << ", $" << i+1; + spirvArgList << " $v" << i; + } + if (argCount > 0) + genericParamList << ">"; + auto params = paramList.toString(); + auto args = argList.toString(); + auto spirvArgs = spirvArgList.toString(); +SLANG_RAW("#line 12230 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_debug_printf)\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, printf)]\n") +SLANG_RAW("void printf") +SLANG_SPLICE(genericParamList.toString() +) +SLANG_RAW("(NativeString format ") +SLANG_SPLICE(paramList +) +SLANG_RAW(")\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"printf\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"debugPrintfEXT($0 ") +SLANG_SPLICE(argList +) +SLANG_RAW(")\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm {\n") +SLANG_RAW(" OpExtension \"SPV_KHR_non_semantic_info\";\n") +SLANG_RAW(" result:$$void = OpExtInst debugPrintf 1 $format ") +SLANG_SPLICE(spirvArgs +) +SLANG_RAW(";\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") + +} +SLANG_RAW("#line 12252 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Tessellation factor fixup routines\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void Process2DQuadTessFactorsAvg(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float2 InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void Process2DQuadTessFactorsMax(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float2 InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void Process2DQuadTessFactorsMin(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float2 InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessIsolineTessFactors(\n") +SLANG_RAW(" in float RawDetailFactor,\n") +SLANG_RAW(" in float RawDensityFactor,\n") +SLANG_RAW(" out float RoundedDetailFactor,\n") +SLANG_RAW(" out float RoundedDensityFactor);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessQuadTessFactorsAvg(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessQuadTessFactorsMax(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessQuadTessFactorsMin(\n") +SLANG_RAW(" in float4 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float4 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float2 RoundedInsideTessFactors,\n") +SLANG_RAW(" out float2 UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessTriTessFactorsAvg(\n") +SLANG_RAW(" in float3 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float3 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float RoundedInsideTessFactor,\n") +SLANG_RAW(" out float UnroundedInsideTessFactor);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessTriTessFactorsMax(\n") +SLANG_RAW(" in float3 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float3 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float RoundedInsideTessFactor,\n") +SLANG_RAW(" out float UnroundedInsideTessFactor);\n") +SLANG_RAW("\n") +SLANG_RAW("[require(hlsl, sm_5_0)]\n") +SLANG_RAW("void ProcessTriTessFactorsMin(\n") +SLANG_RAW(" in float3 RawEdgeFactors,\n") +SLANG_RAW(" in float InsideScale,\n") +SLANG_RAW(" out float3 RoundedEdgeTessFactors,\n") +SLANG_RAW(" out float RoundedInsideTessFactors,\n") +SLANG_RAW(" out float UnroundedInsideTessFactors);\n") +SLANG_RAW("\n") +SLANG_RAW("// Degrees to radians\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T radians(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Radians $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * (T.getPi() / T(180.0f));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> radians(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Radians $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * (T.getPi() / T(180.0f));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> radians(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"radians\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x * (T.getPi() / T(180.0f));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Approximate reciprocal\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T rcp(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rcp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return T(1.0) / x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> rcp(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rcp\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" case wgsl:\n") +SLANG_RAW(" return T(1.0) / x;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, rcp, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> rcp(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rcp\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, rcp, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reflect incident vector across plane with given normal\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T reflect(T i, T n)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Reflect $i $n\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return i - T(2) * dot(n,i) * n;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> reflect(vector<T,N> i, vector<T,N> n)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Reflect $i $n\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"reflect\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return i - T(2) * dot(n,i) * n;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Refract incident vector given surface normal and index of refraction\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> refract(vector<T,N> i, vector<T,N> n, T eta)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Refract $i $n $eta\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let dotNI = dot(n,i);\n") +SLANG_RAW(" let k = T(1) - eta*eta*(T(1) - dotNI * dotNI);\n") +SLANG_RAW(" if(k < T(0)) return vector<T,N>(T(0));\n") +SLANG_RAW(" return eta * i - (eta * dotNI + sqrt(k)) * n;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T refract(T i, T n, T eta)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Refract $i $n $eta\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"refract\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let dotNI = dot(n,i);\n") +SLANG_RAW(" let k = T(1) - eta*eta*(T(1) - dotNI * dotNI);\n") +SLANG_RAW(" if(k < T(0)) return T(0);\n") +SLANG_RAW(" return eta * i - (eta * dotNI + sqrt(k)) * n;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reverse order of bits\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("uint reversebits(uint value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"reversebits\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"bitfieldReverse\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_reversebits($0)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"reverse_bits\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpBitReverse $$uint result $value};\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"reverseBits\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, shader5_sm_5_0)]\n") +SLANG_RAW("vector<uint, N> reversebits(vector<uint, N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(uint, N, reversebits, value);\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"bitfieldReverse\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"reverse_bits\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpBitReverse $$vector<uint, N> result $value};\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"reverseBits\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// round even\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("T rint(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"roundEven\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"rint\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 RoundEven $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" T nearest = round(x);\n") +SLANG_RAW("\n") +SLANG_RAW(" // Check if the value is exactly halfway between two integers\n") +SLANG_RAW(" if (abs(x - nearest) == T(0.5))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // If halfway, choose the even number\n") +SLANG_RAW(" if ((nearest / T(2)) * T(2) != nearest)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // If the nearest number is odd,\n") +SLANG_RAW(" // move to the closest even number\n") +SLANG_RAW(" nearest -= ((x < nearest) ? T(1) : T(-1));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" return nearest;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N:int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> rint(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"roundEven\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"rint\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 RoundEven $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, rint, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Round-to-nearest\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T round(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_round($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_round($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Round $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> round(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"round\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Round $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, round, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> round(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"round\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, round, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Reciprocal of square root\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T rsqrt(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_rsqrt($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_rsqrt($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"inversesqrt($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rsqrt\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"rsqrt\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 InverseSqrt $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return T(1.0) / sqrt(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> rsqrt(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"inversesqrt($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rsqrt\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"rsqrt\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 InverseSqrt $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, rsqrt, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> rsqrt(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"rsqrt\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, rsqrt, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Clamp value to [0,1] range\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T saturate(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return clamp<T>(x, T(0), T(1));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> saturate(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return clamp<T,N>(x,\n") +SLANG_RAW(" vector<T,N>(T(0)),\n") +SLANG_RAW(" vector<T,N>(T(1)));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> saturate(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"saturate\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, saturate, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T:__BuiltinArithmeticType, U:__BuiltinArithmeticType>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_IntCast +) +SLANG_RAW(")\n") +SLANG_RAW("T __int_cast(U val);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T:__BuiltinArithmeticType, U:__BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_IntCast +) +SLANG_RAW(")\n") +SLANG_RAW("vector<T,N> __int_cast(vector<U,N> val);\n") +SLANG_RAW("\n") +SLANG_RAW("// Extract sign of value\n") +SLANG_RAW("__generic<T : __BuiltinSignedArithmeticType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("int sign(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sign\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"int(sign($0))\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"int(sign($0))\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"$P_sign($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %fsign:$$T = OpExtInst glsl450 FSign $x;\n") +SLANG_RAW(" result:$$int = OpConvertFToS %fsign\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return __int_cast<int>(spirv_asm {OpExtInst $$T result glsl450 SSign $x});\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sign\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinSignedArithmeticType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("vector<int, N> sign(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" if(N == 1)\n") +SLANG_RAW(" return vector<int, N>(sign(x[0]));\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sign\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ivec$N0(sign($0))\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"vec<int,$N0>(sign($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %fsign:$$vector<T, N> = OpExtInst glsl450 FSign $x;\n") +SLANG_RAW(" result:$$vector<int, N> = OpConvertFToS %fsign\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return __int_cast<int>(spirv_asm {OpExtInst $$vector<T,N> result glsl450 SSign $x});\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sign\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(int, N, sign, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinSignedArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<int, N, M> sign(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sign\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(int, N, M, sign, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Sine\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T sin(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_sin($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_sin($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Sin $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> sin(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Sin $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, sin, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> sin(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sin\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, sin, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Sine and cosine\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("T __sincos_metal(T x, out T c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sincos($0, *$1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(metal)]\n") +SLANG_RAW("vector<T,N> __sincos_metal(vector<T,N> x, out vector<T,N> c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sincos($0, *$1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("void sincos(T x, out T s, out T c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_sincos($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sincos\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" //__intrinsic_asm \"*($1) = sincos($0, *($2))\";\n") +SLANG_RAW(" s = __sincos_metal(x, c);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" s = sin(x);\n") +SLANG_RAW(" c = cos(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("void sincos(vector<T,N> x, out vector<T,N> s, out vector<T,N> c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sincos\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" //__intrinsic_asm \"*($1) = sincos($0, *($2))\";\n") +SLANG_RAW(" s = __sincos_metal(x, c);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" s = sin(x);\n") +SLANG_RAW(" c = cos(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int, let L1: int, let L2 : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("void sincos(matrix<T,N,M> x, out matrix<T,N,M,L1> s, out matrix<T,N,M,L2> c)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sincos\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" s = sin(x);\n") +SLANG_RAW(" c = cos(x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Hyperbolic Sine\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T sinh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_sinh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_sinh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Sinh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> sinh(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Sinh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, sinh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> sinh(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sinh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, sinh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Sine degree\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T sinpi(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sinpi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return sin(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> sinpi(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sinpi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return sin(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Smooth step (Hermite interpolation)\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T smoothstep(T min, T max, T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 SmoothStep $min $max $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" let t = saturate((x - min) / (max - min));\n") +SLANG_RAW(" return t * t * (T(3.0f) - (t + t));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> smoothstep(vector<T, N> min, vector<T, N> max, vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 SmoothStep $min $max $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_TRINARY(T, N, smoothstep, min, max, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> smoothstep(matrix<T, N, M> min, matrix<T, N, M> max, matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"smoothstep\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_TRINARY(T, N, M, smoothstep, min, max, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Square root\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T sqrt(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_sqrt($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_sqrt($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Sqrt $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> sqrt(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Sqrt $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, sqrt, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> sqrt(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"sqrt\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, sqrt, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Step function\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T step(T y, T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Step $y $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return x < y ? T(0.0f) : T(1.0f);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> step(vector<T,N> y, vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"step\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Step $y $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_BINARY(T, N, step, y, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> step(matrix<T, N, M> y, matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"step\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_BINARY(T, N, M, step, y, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Tangent\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T tan(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_tan($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_tan($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Tan $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> tan(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Tan $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, tan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> tan(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tan\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, tan, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Hyperbolic tangent\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T tanh(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_tanh($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_tanh($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Tanh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> tanh(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T,N> result glsl450 Tanh $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, tanh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T,N,M> tanh(matrix<T,N,M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"tanh\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, tanh, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Tangent degree\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T tanpi(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tanpi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return tan(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T,N> tanpi(vector<T,N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case metal: __intrinsic_asm \"tanpi\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return tan(T.getPi() * x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Matrix transpose\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("[PreferRecompute]\n") +SLANG_RAW("matrix<T, M, N> transpose(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpTranspose $$matrix<T, M, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T,M,N> result;\n") +SLANG_RAW(" for(int r = 0; r < M; ++r)\n") +SLANG_RAW(" for(int c = 0; c < N; ++c)\n") +SLANG_RAW(" result[r][c] = x[c][r];\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("[PreferRecompute]\n") +SLANG_RAW("matrix<T, M, N> transpose(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpTranspose $$matrix<T, M, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T, M, N> result;\n") +SLANG_RAW(" for (int r = 0; r < M; ++r)\n") +SLANG_RAW(" for (int c = 0; c < N; ++c)\n") +SLANG_RAW(" result[r][c] = x[c][r];\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinLogicalType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("[PreferRecompute]\n") +SLANG_RAW("[OverloadRank(-1)]\n") +SLANG_RAW("matrix<T, M, N> transpose(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpTranspose $$matrix<T, M, N> result $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"transpose\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" matrix<T, M, N> result;\n") +SLANG_RAW(" for (int r = 0; r < M; ++r)\n") +SLANG_RAW(" for (int c = 0; c < N; ++c)\n") +SLANG_RAW(" result[r][c] = x[c][r];\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Truncate to integer\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("T trunc(T x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"$P_trunc($0)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"$P_trunc($0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$T result glsl450 Trunc $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("vector<T, N> trunc(vector<T, N> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" OpExtInst $$vector<T, N> result glsl450 Trunc $x\n") +SLANG_RAW(" };\n") +SLANG_RAW(" case wgsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" VECTOR_MAP_UNARY(T, N, trunc, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinFloatingPointType, let N : int, let M : int>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_metal_spirv_wgsl, sm_4_0_version)]\n") +SLANG_RAW("matrix<T, N, M> trunc(matrix<T, N, M> x)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"trunc\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" MATRIX_MAP_UNARY(T, N, M, trunc, x);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Slang Specific 'Mask' Wave Intrinsics\n") +SLANG_RAW("\n") +SLANG_RAW("typedef uint WaveMask;\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("WaveMask WaveGetConvergedMask()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallot(true).x\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveBallot(true).x\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__activemask()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let _true = true;\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallot $$uint4 result Subgroup $_true\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_WaveGetActiveMask +) +SLANG_RAW(")\n") +SLANG_RAW("WaveMask __WaveGetActiveMask();\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot_activemask)]\n") +SLANG_RAW("WaveMask WaveGetActiveMask()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallot(true).x\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveBallot(true).x\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let _true = true;\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallot $$uint4 result Subgroup $_true\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" return __WaveGetActiveMask();\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_basic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic)]\n") +SLANG_RAW("bool WaveMaskIsFirstLane(WaveMask mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupElect()\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"(($0 & -$0) == (WarpMask(1) << _getLaneId()))\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveIsFirstLane()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformElect $$bool result Subgroup\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveMaskAllTrue(WaveMask mask, bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAll($1)\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"(__all_sync($0, $1) != 0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllTrue($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformAll $$bool result Subgroup $condition\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveMaskAnyTrue(WaveMask mask, bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAny($1)\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"(__any_sync($0, $1) != 0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAnyTrue($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformAny $$bool result Subgroup $condition\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("WaveMask WaveMaskBallot(WaveMask mask, bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallot($1).x\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__ballot_sync($0, $1)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveBallot($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallot $$uint4 result Subgroup $condition\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic_ballot)]\n") +SLANG_RAW("uint WaveMaskCountBits(WaveMask mask, bool value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__popc(__ballot_sync($0, $1))\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveCountBits($1)\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return _WaveCountBits(WaveActiveBallot(value));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Waits until all warp lanes named in mask have executed a WaveMaskSharedSync (with the same mask)\n") +SLANG_RAW("// before resuming execution. Guarantees memory ordering in shared memory among threads participating\n") +SLANG_RAW("// in the barrier.\n") +SLANG_RAW("//\n") +SLANG_RAW("// The CUDA intrinsic says it orders *all* memory accesses, which appears to match most closely subgroupBarrier.\n") +SLANG_RAW("//\n") +SLANG_RAW("// TODO(JS):\n") +SLANG_RAW("// For HLSL it's not clear what to do. There is no explicit mechanism to 'reconverge' threads. In the docs it describes\n") +SLANG_RAW("// behavior as\n") +SLANG_RAW("// \"These intrinsics are dependent on active lanes and therefore flow control. In the model of this document, implementations\n") +SLANG_RAW("// must enforce that the number of active lanes exactly corresponds to the programmer429496726642949671684294967193s view of flow control.\"\n") +SLANG_RAW("//\n") +SLANG_RAW("// It seems this can only mean the active threads are the \"threads the program flow would lead to\". This implies a lockstep\n") +SLANG_RAW("// \"straight SIMD\" style interpretation. That being the case this op on HLSL is just a memory barrier without any Sync.\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, memorybarrier)]\n") +SLANG_RAW("void AllMemoryBarrierWithWaveMaskSync(WaveMask mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__syncwarp($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"AllMemoryBarrier()\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" __subgroupBarrier();\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// On GLSL, it appears we can't use subgroupMemoryBarrierShared, because it only implies a memory ordering, it does not\n") +SLANG_RAW("// imply convergence. For subgroupBarrier we have from the docs..\n") +SLANG_RAW("// \"The function subgroupBarrier() enforces that all active invocations within a subgroup must execute this function before any\n") +SLANG_RAW("// are allowed to continue their execution\"\n") +SLANG_RAW("// TODO(JS):\n") +SLANG_RAW("// It's not entirely clear what to do here on HLSL.\n") +SLANG_RAW("// Reading the dxc wiki (https://github.com/Microsoft/DirectXShaderCompiler/wiki/Wave-Intrinsics), we have statements like:\n") +SLANG_RAW("// ... these intrinsics enable the elimination of barrier constructs when the scope of synchronization is within the width of the SIMD processor.\n") +SLANG_RAW("// Wave: A set of lanes executed simultaneously in the processor. No explicit barriers are required to guarantee that they execute in parallel.\n") +SLANG_RAW("// Which seems to imply at least some memory barriers like Shared might not be needed.\n") +SLANG_RAW("//\n") +SLANG_RAW("// The barrier is left here though, because not only is the barrier make writes before the barrier across the wave appear to others afterwards, it's\n") +SLANG_RAW("// also there to inform the compiler on what order reads and writes can take place. This might seem to be silly because of the 'Active' lanes\n") +SLANG_RAW("// aspect of HLSL seems to make everything in lock step - but that's not quite so, it only has to apparently be that way as far as the programmers\n") +SLANG_RAW("// model appears - divergence could perhaps potentially still happen.\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, memorybarrier)]\n") +SLANG_RAW("void GroupMemoryBarrierWithWaveMaskSync(WaveMask mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__syncwarp($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"GroupMemoryBarrier()\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" __subgroupBarrier();\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, memorybarrier)]\n") +SLANG_RAW("void AllMemoryBarrierWithWaveSync()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__syncwarp()\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"AllMemoryBarrier()\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" __subgroupBarrier();\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, memorybarrier)]\n") +SLANG_RAW("void GroupMemoryBarrierWithWaveSync()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"__syncwarp()\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"GroupMemoryBarrier()\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" __subgroupBarrier();\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE! WaveMaskBroadcastLaneAt is *NOT* standard HLSL\n") +SLANG_RAW("// It is provided as access to subgroupBroadcast which can only take a\n") +SLANG_RAW("// constexpr laneId.\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/khr/GL_KHR_shader_subgroup.txt\n") +SLANG_RAW("// Versions SPIR-V greater than 1.4 loosen this restriction, and allow 'dynamic uniform' index\n") +SLANG_RAW("// If that's the behavior required then client code should use WaveReadLaneAt which works this way.\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("T WaveMaskBroadcastLaneAt(WaveMask mask, T value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBroadcast($1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__shfl_sync($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBroadcast $$T result Subgroup $value $ulane;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("vector<T,N> WaveMaskBroadcastLaneAt(WaveMask mask, vector<T,N> value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBroadcast($1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBroadcast $$vector<T,N> result Subgroup $value $ulane;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_ballot)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskBroadcastLaneAt(WaveMask mask, matrix<T,N,M> value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO(JS): If it can be determines that the `laneId` is constExpr, then subgroupBroadcast\n") +SLANG_RAW("// could be used on GLSL. For now we just use subgroupShuffle\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("T WaveMaskReadLaneAt(WaveMask mask, T value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupShuffle($1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__shfl_sync($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformShuffle;\n") +SLANG_RAW(" OpGroupNonUniformShuffle $$T result Subgroup $value $ulane;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("vector<T,N> WaveMaskReadLaneAt(WaveMask mask, vector<T,N> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupShuffle($1, $2)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformShuffle;\n") +SLANG_RAW(" OpGroupNonUniformShuffle $$vector<T,N> result Subgroup $value $ulane;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_shuffle)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskReadLaneAt(WaveMask mask, matrix<T,N,M> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt($1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE! WaveMaskShuffle is a NON STANDARD HLSL intrinsic! It will map to WaveReadLaneAt on HLSL\n") +SLANG_RAW("// which means it will only work on hardware which allows arbitrary laneIds which is not true\n") +SLANG_RAW("// in general because it breaks the HLSL standard, which requires it's 'dynamically uniform' across the Wave.\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("T WaveMaskShuffle(WaveMask mask, T value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveMaskReadLaneAt(mask, value, lane);\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("vector<T,N> WaveMaskShuffle(WaveMask mask, vector<T,N> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveMaskReadLaneAt(mask, value, lane);\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskShuffle(WaveMask mask, matrix<T,N,M> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveMaskReadLaneAt(mask, value, lane);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("uint WaveMaskPrefixCountBits(WaveMask mask, bool value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBallotExclusiveBitCount(subgroupBallot($1))\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__popc(__ballot_sync($0, $1) & _getLaneLtMask())\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixCountBits($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" %mask:$$uint4 = OpGroupNonUniformBallot Subgroup $value;\n") +SLANG_RAW(" OpGroupNonUniformBallotBitCount $$uint result Subgroup 2 %mask\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Across lane ops\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskBitAnd(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupAnd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveAnd($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitAnd($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseAnd $$T result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskBitAnd(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupAnd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveAndMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitAnd($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseAnd $$vector<T,N> result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskBitAnd(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveAndMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitAnd($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskBitOr(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupOr($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveOr($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitOr($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseOr $$T result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskBitOr(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupOr($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveOrMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitOr($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseOr $$vector<T,N> result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskBitOr(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveOrMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitOr($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskBitXor(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupXor($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveXor($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitXor($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseXor $$T result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskBitXor(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupXor($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveXorMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitXor($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformBitwiseXor $$vector<T,N> result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskBitXor(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveXorMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBitXor($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskMax(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMax($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMax($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMax($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMax $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformSMax $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformUMax $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskMax(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMax($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMaxMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMax($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMax $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformSMax $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformUMax $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskMax(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMaxMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMax($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskMin(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMin($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMin($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMin($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMin $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformSMin $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformUMin $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskMin(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMin($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMinMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMin($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMin $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isSignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformSMin $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformUMin $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskMin(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMinMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveMin($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskProduct(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMul($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveProduct($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveProduct($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMul $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$T result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskProduct(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupMul($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveProductMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveProduct($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMul $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$vector<T,N> result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskProduct(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveProductMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveProduct($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskSum(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAdd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveSum($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveSum($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$T result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIAdd $$T result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskSum(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAdd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveSumMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveSum($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$vector<T,N> result Subgroup 0 $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic; \n") +SLANG_RAW(" OpGroupNonUniformIAdd $$vector<T,N> result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskSum(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveSumMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveSum($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveMaskAllEqual(WaveMask mask, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAllEqual($1)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllEqual($1)\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"_waveAllEqual($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAllEqual $$bool result Subgroup $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveMaskAllEqual(WaveMask mask, vector<T,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAllEqual($1)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllEqual($1)\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"_waveAllEqualMultiple($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAllEqual $$bool result Subgroup $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_vote)]\n") +SLANG_RAW("bool WaveMaskAllEqual(WaveMask mask, matrix<T,N,M> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveAllEqualMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveAllEqual($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Prefix\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskPrefixProduct(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveMul($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProduct($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMul $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$T result Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskPrefixProduct(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveMul($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProductMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMul $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$vector<T,N> result Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskPrefixProduct(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProductMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskPrefixSum(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveAdd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSum($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformIAdd Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskPrefixSum(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveAdd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSumMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformIAdd Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskPrefixSum(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSumMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum($1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("T WaveMaskReadLaneFirst(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBroadcastFirst($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveReadFirst($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneFirst($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcastFirst $$T result Subgroup $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("vector<T,N> WaveMaskReadLaneFirst(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupBroadcastFirst($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveReadFirstMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneFirst($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcastFirst $$vector<T,N> result Subgroup $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda, subgroup_ballot)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskReadLaneFirst(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveReadFirstMultiple($0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// WaveMask SM6.5 like intrinsics\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO(JS): On HLSL it only works for 32 bits or less\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_subgroup_partitioned)\n") +SLANG_RAW("__spirv_version(1.1)\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_partitioned)]\n") +SLANG_RAW("WaveMask WaveMaskMatch(WaveMask mask, T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupPartitionNV($1).x\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMatchScalar($0, $1).x\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch($1).x\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformPartitionedNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_subgroup_partitioned\";\n") +SLANG_RAW(" OpGroupNonUniformPartitionNV $$uint4 result $value\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_subgroup_partitioned)\n") +SLANG_RAW("__spirv_version(1.1)\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_partitioned)]\n") +SLANG_RAW("WaveMask WaveMaskMatch(WaveMask mask, vector<T,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupPartitionNV($1).x\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMatchMultiple($0, $1).x\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch($1).x\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return (spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformPartitionedNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_subgroup_partitioned\";\n") +SLANG_RAW(" OpGroupNonUniformPartitionNV $$uint4 result $value\n") +SLANG_RAW(" }).x;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_subgroup_partitioned)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("__cuda_sm_version(7.0)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, subgroup_partitioned)]\n") +SLANG_RAW("WaveMask WaveMaskMatch(WaveMask mask, matrix<T,N,M> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveMatchMultiple($0, $1)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupPartitionNV($1).x\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch($1).x\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskPrefixBitAnd(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveAnd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAnd($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseAnd $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskPrefixBitAnd(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveAnd($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAndMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseAnd $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskPrefixBitAnd(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAndMultiple(_getMultiPrefixMask($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskPrefixBitOr(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveOr($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOr($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseAnd $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskPrefixBitOr(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveOr($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOrMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseOr $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskPrefixBitOr(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOrMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveMaskPrefixBitXor(WaveMask mask, T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveXor($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXor($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseXor $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveMaskPrefixBitXor(WaveMask mask, vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveXor($1)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXorMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformBitwiseXor $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WaveMaskPrefixBitXor(WaveMask mask, matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXorMultiple($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor($1, uint4($0, 0, 0, 0))\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Shader model 6.0 stuff\n") +SLANG_RAW("\n") +SLANG_RAW("// Information for GLSL wave/subgroup support\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/khr/GL_KHR_shader_subgroup.txt\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("T QuadReadLaneAt(T sourceValue, uint quadLaneID)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadLaneAt\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadBroadcast\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformQuadBroadcast Subgroup $sourceValue $quadLaneID;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("vector<T,N> QuadReadLaneAt(vector<T,N> sourceValue, uint quadLaneID)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadLaneAt\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadBroadcast\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformQuadBroadcast Subgroup $sourceValue $quadLaneID;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int> matrix<T,N,M> QuadReadLaneAt(matrix<T,N,M> sourceValue, uint quadLaneID);\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("T QuadReadAcrossX(T localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossX\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapHorizontal($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 0u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("vector<T,N> QuadReadAcrossX(vector<T,N> localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossX\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapHorizontal($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 0u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int> matrix<T,N,M> QuadReadAcrossX(matrix<T,N,M> localValue);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("T QuadReadAcrossY(T localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossY\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapVertical($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 1u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("vector<T,N> QuadReadAcrossY(vector<T,N> localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossY\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapVertical($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 1u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int> matrix<T,N,M> QuadReadAcrossY(matrix<T,N,M> localValue);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("T QuadReadAcrossDiagonal(T localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossDiagonal\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapDiagonal($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 2u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_quad)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subgroup_quad)]\n") +SLANG_RAW("vector<T,N> QuadReadAcrossDiagonal(vector<T,N> localValue)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"QuadReadAcrossDiagonal\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupQuadSwapDiagonal($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint direction = 2u;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformQuad;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformQuadSwap Subgroup $localValue $direction;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int> matrix<T,N,M> QuadReadAcrossDiagonal(matrix<T,N,M> localValue);\n") +SLANG_RAW("\n") +SLANG_RAW("// WaveActiveBitAnd, WaveActiveBitOr, WaveActiveBitXor\n") + +struct WaveActiveBitOpEntry { const char* hlslName; const char* glslName; const char* spirvName; }; +const WaveActiveBitOpEntry kWaveActiveBitOpEntries[] = {{"BitAnd", "And", "BitwiseAnd"}, {"BitOr", "Or", "BitwiseOr"}, {"BitXor", "Xor", "BitwiseXor"}}; +for (auto opName : kWaveActiveBitOpEntries) { +SLANG_RAW("#line 14835 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniform") +SLANG_SPLICE(opName.spirvName +) +SLANG_RAW(" $$T result Subgroup Reduce $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T, N> WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(vector<T, N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniform") +SLANG_SPLICE(opName.spirvName +) +SLANG_RAW(" $$vector<T, N> result Subgroup Reduce $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinIntegerType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T, N, M> WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(matrix<T, N, M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T,N,M> result;\n") +SLANG_RAW(" [ForceUnroll]\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") + +} // WaveActiveBitAnd, WaveActiveBitOr, WaveActiveBitXor +SLANG_RAW("#line 14891 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// WaveActiveMin/Max\n") + +const char* kWaveActiveMinMaxNames[] = {"Min", "Max"}; +for (const char* opName : kWaveActiveMinMaxNames) { +SLANG_RAW("#line 14897 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformF") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$T result Subgroup Reduce $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformU") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$T result Subgroup Reduce $expr};\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformS") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$T result Subgroup Reduce $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T, N> WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("(vector<T, N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformF") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$vector<T, N> result Subgroup Reduce $expr};\n") +SLANG_RAW(" else if (__isUnsignedInt<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformU") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$vector<T, N> result Subgroup Reduce $expr};\n") +SLANG_RAW(" else\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformS") +SLANG_SPLICE(opName +) +SLANG_RAW(" $$vector<T, N> result Subgroup Reduce $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T, N, M> WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("(matrix<T, N, M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" [ForceUnroll]\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveActive") +SLANG_SPLICE(opName +) +SLANG_RAW("(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} // WaveActiveMinMax. +SLANG_RAW("#line 14964 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// WaveActiveProduct/Sum\n") + +struct WaveActiveProductSumEntry { const char* hlslName; const char* glslName; }; +const WaveActiveProductSumEntry kWaveActivProductSumNames[] = {{"Product", "Mul"}, {"Sum", "Add"}}; +for (auto opName : kWaveActivProductSumNames) { +SLANG_RAW("#line 14971 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformF") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW(" $$T result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformI") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW(" $$T result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroup") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW("($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformF") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW(" $$vector<T,N> result Subgroup 0 $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformI") +SLANG_SPLICE(opName.glslName +) +SLANG_RAW(" $$vector<T,N> result Subgroup 0 $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T, N, M> WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(matrix<T, N, M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" [ForceUnroll]\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveActive") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMask") +SLANG_SPLICE(opName.hlslName +) +SLANG_RAW("(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") + +} // WaveActiveProduct/WaveActiveProductSum. +SLANG_RAW("#line 15057 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveActiveAllEqual(T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAllEqual($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllEqual\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAllEqual $$bool result Subgroup $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskAllEqual(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveActiveAllEqual(vector<T,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAllEqual($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllEqual\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAllEqual $$bool result Subgroup $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskAllEqual(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_vote)]\n") +SLANG_RAW("bool WaveActiveAllEqual(matrix<T, N, M> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveAllEqual\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskAllEqual(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveActiveAllTrue(bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAll($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAllTrue($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAll $$bool result Subgroup $condition\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskAllTrue(WaveGetActiveMask(), condition);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_vote)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_vote)]\n") +SLANG_RAW("bool WaveActiveAnyTrue(bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupAny($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveAnyTrue($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformVote;\n") +SLANG_RAW(" OpGroupNonUniformAny $$bool result Subgroup $condition\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskAnyTrue(WaveGetActiveMask(), condition);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("uint4 WaveActiveBallot(bool condition)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallot($0)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveActiveBallot\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallot $$uint4 result Subgroup $condition\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskBallot(WaveGetActiveMask(), condition);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic_ballot)]\n") +SLANG_RAW("uint WaveActiveCountBits(bool value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveCountBits\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return _WaveCountBits(WaveActiveBallot(value));\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskCountBits(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_basic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic)]\n") +SLANG_RAW("uint WaveGetLaneCount()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_SubgroupSize)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(warpSize)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveGetLaneCount()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniform;\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(SubgroupSize:uint)\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_basic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic)]\n") +SLANG_RAW("uint WaveGetLaneIndex()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_SubgroupInvocationID)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_getLaneId()\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveGetLaneIndex()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniform;\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(SubgroupLocalInvocationId:uint)\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_basic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_basic)]\n") +SLANG_RAW("bool WaveIsFirstLane()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupElect()\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"WaveIsFirstLane()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformElect $$bool result Subgroup\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskIsFirstLane(WaveGetActiveMask());\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// It's useful to have a wave uint4 version of countbits, because some wave functions return uint4.\n") +SLANG_RAW("// This implementation tries to limit the amount of work required by the actual lane count.\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, subgroup_basic_ballot)]\n") +SLANG_RAW("uint _WaveCountBits(uint4 value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallotBitCount $$uint result Subgroup Reduce $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" // Assume since WaveGetLaneCount should be known at compile time, the branches will hopefully boil away\n") +SLANG_RAW(" const uint waveLaneCount = WaveGetLaneCount();\n") +SLANG_RAW(" switch ((waveLaneCount - 1) / 32)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" case 0: return countbits(value.x);\n") +SLANG_RAW(" case 1: return countbits(value.x) + countbits(value.y);\n") +SLANG_RAW(" case 2: return countbits(value.x) + countbits(value.y) + countbits(value.z);\n") +SLANG_RAW(" case 3: return countbits(value.x) + countbits(value.y) + countbits(value.z) + countbits(value.w);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// Prefix\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WavePrefixProduct(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveMul($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformFMul $$T result Subgroup ExclusiveScan $expr\n") +SLANG_RAW(" };\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$T result Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixProduct(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WavePrefixProduct(vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveMul($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFMul $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" OpGroupNonUniformIMul $$vector<T,N> result Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixProduct(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T, N, M> WavePrefixProduct(matrix<T, N, M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixProduct\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WavePrefixProduct(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixProduct(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("T WavePrefixSum(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveAdd($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$T result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" result:$$T = OpGroupNonUniformIAdd Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixSum(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("vector<T,N> WavePrefixSum(vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupExclusiveAdd($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" if (__isFloat<T>())\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformArithmetic; OpGroupNonUniformFAdd $$vector<T,N> result Subgroup ExclusiveScan $expr};\n") +SLANG_RAW(" else if (__isInt<T>())\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformArithmetic;\n") +SLANG_RAW(" result:$$vector<T,N> = OpGroupNonUniformIAdd Subgroup ExclusiveScan $expr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else return expr;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixSum(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_arithmetic)]\n") +SLANG_RAW("matrix<T,N,M> WavePrefixSum(matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixSum\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WavePrefixSum(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixSum(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("T WaveReadLaneFirst(T expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBroadcastFirst($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneFirst\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcastFirst $$T result Subgroup $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneFirst(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("vector<T,N> WaveReadLaneFirst(vector<T,N> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBroadcastFirst($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneFirst\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcastFirst $$vector<T,N> result Subgroup $expr};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneFirst(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("matrix<T,N,M> WaveReadLaneFirst(matrix<T,N,M> expr)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneFirst\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveReadLaneFirst(expr[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneFirst(WaveGetActiveMask(), expr);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE! WaveBroadcastLaneAt is *NOT* standard HLSL\n") +SLANG_RAW("// It is provided as access to subgroupBroadcast which can only take a\n") +SLANG_RAW("// constexpr laneId.\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/khr/GL_KHR_shader_subgroup.txt\n") +SLANG_RAW("// Versions SPIR-V greater than 1.4 loosen this restriction, and allow 'dynamic uniform' index\n") +SLANG_RAW("// If that's the behavior required then client code should use WaveReadLaneAt which works this way.\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("T WaveBroadcastLaneAt(T value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBroadcast($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcast $$T result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskBroadcastLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("vector<T,N> WaveBroadcastLaneAt(vector<T,N> value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBroadcast($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformBallot; OpGroupNonUniformBroadcast $$vector<T,N> result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskBroadcastLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("matrix<T, N, M> WaveBroadcastLaneAt(matrix<T, N, M> value, constexpr int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple(_getActiveMask(), $0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T, N, M> result;\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveBroadcastLaneAt(value[i], lane);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskBroadcastLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO(JS): If it can be determines that the `laneId` is constExpr, then subgroupBroadcast\n") +SLANG_RAW("// could be used on GLSL. For now we just use subgroupShuffle\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("T WaveReadLaneAt(T value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupShuffle($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformShuffle; OpGroupNonUniformShuffle $$T result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("vector<T,N> WaveReadLaneAt(vector<T,N> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupShuffle($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformShuffle; OpGroupNonUniformShuffle $$vector<T,N> result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("matrix<T, N, M> WaveReadLaneAt(matrix<T, N, M> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_waveShuffleMultiple(_getActiveMask(), $0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" matrix<T,N,M> result;\n") +SLANG_RAW(" for (int i = 0; i < N; ++i)\n") +SLANG_RAW(" result[i] = WaveReadLaneAt(value[i], lane);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskReadLaneAt(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE! WaveShuffle is a NON STANDARD HLSL intrinsic! It will map to WaveReadLaneAt on HLSL\n") +SLANG_RAW("// which means it will only work on hardware which allows arbitrary laneIds which is not true\n") +SLANG_RAW("// in general because it breaks the HLSL standard, which requires it's 'dynamically uniform' across the Wave.\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("T WaveShuffle(T value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupShuffle($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformShuffle; OpGroupNonUniformShuffle $$T result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskShuffle(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_shuffle)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_shuffle)]\n") +SLANG_RAW("vector<T,N> WaveShuffle(vector<T,N> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" if (__isHalf<T>()) __requireGLSLExtension(\"GL_EXT_shader_subgroup_extended_types_float16\");\n") +SLANG_RAW(" __intrinsic_asm \"subgroupShuffle($0, $1)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let ulane = uint(lane);\n") +SLANG_RAW(" return spirv_asm {OpCapability GroupNonUniformShuffle; OpGroupNonUniformShuffle $$vector<T,N> result Subgroup $value $ulane};\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskShuffle(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, subgroup_shuffle)]\n") +SLANG_RAW("matrix<T, N, M> WaveShuffle(matrix<T, N, M> value, int lane)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveReadLaneAt\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskShuffle(WaveGetActiveMask(), value, lane);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("uint WavePrefixCountBits(bool value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallotExclusiveBitCount(subgroupBallot($0))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WavePrefixCountBits($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" %mask:$$uint4 = OpGroupNonUniformBallot Subgroup $value;\n") +SLANG_RAW(" OpGroupNonUniformBallotBitCount $$uint result Subgroup 2 %mask\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskPrefixCountBits(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_ballot)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_ballot)]\n") +SLANG_RAW("uint4 WaveGetConvergedMulti()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"subgroupBallot(true)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveActiveBallot(true)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"make_uint4(__activemask(), 0, 0, 0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" let _true = true;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformBallot;\n") +SLANG_RAW(" OpGroupNonUniformBallot $$uint4 result Subgroup $_true\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("uint4 WaveGetActiveMulti()\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveGetConvergedMulti();\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Shader model 6.5 stuff\n") +SLANG_RAW("// https://github.com/microsoft/DirectX-Specs/blob/master/d3d/HLSL_ShaderModel6_5.md\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_partitioned)]\n") +SLANG_RAW("uint4 WaveMatch(T value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupPartitionNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformPartitionedNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_subgroup_partitioned\";\n") +SLANG_RAW(" OpGroupNonUniformPartitionNV $$uint4 result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskMatch(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_partitioned)]\n") +SLANG_RAW("uint4 WaveMatch(vector<T,N> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupPartitionNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability GroupNonUniformPartitionedNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_subgroup_partitioned\";\n") +SLANG_RAW(" OpGroupNonUniformPartitionNV $$uint4 result $value\n") +SLANG_RAW(" };\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskMatch(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, subgroup_partitioned)]\n") +SLANG_RAW("uint4 WaveMatch(matrix<T,N,M> value)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMatch\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint4 result = uint4(0xFFFFFFFF);\n") +SLANG_RAW(" [ForceUnroll]\n") +SLANG_RAW(" for (int i = 0; i < N; i++)\n") +SLANG_RAW(" result &= WaveMatch(value[i]);\n") +SLANG_RAW(" return result;\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return WaveMaskMatch(WaveGetActiveMask(), value);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("uint WaveMultiPrefixCountBits(bool value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_popc(__ballot_sync(($1).x, $0) & _getLaneLtMask())\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixCountBits\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("T WaveMultiPrefixBitAnd(T expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAnd(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveAnd($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("vector<T,N> WaveMultiPrefixBitAnd(vector<T,N> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAndMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveAnd($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("matrix<T,N,M> WaveMultiPrefixBitAnd(matrix<T,N,M> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixAndMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitAnd\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("T WaveMultiPrefixBitOr(T expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOr(, _getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveOr($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("vector<T,N> WaveMultiPrefixBitOr(vector<T,N> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOrMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveOr($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("matrix<T,N,M> WaveMultiPrefixBitOr(matrix<T,N,M> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixOrMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitOr\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("T WaveMultiPrefixBitXor(T expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXor(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveXor($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("__glsl_extension(GL_KHR_shader_subgroup_arithmetic)\n") +SLANG_RAW("__spirv_version(1.3)\n") +SLANG_RAW("[require(cuda_glsl_hlsl, waveprefix)]\n") +SLANG_RAW("vector<T,N> WaveMultiPrefixBitXor(vector<T,N> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXorMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subgroupExclusiveXor($0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("matrix<T,N,M> WaveMultiPrefixBitXor(matrix<T,N,M> expr, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixXorMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixBitXor\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("T WaveMultiPrefixProduct(T value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProduct(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixProduct\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("vector<T,N> WaveMultiPrefixProduct(vector<T,N> value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProductMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixProduct\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("matrix<T,N,M> WaveMultiPrefixProduct(matrix<T,N,M> value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixProductMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixProduct\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("T WaveMultiPrefixSum(T value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSum(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixSum\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("vector<T,N> WaveMultiPrefixSum(vector<T,N> value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSumMultiple(_getMultiPrefixMask(($1).x), $0 )\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixSum\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T : __BuiltinArithmeticType, let N : int, let M : int>\n") +SLANG_RAW("[require(cuda_hlsl, waveprefix)]\n") +SLANG_RAW("matrix<T,N,M> WaveMultiPrefixSum(matrix<T,N,M> value, uint4 mask)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"_wavePrefixSumMultiple(_getMultiPrefixMask(($1).x), $0)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WaveMultiPrefixSum\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_EXT_demote_to_helper_invocation)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_metal_spirv, helper_lane)]\n") +SLANG_RAW("bool IsHelperLane()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"IsHelperLane()\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"gl_HelperInvocation\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"simd_is_helper_thread()\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpExtension \"SPV_EXT_demote_to_helper_invocation\";\n") +SLANG_RAW(" OpCapability DemoteToHelperInvocationEXT;\n") +SLANG_RAW(" result:$$bool = OpIsHelperInvocationEXT\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// `typedef`s to help with the fact that HLSL has been sorta-kinda case insensitive at various points\n") +SLANG_RAW("typedef Texture2D texture2D;\n") +SLANG_RAW("\n") + + +// Buffer types + +static const struct { + char const* name; + SlangResourceAccess access; +} kBaseBufferAccessLevels[] = { + { "", SLANG_RESOURCE_ACCESS_READ }, + { "RW", SLANG_RESOURCE_ACCESS_READ_WRITE }, + { "RasterizerOrdered", SLANG_RESOURCE_ACCESS_RASTER_ORDERED }, +}; +static const int kBaseBufferAccessLevelCount = sizeof(kBaseBufferAccessLevels) / sizeof(kBaseBufferAccessLevels[0]); + +for (int aa = 0; aa < kBaseBufferAccessLevelCount; ++aa) +{ + auto access = kBaseBufferAccessLevels[aa].access; + sb << "__generic<T,let format:int=0>\n"; + sb << "typealias "; + sb << kBaseBufferAccessLevels[aa].name; + sb << "Buffer = __TextureImpl<T, __ShapeBuffer, 0, 0, 0, " << aa << ", 0, 0, format>;\n"; + + bool isReadOnly = aa == 0; + + char const* glslTextureSizeFunc = (isReadOnly) ? "textureSize" : "imageSize"; + char const* glslLoadFuncName = (isReadOnly) ? "texelFetch" : "imageLoad"; + char const* spvLoadInstName = (isReadOnly) ? "OpImageFetch" : "OpImageRead"; + char const* requireToSetQuery = (isReadOnly) ? "[require(glsl_hlsl_metal_spirv, texture_size)]" : "[require(glsl_hlsl_metal_spirv, image_size)]"; + char const* requireToSet = (isReadOnly) ? "[require(glsl_hlsl_metal_spirv, texture_sm_4_1)]" : "[require(glsl_hlsl_metal_spirv, texture_sm_4_1_compute_fragment)]"; + char const* requireToSet_onlyHLSL = (isReadOnly) ? "[require(hlsl, texture_sm_4_1)]" : "[require(hlsl, texture_sm_4_1_compute_fragment)]"; +SLANG_RAW("#line 16041 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T, __ShapeBuffer, 0, 0, 0, ") +SLANG_SPLICE(aa +) +SLANG_RAW(", 0, 0, format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" ") +SLANG_SPLICE(requireToSetQuery +) +SLANG_RAW("\n") +SLANG_RAW(" void GetDimensions(out uint dim)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetDimensions\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"($1 = ") +SLANG_SPLICE(glslTextureSizeFunc +) +SLANG_RAW("($0))\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"(*($1) = $0.get_width())\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" dim = spirv_asm {\n") +SLANG_RAW(" OpCapability ImageQuery;\n") +SLANG_RAW(" result:$$uint = OpImageQuerySize $this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_samplerless_texture_functions)\n") +SLANG_RAW(" ") +SLANG_SPLICE(isReadOnly?"[__readNone] ":"" +) +SLANG_RAW("\n") +SLANG_RAW(" ") +SLANG_SPLICE(requireToSet +) +SLANG_RAW("\n") +SLANG_RAW(" T Load(int location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"$c$0.read(uint($1))$z\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"") +SLANG_SPLICE(glslLoadFuncName +) +SLANG_RAW("($0, $1)$z\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" %sampled:__sampledType(T) = ") +SLANG_SPLICE(spvLoadInstName +) +SLANG_RAW(" $this $location;\n") +SLANG_RAW(" __truncate $$T result __sampledType(T) %sampled;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" ") +SLANG_SPLICE(isReadOnly?"[__readNone] ":"" +) +SLANG_RAW("\n") +SLANG_RAW(" ") +SLANG_SPLICE(requireToSet_onlyHLSL +) +SLANG_RAW("\n") +SLANG_RAW(" T Load(int location, out uint status)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Load\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(uint index) -> T {\n") +SLANG_RAW("\n") +SLANG_RAW(" ") +SLANG_SPLICE(isReadOnly?"[__readNone] ":"" +) +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" ") +SLANG_SPLICE(requireToSet +) +SLANG_RAW("\n") +SLANG_RAW(" get { return Load((int)index); }\n") + + if (access != SLANG_RESOURCE_ACCESS_READ) { +SLANG_RAW("#line 16098 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" [nonmutating] \n") +SLANG_RAW(" ") +SLANG_SPLICE(requireToSet +) +SLANG_RAW("\n") +SLANG_RAW(" set\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0)[$1] = $2\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"imageStore($0, int($1), $V2)\";\n") +SLANG_RAW(" case metal: __intrinsic_asm \"$0.write($2, $1)\";\n") +SLANG_RAW(" case spirv: spirv_asm {\n") +SLANG_RAW(" OpImageWrite $this $index __convertTexel(newValue);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // If a 'Texture[index]' is referred to by a '__ref', call 'kIROp_ImageSubscript(index)'.\n") +SLANG_RAW(" // This allows call's to stay aware that the input is from a 'Texture'.\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_ImageSubscript +) +SLANG_RAW(")\n") +SLANG_RAW(" [nonmutating]\n") +SLANG_RAW(" ref;\n") + + } // access != SLANG_RESOURCE_ACCESS_READ +SLANG_RAW("#line 16121 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" }\n") +SLANG_RAW(" \n") +SLANG_RAW("\n") +SLANG_RAW(" }; // end extension\n") + +} +SLANG_RAW("#line 16129 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// DirectX Raytracing (DXR) Support\n") +SLANG_RAW("//\n") +SLANG_RAW("// The following is based on the experimental DXR SDK v0.09.01.\n") +SLANG_RAW("//\n") +SLANG_RAW("// Numbering follows the sections in the \"D3D12 Raytracing Functional Spec\" v0.09 (2018-03-12)\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.1.1 - Ray Flags\n") +SLANG_RAW("\n") +SLANG_RAW("typedef uint RAY_FLAG;\n") +SLANG_RAW("\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_NONE = 0x00;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_FORCE_OPAQUE = 0x01;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_FORCE_NON_OPAQUE = 0x02;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_ACCEPT_FIRST_HIT_AND_END_SEARCH = 0x04;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_SKIP_CLOSEST_HIT_SHADER = 0x08;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_CULL_BACK_FACING_TRIANGLES = 0x10;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_CULL_FRONT_FACING_TRIANGLES = 0x20;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_CULL_OPAQUE = 0x40;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_CULL_NON_OPAQUE = 0x80;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_SKIP_TRIANGLES = 0x100;\n") +SLANG_RAW("static const RAY_FLAG RAY_FLAG_SKIP_PROCEDURAL_PRIMITIVES = 0x200;\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.1.2 - Ray Description Structure\n") +SLANG_RAW("\n") +SLANG_RAW("__target_intrinsic(hlsl, RayDesc)\n") +SLANG_RAW("__target_intrinsic(cuda, RayDesc)\n") +SLANG_RAW("struct RayDesc\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_intrinsic(hlsl, Origin)\n") +SLANG_RAW(" __target_intrinsic(cuda, Origin)\n") +SLANG_RAW(" float3 Origin;\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_intrinsic(hlsl, TMin)\n") +SLANG_RAW(" __target_intrinsic(cuda, TMin)\n") +SLANG_RAW(" float TMin;\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_intrinsic(hlsl, Direction)\n") +SLANG_RAW(" __target_intrinsic(cuda, Direction)\n") +SLANG_RAW(" float3 Direction;\n") +SLANG_RAW("\n") +SLANG_RAW(" __target_intrinsic(hlsl, TMax)\n") +SLANG_RAW(" __target_intrinsic(cuda, TMax)\n") +SLANG_RAW(" float TMax;\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.1.3 - Ray Acceleration Structure\n") +SLANG_RAW("\n") +SLANG_RAW("__builtin\n") +SLANG_RAW("__magic_type(RaytracingAccelerationStructureType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_RaytracingAccelerationStructureType +) +SLANG_RAW(")\n") +SLANG_RAW("struct RaytracingAccelerationStructure {};\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.1.4 - Subobject Definitions\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: We may decide to support these, but their reliance on C++ implicit\n") +SLANG_RAW("// constructor call syntax (`SomeType someVar(arg0, arg1);`) makes them\n") +SLANG_RAW("// annoying for the current Slang parsing strategy, and using global variables\n") +SLANG_RAW("// for this stuff comes across as a kludge rather than the best possible design.\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.1.5 - Intersection Attributes Structure\n") +SLANG_RAW("\n") +SLANG_RAW("__target_intrinsic(hlsl, BuiltInTriangleIntersectionAttributes)\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, raytracing)]\n") +SLANG_RAW("struct BuiltInTriangleIntersectionAttributes\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_intrinsic(hlsl, barycentrics)\n") +SLANG_RAW(" float2 barycentrics;\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.2 Shaders\n") +SLANG_RAW("\n") +SLANG_RAW("// Right now new shader stages need to be added directly to the compiler\n") +SLANG_RAW("// implementation, rather than being something that can be declared in the stdlib.\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3 - Intrinsics\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3.1\n") +SLANG_RAW("\n") +SLANG_RAW("// `executeCallableNV` is the GLSL intrinsic that will be used to implement\n") +SLANG_RAW("// `CallShader()` for GLSL-based targets.\n") +SLANG_RAW("//\n") +SLANG_RAW("[require(glsl, raytracing_raygen_closesthit_miss_callable)]\n") +SLANG_RAW("void __executeCallable(uint shaderIndex, int payloadLocation)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"executeCallableEXT\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Next is the custom intrinsic that will compute the payload location\n") +SLANG_RAW("// for a type being used in a `CallShader()` call for GLSL-based targets.\n") +SLANG_RAW("//\n") +SLANG_RAW("__generic<Payload>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetVulkanRayTracingPayloadLocation +) +SLANG_RAW(")\n") +SLANG_RAW("int __callablePayloadLocation(__ref Payload payload);\n") +SLANG_RAW("\n") +SLANG_RAW("// Now we provide a hard-coded definition of `CallShader()` for GLSL-based\n") +SLANG_RAW("// targets, which maps the generic HLSL operation into the non-generic\n") +SLANG_RAW("// GLSL equivalent.\n") +SLANG_RAW("//\n") +SLANG_RAW("__generic<Payload>\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_raygen_closesthit_miss_callable)]\n") +SLANG_RAW("void CallShader(uint shaderIndex, inout Payload payload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"CallShader\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanCallablePayload]\n") +SLANG_RAW(" static Payload p;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = payload;\n") +SLANG_RAW(" __executeCallable(shaderIndex, __callablePayloadLocation(p));\n") +SLANG_RAW(" payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanCallablePayload]\n") +SLANG_RAW(" static Payload p;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = payload;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExecuteCallableKHR $shaderIndex &p\n") +SLANG_RAW(" };\n") +SLANG_RAW(" payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3.2\n") +SLANG_RAW("\n") +SLANG_RAW("// Some functions only accept a \"struct type\" parameter. The \n") +SLANG_RAW("// following function addresses this issue by transforming non-struct\n") +SLANG_RAW("// parameters into a struct. \n") +SLANG_RAW("// side effect typed use locations (`inout`,`out`, etc.) are managed.\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_ForceVarIntoStructTemporarily +) +SLANG_RAW(")\n") +SLANG_RAW("Ref<T> __forceVarIntoStructTemporarily(inout T maybeStruct);\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<payload_t>\n") +SLANG_RAW("[require(hlsl, raytracing)]\n") +SLANG_RAW("void __traceRayHLSL(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"TraceRay\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl, raytracing_raygen_closesthit_miss)]\n") +SLANG_RAW("void __traceRay(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" float3 Origin,\n") +SLANG_RAW(" float TMin,\n") +SLANG_RAW(" float3 Direction,\n") +SLANG_RAW(" float TMax,\n") +SLANG_RAW(" int PayloadLocation)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"traceRayEXT\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: Slang's parsing logic currently puts modifiers on\n") +SLANG_RAW("// the `GenericDecl` rather than the inner decl when\n") +SLANG_RAW("// using our default syntax, which seems wrong. We need\n") +SLANG_RAW("// to fix this, but for now using the expanded `__generic`\n") +SLANG_RAW("// syntax works in a pinch.\n") +SLANG_RAW("//\n") +SLANG_RAW("__generic<Payload>\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetVulkanRayTracingPayloadLocation +) +SLANG_RAW(")\n") +SLANG_RAW("int __rayPayloadLocation(__ref Payload payload);\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__generic<payload_t>\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_raygen_closesthit_miss)]\n") +SLANG_RAW("void TraceRay(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __traceRayHLSL(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags,\n") +SLANG_RAW(" InstanceInclusionMask,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray,\n") +SLANG_RAW(" __forceVarIntoStructTemporarily(Payload));\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"traceOptiXRay\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW(" __traceRay(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags,\n") +SLANG_RAW(" InstanceInclusionMask,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction,\n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" __rayPayloadLocation(p));\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpTraceRayKHR \n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $RayFlags\n") +SLANG_RAW(" /**/ $InstanceInclusionMask\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ &p;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE!\n") +SLANG_RAW("// The name of the following functions may change when DXR supports\n") +SLANG_RAW("// a feature similar to the `GL_NV_ray_tracing_motion_blur` extension\n") +SLANG_RAW("//\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/nv/GLSL_NV_ray_tracing_motion_blur.txt\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<payload_t>\n") +SLANG_RAW("[require(hlsl, raytracing_motionblur)]\n") +SLANG_RAW("void __traceMotionRayHLSL(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"TraceMotionRay\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_NV_ray_tracing_motion_blur)\n") +SLANG_RAW("[require(glsl, raytracing_motionblur_raygen_closesthit_miss)]\n") +SLANG_RAW("void __traceMotionRay(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" float3 Origin,\n") +SLANG_RAW(" float TMin,\n") +SLANG_RAW(" float3 Direction,\n") +SLANG_RAW(" float TMax,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" int PayloadLocation)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"traceRayMotionNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_motionblur_raygen_closesthit_miss)]\n") +SLANG_RAW("__generic<payload_t>\n") +SLANG_RAW("void TraceMotionRay(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __traceMotionRayHLSL(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags,\n") +SLANG_RAW(" InstanceInclusionMask,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __forceVarIntoStructTemporarily(Payload));\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW(" __traceMotionRay(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags,\n") +SLANG_RAW(" InstanceInclusionMask,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction,\n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __rayPayloadLocation(p));\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW(" \n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW("\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability RayTracingMotionBlurNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_ray_tracing_motion_blur\";\n") +SLANG_RAW("\n") +SLANG_RAW(" OpTraceRayMotionNV\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $RayFlags\n") +SLANG_RAW(" /**/ $InstanceInclusionMask\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $CurrentTime\n") +SLANG_RAW(" /**/ &p;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3.3\n") +SLANG_RAW("\n") +SLANG_RAW("[require(glsl_spirv, raytracing_intersection)]\n") +SLANG_RAW("bool __reportIntersection(float tHit, uint hitKind)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reportIntersectionEXT\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$bool = OpReportIntersectionKHR $tHit $hitKind;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<A>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_intersection)]\n") +SLANG_RAW("bool ReportHit(float tHit, uint hitKind, A attributes)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"ReportHit($0, $1, $2)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" [__vulkanHitAttributes]\n") +SLANG_RAW(" static A a;\n") +SLANG_RAW(" a = attributes;\n") +SLANG_RAW(" return __reportIntersection(tHit, hitKind);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<each T : __BuiltinIntegerType>\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_intersection)]\n") +SLANG_RAW("bool ReportHitOptix(float tHit, uint hitKind, expand each T attribs)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"optixReportIntersection\";\n") +SLANG_RAW(" default:\n") +SLANG_RAW(" return ReportHit(tHit, hitKind, makeTuple(expand each attribs));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3.4\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit)]\n") +SLANG_RAW("void IgnoreHit()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"IgnoreHit\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"ignoreIntersectionEXT;\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixIgnoreIntersection\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" { \n") +SLANG_RAW(" OpIgnoreIntersectionKHR; %_ = OpLabel \n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.3.5\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit)]\n") +SLANG_RAW("void AcceptHitAndEndSearch()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"AcceptHitAndEndSearch\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"terminateRayEXT;\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixTerminateRay\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" { \n") +SLANG_RAW(" OpTerminateRayKHR; %_ = OpLabel \n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.4 - System Values and Special Semantics\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: Many of these functions need to be restricted so that\n") +SLANG_RAW("// they can only be accessed from specific stages.\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.4.1 - Ray Dispatch System Values\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_allstages)]\n") +SLANG_RAW("uint3 DispatchRaysIndex()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"DispatchRaysIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_LaunchIDEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetLaunchIndex\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint3 = OpLoad builtin(LaunchIdKHR:uint3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_allstages)]\n") +SLANG_RAW("uint3 DispatchRaysDimensions()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"DispatchRaysDimensions\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_LaunchSizeEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetLaunchDimensions\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint3 = OpLoad builtin(LaunchSizeKHR:uint3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.4.2 - Ray System Values\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("float3 WorldRayOrigin()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WorldRayOrigin\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_WorldRayOriginEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetWorldRayOrigin\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpLoad builtin(WorldRayOriginKHR:float3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("float3 WorldRayDirection()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WorldRayDirection\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_WorldRayDirectionEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetWorldRayDirection\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpLoad builtin(WorldRayDirectionKHR:float3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("float RayTMin()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"RayTMin\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_RayTminEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetRayTmin\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpLoad builtin(RayTminKHR:float);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Note: The `RayTCurrent()` intrinsic should translate to\n") +SLANG_RAW("// either `gl_HitTNV` (for hit shaders) or `gl_RayTmaxNV`\n") +SLANG_RAW("// (for intersection shaders). Right now we are handling this\n") +SLANG_RAW("// during code emission, for simplicity.\n") +SLANG_RAW("//\n") +SLANG_RAW("// TODO: Once the compiler supports a more refined concept\n") +SLANG_RAW("// of profiles/capabilities and overloading based on them,\n") +SLANG_RAW("// we should simply provide two overloads here, specialized\n") +SLANG_RAW("// to the appropriate Vulkan stages.\n") +SLANG_RAW("//\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("float RayTCurrent()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"RayTCurrent\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_RayTmaxEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetRayTmax\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpLoad builtin(RayTmaxKHR:float);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("uint RayFlags()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"RayFlags\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_IncomingRayFlagsEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetRayFlags\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(IncomingRayFlagsKHR:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.4.3 - Primitive/Object Space System Values\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("uint InstanceIndex()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InstanceIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_InstanceID)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetInstanceIndex\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(InstanceId:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("uint InstanceID()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"InstanceID\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_InstanceCustomIndexEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetInstanceId\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(InstanceCustomIndexKHR:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("uint PrimitiveIndex()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"PrimitiveIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_PrimitiveID)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetPrimitiveIndex\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(PrimitiveId:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float3 ObjectRayOrigin()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ObjectRayOrigin\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_ObjectRayOriginEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetObjectRayOrigin\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpLoad builtin(ObjectRayOriginKHR:float3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float3 ObjectRayDirection()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ObjectRayDirection\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_ObjectRayDirectionEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetObjectRayDirection\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpLoad builtin(ObjectRayDirectionKHR:float3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// TODO: optix has an optixGetObjectToWorldTransformMatrix function that returns 12\n") +SLANG_RAW("// floats by reference.\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float3x4 ObjectToWorld3x4()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ObjectToWorld3x4\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose(gl_ObjectToWorldEXT)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" %mat:$$float4x3 = OpLoad builtin(ObjectToWorldKHR:float4x3);\n") +SLANG_RAW(" result:$$float3x4 = OpTranspose %mat;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float3x4 WorldToObject3x4()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WorldToObject3x4\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose(gl_WorldToObjectEXT)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" %mat:$$float4x3 = OpLoad builtin(WorldToObjectKHR:float4x3);\n") +SLANG_RAW(" result:$$float3x4 = OpTranspose %mat;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float4x3 ObjectToWorld4x3()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"ObjectToWorld4x3\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_ObjectToWorldEXT)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpLoad builtin(ObjectToWorldKHR:float4x3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("float4x3 WorldToObject4x3()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"WorldToObject4x3\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_WorldToObjectEXT)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpLoad builtin(WorldToObjectKHR:float4x3);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// NOTE!\n") +SLANG_RAW("// The name of the following functions may change when DXR supports\n") +SLANG_RAW("// a feature similar to the `GL_NV_ray_tracing_motion_blur` extension\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_extension(GL_NV_ray_tracing_motion_blur)\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_motionblur_anyhit_closesthit_intersection_miss)]\n") +SLANG_RAW("float RayCurrentTime()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"RayCurrentTime\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_CurrentRayTimeNV)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpLoad builtin(CurrentRayTimeNV:float);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Note: The provisional DXR spec included these unadorned\n") +SLANG_RAW("// `ObjectToWorld()` and `WorldToObject()` functions, so\n") +SLANG_RAW("// we will forward them to the new names as a convience\n") +SLANG_RAW("// for users who are porting their code.\n") +SLANG_RAW("//\n") +SLANG_RAW("// TODO: Should we provide a deprecation warning on these\n") +SLANG_RAW("// declarations, so that users can know they aren't coding\n") +SLANG_RAW("// against the final spec?\n") +SLANG_RAW("//\n") +SLANG_RAW("[NonUniformReturn] float3x4 ObjectToWorld() { return ObjectToWorld3x4(); }\n") +SLANG_RAW("[NonUniformReturn] float3x4 WorldToObject() { return WorldToObject3x4(); }\n") +SLANG_RAW("\n") +SLANG_RAW("// 10.4.4 - Hit Specific System values\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda_glsl_hlsl_spirv, raytracing_anyhit_closesthit)]\n") +SLANG_RAW("uint HitKind()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"HitKind\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_HitKindEXT)\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"optixGetHitKind\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(HitKindKHR:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Pre-defined hit kinds (not documented explicitly)\n") +SLANG_RAW("static const uint HIT_KIND_TRIANGLE_FRONT_FACE = 254;\n") +SLANG_RAW("static const uint HIT_KIND_TRIANGLE_BACK_FACE = 255;\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// Shader Model 6.4\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// Treats `left` and `right` as 4-component vectors of `UInt8` and computes `dot(left, right) + acc`\n") +SLANG_RAW("uint dot4add_u8packed(uint left, uint right, uint acc);\n") +SLANG_RAW("\n") +SLANG_RAW("// Treats `left` and `right` as 4-component vectors of `Int8` and computes `dot(left, right) + acc`\n") +SLANG_RAW("int dot4add_i8packed(uint left, uint right, int acc);\n") +SLANG_RAW("\n") +SLANG_RAW("// Computes `dot(left, right) + acc`.\n") +SLANG_RAW("//\n") +SLANG_RAW("// May not produce infinities or NaNs for intermediate results that overflow the range of `half`\n") +SLANG_RAW("float dot2add(float2 left, float2 right, float acc);\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// Shader Model 6.5\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// Mesh Shaders\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// Set the number of output vertices and primitives for a mesh shader invocation.\n") +SLANG_RAW("__glsl_extension(GL_EXT_mesh_shader)\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("[require(glsl_hlsl_metal_spirv, meshshading)]\n") +SLANG_RAW("[noRefInline]\n") +SLANG_RAW("void SetMeshOutputCounts(uint vertexCount, uint primitiveCount)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"SetMeshOutputCounts\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"SetMeshOutputsEXT\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"_slang_mesh.set_primitive_count($1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MeshShadingEXT;\n") +SLANG_RAW(" OpExtension \"SPV_EXT_mesh_shader\";\n") +SLANG_RAW(" OpSetMeshOutputsEXT $vertexCount $primitiveCount;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Specify the number of downstream mesh shader thread groups to invoke from an amplification shader,\n") +SLANG_RAW("// and provide the values for per-mesh payload parameters.\n") +SLANG_RAW("//\n") +SLANG_RAW("// This function doesn't return.\n") +SLANG_RAW("//\n") +SLANG_RAW("// This function cannot be inlined due to a legalization pass happening mid-way through processing\n") +SLANG_RAW("// and later more processing happening to the function which requires eventual inlining.\n") +SLANG_RAW("[KnownBuiltin(\"DispatchMesh\")]\n") +SLANG_RAW("[require(glsl_hlsl_metal_spirv, meshshading)]\n") +SLANG_RAW("[noRefInline]\n") +SLANG_RAW("void DispatchMesh<P>(uint threadGroupCountX, uint threadGroupCountY, uint threadGroupCountZ, __ref P meshPayload)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"DispatchMesh\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" // This intrinsic doesn't take into account writing meshPayload. That\n") +SLANG_RAW(" // is dealt with separately by 'legalizeDispatchMeshPayloadForGLSL'.\n") +SLANG_RAW(" __intrinsic_asm \"EmitMeshTasksEXT($0, $1, $2)\";\n") +SLANG_RAW(" case metal:\n") +SLANG_RAW(" __intrinsic_asm \"*_slang_mesh_payload = *$3; _slang_mgp.set_threadgroups_per_grid(uint3($0, $1, $2)); return;\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability MeshShadingEXT;\n") +SLANG_RAW(" OpExtension \"SPV_EXT_mesh_shader\";\n") +SLANG_RAW(" OpEmitMeshTasksEXT $threadGroupCountX $threadGroupCountY $threadGroupCountZ &meshPayload;\n") +SLANG_RAW(" // OpEmitMeshTasksExt is a terminator, so we need to start a new\n") +SLANG_RAW(" // block to hold whatever comes after this intrinsic\n") +SLANG_RAW(" %_ = OpLabel\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// \"Sampler feedback\" types `FeedbackTexture2D` and `FeedbackTexture2DArray`.\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// https://microsoft.github.io/DirectX-Specs/d3d/SamplerFeedback.html\n") +SLANG_RAW("\n") +SLANG_RAW("// The docs describe these as 'types' but their syntax makes them seem enum like, and enum is a simpler way to implement them\n") +SLANG_RAW("// But slang enums are always 'enum class like', so I use an empty struct type here\n") +SLANG_RAW("\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("[builtin]\n") +SLANG_RAW("interface __BuiltinSamplerFeedbackType {};\n") +SLANG_RAW("\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("__magic_type(FeedbackType, ") +SLANG_SPLICE(int(FeedbackType::Kind::MinMip) +) +SLANG_RAW(")\n") +SLANG_RAW("__target_intrinsic(hlsl, SAMPLER_FEEDBACK_MIN_MIP)\n") +SLANG_RAW("struct SAMPLER_FEEDBACK_MIN_MIP : __BuiltinSamplerFeedbackType {};\n") +SLANG_RAW("\n") +SLANG_RAW("[sealed]\n") +SLANG_RAW("__magic_type(FeedbackType, ") +SLANG_SPLICE(int(FeedbackType::Kind::MipRegionUsed) +) +SLANG_RAW(")\n") +SLANG_RAW("__target_intrinsic(hlsl, SAMPLER_FEEDBACK_MIP_REGION_USED)\n") +SLANG_RAW("struct SAMPLER_FEEDBACK_MIP_REGION_USED : __BuiltinSamplerFeedbackType {};\n") +SLANG_RAW("\n") +SLANG_RAW("// All of these objects are write-only resources that point to a special kind of unordered access view meant for sampler feedback.\n") +SLANG_RAW("__generic<T:__BuiltinSamplerFeedbackType>\n") +SLANG_RAW("extension __TextureImpl<T,__Shape2D, 0, 0, 0, ") +SLANG_SPLICE(kStdlibResourceAccessFeedback +) +SLANG_RAW(", 0, 0, 0>\n") +SLANG_RAW("{\n") +SLANG_RAW(" // With Clamp\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedback<S>(Texture2D<S> tex, SamplerState samp, float2 location, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackBias<S>(Texture2D<S> tex, SamplerState samp, float2 location, float bias, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackGrad<S>(Texture2D<S> tex, SamplerState samp, float2 location, float2 ddx, float2 ddy, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5, $6)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5, $6)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Level\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackLevel<S>(Texture2D<S> tex, SamplerState samp, float2 location, float lod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackLevel($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackLevel($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Without Clamp\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedback<S>(Texture2D<S> tex, SamplerState samp, float2 location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackBias<S>(Texture2D<S> tex, SamplerState samp, float2 location, float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackGrad<S>(Texture2D<S> tex, SamplerState samp, float2 location, float2 ddx, float2 ddy)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T:__BuiltinSamplerFeedbackType>\n") +SLANG_RAW("extension __TextureImpl<T,__Shape2D, 1, 0, 0, ") +SLANG_SPLICE(kStdlibResourceAccessFeedback +) +SLANG_RAW(", 0, 0, 0>\n") +SLANG_RAW("{\n") +SLANG_RAW(" // With Clamp\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedback<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackBias<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float bias, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackGrad<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float3 ddx, float3 ddy, float clamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5, $6)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5, $6)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Level\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackLevel<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float lod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackLevel($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackLevel($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Without Clamp\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedback<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedback($1, $2, $3)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackBias<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackBias($1, $2, $3, $4)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(cpp_hlsl)]\n") +SLANG_RAW(" void WriteSamplerFeedbackGrad<S>(Texture2DArray<S> texArray, SamplerState samp, float3 location, float3 ddx, float3 ddy)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($0).WriteSamplerFeedbackGrad($1, $2, $3, $4, $5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// DXR 1.1 and `TraceRayInline` support\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// Get the index of the geometry that was hit in an intersection, any-hit, or closest-hit shader\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, raytracing_anyhit_closesthit_intersection)]\n") +SLANG_RAW("uint GeometryIndex()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"GeometryIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(gl_GeometryIndexEXT)\";\n") +SLANG_RAW(" case spirv: return spirv_asm {\n") +SLANG_RAW(" result:$$uint = OpLoad builtin(RayGeometryIndexKHR:uint);\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Get the vertex positions of the currently hit triangle in any-hit or closest-hit shader.\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/ext/GLSL_EXT_ray_tracing_position_fetch.txt\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing_position_fetch)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_spirv, raytracing_position)]\n") +SLANG_RAW("float3 HitTriangleVertexPosition(uint index)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"gl_HitTriangleVertexPositionsEXT[$0]\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability RayTracingKHR;\n") +SLANG_RAW(" OpCapability RayTracingPositionFetchKHR;\n") +SLANG_RAW(" OpExtension \"SPV_KHR_ray_tracing\";\n") +SLANG_RAW(" OpExtension \"SPV_KHR_ray_tracing_position_fetch\";\n") +SLANG_RAW(" %_ptr_Input_v3float = OpTypePointer Input $$float3;\n") +SLANG_RAW(" %addr : %_ptr_Input_v3float = OpAccessChain builtin(HitTriangleVertexPositionsKHR:float3[3]) $index;\n") +SLANG_RAW(" result:$$float3 = OpLoad %addr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Status of whether a (closest) hit has been committed in a `RayQuery`.\n") +SLANG_RAW("typedef uint COMMITTED_STATUS;\n") +SLANG_RAW("\n") +SLANG_RAW("// No hit committed.\n") +SLANG_RAW("static const COMMITTED_STATUS COMMITTED_NOTHING = 0;\n") +SLANG_RAW("\n") +SLANG_RAW("// Closest hit is a triangle.\n") +SLANG_RAW("//\n") +SLANG_RAW("// This could be an opaque triangle hit found by the fixed-function\n") +SLANG_RAW("// traversal and intersection implementation, or a non-opaque\n") +SLANG_RAW("// triangle hit committed by user code with `RayQuery.CommitNonOpaqueTriangleHit`\n") +SLANG_RAW("//\n") +SLANG_RAW("static const COMMITTED_STATUS COMMITTED_TRIANGLE_HIT = 1;\n") +SLANG_RAW("\n") +SLANG_RAW("// Closest hit is a procedural primitive.\n") +SLANG_RAW("//\n") +SLANG_RAW("// A procedural hit primitive is committed using `RayQuery.CommitProceduralPrimitiveHit`.\n") +SLANG_RAW("static const COMMITTED_STATUS COMMITTED_PROCEDURAL_PRIMITIVE_HIT = 2;\n") +SLANG_RAW("\n") +SLANG_RAW("// Type of candidate hit that a `RayQuery` is pausing at.\n") +SLANG_RAW("//\n") +SLANG_RAW("// A `RayQuery` can automatically commit hits with opaque triangles,\n") +SLANG_RAW("// but yields to user code for other hits to allow them to be\n") +SLANG_RAW("// dismissed or committed.\n") +SLANG_RAW("//\n") +SLANG_RAW("typedef uint CANDIDATE_TYPE;\n") +SLANG_RAW("\n") +SLANG_RAW("// Candidate hit is a non-opaque triangle.\n") +SLANG_RAW("static const CANDIDATE_TYPE CANDIDATE_NON_OPAQUE_TRIANGLE = 0;\n") +SLANG_RAW("\n") +SLANG_RAW("// Candidate hit is a procedural primitive.\n") +SLANG_RAW("static const CANDIDATE_TYPE CANDIDATE_PROCEDURAL_PRIMITIVE = 1;\n") +SLANG_RAW("\n") +SLANG_RAW("// Handle to state of an in-progress ray-tracing query.\n") +SLANG_RAW("//\n") +SLANG_RAW("// The ray query is effectively a coroutine that user shader\n") +SLANG_RAW("// code can resume to continue tracing the ray, and which yields\n") +SLANG_RAW("// back to the user code at interesting events along the ray.\n") +SLANG_RAW("//\n") +SLANG_RAW("// Note: The treatment of the `RayQuery` type in Slang does not\n") +SLANG_RAW("// perfectly match its semantics in vanilla HLSL in some corner\n") +SLANG_RAW("// cases. Specifically, a `RayQuery` in vanilla HLSL is an\n") +SLANG_RAW("// opaque handle to mutable storage, and assigning a `RayQuery`\n") +SLANG_RAW("// or passing one as a parameter will only copy the *handle*,\n") +SLANG_RAW("// potentially resulting in aliasing of the underlying mutable\n") +SLANG_RAW("// storage.\n") +SLANG_RAW("//\n") +SLANG_RAW("// In contrast, Slang considers a `RayQuery` to own its mutable\n") +SLANG_RAW("// state, and (because the API does not support cloning of queries),\n") +SLANG_RAW("// `RayQuery` values are non-copyable (aka \"move-only\").\n") +SLANG_RAW("//\n") +SLANG_RAW("// The main place where this arises as a consideration is when\n") +SLANG_RAW("// passing a `RayQuery` down into a function that will perform\n") +SLANG_RAW("// mutating operations on it (e.g., `TraceRay` or `Proceed`):\n") +SLANG_RAW("//\n") +SLANG_RAW("// void myFunc( inout RayQuery<FLAGS> q )\n") +SLANG_RAW("// {\n") +SLANG_RAW("// q.Proceed();\n") +SLANG_RAW("// }\n") +SLANG_RAW("//\n") +SLANG_RAW("// In Slang, a parameter like `q` above should be declared `inout`.\n") +SLANG_RAW("// HLSL does not care about whether `q` is declared `inout` or not.\n") +SLANG_RAW("//\n") +SLANG_RAW("//cannot use a cap for struct with unequal target support\n") +SLANG_RAW("//since it will propegate rules to children\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW("[__NonCopyableType]\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_RayQueryType +) +SLANG_RAW(")\n") +SLANG_RAW("struct RayQuery <let rayFlagsGeneric : RAY_FLAG = RAY_FLAG_NONE>\n") +SLANG_RAW("{\n") +SLANG_RAW(" // Create a new ray query, initialized to its default state.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_AllocateOpaqueHandle +) +SLANG_RAW(")\n") +SLANG_RAW(" __init();\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" void __rayQueryInitializeEXT(\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" RAY_FLAG rayFlags,\n") +SLANG_RAW(" uint instanceInclusionMask,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float tMin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float tMax)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryInitializeEXT($0, $1, $2, $3, $4, $5, $6, $7)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm {\n") +SLANG_RAW(" OpRayQueryInitializeKHR &this $accelerationStructure $rayFlags $instanceInclusionMask $origin $tMin $direction $tMax;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Initialize a ray-tracing query.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // This method may be called on a \"fresh\" ray query, or\n") +SLANG_RAW(" // on one that is already tracing a ray. In the latter\n") +SLANG_RAW(" // case any state related to the ray previously being\n") +SLANG_RAW(" // traced is overwritten.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // The `rayFlags` here will be bitwise ORed with\n") +SLANG_RAW(" // the `rayFlags` passed as a generic argument to\n") +SLANG_RAW(" // `RayQuery` to get the effective ray flags, which\n") +SLANG_RAW(" // must obey any API-imposed restrictions.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" [__unsafeForceInlineEarly]\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" void TraceRayInline(\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" RAY_FLAG rayFlags,\n") +SLANG_RAW(" uint instanceInclusionMask,\n") +SLANG_RAW(" RayDesc ray)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".TraceRayInline\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" __rayQueryInitializeEXT(\n") +SLANG_RAW(" accelerationStructure,\n") +SLANG_RAW(" rayFlags | rayFlagsGeneric,\n") +SLANG_RAW(" instanceInclusionMask,\n") +SLANG_RAW(" ray.Origin,\n") +SLANG_RAW(" ray.TMin,\n") +SLANG_RAW(" ray.Direction,\n") +SLANG_RAW(" ray.TMax);\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" \n") +SLANG_RAW(" // Resume the ray query coroutine.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // If the coroutine suspends because of encountering\n") +SLANG_RAW(" // a candidate hit that cannot be resolved with fixed-funciton\n") +SLANG_RAW(" // logic, this function returns `true`, and the `Candidate*()`\n") +SLANG_RAW(" // functions should be used by application code to resolve\n") +SLANG_RAW(" // the candidate hit (by either committing or ignoring it).\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // If the coroutine terminates because traversal is\n") +SLANG_RAW(" // complete (or has been aborted), this function returns\n") +SLANG_RAW(" // `false`, and application code should use the `Committed*()`\n") +SLANG_RAW(" // functions to appropriately handle the closest hit (it any)\n") +SLANG_RAW(" // that was found.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" bool Proceed()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Proceed\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryProceedEXT\";\n") +SLANG_RAW(" case spirv: return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$bool = OpRayQueryProceedKHR &this\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Causes the ray query to terminate.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // This function cases the ray query to act as if\n") +SLANG_RAW(" // traversal has terminated, so that subsequent\n") +SLANG_RAW(" // `Proceed()` calls will return `false`.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" void Abort()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".Abort\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryTerminateEXT\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpRayQueryTerminateKHR &this };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Commit the current non-opaque triangle hit.\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" void CommitNonOpaqueTriangleHit()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CommitNonOpaqueTriangleHit\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryConfirmIntersectionEXT\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpRayQueryConfirmIntersectionKHR &this };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Commit the current procedural primitive hit, with hit time `t`.\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [mutating]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" void CommitProceduralPrimitiveHit(float t)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CommitProceduralPrimitiveHit\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGenerateIntersectionEXT\";\n") +SLANG_RAW(" case spirv: spirv_asm { OpRayQueryGenerateIntersectionKHR &this $t };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Get the type of candidate hit being considered.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" // The ray query coroutine will suspend when it encounters\n") +SLANG_RAW(" // a hit that cannot be resolved with fixed-function logic\n") +SLANG_RAW(" // (either a non-opaque triangle or a procedural primitive).\n") +SLANG_RAW(" // In either of those cases, `CandidateType()` will return\n") +SLANG_RAW(" // the kind of candidate hit that must be resolved by\n") +SLANG_RAW(" // user code.\n") +SLANG_RAW(" //\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" CANDIDATE_TYPE CandidateType()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CandidateType\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionTypeEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint RayQueryCandidateIntersectionKHR = 0;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$CANDIDATE_TYPE = OpRayQueryGetIntersectionTypeKHR &this $RayQueryCandidateIntersectionKHR;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // Get the status of the committed (closest) hit, if any.\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" COMMITTED_STATUS CommittedStatus()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CommittedStatus\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionTypeEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint RayQueryCommittedIntersectionKHR = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$COMMITTED_STATUS = OpRayQueryGetIntersectionTypeKHR &this $RayQueryCommittedIntersectionKHR;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" bool CandidateProceduralPrimitiveNonOpaque()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CandidateProceduralPrimitiveNonOpaque\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"(!rayQueryGetIntersectionCandidateAABBOpaqueEXT($0))\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" %rr:$$bool = OpRayQueryGetIntersectionCandidateAABBOpaqueKHR &this;\n") +SLANG_RAW(" result:$$bool = OpLogicalNot %rr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float CandidateTriangleRayT()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CandidateTriangleRayT\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionTEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpRayQueryGetIntersectionTKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float CommittedRayT()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".CommittedRayT\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionTEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpRayQueryGetIntersectionTKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("///missing hlsl equivlent; only implemented for glsl & spirv\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CandidateRayInstanceCustomIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceCustomIndexEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionInstanceCustomIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CommittedRayInstanceCustomIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceCustomIndexEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionInstanceCustomIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CandidateRayInstanceId()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceIdEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionInstanceIdKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CommittedRayInstanceId()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceIdEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionInstanceIdKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" uint CandidateRayInstanceShaderBindingTableRecordOffset()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceShaderBindingTableRecordOffsetEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpRayQueryGetIntersectionInstanceShaderBindingTableRecordOffsetKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" uint CommittedRayInstanceShaderBindingTableRecordOffset()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionInstanceShaderBindingTableRecordOffsetEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpRayQueryGetIntersectionInstanceShaderBindingTableRecordOffsetKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CandidateRayGeometryIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionGeometryIndexEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionGeometryIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CommittedRayGeometryIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionGeometryIndexEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionGeometryIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CandidateRayPrimitiveIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionPrimitiveIndexEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionPrimitiveIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" int CommittedRayPrimitiveIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionPrimitiveIndexEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$int = OpRayQueryGetIntersectionPrimitiveIndexKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float2 CandidateRayBarycentrics()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionBarycentricsEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float2 = OpRayQueryGetIntersectionBarycentricsKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float2 CommittedRayBarycentrics()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionBarycentricsEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float2 = OpRayQueryGetIntersectionBarycentricsKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" bool CandidateRayFrontFace()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionFrontFaceEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$bool = OpRayQueryGetIntersectionFrontFaceKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" bool CommittedRayFrontFace()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionFrontFaceEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$bool = OpRayQueryGetIntersectionFrontFaceKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 CandidateRayObjectRayDirection()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectRayDirectionEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetIntersectionObjectRayDirectionKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 CommittedRayObjectRayDirection()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectRayDirectionEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetIntersectionObjectRayDirectionKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 CandidateRayObjectRayOrigin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectRayOriginEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetIntersectionObjectRayOriginKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 CommittedRayObjectRayOrigin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectRayOriginEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetIntersectionObjectRayOriginKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float4x3 CandidateRayObjectToWorld()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectToWorldEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpRayQueryGetIntersectionObjectToWorldKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float4x3 CommittedRayObjectToWorld()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionObjectToWorldEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpRayQueryGetIntersectionObjectToWorldKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float4x3 CandidateRayWorldToObject()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionWorldToObjectEXT($0, false)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 0;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpRayQueryGetIntersectionWorldToObjectKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [require(glsl_spirv, rayquery)]\n") +SLANG_RAW(" float4x3 CommittedRayWorldToObject()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersectionWorldToObjectEXT($0, true)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = 1;\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpRayQueryGetIntersectionWorldToObjectKHR $$float4x3 result &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("///~\n") +SLANG_RAW("\n") + + const char* kCandidateCommitted[] = {"Candidate", "Committed"}; + + // Access Candidate and Committed Matrices. + for (uint32_t candidateOrCommitted = 0; candidateOrCommitted < 2; candidateOrCommitted++) + { + auto ccName = kCandidateCommitted[candidateOrCommitted]; + auto ccTF = candidateOrCommitted == 0 ? "false" : "true"; +SLANG_RAW("#line 17976 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing_position_fetch)\n") +SLANG_RAW(" [require(glsl, rayquery_position)]\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" void __glslGetIntersectionTriangleVertexPositions") +SLANG_SPLICE(ccName +) +SLANG_RAW("(out float3 arr[3])\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __intrinsic_asm \"rayQueryGetIntersectionTriangleVertexPositionsEXT($0, ") +SLANG_SPLICE(ccTF +) +SLANG_RAW(", $1)\";\n") +SLANG_RAW(" };\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [require(glsl, rayquery_position)]\n") +SLANG_RAW(" [require(spirv, rayquery_position)]\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" float3[3] ") +SLANG_SPLICE(ccName +) +SLANG_RAW("GetIntersectionTriangleVertexPositions()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" typedef float3[3] float3Arr3;\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" float3 output[3];\n") +SLANG_RAW(" __glslGetIntersectionTriangleVertexPositions") +SLANG_SPLICE(ccName +) +SLANG_RAW("(output);\n") +SLANG_RAW(" return output;\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = ") +SLANG_SPLICE(candidateOrCommitted +) +SLANG_RAW(";\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability RayQueryKHR;\n") +SLANG_RAW(" OpCapability RayQueryPositionFetchKHR;\n") +SLANG_RAW(" OpExtension \"SPV_KHR_ray_query\";\n") +SLANG_RAW(" OpExtension \"SPV_KHR_ray_tracing_position_fetch\";\n") +SLANG_RAW(" result: $$float3Arr3 = OpRayQueryGetIntersectionTriangleVertexPositionsKHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" };\n") +SLANG_RAW("\n") +SLANG_RAW(" // CandidateObjectToWorld3x4, CandidateWorldToObject4x3\n") +SLANG_RAW(" // CommittedObjectToWorld3x4, CommittedObjectToWorld4x3\n") + + const char* kRayQueryMatrixNames[] = {"ObjectToWorld", "WorldToObject"}; + for (auto matName : kRayQueryMatrixNames) { + +SLANG_RAW("#line 18018 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float3x4 ") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(matName +) +SLANG_RAW("3x4()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"transpose(rayQueryGetIntersection") +SLANG_SPLICE(matName +) +SLANG_RAW("EXT($0, ") +SLANG_SPLICE(ccTF +) +SLANG_RAW("))\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(matName +) +SLANG_RAW("3x4\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = ") +SLANG_SPLICE(candidateOrCommitted +) +SLANG_RAW(";\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" %m:$$float4x3 = OpRayQueryGetIntersection") +SLANG_SPLICE(matName +) +SLANG_RAW("KHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" result:$$float3x4 = OpTranspose %m;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__readNone]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float4x3 ") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(matName +) +SLANG_RAW("4x3()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersection") +SLANG_SPLICE(matName +) +SLANG_RAW("EXT($0, ") +SLANG_SPLICE(ccTF +) +SLANG_RAW(")\";\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(matName +) +SLANG_RAW("4x3\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = ") +SLANG_SPLICE(candidateOrCommitted +) +SLANG_RAW(";\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float4x3 = OpRayQueryGetIntersection") +SLANG_SPLICE(matName +) +SLANG_RAW("KHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + } // ObjectToWorld/WorldToObject. + + // Access Candidate and Committed properties. + struct RayQueryMethodEntry + { + const char* type; + const char* hlslName; + const char* glslName; + }; + const RayQueryMethodEntry rayQueryMethods[] = { + {"uint", "InstanceIndex", "InstanceId"}, + {"uint", "InstanceID", "InstanceCustomIndex"}, + {"uint", "PrimitiveIndex", "PrimitiveIndex"}, + {"uint", "GeometryIndex", "GeometryIndex"}, + {"uint", "InstanceContributionToHitGroupIndex", "InstanceShaderBindingTableRecordOffset"}, + {"float3", "ObjectRayOrigin", "ObjectRayOrigin"}, + {"float3", "ObjectRayDirection", "ObjectRayDirection"}, + {"bool", "TriangleFrontFace", "FrontFace"}, + {"float2", "TriangleBarycentrics", "Barycentrics"}, + }; + for (auto method : rayQueryMethods) { +SLANG_RAW("#line 18081 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" ") +SLANG_SPLICE(method.type +) +SLANG_RAW(" ") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(method.hlslName +) +SLANG_RAW("()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".") +SLANG_SPLICE(ccName +) +SLANG_SPLICE(method.hlslName +) +SLANG_RAW("\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetIntersection") +SLANG_SPLICE(method.glslName +) +SLANG_RAW("EXT($0, ") +SLANG_SPLICE(ccTF +) +SLANG_RAW(")\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint iCandidateOrCommitted = ") +SLANG_SPLICE(candidateOrCommitted +) +SLANG_RAW(";\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$") +SLANG_SPLICE(method.type +) +SLANG_RAW(" = OpRayQueryGetIntersection") +SLANG_SPLICE(method.glslName +) +SLANG_RAW("KHR &this $iCandidateOrCommitted;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") + + } // Candidate/Committed properties. + } // for ("Candidate", "Committed") +SLANG_RAW("#line 18104 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" // Access properties of the ray being traced.\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" uint RayFlags()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".RayFlags\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetRayFlagsEXT\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$uint = OpRayQueryGetRayFlagsKHR &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 WorldRayOrigin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".WorldRayOrigin\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetWorldRayOriginEXT\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetWorldRayOriginKHR &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float3 WorldRayDirection()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".WorldRayDirection\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetWorldRayDirectionEXT\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float3 = OpRayQueryGetWorldRayDirectionKHR &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_query)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [NonUniformReturn]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, rayquery)]\n") +SLANG_RAW(" float RayTMin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".RayTMin\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"rayQueryGetRayTMinEXT\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" result:$$float = OpRayQueryGetRayTMinKHR &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" };\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// SubpassInput\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("__magic_type(SubpassInputType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_SubpassInputType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(glsl_hlsl_spirv, subpass)]\n") +SLANG_RAW("struct __SubpassImpl<T, let isMS:int>\n") +SLANG_RAW("{\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T = float4, let isMS:int=0>\n") +SLANG_RAW("typealias SubpassInput = __SubpassImpl<T, isMS>;\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("extension __SubpassImpl<T, 0>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, subpass)]\n") +SLANG_RAW(" T SubpassLoad()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.SubpassLoad()\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subpassLoad($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zeroVec = int2(0);\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability StorageImageReadWithoutFormat;\n") +SLANG_RAW(" result:$$T = OpImageRead $this $zeroVec\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T = float4, let isMS:int=1>\n") +SLANG_RAW("typealias SubpassInputMS = __SubpassImpl<T, isMS>;\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T>\n") +SLANG_RAW("extension __SubpassImpl<T, 1>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, subpass)]\n") +SLANG_RAW(" T SubpassLoad(int sample)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"$0.SubpassLoad($1)\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"subpassLoad($0, $1)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let zeroVec = int2(0);\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability StorageImageReadWithoutFormat;\n") +SLANG_RAW(" result:$$T = OpImageRead $this $zeroVec Sample $sample\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Shader Execution Reordering (SER)\n") +SLANG_RAW("///\n") +SLANG_RAW("/// NOTE! This API is currently experimental and may change in the future as SER is made available\n") +SLANG_RAW("/// in different APIs and downstream compilers.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// Based on the NVAPI on D3D12 only currently.\n") +SLANG_RAW("///\n") +SLANG_RAW("/// White paper on SER on NVAPI https://developer.nvidia.com/sites/default/files/akamai/gameworks/ser-whitepaper.pdf\n") +SLANG_RAW("///\n") +SLANG_RAW("/// The NVAPI headers (R520) required for this functionality to work can be found here...\n") +SLANG_RAW("///\n") +SLANG_RAW("/// https://developer.nvidia.com/rtx/path-tracing/nvapi/get-started\n") +SLANG_RAW("///\n") +SLANG_RAW("/// For VK the specification is currently in this PR\n") +SLANG_RAW("///\n") +SLANG_RAW("/// https://github.com/KhronosGroup/GLSL/pull/196/files\n") +SLANG_RAW("\n") +SLANG_RAW("/// Internal helper functions\n") +SLANG_RAW("\n") +SLANG_RAW("// This is a bit of a hack for GLSL HitObjectAttributes\n") +SLANG_RAW("// It relies on [ForceInline] removing the surrounding function and just inserting the *contained* `t` as a global\n") +SLANG_RAW("// The __ref should indicate the desire for the returned value to not be a copy of t, but *t*.\n") +SLANG_RAW("// In practive __ref doesn't have this effect in practice.\n") +SLANG_RAW("// \n") +SLANG_RAW("// We need this to be able access the payload outside of a function (which is all that TraceRay for example needs)\n") +SLANG_RAW("// We access the HitObjectAttributes via this function for the desired type, and it acts *as if* it's just an access\n") +SLANG_RAW("// to the global t.\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("Ref<T> __hitObjectAttributes<T>()\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__vulkanHitObjectAttributes] \n") +SLANG_RAW(" static T t;\n") +SLANG_RAW(" return t;\n") +SLANG_RAW("}\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("__Addr<T> __allocHitObjectAttributes<T>()\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__vulkanHitObjectAttributes]\n") +SLANG_RAW(" static T t;\n") +SLANG_RAW(" return __get_addr(t);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Next is the custom intrinsic that will compute the hitObjectAttributes location\n") +SLANG_RAW("// for GLSL-based targets.\n") +SLANG_RAW("//\n") +SLANG_RAW("__generic<Attributes>\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetVulkanRayTracingPayloadLocation +) +SLANG_RAW(")\n") +SLANG_RAW("int __hitObjectAttributesLocation(__ref Attributes attributes);\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Immutable data type representing a ray hit or a miss. Can be used to invoke hit or miss shading,\n") +SLANG_RAW(" /// or as a key in ReorderThread. Created by one of several methods described below. HitObject\n") +SLANG_RAW(" /// and its related functions are available in raytracing shader types only.\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("[__NonCopyableType]\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HitObjectType +) +SLANG_RAW(")\n") +SLANG_RAW("struct HitObject\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_AllocateOpaqueHandle +) +SLANG_RAW(")\n") +SLANG_RAW(" __init();\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Executes ray traversal (including anyhit and intersection shaders) like TraceRay, but returns the\n") +SLANG_RAW(" /// resulting hit information as a HitObject and does not trigger closesthit or miss shaders.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject TraceRay<payload_t>(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint RayFlags,\n") +SLANG_RAW(" uint InstanceInclusionMask,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" HitObject hitObj;\n") +SLANG_RAW(" __hlslTraceRay(\n") +SLANG_RAW(" AccelerationStructure, \n") +SLANG_RAW(" RayFlags, \n") +SLANG_RAW(" InstanceInclusionMask, \n") +SLANG_RAW(" RayContributionToHitGroupIndex, \n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex, \n") +SLANG_RAW(" MissShaderIndex, \n") +SLANG_RAW(" Ray, \n") +SLANG_RAW(" __forceVarIntoStructTemporarily(Payload),\n") +SLANG_RAW(" hitObj);\n") +SLANG_RAW(" return hitObj;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslTraceRay(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags, // Assumes D3D/VK have some RayFlags values\n") +SLANG_RAW(" InstanceInclusionMask, // cullMask\n") +SLANG_RAW(" RayContributionToHitGroupIndex, // sbtRecordOffset\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex, // sbtRecordStride\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" __rayPayloadLocation(p));\n") +SLANG_RAW(" \n") +SLANG_RAW(" // Write the payload out\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectTraceRayNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $RayFlags\n") +SLANG_RAW(" /**/ $InstanceInclusionMask\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ &p;\n") +SLANG_RAW(" };\n") +SLANG_RAW("\n") +SLANG_RAW(" // Write the payload out\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Executes motion ray traversal (including anyhit and intersection shaders) like TraceRay, but returns the\n") +SLANG_RAW(" /// resulting hit information as a HitObject and does not trigger closesthit or miss shaders.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject TraceMotionRay<payload_t>( \n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint RayFlags, \n") +SLANG_RAW(" uint InstanceInclusionMask, \n") +SLANG_RAW(" uint RayContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MissShaderIndex, \n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __traceMotionRayHLSL(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags,\n") +SLANG_RAW(" InstanceInclusionMask,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __forceVarIntoStructTemporarily(Payload));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslTraceMotionRay(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" RayFlags, // Assumes D3D/VK have some RayFlags values\n") +SLANG_RAW(" InstanceInclusionMask, // cullMask\n") +SLANG_RAW(" RayContributionToHitGroupIndex, // sbtRecordOffset\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex, // sbtRecordStride\n") +SLANG_RAW(" MissShaderIndex,\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __rayPayloadLocation(p));\n") +SLANG_RAW(" \n") +SLANG_RAW(" // Write the payload out\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_ray_tracing_motion_blur\";\n") +SLANG_RAW(" OpCapability RayTracingMotionBlurNV;\n") +SLANG_RAW(" OpHitObjectTraceRayMotionNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $RayFlags\n") +SLANG_RAW(" /**/ $InstanceInclusionMask\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $CurrentTime\n") +SLANG_RAW(" /**/ &p;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" \n") +SLANG_RAW(" // Write the payload out\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" \n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Creates a HitObject representing a hit based on values explicitly passed as arguments, without\n") +SLANG_RAW(" /// tracing a ray. The primitive specified by AccelerationStructure, InstanceIndex, GeometryIndex,\n") +SLANG_RAW(" /// and PrimitiveIndex must exist. The shader table index is computed using the formula used with\n") +SLANG_RAW(" /// TraceRay. The computed index must reference a valid hit group record in the shader table. The\n") +SLANG_RAW(" /// Attributes parameter must either be an attribute struct, such as\n") +SLANG_RAW(" /// BuiltInTriangleIntersectionAttributes, or another HitObject to copy the attributes from.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeHit<attr_t>(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint InstanceIndex,\n") +SLANG_RAW(" uint GeometryIndex,\n") +SLANG_RAW(" uint PrimitiveIndex,\n") +SLANG_RAW(" uint HitKind,\n") +SLANG_RAW(" uint RayContributionToHitGroupIndex,\n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" attr_t attributes)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" HitObject hitObj;\n") +SLANG_RAW(" __hlslMakeHit(\n") +SLANG_RAW(" AccelerationStructure, \n") +SLANG_RAW(" InstanceIndex,\n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" PrimitiveIndex,\n") +SLANG_RAW(" HitKind,\n") +SLANG_RAW(" RayContributionToHitGroupIndex,\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex,\n") +SLANG_RAW(" Ray,\n") +SLANG_RAW(" attributes,\n") +SLANG_RAW(" hitObj);\n") +SLANG_RAW(" return hitObj;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __hitObjectAttributes<attr_t>() = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslMakeHit(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" InstanceIndex,\n") +SLANG_RAW(" PrimitiveIndex,\n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" HitKind,\n") +SLANG_RAW(" RayContributionToHitGroupIndex, /// sbtRecordOffset?\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex, /// sbtRecordStride?\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" __hitObjectAttributesLocation(__hitObjectAttributes<attr_t>()));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __Addr<attr_t> attr = __allocHitObjectAttributes<attr_t>();\n") +SLANG_RAW("\n") +SLANG_RAW(" *attr = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordHitNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $InstanceIndex\n") +SLANG_RAW(" /**/ $PrimitiveIndex\n") +SLANG_RAW(" /**/ $GeometryIndex\n") +SLANG_RAW(" /**/ $HitKind\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $attr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// See MakeHit but handles Motion \n") +SLANG_RAW(" /// Currently only supported on VK\n") +SLANG_RAW(" [ForceInline] \n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeMotionHit<attr_t>( \n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint InstanceIndex, \n") +SLANG_RAW(" uint GeometryIndex, \n") +SLANG_RAW(" uint PrimitiveIndex, \n") +SLANG_RAW(" uint HitKind, \n") +SLANG_RAW(" uint RayContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex, \n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" attr_t attributes)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"MakeMotionHit\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __hitObjectAttributes<attr_t>() = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslMakeMotionHit(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" InstanceIndex,\n") +SLANG_RAW(" PrimitiveIndex,\n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" HitKind,\n") +SLANG_RAW(" RayContributionToHitGroupIndex, /// sbtRecordOffset?\n") +SLANG_RAW(" MultiplierForGeometryContributionToHitGroupIndex, /// sbtRecordStride?\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __hitObjectAttributesLocation(__hitObjectAttributes<attr_t>()));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __Addr<attr_t> attr = __allocHitObjectAttributes<attr_t>();\n") +SLANG_RAW("\n") +SLANG_RAW(" *attr = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_ray_tracing_motion_blur\";\n") +SLANG_RAW(" OpCapability RayTracingMotionBlurNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordHitMotionNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $InstanceIndex\n") +SLANG_RAW(" /**/ $PrimitiveIndex\n") +SLANG_RAW(" /**/ $GeometryIndex\n") +SLANG_RAW(" /**/ $HitKind\n") +SLANG_RAW(" /**/ $RayContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $MultiplierForGeometryContributionToHitGroupIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $CurrentTime\n") +SLANG_RAW(" /**/ $attr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Creates a HitObject representing a hit based on values explicitly passed as arguments, without\n") +SLANG_RAW(" /// tracing a ray. The primitive specified by AccelerationStructure, InstanceIndex, GeometryIndex,\n") +SLANG_RAW(" /// and PrimitiveIndex must exist. The shader table index is explicitly provided as an argument\n") +SLANG_RAW(" /// instead of being computed from the indexing formula used in TraceRay. The provided index must\n") +SLANG_RAW(" /// reference a valid hit group record in the shader table. The Attributes parameter must either be an\n") +SLANG_RAW(" /// attribute struct, such as BuiltInTriangleIntersectionAttributes, or another HitObject to copy the\n") +SLANG_RAW(" /// attributes from.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeHit<attr_t>(\n") +SLANG_RAW(" uint HitGroupRecordIndex,\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" uint InstanceIndex,\n") +SLANG_RAW(" uint GeometryIndex,\n") +SLANG_RAW(" uint PrimitiveIndex,\n") +SLANG_RAW(" uint HitKind,\n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" attr_t attributes)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" HitObject hitObj;\n") +SLANG_RAW(" __hlslMakeHitWithRecordIndex(\n") +SLANG_RAW(" HitGroupRecordIndex, \n") +SLANG_RAW(" AccelerationStructure, \n") +SLANG_RAW(" InstanceIndex,\n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" PrimitiveIndex,\n") +SLANG_RAW(" HitKind,\n") +SLANG_RAW(" Ray,\n") +SLANG_RAW(" attributes,\n") +SLANG_RAW(" hitObj);\n") +SLANG_RAW(" return hitObj;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __hitObjectAttributes<attr_t>() = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslMakeHitWithIndex(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" InstanceIndex, ///? Same as instanceid ?\n") +SLANG_RAW(" PrimitiveIndex,\n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" HitKind, /// Assuming HitKinds are compatible\n") +SLANG_RAW(" HitGroupRecordIndex, /// sbtRecordIndex\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" __hitObjectAttributesLocation(__hitObjectAttributes<attr_t>()));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __Addr<attr_t> attr = __allocHitObjectAttributes<attr_t>();\n") +SLANG_RAW(" *attr = attributes;\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordHitWithIndexNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $InstanceIndex\n") +SLANG_RAW(" /**/ $PrimitiveIndex\n") +SLANG_RAW(" /**/ $GeometryIndex\n") +SLANG_RAW(" /**/ $HitKind\n") +SLANG_RAW(" /**/ $HitGroupRecordIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $attr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" /// See MakeHit but handles Motion \n") +SLANG_RAW(" /// Currently only supported on VK\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeMotionHit<attr_t>( \n") +SLANG_RAW(" uint HitGroupRecordIndex, \n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint InstanceIndex, \n") +SLANG_RAW(" uint GeometryIndex, \n") +SLANG_RAW(" uint PrimitiveIndex, \n") +SLANG_RAW(" uint HitKind, \n") +SLANG_RAW(" RayDesc Ray, \n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" attr_t attributes)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __hitObjectAttributes<attr_t>() = attributes;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslMakeMotionHitWithIndex(\n") +SLANG_RAW(" __return_val,\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" InstanceIndex, ///? Same as instanceid ?\n") +SLANG_RAW(" PrimitiveIndex, \n") +SLANG_RAW(" GeometryIndex,\n") +SLANG_RAW(" HitKind, /// Assuming HitKinds are compatible\n") +SLANG_RAW(" HitGroupRecordIndex, /// sbtRecordIndex\n") +SLANG_RAW(" Ray.Origin,\n") +SLANG_RAW(" Ray.TMin,\n") +SLANG_RAW(" Ray.Direction, \n") +SLANG_RAW(" Ray.TMax,\n") +SLANG_RAW(" CurrentTime,\n") +SLANG_RAW(" __hitObjectAttributesLocation(__hitObjectAttributes<attr_t>()));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Save the attributes\n") +SLANG_RAW(" __Addr<attr_t> attr = __allocHitObjectAttributes<attr_t>();\n") +SLANG_RAW(" *attr = attributes;\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_ray_tracing_motion_blur\";\n") +SLANG_RAW(" OpCapability RayTracingMotionBlurNV;\n") +SLANG_RAW(" OpHitObjectRecordHitWithIndexMotionNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $AccelerationStructure\n") +SLANG_RAW(" /**/ $InstanceIndex\n") +SLANG_RAW(" /**/ $PrimitiveIndex\n") +SLANG_RAW(" /**/ $GeometryIndex\n") +SLANG_RAW(" /**/ $HitKind\n") +SLANG_RAW(" /**/ $HitGroupRecordIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $CurrentTime\n") +SLANG_RAW(" /**/ $attr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Creates a HitObject representing a miss based on values explicitly passed as arguments, without\n") +SLANG_RAW(" /// tracing a ray. The provided shader table index must reference a valid miss record in the shader\n") +SLANG_RAW(" /// table.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeMiss( \n") +SLANG_RAW(" uint MissShaderIndex, \n") +SLANG_RAW(" RayDesc Ray)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($2=NvMakeMiss($0,$1))\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslMakeMiss(__return_val, MissShaderIndex, Ray.Origin, Ray.TMin, Ray.Direction, Ray.TMax);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordMissNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// See MakeMiss but handles Motion \n") +SLANG_RAW(" /// Currently only supported on VK\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeMotionMiss( \n") +SLANG_RAW(" uint MissShaderIndex, \n") +SLANG_RAW(" RayDesc Ray,\n") +SLANG_RAW(" float CurrentTime)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"($3=NvMakeMotionMiss($0,$1,$2))\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslMakeMotionMiss(__return_val, MissShaderIndex, Ray.Origin, Ray.TMin, Ray.Direction, Ray.TMax, CurrentTime);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let origin = Ray.Origin;\n") +SLANG_RAW(" let direction = Ray.Direction;\n") +SLANG_RAW(" let tmin = Ray.TMin;\n") +SLANG_RAW(" let tmax = Ray.TMax;\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_ray_tracing_motion_blur\";\n") +SLANG_RAW(" OpCapability RayTracingMotionBlurNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordMissMotionNV\n") +SLANG_RAW(" /**/ &__return_val\n") +SLANG_RAW(" /**/ $MissShaderIndex\n") +SLANG_RAW(" /**/ $origin\n") +SLANG_RAW(" /**/ $tmin\n") +SLANG_RAW(" /**/ $direction\n") +SLANG_RAW(" /**/ $tmax\n") +SLANG_RAW(" /**/ $CurrentTime;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Creates a HitObject representing 429496726642949671684294967196NOP429496726642949671684294967197 (no operation) which is neither a hit nor a miss. Invoking a\n") +SLANG_RAW(" /// NOP hit object using HitObject::Invoke has no effect. Reordering by hit objects using\n") +SLANG_RAW(" /// ReorderThread will group NOP hit objects together. This can be useful in some reordering\n") +SLANG_RAW(" /// scenarios where future control flow for some threads is known to process neither a hit nor a\n") +SLANG_RAW(" /// miss.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static HitObject MakeNop()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"($0 = NvMakeNop())\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslMakeNop(__return_val);\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectRecordEmptyNV\n") +SLANG_RAW(" /**/ &__return_val;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [require(hlsl, ser)]\n") +SLANG_RAW(" __generic<payload_t>\n") +SLANG_RAW(" static void __InvokeHLSL(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" HitObject HitOrMiss,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvInvokeHitObject\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Invokes closesthit or miss shading for the specified hit object. In case of a NOP HitObject, no\n") +SLANG_RAW(" /// shader is invoked.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void Invoke<payload_t>(\n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure,\n") +SLANG_RAW(" HitObject HitOrMiss,\n") +SLANG_RAW(" inout payload_t Payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __InvokeHLSL(\n") +SLANG_RAW(" AccelerationStructure,\n") +SLANG_RAW(" HitOrMiss,\n") +SLANG_RAW(" __forceVarIntoStructTemporarily(Payload));\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" __glslInvoke(HitOrMiss, __rayPayloadLocation(p));\n") +SLANG_RAW("\n") +SLANG_RAW(" // Write payload result\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__vulkanRayPayload]\n") +SLANG_RAW(" static payload_t p;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Save the payload\n") +SLANG_RAW(" p = Payload;\n") +SLANG_RAW("\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" { \n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV; \n") +SLANG_RAW(" OpHitObjectExecuteShaderNV\n") +SLANG_RAW(" /**/ &HitOrMiss\n") +SLANG_RAW(" /**/ &p;\n") +SLANG_RAW(" };\n") +SLANG_RAW("\n") +SLANG_RAW(" // Write payload result\n") +SLANG_RAW(" Payload = p;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns true if the HitObject encodes a miss, otherwise returns false.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" bool IsMiss()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".IsMiss\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectIsMissNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$bool = OpHitObjectIsMissNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns true if the HitObject encodes a hit, otherwise returns false.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" bool IsHit()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".IsHit\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectIsHitNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$bool = OpHitObjectIsHitNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns true if the HitObject encodes a nop, otherwise returns false.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" bool IsNop()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".IsNop\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectIsEmptyNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$bool = OpHitObjectIsEmptyNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Queries ray properties from HitObject. Valid if the hit object represents a hit or a miss.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" RayDesc GetRayDesc()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \".GetRayDesc\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" RayDesc ray = { __glslGetRayWorldOrigin(), __glslGetTMin(), __glslGetRayWorldDirection(), __glslGetTMax() };\n") +SLANG_RAW(" return ray;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" %origin:$$float3 = OpHitObjectGetWorldRayOriginNV &this;\n") +SLANG_RAW(" %tmin:$$float = OpHitObjectGetRayTMinNV &this;\n") +SLANG_RAW(" %direction:$$float3 = OpHitObjectGetWorldRayDirectionNV &this;\n") +SLANG_RAW(" %tmax:$$float = OpHitObjectGetRayTMaxNV &this;\n") +SLANG_RAW(" result:$$RayDesc = OpCompositeConstruct %origin %tmin %direction %tmax;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Queries shader table index from HitObject. Valid if the hit object represents a hit or a miss.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetShaderTableIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetShaderTableIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetShaderBindingTableRecordIndexNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetShaderBindingTableRecordIndexNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the instance index of a hit. Valid if the hit object represents a hit.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetInstanceIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetInstanceIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetInstanceIdNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetInstanceIdNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the instance ID of a hit. Valid if the hit object represents a hit.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetInstanceID()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetInstanceID\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetInstanceCustomIndexNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetInstanceCustomIndexNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the geometry index of a hit. Valid if the hit object represents a hit.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetGeometryIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetGeometryIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetGeometryIndexNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetGeometryIndexNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the primitive index of a hit. Valid if the hit object represents a hit.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetPrimitiveIndex()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetPrimitiveIndex\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetPrimitiveIndexNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetPrimitiveIndexNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the hit kind. Valid if the hit object represents a hit.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint GetHitKind()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetHitKind\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetHitKindNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint = OpHitObjectGetHitKindNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float4x3 GetWorldToObject()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetWorldToObject\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetWorldToObjectNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$float4x3 = OpHitObjectGetWorldToObjectNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float4x3 GetObjectToWorld()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".GetObjectToWorld\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetObjectToWorldNV($0)\";\n") +SLANG_RAW(" case spirv: \n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$float4x3 = OpHitObjectGetObjectToWorldNV &this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float GetCurrentTime() {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"hitObjectGetCurrentTimeNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$float = OpHitObjectGetCurrentTimeNV &this\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float3 GetObjectRayOrigin() {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"hitObjectGetObjectRayOriginNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$float3 = OpHitObjectGetObjectRayOriginNV &this\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float3 GetObjectRayDirection() {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"hitObjectGetObjectRayDirectionNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$float3 = OpHitObjectGetObjectRayDirectionNV &this\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" uint2 GetShaderRecordBufferHandle() {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"hitObjectGetShaderRecordBufferHandleNV($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" result:$$uint2 = OpHitObjectGetShaderRecordBufferHandleNV &this\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Returns the attributes of a hit. Valid if the hit object represents a hit or a miss.\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" attr_t GetAttributes<attr_t>()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" attr_t v;\n") +SLANG_RAW(" __hlslGetAttributesFromHitObject(v);\n") +SLANG_RAW(" return v;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Work out the location\n") +SLANG_RAW(" int attributeLocation = __hitObjectAttributesLocation(__hitObjectAttributes<attr_t>());\n") +SLANG_RAW("\n") +SLANG_RAW(" // Load the attributes from the location\n") +SLANG_RAW(" __glslGetAttributes(attributeLocation);\n") +SLANG_RAW("\n") +SLANG_RAW(" // Return the attributes\n") +SLANG_RAW(" return __hitObjectAttributes<attr_t>();\n") +SLANG_RAW(" }\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __Addr<attr_t> attr = __allocHitObjectAttributes<attr_t>();\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpHitObjectGetAttributesNV &this $attr;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" return *attr;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" /// Loads a root constant from the local root table referenced by the hit object. Valid if the hit object\n") +SLANG_RAW(" /// represents a hit or a miss. RootConstantOffsetInBytes must be a multiple of 4.\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, ser)]\n") +SLANG_RAW(" uint LoadLocalRootTableConstant(uint RootConstantOffsetInBytes)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \".LoadLocalRootTableConstant\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// \n") +SLANG_RAW(" /// !!!! Internal NVAPI HLSL impl. Not part of interface! !!!!!!!!!!!!\n") +SLANG_RAW(" /// \n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" void __hlslGetAttributesFromHitObject<T>(out T t)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvGetAttributesFromHitObject($0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __hlslMakeHitWithRecordIndex<attr_t>(\n") +SLANG_RAW(" uint HitGroupRecordIndex, \n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint InstanceIndex, \n") +SLANG_RAW(" uint GeometryIndex, \n") +SLANG_RAW(" uint PrimitiveIndex, \n") +SLANG_RAW(" uint HitKind, \n") +SLANG_RAW(" RayDesc Ray, \n") +SLANG_RAW(" attr_t attributes, \n") +SLANG_RAW(" out HitObject hitObj)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvMakeHitWithRecordIndex\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __hlslMakeHit<attr_t>(RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint InstanceIndex, \n") +SLANG_RAW(" uint GeometryIndex, \n") +SLANG_RAW(" uint PrimitiveIndex, \n") +SLANG_RAW(" uint HitKind, \n") +SLANG_RAW(" uint RayContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex, \n") +SLANG_RAW(" RayDesc Ray, \n") +SLANG_RAW(" attr_t attributes, \n") +SLANG_RAW(" out HitObject hitObj)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvMakeHit\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __hlslTraceRay<payload_t>( \n") +SLANG_RAW(" RaytracingAccelerationStructure AccelerationStructure, \n") +SLANG_RAW(" uint RayFlags, \n") +SLANG_RAW(" uint InstanceInclusionMask, \n") +SLANG_RAW(" uint RayContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MultiplierForGeometryContributionToHitGroupIndex, \n") +SLANG_RAW(" uint MissShaderIndex, \n") +SLANG_RAW(" RayDesc Ray, \n") +SLANG_RAW(" inout payload_t Payload,\n") +SLANG_RAW(" out HitObject hitObj)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvTraceRayHitObject\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// \n") +SLANG_RAW(" /// !!!! Internal GLSL GL_NV_shader_invocation_reorder impl. Not part of interface! !!!!!!!!!!!!\n") +SLANG_RAW(" /// \n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeMiss(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" float3 Origin,\n") +SLANG_RAW(" float TMin,\n") +SLANG_RAW(" float3 Direction,\n") +SLANG_RAW(" float TMax)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordMissNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // \"void hitObjectRecordMissNV(hitObjectNV, uint, vec3, float, vec3, float);\"\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_ray_tracing_motion_blur) \n") +SLANG_RAW(" [require(glsl, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeMotionMiss(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" uint MissShaderIndex,\n") +SLANG_RAW(" float3 Origin,\n") +SLANG_RAW(" float TMin,\n") +SLANG_RAW(" float3 Direction,\n") +SLANG_RAW(" float TMax, \n") +SLANG_RAW(" float CurrentTime)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordMissMotionNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeNop(out HitObject hitObj)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordEmptyNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float3 __glslGetRayDirection()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetObjectRayDirectionNV($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" \n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float3 __glslGetRayWorldDirection()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetWorldRayDirectionNV($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float3 __glslGetRayWorldOrigin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetWorldRayOriginNV($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float __glslGetTMax()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetRayTMaxNV($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" float __glslGetTMin()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetRayTMinNV($0)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // \"void hitObjectRecordHitWithIndexNV(hitObjectNV, accelerationStructureEXT,int,int,int,uint,uint,vec3,float,vec3,float,int);\"\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeHitWithIndex(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" int instanceid,\n") +SLANG_RAW(" int primitiveid,\n") +SLANG_RAW(" int geometryindex,\n") +SLANG_RAW(" uint hitKind,\n") +SLANG_RAW(" uint sbtRecordIndex,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" int attributeLocation)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordHitWithIndexNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // \"void hitObjectRecordHitWithIndexMotionNV(hitObjectNV, accelerationStructureEXT,int,int,int,uint,uint,vec3,float,vec3,float,float,int);\"\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" __glsl_extension(GL_NV_ray_tracing_motion_blur)\n") +SLANG_RAW(" [require(glsl, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeMotionHitWithIndex(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" int instanceid,\n") +SLANG_RAW(" int primitiveid,\n") +SLANG_RAW(" int geometryindex,\n") +SLANG_RAW(" uint hitKind,\n") +SLANG_RAW(" uint sbtRecordIndex,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" int attributeLocation)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordHitWithIndexMotionNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // \"void hitObjectRecordHitNV(hitObjectNV,accelerationStructureEXT,int,int,int,uint,uint,uint,vec3,float,vec3,float,int);\"\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeHit(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" int instanceid,\n") +SLANG_RAW(" int primitiveid,\n") +SLANG_RAW(" int geometryindex,\n") +SLANG_RAW(" uint hitKind,\n") +SLANG_RAW(" uint sbtRecordOffset,\n") +SLANG_RAW(" uint sbtRecordStride,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" int attributeLocation)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordHitNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" // \"void hitObjectRecordHitMotionNV(hitObjectNV,accelerationStructureEXT,int,int,int,uint,uint,uint,vec3,float,vec3,float,float,int);\"\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" __glsl_extension(GL_NV_ray_tracing_motion_blur)\n") +SLANG_RAW(" [require(glsl, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslMakeMotionHit(\n") +SLANG_RAW(" out HitObject hitObj,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" int instanceid,\n") +SLANG_RAW(" int primitiveid,\n") +SLANG_RAW(" int geometryindex,\n") +SLANG_RAW(" uint hitKind,\n") +SLANG_RAW(" uint sbtRecordOffset,\n") +SLANG_RAW(" uint sbtRecordStride,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" float CurrentTime,\n") +SLANG_RAW(" int attributeLocation)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectRecordHitMotionNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" \n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" void __glslGetAttributes(int attributeLocation)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectGetAttributesNV($0, $1)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslTraceRay(\n") +SLANG_RAW(" out HitObject hitObject,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" uint rayFlags,\n") +SLANG_RAW(" uint cullMask,\n") +SLANG_RAW(" uint sbtRecordOffset,\n") +SLANG_RAW(" uint sbtRecordStride,\n") +SLANG_RAW(" uint missIndex,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" int payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectTraceRayNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" __glsl_extension(GL_NV_ray_tracing_motion_blur)\n") +SLANG_RAW(" [require(glsl, ser_motion_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslTraceMotionRay(\n") +SLANG_RAW(" out HitObject hitObject,\n") +SLANG_RAW(" RaytracingAccelerationStructure accelerationStructure,\n") +SLANG_RAW(" uint rayFlags,\n") +SLANG_RAW(" uint cullMask,\n") +SLANG_RAW(" uint sbtRecordOffset,\n") +SLANG_RAW(" uint sbtRecordStride,\n") +SLANG_RAW(" uint missIndex,\n") +SLANG_RAW(" float3 origin,\n") +SLANG_RAW(" float Tmin,\n") +SLANG_RAW(" float3 direction,\n") +SLANG_RAW(" float Tmax,\n") +SLANG_RAW(" float currentTime,\n") +SLANG_RAW(" int payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectTraceRayMotionNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW(" [require(glsl, ser_raygen_closesthit_miss)]\n") +SLANG_RAW(" static void __glslInvoke(\n") +SLANG_RAW(" HitObject hitObj,\n") +SLANG_RAW(" int payload)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"hitObjectExecuteShaderNV\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("};\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Reorders threads based on a coherence hint value. NumCoherenceHintBits indicates how many of\n") +SLANG_RAW(" /// the least significant bits of CoherenceHint should be considered during reordering (max: 16).\n") +SLANG_RAW(" /// Applications should set this to the lowest value required to represent all possible values in\n") +SLANG_RAW(" /// CoherenceHint. For best performance, all threads should provide the same value for\n") +SLANG_RAW(" /// NumCoherenceHintBits.\n") +SLANG_RAW(" /// Where possible, reordering will also attempt to retain locality in the thread429496726642949671684294967193s launch indices\n") +SLANG_RAW(" /// (DispatchRaysIndex in DXR).\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, ser_raygen)]\n") +SLANG_RAW("void ReorderThread( uint CoherenceHint, uint NumCoherenceHintBitsFromLSB )\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvReorderThread\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reorderThreadNV\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpReorderThreadWithHintNV $CoherenceHint $NumCoherenceHintBitsFromLSB;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Reorders threads based on a hit object, optionally extended by a coherence hint value. Coherence\n") +SLANG_RAW(" /// hints behave as described in the generic variant of ReorderThread. The maximum number of\n") +SLANG_RAW(" /// coherence hint bits in this variant of ReorderThread is 8. If no coherence hint is desired, set\n") +SLANG_RAW(" /// NumCoherenceHitBits to zero.\n") +SLANG_RAW(" /// Reordering will consider information in the HitObject and coherence hint with the following\n") +SLANG_RAW(" /// priority:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// 1. Shader ID stored in the HitObject\n") +SLANG_RAW(" /// 2. Coherence hint, with the most significant hint bit having highest priority\n") +SLANG_RAW(" /// 3. Spatial information stored in the HitObject\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// That is, ReorderThread will first attempt to group threads whose HitObject references the\n") +SLANG_RAW(" /// same shader ID. (Miss shaders and NOP HitObjects are grouped separately). Within each of these\n") +SLANG_RAW(" /// groups, it will attempt to order threads by the value of their coherence hints. And within ranges\n") +SLANG_RAW(" /// of equal coherence hints, it will attempt to maximize locality in 3D space of the ray hit (if any).\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, ser_raygen)]\n") +SLANG_RAW("void ReorderThread( HitObject HitOrMiss, uint CoherenceHint, uint NumCoherenceHintBitsFromLSB )\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvReorderThread\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reorderThreadNV\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpReorderThreadWithHitObjectNV &HitOrMiss $CoherenceHint $NumCoherenceHintBitsFromLSB;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Is equivalent to\n") +SLANG_RAW(" /// ```\n") +SLANG_RAW(" /// void ReorderThread( HitObject HitOrMiss, uint CoherenceHint, uint NumCoherenceHintBitsFromLSB );\n") +SLANG_RAW(" /// ```\n") +SLANG_RAW(" /// With CoherenceHint and NumCoherenceHintBitsFromLSB as 0, meaning they are ignored.\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("__glsl_extension(GL_EXT_ray_tracing)\n") +SLANG_RAW("__glsl_extension(GL_NV_shader_invocation_reorder)\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, ser_raygen)]\n") +SLANG_RAW("void ReorderThread( HitObject HitOrMiss )\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvReorderThread\";\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"reorderThreadNV\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" spirv_asm \n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_invocation_reorder\";\n") +SLANG_RAW(" OpCapability ShaderInvocationReorderNV;\n") +SLANG_RAW(" OpReorderThreadWithHitObjectNV &HitOrMiss;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("///\n") +SLANG_RAW("/// DebugBreak support \n") +SLANG_RAW("///\n") +SLANG_RAW("/// There doesn't appear to be an equivalent for debugBreak for HLSL\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__specialized_for_target(glsl)\n") +SLANG_RAW("[[vk::spirv_instruction(1, \"NonSemantic.DebugBreak\")]]\n") +SLANG_RAW("void __glslDebugBreak();\n") +SLANG_RAW("\n") +SLANG_RAW("[ForceInline]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl, breakpoint)]\n") +SLANG_RAW("void debugBreak() \n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"/* debugBreak() not currently supported for HLSL */\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"__brkpt()\";\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"SLANG_BREAKPOINT(0)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __glslDebugBreak();\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("// \n") +SLANG_RAW("// Realtime Clock support\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("// https://github.com/KhronosGroup/GLSL/blob/master/extensions/ext/GL_EXT_shader_realtime_clock.txt\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_realtime_clock)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shaderclock)]\n") +SLANG_RAW("uint getRealtimeClockLow()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvGetSpecial( NV_SPECIALOP_GLOBAL_TIMER_LO)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" return getRealtimeClock().x;\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" __intrinsic_asm \"clock\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return getRealtimeClock().x;\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" __intrinsic_asm \"(uint32_t)std::chrono::high_resolution_clock::now().time_since_epoch().count()\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cpp_cuda, shaderclock)]\n") +SLANG_RAW("int64_t __cudaCppGetRealtimeClock()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cpp: __intrinsic_asm \"std::chrono::high_resolution_clock::now().time_since_epoch().count()\";\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"clock64\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("__glsl_extension(GL_EXT_shader_realtime_clock)\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cpp_cuda_glsl_hlsl_spirv, shaderclock)]\n") +SLANG_RAW("uint2 getRealtimeClock()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"uint2(NvGetSpecial(NV_SPECIALOP_GLOBAL_TIMER_LO), NvGetSpecial( NV_SPECIALOP_GLOBAL_TIMER_HI))\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"clockRealtime2x32EXT()\";\n") +SLANG_RAW(" case cuda:\n") +SLANG_RAW(" case cpp:\n") +SLANG_RAW(" int64_t ticks = __cudaCppGetRealtimeClock();\n") +SLANG_RAW(" return uint2(uint(ticks), uint(uint64_t(ticks) >> 32));\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm\n") +SLANG_RAW(" {\n") +SLANG_RAW(" OpCapability ShaderClockKHR;\n") +SLANG_RAW(" OpExtension \"SPV_KHR_shader_clock\";\n") +SLANG_RAW(" result : $$uint2 = OpReadClockKHR Device\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// \n") +SLANG_RAW("// CUDA specific \n") +SLANG_RAW("// \n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda)]\n") +SLANG_RAW("uint3 cudaThreadIdx()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(threadIdx)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda)]\n") +SLANG_RAW("uint3 cudaBlockIdx()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(blockIdx)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__readNone]\n") +SLANG_RAW("[NonUniformReturn]\n") +SLANG_RAW("[require(cuda)]\n") +SLANG_RAW("uint3 cudaBlockDim()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case cuda: __intrinsic_asm \"(blockDim)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// Workgroup cooperation\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// `saturated_cooperation(c, f, s, u)` will call `f(s, u)` if not all lanes in the\n") +SLANG_RAW("// workgroup are currently executing. however if all lanes are saturated, then\n") +SLANG_RAW("// for each unique `s` across all the active lanes `c(s, u)` is called. The\n") +SLANG_RAW("// return value is the one corresponding to the input `s` from this lane.\n") +SLANG_RAW("//\n") +SLANG_RAW("// Adjacent calls to saturated_cooperation are subject to fusion, i.e.\n") +SLANG_RAW("// saturated_cooperation(c1, f1, s, u1);\n") +SLANG_RAW("// saturated_cooperation(c2, f2, s, u2);\n") +SLANG_RAW("// will be transformed to:\n") +SLANG_RAW("// saturated_cooperation(c1c2, f1f2, s, u1u2);\n") +SLANG_RAW("// where\n") +SLANG_RAW("// c1c2 is a function which calls c1(s, u1) and then c2(s, u2);\n") +SLANG_RAW("// f1f2 is a function which calls f1(s, u1) and then f2(s, u2);\n") +SLANG_RAW("//\n") +SLANG_RAW("// When the input differs, calls are fused\n") +SLANG_RAW("// saturated_cooperation(c1, f1, s1, u1);\n") +SLANG_RAW("// saturated_cooperation(c2, f2, s2, u2);\n") +SLANG_RAW("// will be transformed to:\n") +SLANG_RAW("// saturated_cooperation(c1c2, f1f2, s1s2, u1u2);\n") +SLANG_RAW("// where\n") +SLANG_RAW("// s1s2 is a tuple of s1 and s2\n") +SLANG_RAW("// c1c2 is a function which calls c1(s1, u1) and then c2(s2, u2);\n") +SLANG_RAW("// f1f2 is a function which calls f1(s1, u1) and then f2(s2, u2);\n") +SLANG_RAW("// Note that in this case, we will make a call to c1c2 for every unique pair\n") +SLANG_RAW("// s1s2 across all lanes\n") +SLANG_RAW("//\n") +SLANG_RAW("// (This fusion takes place in the fuse-satcoop pass, and as such any changes to\n") +SLANG_RAW("// the signature or behavior of this function should be adjusted for there).\n") +SLANG_RAW("//\n") +SLANG_RAW("[KnownBuiltin(\"saturated_cooperation\")]\n") +SLANG_RAW("func saturated_cooperation<A : __BuiltinType, B, C>(\n") +SLANG_RAW(" cooperate : functype (A, B) -> C,\n") +SLANG_RAW(" fallback : functype (A, B) -> C,\n") +SLANG_RAW(" A input,\n") +SLANG_RAW(" B otherArg)\n") +SLANG_RAW(" -> C\n") +SLANG_RAW("{\n") +SLANG_RAW(" return saturated_cooperation_using(cooperate, fallback, __WaveMatchBuitin<A>, __WaveReadLaneAtBuiltin<A>, input, otherArg);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// These two functions are a temporary (circa May 2023) workaround to the fact\n") +SLANG_RAW("// that we can't deduce which overload to pass to saturated_cooperation_using\n") +SLANG_RAW("// in the call above\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("func __WaveMatchBuitin<T : __BuiltinType>(T t) -> uint4\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveMatch(t);\n") +SLANG_RAW("}\n") +SLANG_RAW("[__unsafeForceInlineEarly]\n") +SLANG_RAW("func __WaveReadLaneAtBuiltin<T : __BuiltinType>(T t, int i) -> T\n") +SLANG_RAW("{\n") +SLANG_RAW(" return WaveReadLaneAt(t, i);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// saturated_cooperation, but you're able to specify manually the functions:\n") +SLANG_RAW("//\n") +SLANG_RAW("// waveMatch: a function to return a mask of lanes with the same input as this one\n") +SLANG_RAW("// broadcast: a function which returns the value passed into it on the specified lane\n") +SLANG_RAW("//\n") +SLANG_RAW("[KnownBuiltin(\"saturated_cooperation_using\")]\n") +SLANG_RAW("func saturated_cooperation_using<A, B, C>(\n") +SLANG_RAW(" cooperate : functype (A, B) -> C,\n") +SLANG_RAW(" fallback : functype (A, B) -> C,\n") +SLANG_RAW(" waveMatch : functype (A) -> uint4,\n") +SLANG_RAW(" broadcast : functype (A, int) -> A,\n") +SLANG_RAW(" A input,\n") +SLANG_RAW(" B otherArg)\n") +SLANG_RAW(" -> C\n") +SLANG_RAW("{\n") +SLANG_RAW(" const bool isWaveSaturated = WaveActiveCountBits(true) == WaveGetLaneCount();\n") +SLANG_RAW(" if(isWaveSaturated)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" let lanesWithSameInput = waveMatch(input).x;\n") +SLANG_RAW(" // Keep least significant lane in our set\n") +SLANG_RAW(" let ourRepresentative = lanesWithSameInput & -lanesWithSameInput;\n") +SLANG_RAW(" // The representative lanes for all lanes\n") +SLANG_RAW(" var allRepresentatives = WaveActiveBitOr(ourRepresentative);\n") +SLANG_RAW("\n") +SLANG_RAW(" C ret;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Iterate over set bits in mask from low to high.\n") +SLANG_RAW(" // In each iteration the lowest bit is cleared.\n") +SLANG_RAW(" while(bool(allRepresentatives))\n") +SLANG_RAW(" {\n") +SLANG_RAW(" // Broadcast input across warp.\n") +SLANG_RAW(" let laneIdx = firstbitlow(allRepresentatives);\n") +SLANG_RAW(" let uniformInput = broadcast(input, int(laneIdx));\n") +SLANG_RAW("\n") +SLANG_RAW(" // All lanes perform some cooperative computation with dynamic\n") +SLANG_RAW(" // uniform input\n") +SLANG_RAW(" C c = cooperate(uniformInput, otherArg);\n") +SLANG_RAW("\n") +SLANG_RAW(" // Update our return value until it\n") +SLANG_RAW(" if(bool(allRepresentatives & ourRepresentative))\n") +SLANG_RAW(" ret = c;\n") +SLANG_RAW("\n") +SLANG_RAW(" // Clear the lowest bit\n") +SLANG_RAW(" allRepresentatives &= allRepresentatives - 1;\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" return ret;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" else\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return fallback(input, otherArg);\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("\n") + +// The NVAPI operations are defined to take the space/register +// indices of their texture and sampler parameters, rather than +// taking the texture/sampler objects directly. +// +// In order to support this approach, we need intrinsics that +// can magically fetch the binding information for a resource. +// +// TODO: These operations are kind of *screaming* for us to +// have a built-in `interface` that all of the opaque resource +// types conform to, so that we can define builtins that work +// for any resource type. +SLANG_RAW("#line 20055 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetRegisterSpace +) +SLANG_RAW(") uint __getRegisterSpace<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>(__TextureImpl<T,Shape,isArray,isMS,sampleCount,access,isShadow,isCombined,format> texture);\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetRegisterSpace +) +SLANG_RAW(") uint __getRegisterSpace(SamplerState sampler);\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetRegisterIndex +) +SLANG_RAW(") uint __getRegisterIndex<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>(__TextureImpl<T,Shape,isArray,isMS,sampleCount,access,isShadow,isCombined,format> texture);\n") +SLANG_RAW("__intrinsic_op(") +SLANG_SPLICE(kIROp_GetRegisterIndex +) +SLANG_RAW(") uint __getRegisterIndex(SamplerState sampler);\n") +SLANG_RAW("\n") +SLANG_RAW("\n") + +// +// Texture Footprint Queries +// +// This section introduces the types and methods related +// to the `GL_NV_shader_texture_footprint` GLSL extension, +// and the matching NVAPI operations. +// +// Footprint queries are allowed on both 2D and 3D textures, +// and are structurally similar for the two, so we will +// use a meta-loop to deduplicate the code for the two +// cases. +// + +// A footprint query yields a data structure +// that describes blocks of texels that +// conservatively cover the data that might +// be fetched in the query. +// +// A given sampling operation might access two +// mip levels of a texture when, e.g., trilinear +// filtering is on. A footprint query may ask for +// a footprint in either the coarse or fine level +// of the pair. +// +// We first define a `struct` type that closely maps +// to how a footprint is defined for each of the +// implementations we support, and then wrap that +// in a derived `struct` that includes the extra +// data that is returned by the GLSL API via the +// function reuslt. +// +SLANG_RAW("#line 20096 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("[__NoSideEffect]\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, texturefootprint)]\n") +SLANG_RAW("vector<uint, ND> __textureFootprintGetAnchor<let ND:int>(__TextureFootprintData<ND> data, int nd)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvFootprintExtractAnchorTileLoc$!1D($0)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.anchor\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<uint,ND> = OpCompositeExtract $data 1;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("[__NoSideEffect]\n") +SLANG_RAW("[__requiresNVAPI]\n") +SLANG_RAW("[require(glsl_hlsl_spirv, texturefootprint)]\n") +SLANG_RAW("vector<uint, ND> __textureFootprintGetOffset<let ND:int>(__TextureFootprintData<ND> data, int nd)\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvFootprintExtractOffset$!1D($0)\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.offset\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$vector<uint,ND> = OpCompositeExtract $data 2;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_TextureFootprintType +) +SLANG_RAW(")\n") +SLANG_RAW("[require(glsl_hlsl_spirv, texturefootprint)]\n") +SLANG_RAW("struct __TextureFootprintData<let ND:int>\n") +SLANG_RAW("{\n") +SLANG_RAW(" typealias Anchor = vector<uint, ND>;\n") +SLANG_RAW(" typealias Offset = vector<uint, ND>;\n") +SLANG_RAW(" typealias Mask = uint2;\n") +SLANG_RAW(" typealias LOD = uint;\n") +SLANG_RAW(" typealias Granularity = uint;\n") +SLANG_RAW("\n") +SLANG_RAW(" property anchor : Anchor\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" get { return __textureFootprintGetAnchor(this, ND); }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" property offset : Offset\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" get { return __textureFootprintGetOffset(this, ND); }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" property mask : Mask\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvFootprintExtractBitmask\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.mask\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$Mask = OpCompositeExtract $this 3;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" property lod : LOD\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvFootprintExtractLOD\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.lod\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$LOD = OpCompositeExtract $this 4;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" property granularity : Granularity\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"NvFootprintExtractReturnGran\";\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0.granularity\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$Granularity = OpCompositeExtract $this 5;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("struct TextureFootprint<let ND:int> : __TextureFootprintData<ND>\n") +SLANG_RAW("{\n") +SLANG_RAW(" bool _isSingleLevel;\n") +SLANG_RAW("\n") +SLANG_RAW(" property isSingleLevel : bool\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" get\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return _isSingleLevel;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("typealias TextureFootprint2D = TextureFootprint<2>;\n") +SLANG_RAW("typealias TextureFootprint3D = TextureFootprint<3>;\n") +SLANG_RAW("\n") + +// We define the new operations via an `extension` +// on the relevant texture type(s), rather than +// further clutter the original type declarations. +SLANG_RAW("#line 20242 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape: __ITextureShape, let sampleCount:int, let isShadow:int, let format:int>\n") +SLANG_RAW("extension __TextureImpl<T,Shape,0,0,sampleCount,0,isShadow,0,format>\n") +SLANG_RAW("{\n") + +// We introduce a few convenience type aliases here, +// which both keep our declarations simpler and easier +// to understand, but which might *also* be useful to +// users of the stdlib, so that they can write things +// like `Texture2D.Footprint`, and also have auto-complete +// help them find such members. +// +// TODO: The `Coords` type really ought to be something +// defined on the base texture types, rather than via +// this `extension`. +SLANG_RAW("#line 20258 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" typealias Coords = vector<float, Shape.dimensions>;\n") +SLANG_RAW(" typealias Footprint = TextureFootprint<Shape.dimensions>;\n") +SLANG_RAW(" typealias __FootprintData = __TextureFootprintData<Shape.dimensions>;\n") +SLANG_RAW(" typealias FootprintGranularity = Footprint.Granularity;\n") +SLANG_RAW("\n") + +// For the GLSL extension, the choice between the +// coarse and fine level is modeled as a `bool` +// parameter to the query operation(s). We define +// the GLSL functions here as intrinsics, so that +// we can refer to them later in the definitions +// of our stdlib operaitons. +// +// Note: despite the GLSL extension defining the `granularity` +// member of the query result as having type `uint`, the +// function signatures all take `int` parameters for the +// granularity instead. +// +SLANG_RAW("#line 20277 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprint)]\n") +SLANG_RAW(" bool __queryFootprintGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprint)]\n") +SLANG_RAW(" bool __queryFootprintGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint,\n") +SLANG_RAW(" float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel Bias $bias;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprintclamp)]\n") +SLANG_RAW(" bool __queryFootprintClampGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lodClamp,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintClampNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel MinLod $lodClamp;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprintclamp)]\n") +SLANG_RAW(" bool __queryFootprintClampGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lodClamp,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint,\n") +SLANG_RAW(" float bias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintClampNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel Bias|MinLod $bias $lodClamp;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprint)]\n") +SLANG_RAW(" bool __queryFootprintLodGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lod,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintLodNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel Lod $lod;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") + + // Texture sampling with gradient is only available for 2D textures. +SLANG_RAW("#line 20422 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprint)]\n") +SLANG_RAW(" bool __queryFootprintGradGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" Coords dx,\n") +SLANG_RAW(" Coords dy,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintGradNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel Grad $dx $dy;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_NV_shader_texture_footprint)\n") +SLANG_RAW(" __glsl_extension(GL_ARB_sparse_texture_clamp)\n") +SLANG_RAW(" [require(glsl_spirv, texturefootprintclamp)]\n") +SLANG_RAW(" bool __queryFootprintGradClampGLSL(\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" Coords dx,\n") +SLANG_RAW(" Coords dy,\n") +SLANG_RAW(" float lodClamp,\n") +SLANG_RAW(" int granularity,\n") +SLANG_RAW(" bool useCoarseLevel,\n") +SLANG_RAW(" out __FootprintData footprint)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"textureFootprintGradClampNV($p, $*2)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ImageFootprintNV;\n") +SLANG_RAW(" OpCapability MinLod;\n") +SLANG_RAW(" OpExtension \"SPV_NV_shader_image_footprint\";\n") +SLANG_RAW(" %resultVal:$$__FootprintData = OpImageSampleFootprintNV &this $coords $granularity $useCoarseLevel Grad|MinLod $dx $dy $lodClamp;\n") +SLANG_RAW(" OpStore &footprint %resultVal;\n") +SLANG_RAW(" result:$$bool = OpCompositeExtract %resultVal 0;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") + + // End texture2D specific functions. +SLANG_RAW("#line 20484 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") + +// The NVAPI texture query operations encode the choice +// between coarse and fine levels as part of the function +// name, and so we are forced to match this convention +// if we want to provide a more portable API. +// +// TODO: We could conceivably define the functions to use +// a parameter for the coarse/fine choice, which is required +// to be `constexpr` for the HLSL/NVAPI target. +// +static const struct LevelChoice +{ +char const* name; +char const* isCoarseVal; +} kLevelChoices[] = +{ + { "Coarse", "true" }, + { "Fine", "false" }, +}; +for(auto levelChoice : kLevelChoices) +{ + auto CoarseOrFine = levelChoice.name; + auto isCoarseVal = levelChoice.isCoarseVal; + +// We now go ahead and define the intrinsics provided by NVAPI, +// which have a very different signature from the GLSL ones. +// +// Note: the NVAPI functions also support an optional texel +// offset parameter. For now we are not including overloads +// with that parameter, since they have no equivalent in +// the GLSL extension. +// +SLANG_RAW("#line 20519 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, texturefootprint)]\n") +SLANG_RAW(" static __FootprintData __queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("NVAPI(\n") +SLANG_RAW(" int nd,\n") +SLANG_RAW(" uint textureSpace,\n") +SLANG_RAW(" uint textureIndex,\n") +SLANG_RAW(" uint samplerSpace,\n") +SLANG_RAW(" uint samplerIndex,\n") +SLANG_RAW(" float3 coords,\n") +SLANG_RAW(" FootprintGranularity granularity, \n") +SLANG_RAW(" out uint isSingleLod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("($1, $2, $3, $4, NV_EXTN_TEXTURE_$!0D, $*5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, texturefootprint)]\n") +SLANG_RAW(" static __FootprintData __queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("BiasNVAPI(\n") +SLANG_RAW(" int nd,\n") +SLANG_RAW(" uint textureSpace,\n") +SLANG_RAW(" uint textureIndex,\n") +SLANG_RAW(" uint samplerSpace,\n") +SLANG_RAW(" uint samplerIndex,\n") +SLANG_RAW(" float3 coords,\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" float lodBias, \n") +SLANG_RAW(" out uint isSingleLod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Bias($1, $2, $3, $4, NV_EXTN_TEXTURE_$!0D, $*5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, texturefootprint)]\n") +SLANG_RAW(" static __FootprintData __queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("LevelNVAPI(\n") +SLANG_RAW(" int nd,\n") +SLANG_RAW(" uint textureSpace,\n") +SLANG_RAW(" uint textureIndex,\n") +SLANG_RAW(" uint samplerSpace,\n") +SLANG_RAW(" uint samplerIndex,\n") +SLANG_RAW(" float3 coords,\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" float lod, \n") +SLANG_RAW(" out uint isSingleLod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Level($1, $2, $3, $4, NV_EXTN_TEXTURE_$!0D, $*5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [require(hlsl, texturefootprint)]\n") +SLANG_RAW(" static __FootprintData __queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("GradNVAPI(\n") +SLANG_RAW(" int nd,\n") +SLANG_RAW(" uint textureSpace,\n") +SLANG_RAW(" uint textureIndex,\n") +SLANG_RAW(" uint samplerSpace,\n") +SLANG_RAW(" uint samplerIndex,\n") +SLANG_RAW(" float3 coords,\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" float3 dx,\n") +SLANG_RAW(" float3 dy, \n") +SLANG_RAW(" out uint isSingleLod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case hlsl: __intrinsic_asm \"NvFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Grad($1, $2, $3, $4, NV_EXTN_TEXTURE_$!0D, $*5)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + +// We now define the portable operations that will be officially +// supported by the standard library. For each operation, we +// need to provide both a version that maps to the GLSL extension, +// and a version that uses the NVAPI functions. +// +// Some function variations are only available with one extension +// or the other, so we try our best to only define them where +// each is available. +// +// Note that these functions cannot be marked as [ForceInline] for now +// because the texture resource may get removed after DCE, since the only +// use of those resources are done through __GetRegisterIndex/Space, which is +// replaced early with their binding slot in the compilation process. +// Not inlining these function is a quick way to make sure the texture always +// has live uses. +// +SLANG_RAW("#line 20618 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.Sample(sampler, coords);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintGLSL(sampler, coords, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW("\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" uint isSingleLod = 0;\n") +SLANG_RAW(" Footprint footprint = {__queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("NVAPI(\n") +SLANG_RAW(" Shape.dimensions,\n") +SLANG_RAW(" __getRegisterSpace(this), __getRegisterIndex(this),\n") +SLANG_RAW(" __getRegisterSpace(sampler), __getRegisterIndex(sampler),\n") +SLANG_RAW(" __vectorReshape<3>(coords), granularity, /* out */isSingleLod), false};\n") +SLANG_RAW(" footprint._isSingleLevel = (isSingleLod != 0);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleBias(sampler, coords, lodBias);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Bias(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lodBias)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintGLSL(sampler, coords, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint, lodBias);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" uint isSingleLod = 0;\n") +SLANG_RAW(" Footprint footprint = {__queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("BiasNVAPI(\n") +SLANG_RAW(" Shape.dimensions,\n") +SLANG_RAW(" __getRegisterSpace(this), __getRegisterIndex(this),\n") +SLANG_RAW(" __getRegisterSpace(sampler), __getRegisterIndex(sampler),\n") +SLANG_RAW(" __vectorReshape<3>(coords), granularity, lodBias, /* out */isSingleLod), false}; \n") +SLANG_RAW(" \n") +SLANG_RAW(" footprint._isSingleLevel = (isSingleLod != 0);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleClamp(sampler, coords, lodClamp);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Clamp(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lodClamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintClampGLSL(sampler, coords, lodClamp, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleBiasClamp(sampler, coords, lodBias, lodClamp);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("BiasClamp(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lodBias,\n") +SLANG_RAW(" float lodClamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintClampGLSL(sampler, coords, lodClamp, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint, lodBias);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleLevel(sampler, coords, lod);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Level(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" float lod)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintLodGLSL(sampler, coords, lod, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" uint isSingleLod = 0;\n") +SLANG_RAW(" Footprint footprint = {__queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("LevelNVAPI(\n") +SLANG_RAW(" Shape.dimensions,\n") +SLANG_RAW(" __getRegisterSpace(this), __getRegisterIndex(this),\n") +SLANG_RAW(" __getRegisterSpace(sampler), __getRegisterIndex(sampler),\n") +SLANG_RAW(" __vectorReshape<3>(coords), granularity, lod, /* out */isSingleLod), false};\n") +SLANG_RAW(" \n") +SLANG_RAW(" footprint._isSingleLevel = (isSingleLod != 0);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + // TODO: Texture sampling with gradient is only available for 2D textures. +SLANG_RAW("#line 20772 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleGrad(sampler, coords, dx, dy);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect] [ForceInline]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("Grad(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" Coords dx,\n") +SLANG_RAW(" Coords dy)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintGradGLSL(sampler, coords, dx, dy, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" uint isSingleLod = 0;\n") +SLANG_RAW(" Footprint footprint = {__queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("GradNVAPI(\n") +SLANG_RAW(" Shape.dimensions,\n") +SLANG_RAW(" __getRegisterSpace(this), __getRegisterIndex(this),\n") +SLANG_RAW(" __getRegisterSpace(sampler), __getRegisterIndex(sampler),\n") +SLANG_RAW(" __vectorReshape<3>(coords), granularity, __vectorReshape<3>(dx), __vectorReshape<3>(dy), /* out */isSingleLod), false};\n") +SLANG_RAW("\n") +SLANG_RAW(" footprint._isSingleLevel = (isSingleLod != 0);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" /// Query the footprint that would be accessed by a texture sampling operation.\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// This operation queries the footprint that would be accessed\n") +SLANG_RAW(" /// by a comparable call to:\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" /// t.SampleGradClamp(sampler, coords, dx, dy, lodClamp);\n") +SLANG_RAW(" ///\n") +SLANG_RAW(" [__NoSideEffect][ForceInline]\n") +SLANG_RAW(" Footprint queryFootprint") +SLANG_SPLICE(CoarseOrFine +) +SLANG_RAW("GradClamp(\n") +SLANG_RAW(" FootprintGranularity granularity,\n") +SLANG_RAW(" SamplerState sampler,\n") +SLANG_RAW(" Coords coords,\n") +SLANG_RAW(" Coords dx,\n") +SLANG_RAW(" Coords dy,\n") +SLANG_RAW(" float lodClamp)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" Footprint footprint;\n") +SLANG_RAW(" footprint._isSingleLevel = __queryFootprintGradClampGLSL(sampler, coords, dx, dy, lodClamp, granularity, ") +SLANG_SPLICE(isCoarseVal +) +SLANG_RAW(", footprint);\n") +SLANG_RAW(" return footprint;\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") + + // TODO: end texture2D specific functions. +SLANG_RAW("#line 20837 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") + +} +SLANG_RAW("#line 20841 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("} // extension\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("//<T, Shape: __ITextureShape, let isArray:int, let isMS:int, let sampleCount:int, let access:int, let isShadow:int, let isCombined:int, let format:int>\n") +SLANG_RAW("__generic<Shape:__ITextureShape1D2D3D, let format : int>\n") +SLANG_RAW("extension __TextureImpl<float, Shape, 0, 0, 0, ") +SLANG_SPLICE(kStdlibResourceAccessReadWrite +) +SLANG_RAW(", 0, 0, format>\n") +SLANG_RAW("{\n") +SLANG_RAW(" [__requiresNVAPI]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" __glsl_extension(GL_EXT_shader_atomic_float)\n") +SLANG_RAW(" [require(glsl_hlsl_spirv, atomic_glsl_hlsl_nvapi_cuda_metal_float1)]\n") +SLANG_RAW(" void InterlockedAddF32(vector<uint, Shape.dimensions> coord, float value, out float originalValue)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" originalValue = __atomicAdd(this[coord], value);\n") +SLANG_RAW(" return;\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$3 = imageAtomicAdd($0, $1, $2)\";\n") +SLANG_RAW(" case hlsl:\n") +SLANG_RAW(" __intrinsic_asm \"$3 = NvInterlockedAddFp32($0, $1, $2)\";\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" float InterlockedAddF32(vector<uint, Shape.dimensions> coord, float value)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" float originalValue;\n") +SLANG_RAW(" InterlockedAddF32(coord, value, originalValue);\n") +SLANG_RAW(" return originalValue;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("// Buffer Pointer\n") +SLANG_RAW("\n") +SLANG_RAW("namespace vk\n") +SLANG_RAW("{\n") +SLANG_RAW(" // Partial implementation of the vk::buffer_ref proposal:\n") +SLANG_RAW(" // https://github.com/microsoft/hlsl-specs/blob/main/proposals/0010-vk-buffer-ref.md\n") +SLANG_RAW(" struct BufferPointer<T, let Alignment : int = 0>\n") +SLANG_RAW(" {\n") +SLANG_RAW(" T *_ptr;\n") +SLANG_RAW(" [ForceInline] __init(T *ptr) { _ptr = ptr; }\n") +SLANG_RAW(" [ForceInline] __init(uint64_t val) { _ptr = (T *)val; }\n") +SLANG_RAW(" [ForceInline] Ref<T> Get() { return *_ptr; }\n") +SLANG_RAW(" [ForceInline] T *getPtr() { return _ptr;}\n") +SLANG_RAW(" }\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" BufferPointer<U, alignment> static_pointer_cast<U, let alignment : int = 0, T, let a : int>(BufferPointer<T, a> src)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return BufferPointer<U, alignment>((U*)(src.getPtr()));\n") +SLANG_RAW(" }\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" BufferPointer<U, alignment> reinterpret_pointer_cast<U, let alignment : int = 0, T, let a : int>(BufferPointer<T, a> src)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" return BufferPointer<U, alignment>((U *)(src.getPtr()));\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("attribute_syntax[vk_aliased_pointer] : VkAliasedPointerAttribute;\n") +SLANG_RAW("attribute_syntax[vk_restrict_pointer] : VkRestrictPointerAttribute;\n") +SLANG_RAW("\n") +SLANG_RAW("extension uint64_t\n") +SLANG_RAW("{\n") +SLANG_RAW(" __init<T, let alignment : int>(vk::BufferPointer<T, alignment> ptr)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" this = (uint64_t)ptr._ptr;\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, let Alignment : int = 16>\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_HLSLConstBufferPointerType +) +SLANG_RAW(")\n") +SLANG_RAW("__glsl_extension(GL_EXT_buffer_reference)\n") +SLANG_RAW("__magic_type(ConstBufferPointerType)\n") +SLANG_RAW("[require(glsl_spirv, bufferreference)]\n") +SLANG_RAW("struct ConstBufferPointer\n") +SLANG_RAW("{\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_buffer_reference)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" T get()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$0._data\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$T = OpLoad $this Aligned !Alignment;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW(" __subscript(int index) -> T\n") +SLANG_RAW(" {\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" get {return ConstBufferPointer<T>.fromUInt(toUInt() + __naturalStrideOf<T>() * index).get(); }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_shader_explicit_arithmetic_types_int64)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_buffer_reference)\n") +SLANG_RAW(" [require(glsl_spirv, bufferreference_int64)]\n") +SLANG_RAW(" static ConstBufferPointer<T> fromUInt(uint64_t val)\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"$TR($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$ConstBufferPointer<T> = OpConvertUToPtr $val;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_shader_explicit_arithmetic_types_int64)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_buffer_reference)\n") +SLANG_RAW(" [require(glsl_spirv, bufferreference_int64)]\n") +SLANG_RAW(" uint64_t toUInt()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"uint64_t($0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" result:$$uint64_t = OpConvertPtrToU $this;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("\n") +SLANG_RAW(" __glsl_version(450)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_shader_explicit_arithmetic_types_int64)\n") +SLANG_RAW(" __glsl_extension(GL_EXT_buffer_reference)\n") +SLANG_RAW(" [__NoSideEffect]\n") +SLANG_RAW(" [ForceInline]\n") +SLANG_RAW(" [require(glsl_spirv, bufferreference_int64)]\n") +SLANG_RAW(" bool isValid()\n") +SLANG_RAW(" {\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl:\n") +SLANG_RAW(" __intrinsic_asm \"(uint64_t($0) != 0)\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" uint64_t zero = 0ULL;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" %ptrval:$$uint64_t = OpConvertPtrToU $this;\n") +SLANG_RAW(" result:$$bool = OpINotEqual %ptrval $zero;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("//\n") +SLANG_RAW("// HLSL-like dynamic resources\n") +SLANG_RAW("// https://microsoft.github.io/DirectX-Specs/d3d/HLSL_SM_6_6_DynamicResources.html\n") +SLANG_RAW("//\n") +SLANG_RAW("// For Khronos targets, `__DynamicResource` can be used to declare \"untyped\" global bindings as\n") +SLANG_RAW("// usual (e.g. unsized arrays for descriptor indexing), which will then be materialized into \n") +SLANG_RAW("// new aliased bindings for each distinct cast type.\n") +SLANG_RAW("//\n") +SLANG_RAW("\n") +SLANG_RAW("__magic_type(DynamicResourceType)\n") +SLANG_RAW("__intrinsic_type(") +SLANG_SPLICE(kIROp_DynamicResourceType +) +SLANG_RAW(")\n") +SLANG_RAW("struct __DynamicResource<let kind = __DynamicResourceKind.General>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CastDynamicResource +) +SLANG_RAW(")\n") +SLANG_RAW(" T as<T : __IDynamicResourceCastable<kind>>();\n") +SLANG_RAW("}\n") +SLANG_RAW("interface __IDynamicResourceCastable<let kind = __DynamicResourceKind.General>\n") +SLANG_RAW("{\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("enum __DynamicResourceKind\n") +SLANG_RAW("{\n") +SLANG_RAW(" General = 0, // CBV_SRV_UAV\n") +SLANG_RAW(" Sampler = 1\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__generic<T, Shape : __ITextureShape, let isArray : int, let isMS : int, let sampleCount : int, let access : int, let isShadow : int, let isCombined : int, let format : int>\n") +SLANG_RAW("extension __TextureImpl<T, Shape, isArray, isMS, sampleCount, access, isShadow, isCombined, format> : __IDynamicResourceCastable<__DynamicResourceKind.General>\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CastDynamicResource +) +SLANG_RAW(")\n") +SLANG_RAW(" __implicit_conversion(") +SLANG_SPLICE(kConversionCost_GenericParamUpcast +) +SLANG_RAW(")\n") +SLANG_RAW(" __init(__DynamicResource res);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +const char* kDynamicResourceCastableTypes[] = { + "StructuredBuffer<T, L>", "RWStructuredBuffer<T, L>", + "AppendStructuredBuffer<T, L>", "ConsumeStructuredBuffer<T, L>", "RasterizerOrderedStructuredBuffer<T, L>", + "ByteAddressBuffer", "RWByteAddressBuffer", "RasterizerOrderedByteAddressBuffer", + + "SamplerState", "SamplerComparisonState", + + "ConstantBuffer<T>", "TextureBuffer<T>", +}; + +for (auto typeName : kDynamicResourceCastableTypes) { + auto kind = strstr(typeName, "Sampler") ? "Sampler" : "General"; + + if (strstr(typeName, "StructuredBuffer<T, L>")) + sb << "__generic<T, L : IBufferDataLayout = DefaultDataLayout>\n"; + else if (strstr(typeName, "Buffer<T>")) + sb << "__generic<T>\n"; +SLANG_RAW("#line 21053 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("extension ") +SLANG_SPLICE(typeName +) +SLANG_RAW(" : __IDynamicResourceCastable<__DynamicResourceKind.") +SLANG_SPLICE(kind +) +SLANG_RAW(">\n") +SLANG_RAW("{\n") +SLANG_RAW(" __intrinsic_op(") +SLANG_SPLICE(kIROp_CastDynamicResource +) +SLANG_RAW(")\n") +SLANG_RAW(" __implicit_conversion(") +SLANG_SPLICE(kConversionCost_GenericParamUpcast +) +SLANG_RAW(")\n") +SLANG_RAW(" __init(__DynamicResource res);\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") + +} +SLANG_RAW("#line 21064 \"hlsl.meta.slang\"") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("__glsl_extension(GL_ARB_shader_clock)\n") +SLANG_RAW("[require(glsl_spirv, GL_ARB_shader_clock)]\n") +SLANG_RAW("uint2 clock2x32ARB()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clock2x32ARB\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const uint32_t scopeId_subgroup = 3;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ShaderClockKHR;\n") +SLANG_RAW(" OpExtension \"SPV_KHR_shader_clock\";\n") +SLANG_RAW(" result:$$uint2 = OpReadClockKHR $scopeId_subgroup;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("__glsl_version(450)\n") +SLANG_RAW("__glsl_extension(GL_ARB_shader_clock)\n") +SLANG_RAW("__glsl_extension(GL_ARB_gpu_shader_int64)\n") +SLANG_RAW("[require(glsl_spirv, GL_ARB_shader_clock64)]\n") +SLANG_RAW("uint64_t clockARB()\n") +SLANG_RAW("{\n") +SLANG_RAW(" __target_switch\n") +SLANG_RAW(" {\n") +SLANG_RAW(" case glsl: __intrinsic_asm \"clockARB\";\n") +SLANG_RAW(" case spirv:\n") +SLANG_RAW(" const uint32_t scopeId_subgroup = 3;\n") +SLANG_RAW(" return spirv_asm {\n") +SLANG_RAW(" OpCapability ShaderClockKHR;\n") +SLANG_RAW(" OpExtension \"SPV_KHR_shader_clock\";\n") +SLANG_RAW(" result:$$uint64_t = OpReadClockKHR $scopeId_subgroup;\n") +SLANG_RAW(" };\n") +SLANG_RAW(" }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("extension<T, L : IBufferDataLayout> StructuredBuffer<T, L> : IArray<T>\n") +SLANG_RAW("{\n") +SLANG_RAW(" int getCount() { uint count; uint stride; this.GetDimensions(count, stride); return count; }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("extension<T, L : IBufferDataLayout> RWStructuredBuffer<T, L> : IRWArray<T>\n") +SLANG_RAW("{\n") +SLANG_RAW(" int getCount() { uint count; uint stride; this.GetDimensions(count, stride); return count; }\n") +SLANG_RAW("}\n") +SLANG_RAW("\n") +SLANG_RAW("extension<T, L : IBufferDataLayout> RasterizerOrderedStructuredBuffer<T, L> : IRWArray<T>\n") +SLANG_RAW("{\n") +SLANG_RAW(" int getCount() { uint count; uint stride; this.GetDimensions(count, stride); return count; }\n") +SLANG_RAW("}\n") diff --git a/source/slang/slang-check-expr.cpp b/source/slang/slang-check-expr.cpp index 1b9725a8c..bdf8721e9 100644 --- a/source/slang/slang-check-expr.cpp +++ b/source/slang/slang-check-expr.cpp @@ -436,6 +436,26 @@ namespace Slang // of a GLSL buffer interface block which isn't marked as // read_only expr->type.isLeftValue = isMutableGLSLBufferBlockVarExpr(baseExpr) && (expr->type.hasReadOnlyOnTarget == false); + + // Another exception is if we are accessing a property + // that provides a [nonmutating] setter. + if (!expr->type.isLeftValue && + as<PropertyDecl>(declRef.getDecl())) + { + bool isLValue = false; + for (auto member : as<ContainerDecl>(declRef.getDecl())->members) + { + if (as<SetterDecl>(member) || as< RefAccessorDecl>(member)) + { + if (member->findModifier<NonmutatingAttribute>()) + { + isLValue = true; + } + break; + } + } + expr->type.isLeftValue = isLValue; + } } else { diff --git a/source/slang/slang-check-overload.cpp b/source/slang/slang-check-overload.cpp index c738f5b49..d0441397d 100644 --- a/source/slang/slang-check-overload.cpp +++ b/source/slang/slang-check-overload.cpp @@ -1070,10 +1070,29 @@ namespace Slang if(auto subscriptDeclRef = candidate.item.declRef.as<SubscriptDecl>()) { const auto& decl = subscriptDeclRef.getDecl(); - if (decl->getMembersOfType<SetterDecl>().isNonEmpty() || - decl->getMembersOfType<RefAccessorDecl>().isNonEmpty()) + for (auto member : decl->members) { - callExpr->type.isLeftValue = true; + if (as<SetterDecl>(member) || as<RefAccessorDecl>(member)) + { + // If the subscript decl has a setter, + // then the call is an l-value if base is l-value. + if (auto base = GetBaseExpr(baseExpr)) + { + if (base->type.isLeftValue) + { + callExpr->type.isLeftValue = true; + break; + } + } + // Otherwise, if the accessor is [nonmutating], we can + // also consider the result of the subscript call as l-value + // regardless of the base. + if (member->findModifier<NonmutatingAttribute>()) + { + callExpr->type.isLeftValue = true; + break; + } + } } } @@ -2613,7 +2632,7 @@ namespace Slang // If absolutely no viable candidates were extracted from the overloaded expression, // we may be dealing with a composite type or an overloaded expression with composite types. - // + // auto typeExpr = funcExpr; if (auto overloadedExpr = as<OverloadedExpr>(funcExpr)) diff --git a/tests/bugs/gh-4971-2.slang b/tests/bugs/gh-4971-2.slang new file mode 100644 index 000000000..78efded79 --- /dev/null +++ b/tests/bugs/gh-4971-2.slang @@ -0,0 +1,23 @@ +//TEST:SIMPLE(filecheck=CHECK): -target spirv + +struct Test { + RWStructuredBuffer<int> val; + __subscript(int x, int y)->int + { + get { return val[x * 3 + y]; } + set { val[x * 3 + y] = newValue; } + } +} +Test test; + +[numthreads(1, 1, 1)] +void computeMain() +{ + // test[0,0] is not an l-value because `test` is a read-only parameter, + // and the `set` accessor is by-default `mutating`, which means that it is + // only callable when `test` itself is l-value. + + // CHECK: ([[# @LINE+1]]): error 30011 + test[0,0] = 1; + +}
\ No newline at end of file diff --git a/tests/bugs/gh-4971.slang b/tests/bugs/gh-4971.slang new file mode 100644 index 000000000..948e43385 --- /dev/null +++ b/tests/bugs/gh-4971.slang @@ -0,0 +1,25 @@ +//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-vk -output-using-type +//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-d3d11 -output-using-type + +struct Test { + RWStructuredBuffer<int> val; + __subscript(int x, int y)->int + { + get { return val[x * 3 + y]; } + [nonmutating] set { val[x * 3 + y] = newValue; } + } +} +Test test; + +//TEST_INPUT: set test = {out ubuffer(data=[0 0 0 0 0 0 0 0 0], stride=4)}; + +[numthreads(1, 1, 1)] +void computeMain() +{ + // test[0,0] should be a valid l-value here because although `test` is + // a read-only parameter, the `set` accessor is marked as `nonmutating`, + // which means that it can be called even when `test` is not mutable. + + // CHECK: 1 + test[0,0] = 1; +}
\ No newline at end of file |