diff options
| author | Yong He <yonghe@outlook.com> | 2024-02-05 20:07:58 -0800 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-02-05 20:07:58 -0800 |
| commit | af035fb6da2a19ccc647515e9b1edf35777f8c89 (patch) | |
| tree | 6660e1b066e153946b54bda089eb6691582a1664 /source | |
| parent | a88a7db9b3c215be890d4a9da6414959f9f9a3c1 (diff) | |
Add glsl implementation of Texture.InterlockedAddF32 (#3550)
Diffstat (limited to 'source')
| -rw-r--r-- | source/slang/hlsl.meta.slang | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/source/slang/hlsl.meta.slang b/source/slang/hlsl.meta.slang index 29c9d4c51..1e0b95c81 100644 --- a/source/slang/hlsl.meta.slang +++ b/source/slang/hlsl.meta.slang @@ -2528,7 +2528,6 @@ ${{{{ } // FP16x2 - __cuda_sm_version(2.0) [__requiresNVAPI] uint _NvInterlockedAddFp16x2(uint byteAddress, uint fp16x2Value) { @@ -2539,7 +2538,6 @@ ${{{{ } } - __cuda_sm_version(2.0) [__requiresNVAPI] [ForceInline] void InterlockedAddF16(uint byteAddress, half value, out half originalValue) @@ -2555,7 +2553,7 @@ ${{{{ else { byteAddress = byteAddress & ~3; - uint packedInput = asuint16(value) << 16; + uint packedInput = ((uint)asuint16(value)) << 16; originalValue = asfloat16((uint16_t)(_NvInterlockedAddFp16x2(byteAddress, packedInput) >> 16)); } return; @@ -13014,6 +13012,7 @@ __generic<Shape:__ITextureShape1D2D3D, let format : int> extension __TextureImpl<float, Shape, 0, 0, 0, $(kStdlibResourceAccessReadWrite), 0, 0, format> { [__requiresNVAPI] + __glsl_extension(GL_EXT_shader_atomic_float) void InterlockedAddF32(vector<uint, Shape.dimensions> coord, float value, out float originalValue) { __target_switch @@ -13021,6 +13020,8 @@ extension __TextureImpl<float, Shape, 0, 0, 0, $(kStdlibResourceAccessReadWrite) case spirv: originalValue = __atomicAdd(this[coord], value); return; + case glsl: + __intrinsic_asm "$3 = imageAtomicAdd($0, $1, $2)"; case hlsl: __intrinsic_asm "$3 = NvInterlockedAddFp32($0, $1, $2)"; } |