yum-mirror/slang

Making it easier to work with shaders

git clone https://git.yummers.dev/yum-mirror/slang

Sai Praveen BangaruAdded `[AutoPyBindCUDA]` for automatic kernel binding + `[PyExport]` for exporting type information (#3209)739c3a7b5

master
branchesmasterall refs →
1.2 KiB25 linesraw
1//TEST:SIMPLE(filecheck=CUDA): -target cuda -line-directive-mode none
2//TEST:SIMPLE(filecheck=TORCH): -target torch -line-directive-mode none
3
4// CUDA: __global__ void __kernel__myKernel(TensorView inValues_[[#]], TensorView outValues_[[#]])
5[AutoPyBindCUDA]
6[CudaKernel]
7void myKernel(TensorView<float> inValues, TensorView<float> outValues)
8{
9    if (cudaThreadIdx().x > 0)
10        return;
11    outValues.store(cudaThreadIdx().x, sin(inValues.load(cudaThreadIdx().x)));
12}
13
14// TORCH:      {{^SLANG_PRELUDE_EXPORT$}}
15// TORCH-NEXT: void __kernel__myKernel(TensorView {{[[:alnum:]_]+}}, TensorView {{[[:alnum:]_]+}});
16
17// TORCH:      {{^SLANG_PRELUDE_EXPORT$}}
18// TORCH-NEXT: void myKernel(std::tuple<uint32_t, uint32_t, uint32_t> {{[[:alnum:]_]+}}, std::tuple<uint32_t, uint32_t, uint32_t> {{[[:alnum:]_]+}}, torch::Tensor {{[[:alnum:]_]+}}, torch::Tensor {{[[:alnum:]_]+}})
19
20// TORCH:      {{^SLANG_PRELUDE_EXPORT$}}
21// TORCH-NEXT: std::tuple<std::tuple<const char*, const char*, const char*, const char*>, std::tuple<const char*, const char*>, const char*, const char*> __funcinfo__myKernel()
22
23// TORCH:      m.def("myKernel", &myKernel, "myKernel");
24
25// TORCH:      m.def("__funcinfo__myKernel", &__funcinfo__myKernel, "__funcinfo__myKernel");