relu / build /torch-ext /torch_binding.cpp

Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu

e873e70 verified 4 months ago

524 Bytes

	#include <torch/library.h>

	#include "registration.h"
	#include "torch_binding.h"

	TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
	ops.def("relu(Tensor! out, Tensor input) -> ()");
	#if defined(CPU_KERNEL)
	ops.impl("relu", torch::kCPU, &relu);
	#elif defined(CUDA_KERNEL) \|\| defined(ROCM_KERNEL)
	ops.impl("relu", torch::kCUDA, &relu);
	#elif defined(METAL_KERNEL)
	ops.impl("relu", torch::kMPS, relu);
	#elif defined(XPU_KERNEL)
	ops.impl("relu", torch::kXPU, &relu);
	#endif
	}

	REGISTER_EXTENSION(TORCH_EXTENSION_NAME)