v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude)
f30427a verified | [general] | |
| name = "relu" | |
| license = "Apache-2.0" | |
| version = 3 | |
| backends = [ | |
| "cpu", | |
| "cuda", | |
| "xpu", | |
| ] | |
| [general.hub] | |
| repo-id = "SuperexponentialAI/relu" | |
| [torch] | |
| src = [ | |
| "torch-ext/torch_binding.cpp", | |
| "torch-ext/torch_binding.h", | |
| ] | |
| [kernel.relu] | |
| backend = "cuda" | |
| depends = ["torch"] | |
| src = ["relu_cuda/relu.cu"] | |
| [kernel.relu_xpu] | |
| backend = "xpu" | |
| depends = ["torch"] | |
| src = ["relu_xpu/relu.cpp"] | |
| [kernel.relu_cpu] | |
| backend = "cpu" | |
| depends = ["torch"] | |
| src = ["relu_cpu/relu_cpu.cpp"] | |
| cxx-flags = ["-fopenmp"] | |