MCPcopy Create free account
hub / github.com/ashvardanian/less_slow.cpp / tensor_core_scale

Function tensor_core_scale

less_slow.cpp:2272–2277  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

2270};
2271
2272tensor_core_scale_t tensor_core_scale(int sm_capability) {
2273 if (sm_capability >= 90) return tensor_core_scale_t::warpgroup_k;
2274 if (sm_capability >= 80) return tensor_core_scale_t::warp_k;
2275 if (sm_capability >= 70) return tensor_core_scale_t::quadpair_k;
2276 return tensor_core_scale_t::single_k;
2277}
2278
2279/**
2280 * @brief Runs the benchmark loop for precompiled CUDA C++ kernels using

Callers 2

theoretic_tops_cudaFunction · 0.85
theoretic_tops_ptxFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected