Home
last modified time | relevance | path

Searched defs:GmemTile (Results 1 – 1 of 1) sorted by relevance

/aosp_15_r20/external/pytorch/aten/src/ATen/native/transformers/cuda/mem_eff_attention/
H A Dkernel_backward.h68 struct GmemTile { struct
106 using AccessType = cutlass::Array<float, 4>;
107 static constexpr int32_t kBytes = sizeof(AccessType);
108 static constexpr int32_t kStride = kNumThreads * AccessType::kElements;
109 static constexpr int32_t kNumIters =
111 static constexpr int32_t kElementsStored =
117 float* ptr;
119 CUTLASS_DEVICE void load(FragmentType& fragment, int thread_id) { in load()
134 CUTLASS_DEVICE void store(FragmentType const& fragment, int thread_id) { in store()
149 CUTLASS_DEVICE void storeAtomicAdd( in storeAtomicAdd()