EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[Bug] Fix w4afp8 moe kernel (#9392)

Browse Source
This commit is contained in:
Yuhao Yao
2025-08-21 18:45:18 +08:00
committed by GitHub
parent 029e0af31d
commit de4990a5b2
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
sgl-kernel/csrc/cutlass_extensions/gemm/collective/sm90_mma_array_tma_gmma_rs_warpspecialized_mixed_input_.hpp
View File

@@ -1488,6 +1488,10 @@ struct CollectiveMmaArrayMixedInput<
template <class... TMs> template <class... TMs>
CUTLASS_DEVICE void CUTLASS_DEVICE void
tensormaps_cp_fence_release(TensorMapStorage& shared_tensormaps, cute::tuple<TMs...> const& input_tensormaps) { tensormaps_cp_fence_release(TensorMapStorage& shared_tensormaps, cute::tuple<TMs...> const& input_tensormaps) {
if (cute::elect_one_sync()) {
cute::tma_desc_commit_group();
cute::tma_desc_wait_group();
}
// Entire warp must do this (i.e. it's aligned) // Entire warp must do this (i.e. it's aligned)
tma_descriptor_cp_fence_release(get<0>(input_tensormaps), shared_tensormaps.smem_tensormap_A); tma_descriptor_cp_fence_release(get<0>(input_tensormaps), shared_tensormaps.smem_tensormap_A);
tma_descriptor_cp_fence_release(get<1>(input_tensormaps), shared_tensormaps.smem_tensormap_B); tma_descriptor_cp_fence_release(get<1>(input_tensormaps), shared_tensormaps.smem_tensormap_B);