ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-13 18:51:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

skip empty expert

Browse Source
This commit is contained in:
coderfeli
2025-02-10 01:26:08 +00:00
parent e21f36fc24
commit 2e53f9725b
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_b_preshuffle.hpp
View File

@@ -1126,7 +1126,10 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3_b_preshuffle
const index_t m_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(block_m_id * MPerBlock);
const index_t expert_stride = __builtin_amdgcn_readfirstlane(problem.N * problem.K);
const index_t t0 = (p_sorted_token_ids[block_m_id * MPerBlock] & 0xffffff);
if(t0 >= problem.NumTokens)
return;
// N0, K0, Blocksize*KPack
const index_t n_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(block_n_id * NXdlPerWave);