ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-30 11:47:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix schedule

Browse Source
This commit is contained in:
mtgu0705
2025-06-27 03:34:26 -05:00
parent 3684732c49
commit ea65ce0c64
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_mx_moe_v3.hpp
View File

@@ -218,9 +218,9 @@ struct BlockwiseGemmXdlops_pipeline_mx_moe_nbs_v3<BlockGemmPipelineScheduler::In
HotLoopInstList::B_LDS_Read_Width * sizeof(BDataType) == 16 ? 8 : 4;
constexpr auto ds_read_a_mfma_rate =
(mfma_cycle - 4 + 2 * ds_read_a_issue_cycle - 1) / (2 * ds_read_a_issue_cycle);
(mfma_cycle - 8 + 2 * ds_read_a_issue_cycle - 1) / (2 * ds_read_a_issue_cycle);
constexpr auto ds_read_b_mfma_rate =
(mfma_cycle - 4 + 2 * ds_read_b_issue_cycle - 1) / (2 * ds_read_b_issue_cycle);
(mfma_cycle - 8 + 2 * ds_read_b_issue_cycle - 1) / (2 * ds_read_b_issue_cycle);
constexpr auto num_dsread_a_mfma =
(num_ds_read_inst_a + ds_read_a_mfma_rate - 1) / ds_read_a_mfma_rate;