ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-01 20:21:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

Avoid using store_tile_raw() for fp32 tensors (#2072)

Browse Source
This commit is contained in:
Po Yen Chen
2025-04-27 14:07:41 +08:00
committed by GitHub
parent 41541aff7a
commit 3d4d70d2fc
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
example/ck_tile/01_fmha/codegen/ops/fmha_fwd_splitkv.py
View File

@@ -91,10 +91,12 @@ using fmha_pipeline_problem = ck_tile::BlockFmhaFwdSplitKVPipelineProblem<
using fmha_pipeline = {F_pipeline}<
fmha_pipeline_problem>;
/// FIXME: use {F_spad}/{F_dvpad} as kPadM/kPadN parameters after solving
/// store_tile_raw() data corruption issue
using fmha_epilogue =
ck_tile::Default2DEpilogue<ck_tile::Default2DEpilogueProblem<typename FmhaFwdTypeConfig<{F_dtype}>::OaccDataType,
typename FmhaFwdTypeConfig<{F_dtype}>::OaccDataType,
{F_spad}, {F_dvpad}>>;
false, false>>;
using fmha_kernel =
ck_tile::FmhaFwdSplitKVKernel<fmha_pipeline, fmha_epilogue>;