ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-04 05:31:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

[What] Fix bug of verification fail on E Matrix (#371)

Browse Source 
[Why] We need to sync lds even in first loop because Gemm also use the same LDS.
This commit is contained in:
rocking5566
2022-08-22 20:50:28 +08:00
committed by GitHub
parent 9efd033bee
commit c366de553e
3 changed files with 3 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp
View File

@@ -776,8 +776,7 @@ struct GridwiseGemmMultipleDMultipleR_k0mk1_k0nk1_mn_xdl_cshuffle_v1
static_for<0, num_access, 1>{}([&](auto access_id) {
// make sure it's safe to read from LDS
if constexpr(access_id > 0)
block_sync_lds();
block_sync_lds();
// each thread shuffle data from VGPR to LDS
c_thread_copy_vgpr_to_lds.Run(c_thread_desc_m0_n0_m1_n1_m2_m3_m4_n2,