ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-24 23:05:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update AMD buffer coherency (#3403)

Browse Source 
* Update AMD buffer coherency [AICK-421]

* fixes

* fix

* fixes

* fixes

* Add backward compatilibity

* fix

* fixes

* fix

* fix

* fix

* Update grouped_convolution_backward_weight_kernel.hpp

[ROCm/composable_kernel commit: 700b2ec9c0]
This commit is contained in:
Bartłomiej Kocot
2025-12-18 10:16:22 +01:00
committed by GitHub
parent 962966cf5e
commit 80eaeacea5
11 changed files with 268 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
include/ck_tile/ops/flatmm/pipeline/flatmm_pipeline_agmem_bgmem_creg_v1.hpp
View File

@@ -31,7 +31,9 @@ struct BaseFlatmmPipelineAGmemBGmemCRegV1
ck_tile::ignore = K;
if(M <= 416)
{
#if defined(__gfx942__) || defined(__gfx950__)
return ck_tile::amd_buffer_coherence_enum::WAVE_NT1;
#endif
}
return ck_tile::amd_buffer_coherence_enum::coherence_default;
}