mozempk/ggml-org-llama-cpp-mirror
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-11 07:20:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

HIP: fix template skip for DKQ > 256 mfma kernels (#29559)

Browse Source
This commit is contained in:
uvos
2026-09-28 13:47:29 +02:00
committed by GitHub
parent 4364bf7232
commit c2a9e16068
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/fattn-mma-f16.cuh
+1 -1
View File
@@ -1875,7 +1875,7 @@ static __global__ void flash_attn_ext_f16(
#endif // defined(AMD_WMMA_AVAILABLE)
#if defined(AMD_MFMA_AVAILABLE)
if (ncols1*ncols2 < 16 || (DKQ > 256 && ncols1*ncols2 < 64)) {
if (ncols1*ncols2 < 16 || (DKQ > 256 && ncols1*ncols2 < 32)) {
NO_DEVICE_CODE;
return;
}