happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

CUDA: tune GLM 4.7 Flash FA kernel selection logic (#19097)

This commit is contained in:
Johannes Gäßler 2026-01-27 14:28:56 +01:00 committed by GitHub
parent c0204a0893
commit a5bb8ba4c5
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
ggml/src/ggml-cuda/fattn.cu
View File

@ -148,6 +148,10 @@ static void ggml_cuda_flash_attn_ext_mma_f16(ggml_backend_cuda_context & ctx, gg
const int gqa_ratio = Q->ne[2] / K->ne[2];
if (gqa_ratio == 20) { // GLM 4.7 Flash
if (cc >= GGML_CUDA_CC_BLACKWELL) {
if (Q->ne[1] <= 4 && K->ne[1] >= 65536) {
ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 16>(ctx, dst);
break;
}
ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 4>(ctx, dst);
break;
}
@ -161,6 +165,10 @@ static void ggml_cuda_flash_attn_ext_mma_f16(ggml_backend_cuda_context & ctx, gg
}
if (cc >= GGML_CUDA_CC_TURING) {
if (Q->ne[1] <= 4) {
if (K->ne[1] <= 16384) {
ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 16>(ctx, dst);
break;
}
ggml_cuda_flash_attn_ext_mma_f16_switch_ncols1<576, 512, 32>(ctx, dst);
break;
}