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

Merge ec259f9ec0 into 18ddaea2ae

This commit is contained in:
Jiacheng (Jason) Chen 2026-01-03 07:00:33 +09:00 committed by GitHub
parent 18ddaea2ae ec259f9ec0
commit dfdc0062f2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
ggml/src/ggml-cuda/mmq.cu
View File

@ -333,7 +333,16 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11, int64_t
}
if (amd_wmma_available(cc)) {
return true;
if (n_experts > 64 || ne11 <= 128) {
return true;
}
if (type == GGML_TYPE_Q4_0 || type == GGML_TYPE_Q4_1 || type == GGML_TYPE_Q5_0 || type == GGML_TYPE_Q5_1) {
return true;
}
if (ne11 <= 256 && (type == GGML_TYPE_Q4_K || type == GGML_TYPE_Q5_K)) {
return true;
}
return false;
}
return (!GGML_CUDA_CC_IS_CDNA(cc)) || ne11 < MMQ_DP4A_MAX_BATCH_SIZE;