nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 23 Packages Projects Releases Wiki Activity

CUDA: add log line when mxfp4 acceleration is used (#18483)

Browse Source 
* CUDA: add log line when mxfp4 acceleration is used

* add in backend_get_features
This commit is contained in:
Aman Gupta
2025-12-30 17:40:46 +08:00
committed by GitHub
parent a864fb1c14
commit d77d7c5c06
1 changed files with 10 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/ggml-cuda.cu
+10
View File
@@ -4785,6 +4785,16 @@ static ggml_backend_feature * ggml_backend_cuda_get_features(ggml_backend_reg_t
features.push_back({ "FA_ALL_QUANTS", "1" }); features.push_back({ "FA_ALL_QUANTS", "1" });
#endif #endif
{
const auto & info = ggml_cuda_info();
for (int id = 0; id < info.device_count; ++id) {
if (blackwell_mma_available(info.devices[id].cc)) {
features.push_back({ "BLACKWELL_NATIVE_FP4", "1"});
break;
}
}
}
#undef _STRINGIFY #undef _STRINGIFY
#undef STRINGIFY #undef STRINGIFY