llama.cpp

pass cpu-architecture arguments only to host code (C;C++)

#4943

Merged