llama.cpp

Use the same batch size threshold for enabling OpenBLAS and disabling ggml threading

#577

Merged