llama.cpp

41318d70 - llama : use the same threshold for OpenBLAS and ggml thread limiting (#577)