llama.cpp

llama : initial ggml-backend integration

#4520

Merged