llama.cpp

c792dcf4 - ggml : allow loading backend with env variable (ggml/1059)