llama.cpp

875fb428 - ggml-alloc : fix backend assignments of views (#3982)