llama.cpp

ggml-alloc : fix backend assignments of views

#3982

Merged