llama.cpp

cmake: don't force -mcpu=native on aarch64

#2063

Merged