llama.cpp

Use -march=native -mtune=native on x86 (Also Enables AVX512 on macOS)

#609

Merged