llama.cpp

d2b851bf - cmake : only enable GGML_NATIVE and x86 flags if not crosscompiling (ggml/885)