llama.cpp

769266a5 - cmake : compile ggml-rocm with -fpic when building shared library (#3158)