llama.cpp

43ed6f25 - use allocator instead of hop2