llama.cpp

b8872700 - GGUF: backend support, fixed-width I/O, misc fixes