llama.cpp

b2debf65 - parallel : add disabled experimental batch chunking in powers of two