llama.cpp

7dc78764 - compare-llama-bench: tweak output format (#4910)