llama.cpp

c062ffd1 - batched-bench : init warm-up batch