llama.cpp

6961c4bd - batched-bench : print params at start