llama.cpp

f914544b - batched-bench : add "separate text gen" mode (#17103)