Does llama-bench support first token latency?

[kylo5aby]

I'm new to LLM and want to do some research on it. Some other inference frameworks support first token latency in their benchmark tools. In llama-bench, there is a prompt eval time. In the README at https://github.com/ggerganov/llama.cpp/blob/b841d0740855c5af1344a81f261139a45a2b39ee/examples/batched-bench/README.md?plain=1#L31, it seems to indicate this is for the first token time, but I am not sure. I see that some other benchmark tools haven't calculated the prompt->token duration for the first token, and I'm not sure about the difference.