llama-bench-0.8.6 with Xeon Processor

[beartell]

bugra@llamafile:~$ ./llamafile-bench-0.8.6 -p "256,512,1024" -m qwen2-7b-instruct-fp16.gguf,qwen2-7b-instruct-q5_k_m.gguf

cpu_info	model_filename	size	test	t/s
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-fp16	14.19 GiB	pp256	22.18
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-fp16	14.19 GiB	pp512	22.77
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-fp16	14.19 GiB	pp1024	22.65
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-fp16	14.19 GiB	tg16	2.81
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-q5_k_m	5.07 GiB	pp256	26.43
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-q5_k_m	5.07 GiB	pp512	26.28
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-q5_k_m	5.07 GiB	pp1024	25.89
Intel Xeon CPU E5-2673 v3 @ 2.40GHz (haswell)	qwen2-7b-instruct-q5_k_m	5.07 GiB	tg16	7.13

[PauleBertt]

Hi thanks for your result!

How do I read these values? Is the t/s generated tokens or evaluated?
And what are theses tests?

Thanks 😀