[Documentation] Unclear how to run `run_benchmark.py`

[Rohanjames1997]

Describe the documentation issue

From the documentation in run_benchmark.readme, it is unclear how to run the run_benchmark.py script on the onnxruntime_mlas_benchmark program.

Where can I find the onnxruntime_mlas_benchmark executable? I do not see it in the build folder after building onnxruntime from source.

Can someone provide steps to run the script, especially to generate results in the format described in #18031 , #18953, and #19272? (Especially the pattern to be provided to the run_benchmark.py script)

Page / URL

https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/test/run_benchmark.readme.md

[skottmckay]

Intended as an internal tool.

run_benchmark.py is a relatively simple python script that you can read through. Did you run the script with --help as per the doc? What aspect of running it was unclear after reading that help?

You need to enable onnxruntime_BUILD_BENCHMARKS to create onnxruntime_mlas_benchmark.

Calling build.py with --build_micro_benchmarks well set that.

[Rohanjames1997]

Hi @skottmckay !

I had missed the --build_micro_benchmarks flag while building. I figured it out just after but I couldn't close the issue on time 😄

Thanks!