Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Slow when loading customized backend with mkl #907

Open
tz301 opened this issue Apr 9, 2021 · 0 comments
Open

Slow when loading customized backend with mkl #907

tz301 opened this issue Apr 9, 2021 · 0 comments

Comments

@tz301
Copy link

tz301 commented Apr 9, 2021

I have a customized backend, which is similiar as kaldi asr but just using cpu. I have run it successfully but it becomes slow obviously (2~3 times slow) and much more cpu usage. I doubt that it's because server thread function call mkl thread function. The customized backend dynamic library will link mkl library installed in system.

I have used pprof to show functions be called as below figures. The first is local run, and the second is run in server, both using the same data for just one inference.

For local run, mkl_blas is called many times. But for the second, omp_get_num_procs is the highest, and mkl_blas is called only few times. Could anyone give some advices?
1
2

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant