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

Add avx512 flags to makevars #29

Merged
merged 1 commit into from
Jan 27, 2024
Merged

Conversation

shikokuchuo
Copy link
Contributor

Present on recent Intel processors.

@jwijffels jwijffels merged commit 72ca835 into bnosac:master Jan 27, 2024
10 checks passed
@jwijffels
Copy link
Contributor

Thanks. I based myself on https://github.com/ggerganov/whisper.cpp/blob/master/Makefile to construct the Makevars file. Didn't see avx512 in there. Does it make sense to include it upstream at https://github.com/ggerganov/whisper.cpp/blob/master/Makefile?

@shikokuchuo
Copy link
Contributor Author

shikokuchuo commented Jan 27, 2024

I thought I got a decent speedup, but think I was mistaken... I saw it as an option in the output to whisper_benchmark(model). The issue I'm having is that actually every time I run that benchmark (in the same session), the time increases, not sure what the reason could be.

@jwijffels
Copy link
Contributor

I haven't looked myself into details to this benchmark and didn't spend much time to incorporate the newest changes in that since whisper.cpp version 1.5.4. It is still pretty much the benchmark code which was there in whisper.cpp version 1.2.1.
There are some changes in the benchmark script at whisper.cpp which is taken from here: https://github.com/ggerganov/whisper.cpp/blob/master/examples/bench/bench.cpp so not sure what the reason is myself

If you want to compare your machine to others: ggerganov/whisper.cpp#89

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

Successfully merging this pull request may close these issues.

2 participants