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

Regression in version 0.12.130 #436

Closed
karajan9 opened this issue Sep 30, 2022 · 4 comments
Closed

Regression in version 0.12.130 #436

karajan9 opened this issue Sep 30, 2022 · 4 comments

Comments

@karajan9
Copy link

Some of my code isn't able to be @turboed anymore since version 0.12.130. Specifically, I'm getting a

┌ Warning: #= ... =#:
│ `LoopVectorization.check_args` on your inputs failed; running fallback `@inbounds @fastmath` loop instead.

and ~3x slowdown on code that worked fine on version 0.12.129.

I'm not sure if I'll be allowed to share the code. Is there any information I could help with while I try to figure this out?

@karajan9
Copy link
Author

After looking through the changes I think it was making safe=true the default.
Since replacing every instance of a^2 by a*a and ifelse by my_ifelse(decider, a, b) = decider * a + (1 - decider) * b I have achieved the old performance again.
I'm not sure if that was the intention behind the change but if so I can certainly live with the changes I made.

@chriselrod
Copy link
Member

Hi, regressions where @turbo worked before were not the intention.
can_turbo should return true for ifelse and ^.
I'll have changes shortly that make this so.

Please feel free to file issues or create PRs for any more such problems.

@chriselrod
Copy link
Member

Please let me know if your original versions encounter any problems. They should work now.

@karajan9
Copy link
Author

karajan9 commented Oct 2, 2022

I can confirm that the original code now again runs with original speed (and no warning).
Thanks for looking into this and all the work you have been doing on this project! Truly a gamechanger.

This issue was closed.
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

2 participants