-
Notifications
You must be signed in to change notification settings - Fork 699
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
Suppress pass-failed warnings/errors #2404
Conversation
Can you give an example of the failure messages you are seeing? Why would llvm be generating warning about loop vectorization failure unless we are explicitly asking to run that pass? (are we doing that?) |
see also simd-everywhere/simde#1151 |
But if we aren't seeing these warning when we build/run our tests why suppress the warning here? If |
because our clang is too old (it's using clang 14) (this is because we're stuck on ubuntu 22.04 until the next LTS comes out) |
You mean the clang we use in the wabt CI?
In that case I think we should combine this change some amount of CI test for recent versions of clang. Do you know which version of clang is needed in order to reproduce this warning? |
the macos-14 runner (apple silicon, which we aren't using) reproduces the warning, but it also fails due to the other thing. (unimplemented simd read intrinsics or something?) we could also wait a few months for CI to start failing. we know it definitely reproduces on clang 17 |
This sounds like maybe a bug in clang, no? Why would it tell the user about some internal pass failure, unless the user somehow opts into that pass specifically... are we somehow doing that in the generated code? |
Yes. simde expands |
Thanks for context. I think that was what I was missing. How about adding a comment here such as: |
Fixes #2403
simde obviously requests loop vectorization, but we shouldn't fail for missed vectorizations as we're more focused on testing for correctness.