-
-
Notifications
You must be signed in to change notification settings - Fork 25.8k
CI "no OpenMP" build environment actually has OpenMP #24694
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
Comments
The original reason for disability OpenMP was because the default compiler on OSX did not have OpenMP: #15174 I think using linux to test no OpenMP is sufficient and +1 on moving the CI over. |
In #25391 an unprotected Also, it does make sense to keep the OSX with system clang use case build, since this may be something that happens in practice for some users and we prefer that the build works with a warning. That means (at least for now) that the no-OpenMP build has OpenMP, which is a bit weird but oh well 🤷. For completeness I could not find a way to create a conda environment without OpenMP. For example on Linux as soon as you have In a not so far future, we may have a Pyodide build, which will be a "no-OpenMP whatsoever" build. |
To be precise, there's openmp in the env, but scikit-learn is not built against openmp. |
This avoided catching a regression where an unprotected
cimport openmp
was introduced.As a side-comment: Pyodide build needs to be built without OpenMP.
From #24682 (comment), there is OpenMP in the build environment:
From #24682 (comment)
Looking at why we get OpenMP in the "no OpenMP" build:
So it seems like if we want an "no OpenMP" build we need it to be Linux or Windows. Not sure whether there was a good reason to have it on OSX originally.
The text was updated successfully, but these errors were encountered: