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
Disable source build cache on pocl/nvidia #749
base: main
Are you sure you want to change the base?
Disable source build cache on pocl/nvidia #749
Conversation
b9b8091
to
0656451
Compare
a226c17
to
1a9ad0d
Compare
1a9ad0d
to
4ecc57c
Compare
# This must be done before PoCL is actually loaded | ||
from pyopencl.characterize import setup_pocl_cache_dirs | ||
setup_pocl_cache_dirs() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could you explain why this is here? This seems to defeat inter-run caching, which, in my book, is definitely desirable.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a workaround for the intermittent failures we have seen as part of #738.
I couldn't think of a way to ensure PID-local caches that also support inter-run caching, without resorting to using MPI rank numbers (which wouldn't even help for the CI errors seen in #738).
Edit: some sources regarding caching issues in pocl: pocl/pocl#430, pocl/pocl#466
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Before we go in this direction, I think we should engage with the pocl folks and least file an issue to see what they think about fixing concurrent access to their caches.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Reported here: pocl/pocl#1480, including a small reproducer.
Co-authored-by: Matt Smith <mjsmith6@illinois.edu>
4ecc57c
to
4e2ab3e
Compare
Followup of #738