-
Notifications
You must be signed in to change notification settings - Fork 564
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
Tpetra::Map::locallySameAs slow when running computations on GPU #433
Comments
That's totally wrong. The test should go inside HAVE_TPETRA_DEBUG. (Yes, |
I've got a fix ready. Running CUDA tests now, will do non-CUDA tests for the push. |
@csiefer2 would you object to removal of comparable tests from transferAndFillComplete etc.? |
@trilinos/tpetra Make sure Map compatibility tests in doImport and doExport (both forward and reverse mode) only run in a debug build. This fixes #433.
Origin repo remote tracking branch: 'github/master' Origin repo remote repo URL: 'github = git@github.com:TriBITSPub/TriBITS.git' At commit: commit 35d82aaa31fe81ca30a619320bb71fe481e9d4c7 Author: Roscoe A. Bartlett <rabartl@sandia.gov> Date: Tue Dec 14 15:48:02 2021 -0700 Summary: Interpret a raw identifer in TPL_<tplName>_LIBRARIES as a lib name (#299, #433)
Tpetra::DistObject::doImport calls Map::operator!= even when ! defined HAVE_TPETRA_DEBUG. That means isSameAs and locallySameAs must be fast.
If pointer equality, contiguity, uniformity, etc. checks fail, then locallySameAs iterates over GIDs on the host. This is very expensive relative to GPU computations.
I don't require an urgent fix for this problem, as I can work around this, but I'm opening this issue because I think the problem needs to be fixed at some point.
The text was updated successfully, but these errors were encountered: