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
Replace content of MGTransferMatrixFree by that of MGTransferMF #16301
base: master
Are you sure you want to change the base?
Conversation
Thanks for moving so quickly on this! I can confirm that this change fixes the issue discussed in #16292. It does however trigger another assert in some of our tests that is related to handing over a ghosted vector into the Also let me know if there is something I can do to help with this PR, happy to do a review if you say this is ready to be looked at. For later reference: Error message for ASPECT's test
|
@peterrum Two comments before we go this way:
|
@gassmoeller The last commit should fix this. We made a similar change in #15865.
I can run some of the large-scale experiments from the global-coarsening paper.
It should be a 100% replacement. If this is not the case, there aspects not tested in deal.II. |
15fbca8
to
9b505b5
Compare
@peterrum: Thank you! This version can now run all of the ASPECT tests successfully. I cannot check accuracy right now, but at least we didnt trigger any asserts and didnt encounter any crashes. |
As promised, here are some scaling results run with https://github.com/peterrum/dealii-multigrid (annulus, k=1) on SuperMUC-NG. The first plot shows the setup costs and the second plot the computation times. The rows show different number of nodes while the columns show different refinement levels. One can see that the setup costs are more expensive, while the computation is at least as good as in the old implementation. |
@gassmoeller Thank you for checking! |
4a34731
to
fd3b167
Compare
in reference to #16292 (comment)