Skip to content

Conversation

@jeremylt
Copy link
Member

fixes #1862
fixes #1858

This PR does two things

  1. I accidentally got too fancy and provided incorrect values for the libCEED bases being constructed, leading to incorrect convergence. This fixes that

  2. Now this example does not do the extra data copying for BP2 and BP4

@jeremylt
Copy link
Member Author

Removing the copies improves the performance (mild but noticable):

New:

$ build/dealii-bps --resource /cpu/self --bp BP1 --fe_degree 3 --n_refinements 4
ceed: 11 277944 0.00348648
dealii: 11 277944 0.0319286
$ build/dealii-bps --resource /cpu/self --bp BP2 --fe_degree 3 --n_refinements 4
ceed: 11 393072 0.00572181
dealii: 11 393072 0.0601282
$ build/dealii-bps --resource /cpu/self --bp BP3 --fe_degree 3 --n_refinements 4
ceed: 75 4559.6 0.0263361
dealii: 75 4559.6 0.412881
$ build/dealii-bps --resource /cpu/self --bp BP4 --fe_degree 3 --n_refinements 4
ceed: 75 6448.24 0.0481876
dealii: 75 6448.24 0.78902

Old:

$ build/dealii-bps --resource /cpu/self --bp BP1 --fe_degree 3 --n_refinements 4
ceed: 11 277944 0.00345421
dealii: 11 277944 0.0324943
$ build/dealii-bps --resource /cpu/self --bp BP2 --fe_degree 3 --n_refinements 4
ceed: 11 393072 0.00892214
dealii: 11 393072 0.0616518
$ build/dealii-bps --resource /cpu/self --bp BP3 --fe_degree 3 --n_refinements 4
ceed: 75 4559.6 0.028057
dealii: 75 4559.6 0.419625
$ build/dealii-bps --resource /cpu/self --bp BP4 --fe_degree 3 --n_refinements 4
ceed: 75 6448.24 0.0599387
dealii: 75 6448.24 0.834815

@jeremylt jeremylt force-pushed the jeremy/deall-no-pack branch from 4a4c2ab to 704703e Compare July 18, 2025 12:36
@jeremylt jeremylt force-pushed the jeremy/deall-no-pack branch from 1271534 to 7d3b977 Compare July 18, 2025 12:44
@jeremylt jeremylt merged commit df10768 into main Jul 18, 2025
29 checks passed
@jeremylt jeremylt deleted the jeremy/deall-no-pack branch July 18, 2025 13:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Deal.II Examples Bad Bases Deal.II Shouldn't Need copy_to_block_vector

3 participants