Open kronbichler opened 4 years ago
I have seen that, too. For BP1, it is working for some reason.
Then it's probably the Dirichlet boundary conditions (dofs never touched).
That makes sense. I think we discussed a long time ago that you wanted to integrate some compressions in deal.II and I do the renumbering. Should we do it now?
Yes, we should have a strategy in deal.II, all along with the index compression for degree > 2
and for systems (in case of no complicated constraints).
If I enable the renumbering e.g. in
bp3/bench.cc
https://github.com/kronbichler/ceed_benchmarks_dealii/blob/33e5ca1ff9f09ba802d8eacaa44d9a0da079c750/bp3/bench.cc#L83-L86 and run, I get the following error:It seems that the
Renumber::grouping
function misses to add some of the unknowns for this setting.