Martin Kronbichler

I could help because the code is, well, involved, and I would need to somewhere sketch the dependencies. This is not super urgent, so if we run the stuff in...

@peterrum you marked this PR as WIP - is there anything more you want to address?

After discussion with @nfehn, we came up with a more detailed plan for approaching the problem. Regarding the first point: - [ ] Extend the PolynomialsRaviartThomas class https://github.com/dealii/dealii/blob/f541dd4710b82663a96a59ae395609792a91c52d/source/fe/fe_raviart_thomas_nodal.cc#L41-L42 so that...

Apart from the cleanup in https://github.com/dealii/dealii/issues/12728, I think we need to postpone the additional topics in this issue to after the 9.4 release. I think we have a nice "experimental"...

We won't have this until the 9.4 release, either, even though we made progress on some related topics. Adjusting target milestone.

Regarding the namespace, I have no strong opinion. `SparseMatrixTools` sounds OK in my opinion. One question I'm wondering, without having looked at it: Did you compare this to the `sparse_vanka.h`...

I would have said that the hanging node constraints in the ghost layer are not necessarily correct, namely on hanging nodes that are on a face that does not connect...

Is there anything more we plan to do on this topic before the release?

The fix in #14119 does improve things a bit, performance is now halfway between the regression of #14085 and the previous state.  I think we should use this issue...

As a note to future work: Apart from the changes in #14314, there is a number of similar things to be fixed up in the `else` branch of the reads...