Peter Munch

I am in particular thinking about PR #122 and #148. There is no need to cause there merge conflicts if one can wait a couple of weeks.

> On the other hand, the postprocessing time for (l=1, k=4) went down from 18.3 seconds to 2.06 seconds. I like this :)

We have dropped the support of deal.II 9.3 a couple of weeks ago, since deal.II 9.4 is released in the next days. Either you wait till than or you use...

I'll do. @kronbichler could you ping me once you are ready?

> I can make those changes, but before we spend more time on this side topic, let me raise a question: What would you think if we introduce a special...

With https://github.com/dealii/dealii/pull/11803 and https://github.com/dealii/dealii/pull/11802 merged, the following code: https://github.com/exadg/exadg/blob/008cbe87b92ee8a554f22fbfe0c8a5c205588c87/include/exadg/operators/operator_base.cpp#L386-L405 could be written like: ```cpp internal::ComputeDiagonalHelper helper(phi); for (unsigned int cell = 0; cell < matrix_free.n_cell_batches(); ++cell) { helper.reinit(cell); this->reinit_cell(integrator.get_current_cell_index()); for...

Why does this PR influence restart files in the non-hp context? See https://github.com/lethe-cfd/lethe/pull/623. FYI @blaisb

@jh66637 Thanks for working on this topic! Before digging into a user-friendly interface, let's discuss what the class actually does (in comparison to RPE). To set up PRE, you give...

@jh66637 Question what are you communicating during setup? I guess the intersections are never communicated but only the geometric entities that might cut?

Let me allow to shortly summarize the algorithm we talked about. For me a geometric entity is defined as a vector of vertices (2 points for a line, 3 points...