Niklas Fehn

In my opinion its too complex to test all combinations of parameter in ExaDG. Also the code does not have the complexity that all parameters are coupled and need to...

As an addition: It might turn out that the if-statement can be removed entirely. Shouldn't we then actually rethink the addition of many new tests here? So as a suggestion...

For me, the reasoning is not clear why to have "redundant" tests e.g. for AMG and "multigrid". Note that "Multigrid" still has many parameters and that "Multigrid" internally contains AMG,...

I understand the description of this PR as introducing a new numerical method whose benefit has not been demonstrated so far? Is the idea to have this type of solver...

Let me explain from a different perspective why the current design is problematic: - Because of the call `nonlinear_operator.evaluate_rhs_picard(residual_rhs);`, every nonlinear operator has to implement now a function `evaluate_rhs_picard()`, even...

Can you explain why iteration counts are changing, because issue #650 mostly talks about the impact in terms of postprocessing/visualization?

I think we should discuss again the design. When looking at the code changes, it might be that we have forgotten it somewhere. I think this originates from the fact...

Because of your comment "Until then, this is on do not merge.", I converted this PR to a draft.

> Maybe no one solved structure problems with hanging nodes or periodic constraints yet? Yes, this is not unlikely. I wonder right now what periodic BCs would mean for structure...

@kronbichler agree. I should make my statement more precise. I do not want to argue that one could not deal with this technically, but wondered whether we could expect the...