Niklas Fehn

@richardschu thanks for your explanations. What I do not understand is why the operator evaluation is affected by this. During initialization, I thought that MatrixFree receives both AffineConstraints objects and...

I think it makes sense to split this PR into parts as you suggest above. The starting point are probably changes made to `OperatorBase`. If I understand the present PR...

sure. So I wait with the review of this PR until #644 is merged?

This looks good overall! As written in the review comments, I see two major points of discussion: i) computation of the sub time step, ii) throughput calculation

> For me, it was interesting to see the throughput/macro time step but I see your point. I am happy to change this, but I am not sure in which...

The first commit shows which places we need to extend / re-implement for simplex by introducing if-else-branches depending on the `ElementType` and by introducing asserts for element types not supported...

@kronbichler as discussed in person, the current implementation of `MappingDoFVector` might be problematic, because `dof_handler` (and the underlying FE and degree) passed to the function `fill_grid_coordinates_vector()` (and `initialize_mapping_from_dof_vector()`) can be...

@kronbichler thanks for your review. We need some more refactoring in order to integrate Simplex elements easily without having to touch several classes (see https://github.com/exadg/exadg/pull/634/commits/923958b715e0fa06167a80788ed9fe265523fc9f and https://github.com/exadg/exadg/pull/634/commits/7fbcbe8f8cb2a1bb77006ce4b98418c0b93c31df).

I understand this issue as related to classes that we call `OutputGenerator` in ExaDG. Maybe we should restrict the issue name to this type of postprocessing functionality (?), because postprocessing...

Yes, you are right, computing a global vector can have many reasons, more than for writing vtu output. See e.g. the compressible Navier-Stokes postprocessing and the complex if-statements appearing there:...