Martin Kronbichler

It's right, we don't support it right now, but I think the combination `FE_Q`/`FE_DGQ` plus `FE_Nothing` should be relatively simple to address, because we merely need to consider the fe...

Let us remove this from the current release milestone. It would be great to soon provide a fix, but I fear we won't manage to do it in the next...

Let us move the milestone then, we need to live with the current behavior.

I agree, we should run a benchmark and have a look at the proposed changes in #14066. I won't get to it this week, but this PR is not super...

I did have a look at the simplex-throughput benchmark. While the proposed change are certainly visible, the contribution to the overall instruction count is less than 0.5% (18 million out...

Thank you for the detailed experiment @marcfehling! The numbers you present make sense, as the enumeration of unknowns now has to do less work (do not populate the cache), whereas...

@bangerth any plans to still work on that for the release? If not, please adjust the target milestone.

The assumption of a contiguous index range for the locally owned dofs is to render certain operations more efficient. Whichever way we put it, I think there will be some...

I agree with the proposed name. I would have intuitively said we should return the `cell` and `lane` index, but since we are exposing the underlying index already, I think...

One thing that will be a bit of work to port are the matrix-free loops here: https://github.com/dealii/dealii/blob/master/source/matrix_free/task_info.cc#L353-L585 That is essentially implementing a kind of task graph with TBB, but the...