Thomas Baumann
Thomas Baumann
We could run on the macOS environment only tests that are not marked as slow. That would, ofc slightly reduce confidence in the code, but not unreasonably so since there...
To summarise the state of the development branch: We have an implementation that derives from `Mesh` and hence is only compatible with numpy stuff at the moment. What it does...
> Can you also replace the `comp2_mesh`, too? I always wanted to merge all of these numpy-based data types into one core class, with whatever components the "users" want. So,...
I am wondering where we should stop using this. So far, I was only thinking about integration-specific labels for the components, but why stop there? With this it is very...
I just noticed that for some reason this only works with two levels. I'll reopen this PR once I find and fix the bug.
Bug was in the tau correction which was not computed with only two levels. Test for MPI sweeper now goes up to 3 levels.
Either somebody review it again or merge it, I guess :D
face palm... Looks rather similar :D But are you sure your [`prolong f`](https://github.com/Parallel-in-Time/pySDC/blob/a5498546005050b284b77a4f6bac5b1396f4e181/pySDC/projects/parallelSDC/BaseTransfer_MPI.py#L194) works? There's no communication at all in there... How do you want to proceed? We should have...
Using the lovely diff feature of vim, I can see there are some differences after all. Why do you do stuff to the initial conditions in prolong? Eg [here](https://github.com/Parallel-in-Time/pySDC/blob/a5498546005050b284b77a4f6bac5b1396f4e181/pySDC/projects/parallelSDC/BaseTransfer_MPI.py#L183). I...