Feature/fgmres

Implements a right-preconditioned flexible GMRes solver.

BTW this is a great addition I just did not yet find the time to thoroughly test it. Did you test it in parallel, too?

I didn't test it in parallel. There is currently no parallel solver test in istl, no?

no. But one could make a very easy test using Richardson as a preconditioner. With it the results from a serial and parallel run should be very similar if not identical.

added 93 commits

• 1d014192...676d3a38 - 91 commits from branch core:master
• a4174df6 - [gmres][bugfix] Fix double call of prec->post(x) when res is zero
• ff03da48 - [fgmres] Implement a right-preconditioned flexible GMRes solver

Compare with previous version

resolved all discussions

Seems like there is a bugfix hidden in this feature MR. Shouldn't we split this up?

Yes you're right I'll split it up.

added 48 commits

• ff03da48...6e1774cf - 47 commits from branch core:master
• ca592e7e - [fgmres] Implement a right-preconditioned flexible GMRes solver

Compare with previous version

changed the description

I opened !172 (merged) for the gmres bugfix

Is this ready for merge?

yes but it should maybe be checked by someone with solver knowledge.

Could you rebase on top of the current master?

added 26 commits

• ca592e7e...46bf82a5 - 25 commits from branch core:master
• 178e0569 - [fgmres] Implement a right-preconditioned flexible GMRes solver

Compare with previous version

@robert.kloefkorn done

Great. There is only one thing I have to complain about. RestartedGMResSolver RestartedfGMResSolver

These two solvers names only differ in one small letter in the middle of the word. I think this is not good because it's hard to read. Should be something like RestartedFlexibleMResSolver or similar to make the difference more visible.

added 1 commit

• 3a5a43ce - [fgmres] Implement a right-preconditioned flexible GMRes solver

Compare with previous version