[aspect-devel] aspect release planning

[Wolfgang Bangerth]

> Interesting that the curved cells make the problem more difficult for
> the linear solver, though. Do you have an intuition why that is the
> case? I would have expected the linear system to be of similar
> difficulty if not easier because cell sizes are more regular.

I think it's really just a coincidence. The pattern of how the viscosity 
changes with location may just have been fortuitously aligned before and 
now is no longer. It's hard to tell, but I really think it's just a 
coincidence.


> Well, it can be related: we now have a mapping in the interior. Maybe
> there is something fishy happening with the mapping or quadrature
> point locations in the material model? Who knows.

True, you never know. It's always possible that we have bugs. I don't 
know what to do about it, though (that's sort of the software business's 
general problem I suppose). As far as mapping in the interior, I don't 
actually think so. We create the mapping in core.cc like this:
    mapping (parameters.free_surface_enabled?1:4),
so this calls this constructor:
   MappingQ (const unsigned int p,
             const bool use_mapping_q_on_all_cells = false);
I think it still uses the straight mapping in the interior. Of course, 
this may well be a good target for a patch in itself :-)


> I rebased and they both crash the same way:
> http://cdash.kyomu.43-1.org/testDetails.php?test=4480083&build=2541
>
> (I am fixing the test output for the other tests right now)

OK, note it in a comment of the pull request when you've updated 
everything and I will take another look.

Best
  W.


-- 
------------------------------------------------------------------------
Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/