[dune-pdelab] Nonoverlapping solver backend for implicit time-steppers

[Christian Engwer]

> I find the same strategy to work with FEM but apparently it doesn't
> work with DG. Does  DG works in this way or do I need to incorporate
> some things into the code?. I actually do not understand why DG
> fails in this similar way. 

this has nothing to say... the fem code really uses overlap, while the
DG code only requires cell-neighbours, which means ghosts are sufficient.

> I will appreciate some more tips on this.
> Thank you.

here are some tips:

- try the overlapping cell constraints (I didn't look up the name
  ... something like p0-constraints)
>     typedef Dune::PDELab::OverlappingConformingDirichletConstraints CON; 
  you are using the OverlappingConformingDirichletConstraints... to be
  honest, I'm not up-to-date any more regarding the details of the
  different constraints assemblers, but in any case you are not using
  conforming methods and thus you should not expect the FEM example to
  work for DG.

- be sure to use an appropriate preconditioner, your current solver
  should work in principal, but if you can't get it running, you might
  have to look into the details of the implementation, to make sure,
  that the constraints are communicated in the right way and that the
  scalar product matches your discretization

and some general advice concerning software development:

if your code doesn't work, boil it down, if it still fails, try even
simpler code. Or even better... start from the simple case and
increase complexity during your development.

In your case this means, get rid of any systems, get rid of the
non-linearities and solve the static diffusion case, which is exactly
the part you are currently struggling with.

And when writing emails, please try not to cite everything first and
then adding your message at the bottom... if you cite the mail, do it
where appropriate and delete unnecessary content. This helps the
reader to focus on the important parts.

Christian

-- 
Prof. Dr. Christian Engwer 
Institut für Numerische und Angewandte Mathematik
Fachbereich Mathematik und Informatik der Universität Münster
Einsteinstrasse 62
48149 Münster

E-Mail	christian.engwer at uni-muenster.de
Telefon	+49 251 83-35067
FAX		+49 251 83-32729