[Dune] Parallel nonoverlapping instantionary diffusion problem

[Sascha Zelzer]

Hi,

I am trying to do a parallel simulation using nonoverlapping grids 
(ALUGrid with tetrahedra) for calculating the solution to an 
instationary diffusion problem.

My sequential version is working fine. However, the parallel version 
does not seem to propagate the solutions on the grid partition 
boundaries to the other processes correctly. I started using the code in 
"instationarytest.cc" and for the parallel version I am using

typedef Dune::PDELab::NonoverlappingConformingDirichletConstraints CON;
CON con;
...
con.compute_ghosts(gfs);


as the constraints type and set the last template parameter of my 
Dune::PDELab::GridOperator template instantiations to "true". In 
essence, my code still looks pretty much like in "instationarytest.cc" 
with the mentioned modifications.

Did I overlook something important for that kind of problem? How is 
responsible for communicating solution values between grid partition 
boundaries?

Thanks a lot for any hints,

Sascha