[Dune] Parallel solvers in DUNE

[Markus Blatt]

Hi,

On Wed, Jun 23, 2021 at 02:31:57PM +0200, Max Firmbach wrote:
>I actually use Dune for everything right now, so the grid is read via 
>the Gmsh reader
>
>and "constructed" as UGGrid. The assembly process is also done with 
>different Dune modules.
>
>Yes, you are right my degrees of freedom are attached to the vertices.
>
>
>I actually implemented a similar code regarding what Oliver did in his 
>book, which works nice.
>
>But at some point i might need a better preconditioner, so maybe I 
>have a look at the overlapping versions.
>
>
>I had a look at the Dumux implementation and think I missed a major 
>point .... At some point I assume one
>
>has to define some parallel index for the communication/connectivity, 
>which I haven't done. Without this proper
>

>construction, the segmentation fault actually makes sense. I guess the 
>parallel index creation has to be done
>
>manually (like there's no automated thing for this in dune-istl) ?
>

I actually thought that you were doing that and just omitted it in your
test case. If your are not using a discretization tool like dune-pdelab or
others, then you have to do that, indeed.

Yell, if you need help. It is of course easier if the code is shared...

Is there any reason not use one of the discretization modules?

Markus

-- 

Dr. Markus Blatt - HPC-Simulation-Software & Services http://www.dr-blatt.de
Pedettistr. 38, 85072 Eichstätt, Germany
Tel.: +49 (0) 160 97590858