repartition of parallel::shared::Triangulation

[Ce Qin]

Dear all,

When a shared::Triangulation is copied, the partition information is lost. The triangulation is on an invalid state. I think we should reimplement the copy_triangulation function that do the partition operation after the data is copied.

I also have a suggestion that we can add a virtual function repartition to its parent class parallel::Triangulation and reimplement it in shared::Triangulation. Does this make sense? If yes, I can help with this.

Best regards,

Ce Qin

[Wolfgang Bangerth]

Ce,

> When a shared::Triangulation is copied, the partition information is lost. The
> triangulation is on an invalid state. I think we should reimplement the

> *copy_triangulation* function that do the partition operation after the data
> is copied.

Yes, indeed. Can you open a bug at
https://github.com/dealii/dealii/issues
for this? Any help in fixing this would of course be much appreciated!


> I also have a suggestion that we can add a virtual function *repartition*to

> its parent class parallel::Triangulation and reimplement it in
> shared::Triangulation. Does this make sense? If yes, I can help with this.

That's a bit more tricky. A virtual function would only make sense if it could
be called from some of the functions in ::Triangulation. But it's not clear to
me where that would be. One could think about the end of
execute_coarsening_and_refinement(), for example, but that's not how this
currently works with p::d::Triangulation (which has a much more complicated
scheme). So I'm not sure how much making such a function virtual would
actually help.

Best
W.

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

[Ce Qin]

Dear Wolfgang,

 

Yes, indeed. Can you open a bug at
  https://github.com/dealii/dealii/issues
for this? Any help in fixing this would of course be much appreciated!

 

I have opened an issue at #2798. I will try to fix it.

Best regards,

Ce Qin