[petsc-dev] DMPlex redistribution

[Matthew Knepley]

On Mon, Mar 27, 2023 at 6:57 PM Adrian Croucher <a.croucher at auckland.ac.nz>
wrote:

> hi Matt,
> On 28/03/23 00:14, Matthew Knepley wrote:
>
> Yes, it is a new check. I do not expect cells that are shared until we
> create an overlap. We
> can just put in a flag to shut it off, since otherwise it is a great check
> for finding bugs. Can
> you show me how you add cells, so we can make a small example that tests
> this to go in
> PETSc?
>
> Essentially I'm adding (one or more) cells nested "inside" some of the
> original DM cells to simulate dual porosity. As well as new cells I also
> create a new face between the original cell and the new one, and a new edge
> and point for the face (so the DAG is the same depth everywhere - otherwise
> it causes trouble).
>
> In the code, I create a new DMPlex, figure out the chart including the new
> cells etc., set the cone sizes and the cones. I symmetrize it, copy labels
> over from the original DM, set cell types (using DM_POLYTOPE_INTERIOR_GHOST
> for the new points) and set up the point SF for the new DM. If something is
> wrong with that point SF, would it cause the error I'm seeing when I
> redistribute?
>
> Yes, that would cause this. Your routine looks right to me, but it should
be easy to find a problem. Put a check
in at 3460 that h != dim, and also check that the height of minc_leaves(i)
is not 0.

  Thanks,

     Matt

> The code for setting up the new point SF is here (hope this link works):
>
>
> https://github.com/waiwera/waiwera/blob/55d26483b3392b70985fb25f998fbda3297c6154/src/mesh.F90#L3408
>
> - Adrian
>
> --
> Dr Adrian Croucher
> Senior Research Fellow
> Department of Engineering Science
> University of Auckland, New Zealand
> email: a.croucher at auckland.ac.nz
> tel: +64 (0)9 923 4611
>
>

-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20230327/cc06e727/attachment.html>