[petsc-dev] DMPlex overlap redistribution

[Matthew Knepley]

On Tue, Mar 5, 2024 at 10:43 PM Adrian Croucher <a.croucher at auckland.ac.nz>
wrote:

> hi Matt,
>
> I've made a bit of progress with this, using a debugger.
>
> I wasn't setting the overlap for my dual-porosity DM, so it was
> defaulting to zero, and when it got to the DMPlexCheckPointSF() call it
> was doing the check for cells in the overlap when it shouldn't have been
> (this is only triggered "if (!overlap)").
>
> Now I do a DMPlexSetOverlap() to make sure the dual-porosity DM has the
> same overlap as the original DM. That gets rid of the error raised in
> DMPlexCheckPointSF(), as the overlap is now 1.
>
> However it is still falling over further on, when it tries to stratify
> the redistributed DM:
>
> [0]PETSC ERROR: Petsc has generated inconsistent data
> [0]PETSC ERROR: New depth 2 range [0,141) overlaps with depth 1 range
> [0,141)
>
> [0]PETSC ERROR: #1 DMPlexCreateDepthStratum() at
> /home/acro018/software/PETSc/code/src/dm/impls/plex/plex.c:4087
> [0]PETSC ERROR: #2 DMPlexStratify() at
> /home/acro018/software/PETSc/code/src/dm/impls/plex/plex.c:4211
> [0]PETSC ERROR: #3 DMPlexDistributeCones() at
> /home/acro018/software/PETSc/code/src/dm/impls/plex/plexdistribute.c:1164
> [0]PETSC ERROR: #4 DMPlexMigrate() at
> /home/acro018/software/PETSc/code/src/dm/impls/plex/plexdistribute.c:1659
> [0]PETSC ERROR: #5 DMPlexDistribute() at
> /home/acro018/software/PETSc/code/src/dm/impls/plex/plexdistribute.c:1778
>
> I have checked the depth label on my dual-porosity DM for a simple
> example in which this error is raised, and the label looks correct to me.
>
> However I have a memory of trying DMPlexStratify() on my dual-porosity
> DMs in the past and it not liking them. I create the depth label by
> copying its values from the original DM and then explicitly assigning
> depth values for the new dual-porosity cells, vertices, faces & edges.
>
> When the DM is redistributed I'd have thought it shouldn't be necessary
> (at least in my case) to stratify as the depth label is already there
> and just needs to be redistributed as well? But I guess that might not
> always be the case.
>
> Has anything changed in this part of the PETSc code recently that might
> make this stop working?
>

The strange part is that I have not changed anything in the depth
algorithm. It is very old.

Let's do this. Can you send me a dual porosity mesh, and I will figure out
why the stratification
algorithm does not work? I think you can just output the Plex as binary or
HDF5, but you could
also give me code that makes one.

  Thanks!

     Matt


> - Adrian
>
> --
> Dr Adrian Croucher
> Senior Research Fellow
> Department of Engineering Science
> Waipapa Taumata Rau / 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://urldefense.us/v3/__https://www.cse.buffalo.edu/*knepley/__;fg!!G_uCfscf7eWS!foavmxCV-MGSYM4r_v-nLq6CwCQa179y7pSTSYY8YORf6SJ8Djlj6C8Kmw-pzdn5t3BFgNYEg1NGGP5yWJ9T$  <https://urldefense.us/v3/__http://www.cse.buffalo.edu/*knepley/__;fg!!G_uCfscf7eWS!foavmxCV-MGSYM4r_v-nLq6CwCQa179y7pSTSYY8YORf6SJ8Djlj6C8Kmw-pzdn5t3BFgNYEg1NGGO4mBoQ0$ >
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20240306/177d0e54/attachment-0001.html>