[petsc-users] Extracting parts of DMPLEX mesh

[Matthew Knepley]

On Wed, Jun 19, 2019 at 3:49 AM Lawrence Mitchell <wence at gmx.li> wrote:

> On Wed, 19 Jun 2019 at 08:37, Matthew Knepley via petsc-users <
> petsc-users at mcs.anl.gov> wrote:
>
>> On Mon, Jun 17, 2019 at 9:49 PM Swarnava Ghosh <swarnava89 at gmail.com>
>> wrote:
>>
>>> Hi Matthew,
>>>
>>> I am primarily trying to interpolate fields which are defined on
>>> vertices. For a process, the point at which the interpolation to be carried
>>> out can be in an element that is owned by another process. My initial plan
>>> while writing the earlier email was to create local dmplexes in each rank
>>> which contains all the elements that are required for interpolation.
>>> However on second thoughts, this is too cumbersome.
>>>
>>
>> Thanks. This is much clearer to me now.
>>
>>
>>> What might be potentially better approach (ofcourse you will be an
>>> expert to judge) is to simply extend the ghost region to include the
>>> elements of the neighbor processes. In this way the points to interpolate
>>> at lie in the ghost region of the process. Is there a way I could extend
>>> the ghost region if I know the neighbor ranks?
>>>
>>
>> I see at least two cases:
>>
>>   1) Only a few points, anywhere in the domain
>>
>>       If you have a few points that can be anywhere in the domain, I
>> would use
>>
>>
>> https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/DMInterpolationCreate.html#DMInterpolationCreate
>>
>
> Does this work in parallel? If I look at the implementation here:
>
> https://www.mcs.anl.gov/petsc/petsc-current/src/dm/impls/plex/plexgeometry.c.html#DMLocatePoints_Plex
>
> It seems it complains unless run on comm_self.
>

The parallel stuff is shoved into DMInterpolationEvaluate(). I know it is
not pretty, but it is a very old part of Plex
which needs to be rewritten now that more point location stuff is there.

  Thanks,

    Matt


> Cheers,
>
> Lawrence
>
>>

-- 
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-users/attachments/20190619/323ccc7e/attachment.html>