[petsc-dev] DMPlex and scattering

[Matthew Knepley]

On Mon, May 5, 2014 at 12:09 AM, Jed Brown <jed at jedbrown.org> wrote:

> Adrian Croucher <a.croucher at auckland.ac.nz> writes:
>
> > hi
> >
> > I'm writing a finite volume test code, mostly based on the TS ex11
> > tutorial example, but in Fortran. It works ok in serial but not in
> parallel.
> >
> > I don't really understand how the global/local scattering is done in TS
> > ex11. There don't seem to be any DMGlobalToLocalBegin() /
> > DMGlobalToLocalEnd() calls in the main RHS function routine
> > (RHSFunctionLocal_Upwind()), as there are in most of the other
> > (non-DMPlex) examples.
>
> DMTSSetRHSFunctionLocal registers a "local" function.  The
> global-to-local scatters are done in TSComputeRHSFunction_DMLocal().
>
> Matt, WTF is up with the "#if 0" crap strewn throughout this function?
>

That is to remind me where we will need things when we have both FV and FE
here. FE will have communication even in the RHS.


> We need Fortran interfaces for the TS functions.  SNES has Fortran
> interfaces.


Yes. I was getting to that, but first I was pulling in all the ex11 stuff
to TS utils.

  Matt


> > Presumably the setting up of ghost cells is done by the call to
> > DMPlexConstructGhostCells(). Does the ghost scattering happen in the
> > call to DMPlexInsertBoundaryValuesFVM()?
> >
> > Unfortunately it looks like neither of these routines has a Fortran
> > interface as yet, so I can't really try them. In the meantime, I've
> > tried to do it using DMGlobalToLocalBegin() / DMGlobalToLocalEnd()
> > calls, like in the other examples, but this isn't working as yet. So I'm
> > a bit stuck. Should this approach work with DMPLex, or not?
> >
> > Cheers, 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20140505/10c2ba29/attachment.html>