[petsc-dev] DMGetNeighbors and DMPlex

[Pierre Jolivet]

> On 13 Apr 2020, at 4:42 PM, Matthew Knepley <knepley at gmail.com> wrote:
> 
> On Mon, Apr 13, 2020 at 6:14 AM Pierre Jolivet <pierre.jolivet at enseeiht.fr <mailto:pierre.jolivet at enseeiht.fr>> wrote:
> Hello,
> I’m trying to figure out why the result of DMGetNeighbors when a DMPlex is distributed with no overlap can be nonsymmetric.
> 
> Right now, the Plex function only calls PetscSFGetRootRanks(). It should probably call PetscSFGetLeafRanks() as well and merge the lists.
> Have time to make the fix? You could commit this nice test :)

Sure, I’ll try to give it a go.

Thanks,
Pierre

>   Thanks,
> 
>     Matt
>  
> Is this intended? If so, is there some other easy way to get the equivalent list as in the overlapping case (I’m interested in the list of ranks sharing at least one point)?
> Here is a MWE.
> $ patch -p1 < patch.txt
> $ cd src/dm/impls/plex/tests/
> $ make ex1
> $ mpirun -n 2 ./ex1 -dim 2 -cell_simplex 0 -interpolate -domain_shape box -domain_box_sizes 3,3 -test_redistribute
> 1 <— how come rank 0 has 1 neighbor
> 0 <— but rank 1 has 0 neighbor?
> $ mpirun -n 2 ./ex1 -dim 2 -cell_simplex 0 -interpolate -domain_shape box -domain_box_sizes 3,3 -overlap 1
> 1
> 1
> Overlap: 1
> 
> Thanks,
> Pierre
> 
> diff --git a/src/dm/impls/plex/tests/ex1.c b/src/dm/impls/plex/tests/ex1.c
> index 4a5d051a66..ac2ce8eb20 100644
> --- a/src/dm/impls/plex/tests/ex1.c
> +++ b/src/dm/impls/plex/tests/ex1.c
> @@ -3,0 +4 @@ static char help[] = "Tests various DMPlex routines to construct, refine and dis
> +#include <petscsf.h>
> @@ -420,0 +422,8 @@ PetscErrorCode CreateMesh(MPI_Comm comm, AppCtx *user, DM *dm)
> +      PetscInt nranks;
> +      const PetscInt* ranks;
> +      PetscSF sf;
> +      ierr = DMGetPointSF(distributedMesh, &sf);CHKERRQ(ierr);
> +      ierr = PetscSFSetUp(sf);CHKERRQ(ierr);
> +      ierr = DMGetNeighbors(distributedMesh, &nranks, &ranks);CHKERRQ(ierr);
> +      ierr = PetscSynchronizedPrintf(PETSC_COMM_WORLD,"%D\n",nranks);CHKERRQ(ierr);
> +      ierr = PetscSynchronizedFlush(PETSC_COMM_WORLD,PETSC_STDOUT);CHKERRQ(ierr);
> @@ -434,0 +444,8 @@ PetscErrorCode CreateMesh(MPI_Comm comm, AppCtx *user, DM *dm)
> +      PetscInt nranks;
> +      const PetscInt* ranks;
> +      PetscSF sf;
> +      ierr = DMGetPointSF(overlapMesh, &sf);CHKERRQ(ierr);
> +      ierr = PetscSFSetUp(sf);CHKERRQ(ierr);
> +      ierr = DMGetNeighbors(overlapMesh, &nranks, &ranks);CHKERRQ(ierr);
> +      ierr = PetscSynchronizedPrintf(PETSC_COMM_WORLD,"%D\n",nranks);CHKERRQ(ierr);
> +      ierr = PetscSynchronizedFlush(PETSC_COMM_WORLD,PETSC_STDOUT);CHKERRQ(ierr);
> 
> 
> -- 
> 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/20200413/06022a56/attachment.html>