On Wed, May 19, 2021 at 4:36 AM Thibault Bridel-Bertomeu < [email protected]> wrote:
Hello Matt,
Le mar. 18 mai 2021 à 20:31, Matthew Knepley <[email protected]> a écrit :
On Tue, May 18, 2021 at 12:16 PM Thibault Bridel-Bertomeu < [email protected]> wrote:
Hi Matthew,
Thank you very much for your quick answer, as always !
Cool.
Le mar. 18 mai 2021 à 17:46, Matthew Knepley <[email protected]> a écrit :
On Tue, May 18, 2021 at 5:19 AM Thibault Bridel-Bertomeu < [email protected]> wrote:
Dear all,
I am playing around with creating DMPlex from a periodic Gmsh mesh (still in the same finite volume code that solves the Euler equations) because I want to run some classical periodic test cases from the literature (like convecting a vortex and so on). I got from the mailing list and the examples that I gotta use -dm_plex_gmsh_periodic for the reader and the DM creator to do their job properly and read the $Periodic section from the Gmsh mesh. That works fine, I can write the DM in HDF5 format afterwards and the periodicity is clearly there (i can see in Paraview that opposite sides are "attached" to each other). Now, in the context of the finite volume solver, I wrote my one right-hand side term computing routine, and I rely heavily on routines such as DMPlexGetFaceFields. Those routines, in turn rely on the ghost cells and ... I am having an issue with DMPlexConstructGhostCells. When I use -dm_plex_gmsh_periodic, DMPlexCreateFromFile works fine, DMSetBasicAdjacency(dm, true, false) works fine, DMPlexDistribute works fine, but when I get to DMPlexConstructGhostCells, I get
"DM has boundary face 2048 with 2 support cells"
I understand the DM begin periodic, even the "boundary" faces have two neighbors because of the topological folding, so ... yeah, the ghost cells cannot really exist. But then how can I proceed ? Because I need that "ghost" label for all the routines I call from the RHS computing routine ...
Thanks for the example. I have run this myself. It looks to me like the mesh you are providing. Has no boundary. It is doubly periodic (a torus). When I tell ConstructGhostCells() to create a label for the boundary, rather than using "Face Sets", it creates an empty label because no faces have only 1 neighbor.
Yes indeed, up is down and right is left for this mesh, the idea being of emulating an infinite medium. So you manage to get through ConstructGhostCells without a segfault ? What did you specify instead of PETSC_NULL_CHARACTER ?
The NULL tells it to use the label "Face Sets", but that label coming from Gmsh marks edges that are not actually boundaries. I asked PETSc to create a boundary label named "marker" using https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/DMPLEX/DMPlexMa... It was empty, but everything goes through fine.
OK thanks ! I had to write the F90 wrapper for DMLabelCreate but aside from that, everything indeed goes through fine with the ConstructGhostCells.
Also, does it still create a « ghost » label even if it’s empty ?
Yes.
Do you think that will hold for later uses in the GetFaceFields and similar routines ?
Yes.
I read through the source and it seems it won’t like it much if there are no ghost at all, but since I did not pass the ConstructGhostCell step I’m not sure.
It should work fine, you just will not need boundary conditions.
Actually, there is an issue with the corner cells. My mesh is periodic in both directions and it turns out the corner cells accumulate way too much numerical flux and the computation crashes after 4 iterations or so - it happens both with quads and tris. Whether I use my version of the RHS computing routine or the DMPlexTSComputeRHSFunctionFVM leads to the same crash. I'll keep investigating, but if you have any idea what could go wrong ...
Hmm, I have not run a fully periodic case. I guess I could set one up. However, it must be the case that we are conservative, since we first find a flux on every edge, and then update cells. So the corner cells should not be able to accumulate more than the total mass/4. I would first check at each step that total mass is conserved. If not, we can track that down. Thanks, Matt
Thanks !!
Thibault
Thanks,
Matt
Thanks !
Thibault
Thanks,
Matt
I attach to this email the periodic Gmsh mesh I am playing with, and the routine from my code where I create the DM and so on:
subroutine initmesh
PetscErrorCode :: ierr DM :: dmDist, dmGhost integer :: overlap PetscViewer :: vtkViewer
call PetscLogEventBegin(timer_initmesh, ierr); CHKERRA(ierr)
! Number of neighbours taken into account in MP communications(1 - Order 1; 2 - Order 2) overlap = 1
call PetscPrintf(PETSC_COMM_WORLD, "Initializing mesh...\n", ierr) ; CHKERRA(ierr)
! Force DMPlex to use gmsh marker ! call PetscOptionsSetValue(PETSC_NULL_OPTIONS, "-dm_plex_gmsh_use_marker", "true", ierr); CHKERRA(ierr)
! Read mesh from file name 'meshname' call DMPlexCreateFromFile(PETSC_COMM_WORLD, meshname, PETSC_TRUE, dm, ierr); CHKERRA(ierr)
! Distribute on processors ! Start with connectivity call DMSetBasicAdjacency(dm, PETSC_TRUE, PETSC_FALSE, ierr) ; CHKERRA(ierr)
! Distribute on processors call DMPlexDistribute(dm, overlap, PETSC_NULL_SF, dmDist, ierr) ; CHKERRA(ierr)
! Security check if (dmDist /= PETSC_NULL_DM) then ! Destroy previous dm call DMDestroy(dm, ierr) ; CHKERRA(ierr) ! Replace with dmDist dm = dmDist end if
! Finalize setup of the object call DMSetFromOptions(dm, ierr) ; CHKERRA(ierr)
! Boundary condition with ghost cells call DMPlexConstructGhostCells(dm, PETSC_NULL_CHARACTER, PETSC_NULL_INTEGER, dmGhost, ierr); CHKERRA(ierr)
! Security check if (dmGhost /= PETSC_NULL_DM) then ! Destroy previous dm call DMDestroy(dm, ierr) ; CHKERRA(ierr) ! Replace with dmGhost dm = dmGhost end if
if (debug) then ! Show in terminal call PetscPrintf(PETSC_COMM_WORLD, ":: [DEBUG] Visualizing DM in console ::\n", ierr); CHKERRA(ierr) call DMView(dm, PETSC_VIEWER_STDOUT_WORLD, ierr) ; CHKERRA(ierr) ! VTK viewer call PetscViewerCreate(PETSC_COMM_WORLD, vtkViewer, ierr) ; CHKERRA(ierr) call PetscViewerSetType(vtkViewer, PETSCVIEWERHDF5, ierr) ; CHKERRA(ierr) call PetscViewerFileSetMode(vtkViewer, FILE_MODE_WRITE, ierr) ; CHKERRA(ierr) call PetscViewerFileSetName(vtkViewer, "debug_initmesh.h5", ierr) ; CHKERRA(ierr) call DMView(dm, vtkViewer, ierr) ; CHKERRA(ierr) call PetscViewerDestroy(vtkViewer, ierr) ; CHKERRA(ierr) end if
call PetscPrintf(PETSC_COMM_WORLD, "Done !\n", ierr) ; CHKERRA(ierr)
call PetscLogEventEnd(timer_initmesh, ierr); CHKERRA(ierr)
end subroutine initmesh
Thank you very much for your help !
Best regards,
Thibault Bridel-Bertomeu
-- 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/>
-- Thibault Bridel-Bertomeu — Eng, MSc, PhD Research Engineer CEA/CESTA 33114 LE BARP Tel.: (+33)557046924 Mob.: (+33)611025322 Mail: [email protected]
-- 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/>
-- 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/>