[petsc-dev] calling DMPlexPointLocalRef() from fortran

[Adrian Croucher]

hi,

On 08/04/14 23:31, Matthew Knepley wrote:
>
> 1) ex33 will probably get eliminated. It is mostly a copy of ex11 
> which does work. Use that.

OK- to begin with I tried running TS ex11 with the option 
-ufv_vtk_cellgeom (to get it to write out the partition vector to VTK):

mpirun -np 2 ex11 -f sevenside-quad-15.exo -ufv_vtk_cellgeom

but that gave the same error message I was getting yesterday with my 
test code. I tried it with some of the other exo meshes, with the same 
error. So it looks like there might be a problem there (unless I'm 
running it with inappropriate input?) If I don't use -ufv_vtk_cellgeom 
it runs fine.
>
> 2) Are you solving something using FVM?
>
> 3) The VTK/HDF5 output now relies on being able to set the boundary 
> values automatically. Here is it trying
>      to do it for FVM. Right now, it assumes that you have told it 
> about the face geometry, which happens in ex11.
>     If you do not want this, I need to understand what you want to do.

Yes, I'm using FVM. To begin with I'm trying to set up a simple test 
code to solve an advection-diffusion problem on an unstructured FVM 
mesh, to learn how PETSc (particularly DMPlex) works. If that works out 
I'll use it as a basis for solving my real problem.

Cheers, Adrian

-- 
Dr Adrian Croucher
Department of Engineering Science
University of Auckland
New Zealand
tel 64-9-373-7599 ext 84611

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20140409/ddab3495/attachment.html>