How to control mesh partitioning (MPI) so that each partitioned mesh have access to surface where BC are applied


83
views
0
3 months ago by
Hello people,
I am trying to scale-up my model with the more MPI process i can.
I have an error when i request a lot of process (mpirun -n X) i think is related the way FEniCS is partitioning the mesh.
- First, i check that partitioned mesh are well balanced (i.e. they have roughly the same number of DOF). This is the case, great.
- Second, i check the number of DOF, for each partitioned mesh, which are located where the Neumann boundary conditions are applied in my model. Here is the issue.
I have found that some partitioned mesh have 0 node where flux BC ( something*ds(id_) ) are applied.

Then, is this possible to control the mesh partition so that each partitioned mesh are in contact with one part of the surface ds where BC are applied?

Thanks.


Community: FEniCS Project

1 Answer


1
3 months ago by
It's not possible to control the partitioning in way you describe.

However. the application of the boundary conditions should still work just fine. Perhaps you can explain what problem you're having on the boundaries. 
Please login to add an answer/comment or follow this question.

Similar posts:
Search »