| 3c6abd6b | 14-Oct-2014 |
Matthew G. Knepley <knepley@gmail.com> |
Merge branch 'mlange/feature-parallel-distribute'
* mlange/feature-parallel-distribute: (33 commits)
Plex+Label: Fix logic for label distribution - It is restrictive now, but doing it right would
Merge branch 'mlange/feature-parallel-distribute'
* mlange/feature-parallel-distribute: (33 commits)
Plex+Label: Fix logic for label distribution - It is restrictive now, but doing it right would mean a ton of infrastructure I think
Plex ex12: Fix output after label distribution fix
Plex: Only call SetFromOptions() after Distribute, since we cannot avoid it
TS ex11: Since DMPlexDistribute() calls SetFromOptions(), do not call it again, and cleaned up tests
Plex+FV: If using FVM adjacency, do not create ghost cells for faces in the overlap, or set BC on them
Plex: If using FVM adjacency, we must include the closure in the overlap
Plex+Label: All processes with a valid label should participate in label distribution - Formerly it was just proc 0
Plex: Add check that parallel renumbering fully covers the cone points
Plex: For one level adjacency, include the point itself since this makes sense when looking in parallel from the overlap
Plex: Cannot Malloc2() the arrays passed to PetscSFSetGraph()
Plex: Preallocation fixes for indegree > 1 - Again this should be rewritten
SNES ex62: Changes to output for fixes with quads
Plex ex1: Fixed LaTeX output
Plex ex12: Fixed overlap test
Plex: Fixed incorrect usage of PetscSFGatherBegin/End() in preallocation - This should be rewritten using new style
PetscSF: Incorrect check for computed degrees
Builder: Fix for Barry's change
PetscSF: Fixes for sparse leaves - When sf->mine exists, the leaves can be numbered non-sequentially. This causes allocation based upon sf->nleaves to underallocate. Replacing sf->nleaves with the maximum leaf number is sufficient in this case.
Plex: Fix leaks
Configure: How did this break?
...
Conflicts:
config/builder.py
show more ...
|