-
Committer:
Mark Filipiak
-
Date:
2012-08-14 09:46:08 UTC
-
Revision ID:
mjf@staffmail.ed.ac.uk-20120814094608-ndww9ivsto0ke1cu
Use nprocs=4 in petsc_readnsolve. With nprocs=3, flredecomp is making a
part with no volume elements (and another has no surface elements, which
is possibly OK). An alternative is to use a different partitioner in
Zoltan. ParMetis is the default and gives this behaviour. Scotch and
Zoltan's graph and hypergraph partitioners seem to give more a more even
decomposition.