[petsc-users] PCBDDC - PCBDDCSetPrimalVerticesLocalIS

[Stefano Zampini]

Vincent also wrote me directly. I'm forwarding my answer.

Vincent,

to my knowledge there are two cases when this can happen

A) you specify indices (in local numbering) which are not lie on the
interface

B) you specify indices (in local numbering) which are on a part of the
interface shared by 2 subregions only

Note that A) is a natural requirement, whether B) is a bug in the current
release.

If you are in situation B), than you can either use 'master' where th ebug
has been already fixed or (and this is the case if you are also using
SLEPc), to simply comment lines 125, 127, 182 and 187 (4 lines in total) of
src/ksp/pc/bddc/bddcgraph.c. This should solve your problem.

Sorry for slow reply

Stefano
Il 15/ott/2014 22:04 "Jed Brown" <jed at jedbrown.org> ha scritto:

> "De Groof, Vincent Frans Maria" <Vincent.De-Groof at uibk.ac.at> writes:
> > I want to add a few primal vertices to the bddc preconditioner. But
> > when checking the coarse space (using pc_bddc_check_level 1), nothing
> > changes. Is there some special sequence that needs to be followed to
> > add vertices?
>
> Stefano, can you answer this?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20141016/2e1e0a14/attachment.html>