[petsc-users] Propagating a constant marker with AMG

[Jed Brown]

John Mousel <john.mousel at gmail.com> writes:

> If I recall correctly, GAMG has a somewhat high setup cost and this is a
> solver I'll only be using for one single solve. It's not a physical field,
> it's just a marker; 1 inside, 0 outside a surface described by a CAD file.
> This marker is being propagated into the volume of an octree volumetric
> grid. I'm just taking advantage of the fact that the solution of the
> Laplace equation with constant BC's is the constant.

Oh, I see.  Enforce your boundary conditions so that the problems on
each side decouple, then use -pc_gamg_agg_nsmooths 0 (unsmoothed
aggregation).  Profile and let us know if it's a bottleneck.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 818 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150331/d8bb8829/attachment.pgp>