[petsc-dev] Mark, what's the rational for keeping the Chebyshev tuning inside GAMG?

Tobin Isaac tisaac at ices.utexas.edu
Wed Feb 25 10:24:34 CST 2015

On Wed, Feb 25, 2015 at 11:20:59AM -0500, Mark Adams wrote:
> >
> >
> >
> > We could make what you have (finding rows with just diagonals) the
> > default.  For, e.g., vector-valued problems with unaligned partial
> > Dirichlet conditions, the user could specify something else.
> >
> Sure.  But unaligned BCs are fine.  This problem just comes from users
> doing something stupid from our perspective (and not innately stupid) and
> putting a "big" fake number on the diagonal for the BC.
> Note, unaligned BCs require the null space be adjusted accordingly so my
> syntactic sugar (that Jed hates) of taking coordinates and doing this for
> users is not quite correct.  Bit probably does not matter at BCs in
> practice anyway.

MatProjectDirichlet() would be useful here, too: it would allow the
user to inject the right modifications to the near nullspace.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20150225/1ea8fc6d/attachment.sig>

More information about the petsc-dev mailing list