[petsc-dev] stencils, ex2f.F, openmp

[Jed Brown]

Svetlana Tkachenko <svetlana.tkachenko at fastmail.fm> writes:
> I realised I have a set of dummy items to keep around the main matrix
> (1, x_dim, with only items 2 ... x_dim-1 included into matrix A
> itself). They are boundary conditions and the existing (serial)
> preconditioner relies on them.

I'm not sure I parsed this description correctly, but it sounds like
you're talking about Dirichlet boundary points.  The most common
approach is to assemble the matrix so that those rows of the matrix look
like rows of the identity, then just put the known (perhaps 0.0) value
in the right hand side.  The ensures a nonsingular operator.  There is
another trick to preserve symmetry, though it's easiest to express in
the context of nonlinear problems.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20131107/72c36883/attachment.sig>