[petsc-dev] Petsc Support for Bandwidth Minimization

[Dave Nystrom]

Jed Brown writes:
 > On Mon, Jan 2, 2012 at 21:58, Dave Nystrom <dnystrom1 at comcast.net> wrote:
 > 
 > > I'm wondering if petsc has support for reordering the unknowns under the
 > > hood so that for the z-coupling case, the matrix can be cast as a compact
 > > tridiagonal matrix.  Essentially, this would be support for minimizing the
 > > bandwidth.
 > 
 > -pc_factor_mat_ordering_type rcm
 > 
 > You can get this with MatGetOrdering(A,MATORDERINGRCM,&isrow,&iscol)
 > 
 > We need to write man pages for all the MatOrderingTypes (RCM, ND, etc)

Thanks.  I just tried a run with -ksp_type cg -pc_type cholesky
-pc_factor_mat_ordering_type rcm -log_summary and did not see a performance
difference.  I'm trying to reproduce the results of our home grown cg with
cholesky preconditioner but for some reason, I am getting about 3x the
iteration count as I get with our home grown cg with cholesky preconditioner
using the z-coupling.  Not sure what is wrong.  Maybe I have a bug.

Is there any kind of increased level of verbosity that tell me things like
initial bandwidth of the matrix and the bandwidth after reordering?

Thanks,

Dave