[Trilinos-Users] Using the EpetraExt_CrsGraph_MapColoring class
egboman at sandia.gov
Mon Apr 13 18:19:15 MDT 2009
Isorropia supports *partitioning* of nonsymmetric matrices using the
hypergraph partitioner in Zoltan, but coloring is a different issue. A
quick work-around for coloring is to color the symmetrized matrix A+A^t.
This will work but you will usually use more colors than necessary.
On Mon, 2009-04-13 at 16:16 -0600, Bernsen, E. wrote:
> Hi Erik,
> Thanks for the very quick response! I'm trying to color an assymetric
> matrix. The sparsity pattern of the matrix is easily extended to be
> symmetric. For me it's important that the coloring can be used by the
> FiniteDifferenceColoring class to construct a Jacobian matrix and I
> don't think that's the case of assymetric matrices (even if the
> sparsity pattern is symmetric).
> In the user guide I read that for non-symmetric problems hypergraph
> partitioning is recommended. Is support for this in Isorropia
> available? Or is there any other method you can suggest for coloring
> assymetric matrices?
> -----Original Message-----
> From: trilinos-users-bounces at software.sandia.gov on behalf of Erik
> Sent: Mon 4/13/2009 5:47 PM
> To: Bernsen, E.
> Cc: trilinos-users at software.sandia.gov
> Subject: Re: [Trilinos-Users] Using the EpetraExt_CrsGraph_MapColoring
> Hi Erik,
> I'm not sure about your specific question, but you should know that
> more recent work on coloring has been moved to the Zoltan and
> packages. Take a look at Isorropia::Epetra::Colorer at
> This package uses a more scalable parallel algorithm, so you may want
> try it. One drawback is that the current version is limited to
> structurally symmetric matrices. We plan to support nonsymmetric
> matrices in the next release.
> On Mon, 2009-04-13 at 09:24 -0600, Bernsen, E. wrote:
> > Hello,
> > I'm trying to use the EpetraExt_CrsGraph_MapColoring class to color
> > the independent columns of a graph and then use the
> > FiniteDifferenceColoring class to construct a Jacobian matrix.
> > Sometimes the coloring algorithms give unexpected results. For
> > instance consider the attached program (test.C) which creates a 2x2
> > triangular matrix and then colors the corresponding graph. I think
> > coloring of this matrix should result in two colors. However
> > to the output of the program (out.txt) only one color is required.
> > program was run on a single processor in this case.
> > When using the GREEDY algorithm instead of JONES_PLASSMEN the
> > seems to work allright as long as the program is run on a single
> > processor. If more than one processor is used results seem to be
> > incorrect.
> > Is there a bug in the EpetraExt_CrsGraph_MapColoring class or is the
> > use of this class in test.C incorrect?
> > Regards,
> > Erik Bernsen
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
More information about the Trilinos-Users