Hello Martin,

I started by modifying step-33 which was using trilinos solvers. I dont
really need to use trilinos at present.

So my best options would be to use dealii::SolverGMRES with
dealii::PreconditionBlockSSOR.

Thanks
praveen

On Fri, Mar 25, 2011 at 9:48 PM, Martin Kronbichler <
[email protected]> wrote:

> Dear Praveen,
>
>
> > For steady state problems, it is recommended to use a block SSOR
> > preconditioner. This seems to correspond to the AZ_precond = AZ_sym_GS
> > option in aztec. Am I right ?
> >
> >
> > But if I use this option, I get an error that
> >
> >
> > "sym GS preconditioning can only be used with MSR matrices"
>
> The error message basically tells you that you can't use the Aztec
> preconditioners in conjunction with the matrix format used in
> TrilinosWrappers::SparseMatrix, which is of type Epetra_CrsMatrix. What
> you could do is to use a TrilinosWrappers::SolverGMRES object and give
> it a TrilinosWrappers::SparseMatrix and
> TrilinosWrappers::PreconditionSSOR. That should work I think. But then
> you're still using a point-by-point SSOR and not a _block_ SSOR, for
> which you would need to invert some diagonal blocks. For that we do
> currently not have any wrapper classes to Trilinos, and the only option
> would be PreconditionBlock with deal.II sparse matrices. What is the
> motivation for using Trilinos matrices? Do you plan to run your program
> in parallel? If not, you can use deal.II matrices just fine with block
> SSOR.
>
>
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to