ffc team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

Anders Logg wrote:
> On Mon, Jun 08, 2009 at 12:52:41PM +0200, Jed Brown wrote:
> > Anders Logg wrote:
> > > On Fri, Jun 05, 2009 at 06:00:36PM -0500, Brandon Chabaud wrote:
> > > > Is there some way to inform the Krylov solver that constant pressure is in the
> > > > null space of the operator from Dolfin?  The only options that I could find for
> > > > the KrylovSolver command were setting the solver type and setting the
> > > > preconditioner type.
> > > No, that's not included in the interface. You will have to extract the
> > > PETSc Mat pointer using mat() from PETScMatrix and use PETSc calls.
> > Almost.  You need to get the KSP object from the solver and call
> > KSPSetNullSpace().
>
> It looks like the KSP object is not available in the PETScKrylovSolver
> interface anymore. Why has it been removed. Garth?

Not sure that we ever had it. Anyway, I've added it now. 
PETScKrylovSolver may however need some work to make sure that user 
options are not overridden.

Garth


> ------------------------------------------------------------------------
>
> _______________________________________________
> FFC-dev mailing list
> FFC-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/ffc-dev