dolfin team mailing list archive

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

Anders Logg wrote:
> On Sat, Jul 11, 2009 at 02:18:30PM +0200, DOLFIN wrote:
> > One or more new changesets pushed to the primary dolfin repository.
> > A short summary of the last three changesets is included below.
> >
> > changeset:   6476:7cf7eeca31340e13854d69ba8166e969e965ec25
> > tag:         tip
> > user:        "Garth N. Wells <gnw20@xxxxxxxxx>"
> > date:        Sat Jul 11 13:18:15 2009 +0100
> > files:       dolfin/la/EpetraKrylovSolver.cpp dolfin/la/EpetraKrylovSolver.h dolfin/la/EpetraLUSolver.cpp dolfin/la/EpetraLUSolver.h dolfin/la/KrylovSolver.cpp dolfin/nls/NewtonSolver.cpp
> > description:
> > Update some Epetra wrapper and change default Krylov tolerance.
>
> Isn't 1e-6 a little loose? We've had problems with some benchmarks
> giving strange results when the tolerance was not set tight enough.

1e-15 was way too strong - simply wouldn't work for some problems. The 
PETSc default is 1e-5, so 1e-6 seemed pretty reasonable to me. Perhaps 
we should make a habit of always setting the tolerance in the demos to 
indicate that users should consider it.

Garth

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