← Back to team overview

dolfin team mailing list archive

Re: Fixed implementation of NewPreconditioner (pcimpl.h is now used).

 

On Mon, Feb 14, 2005 at 10:46:22AM -0600, Anders Logg wrote:
> Does it work now? If it does, we could think of implementing a simple
> preconditioner for the multi-adaptive Newton-solver (damped fixed
> point iteration with dG(0)).
> 
> /Anders
> 

It doesn't work well yet. It works for some cases, but others not,
giving different errors (erroneous solution, LAPACK errors or
nonconvergence). My test so far has been an identity preconditioner,
which should be the same as not preconditioning at all. I need to do
some more testing though, perhaps with Jaocbi instead.

  Johan



Follow ups

References