dolfin team mailing list archive

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

Jed Brown wrote:
> On Tue 2008-11-11 13:05, mspieg wrote:
> > On a related thread,
> >    I think there is a bug in the Newton solvers (in 0.8.1) in that they now
> > don't   zero the correction dx before each linear solve.  For direct solvers
> > this isn't a problem, however for krylov methods (in particular
> > PETScKrylovSolver), this ends up passing the last correction as the initial
> > guess for the next iteration.  Eventually, when the residual becomes
> > sufficiently small, this throws a divergence_tolerance error in Petsc. (To see
> > this, just change the solver in demo/nls/nonlinearpoisson to something like
> > gmres/ilu and use Petsc as the back end).
>
> This is absolutely correct.  I didn't understand Garth's comment earlier
> in this thread
>
> > > It's also useful for Newton solvers.
>
> How would you obtain a guess of the next Newton correction?

Most problems use an incremental-iterative approach (applying Newton's 
method in a series of steps). The dx for iteration 0 in step n will 
often be a good guess for the iteration 0 dx of step n+1.

Of course, after the initial correction, dx=0 is an appropriate guess.

Garth

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