Trust Region problem

[Matthew Knepley]

On 3/17/07, Gong Ding <gdiso at ustc.edu> wrote:
> Hi all,
>    I use PETSC's nonelinear solver to do
> emiconductor simulation. I found a strange problem
> with trust region solvers. It usually works well
> if LU method is used for inner linear solver. But
> when krylov method i.e. GMRES, is employed, it
> often stops due to pnorm is small, but still
> remains large error of function norm. How to solve
> this problem? LU method is much slower than i.e.
> GMRES.

This sounds like GMRES is not actually converging (you can check this
using -ksp_monitor). You need a more powerful preconditioner, which
unfortunately usually depends a lot on the problem.

  Matt

> G.D.
> --
> USTC Alumni Email System
>
>


-- 
One trouble is that despite this system, anyone who reads journals widely
and critically is forced to realize that there are scarcely any bars to eventual
publication. There seems to be no study too fragmented, no hypothesis too
trivial, no literature citation too biased or too egotistical, no design too
warped, no methodology too bungled, no presentation of results too
inaccurate, too obscure, and too contradictory, no analysis too self-serving,
no argument too circular, no conclusions too trifling or too unjustified, and
no grammar and syntax too offensive for a paper to end up in print. --
Drummond Rennie