[aspect-devel] Skip "cheap" solver phase for Stokes system

[Timo Heister]

Also note that even in the case that the cheap solver does not
converge within the given 30 iterations, the effort is not wasted.
This is because the intermediate result is used as a starting point
for the expensive solver.
That said, it is still likely that it is more efficient to skip the
cheap solver for "difficult" problems.

On Tue, Apr 23, 2013 at 9:24 AM, Wolfgang Bangerth
<bangerth at math.tamu.edu> wrote:
>
> All,
> Ian Rose asked yesterday whether there is a way to skip the 30 "cheap"
> Stokes solver iterations that are attempted at the beginning using a
> "cheaper" preconditioner (details in the Aspect paper, GJI 2012). Trying
> this cheaper preconditioner turns out to be a useful strategy for models
> with constant viscosity, but it's not so successful if you have a
> spatially variable viscosity for which these preconditioners don't work
> so well.
>
> In any case, long story short, there is now a top-level parameter
>    Number of cheap Stokes solver steps
> that defaults to the traditional 30 but can be set to any other value as
> well, including zero -- in the latter case skipping the attempt of using
> the cheap preconditioner altogether.
>
> Best
>   Wolfgang
> --
> ------------------------------------------------------------------------
> Wolfgang Bangerth               email:            bangerth at math.tamu.edu
>                                  www: http://www.math.tamu.edu/~bangerth/
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel



--
Timo Heister
http://www.math.tamu.edu/~heister/