[petsc-users] How to speed up geometric multigrid

[Dave May]

Hey Michele,

I think Galerkin MG, even with a jump in the coefficient of 10^3, can be
made more robust and faster for your problem.
I'd start by trying different smoothers. Try cheby/(bjacobi or asm), or
gmres/(bjacobi or asm). When using cheby, make sure you ask for the
spectrum estimate.
Definitely do more than one application of KSP on each level.

I don't think you should switch to AMG without first trying more robust
smoothers and profiling how much time is spent in the coarse grid solve.
I solve similar problems (geometry and size of the jump) in connection with
variable visocsity stokes problems on structured grids, and I'm yet to see
a case with AMG is more robust and or faster than Galerkin MG.


Cheers,
  Dave



On 23 September 2013 20:52, Mark F. Adams <mfadams at lbl.gov> wrote:

>
> >>
> > The simulation does start with both phases and the geometry is supposed
> to become more complex as the simulation progresses.
> > But so far the run is stopped before there are significant changes in
> the shape of the droplet.
>
> Humm,  not sure why you are seeing degradation then.  I imagine the
> initial geometry is not grid aligned so I'm not sure why you are seeing
> degradation in convergence rate.
>
> > I can give a shot to AMG: which options would you suggest to use.
>
> -pc_type gamg -pc_gamg_type agg -pc_gamg_agg_nsmooths 1
>
> > Also, how can I project out the constant from the rhs? Thanks a lot!
>
>
> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/KSP/KSPSetNullSpace.html#KSPSetNullSpace
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130923/70d9ed1c/attachment-0001.html>