On Oct 25, 2011, at 4:34 PM, Jed Brown wrote:
> Now that lots of people are getting -pc_type mg without using DMMG, we don't
> have FGMRES (or GCR) by default. But the current default for smoothers is to
> use -mg_levels_ksp_type gmres which is nonlinear. This causes -pc_type mg to
> fail even when MG works great.
>
> I think that in the end, most people don't actually want to use GMRES on
> levels (because the synchronization costs dominate on coarse levels), but
> it's more robust when you are playing with smoothers. Should we switch to
> using Richardson on levels by default?
Chebychev?
I agree the right default should be a linear smoother.
Barry
>
> Or should we have some way to automatically identify a preconditioner as
> nonlinear so that the library can default to a flexible Krylov method in such
> cases?
