[petsc-dev] Stokes and saddle point solvers in PETSc?

[Barry Smith]

On Jun 14, 2011, at 8:45 PM, Matthew Knepley wrote:

> On Wed, Jun 15, 2011 at 1:42 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
> On Jun 14, 2011, at 8:36 PM, Matthew Knepley wrote:
> 
> > On Wed, Jun 15, 2011 at 1:20 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> >
> > On Jun 14, 2011, at 7:59 PM, Jed Brown wrote:
> >
> > > On Wed, Jun 15, 2011 at 02:53, Barry Smith <bsmith at mcs.anl.gov> wrote:
> > >  It seems we should provide a DMSDA that is built specifically for staggered grids, with the correct number of "slots" in the correct "locations", it would have appropriate grid hierarchies and interpolation/restriction. Not terribly hard but a bit of basic plumbing code. I would rather have this then try to "tack on" the current DA a bunch more stuff.
> > >
> > > Yeah, though there are many different staggered discretizations so this could end up being a big project. Not conceptually hard, just a lot of code. DMDA is already not especially small.
> >
> >  DMDA started out small. DMSDA will start out small :-)
> >
> >  Actually before we do DMSDA we probably should do a code review of DM, then of DMDA and slim down DMDA a good amount.
> >
> > My vote would be to layer PetscSection on top of DMDA. That would allow arbitrarily complex data layout over a dead simple
> > topology. Its halfway to DMMesh, and it should be the easy half.
> 
>  It won't give the proper layout of degree's of freedom relative to their neighbors of the stagger mesh nor communicate the right ghost points.
> 
> Why not? I have done it before. The trial implementation (in C++) is in src/dm/impls/cartesian. I will just reimplement that completely
> in C.

   If so, why not dump DMDA also and replace it with a dead simple implementation based on PetscSections?

   Barry

> 
>   Matt
>  
> 
>   Barry
> 
> >
> >    Matt
> >
> >
> >   Barry
> >
> >
> >
> >
> >
> > --
> > What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> > -- Norbert Wiener
> 
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> -- Norbert Wiener