[petsc-dev] DMDA_BOUNDARY_GHOSTED

[Mark F. Adams]

On Sep 26, 2013, at 9:17 PM, Matthew Knepley <knepley at gmail.com> wrote:

> On Thu, Sep 26, 2013 at 1:15 PM, Mark F. Adams <mfadams at lbl.gov> wrote:
> >
> > We have ghost cells in TS ex11, and I think we do them the correct way :) We have a function you register that fills them up, and we
> > loop over them explicitly.
> >
> 
> Well damn Matt, this looks like the right way to do FV in PETSc, regardless of the nice BC abstraction, but am I going to get be seduced and crash into this island and turn into stone?  i.e., can stupid people deal with your Sections and Plexes and crap (Barry?).
> 
> So the important thing here is the organization of the boundary functions. You call
> 
>   ModelBoundaryRegister(bdFunc, bdMarker)
> 

OK, I guess I see this as sugar. Don't get me wrong I like sugar.  But I will at least keep the infrastructure for Neumann BCs.

> where the function is
> 
>   PhysicsBoundary_Advect_Inflow(Model mod, PetscReal time, const PetscReal *centroid, const PetscReal *normal, const PetscScalar *xInterior, PetscScalar *xGhost, void *ctx)
> 

Yes, I'm comfortable to this point, I just wonder if I'm getting sucked into a morass of topology or something that I did not even think about taking in grad school.

> Then we loop over all the mesh pieces marked with each boundary marker, and call the function on those faces. You get the interior
> value, and make the ghost accordingly.

I think you are doing the same basic thing that I am used to, where you set the ghost values to satisfy the BCs for the current state, apply an operator of some sort, rinse and repeat ….

> I think you would do exactly the same thing in your simple example below. The only optimization
> would be in handling the topology, which you do not see anyway (unless you are a masochist).
> 


Me a masochist? au contraire :)

This looks good.  I will work on simplifying this example to make a well designed (a la ex11), albeit simple, FV test problem for my current favorite FMG test with an exact solution, and make a converge test out of it.  I really need something simple for SR and this is a coarse grid solver.  I will ask about SR support when I get there.

Mark






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130926/c238a1cc/attachment.html>