dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On Nov 9 2009, Anders Logg wrote:

> On Mon, Nov 09, 2009 at 09:30:44PM +0000, Garth N. Wells wrote:
> > On Nov 9 2009, Anders Logg wrote:
> >
> > >Incidentally, we (myself, Andre Massgin and Mats Larson) discussed a
> > >generalization of assembly over interior facets last week that also
> > >covers the case of intersections between non-matching meshes.
> > >
> > >The central concept in assemble_interior_facets is that of a facet and
> > >its two neighboring cells.
> >
> > If the 'facet' is interior to a cell, why not just assemble it as
> > part of the cell integral? We're doing this already for surfaces
> > which pass through cells.
>
> Since it requires information from two different cells that intersect
> along the "facet". And those cells come from two different meshes.

OK, so the facet really is the facet of a cell and not just part of a 
surface?

Garth

> --
> Anders
>
>
> > Garth
> >
> > Our current implementation assumes that
> > >this facet is a facets for both cells in the pair. But a
> > >generalization would include assembly over triples
> > >
> > > (facet, cell0, cell1)
> > >
> > >where facet is contained in the closure of each of the two cells, but
> > >it may or may not be a facet for the two cells. It can be part of a
> > >facet or part of the interior. If we can implement this (and generate
> > >the appropriate code), we would be able to cover quite a few
> > >interesting cases (like Nitsche on overlapping non-matching meshes).
> > >