[Dune] Internal boundary surfaces via the Dune grid approach ???

[Carsten Gräser]

Jö Fahlke schrieb:
> Am Fri, 13. Nov 2009, 12:29:29 +0100 schrieb Oswald Benedikt:
>> If not, how should we do this ? 
>>
>> Just find out if a triangle is adjacent to 2 specific volume domains ? This may be viable but still be somewhat 
>> inefficient.
> 
> I think that is how it has to be done currently.  For boundary integrals it
> may pay off to find the elements adjacent to the fictitious boundary once and
> store entity pointers to them in a list.  It's possible that dune-subgrid
> (http://numerik.mi.fu-berlin.de/dune-subgrid/index.php) can do just that, but
> I haven't looked at it.
With dune-subgrid you can mark a subset of entities and use them as a hierarchical
grid on their own. This does explicitly incorporate the case where not the whole
domain is covered. In this case the subgrid boundary is an internal boundary
wrt the hostgrid. Subgrid then provides the usual intersection iterators.

However the subgrid leaf intersection iterator is rather slow, since it has
to search the leaf neighbors hierarchicaly to cover the case that the subgrid leaf
level is coarser than the hostgrid leaf level.

This is described in detail in the subgrid paper refered on the dune publications page.
If you have more questions regarding this feel free to ask me or Oliver.

Best regards
Carsten


> I don't think that you can get much more efficient than that for general
> fictitious boundaries unless you build a grid tuned to support just that.  (Of
> course, I'm not an expert on grids).
> 
> Bye,
> Jö.


-- 
----------------------------------------------------------------------
Carsten Gräser           | phone: +49-30 / 838-75349
Freie Universität Berlin | fax  : +49-30 / 838-54977
Institut für Mathematik  | email: graeser at math.fu-berlin.de
Arnimallee 6             |
14195 Berlin, Germany    | URL  : http://page.mi.fu-berlin.de/graeser
----------------------------------------------------------------------