[Dune] Feature request.
robertk at mathematik.uni-freiburg.de
Wed Jun 22 11:10:30 CEST 2005
Christian Engwer wrote:
> We could solve point 2 if we introduce isLeaf for all codims. We then
> could just start all leveliterator for level 0 to maxlevel and return
> only the isLeaf entities. This would be reasonably fast for all grids
> offering fast leveliterators. For alberta this default implementation
> would be horribly slow, so alberta would be one of the first grid to
> become a real implementation.
Sounds good. So entities of every codim get the isLeaf method.
One other possibility is to use the hierarchic iterator and then the
method entity to achieve the LeafIterator functionality. Maybe we
schould test both implementations, as they are possible with the interface.
Another question is, which grids will use the default LeafIterator,
AlbertaGrid not, ALU3dGrid not, SGrid not, YaspGrid not (for those its a
simple typedef LevelIterator = LeafItertor) and what about UGGrid?
> I would prefere setting the default template parameter to
> All_Partition as this is the parameter needed for sequential programs
> and it's the parameter used by leveliterator. Updating existing
> programs is really easy... just as simple query-replace-regexp.
I thought that you do in parallel the same as in serial. For example
using a FV scheme on the leaf entities. Then you have to use the
LeafIterator with Interior_Partition. Now you have 2 possibilties,
one is to make sure that in serial the LeafIterator with
Interior_Partition does the "All_Partition" or the other is to write the
code twice, one for parallel and one for serial applications (which I
Robert Klöfkorn <robertk at mathematik.uni-freiburg.de>
Mathematisches Institut Tel: +49 (0) 761 203 5631
Abt. für Angewandte Mathematik Fax: +49 (0) 761 203 5632
More information about the Dune