dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Wed, Jun 25, 2008 at 10:07:53AM +0200, Johan Hoffman wrote:
> > On Tue, Jun 24, 2008 at 11:00:54AM +0200, Johan Hoffman wrote:
> >> Sounds good! For inspiration on the parallelization I attach some
> >> reading
> >> for the vacation; the MSc thesis of Niclas Jansson (a new phd student of
> >> our group, who will continue to work in this direction):
> >>
> >> http://www.nada.kth.se/utbildning/grukth/exjobb/rapportlistor/2008/rapporter08/jansson_niclas_08051.pdf
> >>
> >> The code is currently in a branch of DOLFIN from earlier this year, so
> >> the
> >> next step is to incorporate this into FEniCS/DOLFIN in a good way.
> >>
> >> /Johan
> >
> > Nice, but it looks like this only works for P1 elements?
> 
> There is some functionality that currently is P1 specific, but most should
> be rather straight forward to extend to general elements.
> 
> I think it is a good idea to merge the parallel functionality even if it
> is not general in terms of all element types, and then generalize the
> functionality from within DOLFIN. Alternatively we could add just the
> general stuff first, and later the rest when generalized, but that may
> take time, and I think the parallel capabilities will boost DOLFIN.
> 
> /Johan

I think we should start with discussing how we should (re-)design the
current DofMap and DofMapSet classes. These will be central for the
parallel assembly.

-- 
Anders

Attachment: signature.asc
Description: Digital signature