Andy, this also could be handled at the FIAT level by providing something thatmeets the "FiniteElement" interface but whose basis happens to consist of macro-elements. FFC (almost) doesn't need to know that the FIAT basis functions are macro-elements in order to build the reference tensor.
By almost, I mean that it would need to get special quadrature rules that are accurate on each of the subdomains. But again, FFC doesn't need to know the correct kind of element, just some kind of flag to force it to use a certain quadrature. It might be worth putting them into FIAT if the would be used. However, I think we need more general element transformations in order to use most of them. (e.g. they have derivative degrees of freedom) Rob On Mon, Nov 3, 2008 at 6:37 AM, Andy Ray Terrel <[EMAIL PROTECTED]>wrote: > On Mon, Nov 3, 2008 at 12:40 PM, Anders Logg <[EMAIL PROTECTED]> wrote: > > On Mon, Nov 03, 2008 at 12:25:12PM +0100, Andy Ray Terrel wrote: > >> On Mon, Nov 3, 2008 at 12:03 PM, Anders Logg <[EMAIL PROTECTED]> wrote: > >> > On Mon, Nov 03, 2008 at 10:38:16AM +0100, Andy Ray Terrel wrote: > >> >> Hello, > >> >> > >> >> I was wondering if anyone has done anything with coupling fields over > >> >> subelements? I have a mixed method that uses a P2 velocity space and > >> >> then couples it to a several P1 subelements for the stress tensor. > >> >> > >> >> Thanks, > >> >> Andy > >> > > >> > What do you mean with coupling? We use mixed elements with coupled sub > >> > fields (through the variational form) all the time (like in Stokes). > >> > > >> > >> > >> I mean the same thing as in the Stokes example, except velocity is on > >> a reference triangle and the stress is on n triangles partitioning the > >> reference. I would handle all the mixed element formulation the same > >> and couple fields in the variational form. > > > > Form on overlapping meshes is not supported, you'll have to do it > > manually (by run-time quadrature). We are working on similar > > formulations (forms on overlapping non-matching meshes) so I might be > > interested in adding support for this at some point. > > > > -- > > Anders > > > > Sounds good. I just thought I would poke the list before I tried > getting it to work on my own. > > -- Andy > _______________________________________________ > DOLFIN-dev mailing list > [email protected] > http://www.fenics.org/mailman/listinfo/dolfin-dev >
_______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
