dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Nov 11, 2005 at 05:59:33PM +0100, Garth N. Wells wrote:

> Quoting Anders Logg <logg@xxxxxxxxx>:
> 
> > Anyone knows why applyBC_3D() is commented out in applyBC(Matrix, ...)
> > and applyBC(Vector, ...)?
> 
> My fault. Forgot to remove comments after testing. Fixed now.

ok.

> There is a lot of stuff for applyBC now in the class FEM (7 different functions,
> 6 of which are 95% the same). Could we simplify this somehow? The only
> difference between 2D and 3D versions is an iterator over edges (2D) and faces (3D).

Yes, this should be simplified. The problem is the current mesh
interface which is not dimension-independent. When we move to Sieve,
we should be able to remove all the duplication and just have one
assembly over the boundary.

The first release of Sieve should be available soon according to Matt
and Dmitry, so we should keep an eye on

    http://www.fenics.org/sieve/

and throw ourselves at the code whenever it appears.

I suspect we will need to go through the following stages:

    0. Getting the code
    1. Getting it to compile (could be non-trivial)
    2. Evaluation and benchmarking against the current mesh
    3. Creating wrappers New* including NewMesh
    4. NewMesh --> Mesh when we're satisfied

/Anders