fiat team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, May 16, 2005 at 03:48:26PM -0500, Robert C.Kirby wrote:
> Anders,  pull from repository.
> 
> On May 13, 2005, at 1:32 PM, Anders Logg wrote:
> 
> >Below is a list of what I need from FIAT to release a new version of
> >FFC and then later a new version of DOLFIN.
> >
> >(I'm thinking we should release FFC 0.1.8 and DOLFIN 0.5.6 pretty soon.
> >When we have general order Lagrange working (and possibly also
> >manuals), it might be time to release FFC 0.2.0 and DOLFIN 0.6.0.)
> >
> >1. Access to the dual basis so I can generate run-time code for the
> >dual basis in DOLFIN. (If it's already there, please give me a pointer
> >where to look.) Given a user-defined function f, FFC has to generate
> >code that can compute n_i(f) for each node n_i. For Lagrange elements,
> >all I need is access to the nodal points x_i so I can compute n_i(f) =
> >f(x_i).
> >
> 
> Done in the following sense:
> 
> U = Lagrange.Lagrange( TRIANGLE , 1 )
> 
> you get the point location by:
> U.dual_basis().pts
> 
> this gives you a tuple of points, each point is a tuple.
> 
> This is a hack for pointwise evaluation.  We'll think of something else 
> some other time.

Looks good. I tried it and it seems to work. I'll start working on
getting the interpolation going (but probably not today).

> >2. Order the dofs for DiscontinuousVectorLagrange in the same style as
> >for VectorLagrange (with num_reps > 1) if you still think this is a
> >good idea. (I think it is.)
> >
> 
> Done.

Works:

dofs[0] = 3*cell.id() + 0
dofs[1] = 3*cell.id() + 1
dofs[2] = 3*cell.id() + 2
int offset = 3*mesh.noCells()
dofs[3] = offset + 3*cell.id() + 0
dofs[4] = offset + 3*cell.id() + 1
dofs[5] = offset + 3*cell.id() + 2

> >3. Confirmation (or suggestion for changes) for the numbering
> >conventions for triangles and tets that I have put in the FFC manual
> >(see docs section on fenics.org).
> >
> 
> Will look more closely soon.  Next on list.

ok.

/Anders