ufl team mailing list archive

[Kristian Oelgaard <k.b.oelgaard@xxxxxxxxxx>]

Quoting Anders Logg <logg@xxxxxxxxx>:

> On Wed, Aug 20, 2008 at 10:44:46AM +0200, Kristian Oelgaard wrote:
> > Quoting Martin Sandve Alnæs <martinal@xxxxxxxxx>:
> > 
> > > 2008/8/20 Kristian Oelgaard <k.b.oelgaard@xxxxxxxxxx>:
> > > >
> > > > Hi,
> > > >
> > > > Just a quick question:
> > > >
> > > > Do you plan on including quadrature code generation in UFL?
> > > >
> > > > Kristian
> > > 
> > > More or less all my uses for UFL will be with quadrature.
> > > Although UFL doesn't care about how you integrate,
> > > this should be possible:
> > > 
> > > e = FiniteElement("Quadrature", triangle)
> > > f = Function(e)
> > > 
> > > I'm also considering
> > > 
> > > e = FiniteElement("Boundary Quadrature", triangle)
> > > f = Function(e)
> > > 
> > > for functions that need to be evaluated on the boundary.
> > > 
> > > Maybe quadrature order could be specified in FiniteElement?
> > > 
> > > e = FiniteElement("Boundary Quadrature", triangle, 4)
> > > 
> > > The default can be up to the form compiler, and the
> > > form compiler is free to override anyway.
> > > But we shouldn't allow this of course:
> > > 
> > > e1 = FiniteElement("Quadrature", triangle, 2)
> > > e2 = FiniteElement("Quadrature", triangle, 4)
> > 
> > This is close to the syntax we have in FFC now:
> > 
> > e1 = QuadratureElement(triangle, 2)
> > e2 = QuadratureElement(triangle, 4)
> > 
> > so it looks fine by me.
> > 
> > But this was not what I meant. I'm talking about the
> > output that UFL will generate in the tabulate_tensor() function. In FFC we
> have
> > tensor representation and quadrature representation. Will we also have both
> in
> > UFL?
> > 
> > Kristian
> 
> UFL will not care about this. FFC will take an UFL object as input and
> then from that generate either quadrature or tensor code. We will need
> to add a predicate is_multilinear to UFL that checks whether or not
> the form is multilinear (in which case tensor representation is
> possible).

OK, I forgot for a second that UFL is only concerned with the interface to the
form compilers.

Kristian
 
> -- 
> Anders
>