dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

Yes, it's correct that the default quadrature rule in DOLFIN is not enough
to compute the mass matrix exactly. Normally, this is not a
problem. For example, if you lump the mass matrix (replace it with a
diagonal matrix where the diagonal entry is the row sum) you will get
the correct result even with a low order quadrature rule.

We're working on replacing the current form evaluation system in
DOLFIN entirely. The new system will be based on FFC, see the page
http://www.fenics.org/ffc/, which is a compiler for variational forms.
FFC uses FIAT (Robert Kirby, Matthew Knepley) to compute integrals and
with FIAT we get all the quadrature rules we need to compute the mass
matrix (and other forms) correctly, even for high-order elements.

There are no plans to add new quadrature rules to DOLFIN (since they
will not be needed once we have replaced the current system with FFC),
but you can always add a new quadrature rule to
src/kernel/quadrature/dolfin/ if you need it.

/Anders

On Sun, Dec 05, 2004 at 02:36:39AM +0100, alex wrote:
> Hello,
>  I am doing my first steps using dolfin and
> I have a question concerning example src/demo/fem :
> 
> the assembled MassMatrix differs from the reference one.
> I have traced this difference to the middlepoint quadrature
> for tetrahedron, that is inaccurate to integrate properly
> quadratic functions.
> 
> I am new also to fem, so I am wondering what should I do :
>  should I live with that or try to implement another quadrature
> for the tetrahedron?
> 
> Thank you in advance for the time you will be able to spend
> to answer such newby's questions.
> Regards
>              Alex
> 
> 
> 
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev
>