dolfin team mailing list archive

[Johan Hoffman <hoffman@xxxxxxxxxxxxxxx>]

This is great news! I haven't had the opportunity to see the details yet, 
but it sounds really good from what you describe. As for added features, I 
guess a natural extension would be to vector elements and corresponding 
operators.

/Johan

On Fri, 22 Oct 2004, Anders Logg wrote:

> I've just added the new project FFC (FEniCS Form Compiler) to the
> FEniCS web site:
>
>       http://www.fenics.org/
>
> FFC is implemented in Python and uses FIAT to precompute integrals on
> the reference element. FFC supports all (scalar) elements that FIAT
> supports.
>
> Although a number of important features are missing in this
> first version, FFC can already generate output for DOLFIN. The initial
> results look promising. Below are results from a simple comparison in
> DOLFIN for assembly of Poisson's equation with piecewise linears on
> tetrahedrons in 3D (assembling 10 times over 25000 tetrahedrons):
>
> Old assembly:   7.380000
> New, optimized: 0.670000
> New, FFC:       0.680000
>
> The new assembly through FFC is more than a factor 10 faster than our
> old run-time system in DOLFIN. Note also that the code generated by
> FFC is almost as fast as the hand-optimized code (w.r.t. symmetries).
> There might be other bottlenecks in the assembly that may affect these
> numbers but it suggests that the C++ compiler (gcc) does a pretty good
> job.
>
> I would appreciate any comments, suggestions or requests for features.
>
> /Anders
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev