← Back to team overview

dolfin team mailing list archive

Re: FIAT timing results

 

Nice plots. The current bottleneck in the system is FFC which is
terribly slow right now. (It takes a week to compile Poisson with
degree 7 Lagrange elements.)

The plan is that I will update FFC for the new faster interface of
FIAT (tabulate mode) as soon as I find the time. This will speedup FFC
by a factor 100 in some cases.

/Anders

On Thu, Apr 14, 2005 at 09:18:20AM -0500, Robert C.Kirby wrote:
> The newest version of FIAT is much faster than the old.  FIAT-L (on the 
> Web site) implements this new architecture, but the full version of 
> FIAT that supports other elements (such as Brezzi-Douglas-Marini) will 
> be released soon.  The timings below are for instantiating (this 
> involves at least inverting a large vandermonde-type matrix and in the 
> case of BDM elements, a possibly large SVD) elements and for tabulating 
> them at a set of quadrature points (the number of points is appropriate 
> for the degree of the polynomials.
> 
> Next on the to-do list is generalizing the generation of dof maps in 
> FFC so as to leverage these higher order Lagrange elements.
> 
> Rob



> 
> 
> 
> Robert C. Kirby
> Assistant Professor
> Department of Computer Science
> The University of Chicago
> http://people.cs.uchicago.edu/~kirby

> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev




References