dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Feb 19, 2008 at 09:20:26AM +0100, Johan Hoffman wrote:
> > Hello all!
> >
> > I did some timing runs on the point-wise evaluation (at non-node points)
> > of a discrete function (vector valued P2 element); and I will not make
> > any wild claims to the underlying complexity here, but it seems that GTS
> > is doing a really good job at the searching! This makes my hopes of
> > computing some line integrals that cut across triangles come to life :)
> >
> >
> > Solve-step, 50x50 grid
> > Elapsed time is 1.715692 seconds.
> > Evaluating discrete function along curve (n points)
> > n = 100
> > Elapsed time is 0.006452 seconds.
> > n = 1000
> > Elapsed time is 0.011296 seconds.
> > n = 10000
> > Elapsed time is 0.060278 seconds.
> > n = 100000
> > Elapsed time is 0.555764 seconds.
> > n = 1000000
> > Elapsed time is 5.737265 seconds.
> >
> > Solve-step, 100x100 grid
> > Elapsed time is 9.487012 seconds.
> > Evaluating discrete function along curve (n points)
> > n = 100
> > Elapsed time is 0.023748 seconds.
> > n = 1000
> > Elapsed time is 0.028949 seconds.
> > n = 10000
> > Elapsed time is 0.080517 seconds.
> > n = 100000
> > Elapsed time is 0.618475 seconds.
> > n = 1000000
> > Elapsed time is 6.065121 seconds.
> >
> > Mesh is UnitSquare; code runs in Matlab through a MEX wrapper.
> >
> >    Dag
> 
> Very nice! Should be very useful.
> 
> /Johan

Better than expected.

-- 
Anders