Re: [Help-gsl] bessel performance

[Leo Razoumov]

On 4/7/09, Gideon Simpson <address@hidden> wrote:
>
>  On Apr 6, 2009, at 4:55 PM, Brian Gough wrote:
>
>
> > At Sat, 4 Apr 2009 20:09:36 -0400,
> > Gideon Simpson wrote:
> >
> > > If I have an array of points at which I want to evaluate a bessel
> > > function, I should expect better performance by using
> > > gsl_sf_bessel_sequence_Jnu_e, as opposed to gsl_sf_bessel_Jnu, right?
> > >
> >
> > Yes, that's the intent.
> >
>
>  Would it surprise you if I said that I found the SciPy python wrapped
> interfaces to the fortran specfun code to be faster than GSL for this?
>
>  -gideon

Faster by how much? Fortran libraries are known to be faster than GSL.
I would not loose my sleep over 10-20%. If, on the other hand, Fortran
Bessel function is several times faster then it is a different matter.

--Leo--