Re: Making some minor contributions to GSL

[linas at austin dot ibm dot com]

On Thu, Jan 15, 2004 at 05:57:31PM -0700, Gerard Jungman wrote:
> On Tue, 2004-01-13 at 17:57, linas@austin.ibm.com wrote:
> 
> > The only "deficiency" is that the GSL implementation generates
> > a "domain error" when gsl_sf_psi_n (1,x) is called with 
> > negative x.  The psi's should be perfectly well defined 
> > for negative, non-integer x's.  
> 
> It might take me a while to get around to that. I don't think
> I ever found a good way, or at least not one that I understood.

?
I'm not sure what you mean.  For small negative values, 
abram & steg 6.4.6 can be applied a few times and for larger
values, 6.4.7.  I'm not sure, I guess you must be worried about
precision/rounding or performance or something like that?
Or maybe large values of n, for which this isn't practical?

>     gsl_sf_zetam1_e(double s, gsl_sf_result * r);

excellent!

--linas

p.s. I don't know if its of interest, but just in case:
the attachment contains some code to compute the determinant
of a matrix.  It wasn't written to fit into gsl, but could
probably be modified to do so easily enough.

Attachment: d.c
Description: determinant.c