Re: [Bug-gsl] [bug #43256] gsl_sf_coupling_6j overflows

[Patrick Alken]

Thanks for the report. If you are able to fix the issue and send a patch 
that would be great, otherwise it could take a while for one of us to 
look into this.

On 09/19/2014 06:45 AM, Anders Søndergaard wrote:
> URL:
>    <http://savannah.gnu.org/bugs/?43256>
>
>                   Summary: gsl_sf_coupling_6j overflows
>                   Project: GNU Scientific Library
>              Submitted by: andersas
>              Submitted on: Fri 19 Sep 2014 12:45:52 PM GMT
>                  Category: Runtime error
>                  Severity: 3 - Normal
>          Operating System: Ubuntu linux
>                    Status: None
>               Assigned to: None
>               Open/Closed: Open
>                   Release: 1.16 (ubuntu latest)
>           Discussion Lock: Any
>
>      _______________________________________________________
>
> Details:
>
> I need to calculate 6-j symbols for somewhat large J.
> For example the symbol:
>
> (14, 16, 16
>   78, 62, 76) = 6/1739*sqrt(16971005238954/1556382731177197) = 0.000360286
>
> but this fails because gsl_sf_fact overflows.
> It is used internally in gsl_sf_coupling_6j.
>
>
> I have looked a bit at the code, and I think in these large
> cases it is possible to work with the logarithm of the factorials
> and in the end exponentiate.
>
> Other libraries I have tested also gets these symbols wrong (but silently
> returns the wrong value),
> except pythons sympy, which calculates the symbols as the square root of a
> rational number.
>
> I would rather not interface my Fortran program with Python.
>
>
>
>
>
>
>      _______________________________________________________
>
> Reply to this item at:
>
>    <http://savannah.gnu.org/bugs/?43256>
>
> _______________________________________________
>    Message sent via/by Savannah
>    http://savannah.gnu.org/