Re: [Help-gsl] Recourse, overflowing lngamma

[Martin Jansche]

On 8/21/06, Marc <address@hidden> wrote:

> I've been using GSL for an intensive numerical calculation involving large
> numbers. I rely on gsl_sf_lngamma, however it overflows at points of my
> program where the argument is large.

How large?  In principle, you should be able to evaluate lngamma for
arguments up to around 10^305 without overflow.  In double precision,
there aren't many argument values for which lngamma is expected to
overflow (roughly 1.28e305 to 1.79e308).  If you get an unexpected
overflow for smaller (< 1.28e305)  argument values, you may want to
post the details here, since this would likely indicate a bug in
gsl_sf_lngamma.

-- mj