Re: gsl_sf_hyperg_U_int_e10_e function

[Brian Gough <bjg at network-theory dot co dot uk>]

Loredo, Elvira writes:
 > Below is a C program that uses the gsl_sf_hyperg_U_int_e10_e
 > function.  The program takes the user supplied parameter values,
 > (n,k,z and r) and generates the correct solution over several
 > ranges of these parameters when compared to the results generated
 > by Mathematica.  However, the C call begins to break down when the
 > combination of parameters exceeds a certain threshold, while
 > Mathematica continues to work.
 >  I am running Redhat Linux 7.2 on a Dell Optiplex GX-150 desktop
 > with an i686 processor.
 >  Any ideas as to why this is happening and any suggested solutions
 > would be appreciated.

By default there is a trap on all internal errors for safety,
including some which do not affect the final result (e.g. overflow in
the denominator of an expression).

If you add a call to

  gsl_set_error_handler_off();

at the start of the program you should get the correct results.

regards
Brian Gough