>Split up the integral (or integrand) to compute the part near x=0 >using the asymptotic form of j(x,n) for small x to avoid underflow. >Or disable the underflow error if it doesn't affect the final results. > >-- >Brian Gough
Hi. Thanks for your reply. I implemented a solution equivalent to your second suggestion: it turns out that I can prevent gnu_sf_bessel_Jnu(x,n) from collapsing with an underflow error if I ensure that (n-x)<~70. (n, x > 0.) I did this by setting the integrand equal to zero if this condition is not satisfied. Then I checked my integral for a few relevant values of n and x. (I used matlab for cross-checking.) The integral did not change. Thanks again. Jorge --- Jorge Talamantes, Physics and Geology, California State University, Bakersfield, CA 93311, jtalamantes at csub dot edu, 661.654.2335 _______________________________________________ Help-gsl mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-gsl
