Re: binomial_tpe

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

James Theiler writes:
 > there are a few other gsl routines (multinomial, poisson, and
 > negative_binomial [via poisson]) that call gsl_ran_binomial,
 > so if we did swap in the faster algorithm, then those would also
 > be speeded up.

Ok, I've changed it over for the next release.

 > by the way, there is a gsl_ran_poisson_array function (which 
 > just calls gsl_ran_poisson multiple times).  for consistency's
 > sake, we might want to either eliminate that one (i don't think 
 > it's documented), or think about providing *_array functions for
 > some of the other random number distributions.  particularly for
 > the discrete distributions, there are often more efficient ways
 > to compute arrays than through multiple calls of the single 
 > variate generator.

I'll leave it in to avoid any problems with shared libraries.

Unless there are huge savings I'd be inclined to let people roll their
own array functions as needed.  

-- 
Brian Gough