Re: circular random number distribution

[Jochen Küpper <jochen at fhi-berlin dot mpg dot de>]

Ok, I did some quick tests, the times are for the creation of 1e5 xy
pairs plus some constant cruft done in all runs -- runtime is always
3.4s (+-0.5) for all approaches.... (This is not a good benchmark(!),
it's only meant to give me an idea of what is going on.)

I tried:
  rejection, 
  trigonomtric without sqrt, 
  trigonomtrix + sqrt,
  gsl_ran_dir_2d + sqrt, 
  gsl_ran_dir_2d_trig_method + sqrt.

"Robert G. Brown" <rgb@phy.duke.edu> writes:

>> | phi = gsl_ran_flat(rng, 0, 2.0*PI);
>> | r = gsl_ran_flat(rng, 0, 1);
>> | x = sqrt(r)*cos(phi);
>> | y = sqrt(r)*sin(phi);

> I'm probably late with this (sorry, busy weekend) but: Your
> accept-reject method is almost certainly much faster than using
> ANYTHING with transcendental calls in it.

Well, I am told that some machines (i.e. PCs) have these functions in
hardware.

> In fact, your sqrt call is redundant

Well, that's the deal: It isn't!

Without the sqrt, the functions are peaked at the center. With the
sqrt they are uniform.

I guess I am going with the reject anyway... Thanks for everybodies
suggestions and help.

Greetings,
Jochen
-- 
Einigkeit und Recht und Freiheit                http://www.Jochen-Kuepper.de
    Liberté, Égalité, Fraternité                GnuPG key: CC1B0B4D
        (Part 3 you find in my messages before fall 2003.)

Attachment: pgp00000.pgp
Description: PGP signature