I just had occaison to test random number generators in my work.
I also read a review paper from the physics literature on the subject. I
did extensive tests (on my problem) on all four of the best Numerical
Recipes generators and the linux system 'random', which is the longer
period generator. Even when given the largest table size, the linux
generator performed perceptibly worse than the others. When given a
somewhat smaller table size, it performed abysmally. I believe the N.R.
people did a good job in digging up good algorithms.
What are the considerations for choosing a system random number
generator? What if debian were the first unix to have a good random
number generator ? (is there another?) At least I should upload GPL'd
versions of the good generators (there is not much code) for people that
need them .
G John Lapeyre <email@example.com>
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
Trouble? e-mail to firstname.lastname@example.org .