Re: CCL:random number generators

 My best advice is to beware of using random number
 generators as black-boxes, especially in the context
 of Monte Carlo integrations and simulations.
 When using random number generators in Monte Carlo, you
 have to not only worry about whether the samples are
 uniform on the plane and uncorrelated in hyperdimensions
 but also about the period of the random number sequence
 (all random number generators, as you know, have a period
 for a given seed -
 and if your program requires a # of random numbers which
 is on the order of the period of the algorithm, better
 watch out).  What's more, RNGs which perform perfectly
 well against a battery of statistical tests have been
 known to fail when used, for example, in 3D Ising lattice
 calculations...there is a good bit of literature on this
 subject.
 I personally wouldn't use any random number generator in
 my Monte Carlo work unless I knew exactly what the algorithm
 was, had a reference for it, and had subjected it to my personal
 battery of tests, which includes statistical tests ala Knuth
 and application tests (path-integral and classical Monte Carlo
 computations). At least then, if there is a bug in the
 algorithm I know who to blame for all those wasted cycles...
 Finally, in every Monte Carlo calculation I have ever carried out,
 the limiting factor has not been the speed of the random number
 generator, but the evaluation of the potential energy. But then
 I've never done QMC calculations, so I don't know what the
 "rate-determining step" is for such calcs. Because of this,
 I tend to err on the side of slow, but solid random number generation.
 hope this helps, best to all-
 rqt
 ************************************************************************
  Prof. Robert Q. Topper               internet:   topper at.at cooper.edu
  Department of Chemistry              phone:      (212) 353-4378
  The Cooper Union                     FAX:        (212) 353-4341
  51 Astor Place                       subway:     take the 6 to Astor Place
  New York, NY 10003 USA                           or the N/R to 8th St/NYU
 ************************************************************************
 The Cooper Union for the Advancement of Science and Art, established by
 Peter Cooper in 1859, is a private institution of higher learning where
 all students receive full-tuition scholarships.