Random Number Generatoration in High Performance Fortran



Abstract:

The generation of large scale series of ``Pseudo-random'' numbers is essential to the use of Monte Carlo simulations. All good Pseudo-random number generators (PRNG's) have a very large period, for which they satisfy discreet mathematical principles of uniform distribution. The size of the period is important for us to consider, since large amounts of random numbers, on the order of $10^{18}$, could be used in these simulations. We implemented a parallel PRNG for use in High Performance Fortran.


A scatter plot of random tuples produced by the parallel random number generator. The different colors denote random numbers produced on different processors.



1994 Undergraduate Research performed by Christopher A. Stoner.

Hypertext and Postscript versions of the final paper are also available.



stoner@cs.buffalo.edu