Random number generation and simulation on vector and parallel computers

185. R. P. Brent, Random number generation and simulation on vector and parallel computers (invited paper), Proc. Fourth International Euro-Par Conference (Southampton, UK, 1-4 Sept 1998), D. Pritchard and J. Reeve (editors), Lecture Notes in Computer Science, Vol. 1470, Springer-Verlag, Berlin, 1998, 1-20.

Preprint: dvi (33K), pdf (261K), ps (89K).

Paper: available from SpringerLink.

Abstract

Pseudo-random numbers are often required for simulations performed on parallel computers. The requirements for parallel random number generators are more stringent than those for sequential random number generators. As well as passing the usual sequential tests on each processor, a parallel random number generator must give different, independent sequences on each processor.

We consider the requirements for a good parallel random number generator, discuss generators for the uniform and normal distributions, and describe a new class of generators for the normal distribution, based on a proposal by Chris Wallace [ ACM Transactions on Mathematical Software 22 (1996), 119-127]. These generators can give very fast vector or parallel implementations. Implementations of uniform and normal generators on vector and vector/parallel computers are discussed.

Comments

Related papers are [132, 133, 141, 170, 211, 213, 217, 224].

Fortran 77 implementations of some uniform and normal random number generators are available here.

Go to next publication

Return to Richard Brent's index page