DMCA
Techniques for Testing the Quality of Parallel Pseudorandom Number Generators (1995)
| Venue: | in Proc. of the 7th SIAM Conf. on Parallel Processing for Scientific Computing, SIAM |
| Citations: | 13 - 5 self |
BibTeX
@INPROCEEDINGS{Cuccaro95techniquesfor,
author = {Steven Cuccaro and Michael Mascagni and Daniel V. Pryor},
title = {Techniques for Testing the Quality of Parallel Pseudorandom Number Generators},
booktitle = {in Proc. of the 7th SIAM Conf. on Parallel Processing for Scientific Computing, SIAM},
year = {1995},
pages = {279--284},
publisher = {SIAM}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Ensuring that pseudorandom number generators have good randomness properties is more complicated in a multiprocessor implementationthan in the uniprocessor case. We discuss simple extensions of uniprocessor testing for SIMD parallel streams, and develop in detail a repeatability test for the SPMD paradigm. Examples of the application of these tests to an additive lagged-Fibonacci generator are also given. 1 Introduction Random numbers are used in applications ranging from scientific simulations to statistical sampling procedures. The impracticality in most cases of using true random numbers led to the development of pseudorandom number generators (PRNGs). In the course of this development, testing procedures were designed to ensure that the necessarily deterministic sequence of numbers produced by these PRNGs had analytical and statistical properties which compared well with those of a true random stream. With the tests described by Knuth [1] and Marsaglia [2, 3], an intelligent user ...
Keyphrases
parallel pseudorandom number generator pseudorandom number generator additive lagged-fibonacci generator statistical sampling procedure good randomness property true random stream introduction random number uniprocessor case simd parallel stream intelligent user repeatability test multiprocessor implementationthan spmd paradigm scientific simulation true random number deterministic sequence statistical property simple extension
