Knuth [175] provides a wealth of material on random number
generation, including a table of appropriate values for * a*
and
* m*
and tests that can be used to determine the quality of a
particular generator. See in particular the table on page 102.
Anderson [13] provides a more up-to-date survey of random
number generation algorithms. He includes a short but useful section
on parallel
computers and provides numerous references. The random tree method
was first described by Frederickson et al. [114].

Random numbers are used extensively in the Monte Carlo method, in which a large, statistically valid sequence of ``samples'' is used to compute properties of mathematical functions or physical processes. Koonin [177] provides a good introduction to the computational issues associated with Monte Carlo methods on sequential computers, and Kalos [163] provides a more detailed discussion.

Here is a
Web Tour
providing access to additional information on random number generation
on parallel computers.