Random Number Geuerators: Good Ones Are Hard to Fin

An important utility that digital computer systems should provide is the ability to generate random numbers. Certainly this is true in scientific computing where many years of experience has demonstrated the importance of access to a good random number generator. And in a wider sense, largely due to the encyclopedic efforts of Donald Knuth [18], there is now a realization that random number generation is a concept of fundamental importance in many different areas of computer science. Despite that, the widespread adoption of good, portable, industry standard software for random number generation has proven to be an elusive goal. Many generators have been written, most of them. have demonstrably non-random characteristics, and some are embarrassingly bad. In fact, the current state of random number generation software is accurately described by Knuth [18, p. 1761 who advises “. . . look at the subroutine library of each computer installation in your organization, and replace the random number generators by good ones. Try to avoid being too shocked at what you find.” Knuth’s advice applies equally well to most recently published computer science textbooks, particularly those written for the undergraduate market. Indeed, during the preparation of this article we reviewed more than 50 computer science textbooks that contained software for at least one random number generator. Most of these generators are unsatisfactory. This article was motivated by practical software con..