The authors write about pseudorandom number generators as though they are generating true random numbers. As the techniques get better and better, it is easy to talk about “random number generators,” but we must be careful to keep in mind what we are doing. Aside from this, the paper is well done. It discusses the Tausworthe system, the improvements from combining two or more processes, and the benefits of portability. Portability is becoming more and more important. We need a single program that will produce the same series on any word size computer.
From the evidence in this paper, the Tausworthe system does seem to be better than the more common linear congruential methods. This paper gives strong theoretical support for the authors’ conclusions. Summary information about extensive results shows the goodness of the technique. A C program of only 11 lines implements the program. It is fast, generating one million pseudorandom numbers in 154 seconds on a 16 MHz 386 with a math coprocessor.
Several assumptions made throughout the paper require the reader to go back and review some earlier papers by L’Ecuyer. Access to those papers is almost essential to fully understanding this paper. The authors provide 26 good references.