Table-Lookup Methods for Generating Arbitrary Random Numbers

Extremely fast table-lookup procedures are described for generating random numbers from an arbitrary distribution on a digital computer. The procedures sacrifice some accuracy in order to achieve this speed, but the accuracy is sufficient for most types of Monte-Carlo simulations. A method is given to eliminate the error associated with truncation for distribution functions with an infinite domain. The performance of the table-lookup methods is examined for a few common distribution functions.