Converter portable random number generators

In the mid 1960's the publication of the ANSI FORTRAN standard attempted to standardize software. The effort was not entirely successful. Renewed interest in the project can be evidenced by the appearance of transportability studies. Robust codes are growing in importance, in computing. Such codes have good behavioral characteristics in normal situations and react gracefully under abnormal situations. Studies of robustness and standardization have avoided the topic of random number generation because of its highly machine dependent nature. This thesis examines the topic of portability as it is related to random number generation, by examining the possibility of producing identical deviates on a set of computers which vary in architecture. Included is a set of robust "converter portable" random number generators which produce identical deviates on IBM, UNIVAC and Control Data Corporation (CDC) computers. The algorithms employed are representative of commonly used generation techniques. It is felt that the principles established apply equally well to most other generation algorithms.