GENERATOR FOR PSEUDO-RANDOM NUMBERS

Pseudo-random numbers (PRNs) have great importance in data processing and encryption. The standard technique for generating PRNs on computers at the present time involves software implementation of recursively computing PRN.sub.n+1 =(PRN.sub.n)(b) mod M, in real number field which means that a relatively slow and involved multiplication must occur for each PRN, and each PRN calculation is strictly subsequent to previous PRM calculations. This disclosure shows a systolic multiplier implemented by a series of exclusive OR operations which generates simultaneously a new set of the PRN's from the current set of PRN's in parallel. The operation of multiplication mod M is performed over a finite field instead of a real number field. To facilitate the multiplication, the input operands are first transformed into a different representation that consists of more bits than that of the original ones. The operation of multiplication is then carried out in this transformed representation. The number of bits of the intermediate product is then reduced to that specified for each PRN through an inverse transform network, thereby forming a new PRN. The cheapness, speed, and simplicity of the required logic allows at least one PRN to be generated per computer cycle. If desired, several logic networks can be paralleled for simultaneous generation of multiple PRNs. This technique allows PRN.sub.n+k, . . . to be generated directly from PRN.sub.n, . . . where k is some proper integer.