An Overlapped Prolog Processor

Abstract

This report describes the design of a Prolog machine organization implementing D. Warren’s architecture. The objective was to determine the maximum performance attainable by a sequential Prolog machine for "reasonable" cost. The report compares the organization to both general-purpose, microcoded machines and reduced-instruction-set machines. Hand timings indicate that a peak performance rate of 450 K LIPS (logical inferences per second) is well within current technology limitations and 1 M LIPS is potentially feasible.