A case study of parallel execution of a rule-based expert system

Abstract

We report on a case study of the potentials for parallel execution of the inference engine of EMYCIN, a rule-based expert system. Multilisp, which supports parallel execution of tasks by means of the future construct, is used to implement the parallel version of the backwards-chaining inference engine. The study uses explicit specification of parallel execution and synchronization to attain parallel execution. It suggests some general techniques for obtaining parallel execution in expert systems and other applications.

Overview of concert multilisp: a multiprocessor symbolic computing system

“…A major focus of our effort is in applying this strategy to general-purpose symbolic computing. Several application programs have been and are being written and studied, including sorting [12], simulation of logic circuits [6], event-based simulation [23], part of a speech recognition system [22], semantic net retrieval [4], a subset of Prolog not including "cut" [27], a kernel of EMYCIN [21] (developed at MCC), the Boyer benchmark from the Gabriel benchmark set [9], polynomial manipulation, traveling salesman, parallel parsing of Lisp expressions, and the Multiiisp compiler itself.…”