Date of Award

12-1990

Document Type

Thesis

Degree Name

Master of Science in Computer Engineering

Department

Department of Electrical and Computer Engineering

First Advisor

Gary B. Lamont, PhD

Abstract

Productions systems provide a flexible and powerful means for expressing and solving problems using high-level reasoning approaches. Unfortunately applications based on the production system paradigm are very compute intensive and therefore execute very slowly on conventional computer architectures. Lack of execution speed therefore limits the use of production systems in many time sensitive and complex tasks. Researchers have demonstrated that increasing the execution speed of production systems through the use of small scale parallel computer architecture is possible, however demonstrated execution speed is still several orders of magnitude below the desired level of performance. Past research indicates that there exists an irrevocable tie between the specific production system application and the level of parallelism inherent in that application. To date, this parallelism has been quantified only in very lose terms based on specific problem decomposition and implementation on a specific parallel architecture. This thesis examines many of the possible approaches to extracting parallelism from productions systems and quantifying the level of parallelism in specific applications. This research also investigates previously unexplored aspects of several important decomposition approaches. The software products developed to support this research are Software Tool for Evaluating Parallelism in Production Systems (STEPPS) and a modified version of the C-Language Integrated Production System (CLIPS) called Multi-CLIPS.

AFIT Designator

AFIT-GCE-ENG-90D-04

DTIC Accession Number

ADA230498

Comments

The author's Vita page is omitted.

Share

COinS