Parallelizing the OPS5 Matching Algorithm in Qlisp,
Abstract
In recent years, production systems have become a popular framework within which to implement large scale expert systems. Unfortunately, production systems are often characterized by slow running times, because of the large amount of matching that must be done during their execution. For the production system language OPS5, there is a highly efficient matching algorithm known as the Rete algorithm which gives a large speedup over a naive implementation of production systems. In this paper, we describe our attempts to speed up OPS5 even further by parallelizing the Rete algorithm in Qlisp, a parallel Lisp language. We give details on the Qlisp constructs we used to parallelize the Rete algorithm and provide actual timing results on various OPS5 rule sets.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1991
- Accession Number
- ADA323588
Entities
People
- Daniel J. Scales
Organizations
- Stanford University