A parallel algorithm for execution of production systems on HMESH architecture

In this paper, the parallelism of production systems is studied and divided into 6 levels — attribute level, precondition level, production (rule) level, working memory element level, parallel action level, and multiple rule firing. An algorithm for parallel execution of production systems to run on a HMESH structure[Ragh 1986] is proposed. It uses large number of PE's (4,096 to 65,536) to accommodate large production systems with one thousand to 16 thousand rules. The hierarchy of buses in HMESH structure provides flexible partitioning of PE's into different sizes such that different partitions can perform their processing simultaneously. This matches the need in executing the rule-based production systems by partitioning the rules into disjoint and independent sets. The algorithm, using the HMESH structure, can make the communication among PE's simple and regular to avoid overheads such as bus contention, access conflicts, etc. It exploits various levels of parallelism in production systems.