Design and implementation of a parallel inference machine for first order logic: an overview

Abstract

Several approaches to parallel architectures and languages for advanced information processing are being investigated within ESPRIT Project 415. This overview presents one of these approaches, carried out in one subproject of ESPRIT 415 ("Subproject F"), which aims at the design and implementation of a parallel inference machine.

• -This machine, called LCM, for Logical Connection Machine, will be a highly parallel system for problem solving, mostly directed toward symbolic processing and artificial intelligence applications. It is based on the "connection method", which is a deductive inference technique for first order logic with significant possibilities for introducing parallelism.

• -The design of such a complex system requires the use of formal means. The abstract specification of the LCM architecture reflects the parallel algorithmic description of the connection method and is expressed with FP2, a "Functional and Parallel Programming" language based on term rewriting techniques. Formal treatment and systematic transformations can be applied to FP2 programs.

• -The actual implementation of the LCM requires a highly parallel multi-processor machine. Once such a machine is chosen, systematic transformations of FP2 programs can be applied for mapping the abstract process structure of the LCM specification onto the concrete processor structure of the machine.

This paper is an informal introduction to the bases of this work. It presents the main ideas and design decisions of the project through a series of simple examples.

This research is partially funded by ESPRIT Project 415 : Parallel Architectures and Languages for Advanced Information Processing — A VLSI-Directed Approach.