An associative parallel processing system for non-numerical computation

Abstract

After reviewing the problems and techniques of non-numerical computation using current technology, a systems design is postulated for a symbol processing machine (SPM) employing associative storage principles. The system design commences with an examination of the basic attributes of a SPM, in particular the storage requirements, such as data structures and their manipulation, and the need for internal and external communication and control. A storage system is then proposed based on a linear array (or string) of storage cells each holding one symbol plus control logic, which may be associatively addressed in parallel. Using this concept a set of cell commands has been developed and the overall storage system shown (using simulation techniques) to have considerable advantages when performing symbolic processing algorithms. The paper concludes with comments on the feasibility of building a hardware model of the SPM, and a comparison of the system with other highly parallel computer structures.