A General-Purpose Cam-Based System

Abstract

VLSI makes feasible the massive use of content addressable memory in a general purpose computer. We present a design for memory which is addressable conventionally and by content, and which supports low-level bit-serial word-parallel algorithms. CAM provides one of the most easily understood and programmed frameworks for massively parallel computations. We present a programming methodology for the use of our design. This includes a programming language, CAML; a number of algorithms from various fields which demonstrate the generality of the design and the language; and techniques for transforming algorithms from conventional to CAM-based structures and methods.

We do not attempt to better the performance of specialized hardware for particular tasks. Our contention is that CAM is a practical technique for the application to general-purpose use of the massive parallelism available with VLSI.