Abstract

The use of optics in computing has taken many forms over the last 25 years, it ranges from optical data storage and fiber-optic local area networks to all-optical digital computers and optical realizations of neural net models. The first two examples represent a mature, widely accepted use of optics in computing while the latter examples represent speculative and highly controversial use of optics in computing. In between the two extremes are systems with various levels of development and acceptability, such as matched filter correlators, acoustooptic processors, and optoelectronic interconnects for electronic computing systems. I seek to categorize various computing systems—optical, electronic, or hybrid—according to the operations that they perform (e.g., number crunching, symbolic inferences, signal preprocessing, etc.) and the class of computing systems (e.g., analog, linear, digital, or quasilinear) that they belong to. Within each category, a hierarchical organization of algorithms/architectures, higher-order operations, elementary operations, and computational primitives can be introduced. This framework will then be useful in organizing the various aspects of the use of optics in computing. Such a taxonomy will presumably help in analyzing the comparative advantages of optics over electronics and of one scheme for optical computing over another. It is also hoped that such a framework will serve as a guide for future directions of research.

© 1986 Optical Society of America