Abstract

The design constraints and resultant performance boundaries applicable to the developemnt of optical information processing and computing systems derive directly from certain fundamental limitations imposed by physical laws, as well as from specific algorithmic (and associated architectural) choices implemented within a given device or component technology base.¹ In general, a given processing or computation function can be partitioned into the cost (in energy or otherwise) of representation, the cost of computation, and the cost of detection and utilization of the desired computational result. For operation at the quantum limits, for example, analog representations are favored for architectures and algorithms that implement a high degree of computational complexity (irreducible number of equivalent binary operations) per unit detected output resolution element, whereas binary representations favor operations with a somewhat lower degree of complexity. The relevant performance tradeoffs among the various possible computational paradigms, however, are often strongly influenced by the current and projected technological status of the requisite photonic components, including both coherent and dynamically reconfigurable volume holographic optical interconnection elements and both one- and two-dimensional array detectors. In this presentation, we examine a broad range of interrelated—and at times conflicting—device requirements from three complementary perspectives: the fundamental physical limitations that affect the performance of any photonic component function; the current status of component development with respect to such fundamental limits; and the technological considerations that impact present and future device design and development. The implications of these considerations on the evolution of practically implementable algorithms and architectures that exhibit significant computational advantages will be addressed.

© 1990 Optical Society of America