Abstract

Many problems exist that available methods of computation, including those implemented by advanced parallel electronic computers, cannot solve within the bounds established by the immutable requirements of numerous significant applications. The bounds for some of these applications are generally described in terms of environmental stresses, input format, desired output, computational throughput, accuracy, size and power. In recent years, the potential capabilities of many optically implementable algorithms, architectures and technologies have been considerably enhanced. What must be done now is to construct composite system performance metrics in terms of individual algorithmic, architectural and technological capabilities that can predict whether or not a given optical system can satisfy the requirements imposed by these applications. (Clearly, a given application can be solved by more than one algorithm, each implemented on several architectures, using a wide variety of technology.) The method used to construct these composite metrics should reveal what individual performance gains must be achieved before an optical system can be applied to a given problem. Thus, a methodology that unifies heretofore unrelated aspects of algorithms, architectures, and technology will serve as a tool to point out fruitful avenues of .research to the optical computing community.

© 1987 Optical Society of America