Abstract

Optical techniques have a number of potential benefits for information processing. They include a high degree of parallelism, high speeds, short pulses and non-interference of signals. Many of the attempts to use the non-linear optical technology which is beginning to emerge in the device physics area has concentrated on the spatial parallelism available in optics. Problems which have arisen with this approach have included design and fabrication of arrays of optical elements, accuracy of data representation, controlled permutation of data, synchronization with a master clock and optical to electronic interfacing. But spatial parallelism is not the only advantage of optical computing. The extremely high speeds and short pulses which are possible allow the exploitation of the time domain to obtain significant processing power. Serial computer designs are based on the time domain. Exploiting the time domain is not an alternative to spatial parallelism in optical computing but is complementary to it. Since information in an optical computer is represented by pulses propagating at the speed of light there is a homogeneity between time and space which is not present in an electronic computer. Difficult serial design issues must be addressed in optical computer architectures even in predominantly parallel designs. Finally, there are reasons why a serial architecture may lead more rapidly to functional optical computers than one relying heavily on spatial parallelism.

© 1987 Optical Society of America