Abstract

In this paper we study the resource requirements of electrooptical organizations in performing digital computing tasks. We define a generic model of parallel computation using optical interconnects, called the optical model of computation (OMC). In this model, computation is performed in digital electronics and communication is performed using free space optics. Using this model we derive relationships between information transfer and computational resources in solving a given problem. To illustrate our results, we concentrate on a computationally intensive operation, 2-D digital image convolution. Irrespective of the input/output scheme and the order of computation, we show a lower bound of Ω(nw) on the optical volume required for convolving a w × w kernel with an n × n image, if the input bits are given to the system only once.

© 1991 Optical Society of America

Area–time trade-offs in arrays with optical pipelined buses

Sandy Pavel and Selim G. Akl
Appl. Opt. 35(11) 1827-1835 (1996)

Lower bounds on the computational efficiency of optical computing systems

Richard Barakat and John Reif
Appl. Opt. 26(6) 1015-1018 (1987)

Efficient parallel algorithms for optical computing with the discrete Fourier transform (DFT) primitive

John H. Reif and Akhilesh Tyagi
Appl. Opt. 36(29) 7327-7340 (1997)

Optical digital processor using arrays of symmetric self-electrooptic effect devices

Michael E. Prise, Nicholas C. Craft, M. M. Downs, R. E. LaMarche, L. A. D’Asaro, Leo M. F. Chirovsky, and Miles J. Murdocca
Appl. Opt. 30(17) 2287-2296 (1991)

Optical systems for digit-serial computation

Caroline J. Perlee and David P. Casasent
Appl. Opt. 28(3) 611-626 (1989)