Abstract

The fundamental advantage which optics has is its parallel communications capability. Lenses, mirrors, and prisms can easily communicate millions of wide bandwidth channels in parallel. The question arises as to how to take advantage of this capability. This difficulty is more a computer science problem than an optics problem. Large arrays of processors have yet to be built and successfully applied. The two basic approaches being pursued involve either a single instruction multiple data (SIMD) or a multiple instruction multiple data (MIMD) architecture. The single instruction multiple data architecture is the most prevalent version since it is simpler conceptually, however it is limited in terms of application and difficult to use. The more interesting architecture is the multiple instruction and multiple data configuration. The main difficulty in applying this architecture is finding a means of coordinating the various asynchronous processes. This need for coordination brings alone with it a large communications burden. The classification of parallel processors into SIMD and MIMD processors is formal and only hints at how the parallel communications capability of optics might be used.

© 1985 Optical Society of America