Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical and Computer Engineering

First Advisor

Jerry L. Trahan


Recently, many models using reconfigurable optically pipelined buses have been proposed in the literature. A system with an optically pipelined bus uses optical waveguides, with unidirectional propagation and predictable delays, instead of electrical buses to transfer information among processors. These two properties enable synchronized concurrent access to an optical bus in a pipelined fashion. Combined with the abilities of the bus structure to broadcast and multicast, this architecture suits many communication-intensive applications. We establish the equivalence of three such one-dimensional optical models, namely the LARPBS, LPB, and POB. This implies an automatic translation of algorithms (without loss of speed or efficiency) among these models. In particular, since the LPB is the same as an LARPBS without the ability to segment its buses, their equivalence establishes reconfigurable delays (rather than segmenting ability) as the key to the power of optically pipelined models. We also present simulations for a number of two-dimensional optical models and establish that they possess the same complexity, so that any of these models can simulate a step of one of the other models in constant time with a polynomial increase in size. Specifically, we determine the complexity of three two-dimensional optical models (the PR-Mesh, APPBS, and AROB) to be the same as the well known LR-Mesh and the cycle-free LR-Mesh. We develop algorithms for the LARPBS and PR-Mesh that are more efficient than existing algorithms in part by exploiting the pipelining, segmenting, and multicasting characteristics of these models. We also consider the implications of certain physical constraints placed on the system by restricting the distance over which two processors are able to communicate. All algorithms developed for these models assume that a healthy system is available. We present some fundamental algorithms that are able to tolerate up to N/2 faults on an N-processor LARPBS. We then extend these results to apply to other algorithms in the areas of image processing and matrix operations.