Brian J. d'Auriol, Ph.D.

Student Research Details
Return to d'Auriol's Students Supervised Page

Contact

Rajesh Molakaseema, A Parameterized Linear Array With A Reconfigurable Pipelined Bus Systems: LARPBS(p), Department of Computer Science, The University of Texas at El Paso, December 2002. Advisor: Brian J. d'Auriol

Abstract
The Linear Array with a Reconfigurable Pipelined Bus System (LARPBS) is an optical bus-based theoretical parallel computaiton model. This model relies on optical communication between multiple processors. Many parallel algorithms have been proposed using this model of computation. Also, many of these algorithms have exhibited successful unique properties. A survey of parallel computation models is conducted. In particular, it is necessary to evaluate the predictive capability of the model for algorithms running on real machines. One way to do this is to parameterize the parallel computaiton models. Specifically, the LARPBS model lacks such practical basis; moreover, LARPBS model is not parameterized.

In an attempt to move the LARPBS model from theory to practice, in this thesis, a new parameterized parallel computation model, the LARPBS(p) is proposed. The LARPBS(p) model captures computational costs through a set of abstract parameters. The proposed parameters, and subsequently, the cost analysis on the LARPBS(p) model and the cost analysis of some of the primitive operations on LARPBS(p) are presented. The utility of the LARPBS(p) model is demonstrated through the cost analysis of two matrix multiplication algorithms, a naive matrix multiplication algorithm and Cannon's matrix multiplication algorithm.

This model is intended to serve as a basis for a more accurate estimation of execution time of algorithms. We hope the parameterization should be able to provide a feelback for the design characteristics to build a practical LARPBS parallel computer in the near future.


Last Updated: August 3, 2007