Masters Theses

Date of Award

8-2003

Degree Type

Thesis

Degree Name

Master of Science

Major

Electrical Engineering

Major Professor

Dr. Gregory D. Peterson

Committee Members

Dr. Donald W. Bouldin, Dr. Michael A. Langston, Dr. Seong G. Kong

Abstract

A lot of raw computing power is needed in many scientific computing applications and simulations. MATLAB®† is one of the popular choices as a language for technical computing. Presented here are approaches for MATLAB based applications acceleration using High Performance Reconfigurable Computing (HPRC) machines. Typically, these are a cluster of Von Neumann architecture based systems with none or more FPGA reconfigurable boards. As a case study, an Image Correlation Algorithm has been ported on this architecture platform. As a second case study, the recursive training process in an Artificial Neural Network (ANN) to realize an optimum network has been accelerated, by porting it to HPC Systems. The approaches taken are analyzed with respect to target scenarios, end users perspective, programming efficiency and performance. Disclaimer: Some material in this text has been used and reproduced with appropriate references and permissions where required. † MATLAB® is a registered trademark of The Mathworks, Inc. ©1994-2003.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS