Author

Sheri Zimmel

Date of Award

8-1-1994

Document Type

Thesis

Degree Name

Master of Arts (MA)

Department

Computer Science

First Advisor

Dr. Betty Hickman

Abstract

The objective of this thesis is to implement and empirically evaluate the perfor­mance of serial and parallel algorithms for the multidimensional convex hull problem. In recent years several new serial algorithms and parallel variants thereof have been developed. However, the algorithms were implemented on machines having very dif­ferent architectures with speeds differing by orders of magnitude. As a result, no comparison of the implementations can be done. Inefficiencies, ranging from major to minor in scope, are also inherent in each of these previous implementations. In this study we designed and implemented improved versions of existing serial algorithms for the multidimensional convex hull problem, designed and implemented improved parallel versions of these algorithms, and performed extensive computa­tional testing to determine the relative merits of each. All of the algorithms are based on mathematical programming formulations - those in which an objective and a set of limitations, or constraints, are stated in the form of mathematical functions. The computational results, obtained on a shared-memory multiprocessor, show that the algorithms vary in performance depending on certain characteristics of the problem data. However, one of the implementations was clearly superior to the others in all cases.

Comments

A Thesis Presented to the Department of Computer Science and the Faculty of the Graduate College University of Nebraska In Partial Fulfillment of the Requirements for the Degree Master of Arts University of Nebraska at Omaha. Copyright 1994 Sheri Zimmel

COinS