Você está na página 1de 5

MULTITHREADED CONSTRAINT PROGRAMMING

AND APPLICATIONS
by Fabian Zabatta
A dissertation submitted to the Graduate Faculty in Computer Science i
n partialfulfillment of the requirements for the degree of Doctor of
Philosophy, The City University of New York
1999
ii (C) 1999 FABIAN ZABATTA
All Rights Reserved
iii This manuscript has been read and accepted for the Graduate Facult
y in ComputerScience in satisfaction of the dissertation requirement fo
r the degree of Doctor of Philosophy.
___________________________ ____________________________________Date Professor K
enneth McAloon
Chair of Examining Committee
___________________________ ____________________________________Date Professor S
tanley Habib
Executive Officer
Professor David ArnowProfessor Jim Cox Dr. Lioun Wee Chen_______________________
_____________ Supervisory Committee
THE CITY UNIVERSITY OF NEW YORK
iv Abstract MULTITHREADED CONSTRAINT PROGRAMMING AND APPLICATIONS
by Fabian Zabatta
Adviser: Professor Kenneth McAloon Constraint programming is a powerful and robu
st software technology for modeling and solving difficult combinatorial problems
. However, as problems grow larger in scale and complexity, they become incr
easingly difficult to solve and their execution times can exhibit inordi
nate growth. One solution is multithreaded parallelization. Multithreaded p
arallelization can greatly improve performance, not only increasing the size of
solvable problems but also improving solutions that were previously limited by t
ime.
This thesis addresses the fundamental issues concerning the problem of
application-based multithreaded constraint programming. The focus of th
e thesis is on methods for achieving application-based parallelism in c
onstraint programming using threads. Included are the presentations of algori
thms such as Dynamic Thread Creation, a new dynamic load balancing scheme spe
cific to multithreading and the Multithreaded Least Discrepancy Search an
d the Multithreaded Best-First With Backtracking Search, two new parallel
search strategies that break away from the classic backtracking schemes normally
associated with constraint programming. These new algorithms are applied to be
nchmark problems with data from the literature and empirical results a
re reported.
v Acknowledgements I first would like to thank Professors Kenneth McAloon, David
Arnow, James Cox, and Dr. Lioun Wee Chen, the members of my dissertation commit
tee. I would like to express special gratitude to my advisor Kenneth McAl
oon for the introduction to constraint programming, the consistent guidance
, and for always keeping my best interests in mind. I would also like to thank
Dr. Carol Tretkoff, a mentor who helped inspire me to study parallel sea
rch.

Você também pode gostar