Parallelism through multithreaded programming in a constraint-based system.
Item
-
Title
-
Parallelism through multithreaded programming in a constraint-based system.
-
Identifier
-
AAI9732951
-
identifier
-
9732951
-
Creator
-
Muller, Lawrence Gerard.
-
Contributor
-
Adviser: Kenneth McAloon
-
Date
-
1997
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Computer Science
-
Abstract
-
This research explores the applicability of multithreaded programming to constraint based programming. To that end, we used 2LP, a constraint based linear programming language whose framework has an imperative and declarative nature, and our current research extends the system to include language level control over parallelism. We also demonstrate the inherent scalability of multithreaded programming. In addition, insights into the issues of reorganizing a single-threaded system into a multithreaded system will be discussed. To benchmark success and failure, a number of applications are used; these include netlib problems such as set covering. This research is the first time a multithreaded constraint system has been implemented.;2LP has previously benefited from transforming a local area network into a parallel machine. This work explored distributed programming models using an embedded itinerary mechanism on the one hand and explicitly coded parallelism on the other. Further, this research demonstrated that significant load-balancing can be obtained using cooperative threads. The hardware architectures that this research employes as a test-bed are two currently available symmetric multiprocessor (SMP) systems. The operating systems utilized are Microsoft's NT system, and Sun's Solaris system (including Sun's multithreaded library).
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
