University of Houston
Department of Computer Science
In
partial fulfillment of the Requirements for the Degree of
Master of
Science
Amit J. Patil
will defend his
thesis
Performance Oriented Programming for Distributed Shared Memory Machines
Abstract
OpenMP is emerging as viable high-level programming model for shared memory parallel systems. It was conceived to enable easy, portable application development on this range of systems, and it has also been implemented on ccNUMA architectures. Unfortunately, it is hard to obtain high performance on the latter architecture, particularly when large number of threads is involved. This thesis work discusses the difficulties faced when writing OpenMP applications for ccNUMA systems and explains how vendors have tried to overcome them. The focus is mainly on SGI's Origin 2000 system on which various experiments were conducted to illustrate the impact of the vendor's efforts. We compare codes written in a standard, loop-level parallel style under OpenMP with alternative versions written in an SPMD fashion, also realized via OpenMP, and show that the latter consistently provides superior performance. A carefully chosen set of language extensions similar to HPF can help us translate programs from the former style to latter. An extended language, if well compiled, should improve the attractiveness of OpenMP as a language for high performance on an important class of modern architectures.
Date: Monday, May 3, 2004
Time: 11:30 AM
Place: 218 D - PGH
Faculty, students, and the general public are invited.
Thesis Advisor: Dr. Barbara Chapman