UPC:  A Distributed Shared Memory Programming Paradigm

Tarek El-Ghazawi

Department of Electrical and Computer Engineering
The George Washington University

Abstract

UPC, or Unified Parallel C, is an explicit parallel extension of ANSI C.
As in message passing, UPC allows the explicit distribution of data
and work.  As in the shared memory programming paradigm, however, UPC
provides an easy to use syntax and semantics. UPC accomplishes this by
following a distributed shared-memory programming model, and by providing
constructs that facilitate explicit control of data and work distribution
among threads, so that remote memory accesses can be minimized. Thus, UPC
maintains the C language heritage of keeping programmers in control of and
close to the hardware. In this talk the underlying concepts of UPC along
with the main language features will be introduced. Initial performance
results comparing UPC and MPI will be presented.  It will be shown that
with proper optimizations, UPC can compare favorably with other paradigms.
UPC has been gaining acceptance from several vendors who are producing
exploratory compilers.   For more information see upc.gwu.edu.

------------------------------------------------------------

Tarek El-Ghazawi has received the Ph.D. degree in 1988 from New
Mexico State University in Electrical and Computer Engineering. He
is currently an Associate Professor at the Department of Electrical
and Computer Engineering of the George Washington University.
Dr. El-Ghazawi has also held faculty positions at the George Mason
University and Florida Institute of Technology. His research interests
include high-performance computing and architectures, parallel I/O,
and performance evaluations. Dr. El-Ghazawi has published extensively in
these areas and his research has been frequently supported by the DoD,
NASA, and NSF. He has also served as a visiting scientist with CESDIS at
NASA GSFC, and RIACS at NASA Ames Research Center.  He is currently an
associate editor for the International Journal on Parallel and Distributed
Systems and Networking, and was a guest editor for the IEEE Concurrency,
special track on High-Performance Data Mining. He is a Senior Member of
the IEEE, a member of the ACM, and a member of Phi Kappa Phi.