ANN ARBOR --- Philip Emeagwali, a
graduate student in The University of Michigan College
of Engineering, received the Gordon Bell Prize for
price/performance in supercomputer research. The
annual Gordon Bell competition is administered by
IEEE Software, a Computer Society publication.
Emeagwali won the prize for oil reservoir simulation
algorithms he developed using Connection Machine
supercomputers at the Los Alamos National
Laboratory, the Argonne National
Laboratory/California Institute of Technology, the
National Center for Supercomputer Applications and
the Thinking Machines Corp.
"One of the fastest supercomputers ever developed,
the Connection Machine has more than 65,000
separate processors cooperating simultaneously to
solve one complex problem," Emeagwali said. "The
technical challenge is getting all 65,000 processors to
work together efficiently."
Emeagwali's goal was to harness the Connection
Machine's potential to solve real-life engineering
problems faster than is possible with other types of
supercomputers. He selected petroleum reservoir
simulation for his research "because of the problem's
technical challenge and because there is a huge
economic benefit to be derived from improving oil
recovery techniques."
By reducing Connection Machine inter-processor
communication time and taking a new approach to the
simulation algorithms, Emeagwali was able to run his
model on the Connection Machine with an operating
speed of 3.1 billion gigaflops, or 3.1 billion calculations
per second.
"This execution speed is twice that of the 1988
Gordon Bell Prize winning entry," Emeagwali said,
"and even exceeds the theoretical peak calculation
speed of more expensive supercomputers, including
the widely used CRAY Y/MP."
Emeagwali said his simulation also could be used to
study the impact of buried nuclear waste on
underground water supplies. "The equations should
apply to any situation where fluid flows through porous
media," he said.
Electronically distributed to the media on April 13, 1990 by Sally Pobojewski for the
News and Information Services department of
the University of Michigan at Ann Arbor.
