RELATED |
When Philip Emeagwali
was a child in Nigeria, his classmates thought he was a genius.
Today, much of the academic and computer science community would agree.
In February, he won the prestigious 1989 Gordon Bell Prize for programming supercomputers to work faster than ever before and for helping to solve one of the nation's toughest oil recovery problems. Even more impressive, in a field in which most breakthroughs are a collaborative effort, the 35-year-old University of Michigan doctoral candidate did it on his own.
To capture the $1,000 prize, awarded each year by the Institute of for Electrical and Electronics Engineers (IEEE), Emeagwali programmed the Connection Machine, a supercomputer made by Thinking Machines Corp. in Cambridge, Mass., to perform calculations at the rate of 3.1 billion per second. That's twice the speed of the previous Bell prize winner, 24 times faster than the `87 winner, surpasses even the theoretical peak speed of the more expensive and widely accepted supercomputer marketed by Cray Research Inc.
Oil usually is found underground, trapped in rocks. Engineers, able to recover only about 30 percent of the oil in a reservoir, use computers to simulate oil fields and track oil's flow. They're like space-age divining rods, and they'll operate much more swiftly and accurately as a result of Emeagwali's work.
"I chose this problem to tackle because in this field, in order to have the most impact, you have to work on the most serious problems," Emeagwali explains. "I wanted to work on something useful." Emeagwali, who typically works 13-hour days, 7 days a week, avoided computer downtime by linking up with several Connection Machines from New Mexico to Massachusetts via Internet, an electronic, computerized "data highway" that researchers use to communicate and share new discoveries. His interest in parallel computations led him to choose the Connection Machine for his research because it uses 65,000 processors working simultaneously on calculations. It is radically different from conventional supercomputers, which use eight high-powered processors. "It will be interesting to see how this publicity influences the architecture of the supercomputer of the future," Emeagwali says. He developed equations to simulate movement of fluid in underground reservoirs by modifying a set of equations designed and then abandoned by a Russian mathematician in 1938. Speed is the key in solving the equations. Desktop personal computers would take years to solve problems that supercomputers can tackle in minutes. Emeagwali's equations drastically speed up supercomputers designed like the Connection Machine. Emeagwali expects his computations to be applicable to a number of other significant problems. Calculating the movement of buried nuclear waste, tracking the spread of AIDS, improving the accuracy of weather prediction and testing for automobile safety all can be accomplished inexpensively and quickly through computer simulation, he says. A relative newcomer to the field of scientific computation, Emeagwali admits his choice of the Connection Machine and his solitary working style are unorthodox. "It has something to do with my personality. If everyone goes east, I feel I should go west, so I will be among the first to find the solution." He has a history of walking and working alone. Classmates in Nigeria recognized Emeagwali's unusual ability in math in seventh grade and nicknamed him Calculus. "I think some of them didn't even know my real name," he laughs. "To this day, if I hear someone call me "Calculus," I know without looking it's an old friend from home." The oldest of nine children, Emeagwali was forced to quit school after eighth grade because of financial constraints. High schools in Nigeria were private, expensive and not located in every city. His father, James, a nurse, encouraged his son's interest and talent in math and tutored him at night. Emeagwali studied on his own to finish high school and earn a General Certificate of Education from the University of London. When he was 17, he won a scholarship to Oregon State University, where he majored in math. He since has earned two master's from George Washington University, one in civil engineering and the other in ocean, coastal and marine engineering, and a master's in mathematics from the University of Maryland. He has worked as a civil engineer in Maryland and Wyoming, and for the National Weather Service, for which he wrote a thesis on mathematical calculations used in forecasting floods. As is traditional in Nigerian culture, Emeagwali has assumed responsibility for his younger siblings. He's brought seven brothers and sisters to the Washington, D.C., area, where they attend school. His mother also lives there, and his father plans to move to the United States soon. One sister also is still in Nigeria. Emeagwali's wife, Dale, is a molecular biologist on the research faculty at the University of Michigan Medical School. "She provides lots of support," he says. The couple expect their first child this summer. There's not a lot of free time in Emeagwali's schedule, but he tries to fit exercise in daily.
"My work is challenging, tiring, rewarding and engrossing," he explains. "I exercise to keep my sanity; otherwise I can't sleep or work well. I like to get outside, so I took up tennis. If the weather is bad, I swim or jog." Emeagwali expects to receive his doctorate in June, and hopes to pursue a research position in an academic community. But it will take two or three months to catch up with the requests for information about his discoveries from across the country and around the world. "Since I walked alone on this project, I am the sole source on it," he says. He doesn't mind the attention. "Recognition is what every scientist wants. I would like to stay in research and contribute to the solving of important problems.
Some research is so esoteric that researchers are reluctant to step down and explain it to the average person,"
he goes on, "but we have to be able to explain our goals and relate
them to everyday life."
Click on emeagwali.com for more information.
|
