PHILIP EMEAGWALI, a doctoral student in the U-M College of Engineering, was awarded the 1989 Gordon Bell Prize for supercomputing research. The annual Gordon Bell Prize is awarded by the Institute for Electrical and Electronic Engineers (IEEE) Computer Society in recognition of outstanding achievement in the use of supercomputers to solve practical and important scientific and engineering problems. The prize is considered to be the annual high point of supercomputer research. The 1989 award was presented to Philip Emeagwali at the IEEE Computer Society's 35th CompCon conference in San Francisco, California on February 28, 1990.
Emeagwali, who says that "the best way to solve computation-intensive problems is to use massively parallel computers," performed his award-winning computation entirely using CAEN systems to connect to several national laboratories and supercomputer centers. CAEN's connection to the Internet allowed him to conduct research on Connection Machine supercomputers at the Los Alamos National Laboratory, the Argonne National Laboratory/California Institute of Technology, the National Center for Supercomputer Applications (NCSA), and the Thinking Machines Corporation. Interestingly, Emeagwali has never physically seen a Connection Machine.
The Connection Machine, one of the fastest supercomputers ever built, uses 65,000 separate processors to perform 65,000 calculations simultaneously. The computational model developed by Emeagwali allowed the Connection Machine to run at an operating speed of 3.1 gigaflops (3.1 billion calculationsper second), which exceeds the theoretical peak calculation speed of much more expensive supercomputers, including the $30 million CRAY Y/MP.
Emeagwali designed and implemented a highly efficient and parallel algorithm for petroleum reservoir simulation on massively parallel computers. "I have checked with several reservoir engineers who feel that his calculation is of real importance and very fast," explained Alan Karp, one of the judges. "His explicit method not only generates lots of megaflops, but solves problems faster than implicit methods," said Karp.
Beyond the highly practical benefits of such simulations, Emeagwali's theoretical approach is applicable to a wide range of scientific and engineering problems. His work with the Connection Machine has greatly reduced inter-processor communication time and thus has significantly expanded the horizons for supercomputing in the world of scientific and engineering computing.
By Ben Vonk for CAEN (Computer Aided Engineering Network) Newsletter, (a University of Michigan publication) May/June 1990
Click on emeagwali.com for more information.