Parallel Object-Oriented Scientific Computing Today

Abstract

High-performance scientific computing is arguably of greater importance than heretofore in the history of computing. Traditionally used primarily in support of fission- and fusion bomb design and weather forecasting, numerically-intensive applications running on massively parallel machines may be found modeling any number of natural phenomena from the smallest scale to the largest—from nuclear dynamics to the formation and evolution of the universe—as well as for the design of human artifacts—automobiles, aircraft, internal combustion engines, pharmaceuticals—the list seems endless. With the announcement of the Japanese Earth Simulator—a machine ostensibly designed expressly for global environmental modeling—and subsequent announcements by other institutions and governments of intents to build comparable machines for similar purposes, it is arguable that such simulations will be the primary consumer of supercomputing capacity worldwide.