Concepts familiar from grade-school algebra have broad ramifications in computer science.
A $9 million grant to support a National Consortium for High Performance Computing has been made by the Defense Advanced Research Projects Agency (DARPA) to MIT, which will be the consortium's lead organization through the Laboratory for Computer Science (LCS).
The grant's purpose is to accelerate advances in high-performance computing technology and stimulate collaborative partnerships to solve important computational problems that can affect the nation's security and productivity.
The major MIT figures involved in the effort are Albert Vezza, LCS associate director, who is the principal investigator, and Professor Michael L. Dertouzos, LCS director.
The new consortium will focus on DOD problems in many areas, including climatology, physics, chemistry and biology. It will also seek ways in which high-performance computing can help solve engineering problems in industries, including the airliner, automotive, computer, defense, electronic, pharmaceutical and process-control industries.
"High performance computing will be a key tool allowing government, academia and industry to solve problems critical to our national needs, such as improving product quality through better design, faster design turn-around time, better manufacturing and manufacturability," Vezza said.
Professor Dertouzos, chairman of the MIT Commission on Industrial Productivity which produced the book Made in America: Regaining the Productive Edge (MIT Press 1989), said, "Improvements in the manner in which American industry designs and manufactures products are essential at this stage of US industrial performance and will have a significant positive impact on American productivity."
In addition, the new consortium will seek to bring high-performance computing to local communities and community school systems. Consortium members are: Defense Supercomputing Research Alliance (DeSRA), the Front Range Consortium, SCOUT, the Parallel Supercomputing Consortium (SuperPar), TeraComputing and the US Chamber of Commerce.
The members are smaller collaborative groups whose major partners are: DeSRA-Sandia National Laboratory, Army Research Laboratory, Phillips Laboratory; FrontRange-National Center for Atmospheric Research, Center for Applied Parallel Processing at the University of Colorado, Boulder; SCOUT (Supercomputing Triad)-the LCS collaborative comprised of scientists from many departments at Harvard and MIT; SuperPar-San Diego Supercomputer Center and the Naval Command Control and Ocean Surveillance Center; TeraComputing-National Center for Supercomputing Applications, Pittsburgh Supercomputer Center, Naval Research Laboratory; US Chamber of Commerce-the Chamber and Community Learning and Information Network.
DARPA selected those six to initiate the consortium from a larger group of proposals submitted last year. Representatives of the six met in January with DARPA officials, who had invited the National Science Foundation to participate. The six consortia were asked to form a national consortium. An interim council was established at the January meeting and MIT-LCS was elected to chair the new effort.
The new consortium's goals and objectives include:
- To accelerate the process of weaving high-performance computing into the every day fabric of all American communities through their schools "and with significant benefit to science and engineering for both defense and non-defense purposes and to increase our national competitiveness by reducing design, development and testing time through modeling and simulation.
- To accelerate the process of making high-performance computing available to all by "educating a critical mass of scientists" in its use "who in turn can educate others and/or make high-performance-computers easier to use."
- To stimulate development of interdisciplinary high-performance computing collaborations among government, academic and business partners.
- To give scientists from diverse disciplines access to advanced large-scale, high-performance computing systems capable of from billions to tens of trillions of operations per second and beyond.
A version of this article appeared in the November 18, 1992 issue of MIT Tech Talk (Volume 37, Number 14).