Researchers' push: reinvent computing

UC-BERKELEY 1 OF 2 UNIVERSITIES TO RECEIVE GRANT

Intel and Microsoft said last week that they planned to finance two groups of university researchers to start over and design a new generation of computing systems intended to break the industry out of a technological cul-de-sac that threatens to end decades of performance increases in computers.

If the research efforts succeed, this would enable the development of new kinds of portable computers and would help computer engineers tackle areas as diverse as speech recognition, image processing, health care systems and music. For example, a music professor at the University of California-Berkeley, David L. Wessel, envisions a new era of digital musical instruments that would begin to match the rich versatility of acoustic instruments like violins and pianos.

The research grant, worth $20 million over five years, will create independent laboratories at Berkeley and at the University of Illinois, Urbana-Champaign, that will be charting a way to reinvent computing. Each will work on hardware, software and a new generation of applications powered by computer chips containing multiple processors. The University of Illinois plans to contribute an additional $8 million to the project and the Berkeley project is applying for an additional $7 million from a state-supported program to match the industry grants.

The computer industry has generally stopped relying on regular increases in the processing speed of chips. In recent years, it has bet instead that future advances in speed and energy efficiency will come from putting multiple processors on a single silicon chip. A number of computer functions can then be done in parallel rather than sequentially.

The new research agenda was motivated in part by an increasing sense that the industry is in a crisis of a sort because advanced parallel software has failed to emerge quickly. Most programmers today still write programs that solve problems in a serial fashion.

The most advanced consumer-oriented microprocessors have up to eight processors, or cores, on a chip, but the industry is moving toward chips with 100 or more. The problem, according to academic researchers and industry executives, is that the software to keep dozens of processors busy simultaneously for all kinds of computing problems does not exist.

Although the amounts of the grant are modest, both universities have a reputation for early-stage research that has had notable impacts on the computer industry. The director of the new Universal Parallel Computing Research Center at Berkeley, the computer scientist David Patterson, has been associated with significant breakthroughs both in microprocessor and computer storage system design. The University of Illinois laboratory will be led by Marc Snir, a professor of computer science, and Wen-mei Hwu, professor of electrical and computer engineering. The laboratory will include the participation of David Kuck, a University of Illinois researcher who was a pioneer in the field of parallel computing and who is presently an Intel Fellow.

http://www.mercurynews.com/business/ci_8676027