DOE's supercomputer man

Gil Weigand, the Energy Department's deputy assistant secretary for strategic computing and simulation since 1996, has been thinking big for more than two decades.

In the early 1970s, before supercomputers had even been invented, Weigand chose to pursue his doctorate in engineering at Purdue University because, as he said in a recent interview, "It had one of the biggest computers in the world.''

Now he presides over the $900 million DOE Accelerated Strategic Computing Initiative, overseeing development of the largest and fastest computers on Earth. Under ASCI, the government plans to build supercomputers capable of 100 trillion calculations per second— fast enough to simulate nuclear explosions— by 2004.

If Weigand and DOE ultimately succeed in this endeavor, the government would be able to determine whether its atomic weapons still work without live testing, which is prohibited under the 1995 Comprehensive Test Ban Treaty.

In 1978 Weigand took a job at General Motors Corp. to build a program in computational fluid dynamics— the use of computers to model the behavior of liquids and gases. A year later, the U.S. automobile industry was flagging and there was not enough money to continue the project, so Weigand left to do similar work at DOE's Sandia National Laboratories.

Weigand said he realized he had made the correct career decision soon after his arrival at the lab, when his supervisor calmly instructed him to fix a $10,000 programming error. ''I knew I had made it home when I got to Sandia,'' he said. ''That [$10,000] was about one-third of my total computing budget at GM.''

At Sandia, Weigand found fertile ground for invention. Although he said his Fluid Mechanics and Heat Transfer Group ''did very little experimentation,'' he used part of his budget to buy the group's first massively parallel computer. The system, from a now-defunct company called Ncube, used a then-experimental technique solving complex computing problems by splitting them up among numerous processors.

In 1988 Weigand's group won three prestigious awards for its research in massively parallel processing. The approach has become the dominant architecture for supercomputers today, and the ASCI computers trace their roots to those early models.

Weigand has spent the past decade in positions in which he could promote the parallel computing technology that he helped to pioneer at Sandia. ''Since '89 or '90, I have been associated with, party to or funder of all the largest computers in the world,'' he said.

His association with the world's largest computers resulted from his work at the Defense Advanced Research Projects Agency, where he worked following his stint at Sandia and his subsequent work at DOE. He is an unflinching advocate for supercomputing technology, with ready arguments for why vendors need a cash infusion from the government.

''I'm not as concerned about which companies we're picking,'' he said. ''What I'm really concerned about is whether the U.S. national security community will be able to buy [equipment that is] without peer. It's important to have strong, robust computer companies to work with.''

With its ultimate goal of building more speed-record-breaking computers, ASCI offered Weigand the chance to continue a career that has been spent ''building something from scratch.'' But Weigand also saw a bigger challenge: addressing the question of whether simulation actually will work.

''Can you really come so close...that you can't tell the difference between Mother Nature and what you've simulated on the computer?" he asked. "It's a new discipline. And when you think about that, it's pretty heavy stuff.''

While vendors deliver more high-powered hardware— the latest box works at speeds of more than 3 trillion floating- point operations per second— scientists are writing new software intended to create more lifelike models.

Although most of the hype surrounds the continuous series of announcements touting the world's fastest computers, Weigand is careful not to downplay the significance of software.

"The software is a big deal," he said. "We tend to glamorize the hardware."