Forbidden by treaty from exploding any more actual nuclear bombs, the U.S. Department of Energy plans to begin simulating nuclear explosions on computer with the assistance of several University professors.
The 10-year, $1 billion project, called the Accelerated Strategic Computing Initiative, comes in response to the Comprehensive Nuclear Test Ban Treaty, an agreement signed last year by the United States.
“The big push in the (strategic computing) program is on computer simulation of the processes that go on in nuclear bombs and explosions,” said computer science professor Vipin Kumar.
Such simulation is needed because the United States can no longer test its nuclear weapons directly in underground explosions.
Much of the project’s computing will be done on powerful supercomputers that will soon be installed at the Lawrence Livermore National Laboratories in California, and the Sandia and Los Alamos National Laboratories in New Mexico. The computers will be able to perform several trillion calculations per second. That compares to 150 billion calculations per second for the Cray T3E, today’s top-of-the-line supercomputer.
“Some of the programs we developed here for graph partitioning and simulation (are already being used by) the people at Los Alamos National Laboratory and at Livermore,” Kumar said.
Kumar said he hopes that his group will receive a Department of Energy grant of $300,000 to $400,000 later this year.
The supercomputers used in the project will be massively parallel, which means that they can process many instructions simultaneously.
“Processing on a parallel computer requires specific kinds of programs. Our expertise is in developing parallel programs — programs designed to work effectively on massively parallel computers,” Kumar said.
In the 1980s, Japan’s much-publicized Fifth Generation Project attempted to create a program that would automatically optimize code for use on massively parallel machines. The Japanese compiler was intended to turn programming code, written by programmers, into machine language that the computers could understand — essentially a string of ones and zeros.
“That was a very expensive experiment. The results were negative, basically,” said Kumar. “They were trying to do this automatically. When we write programs, we explicitly decompose the computation ourselves. It makes our job harder, but it guarantees success.”
While Japan used a highly specialized computer language that few outside their project understood, Kumar writes code in conventional computer languages like C and FORTRAN.
Not all those who seek funds from the strategic computing initiative see a need for public scrutiny.
Astronomy professor Paul Woodward, who is applying for a strategic computing grant, said the program is not fundamentally different from any other grant-supported research.
Woodward’s proposal is separate from Kumar’s and focuses on the issue of graphic display.
Kumar said that he does not see any ethical problems with his research.
“In this case they are not designing nuclear bombs. They are trying to ensure the safety of their nuclear stockpile. So the goals are much more noble. If they were using (the research) to design nuclear bombs, I would have to think about it more carefully,” he said.