Scientific Computing Group members (from left) Sherry Li, Andrew Canning, and Tammy
Welcome work with NERSC users on scientific applications in such fields as physics,
chemistry, fluid dynamics, earth sciences, materials science, and mathematics.
Scientific Computing Group staff members work with NERSC clients to make the scientist-supercomputer interface as productive as possible. They also help develop new tools for the benefit of the larger supercomputing community.
For example, group members Xiaoye "Sherry" Li and Osni Marques (not shown) played a key supporting role in the development of a new algorithm library for performing dense linear algebra calculations on distributed memory computers. Besides fine-tuning and testing various parts of the package, they also recorded the fastest performance for running the algorithms among the various supercomputers used. The library of algorithms, called ScaLAPACK for Scaled Linear Algebra Package, was placed in the public domain for researchers in such areas as physics, chemistry, engineering, and materials science.
As NERSC's 512-processor Cray T3E was being put through its specified acceptance tests, a small number of codes from NERSC clients were also being run to test the overall capabilities of the machine. Group member Andrew Canning, working with the Materials Science Grand Challenge team headed by Malcolm Stocks of Oak Ridge National Laboratory, helped produce some of the computer's first scientific results (see the Science Highlight) -- even before it officially went on line. The code, which simulates 498 atoms and 996 electronic states, is a first-principles quantum calculation, "a state-of-the-art calculation that would push the memory and performance limits of any supercomputer," Canning said. "The T3E, with 512 processors, is one of the few machines capable of handling this type of calculation."
To work with the eight Grand Challenge teams using NERSC, the group rolled out a virtual red carpet to participating researchers around the country. "Our aim was to make NERSC an indispensable partner in solving these problems," said Scientific Computing Group Leader Tammy Welcome. "By assigning group members as points of contact and meeting with each team, we learned about their computational requirements and how they intend to use NERSC. At the same time, we let them know about our capabilities and the level of collaboration we can provide."