April 2009

Project weds SciDAC codes, expertise to INCITE resources

The INCITE (Innovative and Novel Computational Impact on Theory and Experiment) project at the SLAC National Accelerator Laboratory (SLAC) is highly integrated with another Department of Energy Office of Science program: Scientific Discovery through Advanced Computation, or SciDAC.

Advanced Computing for 21st Century Accelerator Science and Technology (AST), a project in the first round of SciDAC from 2001 through 2005, developed multiple computer codes to calculate electromagnetic wave models. The multi-institutional Community Petascale Project for Accelerator Science and Simulation (ComPASS) project, part of the latest SciDAC program, is improving and expanding those codes.

SLAC has been heavily involved in developing these codes. Cho Ng and Lie-Quan Lee, two of the investigators on the INCITE project to model the International Linear Collider, also participate in ComPASS.

The main charge of ComPASS “is to develop state-of-the-art parallel codes to run on DOE supercomputers and advance accelerator science with simulation,” Ng says. “We use ComPASS as support to develop the code – for example, to develop a scalable algorithm that can run on supercomputers with tens of thousands of processors.” INCITE, in turn, gives researchers a chance to try their SciDAC-developed codes on the largest available computers.

T3P, Omega3P, TEM3P and other accelerator simulation codes developed under SciDAC have broad applications. Researchers have used them to design future accelerators and improve existing ones across the offices of High Energy Physics, Basic Energy Sciences and Nuclear Physics under the DOE Office of Science.

The INCITE team also has worked with SciDAC Centers for Enabling Technology and SciDAC Institutes, such as TOPS (Towards Optimal Petascale Simulations) to help efficiently solve the equations behind the simulations, and ITAPS (Interoperable Technologies for Advanced Petascale Simulations to Improve Accuracy and Efficiency) to improve the data mesh. The Institute for Combinatorial Scientific Computing and Petascale Simulations (CSCAPES), a SciDAC Institute based at Purdue University, also has helped balance the computational load among processors in supercomputers the INCITE researchers use. And the Institute for Ultra-Scale Visualization at the University of California at Davis helps the SLAC researchers represent their data graphically.

“This relationship is crucial,” Lee says. “Without SciDAC support our INCITE project cannot really do such a device simulation. In one sense INCITE is providing computing resources to tackle a more challenging problem. SciDAC is providing the algorithms and the necessary collaboration to enable the simulation in the larger scale.”