Abstract

April 9, 2007

Title: SciDAC-2 Performance Engineering Research Center
Presented by: Bob Lucas
ISI / USC, Los Angeles, CA

Video Podcast

Abstract

As we look to the future, achieving good performance on high-end computing (HEC) systems is growing ever more challenging due to enormous scales, increasing architectural complexity, and increasing application complexity. The SciDAC-2 Performance Engineering Research Institute (PERI) was formed to address these challenges. It is a unified, tripartite research plan encompassing: (1) performance modeling and prediction, (2) automatic performance optimization, and (3) performance engineering of high-profile applications. First, the PERI performance modeling and prediction activity will develop and refine performance models, significantly reducing the cost of collecting the data upon which the models are based and increasing model fidelity, speed, and generality. Second, a principal new component is a research activity in automatic tuning software, which is spurred by the strong user preference for automatic tools. This work will be based on previous successful activities such as ATLAS, which has automatically tuned components of the LAPACK linear algebra library, the highly successful FFTW library, and other recent work. Finally, the third major component, application engagement, will involve direct interactions with SciDAC applications, devoting 30 percent of PERI’s total resources to this activity, including “tiger teams” to focus on particular codes.