As we enter the exascale computing era, efficiently utilizing power and optimizing the performance of scientific applications under power and energy constraints has become critical and challenging. We propose a low-overhead autotuning framework to autotune performance and energy for various hybrid MPI/OpenMP applications at large scales, then use this framework to autotune four ECP proxy applications---XSBench, AMG, SWFFT, and SW4lite. Our approach uses Bayesian optimization with a Random Forest surrogate model to effectively search parameter spaces with up to 6 million different configurations on two large-scale production systems, Cray XC40 Theta at Argonne National Laboratory and Summit at Oak Ridge National Laboratory. The experimental results show that our autotuning framework at large scales has low overhead and achieves good scalability. Using the proposed autotuning framework to identify the best configurations, we achieve up to 91.59% performance improvement, up to 21.2% energy savings and up to 37.84% EDP (energy delay product) improvement on up to 262,144 cores.

Xingfu Wu

Dr. Wu joined the Mathematics and Computer Science Division at Argonne National Lab as a staff scientist in July 2017 and has a joint appointment at the University of Chicago as CASE Senior Scientist. He worked as the faculty member in Department of Computer Science & Engineering at Texas A&M University from 2003 to 2017. He was a post-doc researcher in the Department of Electrical & Computer Engineering at Northwestern University from 1999 to 2003. He is the author of a monograph Performance Evaluation, Prediction and Visualization of Parallel Systems (Kluwer, 1999). His research interests are performance modeling and analysis, autotuning, high performance computing, energy efficient computing, and power modeling and analysis.