The High Performance GeoComputing Laboratory (HPGeoC) at the San Diego Supercomputer Center (SDSC), an organized research unit at the University of California, San Diego, was named a winner of the HPC Innovation Excellence Award by the International Data Corporation (IDC) for developing a highly-scalable computer code that promises to dramatically cut both research times and energy costs in simulating seismic hazards throughout California and elsewhere.
The announcement of the award’s recipients came during the ISC’13 supercomputing industry conference in Leipzig, Germany. The award recognizes “noteworthy achievements by users of high performance computing (HPC) technologies”, particularly “scientific success stories involving HPC.”
“We are grateful to IDC for this award which recognizes our efforts to speed regional earthquake simulations for use in earthquake engineering and disaster management as part of a larger computational effort coordinated by the Southern California Earthquake Center (SCEC),” said Yifeng Cui, a computational scientist at SDSC and the leader of the HPGeoC lab, jointly funded by SCEC.
The IDC award cites Cui’s team for the development of a highly scalable and efficient code — using GPUs (graphical processing units) as opposed to CPUs, or central processing units — that accelerate a widely-used wave propagation code called AWP-ODC, which stands for Anelastic Wave Propagation by Olsen, Day, and Cui (named for Kim Olsen and Steven Day, geological science professors at San Diego State University and SDSC’s Cui). This community code, developed and supported by SCEC, simulates the dynamic rupture and wave propagation that occurs during large-scale earthquakes.
The GPU code was restructured to maximize throughput for reduced time-to-solution, achieving a sustained 2.3 petaflops (quadrillion calculations per second) on a recent benchmark run for a problem twice as large as the previous M8, a magnitude-8 earthquake simulation on the southern San Andreas fault (http://www.sdsc.edu/News%20Items/PR081910_m8_earthqua.html). The Titan supercomputer, based at Oak Ridge National Laboratories and funded by the U.S. Department of Energy, was used for this simulation. Titan is equipped with Cray XK7 systems and NVIDIA’s Tesla K20X GPU accelerators.
Moreover, the restructured code resulted in a 110-fold speedup over a heavily optimized CPU code for a key strain tensor calculation critical to probabilistic seismic hazard analysis (PSHA), a standard used by decision-makers to help reduce seismic risk and increase community response to earthquakes.
“The results are expected to take physics-based seismic hazard analysis to a new level with petascale computers, with the potential of saving hundreds of millions of core-hours of computing time,” said Cui.
Also contributing to the code’s development was Jun Zhou and Efecan Poyraz, graduate students of Electrical and Computer Engineering (ECE) with the Jacobs School of Engineering at UC San Diego; Dong Ju Choi, a computational scientist at SDSC; and Clark Guest, associate professor of ECE.
Other winners of this year’s HPC Innovation Excellence Awards are Alenia Aermacchi (Italy); DOD High Performance Computing Modernization Program; ESTECO and Airworks Engineering (Italy); University College London and NAG HECTOR dCSE (UK); Bottero S. p. A. (Italy); Polestar Racing (Sweden); and RENCI (US).