Earthquake ground motions pose an ever present risk to engineered structures and the infrastructure that modern life depends on. Civilization has evolved in close proximity to active earthquake faults and sedimentary basins that amplify seismic motions.
However, many cities at high ground motion risk haven’t experienced damaging motions due to long time intervals between large earthquake events.
In the September/October issue of Computing in Science & Engineering, researchers say application modernization for massively parallel time-domain simulations (login may be required for full text) of earthquake ground motion in 3D models is increasing application resolution and providing ground motion estimates for critical infrastructure risk evaluations.
Improvements to the geophysics application code SW4 algorithms, developed while porting the code to systems at Lawrence Berkeley National Laboratory, revealed that reorganizing operation order can improve performance for massive problems.
“In the absence of empirical data for large earthquakes in the near-fault region, seismologists are using high-performance computing (HPC) to simulate ground motions,” the researchers said.
“In this study, we describe SW4 (seismic waves 4th order), a summation-by-parts finite-difference code for parallel simulations of seismic wave propagation. We describe how SW4 is being enhanced to run simulations on the next generation of HPC systems and how mesh refinement provides a remarkable reduction in the computational effort required to simulate a given earthquake. We also describe preliminary efforts to optimize SW4 for the Cori Phase 2 architecture at the National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory,” the authors said.
About Lori Cameron
Lori Cameron is a Senior Writer for the IEEE Computer Society and currently writes regular features for Computer magazine, Computing Edge, and the Computing Now and Magazine Roundup websites. Contact her at l.cameron@computer.org. Follow her on LinkedIn.
