Search

Earthquakes are among the most complex terrestrial phenomena, and modeling of earthquake dynamics is one of the most challenging computational problems in science. Computational capabilities have advanced to a state where we can perform wavefield simulations for realistic three-dimensional earth models, and gain more insights into the earthquakes that threaten California and many areas of the world. The Southern California Earthquake Center initiated a major earthquake research program called TeraShake to perform physics-based numerical simulations of earthquake processes for large geographical regions, at high resolution, and for high frequencies. For a large scale simulation such as TeraShake, optimization problems tend to emerge that are not significant in smaller scale simulations. This involves both large parallel computation and also massive data management and visualization coordination. In this paper, we describe how we performed single-processor optimization of the TeraShake AWM application, optimization of the I/O handling, and optimization of initialization. We also look at the challenges presented by run-time data archive management and visualization. The improvements made to the TeraShake AWM code enabled execution on the 40k IBM Blue Gene processors and have created a community code that can be used by seismologists to perform petascale earthquake simulations.