Collaborative Research: SGER--Profiling and Analysis of Southern California Earthquake Center (SCEC) HPC Code for Petascale Simulations

该项目是一项探索性活动，旨在测试在性能建模和优化方面的新专业知识和技术在地震数值模拟中的应用。它包括初步工作，以确定未来努力开发千兆级版本的非弹性波建模（AWM）代码的可行性，该代码构成了南加州地震中心社区建模环境的一部分。如果AWM和类似的模型可以在大规模的计算资源上很好地运行，比如即将到来的千兆级系统，那么研究人员就有可能使用该模型来研究南加州和其他地方地震的“最坏”情况下的地面运动，从而确定对估计建筑结构破坏很重要的频率。该项目是2006年春季“千万亿次计算与地球科学”研讨会产生的一系列试点项目之一。这些试点活动的目的都与在地球科学不同领域使用高性能计算建模有关，目的是确定如何将前沿地球科学研究中使用的模拟工具应用于未来几年将部署的大规模计算系统。这些分析的结果将在2006年秋季举行的第二届“千万亿次计算与地球科学”研讨会上进行讨论。这个SGER项目涉及加州大学圣地亚哥分校（UCSD）的计算机科学家（专门从事数值应用的详细分析和优化）和南加州大学（USC）和圣地亚哥州立大学（SDSU）的地球科学家（南加州地震中心的一部分）之间的合作。这项工作的更广泛影响源于千万亿次计算机地震建模的潜力，它能够比目前更准确地预测地震危险的地理分布。

This project is an exploratory activity intended to test the application of new expertise and techniques in performance modeling and optimization to the numerical simulation of earthquakes. It encompasses preliminary work to determine the feasibility of a future effort to develop a petascale version of the Anelastic Wave Modeling (AWM) code that forms part of the Southern California Earthquake Center's community modeling environment. If the AWM and similar models can be adapted to run well on large-scale computational resources, such as forthcoming petascale systems, then it will be possible for researchers to use the model to study the ground motion that would accompany "worst-case" scenarios of earthquakes in Southern California, and elsewhere, resolving frequencies that are important in estimating the damage to built structures. The project is one of a series of pilot projects spawned by a workshop on "Petascale Computing and the Geosciences" in Spring 2006. The purpose of these pilot activities, all related to the use of high-performance computational modeling in different parts of the geosciences, is to determine how the simulation tools used in leading-edge geoscience research can be adapted for the very large-scale computing systems that will be deployed in the next few years. The results of these analyses will be discussed at a second workshop on "Petascale Computing and the Geosciences" in Fall 2006. This SGER project involves collaboration between computer scientists from University of California-San Diego (UCSD), who specialize in the detailed analysis and optimization of numerical applications, and geoscientists from University of Southern California (USC) and San Diego State University (SDSU) who are part of the Southern California Earthquake Center.Broader impacts of this work stem from the potential of earthquake modeling on petascale computers to be able to predict the geographic distribution of seismic hazards more accurately than is presently possible.