Analytical and Computational Framework for n-Body Simulations

这个项目将探讨一系列的计算问题有关的大规模模拟的n体系统使用平行配方的多极和埃瓦尔德求和技术。 如今，这些技术已经可以扩展到数千万个天体的系统，包括模拟真实的物质和整个星系的分子动力学。然而，越来越多的数据分析、解释和可视化这些结果带来了严重的瓶颈。 有效的算法预处理的数据，因为它是产生和快速的技术查询预处理的数据是必要的。 该项目将为这些算法提供以下方法：* 分析量化各种应用背景下的一系列高级n体方法的性能和并行处理开销。*动态地将模拟转向具有“有趣”属性的子域的方法。通过分子动力学模拟预测宏观性质的技术，如杨氏模量、介电常数和磁导率。*基于选择性聚焦和细节层次的数据分析和可视化方法紧密集成到模拟中。这些方法将在材料模拟、蛋白质合成和静电学中的分子动力学应用中得到验证。

This project will explore a range of computational issues related tolarge-scale simulation of n-body systems using parallel formulations ofmultipole and Ewald summation techniques. Today, these techniques areknown to scale to systems with tens of millions of bodies, includingsimulations of molecular dynamics of real materials and full galaxies.However, increasingly data analysis, interpretation, and visualization ofthese results present serious bottlenecks. Effective algorithms forpre-processing the data as it is generated and fast techniques for queryingthe pre-processed data are necessary. This project will contribute thefollowing methods to those algorithms:* Analytical quantification of performance and parallel processingoverheads of a range of advanced n-body methods in the context of variousapplications.* Methods for dynamically steering the simulation to subdomains thatexhibit "interesting" properties.* Techniques for predicting macro properties such as Young's modulus,permittivity, and permeability from molecular dynamics simulations.* Methods for data analysis and visualization based on selective focus andlevel of detail that are tightly integrated into the simulation.The methods will be validated on molecular dynamics applications inmaterials simulation, protein synthesis, and electrostatics.