计算物理 ›› 2004, Vol. 21 ›› Issue (5): 377-385.

• 研究论文 •    下一篇

基于“块-单元”数据结构的分子动力学并行计算

曹小林, 莫则尧, 张景琳, 陈其峰   

  1. 北京应用物理与计算数学研究所高性能计算中心, 计算物理实验室, 北京 100088
  • 收稿日期:2004-04-28 修回日期:2004-05-28 出版日期:2004-09-25 发布日期:2004-09-25
  • 作者简介:曹小林(1974-),男,江西瑞昌,副研究员,博士,主要从事大规模科学与工程计算中并行算法与并行应用软件的研究,北京8009信箱17分箱100088.
  • 基金资助:
    国家自然科学基金(60273030);中国工程物理研究院重点基金(2002Z01041)资助项目.

Parallel Computation for Molecular Dynamics Simulation Based on Cell-Block Data Strucutres

CAO Xiao-lin, MO Ze-yao, ZHANG Jing-lin, CHEN Qi-feng   

  1. High Performance Computing Center;State Key Laboratory of Computational Physics, Institute of Applied Physics and Computational Mathematics, Beijing 100088, China
  • Received:2004-04-28 Revised:2004-05-28 Online:2004-09-25 Published:2004-09-25

摘要: 开发了一种基于"块-单元"数据结构的可扩展并行算法,以实现大规模、非均匀分子动力学模拟.它采用空间填充曲线将三维区域分解转换为一维负载平衡问题,然后用基于实测的多层均权法求解,以保持处理机间负载均衡.在一个MPP并行机的500个CPU上,模拟包含2.1×108个粒子的三维金属微喷射模型,该算法获得了420倍的加速比.

关键词: 分子动力学, “块-单元”数据结构, 空间填充曲线, 多层均权法

Abstract: A scalable parallel algorithm especially for large-scale three dimensional simulations with seriously non-uniform molecular distributions is presented. In particular, based on cell-block data structures, this algorithm uses space filling curve to convert three-dimensional domain decomposition for load distribution across processors into one-dimensional load balancing problems for which run-time measurement-based multilevel averaging weights method can be applied successfully. Parallel numerical experiments for simulations of three-dimensional metal-eject physical models with 210 million particles using 500 processors of one massively parallel processors have shown that this algorithm has achieved speedup about 420.

Key words: molecular dynamics, cell-block data structure, space filling curve, multilevel averaging weights method

中图分类号: