中国物理B ›› 2014, Vol. 23 ›› Issue (2): 28903-028903.doi: 10.1088/1674-1056/23/2/028903

• INTERDISCIPLINARY PHYSICS AND RELATED AREAS OF SCIENCE AND TECHNOLOGY • 上一篇    下一篇

MDSLB：A new static load balancing method for parallel molecular dynamics simulations

武云龙, 徐新海, 杨学军, 邹顺, 任小广   

  1. State Key Laboratory of High Performance Computing, National University of Defense Technology, Changsha 410073, China
  • 收稿日期:2013-05-30 修回日期:2013-09-04 出版日期:2013-12-12 发布日期:2013-12-12
  • 基金资助:
    Project supported by the National Natural Science Foundation of China (Grant Nos. 61303071 and 61120106005) and the Natural Science Fund from the Guangzhou Science and Information Technology Bureau (Grant No. 134200026).

MDSLB：A new static load balancing method for parallel molecular dynamics simulations

Wu Yun-Long (武云龙), Xu Xin-Hai (徐新海), Yang Xue-Jun (杨学军), Zou Shun (邹顺), Ren Xiao-Guang (任小广)   

  1. State Key Laboratory of High Performance Computing, National University of Defense Technology, Changsha 410073, China
  • Received:2013-05-30 Revised:2013-09-04 Online:2013-12-12 Published:2013-12-12
  • Contact: Xu Xin-Hai E-mail:xuxinhai@nudt.edu.cn
  • About author:89.20.-a; 89.20.Ff; 02.70.Ns; 31.15.-p
  • Supported by:
    Project supported by the National Natural Science Foundation of China (Grant Nos. 61303071 and 61120106005) and the Natural Science Fund from the Guangzhou Science and Information Technology Bureau (Grant No. 134200026).

PDF (PC)

448

赞

摘要： Large-scale parallelization of molecular dynamics simulations is facing challenges which seriously affect the simulation efficiency, among which the load imbalance problem is the most critical. In this paper, we propose, a new molecular dynamics static load balancing method (MDSLB). By analyzing the characteristics of the short-range force of molecular dynamics programs running in parallel, we divide the short-range force into three kinds of force models, and then package the computations of each force model into many tiny computational units called “cell loads”, which provide the basic data structures for our load balancing method. In MDSLB, the spatial region is separated into sub-regions called “local domains”, and the cell loads of each local domain are allocated to every processor in turn. Compared with the dynamic load balancing method, MDSLB can guarantee load balance by executing the algorithm only once at program startup without migrating the loads dynamically. We implement MDSLB in OpenFOAM software and test it on TianHe-1A supercomputer with 16 to 512 processors. Experimental results show that MDSLB can save 34%–64% time for the load imbalanced cases.

关键词: molecular dynamics, static load balancing, parallel computing

Abstract: Large-scale parallelization of molecular dynamics simulations is facing challenges which seriously affect the simulation efficiency, among which the load imbalance problem is the most critical. In this paper, we propose, a new molecular dynamics static load balancing method (MDSLB). By analyzing the characteristics of the short-range force of molecular dynamics programs running in parallel, we divide the short-range force into three kinds of force models, and then package the computations of each force model into many tiny computational units called “cell loads”, which provide the basic data structures for our load balancing method. In MDSLB, the spatial region is separated into sub-regions called “local domains”, and the cell loads of each local domain are allocated to every processor in turn. Compared with the dynamic load balancing method, MDSLB can guarantee load balance by executing the algorithm only once at program startup without migrating the loads dynamically. We implement MDSLB in OpenFOAM software and test it on TianHe-1A supercomputer with 16 to 512 processors. Experimental results show that MDSLB can save 34%–64% time for the load imbalanced cases.

Key words: molecular dynamics, static load balancing, parallel computing

中图分类号:  (Interdisciplinary applications of physics)

  • 89.20.-a
89.20.Ff (Computer science and technology) 02.70.Ns (Molecular dynamics and particle methods) 31.15.-p (Calculations and mathematical techniques in atomic and molecular physics)