LARGE SCALE AND NON-UNIFORM SIMULATION OF PARALLEL MOLECULAR DYNAMICS
This paper presents a HSFC-based, measured-based dynamic load balancing strat-egy for a parallel molecular dynamics method. It reduces largely bookkeeping overhead by introducing block structure. Moreover, it redistributes blocks by monitoring change of total workload in single processor instead of monitoring change of workload in single block. Two tests have showed that it can be efficient enough to load fluctuations in each processor as the simulation progresses. Excellent scalabil-ity was demonstrated, with a speedup larger than 400 with 500 processors of one MPP, by simulating non-uniform systems including 210,000,000 particles in 3-D.
CAO XIAOLIN MO ZEYAO ZHANG JINGLIN
High performance Computing center,State Key Laboratory of Computational Physics,Institute of Applied Physics and Computational Mathematics,P.O. Box 8009, Beijing, 100088, P.R. China
国际会议
The Joint Conference of ICCP6 and CCP2003(第6届国际计算伦理会议)
北京
英文
316-320
2004-05-23（万方平台首次上网日期，不代表论文的发表时间）