EFFICIENT PARALLEL ALGORITHM FOR LARGE-SCALE MOLECULAR DYNAMICS SIMULATION IN MICROSCALE THERMOPHYSICS

摘要

Molecular dynamics (MD) simulation is an important research method in thermophysics. But it is difficult to implement the simulation with traditional serial algorithms because of a complex numerical calculation. In this paper, we propose algorithms based on a new force decomposition approach called Half Force-Block Decomposition (HFBD). The HFBD approach greatly reduces the memory usage and the communication cost, so it can be more easily to simulate a large-scale particle system. Furthermore, we propose two new strategies to maintain load balance, which is the main problem when parallel algorithms based on force decomposition apply to short-range MD simulation. The first technique which we called Random Redistribution strategy (RRD), randomly permuting the particle ordering when load is imbalanced and the other approach, called Optimal Redistribution strategy (ORD), makes a simple load-balance calculation based on the computing time of all processors, and achieves the optimal particle ordering. The parallel algorithm based on the above approaches was tested on a system of 4,000,000 particles and implemented on an SMP-cluster and achieved an efficiency of 67.2% on 120 processors. The numerical results show that the proposed parallel algorithm can simulate a thermophysical system with more particles than before efficiently.