Parallelization And Performance Optimization Of Mesoscale Dissipative Particle Dynamics Simulation Program

Dissipative particle dynamics (DPD) is a simulation method used within the range from atomistic scale to mesoscopic scale. This method is widely used because of its effectivity in dealing with problems on the accumulation and dispersing of molecules. This simulation method can be applied not only to simulation systems with much more particles, but also to the study of longer-time motion behavior of particles in certain systems. However, the sequential simulation is not able to handle efficienty the situation that the simulation system is becoming larger and the systemic complexity is increasing.In this paper, the research background of molecule simulation and parallel computing is briefly introduced. And a comprehensive introduction of the basic theories and computing method of dissipative particle dynamics is provided. Then a parallel algorithm of the serial simulation program based on spatial domain decomposition is proposed. The algorithm is implemented using MPI and the simulation tests under various conditions are accomplished on a high performance cluster platform equipped with SGE (SUN Grid Engine). The results demonstrate that the parallel algorithm can not only cut the computing time sharping but also be more applicable to larger simulation systems.With the former research work mentioned above, the fundamental knowledge of wetting phenomena and the related theories are introduced. And an improved method of the traditional dissipative particle dynamics many-body dissipative particle dynamics (MDPD) simulation method is proposed. Taken the capillary wetting phenomena as the simulation instance, the fact that the MDPD method can simulate the capillary wetting phenomena much better is further discussed, which finally come to a conclusion that the spatial domain decomposition algorithm works well for the MDPD simulation program and serves a high simulation efficiency.