详细介绍了单晶硅磨削过程分子动力学并行化涉及到的并行算法设计基础等基础理论,分析了现有的几种并行算法,确定采用区域分解法作为本文的并行算法,在此基础上提出了基于区域二次划分的分子动力学并行仿真算法。介绍了分子动力学并行仿真计算的软硬件环境,设计了分子动力学并行仿真程序,在联想深腾1800机群系统上分别应用2、3、4台结点机上进行仿真实验,运行结果表明:与串行程序仿真结果在瞬间位置图和总能量变化方面相似,证明并行程序的结果是可靠的。加速比随着结点数的增加而增加,并行效率所略有下降但都在87.5%以上,并行效率并没有随着结点的增加有明显的降低,说明并行程序具有很好的扩展性。
The basic theories in the MD parallel simulation of monocrystal silicon grinding was described including the design methodology of parallel algorithm, etc. Then some popular parallel algorithms were analyzed and the domain decomposition was selected for the parallel algorithm of MD simulation. Based on the domain decomposition, a twice domain partition algorithm was proposed for MD parallel simulation. Moreover, the hardware and software environments of the MD parallel simulation were introduced. Lastly, MD parallel program was made and executed on 2, 3, 4 nod computers respectively. Through the comparison of the results between the serial and the parallels-gimulations regarding the instantaneous distribution of atoms and the total energy of the system, the results of the parallel simulation are reliable. In addition, it is shown that when the number of process increases, the accelerating ratio increases and the efficiency decreases a little but still more than 87.5%. That means the efficiency doesn't reduce obviously with the increase of the process and the parallel program has good expansibility.