基于OpenMP的Multi-Critical分子动力学并行算法优化

Optimization of parallel algorithm for molecular dynamics based on OpenMP and Multi-Critical

为提高分子动力学模拟在多核共享内存式服务器上的运算速度,在现有的分子动力学并行算法基础上提出了Multi-Critical算法。该算法使用手动划分力矩阵的方法,使多个线程进入不同名的临界区,并使用分块叠加的方法优化了并行算法,提高了并行效率。实验结果表明,对比之前的Critical算法,该算法的加速比和并行效率均有较大幅度的提高。

In order to improve the computing speed of molecular dynamics simulation on mutil-core shared memory server,this paper proposed Multi-Critical algorithm based on existing parallel algorithm.This algorithm used manual division force matrix,to make mutilple threads enter different critical section with different name,and the method of the superposition block optimized parallel algorithm,improved the parallel efficiency.The experimental results show that,compared to the previous critical algorithm,proposed algorithm’s speedup and parallel efficiency has a relatively large increase.