Improved neighbor list algorithm in molecular simulations using cell decomposition and data sorting method
Improved neighbor list algorithm in molecular simulations using cell decomposition and data sorting method
复制标题
DOI：
10.1016/j.cpc.2004.04.004
复制
发表时间：
2003-11
期刊：
影响因子：
--
通讯作者：
Z. Yao;Jian-Sheng Wang;Gui-rong Liu;Min Cheng
中科院分区：
文献类型：
--
作者：
Z. Yao;Jian-Sheng Wang;Gui-rong Liu;Min Cheng
为了减少分子模拟中不必要的原子间距离计算，提出了一种改进的近邻列表算法。该算法结合了Verlet表和单元链表算法的优点，采用单元分解的方法加快邻居表的构建速度，采用数据排序的方法降低CPU数据缓存的丢失率，采用部分更新的方法减少不必要的邻居表重构。利用该算法对分子动力学模拟的串行和并行性能进行了评价，并与传统的Verlet表和单元链表算法进行了比较。结果表明，无论是在原子数较少的情况下，还是在原子数较多的情况下，新算法的性能都是传统算法的2/3。
An improved neighbor list algorithm is proposed to reduce unnecessary interatomic distance calculations in molecular simulations. It combines the advantages of Verlet table and cell linked list algorithms by using cell decomposition approach to accelerate the neighbor list construction speed, and data sorting method to lower the CPU data cache miss rate, as well as partial updating method to minimize the unnecessary reconstruction of the neighbor list. Both serial and parallel performance of molecular dynamics simulation are evaluated using the proposed algorithm and compared with those using conventional Verlet table and cell linked list algorithms. Results show that the new algorithm outperforms the conventional algorithms by a factor of 2∼3 in cases of both small and large number of atoms.