基于空位跃迁的KMC并行实现  

A PARELLEL IMPLEMENTATION OF KINETIC MONTE CARLO SIMULATION FOR VACANCY TRANSITION

作　　者：宋梦召 冯仰德[1] 聂宁明[1] 王武[1] Song Mengzhao Feng Yangde Nie Ningming Wang Wu(Department of High Performance Computing Technology and Application Development, Computer Network Information Center, CAS, Beijing 100190, China University of Chinese Academy of Sciences, Beijing 100049, Chin)

机构地区：[1]中国科学院计算机网络信息中心,高性能计算技术与应用发展部,北京100190 [2]中国科学院大学,北京100049

出　　处：《数值计算与计算机应用》2017年第2期130-142,共13页Journal on Numerical Methods and Computer Applications

基　　金：863项目(2015AA01A303);自然科学基金项目(91430214)资助

摘　　要：动力学蒙特卡洛方法是一种随机方法,它将系统的演化过程抽象为状态到状态之间的跃迁,这种方法被广泛地应用到各种材料的演化模拟中.本文采用同步子域方法实现了空位跃迁过程的MPI并行模拟,通过自适应同步时间步长方法,在保证准确性的情况下获得较好并行性能.设计动态添加通信数据的通信方式,有效降低通信开销.数值实验表明,采用同步子域的串、并行方法同原始串行算法结果一致.同时算法在大规模问题的模拟上也表现出很好的并行效率.Kinetic Monte Carlo （KMC） simulation is a stochastic method that characterizes system evolution by occasional transitions from one state to another. It is widely used to simulate the evaluation of many physical materials. This paper presents a parallel implementation of vacancy transition based on Synchronous Sublattice Algorithm using MPI. We adapt an adaptive time synchronization method so as to achieve both accuracy and parallel efficiency. A new communication mechanism which dynamically adds communication data to the send queues is designed to decrease communication load. The computer simulation turns that the parallel KMC based on SL algorithm achieves consistent results with the serial implementation, and it shows great parallel efficiency on the large-scale numerical experiments.

关 键 词：空位跃迁 KMC 并行计算 同步子域算法 

分 类 号：O242.2[理学—计算数学]