• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

核辐照损伤金属材料的大规模KMC模拟

宋梦召,冯仰德   

  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)
  • 收稿日期:2017-01-11 修回日期:2017-03-16 出版日期:2017-07-25 发布日期:2017-07-25
  • 基金资助:

    国家863计划(2015AA01A303)

Large-scale kinetic Monte Carlo simulation
 of metal material under irradiation
 

SONG Meng-zhao,FENG Yang-de   

  1. (Department of High Performance Computing Technology and Application Development,
    Computer Network Information Center,Chinese Academy of Sciences,Beijing 100190,China)

     
  • Received:2017-01-11 Revised:2017-03-16 Online:2017-07-25 Published:2017-07-25

摘要:

动力学蒙特卡洛方法可用来模拟核反应堆第一壁材料的辐射效应和缺陷扩散,有助于理解和预测材料在辐照损伤下的微观性质和宏观变化。采用同步子域方法实现了空位跃迁过程的并行模拟。通过采用动态更新通信数据和自适应同步时间步长方法,减少通信次数和通信量,在保证准确性的情况下获得了较好并行性能。实验表明,基于同步子域的串行计算比原始串行算法时间缩短60.31%,并行算法在80核时达到39倍加速比。对于大规模问题,算法也表现出很好的并行效率,适用于大规模问题的模拟。
 

关键词: KMC, 空位跃迁, SPPARKS, MPI, 同步子域算法

Abstract:

Kinetic Monte Carlo (KMC) simulation can be used to simulate the radiation effects and defect diffusion of reactor materials in nuclear power plants, which is helpful to understand and predict how microstructures and mechanical properties evolve under irradiation. We present a parallel simulation method of vacancy diffusion based on synchronous sub-lattice (SL) algorithm. We adopt a dynamic communication data update method and an adaptive time synchronization method to decrease communication frequency and load so as to achieve both accuracy and parallel efficiency. Simulation results show that the serial implementation by the SL algorithm saves the total execution time by 60.31% in comparison with the original algorithm, and the parallel implementation achieves 39X speedup with 80 cores. The algorithm also performs well on large scale problems.
 

Key words: KMC, vacancy diffusion, SPPARKS, MPI, synchronous sub-lattice algorithm