首页 | 本学科首页   官方微博 | 高级检索  
     

NECP-X中角度与特征线并行的研究与实现
引用本文:马党伟,刘宙宇,赵晨,陈军,曹良志,吴宏春,唐霄,柴晓明. NECP-X中角度与特征线并行的研究与实现[J]. 原子能科学技术, 2018, 52(3): 405-411. DOI: 10.7538/yzk.2017.youxian.0362
作者姓名:马党伟  刘宙宇  赵晨  陈军  曹良志  吴宏春  唐霄  柴晓明
作者单位:1.西安交通大学 核科学与技术学院,陕西 西安710049;2.中国核动力研究设计院 核反应堆系统设计技术重点实验室,四川 成都610041
摘    要:特征线方法在应用于全堆芯三维输运计算时面临着计算时间长、内存需求量大的问题,而大规模并行是最有效的解决办法。传统的并行策略是进行空间的区域分解,但当问题的几何规模较小时,其并行度有限,无法充分利用并行资源。本文在高保真物理计算程序NECP-X空间区域分解的基础上研究了角度和特征线的三重并行计算。为实现角度并行的负载平衡,采用了考虑权重的贪婪算法角度并行策略;为节省内存,在共享内存的并行模式下采用动态调度的分配方案实现特征线并行。数值结果表明,NECP-X中的角度和特征线并行效率较高,可在空间区域分解并行的基础上进一步扩大并行规模,提高计算速度。

关 键 词:特征线方法  角度并行  贪婪算法  负载平衡  特征线并行  多重并行

Development and Application of Angle and Ray Parallel in NECP-X
MA Dangwei,LIU Zhouyu,ZHAO Chen,CHEN Jun,CAO Liangzhi,WU Hongchun,TANG Xiao,CHAI Xiaoming. Development and Application of Angle and Ray Parallel in NECP-X[J]. Atomic Energy Science and Technology, 2018, 52(3): 405-411. DOI: 10.7538/yzk.2017.youxian.0362
Authors:MA Dangwei  LIU Zhouyu  ZHAO Chen  CHEN Jun  CAO Liangzhi  WU Hongchun  TANG Xiao  CHAI Xiaoming
Affiliation:1.School of Nuclear Science and Technology, Xi’an Jiaotong University, Xi’an 710049, China; 2.Science and Technology on Reactor System Design Technology Laboratory,Nuclear Power Institute of China, Chengdu 610041, China
Abstract:Long computing time and abundant memory requirement are two most critical issues for the method of characteristics (MOC) when applied for the three-dimensional whole-core transport calculation. The massive parallel is an effective method to solve these problems. The traditional parallel strategy based on space decomposition is limited by its degree of parallelism when solving the smaller size problem, which means it can not take full advantage of parallel resources. In this paper, a multilevel parallel of angle and characteristic ray was developed based on the space parallel implemented in the high-fidelity neutronics code NECP-X. To get good load balance of angle parallel, a greedy algorithm was applied, and to save memory in the ray parallel, a dynamic scheduling feature within shared memory architectures was used. The numerical results show that the parallel efficiency of the angle and characteristic ray in NECP-X is high. And it further expands the parallel degree based on the special decomposition parallelism.
Keywords:method of characteristics  angle parallel  greedy algorithm  load balance  characteristic ray parallel  multilevel parallel
本文献已被 CNKI 等数据库收录!
点击此处可从《原子能科学技术》浏览原始摘要信息
点击此处可从《原子能科学技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号