基于申威异构众核处理器架构的模态并行算法

摘要
图/表
参考文献(18)
相关文章 (12)

全文: PDF (2602 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要根据国产申威异构众核处理器架构特点，提出了一种结构有限元模态分层通信并行计算方法，对于提高国产申威异构众核分布式存储并行计算机下重大装备系统级模态分析的并行效率具有重要意义。该方法在分层通信策略和加速子空间迭代法的基础上构建了大规模模态分析并行计算体系，不仅实现了计算过程和数据通信的分层，有效提高了通信效率；而且实现了计算数据的分布式存储，显著改善了数据访存效率。将并行计算体系应用于某超深钻机制动系统主体结构和某跨江隧道，实现了上千万自由度、数万核的模态分析并行计算；结合算例对该算法的正确性和有效性进行了评估。研究结构结果表明：该算法能够充分利用国产申威异构众核分布式存储并行计算机的体系结构特点提高重大装备系统级模态并行计算效率。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	喻高远1
	2
	马志强1
	2
	3
	李俊杰1
	2
	金先龙1
	2

关键词 ：异构众核 , 分布式存储 , 分层通信大规模模态分析 , 并行计算

Abstract：According to the architecture characteristics of domestic heterogeneous multi-core processor, a hierarchical communication parallel computing algorithm for structural modal finite element analysis is proposed, which has important significance to improve the parallel efficiency of the system modal analysis on the entire large structure under the domestic heterogeneous multi-core and distributed memory parallel computers. Based on hierarchical communication and accelerating subspace iteration, a parallel computing system for a large-scale modal analysis was established, which can not only significantly improve communication rate through the hierarchical computing and communication, but also improve data access and storage rate through the distributed storage of a large amount of data. As a typical application, it was used to get solutions to the main structure of a certain brake system on ultra-deep drilling rig and the over-river tunnel, a parallel modal analysis with over ten-million-DOF was performed and ten thousands of core processors were applied. Then, the correctness and efficiency of the proposed method were validated. The results showed that the proposed parallel solving system can significantly improve the parallel efficiency of the modal analysis on the major equipment system with the domestic heterogeneous multi-core and distributed memory parallel computers.

Key words： Heterogeneous multi-core Distributed memory parallel computer Hierarchical communication Large-scale modal analysis Parallel computation

收稿日期: 2020-08-27 出版日期: 2022-02-15

引用本文:

喻高远1,2,马志强1,2,3,李俊杰1,2,金先龙1,2. 基于申威异构众核处理器架构的模态并行算法[J]. 振动与冲击, 2022, 41(3): 224-230.
YU Gaoyuan1,2, MA Zhiqiang1,2,3, LI Junjie1,2, JIN Xianlong1,2. Modal parallel algorithm based on Shenwei heterogeneous multi-core processor architecture. JOURNAL OF VIBRATION AND SHOCK, 2022, 41(3): 224-230.

链接本文:

http://jvs.sjtu.edu.cn/CN/ 或 http://jvs.sjtu.edu.cn/CN/Y2022/V41/I3/224

[1] 范宣华,肖世富,陈璞.千万自由度量级有限元模态分析并行计算研究[J].振动与冲击,2015,34(17):77-82.
FAN Xuanhua, XIAO Shifu, CHEN Pu. Parallel computing for finite element modal analysis with over ten-million-DOF[J]. Journal of Vibration and Shock, 2015, 34(17)：77-82.
[2] 朱彬,张宜生,王梁,田晓薇.基于GPU的模态分析并行算法[J].华中科技大学学报(自然科学版),2012,40(05):33-36.
ZHU Bin,ZHANG Yisheng, WANG Liang,Tian Xiaowei. Parallel algorithms for modal analysis based on GPU[J]. Journal of Huazhong University of Science and Technology.Nature Science, 2012, 40(5)：33-36.
[3] 何王全,刘勇,方燕飞,魏迪,漆锋滨.面向国产异构众核系统的Parallel C 语言设计与实现[J]. 软件学报,2017,28(4): 764−785.
HE Wanquan, LIU Yong, FANG Yanfei, Wei Di, Qi Fengbin. Design and implementation of Parallel C programming language for domestic heterogeneous many-core systems[J]. Ruan Jian Xue Bao/Journal of Software, 2017,28(4):764-785.
[4] Seid Koric, Qiyue Lu, Erman Guleryuz. Evaluation of massively parallel linear sparse solvers on unstructured finite element meshes[J]. Computers and Structures, 2014,141:19-25.
[5] Jesús Martínez-Frutos, Pedro J.Martínez-Castejón, David Herrero-Pérez. Fine-Grained GPU implementation of assembly-free iterative solver for finite element problems[J]. Computers and Structures, 2015,157:9-18.
[6] 杨梅芳,车永刚,高翔.基于OpenMP 4.0的发动机燃烧模拟软件异构并行优化[J]. 计算机研究与发展,2018,55(2): 400−408.
YANG Meifang, CHE Yonggang, Gao Xiang. Heterogeneous parallel optimazation of an engine combustion simulation application with the OpenMP 4.0 standard[J]. Journal of Computer Research and Development, 2018, 55(2):400-408. DOI:10.7544/issn 1000-1239.2018.20160872
[7] B.C.P. Heng, R.I. Mackie. Parallel modal analysis with concurrent distributed objects[J]. Computers and Structures, 2010,88:1444-1458.
[8] 刘芳芳,杨超,袁欣辉,吴长茂,敖玉龙.面向国产申威26010众核处理器的SpMV实现与优化[J]. 软件学报,2018,29(12): 3921−3932.
Liu Fangfang, Yang Chao, Yuan Xinhui, Wu Changmao, Ao Yulong. General SpMV implementation in many-core domestic sunway 26010 processor[J]. Ruan Jian Xue Bao/Journal of Software,2018,29(12):3921-3932.
[9] Watkins D S. The matrix eigenvalue problem: GR and krylov subspace methods[M]. SIAM, Philadelphia, 2007.
[10] Sorensen D C. Implicit application of polynomial filters in a k-step Arnoldi method[J]. SIAM Journal on Matrix Analysis and Applications, 1992,13:357-385.
[11] Stewart G W. A Krylov-Schur algorithm for large eigenproblems[J]. SIAM Journal on Matrix Analysis and Applications, 2001, 23(3):601-614.
[12] Zhongxiao Jia*. A refined subspace iteration algorithm for large sparse eigenproblems[J]. Applied Numerical Mathematics, 2000, 32:35-52.
[13] ZAIWEN WEN, YIN ZHANG. Accelerating convergence by augmented rayleigh-ritz projections for large-scale eigenpair computation[J]. Siam Journal on matrix analysis and applications, 2017, 38(2):273-296.
[14] PING TAK PETER TANG, ERIC POLIZZI. Feast as a subspace iteration eigensolver accelerated by approximate spectral projection[J]. Siam Journal on matrix analysis and applications, 2014, 35(2):354-390.
[15] R.G. Grimes, J.G. Lewis, and H.D.Simon. A shifted Block lanczos algorithm for solving sparse symmetric generalized eigenproblems[J]. SIAM Journal on Matrix Analysis and Applications, 1996,15(1):228-272.
[16] 范宣华.基于Panda框架的大规模有限元模态分析并行计算机应用[D]. 北京：北京大学，2013.
[17] 苗新强,金先龙,丁峻宏.结构静力有限元分层并行计算方法.力学学报,2014,46(4):611-618.
Miao XQ, Jin XL, Ding JH. A hierarchical parallel computing approach for structural static finite element analysis. Acta Mechanica Sinica, 2014,46(4):611-618.
[18] Hernandez V , Roman J E , Vidal V . SLEPc: A scalable and flexible toolkit for the solution of eigenvalue problems[J]. ACM Transactions on Mathematical Software, 2005, 31(3):351-362.