数值计算与计算机应用
       首页 |  期刊介绍 |  编委会 |  投稿指南 |  期刊订阅 |  下载中心 |  联系我们 |  在线办公 | 
数值计算与计算机应用  2017, Vol. 38 Issue (3): 197-214    DOI:
论文 最新目录 | 下期目录 | 过刊浏览 | 高级检索 Previous Articles  |  Next Articles  
基于CPU-MIC异构众核环境的行星流体动力学数值模拟
吴长茂1, 杨超1,2, 尹亮3, 刘芳芳1, 孙乔1, 李力刚4
1. 中国科学院软件研究所并行软件与计算科学实验室, 北京 100190;
2. 中国科学院软件研究所计算机科学国家重点实验室, 北京 100190;
3. 中国科学院大学工程科学学院, 北京 100049;
4. 中国科学院上海天文台, 上海 200030
NUMERICAL SIMULATION OF PLANETARY FLUID DYNAMICS ON CPU-MIC HETEROGENEOUS MANY-CORE SYSTEMS
Wu Changmao1, Yang Chao1,2, Yin Liang3, Liu Fangfang1, Sun Qiao1, Li Ligang4
1. Laboratory of Parallel Software and Computational Science, Institute of Software, Chinese Academy of Sciences, Beijing 100190, China;
2. State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing 100190, China;
3. School of Engineering Science, University of Chinese Academy of Sciences, Beijing 100049, China;
4. Shanghai Astronomical Observatory, Chinese Academy of Sciences, Shanghai 200030, China
 全文: PDF (850 KB)   HTML (1 KB)   输出: BibTeX | EndNote (RIS)      背景资料
摘要 数值模拟是行星流体动力学研究的主要工具.本文介绍CPU-MIC异构众核平台的行星流体动力学数值模拟,计算并模拟地球外核的磁流体运动.本文在已有工作的基础上[1-3],添加了CPU-MIC异构众核环境的数值模拟支持.首先描述了CPU-MIC异构众核环境的上的数值模拟流程,然后给出了MIC上的分布式并行GMRES(m)众核解法器的实现算法.其次,实现了解法器的计算核心稀疏矩阵向量乘(SpMV)在MIC上的分布式并行算法,该SpMV实现了计算-通信重叠、数据传输-计算重叠.再次,为加速行星流体动力学方程收敛,给出了MIC上以SpMV为基本操作的分布式并行多项式预条件子.最后,提出了一些MIC众核平台的优化措施,如多线程、流存储和数据传输优化等.天河2号数值模拟表明相比CPU版的数值模拟, CPU-MIC异构众核环境下数值模拟在单MIC卡和64块MIC卡分别取得了6.93和6.0倍的加速比.
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
关键词行星流体动力学   解法器   并行计算   众核加速   GMRES   性能优化     
Abstract: Massively parallel computing is becoming a primary tool for the numerical simulation of planetary fluid dynamics. In this paper, Numerical simulation of the planetary fluid dynamics for distributed memory Xeon Phi-accelerated systems is studied. Firstly, we start from a legacy parallel code[1-3] using PETSc software package, which employs a pure MPI approach for parallel computing, to date, is in lack of support for multi-threaded parallelism on manycore accelerated systems, and then we extend the legacy code to multi-threaded parallelism on Xeon Phi-accelerated systems. Furthermore, based on PETSc software package, a sparse linear solver for Xeon Phi-accelerated cluster, which utilizes restarted generalized residual method(GMRES(m)), is presented and optimized. Secondly, a novel sparse matrix-vector multiplication(SpMV) algorithm for Xeon Phi-accelerated cluster is proposed, it combines highly aggressive use of asynchrony with offload, compute, communication, all of which serve the overlap of computation and communication. What's more, based on our SpMV algorithm, a polynomial preconditioner is given, which mainly consists of SpMV operations, hide and reduce communication, whether to local memory, across the network, or over PCIe. Finally, some optimized measures are taken to the extended code. Experiments on Tianhe- 2 Supercomputer show that as compared to the original code, our Xeon Phi-accelerated design is able to deliver 6.93x and 6.00x speedups for single MIC device and 64 MIC devices, respectively.
Key wordsplanetary fluid dynamics   solver   parallel computing   Xeon Phi acceleration   GMRES   performance optimization   
收稿日期: 2016-10-25;
基金资助:

国家高技术研究发展计划(863计划)(2015AA01A302);国家自然科学基金重大研究计划集成项目(91530323);国家重点研发计划(2016YFB0200603);中国科学院信息化专项资助项目(XXH-12503-02-02-2);中国科学院前沿科学重点研究项目(QYZDB-SSW-SYS006).

引用本文:   
. 基于CPU-MIC异构众核环境的行星流体动力学数值模拟[J]. 数值计算与计算机应用, 2017, 38(3): 197-214.
. NUMERICAL SIMULATION OF PLANETARY FLUID DYNAMICS ON CPU-MIC HETEROGENEOUS MANY-CORE SYSTEMS[J]. Journal of Numerical Methods and Computer Applicat, 2017, 38(3): 197-214.
 
[1] Chan K H, Li Ligang and Liao Xinhao. Modelling the core convection using finite element and finite difference methods[J]. Physics of the Earth and Planetary Interiors, 2006, 157(1-2):124-138.
[2] Yang Chao, Zhang Yunquan and Li Ligang. Numerical Simulation of the Thermal Convection in the Earth's Outer Core[C]. 12th IEEE International Conference on High Performance Computing and Communications, HPCC. 2010, 552-555.
[3] Yang Chao, Li Ligang and Zhang Yunquan. Development of a Scalable Solver for theEarth's Outer Core[C]. High Performance Computing and Applications, Second International Conference, HPCA 2009, Shanghai, China, August 10-12, 2009, Revised Selected Papers
[4] Satish Balay and Shrirang Abhyankar and Mark F. Adams and Jed Brown and Peter Brune and Kris Buschelman and Lisandro Dalcin and Victor Eijkhout and William D. Gropp and Dinesh Kaushik and Matthew G. Knepley and Lois Curfman McInnes and Karl Rupp and Barry F. Smith and Stefano Zampini and Hong Zhang and Hong Zhang. PETSc Web page[OL]. http://www.mcs.anl.gov/petsc, 2016.
[5] Zhang, Keke and Schubert, Gerald. Magnetohydrodynamics in Rapidly Rotating spherical Systems[J]. Annual Review of Fluid Mechanics. 2000, 32(1):409-443.
[6] Gilman P A. Dynamically consistent nonlinear dynamos driven by convection in a rotating spherical shell[J]. Astrophysical Journal Supplement. 1983, 53(53):243-268.
[7] Zhang K K and Busse F H. Convection driven magnetohydrodynamic dynamos in rotating spherical shells[J]. Geophysical and Astrophysical Fluid Dynamics. 1989, 49(1):97-116.
[8] Glatzmaier G A, Roberts P H. A three-dimensional convective dynamo solution with rotating and finitely conducting inner core and mantle[J]. Physics of the Earth and Planetary Interiors, 1995, 91(1-3):63-75.
[9] Kageyama A, Sato T. Generation mechanism of a dipole field by a magnetohydrodynamic dynamo[J]. Physical Review E Statistical Physics Plasmas Fluids and Related Interdisciplinary Topics, 1997, 55(4):4617-4626.
[10] Moritz Heimpel and Jonathan Aurnou. Turbulent convection in rapidly rotating spherical shells:A model for equatorial and high latitude jets on Jupiter and Saturn[J]. 2007, 187(2):540-557.
[11] Liao X, Feng T, Zhang K. On the Saturation and Temporal Variation of Mean Zonal Flows:An Implication for Equatorial Jets on Giant Planets[J]. Astrophysical Journal, 2008, 666(1).
[12] Heimpel M, Aurnou J, Wicht J. Simulation of equatorial and high-latitude jets on Jupiter in a deep convection model.[J]. Nature, 2005, 438(7065):193-6.
[13] Showman A P, Kaspi Y, Flierl G R. Scaling laws for convection and jet speeds in the giant planets[J]. Icarus, 2010, 211(2):1258-1273.
[14] Erich Strohmaier and Horst Simon and Hans Meuer and Jack Dongarra. TOP500 Web page[OL]. http://www.top500.org/, 2016.
[15] Chandrasekhar S. Hydrodynamic and Hydromagnetic Stability[M]. Dover edition. New York:Dover Publications, Inc., 1981.
[16] Christensen U R, Aubert J, Cardin P, Dormy E, Gibbons S, Glatzmaier G A, Grote E, Honkura Y, Jones C, Kono M, Matsushima M, Sakuraba A, Takahashi F, Tilgner A, Wicht J, Zhang K. A numerical dynamo benchmark[J]. Physics of the Earth and Planetary Interiors. 2001, 128(1-4):25-34.
[17] Dukowicz J K, Dvinsky A S. Approximate factorisation as a high order splitting for the incompressible flow equations[J]. Journal of Computational Physics, 1992, 102(2):336-347.
[18] Saad Y, Schultz M H. GMRES:a generalized minimal residual algorithm for solving nonsymmetric linear systems[J]. SIAM Journal on Scientific and Statistical Computing, 2006, 7(3):856-869.
[19] Yousef Saad. Iterative Methods for Sparse Linear Systems[M]. Society for Industrial and Applied Mathematics, 2003.
[20] Philippe Tillet and Karl Rupp and Siegfried Selberherr and Chin-Teng Lin. Towards PerformancePortable, Scalable, and Convenient Linear Algebra[C]. Presented as part of the 5th USENIX Workshop on Hot Topics in Parallelism, 2013, Berkeley, CA.
[21] CUSP library[OL]. https://github.com/cusplibrary, 2016.
[22] PARALUTION-the library for iterative sparse methods on CPU and GPU[OL]. http://www.paralution.com/, 2016.
[23] Minden, Victor and Smith, Barry and Knepley, Matthew G. Preliminary Implementation of PETSc Using GPUs[M]. Springer Berlin Heidelberg, 2013.
[24] Baker C G and Heroux, M A. Tpetra, and the Use of Generic Programming in Scientific Computing[J]. Sci. Program, 2012, 20(2):115-128.
[25] Yamazaki I, Anzt H, Tomov S, et al. Improving the Performance of CA-GMRES on Multicores with Multiple GPUs[C]. IEEE International Parallel and Distributed Processing Symposium, IPDPS. 2014, 382-391.
[26] Cramer T, Schmidl D, Klemm M, et al. OpenMP Programming on Intel Xeon Phi Coprocessors:An Early Performance Comparison[J]. 2012.
[27] Liu X, Smelyanskiy M, Chow E, et al. Efficient sparse matrix-vector multiplication on x86-based many-core processors[C]. International ACM Conference on International Conference on Supercomputing. 2013:273-282.
[28] Liu W, Vinter B. CSR5:An Efficient Storage Format for Cross-Platform Sparse Matrix-Vector Multiplication[C]. ACM on International Conference on Supercomputing. ACM, 2015, 339-350.
[29] Yan S, Li C, Zhang Y, et al. yaSpMV:Yet Another SpMV Framework on GPUs[C]. ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP, 2014, 107-118.
[1] 宋梦召, 冯仰德, 聂宁明, 王武. 基于空位跃迁的KMC并行实现[J]. 数值计算与计算机应用, 2017, 38(2): 130-142.
[2] 许竞劼, 谢妍, 卢本卓. 一个基于PHG平台的并行有限元生物分子模拟解法器[J]. 数值计算与计算机应用, 2016, 37(1): 67-82.
[3] 刘辉, 冷伟, 崔涛. 并行自适应有限元计算中的负载平衡研究[J]. 数值计算与计算机应用, 2015, 36(3): 166-184.
[4] 刘妍, 田保林, 申卫东, 茅德康. MFCAV近似Riemann解法器在相容拉氏方法中的熵条件分析[J]. 数值计算与计算机应用, 2015, 37(3): 286-298.
[5] 刘辉, 崔涛, 冷伟. hp自适应有限元计算中一种新的自适应策略[J]. 数值计算与计算机应用, 2015, 36(2): 100-112.
[6] 刘皞, 李超. 多右端线性方程组的块种子投影方法[J]. 数值计算与计算机应用, 2014, 35(3): 221-228.
[7] 朱雪芳. 求解鞍点问题的一类广义SSOR预条件子[J]. 数值计算与计算机应用, 2014, 35(2): 117-124.
[8] 吴立飞, 杨晓忠, 张帆. 非线性Leland方程的一种并行本性差分方法[J]. 数值计算与计算机应用, 2014, 35(1): 69-80.
[9] 王倩, 戴华. 求解离散不适定问题的正则化GMERR方法[J]. 数值计算与计算机应用, 2013, 35(2): 195-204.
[10] 纪国良, 冯仰德. 大规模有限元刚度矩阵存储及其并行求解算法[J]. 数值计算与计算机应用, 2012, 33(3): 230-240.
[11] 段班祥, 朱小平, 张爱萍. 解线性互补问题的并行交替迭代算法[J]. 数值计算与计算机应用, 2011, 32(3): 183-195.
[12] 吴树林, 王志勇, 黄乘明. Parareal 算法的均方稳定性分析[J]. 数值计算与计算机应用, 2011, 33(2): 113-124.
[13] 王顺绪, 戴华. 二次特征值问题的并行Jacobi-Davidson方法及其应用[J]. 数值计算与计算机应用, 2008, 29(4): 313-320.
[14] 姜金荣,迟学斌,陆忠华,刘洪利. 在MM5中实现IKAWA参数自适应选择[J]. 数值计算与计算机应用, 2007, 28(3): 215-220.
[15] 李长峰,袁益让,. 三维两相渗流驱动问题迎风区域分裂显隐差分法[J]. 数值计算与计算机应用, 2007, 29(2): 113-136.
Copyright © 2008 数值计算与计算机应用 版权所有
中国科学院数学与系统科学研究院 《数值计算与计算机应用》编辑部
北京2719信箱 (100190) Email: szjs@lsec.cc.ac.cn
Support by Beijing Magtech Co.ltd   E-mail:support@magtech.com.cn
京ICP备05002806号-10