计算机科学 ›› 2017, Vol. 44 ›› Issue (4): 197-201.doi: 10.11896/j.issn.1002-137X.2017.04.043
王明清,李明,张清,张广勇,吴韶华
WANG Ming-qing, LI Ming, ZHANG Qing, ZHANG Guang-yong and WU Shao-hua
摘要: 广义极小残量法(GMRES)是最常用的求解非对称大规模稀疏线性方程组的方法之一,其收敛速度快且稳定性良好。Intel Xeon Phi众核协处理器(MIC)具有计算能力强、易编程、易移植等特点。采用MPI+OpenMP+offload混合编程模型将GMRES算法移植到MIC集群平台上。采用进程间集合通信异步隐藏、数据传输优化、向量化以及线程亲和性优化等多种手段,大幅提升了GMRES算法的求解效率。最后将并行算法应用到“局部径向基函数求解高维偏微分方程”问题的求解中。测试表明,CPU节点集群上开启32个进程,并行效率高达71.74%,4块MIC卡的最高加速性能可达单颗CPU的7倍。
[1] SAAD Y,SCHULTZ M H.GMRES:a generalized minimal residual algorithm for solving nonsymmetric linear systems[J].Siam Journal on Scientific & Statistical Computing,1986,7(3):856-869. [2] SAAD Y.Iterative methods for sparse linear systems(2nd ed)[M].Philadelphia:SIAM,2003:151-126. [3] NACHTIGAL N M,REICHEL L, TREFETHENK L NL N.A hybrid GMRES algorithm for nonsymmetric linear systems[J].Siam Journal on Scientific & Statistical Computing,1992,13(3):796-825. [4] QUAN Z,XIANG S H.A GMRES based polynomal preconditioning algorithm[J].Mathematical Numerical Sinica,2006,28(4):365-376.(in Chinese).全忠,向淑晃.基于GMRES的多项式预处理广义极小残差法[J].计算数学,2006,28(4):365-376. [5] GHAEMIAN N,ABDOLLAHZADEH A,HEINEMANN Z.Accelerating the GMRES Iterative Linear Solver of an Oil Re-servoir Simulator using the Multi-Processing Power of Compute Unified Device Architecture of Graphics Cards[C]∥Proceedings of the 9th International Workshop on State of the Art in Scientific and Parallel Computing.Heidelberg:Springer,2008:156-159. [6] WANG M L,KLIE H,PARASHAR M,et al.Solving Sparse Linear Systems on NVIDIA Tesla GPUs[C]∥Computational Science-ICCS 2009 Lecture Notes in Computer Science.2009,5544:864-873. [7] LIU Y Q,YIN K X,WU E H.Fast GMRES-GPU Solver Large Scale Sparse Linear Systems[J].Journal of Computer-Aided Design & Computer Graphics,2011,23(4):553-560.(in Chinese) 柳有权,尹康学,吴恩华.大规模稀疏线性方程组的GMRES-GPU快速求解算法[J].计算机辅助设计与图形学学报,2011,23(4):553-560. [8] GHYSELS P,ASHBY T J,MEERBERGEN K,et al.Hidingglobal communication latency in the GMRES algorithm on massively parallel computers[J].Siam Journal on Scientific Computing,2013,35(1):48-71. [9] 王恩东,张清,等.MIC高性能计算编程指南[M].北京:中国水利水电出版社,2012. [10] JEFFERS J,REINDERS J.Intel Xeon Phi Coprocessor HighPerformance Programming[R].morgan kaufmann.2013. [11] LI M,CHEN W,CHEN C S.The localized RBFs collocationmethods for solving high dimensional PDEs[J].Engineering Analysis with Boundary Elements,2013,37(10):1300-1304. [12] BELLALIJ M,REICHEL L,SADOK H.Some properties ofrange restricted GMRES methods[J].Journal of Computational &Applied Mathematics,2015,290:310-318. [13] HE K,TAN X D,ZHAO H Y,et al.Parallel GMRES solver for fast analysis of large linear dynamic systems on GPU platforms[J].Integration the VLSI Journal,2016,52(c):10-22. |
No related articles found! |
|