HPSEPS在元与神威·太湖之光上的移植和性能分析
|
摘要
"神威·太湖之光"高效能计算机系统是世界上首台峰值运算速度超过10亿亿次量级的超级计算机,HPSEPS (High Performance Symmetric Eigenproblem Solvers)是自主开发的大规模对称稠密矩阵特征问题并行求解器,包括标准对称稠密矩阵特征问题的并行计算方法,对大规模数据问题的计算,表现出较好的性能,本文分别在中科院的"元"超级计算机上和神威·太湖之光超级计算机上进行了移植,对比了两种超级计算机的系统性能,并且在"神威·太湖之光"上分别链接适合其异构众核结构的xMath数学库和mkl数学库,对求解器在链接两种不同数学库的计算机效果进行了测试与分析。
The Sunway TaihuLight supercomputer is the world's first system with a peak performance greater than 100 Pflops. High Performance Symmetric Eigenproblem Parallel Solver(HPSEPS) is a large scale symmetric eigenproblem parallel solvers which include parallel computing method of standard symmetric dense matrix eigenvalue problem, the calculation of large-scale data problem, it performance good, based at the Chinese Academy of Sciences "Yuan" and "Taihu supercomputer Shenwei light" super computer for trans-plant and performance test, at the same time we compare the performance of two kinds of supercomputers.In this paper,we also analysis and test HPSEPS on the Sunway TaihuLight about linking two kinds of mathematical library.
The Sunway TaihuLight supercomputer is the world's first system with a peak performance greater than 100 Pflops. High Performance Symmetric Eigenproblem Parallel Solver(HPSEPS) is a large scale symmetric eigenproblem parallel solvers which include parallel computing method of standard symmetric dense matrix eigenvalue problem, the calculation of large-scale data problem, it performance good, based at the Chinese Academy of Sciences "Yuan" and "Taihu supercomputer Shenwei light" super computer for trans-plant and performance test, at the same time we compare the performance of two kinds of supercomputers.In this paper,we also analysis and test HPSEPS on the Sunway TaihuLight about linking two kinds of mathematical library.
引文
[1]惠新忠.Xen虚拟I/O优化策略[硕士学位论文].大连:大连理工大学,2010.
[2]赵永华,迟学斌,程强,陈江,赵涛.对称矩阵三对角化的有效并行块算法设计[J].数值计算与计算机应用,2006(02):123-132.
[3]赵永华,迟学斌,王武.HPSEPS软件包及其千核应用[J].华中科技大学学报(自然科学版),2011,39(S1):59-62+75.
[4]廖文诗.矩阵特征值的估计及其展形的界作者[D].重庆大学,2013.
[5]杨广文,赵文来,丁楠,段芳.“神威·太湖之光”及其应用系统[J].科学,2017,69(03):12-16+2.[2017-09-01].
[6]姚文军,陈俊仕,苏志超,余洋,廖陈志,安虹.基于神威太湖之光的NAMD软件的移植与优化[J].计算机工程与科学,2017,39(06):1022-1030.[2017-09-01].
[7]高军晖,管哲人,刘钊,钟盈福,高竹,魏冬青.分子对接软件AutoDockVina在太湖之光操作系统上的移植[J].科技创新导报,2016,13(23):89-90.[2017-09-01].DOI:10.16660/j.cnki.1674-098X.2016.23.089
[8]龚晓军.OSPF的移植和实现[D].电子科技大学,2001.
[9]张立博,漆锋滨,卞卫峰,姜小成.一种适合众核MPI的托管式消息模型[J].计算机工程与科学,2009,31(S1):160-162+209.[2017-09-01].
[10]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/OL].软件学报,2017,28(04):764-785.(2017-02-20)[2017-09-01].
[11]陈一昭.并行计算在矩阵运算中的应用[D].昆明理工大学,2011.
[12]Haohuan FU,Junfeng LIAO,Jinzhe YANG,Lanning WANG,Zhenya SONG,Xiaomeng HUANG,Chao YANG,Wei XUE,Fangfang LIU,Fangli QIAO,Wei ZHAO,Xunqiang YIN,Chaofeng HOU,Chenglong ZHANG,Wei GE,Jian ZHANG,Yangang WANG,Chunbo ZHOU,Guangwen YANG.The Sunway Taihu Light supercomputer:system and applications[J].Science China(Information Sciences),2016,59(07):113-128.[2017-09-01].
[13]Jack Dongarra.Sunway Taihu Light supercomputer makes its appearance[J].National Science Review,2016,3(03):265-266.[2017-09-01].
[14]孟德龙,文敏华,韦建文,林新华.神威太湖之光上OpenFOAM的移植与优化[J].计算机科学,2017,44(10):64-70.
[15]赵伟,雷晓燕,陈德训,赵春,宋振亚.全球涡分辨率并行海洋模式POP在神威蓝光上的移植和应用[J].计算机应用与软件,2014,31(05):42-45.
[16]许瑾晨,郭绍忠,黄永忠,王磊.面向异构众核从核的数学函数库访存优化方法[J].计算机科学,2014,41(06):12-17.
[17]吴少刚,许解峰,杨耀忠,任钢.高性能BLAS在类Beowulf机群系统上的实现[J].小型微型计算机系统,2001(08):897-900.
[18]伍明川,黄磊,刘颖,何先波,冯晓兵.面向神威·太湖之光的国产异构众核处理器OpenCL编译系统[J/OL].计算机学报,2017:1-15[2018-01-.http://kns.cnki.net/kcms/detail/11.1826.TP.20171003.1550.002.html.
[2]赵永华,迟学斌,程强,陈江,赵涛.对称矩阵三对角化的有效并行块算法设计[J].数值计算与计算机应用,2006(02):123-132.
[3]赵永华,迟学斌,王武.HPSEPS软件包及其千核应用[J].华中科技大学学报(自然科学版),2011,39(S1):59-62+75.
[4]廖文诗.矩阵特征值的估计及其展形的界作者[D].重庆大学,2013.
[5]杨广文,赵文来,丁楠,段芳.“神威·太湖之光”及其应用系统[J].科学,2017,69(03):12-16+2.[2017-09-01].
[6]姚文军,陈俊仕,苏志超,余洋,廖陈志,安虹.基于神威太湖之光的NAMD软件的移植与优化[J].计算机工程与科学,2017,39(06):1022-1030.[2017-09-01].
[7]高军晖,管哲人,刘钊,钟盈福,高竹,魏冬青.分子对接软件AutoDockVina在太湖之光操作系统上的移植[J].科技创新导报,2016,13(23):89-90.[2017-09-01].DOI:10.16660/j.cnki.1674-098X.2016.23.089
[8]龚晓军.OSPF的移植和实现[D].电子科技大学,2001.
[9]张立博,漆锋滨,卞卫峰,姜小成.一种适合众核MPI的托管式消息模型[J].计算机工程与科学,2009,31(S1):160-162+209.[2017-09-01].
[10]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/OL].软件学报,2017,28(04):764-785.(2017-02-20)[2017-09-01].
[11]陈一昭.并行计算在矩阵运算中的应用[D].昆明理工大学,2011.
[12]Haohuan FU,Junfeng LIAO,Jinzhe YANG,Lanning WANG,Zhenya SONG,Xiaomeng HUANG,Chao YANG,Wei XUE,Fangfang LIU,Fangli QIAO,Wei ZHAO,Xunqiang YIN,Chaofeng HOU,Chenglong ZHANG,Wei GE,Jian ZHANG,Yangang WANG,Chunbo ZHOU,Guangwen YANG.The Sunway Taihu Light supercomputer:system and applications[J].Science China(Information Sciences),2016,59(07):113-128.[2017-09-01].
[13]Jack Dongarra.Sunway Taihu Light supercomputer makes its appearance[J].National Science Review,2016,3(03):265-266.[2017-09-01].
[14]孟德龙,文敏华,韦建文,林新华.神威太湖之光上OpenFOAM的移植与优化[J].计算机科学,2017,44(10):64-70.
[15]赵伟,雷晓燕,陈德训,赵春,宋振亚.全球涡分辨率并行海洋模式POP在神威蓝光上的移植和应用[J].计算机应用与软件,2014,31(05):42-45.
[16]许瑾晨,郭绍忠,黄永忠,王磊.面向异构众核从核的数学函数库访存优化方法[J].计算机科学,2014,41(06):12-17.
[17]吴少刚,许解峰,杨耀忠,任钢.高性能BLAS在类Beowulf机群系统上的实现[J].小型微型计算机系统,2001(08):897-900.
[18]伍明川,黄磊,刘颖,何先波,冯晓兵.面向神威·太湖之光的国产异构众核处理器OpenCL编译系统[J/OL].计算机学报,2017:1-15[2018-01-.http://kns.cnki.net/kcms/detail/11.1826.TP.20171003.1550.002.html.