面向异构超算的结构分析高效并行计算方法  

An efficient parallel computing method for structural analysis based on heterogeneous supercomputer

作　　者：丁峻宏 苗新强 李根国[3] DING Jun-hong;MIAO Xin-qiang;LI Gen-guo(Shanghai Supercomputing Technology Co.,Ltd.,Shanghai 201203;The 51st Research Institute of China Electronics Technology Group Corporation,Shanghai 201802;Shanghai Supercomputer Center,Shanghai 201203,China)

机构地区：[1]上海超算科技有限公司,上海201203 [2]中国电子科技集团公司第五十一研究所,上海201802 [3]上海超级计算中心,上海201203

出　　处：《计算机工程与科学》2020年第12期2133-2140,共8页Computer Engineering & Science

基　　金：国家重点研发计划(2018YFB0204004);上海市工业互联网创新发展专项(2018-GYHLW-02008)。

摘　　要：为充分利用异构超算众核环境下的高效计算性能,提出一种针对有限元结构力学分析的多层次多粒度协同并行计算方法。该方法将每个计算作业从结点间、设备间和核间3个层次进行划分。每个计算作业被分解映射到异构超算的各硬件层面运行,在有效处理CPU与MIC之间负载均衡问题的基础上显著降低了异构超算的通信成本。在“天河二号”上开展了针对几个大规模工程案例的计算测试,所启动的CPU+MIC核数总计达39000个,计算模型有限元规模超过1亿单元。测试结果表明,该方法进行有限元结构分析大规模并行计算时能够获得良好的加速比和并行计算效率,实现了有限元结构分析与异构众核计算平台的优化适配,可为同类应用问题的并行移植和性能优化提供借鉴与参考。In order to exploit the efficient computing power of many integrated cores on heterogeneous cluster,a multi-level and multi-granularity collaborative parallel computing method is proposed for finite element structural mechanical analysis.Computing tasks are divided into three levels:inter-node parallelism,inter-device parallelism and inter-core parallelism.Through mapping decomposable comput-ing jobs to different hardware layers of heterogeneous MIC system,the proposed method not only effectively resolves the load balancing problem between CPU and MIC devices,but also significantly reduces the communication overheads of the system.Different engineering simulation case experiments for large scale parallel computing were conducted on“Tianhe 2”supercomputer.Up to 39000 CPU+MIC cores were employed and the finite element size of the analysis was more than 100 million units.Test results show that the proposed method can achieve good speedup and parallel computing efficiency in large scale parallel computing of finite element structural analysis.The optimized adaptation of finite element structural analysis and heterogeneous MIC computing platform is realized,which can provide reference for parallel porting and performance optimization of similar applications.

关 键 词：并行计算 异构超算 结构分析 负载均衡 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] O246[自动化与计算机技术—计算机科学与技术]