首页 | 本学科首页   官方微博 | 高级检索  
     

龙芯3A多核处理器系统级性能优化与分析
引用本文:孟小甫,高翔,从明,张爽爽. 龙芯3A多核处理器系统级性能优化与分析[J]. 计算机研究与发展, 2012, 0(Z1): 137-142
作者姓名:孟小甫  高翔  从明  张爽爽
作者单位:中国科学院计算技术研究所;中国科学院研究生院
基金项目:国家“八六三”高技术研究发展计划基金项目(2008AA010901);国家自然科学基金项目(60736012,60921002,61070025,61003064,61100163);国家“九七三”重点基础研究发展计划基金项目(2005CB321600);“核高基”国家科技重大专项课题基金项目(2009ZX01028-002-003,209ZX01029-001-003)
摘    要:多核处理器的性能与系统软件有着密切的联系:操作系统是处理器与应用程序之间的接口,对于充分利用处理器特性和提高应用程序的性能起着极其重要的作用;编译器与处理器体系结构密切相关,一方面要产生处理器支持的二进制代码,另一方面还要结合处理器特性产生高效运行的代码,其性能好坏直接影响着系统的整体性能.为了提高龙芯3A系统的实际性能,从操作系统和编译器着手,结合龙芯3A微结构特征,进行了一系列有效的优化.这些措施包括CC-NUMA多核操作系统的实现、操作系统二级Cache锁机制、操作系统调度共享二级Cache分配、自动向量化编译和支持预取机制的编译等.实验结果表明,在系统软件中增加对处理器特性的支持,能够充分挖掘体系结构的优势,对系统性能有较大的好处.其性能优化技术对于其他处理器的优化也有一定的借鉴价值.

关 键 词:龙芯3A  自动向量化  预取  锁Cache  性能分析

System Performance Optimization and Analysis of Godson-3A Multiprocessor
Meng Xiaofu,Gao Xiang,Cong Ming,and Zhang Shuangshuang. System Performance Optimization and Analysis of Godson-3A Multiprocessor[J]. Journal of Computer Research and Development, 2012, 0(Z1): 137-142
Authors:Meng Xiaofu  Gao Xiang  Cong Ming  and Zhang Shuangshuang
Affiliation:1,2 1(Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190) 2(Graduate University of Chinese Academy of Sciences, Beijing 100049)
Abstract:As the interface between processor and applications, operating system plays an important role in taking advantage of processor features and improving the performance of applications. Compiler is closely related with processor architecture, it has ability to generate the efficient binary code by combining the processor features, so it has direct influence on the overall performance of the system. In this paper, critical techniques of system performance optimization for Godson-3A processor are described:how the compiler supports the automatic vectorization and prefetch mechanism for Godson-3A, and how Linux kernel supports the lock cache mechanism and distributed shared L2 cache partion. And CC-NUMA operating system for Godson-3A is realized. Experimental results show that support for processor architecture in system software can fully exploit the advantages of architecture, and benefit the performance of the computer system.
Keywords:godson-3A  automatic vectorization  prefetch  lock Cache  performance analysis
本文献已被 CNKI 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号