作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2008, Vol. 34 ›› Issue (10): 30-32. doi: 10.3969/j.issn.1000-3428.2008.10.011

• 博士论文 • 上一篇    下一篇

基于部分调用图的线程敏感Profiling技术

刘 弢1,2,吴承勇1,张兆庆1   

  1. (1. 中国科学院计算技术研究所计算机系统结构重点实验室,北京 100080;2. 中国科学院研究生院,北京100039)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2008-05-20 发布日期:2008-05-20

Thread-sensitive Profiling Technique Based on Partial Call Graph

LIU Tao1,2, WU Cheng-yong1, ZHANG Zhao-qing1   

  1. (1. Key Laboratory of Computer System and Architecture, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100080;2. Graduate University of Chinese Academy of Sciences, Beijing 100039)
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-05-20 Published:2008-05-20

摘要: Profiling技术能提供程序实际执行时的相关信息。在动态编译环境中,Profiling的运行时开销导致难以收集较为复杂的运行时信息。该文提出一种基于部分调用图的Profiling技术,在收集多线程程序中线程相关的各种执行信息时,能有效减少运行时开销。在开放源码的虚拟机上实现了该Profiling技术。实验表明,其运行时开销只有原来的2%~4%。

关键词: Profiling技术, 在线Profiling, 运行时环境

Abstract: Profiling technique can provide accurate program execution information. But an online Profiling can only collect simple information in runtime environment due to the overhead. This paper proposes Profiling technique based on partial call graph, which can reduce the runtime overhead when collecting thread-specific profile information of multithread applications. The Profiling technique is implemented in a Java virtual machine. The experiments show its runtime overhead is only 2%~4% compared with that of old ones.

Key words: Profiling technique, online Profiling, runtime environment

中图分类号: