计算机工程 ›› 2012, Vol. 38 ›› Issue (20): 243-246.doi: 10.3969/j.issn.1000-3428.2012.20.063

• 开发研究与设计技术 • 上一篇    下一篇

公平性考虑的短作业优先内存调度策略

金瑛棋 1,吴俊敏 1,2,赵小雨 2   

  1. (1. 中国科学技术大学计算机科学与技术学院,合肥 230027;2. 中国科学技术大学苏州研究院,江苏 苏州 215123)
  • 收稿日期:2011-12-15 修回日期:2012-02-23 出版日期:2012-10-20 发布日期:2012-10-17
  • 作者简介:金瑛棋(1987-),男,硕士研究生,主研方向:多核系统结构;吴俊敏,副教授;赵小雨,硕士
  • 基金项目:
    中央高校基本科研业务费专项基金资助项目(WK0110000020)

Fairness-considered Shortest Job First Strategy for Memory Scheduling

JIN Ying-qi 1, WU Jun-min 1,2, ZHAO Xiao-yu 2   

  1. (1. School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027, China; 2. Suzhou Institute for Advanced Study, University of Science and Technology of China, Suzhou 215123, China)
  • Received:2011-12-15 Revised:2012-02-23 Online:2012-10-20 Published:2012-10-17

摘要: 针对片上多核平台下多线程访问共享内存资源的不公平性、低效性问题,提出公平性考虑的短作业优先内存调度策略,采用设置请求最大等待时间保证线程访问公平性,短作业优先策略缩小请求平均等待时间,关注线程本身固有的并行性。实验结果证明,该策略在多线程访问时IPC性能提升明显,最高性能提升达到43%。

关键词: 多核, 多线程, 共享内存, 公平性, 短作业优先, 并行性

Abstract: It is unfair, ineffective to access shared memory for multi-thread on the chip of multi-core. This paper proposes fairness-aware shortest job first scheduling which sets request for the maximun waiting time to guarantee fairness, uses shortest job first priority to cut down the average waiting time, pays attention to parallelism of threads to solve above problems. Experimental result shows that this method improves IPC performance in multi-thread visit, and the highest performance improvement can reach 43%.

Key words: multi-core, multi-thread, shared memory, fairness, shortest job first, parallelism

中图分类号: