期刊文献+

基于门限和数据Cache预测的同时多线程调度策略 被引量:1

Improving SMT Thread Schedule Policy Based on Threshold and Data Cache Prediction
下载PDF
导出
摘要 同时多线程处理器每周期从多个活动线程取指令执行,极大地提高了处理器的性能,线程调度策略直接影响SMT处理器的性能。详细分析了SMT中导致指令队列阻塞的各种因素,以及它们对系统性能的影响。在ICOUNT取指调度策略的基础上,提出了一种基于门限的线程调度策略,它考虑了各种导致IQ阻塞的因素及其对IQ效率的影响程度,相对ICOUTNT策略,将IPC平均提高了4.141%。在此基础上,通过统计分析两次数据Cache缺失之间的距离,提出了数据Cache预测技术,又将IPC提高了0.65%。 In SMT architecture, instructions can be fetched from multi-thread simultaneously to increase system performance. Thread schedule policy is an important case in point. Factors affect instruction queue efficiency are analyzed in detail. And on the base of the popular ICOUNT policy, an improving thread schedule policy based on threshold is proposed. In the policy, main factors affect queue efficiency are taken into account according to their influence degree. After simulation, the IPC value is increased by 4. 141% compared with ICOUTNT policy. Afterwards, the distance between twice data cache miss are statistically analyzed, and a new data cache miss prediction method is presented and used in the threshold based thread schedule policy, which allows an additional IPC value increase by 0.65%.
出处 《科学技术与工程》 2008年第8期2044-2048,共5页 Science Technology and Engineering
基金 国家自然科学基金(60573107)资助
关键词 同时多线程 IQ阻塞 线程调度策略 线程取值策略 Simultaneously Multi-Threading (SMT) IQ Clog thread schedule policy instruction fetch policy
  • 相关文献

参考文献2

二级参考文献24

  • 1D Tullsen,S Eggers,et al.Simultaneous multithreading:Maximizing on-chip parallelism[A].Proceedings of the 22nd Annual International Symposium on Computer Architecture[C].Santa Margherita Ligure,Italy:ACM Press,1995.23(2):392-403.
  • 2D Tullsen,S Eggers,J Emer,et al.Exploiting choice:Instruction fetch and issue on an implementable simultaneous multithreading processor[A].Proceedings of the 23rd Annual International Symposium on Computer Architecture[C].PA,USA:ACM Press 1996.24(2):191-202.
  • 3S J Eggers,J Emer,H M Levy,et al.Simultaneous multithreading:a platform for next-generation processors[J].IEEE Micro,IEEE Computer Society Press,1997,17(5):12-19.
  • 4D Tullsen,J Brown.Handling long-latency loads in a simultaneous multithreaded processor[A].Proceedings of the 34th Annual ACM/IEEE International Symposium on Microarchitecture[C].Texas,USA:IEEE Computer Society Press,2001.318-327.
  • 5A El-Moursy,D Albonesi.Front-end policies for improved issue efficiency in SMT processors[A].Proceedings of the 9th International Conference on High Performance Compute Architecture[C].California,USA:IEEE Computer Society Press,2003.31-42.
  • 6A Yoaz,M Erez,et al.Speculation techniques for improving load related instruction scheduling[A].Proceedings of the 26th Annual International Symposium on Computer Architecture[C].Georgia,USA:ACM Press,1999.27(2):42-53.
  • 7F J Cazorla,A Ramirez,et al.DCache Warn:an I-Fetch policy to increase SMT efficiency[A].Proceedings of the 18th International Parallel and Distributed Processing Symposium[C].Santa Fe,New Mexico:IEEE Computer Society Press,2004.74-83.
  • 8F J Cazorla,A Ramirez,et al.Dynamically controlled resource allocation in SMT processors[A].Proceedings of the 34th Annual ACM/IEEE International Symposium on Microarchitecture[C].Portland,Oregon:IEEE Computer Society Press,2004.171-182.
  • 9D Tullsen.Simulation and modeling of a simultaneous multithreading processor[A].Proceedings of the 22nd Annual Computer Measurement Group Conference[C].San Diego,CA,USA:Computer Measurement Group,1996.819-828.
  • 10The standard performance evaluation corporation[Z].http://www.specbench.org.

共引文献5

同被引文献6

引证文献1

二级引证文献5

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部