期刊文献+

一种选择性冲突预测高缓方案

A Cache Scheme by Selective Conflict Prediction
下载PDF
导出
摘要 NTS高缓为直接映象高缓扩充了一个小的全相联高缓,其中保存被预测为具有非时间局部性的数据块.与牺牲高缓的区别在于NTS高缓在两个高缓之间没有直接的数据通路,因此结构设计简单,功耗低.本文提出了一个NTS高缓的改进方案,称为选择性冲突预测高缓(SCP高缓)设计方案.SCP高缓利用冲突预测算法监测高缓中数据块局部性,并有选择性地将数据块填充到直接映象高缓或全相联高缓中.仿真结果显示,容量相当的SCP高缓性能优于NTS高缓. NTS cache augments the direct-mapped main cache with a small fully-associative cache that holds those blocks predicted as holding non-temporal locality characteristics. The most outstanding difference of the NTS cache from the victim cache lies in the NTS cache doesn' t have direct data path between the two caches, so its advantages are lower power and easier structure design. In this paper, an improvement of the NTS cache scheme, called Selective Conflict Prediction cache (SCP cache) ,is proposed. In this scheme,incoming blocks into the cache are placed selectively in the direct-mapped cache or the fully-associative cache by the use of a conflict prediction algorithm which detects the locality of data blocks in the cache. Simulation results show that the performance of SCP cache is always better than that of NTS cache with similar area.
出处 《电子学报》 EI CAS CSCD 北大核心 2018年第2期473-478,共6页 Acta Electronica Sinica
关键词 高缓性能 冲突预测 缺失率 cache performance conflict prediction miss ratio
  • 相关文献

参考文献1

二级参考文献13

  • 1D Burger, T MAustin. Evaluating future processors: The SimpleScalar Tool Set. Technical Report # 1342 [ R]. University of Wisconsin,1997.
  • 2E S Tam,et al.mlcache: A flexible multilateral cache simulater [ A].Proc. MASCOTS'98 [ C]. Motreal, Canada, 1995.19 - 26.
  • 3D A Patterson, J L Hennessy. Computer Architecture a Quantitative Approach [ M].北京:机械工业出版社, 1999.396 - 398.
  • 4N P Jouppi. Improving direct mapping cache performance by the addition of a small full associative cache and prefetch buffers [A]. Proc. of ISCA - 17 [C]. Seattle, USA, 1990.364 - 373.
  • 5J-K Peir, et al. Functional implementation techniques for CPU cache memories [J]. IEEE Trans on Computers, 1999,48(2) : 100 - 110.
  • 6G Kurpanek, et al. PA7200: A PA-RISC Processor with Integrated High Performance MP Bus Interface [ R]. COMPCON Digest of Papers, San Francisco, USA, 1994.375 - 382.
  • 7E S Tam, et al. Active management of data caches by exploiting reuse information [J]. IEEE Trans on Computers, 1999,48(11): 1244 -1258.
  • 8J A Rivers, et al. Evaluating the performance of active cache management schemes [A]. Proc. of ICCD' 98 [ C]. Austin, USA, 1998: 368-375.
  • 9T Johnson, W W Hwu. Run-time adaptive cache hierarchy management via reference analysis [ A]. Prec. of ISCA - 24 [ C]. Boulder, USA,1997. 315 - 326.
  • 10A Gonzalez, et al. Data cache with multiple caching strategies tuned to different types of locality [ A]. Supercomputing' 95 [ C], San Diego,USA, 1995.338 - 347.

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