[1]凌明,张阳,梅晨,等.一种面向能耗的可重构片上统一存储架构[J].东南大学学报(自然科学版),2011,41(6):1137-1145.[doi:10.3969/j.issn.1001-0505.2011.06.004]
 Ling Ming,Zhang Yang,Mei Chen,et al.Energy-oriented reconfigurable on-chip unified memory architecture[J].Journal of Southeast University (Natural Science Edition),2011,41(6):1137-1145.[doi:10.3969/j.issn.1001-0505.2011.06.004]
点击复制

一种面向能耗的可重构片上统一存储架构()
分享到:

《东南大学学报(自然科学版)》[ISSN:1001-0505/CN:32-1178/N]

卷:
41
期数:
2011年第6期
页码:
1137-1145
栏目:
电子科学与工程
出版日期:
2011-11-20

文章信息/Info

Title:
Energy-oriented reconfigurable on-chip unified memory architecture
作者:
凌明张阳梅晨武建平王欢
(东南大学国家专用集成电路系统工程技术研究中心,南京 210096)
Author(s):
Ling MingZhang YangMei ChenWu JianpingWang Huan
(National ASIC System Engineering Research Center, Southeast University, Nanjing 210096, China)
关键词:
可重构片上统一存储器 SPM动态布局 能耗优化
Keywords:
reconfigurable on-chip unified memory dynamic SPM(scratch-pad memory) allocation low energy consumption
分类号:
TN302
DOI:
10.3969/j.issn.1001-0505.2011.06.004
摘要:
研究了一种新型的针对指令的可重构片上统一存储器架构,能通过配置信息动态地实现Cache和SPM的相互转换,并设计了一套基于Cache相变行为图的动态配置管理算法.为了满足程序执行不同阶段对片上存储资源的需求,对程序的执行特征进行研究并采用了一种基于程序跳转块的程序阶段动态监测与预测技术.通过对程序阶段的预测实现配置信息的快速上下文切换,缩短了重构时间.实验结果表明,采用该优化策略,使用4kB的架构对程序指令段进行优化,与4 kB四路组关联的Cache相比,在保证系统性能的前提下,系统能耗平均降低15.98%,最高能耗降低34.03%.
Abstract:
A dynamic reconfigurable on-chip unified memory hierarchy (RcfgMem) is explored. The given resource of RcfgMem can be divided into certain size of Cache and SPM (scratch-pad memory) by the configuration. And a dynamic configuration management algorithm is provided based on Cache behavior phase graph. The characteristics of program execution are studied and a phase detect logic based on basic block vector is used. The configuration context can be shifted quickly by predicting the program phase. The experiment results show that compared with 4 kB 4-way associate instruction Cache, the total energy consumption can be reduced by 15. 98% on average 34. 03% to the utmost without performance degradation.

参考文献/References:

[1] Wulf W A, Mckee S A. Hitting the memory wall implications of the obvious[J]. Computer Architecture News, 1995, 23(1):20-24.
[2] Balasubramonian R, Albonesi D, Buyuktosunoglu A, et al. Memory hierarchy reconfiguration for energy and performance in general-purpose processor architectures[C]//Proceedings of the 33rd Annual International Symposium on Microarchitecture. Monterey, CA, USA, 2000:245-257.
[3] Viana P, Gordon A, Barros E, et al. A table-based method for single-pass cache optimization[C]//Proceedings of the 18th ACM Great Lakes Symposium on VLSI. New York, 2008:71-76.
[4] Peng Manman, Sun Jiaguang, Wang Yuming. A phase-based self-tuning algorithm for reconfigurable cache[C]//Proceedings of the First International Conference on the Digital Society. Guadeloupe, Caribbean, French, 2007:27-32.
[5] Nakamura H, Kondo M, Boku T. Software controlled reconfigurable on-chip memory for high performance computing[C]//Proceedings of the 2nd Workshop on Intelligent Memory Systems.Denpasar, Bali, Indonesia, 2000:15-32.
[6] Ge Z G, Mitra T, Wong W F. A DVS-based pipelined reconfigurable instruction memory[C]//Proceedings of the 46th Design Automation Conference. San Francisco, CA, USA, 2009:897-902.
[7] Banerjee S. On the effectiveness of runtime resource reconfiguration exploiting program behavior in superscalar microprocessor[D]. Bangalore, India: Philosophy in the Faculty of Engineering, Indian Institute of Science, 2006.
[8] Abella J, Gonzalez A. Heterogeneous way-size cache[C]//Proceedings of the 20th Annual International Conference on Supercomputing. New York, 2006:239-248.
[9] Verma M, Wehmeyer L, Marwedel P. Cache-aware scratchpad allocation algorithm[C]//Proceedings of the Conference on Design, Automation and Test in Europe. Paris, France,2004:1264-1269.
[10] Kuipers K. BNB20 solves mixed integer nonlinear optimization problems[EB/OL]. (2003-08-15)[2010-11-20]. http://www.mathworks.com/matlabcentral/fileexchange/95-bnb.
[11] Guthaus M R, Ringenberg J S, Emst D. MiBench: a free, commercially representative embedded benchmark suite[C]//Proceedings of the Workload Characterization. San Diego, CA, USA, 2001:3-14.
[12] Lee C, Potkonjak M, Mangione W H. MediaBench: a tool for evaluating and synthesizing multimedia and communications systems[C]//The 30th Annual ACM/IEEE International Symposium on Microarchitecture. Research Triangle Park, NC, USA,1997:330-335.

备注/Memo

备注/Memo:
作者简介: 凌明(1972—),男,博士,副教授,trio@seu.edu.cn.
基金项目: 江苏省“青蓝工程”资助项目.
引文格式: 凌明,张阳,梅晨,等.一种面向能耗的可重构片上统一存储架构[J].东南大学学报:自然科学版,2011,41(6):1137-1145. [doi:10.3969/j.issn.1001-0505.2011.06.004]
更新日期/Last Update: 2011-11-20