Relinquishment Coherence for Enhancing Directory Efficiency in Chip Multiprocessors

被引:0
作者
Shu, Wei [1 ]
Tzeng, Nian-Feng [1 ]
机构
[1] Univ Louisiana Lafayette, Ctr Adv Comp Studies, Lafayette, LA 70504 USA
来源
PROCEEDINGS OF THE 34TH IEEE INTERNATIONAL CONFERENCE ON COMPUTER DESIGN (ICCD) | 2016年
基金
美国国家科学基金会;
关键词
D O I
暂无
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
A directory-based chip multiprocessor (CMP) suffers from excessive directory area overhead when its size grows. This work leverages novel relinquishment coherence and superior directory efficiency (RECODE) to lower area overhead. Relinquishment coherence boosts the utilization of a hash-based, set-associative table which holds distinct present-bit vectors (PVs), as it transforms a conflict PV to its variations after sharer relinquishment for hashing alternative sets. Superior directory efficiency is resulted from both boosted table utilization and table width shrunk via dropping "runs of zeros" commonly found in PVs. RECODE Table utilization is elevated by relinquishment coherence, which transforms a conflict PV to its variations after sharer relinquishment for hashing alternative sets. RECODE maintains "exact" directory representations for simple coherent logics and low coherent traffic. RECODE is found to enjoy 3.21x more storage efficiency than conventional bit-tracking directories for a CMP with 64 cores and it is 2.64x more storage efficient than the best directory SCD known so far.
引用
收藏
页码:372 / 375
页数:4
相关论文
共 15 条
[1]  
[Anonymous], 2008, P 17 INT C PAR ARCH
[2]  
Binkert Nathan, 2011, Computer Architecture News, V39, P1, DOI 10.1145/2024716.2024718
[3]   SPACE/TIME TRADE/OFFS IN HASH CODING WITH ALLOWABLE ERRORS [J].
BLOOM, BH .
COMMUNICATIONS OF THE ACM, 1970, 13 (07) :422-&
[4]  
Bovet D.P., 2005, UNDERSTANDING LINUX, P258
[5]  
Demetriades S, 2014, INT S HIGH PERF COMP, P177, DOI 10.1109/HPCA.2014.6835928
[6]  
Ferdman M, 2011, INT S HIGH PERF COMP, P169, DOI 10.1109/HPCA.2011.5749726
[7]   Coherence Domain Restriction on Large Scale Systems [J].
Fu, Yaosheng ;
Nguyen, Tri M. ;
Wentzlaff, David .
PROCEEDINGS OF THE 48TH ANNUAL IEEE/ACM INTERNATIONAL SYMPOSIUM ON MICROARCHITECTURE (MICRO-48), 2015, :686-698
[8]   Eliminating conflict misses using prime number-based cache indexing [J].
Kharbutli, M ;
Solihin, Y ;
Lee, J .
IEEE TRANSACTIONS ON COMPUTERS, 2005, 54 (05) :573-586
[9]  
Laudon J, 1997, ACM COMP AR, P241, DOI 10.1145/384286.264206
[10]   Runtime Energy Consumption Estimation for Server Workloads Based on Chaotic Time-Series Approximation [J].
Lewis, Adam Wade ;
Tzeng, Nian-Feng ;
Ghosh, Soumik .
ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION, 2012, 9 (03)