一种1 GHz多端口低功耗寄存器堆设计

J4 ›› 2015, Vol. 37 ›› Issue (12): 2222-2227.

一种1 GHz多端口低功耗寄存器堆设计

李娇1,2，王良华1，毕卓1,3，刘鹏1

(1.上海大学微电子研究与开发中心，上海 200072；
2.上海大学新型显示技术及应用集成教育部重点实验室，上海200072；
3.上海大学机电工程与自动化学院自动化系，上海 200072)

收稿日期:2014-12-15 修回日期:2015-03-31 出版日期:2015-12-25 发布日期:2015-12-25

A 1 GHz multi-port low-power register file design

LI Jiao1,2,WANG Lianghua1,BI Zhuo1,3,LIU Peng1

(1.Microelectronics R&D Center,Shanghai University,Shanghai 200072;
2.Key Laboratory of Advanced Display and System Application,Shanghai University,Shanghai 200072;
3.School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200072,China)

Received:2014-12-15 Revised:2015-03-31 Online:2015-12-25 Published:2015-12-25

摘要/Abstract

摘要：

超标量处理器中的寄存器堆通常采用多端口结构以支持宽发射，这种结构对寄存器堆的速度、功耗和面积提出了很大的挑战。设计了一个64*64 bit多端口寄存器堆，该寄存器堆能够在同一个时钟周期内完成8次读操作和4次写操作，通过对传统单端读写结构的存储单元进行改进，提出了电源门控与位线悬空技术相结合的单端读写结构的存储单元，12个读写端口全部采用传输门以加快访问速度。采用PTM 90 nm、65 nm、45 nm和32 nm仿真模型，在Hspice上进行仿真，与传统单端读写结构相比较，所提出的方法能够显著提升寄存器堆的性能，其中写1操作延时降低超过32%，总功耗降低超过45%，而且存储单元的稳定性也得到明显改善。

关键词: 寄存器堆, 单端结构, 电源门控, 位线悬空

Abstract:

Register files in superscalar processors usually adopt the multiport structure to support the wide issue, however, this structure brings in problems such as prolonging access speed, increasing in silicon areas and higher power consumption.We design a 64*64 bit multiport register file which can concurrently accomplish 8 read operations and 4 write operations in one single clock cycle.We improve the conventional singleended memory cell structure and purpose a new structure, which combines the powergating and the bitline floating techniques, and the transmission gate is used in all ports to accelerate the access speed.Simulations are conducted on Hspice with PTM 90 nm, 65 nm, 45 nm and 32 nm technology models compared with the conventional singleended structure, the proposed method can significantly improve the performance of register files, the delay of write logic 1 decreases more than 32%, and the total power consumption decreases more than 45%; the stability of memory cells is also improved.

Key words: register file;single-ended;power-gating;bit-line floating

李娇1,2，王良华1，毕卓1,3，刘鹏1. 一种1 GHz多端口低功耗寄存器堆设计[J]. J4, 2015, 37(12): 2222-2227.

LI Jiao1,2,WANG Lianghua1,BI Zhuo1,3,LIU Peng1. A 1 GHz multi-port low-power register file design [J]. J4, 2015, 37(12): 2222-2227.

参考文献

［1］Patwary A R,Greub H,Feng Zhong. Bitline organization in register files for lowpower and highperformance application［C］∥Proc of Electrical and Computer Engineering，2006:505508．
［2］Roy S,Ranganathan N. Stateretentive power gating of register files in multicore processors featuring multithreaded inorTable 1Register file performance comparison under four technique models表1四种工艺下寄存器堆性能对比单位90 nm传统改进提升65 nm传统改进提升45 nm传统改进提升32 nm传统改进提升写0延时ps 107108-0.9%90.689.90.7% 88.880.69.2%92.474.219.7% 写1延时ps 18212133.5%15010132.7%22490.159.8%25383.367.1%读0延时ps 30620533.0%25218725.8%2101919.0%1911872.1%读1延时ps0 262 /0245 / 0227 /0240 / 写0容限mv43052522.1%40550524.7%39048524.3%36544523.3%写1容限mv37559057.3%38557549.4%34050548.5%30546552.5%读0容限mv375375保持380380保持 325325保持 320320保持读1容限mv375375保持380380保持 325325保持 320320保持总功耗 mw48.913.971.2%29.99.767.6%13.26.947.7% 7.23.945.8%der cores［J］.Transactions on Computers，2011，60（11）:15471560．
［3］Li Shenglong，Li Zhaolin，Wang Fang. Design of a highspeed lowpower multiport register file［C］∥Proc of Microeletronics &Electronics，2009:408411．
［4］Wang Fang,Ji Lijiu. Design of high speed 2Write/6Read eightport register file［C］∥Proc of ICASIC，2003:498501．
［5］Sarfraz K,Chan M. A lownoise local bitline technique for dualVt register files［C］∥Proc of Faible Tension Faible Consommation，2014:14．
［6］Gong N,Wang J,Sridhar R. Clockbiased local bit line for high performance register files［J］.Electronics Letters，2012，48（18）:11041105．
［7］Tseng J H,Asanovic K. Banked multiported register files for highfrequency superscalar microprocessors［C］∥Proc of Computer Architecture，2003:6271．
［8］Wang Daoping,Lin HonJarn,Hwang Wei. Lowpower multiport SRAM with crosspoint write wordlines,shared write bitlines,and shared write rowaccess transistors［J］.Circuits and Systems II，2014，61（3）:188192．
［9］Xiong Baoyu. High performance low power multiport register file research and full custom implementation［D］.Shanghai:Fudan University，2011．(in Chinese)
［10］He Peng. Full custom design and realization of largescale multiport high speed register File［D］.Changsha:National University of Defense Technology，2005．(in Chinese)
［11］Konstadinidis G K,Tremblay M,Chaudhry S. Architecture and physical implementation of a third generation 65nm，16Core，32Thread chipmultithreading SPARC processor［J］.JSSC，2009，44（1）:717．
［12］Hassanzadeh S,Zamani M,Hajsadeghi K. A 32kb 90nm 10Tcell subthreshold SRAM with improved read and write SNM［C］∥Proc of Eletrical Engineering,2013:15．
附中文参考文献：
［9］熊保玉. 高性能低功耗多端口寄存器文件研究与全定制实现［D］.上海:复旦大学，2011．
［10］贺鹏. 大规模多端口高速寄存器文件全定制设计与实现［D］.长沙:国防科学技术大学，2005．