标题:Energy Optimization for Multi-Level Cell STT-MRAM Using State Remapping
作者:Zang, Xiangteng; Li, Xin; Sun, Yuqing; Zhao, Mengying; Dou, Lei
通讯作者:Li, X
作者机构:[Zang, Xiangteng; Li, Xin; Sun, Yuqing; Zhao, Mengying] Shandong Univ, Sch Comp Sci & Technol, Jinan, Peoples R China.; [Dou, Lei] Shandong Normal U 更多
会议名称:18th IEEE International Conference on High Performance Computing and Communications (HPCC) / 14th IEEE International Conference on Smart City (Smart City) / 2nd IEEE International Conference on Data Science and Systems (DSS)
会议日期:DEC 12-14, 2016
来源:PROCEEDINGS OF 2016 IEEE 18TH INTERNATIONAL CONFERENCE ON HIGH PERFORMANCE COMPUTING AND COMMUNICATIONS; IEEE 14TH INTERNATIONAL CONFERENCE ON SMART CITY; IEEE 2ND INTERNATIONAL CONFERENCE ON DATA SCIENCE AND SYSTEMS (HPCC/SMARTCITY/DSS)
出版年:2016
页码:546-553
DOI:10.1109/HPCC-SmartCity-DSS.2016.12
摘要:Multi-level Cell Spin-Transfer Torque Magnetic RAM (MLC STT-MRAM), is emerging as a promising candidate to build L2 cache. However, the high write energy impedes the adoption of MLC STT-MRAM. In this paper, we focus on reducing the write energy consumption. For a 2-bit STT-MRAM cell, the energy consumption of flipping the left bit is greater than that of flipping the right bit. Beside, writing state '01' and state '10' consumes more energy than writing states ;00' and '11'. To reduce the energy consumption during MLC STT-MRAM write operations, we propose an encoding scheme to minimize the number of the left bit flips and reduce the write count of states '01' and '10' by exchanging two states with different left bit, such as states ` 00' and '10', when exchanging two states can reduce the total write energy consumption of a cache line. The main idea of this scheme is state remapping. The experimental results show that the encoding scheme reduces 12.17% write energy consumption on average and up to 27.58% based on the complex write scheme.
收录类别:CPCI-S
WOS核心被引频次:2
资源类型:会议论文
TOP