用于处理神经网络的电流模式8T SRAM内存计算宏的设计

Taegeun Yoo, T. T. Kim, Bongjin Kim, Chengshuo Yu, Kevin Chai Tshun Chuan
{"title":"用于处理神经网络的电流模式8T SRAM内存计算宏的设计","authors":"Taegeun Yoo, T. T. Kim, Bongjin Kim, Chengshuo Yu, Kevin Chai Tshun Chuan","doi":"10.1109/ISOCC50952.2020.9332992","DOIUrl":null,"url":null,"abstract":"A novel 8T SRAM bitcell is proposed for computing dot-products using current-mode accumulation. A write disturb issue has been eliminated by adding two extra transistors into a standard 6T SRAM bitcell. Besides, we embed a column ADC in each column-based neuron to address the ADC overhead issue of conventional analog compute-in-memory macros. The resolution of ADC is reconfigurable from 1 to 5bit. A test-chip is fabricated using 65nm, and the energy-efficiency of bitwise operation is 490-to-15.8TOPS/W at 1-5bit.","PeriodicalId":270577,"journal":{"name":"2020 International SoC Design Conference (ISOCC)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2020-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Design of Current-Mode 8T SRAM Compute-In-Memory Macro for Processing Neural Networks\",\"authors\":\"Taegeun Yoo, T. T. Kim, Bongjin Kim, Chengshuo Yu, Kevin Chai Tshun Chuan\",\"doi\":\"10.1109/ISOCC50952.2020.9332992\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"A novel 8T SRAM bitcell is proposed for computing dot-products using current-mode accumulation. A write disturb issue has been eliminated by adding two extra transistors into a standard 6T SRAM bitcell. Besides, we embed a column ADC in each column-based neuron to address the ADC overhead issue of conventional analog compute-in-memory macros. The resolution of ADC is reconfigurable from 1 to 5bit. A test-chip is fabricated using 65nm, and the energy-efficiency of bitwise operation is 490-to-15.8TOPS/W at 1-5bit.\",\"PeriodicalId\":270577,\"journal\":{\"name\":\"2020 International SoC Design Conference (ISOCC)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-10-21\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 International SoC Design Conference (ISOCC)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ISOCC50952.2020.9332992\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 International SoC Design Conference (ISOCC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISOCC50952.2020.9332992","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

提出了一种新型的8T SRAM位单元,利用电流模式累加计算点积。通过在标准的6T SRAM位单元中添加两个额外的晶体管,消除了写入干扰问题。此外,我们在每个基于列的神经元中嵌入一个列ADC,以解决传统模拟内存中计算宏的ADC开销问题。ADC的分辨率从1位到5位可重新配置。采用65nm工艺制作了测试芯片,1-5bit的逐位操作能量效率为490 ~ 15.8 tops /W。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
Design of Current-Mode 8T SRAM Compute-In-Memory Macro for Processing Neural Networks
A novel 8T SRAM bitcell is proposed for computing dot-products using current-mode accumulation. A write disturb issue has been eliminated by adding two extra transistors into a standard 6T SRAM bitcell. Besides, we embed a column ADC in each column-based neuron to address the ADC overhead issue of conventional analog compute-in-memory macros. The resolution of ADC is reconfigurable from 1 to 5bit. A test-chip is fabricated using 65nm, and the energy-efficiency of bitwise operation is 490-to-15.8TOPS/W at 1-5bit.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
期刊最新文献
A Quadcopters Flight Simulation Considering the Influence of Wind Design of a CMOS Current-mode Squaring Circuit for Training Analog Neural Networks Instant and Accurate Instance Segmentation Equipped with Path Aggregation and Attention Gate 13.56 MHz High-Efficiency Power Transmitter and Receiver for Wirelessly Powered Biomedical Implants Investigation on Synaptic Characteristics of Interfacial Phase Change Memory for Artificial Synapse Application
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1