Multilevel Storage Cell Characterization and Behavior Modeling of a Crossbar Computational Array in ESF3 Flash Technology : (Invited Paper)

Jonah P. Sengupta, Gaspar Tognetti, P. Pouliquen, A. Andreou
{"title":"Multilevel Storage Cell Characterization and Behavior Modeling of a Crossbar Computational Array in ESF3 Flash Technology : (Invited Paper)","authors":"Jonah P. Sengupta, Gaspar Tognetti, P. Pouliquen, A. Andreou","doi":"10.1109/CISS.2019.8693054","DOIUrl":null,"url":null,"abstract":"Analog and mixed signal vector-matrix multipliers (VMMs) in non-volatile memory cells provide a pathway way to energy and area efficient computation that rivals traditional DSP units. The ability to store multiple bits of data in each floating-gate device for a long amount of time allows for ideal use as hardware accelerators for neural networks. However, current implementations of the technology are impeded by large peripheral circuit overhead. In this work, steps towards a fully-realized NVM-based VMM ASIC are taken. In the 55-nm process, a test array of embedded NOR-flash memory (ESF3) from SST was implemented and fabricated. Multi-bit storage has been demonstrated within the 2x2 array via voltage-mode programming. Secondly, behavioral emulation of the VMM is implemented on a Kintex-7 FPGA to prototype the digital periphery, algorithms, and configurations necessary to embed the NVM-based VMM in a system on chip (SoC). Two quadrant kernel processing is demonstrated at a throughput of 1.8 GOPs for a 72x16 array which can process a 128x128 image @ 250 fps.","PeriodicalId":123696,"journal":{"name":"2019 53rd Annual Conference on Information Sciences and Systems (CISS)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2019-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 53rd Annual Conference on Information Sciences and Systems (CISS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/CISS.2019.8693054","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Analog and mixed signal vector-matrix multipliers (VMMs) in non-volatile memory cells provide a pathway way to energy and area efficient computation that rivals traditional DSP units. The ability to store multiple bits of data in each floating-gate device for a long amount of time allows for ideal use as hardware accelerators for neural networks. However, current implementations of the technology are impeded by large peripheral circuit overhead. In this work, steps towards a fully-realized NVM-based VMM ASIC are taken. In the 55-nm process, a test array of embedded NOR-flash memory (ESF3) from SST was implemented and fabricated. Multi-bit storage has been demonstrated within the 2x2 array via voltage-mode programming. Secondly, behavioral emulation of the VMM is implemented on a Kintex-7 FPGA to prototype the digital periphery, algorithms, and configurations necessary to embed the NVM-based VMM in a system on chip (SoC). Two quadrant kernel processing is demonstrated at a throughput of 1.8 GOPs for a 72x16 array which can process a 128x128 image @ 250 fps.
ESF3闪存技术中交叉杆计算阵列的多层存储单元表征和行为建模(特邀论文)
非易失性存储单元中的模拟和混合信号矢量矩阵乘法器(vmm)提供了一种与传统DSP单元相媲美的节能和面积高效计算途径。在每个浮栅设备中长时间存储多位数据的能力,使其成为神经网络硬件加速器的理想用途。然而,目前该技术的实现受到大型外围电路开销的阻碍。在这项工作中,采取了完全实现基于nvm的VMM ASIC的步骤。在55纳米工艺下,实现并制作了SST的嵌入式nor闪存(ESF3)测试阵列。通过电压模式编程在2x2阵列中演示了多比特存储。其次,在Kintex-7 FPGA上实现了VMM的行为仿真,以实现将基于nvm的VMM嵌入片上系统(SoC)所需的数字外围、算法和配置的原型。二象限内核处理的吞吐量为1.8 GOPs,用于72x16阵列,该阵列可以以250 fps的速度处理128x128图像。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
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
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
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 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信