基于内存指数计算异构计算架构的高能效浮点DNN处理器

2021 IEEE Hot Chips 33 Symposium (HCS) Pub Date : 2021-08-22 DOI:10.1109/HCS52781.2021.9566881

Juhyoung Lee, Jihoon Kim, Wooyoung Jo, Sangyeob Kim, Sangjin Kim, Donghyeon Han, Jinsu Lee, H. Yoo

{"title":"基于内存指数计算异构计算架构的高能效浮点DNN处理器","authors":"Juhyoung Lee, Jihoon Kim, Wooyoung Jo, Sangyeob Kim, Sangjin Kim, Donghyeon Han, Jinsu Lee, H. Yoo","doi":"10.1109/HCS52781.2021.9566881","DOIUrl":null,"url":null,"abstract":"Abstract of Proposed FP CIM Processor (1) Heterogeneous FP Computing Arch. : Separate optimization of FP computing: Realize 2 cycles FP MAC w/ CIM (2) Exponent Computing-in-Memory: In-memory AND/NOR + BL charge reusing: Total memory power 46.4% 2) Mantissa Free Exponent Calculation: Removing redundant normalization: Total MAC power 14.4%","PeriodicalId":246531,"journal":{"name":"2021 IEEE Hot Chips 33 Symposium (HCS)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"An Energy-efficient Floating-Point DNN Processor using Heterogeneous Computing Architecture with Exponent-Computing-in-Memory\",\"authors\":\"Juhyoung Lee, Jihoon Kim, Wooyoung Jo, Sangyeob Kim, Sangjin Kim, Donghyeon Han, Jinsu Lee, H. Yoo\",\"doi\":\"10.1109/HCS52781.2021.9566881\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Abstract of Proposed FP CIM Processor (1) Heterogeneous FP Computing Arch. : Separate optimization of FP computing: Realize 2 cycles FP MAC w/ CIM (2) Exponent Computing-in-Memory: In-memory AND/NOR + BL charge reusing: Total memory power 46.4% 2) Mantissa Free Exponent Calculation: Removing redundant normalization: Total MAC power 14.4%\",\"PeriodicalId\":246531,\"journal\":{\"name\":\"2021 IEEE Hot Chips 33 Symposium (HCS)\",\"volume\":\"15 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-08-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 IEEE Hot Chips 33 Symposium (HCS)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/HCS52781.2021.9566881\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE Hot Chips 33 Symposium (HCS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HCS52781.2021.9566881","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

提出的FP CIM处理器摘要(1)异构FP计算架构。:独立优化FP计算:实现2周期FP MAC w/ CIM(2)内存指数计算:内存和/NOR + BL电荷重用:内存总功率46.4% 2)无尾数指数计算:去除冗余归一化:MAC总功率14.4%

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

微信好友朋友圈 QQ好友复制链接

本刊更多论文

An Energy-efficient Floating-Point DNN Processor using Heterogeneous Computing Architecture with Exponent-Computing-in-Memory

Abstract of Proposed FP CIM Processor (1) Heterogeneous FP Computing Arch. : Separate optimization of FP computing: Realize 2 cycles FP MAC w/ CIM (2) Exponent Computing-in-Memory: In-memory AND/NOR + BL charge reusing: Total memory power 46.4% 2) Mantissa Free Exponent Calculation: Removing redundant normalization: Total MAC power 14.4%

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

2021 IEEE Hot Chips 33 Symposium (HCS)

自引率

0.00%

发文量

期刊最新文献

Multi-Million Core, Multi-Wafer AI Cluster Next Generation “Zen 3” Core Intel’s Hyperscale-Ready Infrastructure Processing Unit (IPU) Sapphire Rapids SambaNova SN10 RDU:Accelerating Software 2.0 with Dataflow