MIX-ACIM：基于矢量矩阵乘法数字特征恢复的28纳米混合精度内存模拟计算

IF 2 Q3 COMPUTER SCIENCE, HARDWARE & ARCHITECTURE

IEEE Solid-State Circuits Letters Pub Date : 2025-07-22 DOI:10.1109/LSSC.2025.3590757

Wei-Chun Wang;Shida Zhang;Laith Shamieh;Narasimha Vasishta Kidambi;Isha Chakraborty;Saibal Mukhopadhyay

{"title":"MIX-ACIM：基于矢量矩阵乘法数字特征恢复的28纳米混合精度内存模拟计算","authors":"Wei-Chun Wang;Shida Zhang;Laith Shamieh;Narasimha Vasishta Kidambi;Isha Chakraborty;Saibal Mukhopadhyay","doi":"10.1109/LSSC.2025.3590757","DOIUrl":null,"url":null,"abstract":"A mixed-precision analog compute-in-memory (Mix-ACIM) is presented for mixed-precision vector-matrix multiplication (VMM). The design features an all-analog current-domain fixed-point (FxP) VMM with floating-point conversion and feature restoration. A 28 nm CMOS test chip shows 41 TOPS/W and 24 TOPS/mm2 for FxP (8-bit input/weight and 12-bit output) and 24.18 TFLOPS/W and 3.3 TFLOPS/mm2 for 16-bit floating-point equivalent operation.","PeriodicalId":13032,"journal":{"name":"IEEE Solid-State Circuits Letters","volume":"8 ","pages":"213-216"},"PeriodicalIF":2.0000,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"MIX-ACIM: A 28-nm Mixed-Precision Analog Compute-in-Memory With Digital Feature Restoration for Vector-Matrix Multiplication\",\"authors\":\"Wei-Chun Wang;Shida Zhang;Laith Shamieh;Narasimha Vasishta Kidambi;Isha Chakraborty;Saibal Mukhopadhyay\",\"doi\":\"10.1109/LSSC.2025.3590757\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"A mixed-precision analog compute-in-memory (Mix-ACIM) is presented for mixed-precision vector-matrix multiplication (VMM). The design features an all-analog current-domain fixed-point (FxP) VMM with floating-point conversion and feature restoration. A 28 nm CMOS test chip shows 41 TOPS/W and 24 TOPS/mm2 for FxP (8-bit input/weight and 12-bit output) and 24.18 TFLOPS/W and 3.3 TFLOPS/mm2 for 16-bit floating-point equivalent operation.\",\"PeriodicalId\":13032,\"journal\":{\"name\":\"IEEE Solid-State Circuits Letters\",\"volume\":\"8 \",\"pages\":\"213-216\"},\"PeriodicalIF\":2.0000,\"publicationDate\":\"2025-07-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE Solid-State Circuits Letters\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://ieeexplore.ieee.org/document/11087240/\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Solid-State Circuits Letters","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/11087240/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}

引用次数: 0

摘要

提出了一种用于混合精度向量矩阵乘法（VMM）的混合精度内存模拟计算（Mix-ACIM）。该设计具有全模拟电流域定点（FxP） VMM，具有浮点转换和特征恢复功能。28纳米CMOS测试芯片显示41 TOPS/W和24 TOPS/mm2的FxP（8位输入/重量和12位输出）和24.18 TFLOPS/W和3.3 TFLOPS/mm2的16位浮点等效操作。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

MIX-ACIM: A 28-nm Mixed-Precision Analog Compute-in-Memory With Digital Feature Restoration for Vector-Matrix Multiplication

A mixed-precision analog compute-in-memory (Mix-ACIM) is presented for mixed-precision vector-matrix multiplication (VMM). The design features an all-analog current-domain fixed-point (FxP) VMM with floating-point conversion and feature restoration. A 28 nm CMOS test chip shows 41 TOPS/W and 24 TOPS/mm2 for FxP (8-bit input/weight and 12-bit output) and 24.18 TFLOPS/W and 3.3 TFLOPS/mm2 for 16-bit floating-point equivalent operation.

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

IEEE Solid-State Circuits Letters Engineering-Electrical and Electronic Engineering

CiteScore

4.30

自引率

3.70%

发文量