一种基于近似压缩器的低功耗近似乘法器  

A Low Power Approximate Multiplier Based on Approximate Compressors

作　　者：方宝 梁华国[1] 盛勇侠 蒋翠云[2] 易茂祥[1] 黄正峰[1] 鲁迎春[1] 徐辉[3] FANG Bao;LIANG Huaguo;SHENG Yongxia;JIANG Cuiyun;YI Maoxiang;HUANG Zhengfeng;LU Yingchun;XU Hui(School of Electronic Science&Applied Physics,Hefei University of Technology,Hefei 230009,P.R.China;School of Mathematics,Hefei University of Technology,Hefei 230009,P.R.China;School of Computer Science&Engineering,Anhui University of Science&Technology,Huainan,Anhui 232001,P.R.China)

机构地区：[1]合肥工业大学电子科学与应用物理学院,合肥230009 [2]合肥工业大学数学学院,合肥230009 [3]安徽理工大学计算机科学与工程学院,安徽淮南232001

出　　处：《微电子学》2021年第5期678-684,共7页Microelectronics

基　　金：国家自然科学基金资助项目(61834006);国家重大科研仪器研制项目(62027815)。

摘　　要：随着物联网的快速发展,智能终端设备在硬件资源和供电上受到较强限制,迫切需要低功耗的新型运算单元。针对运算单元功耗高的问题,提出了一种基于近似压缩器的低功耗近似乘法器,用于图像处理、深度学习等可容错应用领域。实验结果表明,相比于现有近似乘法器,该近似乘法器降低了30.70%的功耗和26.50%的延迟,节省了30.23%的芯片面积,在功耗延迟积(PDP)和能量延迟积(EDP)方面均优化了43%以上。在计算精度方面同样具有一定优势。最后,在图像滤波应用中验证了该近似乘法器的有效性。With the rapid development of the Internet of Things(IoT),the demand for low power consumption by smart mobile devices is increasingly urgent,owing to the limit of their hardware resources and battery life.To solve the problem,a low power approximate multiplier was proposed based on approximate compressors.It could be used in fault-tolerant applications such as image processing and deep learning.The experimental results indicated that,compared to the previous designs,it reduced the power by 30.70%,the delay by 26.50%,and the chip area by 30.23%.It also improved the power-delay product(PDP)or energy-delay product(EDP)by more than 43%.Meanwhile,it also took some advantages of accuracy.The application for image filtering was presented to show the validity of it.

关 键 词：近似计算 物联网 低功耗 乘法器 

分 类 号：TN432[电子电信—微电子学与固体电子学] TN79