光计算的发展趋势:模拟或数字?  

Future of Optical Computing:Analog or Digital?

作　　者：马国庆 周常河 朱镕威 郑奉禄 余俊杰[1,2] 司徒国海 Ma Guoqing;Zhou Changhe;Zhu Rongwei;Zheng Fenglu;Yu Junjie;Situ Guohai(Laboratory of Information Optics and Optoelectronic Technology,Shanghai Institute of Optics and Fine Mechanics,Chinese Academy of Sciences,Shanghai 201800,China;Center of Materials Science and Optoelectronics Engineering,University of Chinese Academy of Sciences,Beijing 100049,China;The Institute of Photonics Technology,Jinan University,Guangzhou 510632,Guangdong,China)

机构地区：[1]中国科学院上海光学精密机械研究所信息光学与光电技术实验室,上海201800 [2]中国科学院大学材料与光电学院,北京100049 [3]暨南大学光子技术研究院,广东广州510632

出　　处：《中国激光》2023年第5期1-13,共13页Chinese Journal of Lasers

基　　金：中国科学院前沿科学重点项目(QYZDJSSWJSC014);上海市自然科学基金(19JC1415400、19DZ2291102、20ZR1464700)。

摘　　要：受益于光子独特的优势,光计算技术在构建高速、高算力和高能效比的专用计算加速器方面被寄予厚望,目前已经涌现出了许多极具吸引力的方案。特别是对于涉及运算量巨大的二维矩阵矩阵乘加操作的专用场景,光计算有望在算力和能效比等方面实现超越当前最先进电子计算机几个数量级的性能提升。不同于电子计算通过构建逻辑门实现通用数字计算,主要受深度学习驱动而复兴的光计算更倾向于模拟计算。本文从模拟和数字光计算的角度出发对主流的光计算架构进行分析和讨论,指出了目前光计算技术发展面临的瓶颈,并对光计算未来的发展趋势进行了展望。Significance Deep learning(DL)has become a powerful driving force in the era of intelligence and has been widely used in computer vision,speech recognition,natural language processing,etc.However,more than 80%of calculations in DL are matrixmatrix multiplyaccumulate(MMMAC)operations.Largescale MMMAC operations result in a large number of memory access requirements when the algorithm is converted into CPUexecutable code.Limited by the von Neumann architecture of electronic computers and the physical constraints of the interconnection limit of copper wires on a chip,the training efficiency and speed of deep neural networks(DNNs)are severely restricted.According to the research of de Lima et al.,the computing power required to train stateoftheart DNNs doubles approximately every 3.5 months,far exceeding the computing power supply of electronic integrated circuits(EICs)that follow Moore’s Law.Compared to traditional electronic computing,optical computing is expected to build artificial intelligence(AI)accelerators with high computing power and energy efficiency ratio owing to the high parallelism,high speed,and low power consumption of photons.Currently,various optical computing architectures have demonstrated advantages in terms of high computing power and energy efficiency ratio,and their development routes can be divided into two types.One route is to realize dedicated optical information processing based on multidimensional optical signal modulation and to primarily focus on analog optical computing,such as multiplyaccumulate(MAC)operations,convolutions and correlations,differentiation and integration,Fourier transform,and optical neural networks(ONNs).The other route is to use the conception of electronic computers to design digital optical computers,such as optical transistors,optical logic devices,optical directed logic operations,spacetime parallel coding,and ternary optical computers.Additionally,some important supporting technologies such as optical interconnections,optoelectronic copackaging,optoelectron

关 键 词：光计算 模拟光计算 数字光计算 光计算架构 光学矩阵计算 光学神经网络 光电智能计算 光学信号处理 

分 类 号：TN45[电子电信—微电子学与固体电子学]