摘要
自1971年依据电路理论完备性预测有忆阻器的存在,到2008年惠普实验室将忆阻器和现实器件联系起来以来,人们围绕忆阻器的研究层出不穷,而针对高效能计算面临的存储墙和功耗墙问题,忆阻计算体系和电路便是重要的候选者。以基于忆阻器的数值表达、存储、传输和计算方式为基础,阐明了忆阻计算中的蕴含逻辑、布尔逻辑、阈值逻辑等的发展现状和优缺点,总结了忆阻计算在神经形态计算、精确计算、近似计算等方面的突出研究工作。现阶段,忆阻计算仍然处在相对初级的阶段,其中的逻辑电路基础体系、信号退化、良好扩展性等问题还需要深入研究以迎接忆阻计算时代的到来。
Since memristor was postulated based on the symmetry and completeness of circuit theory in 1971 and it was related to the real device by HP lab in 2008, memristor centered research has outbroken, and to attack the memory wall and power wall and for high-efficiency computing, memristor-based computing architecture and circuit are strong and critical candidate. Based on the introduction of memristor-based number representation, storage, transmission, and computing methods, the actual state of the art of development, pros, and cons of implication logic, Boolean logic, threshold logic for memristor-based computing are addressed, the research works are deeply reviewed for neuromorphic computing, accurate computing, and approximate computing. Currently,memristor-based computing is on an infancy stage, and its basic logic circuit architecture, signal degradation, good scalability, etc.,should be further investigated in detail to welcome the era of memristor-based computing.
作者
江先阳
容源
王永甲
张惟
JIANG Xianyang;RONG Yuan;WANG Yongjia;ZHANGWei(School of Physics and Technology,Wuhan University,Wuhan 430072,China;National Demonstration Center for Experimental Physics and Education,Wuhan University,Wuhan 430072,China)
出处
《微纳电子与智能制造》
2020年第1期12-36,共25页
Micro/nano Electronics and Intelligent Manufacturing
基金
国家自然科学基金(61072135,81971702)
中央高校基本科研业务费专项资金(2042017 gf 0075,2042019 gf0072)
湖北省自然科学基金(2017CFB721)项目资助。