← 返回 JSSC 论文列表JSSC 2020第1期Memory55nmNeural Network AcceleratorCIM
Embedded 1-Mb ReRAM-Based Computing-in- Memory Macro With Multibit Input and Wei
提出基于1Mb ReRAM的内存计算宏，支持多比特输入和权重，实现高效MAC操作。
55nm工艺, 14.6ns MAC延迟, 53.17 TOPS/W能效
内存计算(CIM)阻变存储器(ReRAM)多比特MAC边缘AI能效优化
▸串行输入非加权乘积结构(SINWP)
▸下缩放加权电流转换器(DSWCT)和正负电流减法器(PN-ISUB)
▸电流感知位线钳位方案(CABLC)
▸三重裕度小偏移电流模式感放(TMCSA)
Abstract
Computing-in-memory (CIM) based on embedded
nonvolatile memory is a promising candidate for energy-efficient
multiply-and-accumulate (MAC) operations in artificial intelli-
gence (AI) edge devices. However, circuit design for NVM-based
CIM (nvCIM) imposes a number of challenges, including an area-
latency-energy tradeoff for multibit MAC operations, pattern-
dependent degradation in signal margin, and small read margin.
To overcome these challenges, this article proposes the follow-
ing: 1) a seri