← 返回 JSSC 论文列表JSSC 2024第1期Memory22nmCIM
A Nonvolatile AI-Edge Processor With SLCMLC Hybrid ReRAM Compute-in-Memory Macro
提出一种采用混合模式ReRAM的非易失性AI边缘处理器，解决能效与计算延迟问题。
22nm工艺, 51.4 TOPS/W, 472.7µs唤醒响应时间, 67.2 TOPS/W（8位输入/8位权重/22-24位输出）
非易失性存内计算AI边缘处理器混合模式ReRAM能效优化动态累加
▸多模式nvCIM引擎控制器（mmCIM-EC）
▸位输入稀疏与位值感知动态累加（BIS-PVA-DA）
▸位权重列反转（BWCI）与动态累加感知电流量化（DACQ）
Abstract
On-chip non-volatile compute-in-memory (nvCIM)
enables artificial intelligence (AI)-edge processors to perform
multiply-and-accumulate (MAC) operations while enabling the
non-volatile storage of weight data in power-off mode to enhance
energy efficiency. However, the design challenges of nvCIM-
based AI-edge processors include: 1) lack of a nvCIM-friendly
computing flow; 2) a tradeoff between usage of memory devices
versus process variations, computing yield and area overhead;
3) long computing