XB-SIM*:A Simulation Framework for Modeling and Exploration of ReRAM-Based CNN Acceleration Design

作     者：Xiang Fei Youhui Zhang Weimin Zheng Xiang Fei;Youhui Zhang;Weimin Zheng

作者机构：Department of Computer Science and TechnologyTsinghua UniversityBeijing 100084China Beijing National Research Center for Information Science and TechnologyBeijing 100084China 

出 版 物：《Tsinghua Science and Technology》 (清华大学学报（自然科学版（英文版）)

年 卷 期：2021年第26卷第3期

页      面：322-334页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 081201[工学-计算机系统结构] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported in part by Beijing Academy of Artificial Intelligence(BAAI)(No.BAAI2019ZD0403) Beijing Innovation Center for Future Chip,Tsinghua University the Science and Technology Innovation Special Zone Project,China 

主　　题：deep neural network Resistive Random Access Memory(Re RAM) simulation accelerator processing in memory 

摘      要：Resistive Random Access Memory(ReRAM)-based neural network accelerators have potential to surpass their digital counterparts in computational efficiency and ***,design of these accelerators faces a number of challenges including imperfections of the Re RAM device and a large amount of calculations required to accurately simulate the *** present XB-SIM,a simulation framework for Re RAM-crossbar-based Convolutional Neural Network(CNN)***-SIM can be flexibly configured to simulate the accelerator’s structure and clock-driven behaviors at the architecture *** framework also includes an Re RAM-aware Neural Network(NN)training algorithm and a CNN-oriented mapper to train an NN and map it onto the simulated design *** of the simulator has been verified by the corresponding circuit simulation of a real ***,a batch processing mode of the massive calculations that are required to mimic the behavior of Re RAM-crossbar circuits is proposed to fully apply the computational concurrency of the mapping *** CPU/GPGPU,this batch processing mode can improve the simulation speed by up to 5.02 or *** this framework,comprehensive architectural exploration and end-to-end evaluation have been achieved,which provide some insights for systemic optimization.