COPPER:a combinatorial optimization problem solver with processing-in-memory architecture

作     者：Qiankun WANG Xingchen LI Bingzhe WU Ke YANG Wei HU Guangyu SUN Yuchao YANG Qiankun WANG;Xingchen LI;Bingzhe WU;Ke YANG;Wei HU;Guangyu SUN;Yuchao YANG

作者机构：School of Software&MicroelectronicsPeking UniversityBeijing 100871China School of Computer SciencePeking UniversityBeijing 100871China School of Integrated CircuitsPeking UniversityBeijing 100871China Tencent AI LabShenzhen 518057China College of Physics and Information EngineeringFuzhou UniversityFuzhou 350116China Beijing Advanced Innovation Center for Integrated CircuitsBeijing 100871China Beijing Academy of Artificial IntelligenceBeijing 100080China 

出 版 物：《Frontiers of Information Technology & Electronic Engineering》 (信息与电子工程前沿（英文版）)

年 卷 期：2023年第24卷第5期

页      面：731-741页

核心收录：

学科分类：0810[工学-信息与通信工程] 12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 0808[工学-电气工程] 0809[工学-电子科学与技术（可授工学、理学学位）] 0839[工学-网络空间安全] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 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：Project supported by the National Natural Science Foundation of China(Nos.61832020,62032001,92064006,and 62274036) the Beijing Academy of Artificial Intelligence(BAAI)of China the 111 Project of China(No.B18001) 

主　　题：Combinatorial optimization Chaotic simulated annealing Processing-in-memory 

摘      要：The combinatorial optimization problem(COP),which aims to find the optimal solution in discrete space,is fundamental in various ***,many COPs are NP-complete,and require much more time to solve as the problem scale *** by this,researchers may prefer fast methods even if they are not exact,so approximation algorithms,heuristic algorithms,and machine learning have been *** works proposed chaotic simulated annealing(CSA)based on the Hopfield neural network and did a good ***,CSA is not something that current general-purpose processors can handle easily,and there is no special hardware for *** efficiently perform CSA,we propose a software and hardware *** software,we quantize the weight and output using appropriate bit widths,and then modify the calculations that are not suitable for hardware *** hardware,we design a specialized processing-in-memory hardware architecture named COPPER based on the *** is capable of efficiently running the modified quantized CSA algorithm and supporting the pipeline further *** results show that COPPER can perform CSA remarkably well in both speed and energy.