摘要
随着大数据应用的出现以及传统架构在大数据应用处理过程中诸多局限的产生,体系结构研究逐渐从处理器向异构架构偏移,其中存内处理(Processing-In-Memory,PIM)架构受到广泛关注。为了研究基于粗粒度可重构阵列的存内处理架构,设计实现CGRA-PIMSim,一个支持高度可配置的CGRA-PIM仿真器,以实现对该架构和系统级仿真的研究。该仿真器基于Gem5,采用面向对象的层次化建模技术,通过Python接口配置参数,实现按需求快速综合仿真。
With the emergence of big data applications and the many limitations of traditional architectures in the processing of big data applications,architecture research has gradually shifted from the processor to the heterogeneous architecture,in which the Processing-In-Memory(PIM)architecture Widespread concern.In order to study the in-memory processing architecture based on coarse-grained reconfigurable arrays,propose CGRA-PIMSim,a highly configurable CGRA-PIM simulator to realize the research on the architecture and system-level simulation.The simulator is based on Gem5,uses object-oriented hierarchical modeling technology,configures parameters through the Py⁃thon interface,and realizes rapid comprehensive simulation on demand.
作者
刘硕
LIU Shuo(School of Electronic Information and Electrical Engineering,Shanghai Jiaotong University,Shanghai 200240)
出处
《现代计算机》
2021年第5期97-101,共5页
Modern Computer
关键词
粗粒度可重构阵列
存内处理
仿真器
Coarse-Grained Reconfigurable Array
Processing In Memory
Simulator
