基于指令和存储器分析模型的性能分析方法  

An performance analysis method based on instruction and memory analysis model

作　　者：金晶[1] 浦汉来[1] 凌明[1] 

机构地区：[1]东南大学国家专用集成电路系统工程技术研究中心,江苏南京210096

出　　处：《电路与系统学报》2006年第3期42-47,共6页Journal of Circuits and Systems

基　　金：国家"863"计划"个人信息终端SoC"项目(2003AA1Z1340)

摘　　要：随着面向特定应用的系统集成芯片(Application-Specific SoC)大量涌现,在高层抽象级上分析应用软件在目标系统中的性能,对于指导底层设计,减少设计空间探索有着重要意义。本文基于指令执行流和内存访问流的分析,提出一种性能分析的新方法。该方法首先记录指令流和内存访问流,通过元素选取、匹配、解析,最终统计出指令执行时间和内存访问频度,进而分析系统性能。在系统设计早期应用该方法,不仅可以根据系统性能瓶颈针对性的指导底层硬件架构设计,而且可以精确分析系统性能。with the increasing use of Application _Specific SoC, in order to direct lower level design and reduce time spent on exploring design space, target system analysis at abstraction level is developed. In this paper, we propose a performance analysis method on the study for instruction stream and memory access. It firstly registers all of the instruction stream and memory access. Then, by selecting matching and analyzing the elements, it figures out time cost for executing instruction and frequency of accessing memory. Finally, it gets analysis of the whole system performance. It has proved to be an effective way for it not only can guide structure design in lower level hardware but also can provide accurate analysis for the system performance.

关 键 词：系统集成芯片(SoC) 系统性能评估 指令流 内存访问 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]