面向数据流的嵌入式系统中缓存需求量与反应时间的同步优化  

Synchronous Optimization to Buffer Requirement and Response Time in Data Flow Oriented Embedded System

作　　者：朱琦[1] 吴百锋[1] 焦毅[1] 吕晖[1] 

机构地区：[1]复旦大学计算机与信息技术系,上海200433

出　　处：《计算机工程与应用》2005年第12期58-60,91,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(编号:69873010);教育部高等学校骨干教师资助计划资助

摘　　要：同步数据流图(SynchronousDataflowGraph,SDF)是一个在嵌入式系统设计中应用很广泛的模型。基于该模型人们设计出了多种调度算法,针对信号处理领域中的实际问题作了大量的优化工作,提高了系统的性能。但是原先存在的调度算法是分别以缓存优化,或者反应时间优化为其优化目标,从而导致了一些矛盾结果:比如说减少缓存需求量的同时,增大了反应时间;又或者减少反应时间的时候,增大了缓存的需求量。而该文在EDF(最早最终期限优先算法,一种反应时间优算法)的基础上,不以增大系统优化后的反应时间为代价,进一步对其缓存进行优化。从而达到在减少系统反应时间的同时,也能够减少实际的缓存开销的目的。Synchronous Data flow Graph(SDF) is a model of computation that has been widely used in block diagram Digit Signal Processing(DSP) programming environment.Many existing schedule methods of SDF have been developed to reduce run-time costs including the amount of memories used by executing codes and the response time of the target system.But because these methods deal with amount of memories and response time independently,they often generate contradicting results about these two kinds of run-time costs.For example,these methods reduce the amount of memory consumption but increasing response time.By contrast,Earliest Deadline First(EDF) reduce the response time,without considering the buffer requirement of the system.However,by adding a judging condition to EDF,on the one hand,it could schedule the SDF with an efficient response time,one the other hand,could schedule the system with an efficient buffer requirement,while keeping the same response time.Hence,this improved EDF effectively makes up the deficiencies of SDF and EDF,thereby enhancing the utility of SDF and EDF.

关 键 词：嵌入式系统 软硬件协同设计 SDF图 EDF 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]