{"title":"Hardware-software cosynthesis of multitask MPSoCs with real-time constraints","authors":"Choonseung Lee, S. Ha","doi":"10.1109/ICASIC.2005.1611458","DOIUrl":null,"url":null,"abstract":"The cosynthesis problem addressed in this paper involves three subproblems: selection of appropriate processing elements, mapping and scheduling of function blocks to the selected processing elements, and schedulability analysis. We have presented a cosynthesis framework in our earlier work (Oh and Ha, 1999) that defines an iteration loop of three steps that attack the subproblems separately. Despite many good features, our previous technique has a severe restriction that a task monopolizes the entire system once activated. But in general we may obtain higher performance if multiple tasks can be running concurrently on different processor cores. It is the main contribution of this paper that we extend the previous framework to be applicable for general multiprocessor systems with diverse operating policies. We demonstrate the performance improvement from the proposed work with a multi-media real-time application, DVR system, and randomly generated multi-task graphs","PeriodicalId":431034,"journal":{"name":"2005 6th International Conference on ASIC","volume":"53 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2005-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"13","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2005 6th International Conference on ASIC","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICASIC.2005.1611458","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 13
Abstract
The cosynthesis problem addressed in this paper involves three subproblems: selection of appropriate processing elements, mapping and scheduling of function blocks to the selected processing elements, and schedulability analysis. We have presented a cosynthesis framework in our earlier work (Oh and Ha, 1999) that defines an iteration loop of three steps that attack the subproblems separately. Despite many good features, our previous technique has a severe restriction that a task monopolizes the entire system once activated. But in general we may obtain higher performance if multiple tasks can be running concurrently on different processor cores. It is the main contribution of this paper that we extend the previous framework to be applicable for general multiprocessor systems with diverse operating policies. We demonstrate the performance improvement from the proposed work with a multi-media real-time application, DVR system, and randomly generated multi-task graphs
本文研究的协同综合问题包括三个子问题:适当加工元素的选择、功能块到所选加工元素的映射和调度以及可调度性分析。我们在早期的工作(Oh and Ha, 1999)中提出了一个共合成框架,它定义了一个由三个步骤组成的迭代循环,分别攻击子问题。尽管我们以前的技术有很多好的特性,但它有一个严重的限制,即任务一旦被激活就会垄断整个系统。但一般来说,如果多个任务可以在不同的处理器内核上并发运行,我们可能会获得更高的性能。本文的主要贡献在于我们扩展了以前的框架,使其适用于具有不同操作策略的通用多处理器系统。我们通过多媒体实时应用程序、DVR系统和随机生成的多任务图演示了所提出的工作的性能改进