Tiago Rogerio Mück;Zana Ghaderi;Nikil D. Dutt;Eli Bozorgzadeh
{"title":"Exploiting Heterogeneity for Aging-Aware Load Balancing in Mobile Platforms","authors":"Tiago Rogerio Mück;Zana Ghaderi;Nikil D. Dutt;Eli Bozorgzadeh","doi":"10.1109/TMSCS.2016.2627541","DOIUrl":null,"url":null,"abstract":"The pervasiveness of heterogeneous multiprocessors (HMP) in the mobile domain enables more energy efficient systems. However, current approaches to exploit the energy efficiency of HMPs results in unbalanced usage of resources, which leads to higher aging rates and delay degradation when compared to homogeneous architectures. In this paper, we propose ADAMANT, an aging-aware task mapping algorithm for HMPs. ADAMANT exploits on-chip sensing of aging, performance, and power in orderto enable on-line workload characterization to select task-to-core mappings that yield both increased system lifetime and energy efficiency. Experimental evaluation using a typical mobile workload demonstrates an improvement in chip lifetime by up to 2x on a big.LITTLE architecture.","PeriodicalId":100643,"journal":{"name":"IEEE Transactions on Multi-Scale Computing Systems","volume":"3 1","pages":"25-35"},"PeriodicalIF":0.0000,"publicationDate":"2016-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TMSCS.2016.2627541","citationCount":"23","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Multi-Scale Computing Systems","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/7740903/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 23
Abstract
The pervasiveness of heterogeneous multiprocessors (HMP) in the mobile domain enables more energy efficient systems. However, current approaches to exploit the energy efficiency of HMPs results in unbalanced usage of resources, which leads to higher aging rates and delay degradation when compared to homogeneous architectures. In this paper, we propose ADAMANT, an aging-aware task mapping algorithm for HMPs. ADAMANT exploits on-chip sensing of aging, performance, and power in orderto enable on-line workload characterization to select task-to-core mappings that yield both increased system lifetime and energy efficiency. Experimental evaluation using a typical mobile workload demonstrates an improvement in chip lifetime by up to 2x on a big.LITTLE architecture.
京公网安备 11010802042870号
京ICP备2023020795号-1
分享
求助内容：
应助结果提醒方式：
扫码关注我们
