欢迎来到天天文库
浏览记录
ID:31978515
大小:6.97 MB
页数:65页
时间:2019-01-30
《基于化学反应优化和优先级的异构多核低功耗调度.研究》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、硕士学位论文略在显著降低系统能耗的同时也有效地降低了算法的时间复杂度。关键词:异构多核;CRO;任务划分;关键任务;任务调度;缩放优先级;DVSIII基于化学反应优化和优先级的异构多核低功耗调度研究AbstractAsthecomputingpowerofheterogeneousmulti-coreprocessorsincreasessignificantly,energyconsumptionhasalsobecomeaperformancebottleneckrestrictingitsdevelopmentandap
2、plicationpromotion.Thelow-powerdesigntechniquesclosertothetop,thehigherthelevelofabstraction,thegreatertheimpactonpowerconsumption.Sotheresearchofreal·timelow-powerschedulingalgorithmforheterogeneousmulti—coreprocessorsystemshasgotextensiveattentionfromresearchersat
3、homeandabroad.Heterogeneousmulti·-corelow—-powerschedulingproblemisatypicalNP(non—polynomial)hardproblem.Existingalgorithmstosolvetheproblemmostlyuseatwo-stageheuristicstrategies:First,tasksareassignedtoeachprocessingunitbyapartitioningstrategy.Andthen,thetask’Sexec
4、utionorderandvoltageofprocessingunitaredeterminedbylow-powertaskschedulingpolicy.However,thesealgorithmsmostlyignoredthetimecomplexityofthealgorithmandthetask—relatedcharacteristics’Simpactonthelow-powertechnologyenergysaving.Therefore,inthelightofthecharacteristics
5、ofheterogeneousmulti-coreprocessorsystemmodel,basedonthestronggolbaloptimizationcapabilityofchemicalreactionsoptimizationalgorithm,combinedwitheffectivecontrolofshedulinglengthofkeytasks,taskpowerconsumptionandtimeproperties’impactonlow-powerenergy—savingtechnology,
6、weproposeahighenergy-savingefficientandafastconvergentheterogeneousmulti-corelow-powerschedulingalgorithm.Themainworkareasfollows:First,wedesignataskportioningalgorithmbasedonchemicalreactionoptimization,CROTM(ChemicalReactionOptimizationforTaskMapping).CROisameta—h
7、euristicalgorithmbysimulatingchemicalreactionsinmolecularmotion.ThecoreofCROisfourelementaryreactions:On—wallIneffectiveCollision,Decomposition,Inter—molecularIneffectiveCollisionandSynthesis.Inter-molecularIneffectiveCollisionandOn—wallIneffectiveCollisioncansearch
8、thelocaloptimalvalueeffectively.Inordertopreventsearchstagnationinlocaloptimalvalue,SynthesisandDecompositioncaneffectivelyexpandthesearch
此文档下载收益归作者所有