基于任務(wù)負(fù)載預(yù)測(cè)的實(shí)時(shí)系統(tǒng)低功耗策略的設(shè)計(jì).pdf

1、東南大學(xué)碩士學(xué)位論文基于任務(wù)負(fù)載預(yù)測(cè)的實(shí)時(shí)系統(tǒng)低功耗策略的設(shè)計(jì)姓名：陶倩倩申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：電子科學(xué)與技術(shù)；微電子學(xué)與固體電子學(xué)指導(dǎo)教師：胡晨20110409東南大學(xué)碩士學(xué)位論文AbstractEmbeddedrealtimesystemhadbeenwidelyusedinportableproductswhichalwaystendedtelyonbaacrypoweredBecausetheconstraintsinthev

2、olumeandqualityoftheproducts，baRerycapacitywaslimitedTherefore，lowpowerdesignbecameamajorconcernintheembeddedrealtimesystemsdesignDynamicVoltageScaling(DVS)technologywasaneffectivelowpowertechnologyThecoreideaswere：dynam

3、icallychangedtheprocessorvoltageandfrequencyduringtheexecutionoftaskstomakefulluseofthesystemidletime，whilekeepingthesystemperformanceThestochasticintrataskDVSalgorithmsadjustedthevoltageandfrequencybasedontheprobability

4、distributionofthetaskworkloadAtthebeginningofthetaskoperationalowerspeedcanbeassignedteducepowerWhenthetaskwasnotOVeratacertainstageofthetaskoperationahigherspeedmustbeassignedtomeetthedeadlinerequirementsSSWPalgorithmde

5、signedinthisthesis，whichpredictedthespeedoftasktunbasedontheprobabilitydistributionofthetaskworkload；thespeedwasgraduallyincreasingduringthetaskexecutionSSWPalgorithmadjustedtheoptimalfrequencyformula，selectedtheappropri

6、atesetoffrequencyvariationpointstoaccomplishtaskfrequencyschedulingTheeffectivenessofreducingpowerconsumptionvariedindifferentdistributionmodelsComparingtOOPDVSalgorithmSSWPalgorithmloweredpowerbetterbecameitdidspeedscal

7、ingdirectlyandconsideredmorepracticalfactorsOnthesimdationplatformunderthedifferentquantityoftasks，differentworkloadprobabilitydistributionanddifferentworkloadutilizationtheeffectivenessofreducingpowerconsumptionusingSSW

8、PalgorithmwasrespectivelyverifiedandcomparedwithOPDVSalgorithmTheconclusionswereasfollows：thequantityoftaskhadnoobviouseffectivenessonSSWPalgorithmtheextentofreductioninpowerconsumptionremainedaround38％Whiletaskworkloadp

9、robabilitydistributionhadasignificantinfluence，gammadistributionmodelhadahighestextentofreductioninpowerconsumptionto45％Atlast，whentheworkloadutilizationwaslowSSWPalgorithmhadaverylimitedreductioninpowerconsumptionbutthe

10、extentofreductionupto38％whentheworkloadutilizationwashighIntheabovethreecases，SSWPalgorithmloweredabout10％powerconsumptionthanOPDVSalgorithmAndSSWPalgorithmalsohadahigherutilizationofsystemidletimethanOPDVSalgorithmthuss