资源描述：
《AUTOMATED DISTRIBUTED SIMULATION》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、AUTOMATEDDISTRIBUTEDSIMULATIONINPTOLEMYIIbyDANIELLÁZAROCUADRADOEMBEDDEDSYSTEMSGROUPCenterforTeleInfratructure(CTIF)AalborgUniversity(AaU)FredrikBajersVej7,A3DK-9220AalborgDenmarkSupervisorsProf.AndersP.Ravn,AalborgUniversity,DenmarkAssoc.Prof.PeterKoch,AalborgUniversity,Denm
2、arkOpponentsProf.BrianVinter,UniversityofCopenhagen,DenmarkAssoc.Prof.KallolBagchi,UniversityofTexasatElPaso,USAAssoc.Prof.JosvaKleist,AalborgUniversity,Denmark(chairman)ISSN0908-1224ISBN87-92078-08-7Amispadres,quevalenmásquecienmaestros.ABSTRACTABSTRACTDespitethewellknownad
3、vantagesofdistributedprocessingforintensivecomputationslikesimulation,frameworksoftenfailtoexploitthem.Adistributedsimulationishardertodevelopthanasequentialone,becauseitisnecessarytointerfaceandmapactivitiestoprocessorsandhandletheensuingcommunicationandsynchronizationprobl
4、ems.Veryoftenthedesignerhastoexplicitlyspecifyinformationaboutdistributionfortheframeworktoexploitparallelism.ThisdissertationpresentsAutomatedDistributedSimulation(ADS),whichallowsdesignerstoforgetaboutdistributionconcernswhilebenefitingfromtheadvantages.Thisstudyshowsabstr
5、actionsthathelpautomatethedistributionofasimulationbytakingmaximumadvantageofinherentparallelism.ADSreliesonanactorformalismtoprovideencapsulationofcomponentbehaviorandreusability.Modelsofcomputationgoverncomponentinteractionsbydefiningexecutionandcommunicationmechanismsandt
6、henotionoftime;thusprovidingsemantics.Differentmodelsofcomputationusefulforembeddedsystemsaresurveyedtodiscusspossibilitiesfordistribution.SynchronousDataflowallowsforstaticscheduling,thereforeeliminatingruntimeoverheadsandhavingahigherpotentialforefficientparallelization.Mo
7、reoveritisapopularformalismsuitedforalargenumberofembeddedapplications,thusischosenfortheinitialimplementationofADS.Wepresentanovelexecutionmechanismthatproducesoptimalperiodicadmissibleparallelschedulesallowingthedispatchingmechanismtoapplypipeliningtechniques.Theimplementa
8、tionisdescribedwithmajoremphasisondistributionissuesasinterfacingandmapping
