everyprocessorinthesystem.Interruptlatency,reliability,errorrecoverabi - PDF document

everyprocessorinthesystem.Interruptlatency,reliability,errorrecoverability,andotherrelatedaspectsofthissystemarediscussed.7.Davis,A.L.,"TheArchitectureandSystemMethodofDDM1:ARecursivelyStructuredDataDrivenMachine,"ConfProc.5thAnn.Symp.ComputerArchitecture,PaloAlto,Calif.,Apr.1978,pp.210-215.*Thisarchitecturepresentsafacadeofunlimitedconcur-rencyandpipeliningtoprograms,expressedinaformalismcalleddatadrivennets.Multiprocessingisusedtoapprox-imatethisunlimitedparallelism,withtheassignmentofprocessestoavailableprocessorsbeingdonebythehard-ware.8.delaGuardia,M.F.andJ.A.Field,"AHighLevelLanguageOrientedMultiprocessor,"Proc.1976Int'lConfParallelProcessing,WaldenWoods,Mich.,Aug.1976,pp.256-262.*9.Ellis,C.A.andG.J.Nutt,"PreliminaryThoughtsonDegreesofSecurityinMultiprocessorSystems,"Tech.ReportCU-CS-036-74,UniversityofColorado,Boulder,Colo.,June1974.Discussesprotectionandsecurityissuesinamultiprocessingenvironmentandexaminestheimpactofmultiprocessingonproblemssuchasdeadlocks,schedul-ing,andequitableresourceallocation.10.England,D.M.,"ArchitecturalFeaturesoftheSystem250,"Proc.Symp.OperationalOn-LineComputingforDefence,Malvern,England,Nov.1972.Describesthearchitectureandsoftwareofamultiprocessorsystemdesignedforreliable,real-timeoperation.11.Filene,R.J.andA.I.Green,"ASimpleExecutiveforaFault-TolerantReal-TimeMultiprocessor,"Hardware,Software,FirmwareandTradeoffs-DigestofPapers-COMPCONFall71,Boston,Mass.,Sept.1971.Anoperatingsystemintendedforafault-tolerantmultiprocessorinareal-timeenvironmentisdescribed.Ar-chitecturalfeaturestosupportsuchanoperatingsystemarediscussed.12.Fuller,S.H.etal.,"Multi-Microprocessors:AnOverviewandWorkingExample,"Proc.IEEE,Vol.66,No.2,Feb.1978,pp.216-228.Discussestheproblemsandadvantagesofusingmicroprocessorsforbuildingmultiprocessors.DescribesCm*fromthisviewpointandgivessomeperformancemeasurements.13.Gilliland,M.C.,B.J.Smith,andW.Calvert,"HEP:ASemaphore-SynchronizedMultiprocessorwithCentralControl,"Proc.1976SummerComputerSimulationConf.Washington,D.C.,July1976,pp.57-62.HEP-heterogeneouselementprocessor-isamulti-processordesignedspecificallyforhighperformanceinsimulationapplications.14.Grimsdale,R.L.andD.M.Johnson,"AModularExecutiveforMultiprocessorSystems,"Proc.Conf.TrendsinOn-LineComputerControlSystems,Sheffield,England,Apr.1972.Anoperatingsystemforasymmetricmultiprocessorisdescribed.Thereliabilityanderrorrecoveryfeaturesofthissystemareemphasized.15.Grushcow,M.S.,"TheKerneloftheSUEOperatingSystem,"Proc.CanadianComputerConf.,Montreal,June1972.SUEisanexperimentalmultiprocessoroperatingsystembasedonthenotionofvirtualparallelprocessors-eachpro-cessconsidersitselftoberunningonitsown,privatepro-cessor.Themultiplexingoftheactualhardwarebetweenprocesses,theinitiationandterminationofprocesses,andlow-leveloperatingsystemfunctionssuchasI/Oandtimingareperformedbythekernel.16.Gula,J.L.,"OperatingSystemConsiderationsforMulti-processorArchitectures,"Proc.SeventhTexasConf.Com-putingSystems,Houston,Tex.,Nov.1978,pp.7-1to7-6.*17.Holley,L.H.etal.,"VM/370AsymmetricMultiprocessing,"IBMSystemsJ.,Vol.18,No.1,1979,pp.47-70.DescribestheuseofattachedprocessinginIBM'sVirtualMachineOperatingSystem.Performancegoalsaredis-cussedandcomparedwithactualperformance.18.Jones,A.K.etal.,"SoftwareManagementofCm*-ADistributedMultiprocessor,"AFIPSConfProc.,Vol.46,1977NCC,pp.657-663.ContainsanexpositionofthesoftwareissuesinvolvedinbuildinganoperatingsystemforCm*.StarOS,oneofthetwooperatingsystemsforCm*,evolvedfromtheideasandstrategiesdevelopedhere.19.Kamiuchi,T.andH.Nakanishi,"H-80-ALoadsharing,N:1BackupMultisystem,"ComputerTechnology:Status,Limits,Alternatives-DigestofPapers-COMPCONSpring78,SanFrancisco,Calif.,Mar.1978,pp.261-264.*Intendedforon-linecontrolapplications,thissystemusesmultiprocessingforimprovingthroughputandenhancingreliabilityandavailability.20.Kober,R.,"TheMultiprocessorSystemSMS201-Combin-ing128MicroprocessorstoaPowerfulComputer,"Micros,Minis,&Maxis,TechnologyThrustvs.UserRequirement-DigestofPapers-COMPCONFall77,Washington,D.C.,Sept.1977,pp.225-230.*Amulti-microprocessorsystemdesignedbySiemensisdescribed.21.Kraley,M.F.,"ThePluribusMultiprocessor,"DigestofPapers-1975Int'lSymp.Fault-TolerantComputing,Paris,France,June1975,p.251.*Thisfault-tolerantmultiprocessorisdesignedprimarilyformessageroutingandbufferingintheArpanet.Itusesmultiprocessingforreliabilityaswellastoallowmodularperformancegrowth.22.Kuznia,C.H.,R.Kober,andH.Kopp,"SMS101-AStruc-turedMultimicroprocessorSystemwithDeadlock-FreeOperationScheme,"Conf.Proc.3rdAnn.Symp.ComputerArchitecture,Clearwater,Fla.,Jan.1976,p.122.*Amultiprocessordesignusingmicroprocessorsisdescribed.Adistinctivefeatureofthissystemistheuseofsimultaneouslyupdatableprivatememoriesinsteadofasharedglobalmemory.23.Murakami,K.,S.Nishikawa,andM.Sato,"Poly-ProcessorSystemAnalysisandDesign,"ConfProc.4thAnn.Symp.ComputerArchitecture,SilverSpring,Md.,Mar.1977,pp.49-56.*Describesamultiprocessorsystembuiltwithmicroprocessorsdedicatedtoperformingspecializedfunc-tions.Discussesthefunctionalpartitioningofthesystemandpresentssimulationresultspredictingitsperformance.24.Newton,R.S.,"TheDesignofaReentrantExecutiveforaMulti-CPUShared-StoreOperatingSystem,"Proc.Symp.OperationalOn-LineComputingforDefence,Malvern,England,Nov.1972.-Examinestheminimalhardwareandsoftwaresupportneededformultiprocessinganddescribesthoseaspectsofthesoftwarewhichdifferfromtheuniprocessorcase.Alsosuggestsprogrammingtechniquesforimprovingreliabil-ity.25.Newton,R.S.,"AnExerciseinMultiprocessorOperatingSystemDesign,"AGARDConfProc.,No.149,Real-TimeComputer-BasedSystems(NATOAdvisoryGroupforAerospaceR&D),Athens,Greece,May1974.COMPUTER102 Describesamultiprocessorwhoseperformancemaybevariedwidelywithoutsoftwaremodification,thuspermit-tingthehardwaredesignertodefercertaindesigndecisions.42.Wecker,S.,"ADesignforaMultipleProcessorOperatingEnvironment,"ComputingNetworksfromMinisthroughMaxis-AreTheyforReal?-DigestofPapers-COMP-CON73,SanFrancisco,Calif.,Mar.1973,pp.143-146.*Interprocesscommunicationinmultiprocessorsystemsandcomputernetworks,thesoftwareanalogofinter-processorcommunication,isdiscussed.Alsopresentsaunifiedcommunicationstructureandtechniquesforusingthisstructuretointerfaceoperatingsystemsinmultipleprocessorenvironments.43.Widdoes,L.C.,"TheMinervaMulti-Microprocessor,"Conf.Proc.3rdAnn.Symp.ComputerArchitecture,Clear-water,Fla.,Jan.1976,pp.34-39.*Thismultiprocessorconsistsofmicroprocessorscon-nectedbyasharedbus.Discussesinterrupthandlingandtechniquesforreducingprocessorcontentionforthesharedbus.44.Wulf,W.A.andC.G.Bell,"C.mmp-AMulti-Mini-Processor,"AFIPSConf.Proc.,Vol.41,PartII,1972FJCC,pp.765-777.ThispaperdescribestheoriginaldesignofC.mmpanddiscussesissuessuchasitsviabilityandthefeasibilityofbuildingitusingaminimumofcustom-builthardware.ThefinalC.mmpdesigndifferssignificantlyfromthedesigndescribedhere,butthepaperisworthreadingforthepic-tureofthedesigners'originalmotives.45.Wulf,W.etal.,"HYDRA:TheKernelofaMultiprocessorOperatingSystem,"Comm.ACM,Vol.17,No.6,June1974,pp.337-345.DescribestheoriginaldesignoftheoperatingsystemforC.mmp.Thehardwarestructureassumedistheonede-scribedinWulfandBell'sfirstC.mmppaper(44).46.Wulf,W.A.,S.P.Harbison,andR.Levin,Hydra:An'Ex-perimentalOperatingSystem,tobepublishedbyMcGraw-HillInc.,1980.Adetaileddescriptionofthephilosophy,design,andim-plementationofHydra,similarinasensetoOrganick'smonographonMultics.HighlyrecommendedtoanyonedesiringanunderstandingofmultiprocessoroperatingsystemsingeneralandHydrainparticular.PerformanceTheperformanceofmultiprocessorsystemshasbeenatopicofconsiderableinteresttobothresearchersandusers.Someofthepapersinthissectionattempttodefinemean-ingfulmeasuresofperformanceformultiprocessorsystems.Othersusemodelsbasedonanalyticaltechniquessuchasqueueingtheorytopredictperformance.Yetothersestimateperformancebyemployingsimulationandactualmeasurements.Thatmultiprocessorssharememoryim-plies,almostbydefinition,thattherewillbeconflictsbe-tweenprocessorstryingtoaccessthesamememorymodules.Manystudieshaveinvestigatedthemagnitudeofthisproblemandhaveprovidedtechniquestominimizetheperformancedegradationcausedbyit.Mostoftheworkscitedhereaddresstheproblemofconflicts,eitherdirectlyorindirectly.47.Baskett,F.andA.J.Smith,"InterferenceinMultipro-cessorComputerSystemswithInterleavedMemory,"Comm.ACM,Vol.19,No.6,June1976,pp.327-334.Investigatestheeffectsofmemoryinterferenceinmulti-processorsystemsandderivesananalyticalmodeltodescribethis.Simulationresultsvalidatethismodel.48.Bhandarkar,D.P.,"SomePerformanceIssuesinMultiprocessorSystemDesign,"IEEETrans.Computers,Vol.C-26,No.5,May1977,pp.506-511.Usesanalyticandsimulationmodelstostudytheeffectofalternativemultiprocessordesignsonperformance.Alsopresentsguidelinesfordesigningmultiprocessorsystems.49.Chandy,K.M.andM.Reiser,eds.,ComputerPerformance,North-HollandPublishers,Amsterdam,Netherlands,1977.Containsarticlesontheperformanceofmultiprocessorsandcomputernetworks.50.Chow,Yuan-ChiehandW.H.Kohler,"PerformanceofSeveralQueueingModelsforMultiprocessorMulti-programmingSystems,"Computers...bytheMillions,fortheMillions-DigestofPapers-COMPCONFall76,Wash-ington,D.C.,Sept.1976,pp.66-71.*Theperformancesofasinglelargeprocessorandanumberofcooperatingprocessorsarecomparedusingthreedifferentmodels.Indicatestheconditions,whenusingmultipleprocessors,yieldingasmallperformancedegrada-tion.51.Covo,A.A.,"AnalysisofMultiprocessorControlOrganiza-tionswithPartialProgramMemoryReplication,"IEEETrans.Computers,Vol.C-23,No.2,Feb.1974,pp.113-120.Replicatingcopiesofaprogramrunbydifferentpro-cessorsinamultiprocessingsituationhelpsreducememoryinterferenceandqueueingdelays,attheexpenseofin-creasedmemoryrequirements.Describesadynamicpro-grammingsolutionwhichfindsanoptimaldegree-ofreplica-tion.52.DalCin,M.,"PerformanceEvaluationofSelf-Diagnosing,MultiprocessingSystems,"DigestofPapers-EighthAnn.Int'lConf.Fault-TolerantComputing,Toulouse,France,June1978,pp.59-64.*Usesagraph-theoreticmodelandqueueinganalysistoin-vestigatetheperformanceofself-diagnosingmultipro-cessorsystems.53.Ferrari,D.,E.Gelenbe,andR.Mahl,"AnAnalyticStudyofMemoryAllocationinMultiprocessorSystems,"Proc.Conf.ComputerArchitectureandNetworks,Rocquencourt,France,Aug.1974.Discussesanalyticaltechniquesforoptimallypartition-ingmainmemoryamongprocessesrunninginamultipro-cessorenvironmentandoutlinestheissuesinvolvedinap-plyingtheseideastoapracticalsituation.54.Franta,W.R.andP.A.Houle,"CommentsonModelsofMultiprocessorMulti-MemoryBankComputerSystems,"Proc.1974WinterSimulationConf,Vol.I,Washington,D.C.,Jan.1974,pp.86-97.Containsadiscussionofthehardwareandsoftwarefac-torscontributingtomemoryinterferenceinmultiprocessorsystems.Describesasimulationmodelofmemoryin-terferenceandcomparesittoothermodels.55.Fuller,S.H.,"Price/PerformanceComparisonofC.mmpandthePDP-10,"Conf.Proc.3rdAnn.Symp.ComputerAr-chitecture,Clearwater,Fla,Jan.1976,pp.195-202.*Comparestheprice/performancetrade-offsinvolvedinchoosingbetweenalargemainframecomputer(thePDP-10)andamultiprocessorbuiltwithminicomputers(theC.mmp).Yieldsvaluableinsightsintotheeconomicandtechnicalissuesinvolvedinsuchachoice.56.Hoogendoorn,C.H.,"AGeneralModelforMemoryInter-ferenceinMultiprocessors,"IEEETrans.Computers,Vol.C-26,No.10,Oct.1977,pp.998-1005.PresentsamathematicalmodelofmemoryinterferenceinCOMPUTER104 Carnegie-MellonUniversity,Pittsburgh,Pa.,Aug.1978.ThemajorpublishedsourceofinformationonCm*'sper-formance.Describesexperimentsdoneonanearly,one-clusterversionofCm*aswellasanalyticmodelsforlargerconfigurations.72.Raynor,R.J.andJ.M.Gwynn,"MinimizationofSuper-visorConflictforMultiprocessorComputerSystems,"Proc.Symp.SimulationofComputerSystems,Boulder,Colo.,Aug.1976.73.Sastry,K.V.andR.Y.Kain,"OnthePerformanceofCer-tainMultiprocessorComputerOrganizations,"IEEETrans.Computers,Vol.C-24,No.11,Nov.1975,pp.1066-1074.Developsanalyticmodelsforevaluatingtheperformanceofamultiprocessorcomputersystem.Studieshowvaryingthenumberofprocessors,interleavingmemory,andvary-ingthenumberofmemorymodulesaffectstheinstructionexecutionrate.74.Sethi,A.S.andN.Deo,"InterferenceinMultiprocessorSystemswithLocalizedMemoryAccessProbabilities,"IEEETrans.Computers,Vol.C-28,No.2,Feb.1979,pp.157-163.Usingmorerealisticassumptionsthanearlierwork,con-siderstheproblemofmemoryinterference.Developsamodelofthememoryreferencepatternoftypicalprogramsandusesittoanalyzetheperformanceofamultiprocessorsystem.Alsopresentssimulationdatavalidatingtheseanalyticalresults.75.Smith,A.J.,"MultiprocessorMemoryOrganizationandMemoryInterference,"Comm.ACM,Vol.20,No.10,Oct.1977,pp.754-761.Examinestheeffectofvariousmemoryorganizationsoninterference.Demonstratesthatlocalizingeachprocessor'smemoryreferencestooneormorememorymodulesresultsinloweredmemoryinterference.Theseresultsarederivedanalyticallyandverifiedbysimulation.76.Towsley,D.F.,"TheEffectsofCPU:I/OOverlaponCom-puterSystemConfigurations,"Conf.Proc.5thAnn.Symp.ComputerArchitecture,PaloAlto,Calif.,Apr.1978,pp.238-241.*PresentsamodeldescribingtheoverlapofprocessingandI/Oinmultiprocessorsystems.Themodelallowsanevalua-tionoftheeffectivenessofanumberofmultiprocessordesigns.TheoreticalresultsOnegoalofmultiprocessingistospeedupprogramex-ecutionbyeXploitingparallelism.Someofthepapersin-cludedinthissectionpresentresultsontherecognitionofparallelisminsequentialprograms.Othersdealwithissuessuchassynchronizationofcooperatingprocessesandequitablesharingofresourcesbetweencompetingpro-cesses.Theverificationofparallelprogramsformsyetanotherdimensionofresearchinthisarea.Thephrase"multiprocessing,"inthecontextofthissection,isoftenusedasasynonymfor"parallelprocessing,"ratherthaninthemorerestrictedsenseusedintheforegoingsections.Sincemultiprocessorsarejustaparticulartypeofparallelprocessingsystem,theresultsreportedinthesepapersap-plytomultiprocessorsystemstoo.77.Andrews,G.J.andJ.R.McGraw,"LanguageFeaturesforParallelProcessingandResourceControl,"Proc.Conf.De-signandImplementationofProgrammingLanguages,Ithaca,N.Y.,Oct.1976.Describesrequirementsthatalanguageforparallelpro-cessingshouldmeet.Asetoflanguagefeaturesfordescrib-ingprocesses,andtheinteractionsbetweenthem,ispresentedandevaluatedinlightoftheserequirements.Alsopresentslanguagefeaturestocontroltheuseofsharedresourcesinaparallelprocessingenvironment.78.Chen,Shyh-Ching,"SpeedupofInteractiveProgramsinMultiprocessingSystems,"Tech.ReportUIUCDCS-R-75-694,Dept.ofComputerScience,UniversityofIllinois,Urbana,1975.Discussescertainaspectsofrecognizingparallelisminse-quentialprograms.Presentstechniquesforexposingvectoroperationsinloopcomputations,foruseinmultiprocessorsaswellaspipelinedsystems.Derivestheoreticalresultsforthetimeandprocessorboundsofthesetechniquesandforthemaximumspeedupachievablefordifferentloopcom-putations.79.Cohen,T.,"StructuredFlowchartsforMultiprocessing,"ComputerLanguages,Vol.3,No.4,1978,pp.209-226.Describesflowchartconstructsforparallelprocessingprimitivesanddiscussesthetransformationofprogramsusingtheseprimitivesintoequivalentstructuredpro-grams.80.Dowsing,R.D.,"ProcessorManagementinaMultiproces-sorSystem,"ElectronicLetters,Vol.12,No.24,Nov.1976.Comparesmemorymanagementinuniprocessorsystemswithprocessormanagementinmultiprocessorsystemsandshowsthatmanyresultsapplytobothcases.81.Flon,L.andN.Suzuki,"NondeterminismandtheCorrect-nessofParallelPrograms,"Proc.IFIPWorkingConf.For-malDescriptionsofProgrammingConcepts,St.Andrews,N.B.,Canada,Aug.1977.Describesatechniquefortransformingparallelprogramsintoequivalentnondeterministicprogramsandthenprov-ingtheformerbyprovingthelatter.Presentstheproofsofsomecommonlyencounteredparallelprograms.82.Friedman,D.P.andD.S.Wise,"TheImpactofApplicativeProgrammingonMultiprocessing,"Proc.1976Int'lConfParallelProcessing,WaldenWoods,Mich.,Aug.1976,pp.263-272.*83.Gonzalez,M.J.andC.V.Ramamoorthy,"RecognitionandRepresentationofParallelProcessableStreamsinCom-puterPrograms,"inParallelProcessorSystems,TechnologiesandApplications,MacmillanLtd.,London,England,1970.Twotechniquesforexploitingparallelisminprogramsaredevelopedhere.Onetriestodecomposeasequentialpro-gramintotasksthatmaybeexecutedinparallel;theothertriestorecognizeparallelismatafinergrain,withinonetask.84.Gonzalez,M.J.andC.V.Ramamoorthy,"ParallelTaskEx-ecutioninaDecentralizedSystem,"IEEETrans.Com-puters,Vol.C-21,No.12,Dec.1972,pp.1310-1322.Developsatechniqueforrepresentingprogramsegmentsthatmaybeexecutedinparallelanddescribestheuseofthistechniqueinmultiprocessorsystems.Theoverheadsin-volvedinexecutingtasksinparallelareinvestigatedinbothcentralizedanddecentralizedcontrolenvironments.85.Graham,R.L.,"BoundsonMultiprocessingAnomaliesandRelatedPackingAlgorithms,"AFIPSConf.Proc.,Vol.40,1972SJCC,pp.205-217.Surveystheoreticalresultsinmultiprocessinganddiscussesalgorithmsforimprovingmultiprocessorperfor-mance.86.Gries,D.,"AnExerciseinProvingParallelProgramsCor-rect,"Comm.ACM.,Vol.20,No.12,Dec.1977,pp.921-930.COMPUTER106 95.Strovink,E.F.,"CompilationStrategiesforMultiprocessorMessage-PassingSystems,"Proc.SeventhTexasConf.ComputingSystems,Houston,Tex.,Nov.1978,pp.7-15to7-20.*DescribesthesyntaxandimplementationofalanguagefortheMuNetmultiprocessor.Animportantdesigngoalofthislanguageistheeasyexpressionofparallelism.96.Subrahmanyam,P.A.andR..D.Kieburtz,"InterprocessCommunicationandBlockagePropagationinMulti-processorConfigurations,"Proc.TwelfthHawaiiInt'lConfSystemSciences,PartI,Honolulu,Hawaii,Jan.1979.Considerstheinteractionofprocessesandclassifiesinter-processcommunicationtechniquesaccordingtothedegreeofinterprocesssynchronization.MultiprocessorapplicationsPapersinthissectionexaminetheuseofmultiprocessorsinspecificapplications.Thepotentialreliabilityandrobustnessofmultiprocessorsmakethemi,dealcandidatesforaerospaceapplications,whererepairsaredifficultifnotimpossible.Someworkscitedheredescribesuchmultipro-cessorapplications.Othersinvestigateapplicationssuchasdigitalsignalprocessing,simulation,andmessageswitching.Afewdescribeandevaluatespecificalgorithmsimplementedonmultiprocessors.97.Barnwell,T.P.,S.Gaglio,andC.J.M.Hodges,"EfficientImplementationofOneandTwoDimensionalDigitalSignalProcessingAlgorithmsonaMulti-ProcessorAr-chitecture,"Rec.IEEEInt'lConf.Acoustics,SpeechandSignalProcessing,Washington,D.C.,Apr.1979,pp.698-701.Describestheefficientimnplementationofcertaindigitalsignalprocessingalgorithmsonamultiprocessorarchitec-turespecificallydesignedforsuchapplications.98.Baudet,G.M.,"AsynchronousIterativeMethodsforMulti-processors,"J.ACM,Vol.25,No.2,Apr.1978,pp.226-244.Presentstechniquesforexploitingtheparallelismofmultiprocessingsystemsinsolvingasetofequations.Aninterestingfeatureofthealgorithmsdescribedisthattheydonotrequiresynchronizationamongthecooperatingpro-cessesworkingontheproblem.99.Davis,E.W.,"AMicroprocessor-BasedSimulationMa-chine,"Proc.SOUTHEASTCON'78Region3Conf.,Atlan-ta,Ga.,Apr.1978.Thismultiprocessorsystemwasintendedspecificallyforsimulationapplications.Presentstechniquesforusingitsparallelismtospeedupsimulations.100.Erman,L.D.etal.,"SystemOrganizationsforSpeechUn-derstanding:ImplicationsofNetworkandMultiprocessorComputerArchitecturesforAl,"IEEETrans.Computers,Vol.C-25,No.4,Apr.1976,pp.414-421.Discussestheimplementationofaspeech-understandingsystemonmultiprocessorsandonnetworkarchitectures.101.Jones,A.K.etal.,"ProgrammingIssuesRaisedbyaMulti-processor,"Proc.IEEE,Vol.66,No.2,Feb.1978,pp.229-237.ThereareproblemsinherentinwritingsoftwareforCm*.DescribestheStarOSoperatingsystemandtheimplemen-tationofaspecificapplicationunderit.102;Mazare,G.,"AFewExamplesofHowtoUseaSymmetricalMulti-Micro-Processor,"ConfProc.4thAnn.Symp.Com-puterArchitecture,SilverSpring,Md.,Mar.1977,pp.57-62.*-Describesamicroprocessor-basedmultiprocessorsystemandillustratesitsusewithtwoprogramsthatexploittheparallelismpresentinthesystem.103.McGill,R.andJ.Steinhoff,"AMultiprocessorApproachtoNumericalAnalysis:AnApplicationtoGamingProblems,"Conf.Proc.3rdAnn.Symp.ComputerArchitecture,Clear-water,Fla.,Jan.1976,pp.46-51.*Thismultiprocessororganizationisorientedtowardssolvingacertainclassofnumericalanalysisproblems.Thesystemusesaminicomputerhostandasetofmicroproces-sormodules.Describesalgorithmsthatexploitthestruc-tureofthissystemtospeedupcertainproblemsolutions.104.Oleinick,P.N.,"TheImplementationandEvaluationofParallelAlgorithmsonC.mmp,"doctoralthesis,Dept.ofComputerScience,Carnegie-MellonUniversity,Pitts-burgh,Pa.Describesexperimentalperformanceresultsanddis-cussesthefactorsaffectingtheperformance.105.Robinson,J.T.,"AnalysisofAsynchronousMultipro-cessorAlgorithmswithApplicationstoSorting,"Proc.1977Int'lConfParallelProcessing,Detroit,Mich.,Aug.1977,pp.128-135.*106.Soubiron,J.,"Multiple-MicroprocessorSystemsinAt-titudeandOrbitControlSystems,"Proc.FirstInt'lConfAttitudeandOrbitControlSystems,Noordwijk,Netherlands,Oct.1977.Examinestheuseofmulti-microprocessorsystemsforsignalanddataprocessinginspaceapplications,wherereliabilityiscrucial.107.Steele,G.L.,"MultiprocessingCompactifyingGarbageCollection,'"Comm.ACM,VoL18,No.9,Sept.1975,pp.495-508.Presentsalgorithmsforconcurrentlydoingusefulcom-putationandgarbagecollectioninalist-processingenviron-ment,suchasLisp.108.Thomas,T.B.andW.L.Arbuckle,"MultiprocessorSoft-ware:TwoApproaches,"Proc.6thAnn.ConfUseofDigitalComputersinProcessControl,BatonRouge,La.,Feb.1971.Twousesofmultiprocessinginreal-timesystemsarediscussed.Oneapproachusesmultipleprocessorsasbackupstoprovideenhancedreliability,whiletheotherusesthemforimprovingperformance.Discussesthecharacteristicsoftheoperatingsystemineachcase.109.Tippie,J.W.andJ.E.Kulaga,"DesignConsiderationsforaMultiprocessorBasedDataAcquisitionSystem,"IEEETrans.NuclearScience,Vol.NS-26,No.4,Aug.1979,pp.4548-4551.Discussestheuseofamultiprocessorsystemtoachieveahighthroughputindataacquisitionapplications.110.Zellweger,A.,"ComputerArchitecturesforAdvancedAirTrafficControlApplications,"Proc.1976Int'lConf.ParallelProcessing,WaldenWoods,Mich.,Aug.1976,pp.132-139.*Thesemultiprocessorsareusedinhighlyreliableairtraf-ficcontrolsystems;twospecificexamplesaredescribed.SchedulingMultiprocessorschedulinghasbeenafertilesourceofin-terestingproblemsforresearchersincomputerscience,mathematics,andoperationsresearch.Initsmostgeneralform,theprobleminvolvestheschedulingofasetoftaskswitharbitraryprecedenceconstraintsonasetofprocessors*ThesedigestsandproceedingsareavailablefromtheIEEEComputerSoci-etyPublicationsOffice,5855NaplesPlaza,Suite301,LongBeach,CA90803.COMPUTER108 Presentsalineartimealgorithmforcomputingtheop-timalfinishtimeforasetofindependenttasks,orderedbylength,onasetofprocessors,orderedbyspeed.129.Gwynn,J.M.andR.J.Raynor,"SchedulinginaMultipro-cessorEnvironment,"Proc.1973SagamoreComputerConfParallelProcessing,Sagamore,N.Y.,Aug.1973,p.139.*Examinesinterrupthandlingtechniquesinmultiproces.sorsystems.Describesschemessuchasmaster-slavecon-trolandfloatingexecutivecontrolanddiscussestheirim-pactonthequeueingdelaysinhandlinginterrupts.130.Horvath,E.C.,S.Lam,andR.Sethi,"ALevelAlgorithmforPreemptiveScheduling,"J.ACM,Vol.24,No.1,Jan;1977,pp.32-43.Extendsearlierworkonthepreemptiveschedulingofasetoftaskswhoseprecedencegraphisatree,onprocessorswithdifferentspeeds.Presentsanalgorithmwhichisop-timalforthetwo-processorcaseandshowsthat,inothercases,thealgorithmisoptimalonlywhenthetasksarein-dependent.131.Jensen,J.E.,"AFixed-VariableSchedulingModelforMultiprocessors,"Proc.1977Int'lConf.ParallelProcess-ing,Detroit,Mich.,Aug.1977,pp.108-117.*Thisadaptiveschedulingalgorithmselectsaschedulingstrategyappropriatetotheloadonthesystem.Thealgorithmusesasimplisticbutcomputationallycheapap-proachwhenprocessorcyclesarescarce.Itadoptsmorecomplexschedulingstrategiesasprocessorcyclesbecomeavailable.132.Kafura,D.G.,"RelationshipBetweenWorst-CaseandEx-pectedSchedulingPerformanceforaModelofaMultipro-cessorSystem,"Proc.ComputerScienceConf.,Wash-ington,D.C.,Feb.1975.133.Kafura,D.G.andV.Y.Shen,"TaskSchedulingonaMultiprocessorSystemwith-IndependentMemories,"SIAMJ.Computing,Vol.6,No.1,Mar.1977,pp.167-187.Considersschedulinginamultipleprocessorsystemwhereallprocessorsareidenticalandeachprocessorhasitsown,privatememory.Examinespreemptiveandnon-preemptiveschedulingschemesfortaskswithknownresourcerequirements,andevaluatestheseschemesbothanalyticallyandbysimulation.134.Kohler,W.H.,"APreliminaryEvaluationoftheCriticalPathMethodforSchedulingTasksonMultiprocessorSystems,"IEEETrans.Computers,Vol.C-24,No.12,Dec.1975,pp.1235-1238.Expressestheproblemof-schedulingtasksona-setofin-dependent,identicalprocessorsinagraph-theoreticframe.workand,usingacriticalpathmethod,derivesanapprox-imationtotheoptimalsolution.135.Lang,T.andE.B.Fernandez,"SchedulingofUnit-LengthIndependentTaskswithExecutionConstraints,"Informa-tionProcessingLetters,Vol.4,No.4,Jan.1976,pp.95-98.136.Liu,J.W.S.andC.L.Liu,"PerformanceAnalysisofMultiprocessorSystemsContainingFunetionallyDedicatedProcessors,"ActaInformatica,Vol.10,No.1,1978,pp.95-104Presentsmodelstodescribetheschedulingofmultiprocessingsystemshavingspecializedprocessors.137.Martin-Vega,L.A.andH.D.Ratliff,"SchedulingParallelProcessors,"BulLOperationsResearchSoc.America,Vol23,Suppl.2,Nov.1975,-p.B-418.Investigatesthedifferencesbetweenpreemptiveand-non-preemptiveschedulingalgorithms.Showsthat,forasubclassofschedulingproblems,therelationshipbetweenpreemptiveandnon-preemptiveoptimalsolutionsisisomorphictotherelationshipamongoptimalsolutionstoacertaintypeoflinearprogrammingproblem.138.Mullen,J.K.,"ParallelDerivativesforMultiprocessorTaskScheduling,"ReportNo.MTR-2922,MitreCorp.,McLean,Va.,May1975.Thistechniquetransformsanarbitraryflowchartintotheequivalentmaximallyparallelflowchartandusesittoderivealternativeschedulingstrategiesforamulti-miniprocessor.139.Rowicki,A.,"ANoteonOptimal-'SchedulingforTwo-ProcessorSystems,"Informaiion,ProcessingLetters,Vo1.4,No.2,Nov.1975,pp.27-30.Describesanalgorithmfortheoptimalnon-preemptiveschedulingofidenticaltasksonidenticalprocessors.140.Schindler,S.andH.Ludtke,''AnApproachtoaRestrictedScheduling-ProblemforMultiprocess6rSystems,."Proc.1973SagamoreComputerConfParallelProcessing,Sagamore,N.Y.,Aug.1973,pp.121-129.*Discussesthepreemptiveschedulingoftaskswithprecedenceconstraintsonafixedsetofidenticalpro-cessors,withallotherresourcesunlimited.141.Sethi,R.,"SchedulingGraphsonTwoProcessors,"SIAMJ.Computing,Vol.5,No.1,Mar.1976,pp.73-82.Presentstwoalgorithmsforschedulingequal-lengthtaskswithprecedenceconstraintsontwo-processorsystems.142.Stone,H.S.,"MultiprocessorSchedulingwiththeAidofNetworkFlowAlgorithms,"IEIEETrans.SoftwareEng.,Vol.SE-3,No.1,Jan.1977,pp.85-93.Theseschedulingalgorithms,basedontheFord-Fulkersonnetworkflowalgorithm,minimiieinterprocessorcommunicationinamultiprocessorsystem.Thetwo-processorcaseistreatedindetail,andpartialresultsarepresentedforthegeneraicase.143.Yang,Chao-Chih,"FastAlgorithmsforBoundingthePer-formanceofMultiprocessorSystems,"Proc.1976Int'lConfParallelProcessing,WaldenWoods,Mich.,Aug.1976,pp.73-82.*Presentsefficientalgorithmsforfindingthelowerboundonperformancewhenschedulingasetoftasks.Theperfor-mancemeasuresherearethenumberofprocessorsusedandthetotalexecutiontimeforthesetoftasks.144.Yao,A.Chi-Chih,"SchedulingUnit-TimeTaskswithLimitedResources,"Proc.ConfParallelProcessing,Ra-quetteLake,N.Y.,Aug.1974.Givesheuristicalgorithmsforoptimallyschedulingasetoftaskswithdifferentresourcerequirements,butwithiden-ticalprocessingtimes.ReliabilityanderrorrecoveryPapersinthissectioninvestigatethereliability.ofmultiprocessorsystemsanddescribetechniquestoutilizeavailableredundancy.Sometrytodefinefiguresofmeritforsuchentitiesas"reliability"and"fault-tolerance,"sothatquantitivecomparisonsmaybemade.Techniquestohandlehardwareandsoftwarefailures,andtopermitgracefuldegradationincaseofsuchfailures,arealsodescribed.Reliabilitymodelsandtheiruseinpredictingfailureratesarethefocusofanumberoftheworkscitedhere.145.Bertera,F.,A.Boccalatte,andM.DiManzco,"AFault-Tol-erantApproachtoProcessorSynchronizationinaMulti-*ThesedigestsandproceedingsareavailablefromtheIEEEComputerSoci-etyPublicationsOffice,5855NaplesPlaza,Suite301,LongBeach,CA90803.112CCOMPUTER Asoftwaretechniqueusesvotingamongthreeprocessorsperformingidenticaltaskstodetectcomputationalerrorsandprocessorfailure.Differentapproachestothisproblemareproposedandtheirrelativemeritsdiscussed.MiscellaneoustopicsinmultiprocessingIncludedhereareworksIfeltwereofinterest,butwhichdidnotfallintoanyoftheearliercategories.Somearegeneralsurveysofmultiprocessing.Othersdescribeex-periencesinusingmultiprocessors.Afewgiveadviceontheappropriateuseofmultiprocessing.164.Baer,J.L.,"MultiprocessingSystems,"IE-EETrans.Com-puters,Vol.C-25,No.12,Dec.1976,pp.1271-1277.Examinesandclassifiesarrayprocessorsandmultiprocessorsaccordingtothetightnessofcouplingbe-tweenindividualprocessors.Discussessoftwareenhance-mentsneededtoexploitthepotentialofmultiprocessorsystems.165.Childs,R.E.,"MultipleMicroprocessorSystems:Goals,LimitationsandAlternatives,"ExplodingTechnology,ResponsibleGro`wth-DigestofPapers-COMPCONSpring79,SanFrancisco,Calif.,Mar.1979,pp.94-97.*Surveysthemeritsandappropriatenessofusingmulti-processinginmicroprocessorsystems.166.Enslow,P.H.,ed.,MultiprocessorsandParallelProcessing,JohnWileyandSons,NewYork,1974.Describeshardwareandsoftwareissuesinvolvedinmultiprocessing.Anappendixincludesbriefdescriptionsofanumberofmultiprocessingsystems.167.Enslow,P.H.,"MultiprocessorOrganization-ASurvey,"ComputingSurveys,Vol.9,No.1,Mar.1977,pp.103-129.Awell-writtenandoft-quotedsurveyofmultiprocessing.Describesavarietyofprocessor-memoryinterconnectionschemesaswell'asdifferentopera-tingsystemstructures.168.Ferreira,R.C.andD.Vojnovic,"Multiminicomputers:APerspectiveontheNextFiveYears,"inFutureSystems,StateoftheArtReportInfotechLtd.,Maidenhead,England,1977.Predictsthefutureofmultiprocessorsthatarebuiltwithminicomputers,e.g.,C.mmp.169.Franklin,M.A.,S.A.Kahan,andM.J.Stucki,"DesignIssuesintheDevelopmentofaModularMultiprocessorCommunicationsNetwork,"Conf.Proc.6thAnn.Symp.ComputerArchitecture,Philadelphia,Pa.,Apr.1979,pp.182-187.*Describesamodular,easilyexpandablecrossbarnetworkforuseinmultiprocessorsystems.170.Hardcastle,A.R.K.,"Multi-MinisversusLargeMain-frames,"inMinisversusMainframes,StateoftheArtReport,InfotechLtd.,Maidenhead,England,1978.Discussestheprosandconsofusingmulti-miniproces-sorsinlieuoflargemainframesforcommercialdataprocess-ing.171.Harris,J.A.andD.R.Smith,"HierarchicalMultiprocessorOrganizations,"Conf.Proc.4thAnn.Symp.ComputerAr-chitecture,SilverSpring,Md.,Mar.1977,pp.41-48.*Examinesthebenefitsofmulti-microprocessororganiza-tionsandtheproblemsencounteredinsuchsystems.Describesaspecificmultiprocessordesign,basedonmicro-processors,andinvestigatesitssuitabilityforsomeclassesofproblems.172.Jones,A.K.andP.Schwarz,"ExperienceUsingMultipro-cessorSystems:AStatusReport,"Tech.ReportCMU-CS-79-146,Dept.ofComputerScience,Carnegie-MellonUniversity,Pittsburgh,Pa.,Oct.1976.Thediscussionhereisbasedonactualexperienceinbuildingandusingmultiprocessors;itthusfallsintoarelativelysmallgroupofexperience-orientedpapers.173.Leiberman,M.,"ToMultiprocessorNottoMultiprocess,"Proc.FourteenthMeetingComputerPerformanceEvalua-tionUsersGroup,Boston,Mass.,Oct.1978.Considersthedecisiontorunmultipleprocessorsasamultiprocessingsystemorasindependentsystems.174.Mazare,G.,"MultiprocessorSystems,"Proc.1974CERNSchoolofComputing(EuropeanOrganizationforNuclearResearch),Godysund,Norway,Aug.1974.Adetaileddescriptionofthedevelopmentandprinciplesofmultiprocessingsystems.175.Patel,J.H.,"Processor-MemoryInterconnectionsforMultiprocessors,"ConfProc.6thAnn.Symp.ComputerArchitecture,Philadelphia,Pa.,Apr.1979,pp.168-177.*Describesaninterconnectionstructure-morecost-effectivethanacrossbarswitch-thatpermitseverypro-cessortobeconnectedtoeverymemorymodule.176.Siewiorek,D.P.andM.R.Barbacci,"ModularityandMultiprocessorStructures-SomeOpenProblemsintheConstructionandUtilizationofMini-andMicroprocessorNetworks,"inDistributedSystems,InternationalStateoftheArtReport,InfotechLtd.,Maidenhead,England,1976.Presentsataxonomyofthemultiprocessordesignspaceanddiscussesthreemultiprocessordesignsinthislight.Discussesbothhardwareandsoftwareissuesrelatedtothethreedesigns.177.Srodawa,R.J.,"PositiveExperienceswithaMultiprocess-ingSystem,"ComputingSurveys,Vol.10,No.1,Mar.1978,pp.73-82.Examinestheexperiencegainedfromusingmultiprocess-ingintheMichiganTerminalSystem.Basedonthisex-perience,evaluatestheoptionsavailabletothedesignerofamultiprocessingsystem.Thispaperisoneofthefewpiecesofwidelycirculatedliteraturedescribingactual.multipro-cessorexperienceratherthansystemdesignormodeling.178.White,C.H.,ed.,MultiprocessorSystems,InternationalComputerStateoftheArtReport,InfotechLtd.,Maid-enhead,England,1976.Acollectionofarticlesonthestateoftheartofmultiprocessingin1976.Mostoftheinformationisstillvalidandrelevant.179.Wulf,W.A.andS.P.Harbison,"ReflectionsinaPoolofProcessors,"Tech.ReportNo.CMU-CS-78-103,Dept.ofComputerScience,Carnegie-MellonUniversity,Pitts-burgh,Pa.,Feb.1978.AcandidevaluationofthesuccessesandshortcomingsoftheC.mmpproject.Thereportisrefreshinglyfrankandpullsnopunches.*ThesedigestsandproceedingsareavailablefromtheIEEEComputerSoci-etyPublicationsOffice,5855NaplesPlaza,Suite301,LongBeach,CA90803.M.Satyanarayananistheauthorofthesurveyaccompanyingthisbibliography;hisphotographandbiographyappearonpage96.COMPUTER116