计算机组成原理课件英文版第一章

Chapter1,ComputerAbstractionsandTechnology,Chapter1ComputerAbstractionsandTechnology2,TheComputerRevolution,ProgressincomputertechnologyUnderpinnedbyMooresLawWhatisMooresLaw?Mooreslawdescribesalong-termtrendinthehistoryofcomputinghardware.Thequantityoftransistorsthatcanbeplacedinexpensivelyonanintegratedcircuithasdoubledapproximatelyeverytwoyears.,1.1Introduction,MooresLaw,Chapter1ComputerAbstractionsandTechnology3,Thetrendhascontinuedformorethanhalfacenturyandisnotexpectedtostopuntil2015orlater.,Chapter1ComputerAbstractionsandTechnology4,TheComputerRevolution,MakesnovelapplicationsfeasibleComputersinautomobiles,1.1Introduction,Chapter1ComputerAbstractionsandTechnology5,TheComputerRevolution,MakesnovelapplicationsfeasibleCellphones,1.1Introduction,Chapter1ComputerAbstractionsandTechnology6,TheComputerRevolution,MakesnovelapplicationsfeasibleHumangenomeprojectWorldWideWebSearchEngines,1.1Introduction,Chapter1ComputerAbstractionsandTechnology7,TheComputerRevolution,Computersarepervasive,1.1Introduction,Chapter1ComputerAbstractionsandTechnology8,ClassesofComputers,Question:Howdoyouclassifycomputers?DesktopComputersServerComputersEmbeddedComputers,Chapter1ComputerAbstractionsandTechnology9,ClassesofComputers,DesktopcomputersPCGeneralpurpose,varietyofsoftwareSubjecttocost/performancetradeoff,Chapter1ComputerAbstractionsandTechnology10,ClassesofComputers,ServercomputersNetworkbasedHighcapacity,performance,reliabilityRangefromsmallserverstobuildingsized,Worldssmallestwebserver,Chapter1ComputerAbstractionsandTechnology11,ClassesofComputers,EmbeddedcomputersHiddenascomponentsofsystemsStringentpower/performance/costconstraints,Chapter1ComputerAbstractionsandTechnology12,TheProcessorMarket,Chapter1ComputerAbstractionsandTechnology13,WhatYouWillLearn,HowprogramsaretranslatedintothemachinelanguageAndhowthehardwareexecutesthemThehardware/softwareinterfaceWhatdeterminesprogramperformanceAndhowitcanbeimprovedHowhardwaredesignersimproveperformanceWhatisparallelprocessing,Chapter1ComputerAbstractionsandTechnology14,LevelsofProgramCode,High-levellanguageLevelofabstractionclosertoproblemdomainProvidesforproductivityandportabilityAssemblylanguageTextualrepresentationofinstructionsHardwarerepresentationBinarydigits(bits)Encodedinstructionsanddata,Chapter1ComputerAbstractionsandTechnology15,BelowYourProgram,ApplicationsoftwareWritteninhigh-levellanguage(HLL)SystemsoftwareCompiler:translatesHLLcodetomachinecodeOperatingSystem:servicecodeHandlinginput/outputManagingmemoryandstorageSchedulingtasks&sharingresourcesHardwareProcessor,memory,I/Ocontrollers,1.2BelowYourProgram,Chapter1ComputerAbstractionsandTechnology16,UnderstandingPerformance,AlgorithmDeterminesnumberofoperationsexecutedProgramminglanguage,compiler,architectureDeterminenumberofmachineinstructionsexecutedperoperationProcessorandmemorysystemDeterminehowfastinstructionsareexecutedI/Osystem(includingOS)DetermineshowfastI/Ooperationsareexecuted,Chapter1ComputerAbstractionsandTechnology17,ComponentsofaComputer,SamecomponentsforallkindsofcomputerDesktop,server,embeddedInput/outputincludesUser-interfacedevicesDisplay,keyboard,mouseStoragedevicesHarddisk,CD/DVD,flashNetworkadaptersForcommunicatingwithothercomputers,1.3UndertheCovers,TheBIGPicture,Chapter1ComputerAbstractionsandTechnology18,Anatomy(结构)ofaComputer,Outputdevice,Inputdevice,Inputdevice,Networkcable,Chapter1ComputerAbstractionsandTechnology19,AnatomyofaMouse,OpticalmouseLEDilluminatesdesktopSmalllow-rescameraBasicimageprocessorLooksforx,ymovementButtons&wheelSupersedesroller-ballmechanicalmouse,Chapter1ComputerAbstractionsandTechnology20,ThroughtheLookingGlass,LCDscreen:pictureelements(pixels)Mirrorscontentofframebuffermemory,Chapter1ComputerAbstractionsandTechnology21,OpeningtheBox,Chapter1ComputerAbstractionsandTechnology22,InsidetheProcessor(CPU),Datapath:performsoperationsondataControl:sequencesdatapath,memory,.CachememorySmallfastSRAMmemoryforimmediateaccesstodataSRAMStaticRandomAccessMemory,Chapter1ComputerAbstractionsandTechnology23,InsidetheProcessor,AMDBarcelona:4processorcores,Chapter1ComputerAbstractionsandTechnology24,ASafePlaceforData,Volatile(易变的)mainmemoryLosesinstructionsanddatawhenpoweroffNon-volatilesecondarymemoryMagneticdiskFlashmemoryOpticaldisk(CDROM,DVD),Chapter1ComputerAbstractionsandTechnology25,Networks,CommunicationandresourcesharingLocalareanetwork(LAN):EthernetWithinabuildingWideareanetwork(WAN):theInternet,Chapter1ComputerAbstractionsandTechnology26,Abstractions,InstructionSetArchitecture(ISA)Aninterfacebetweenthehardwareandthelowest-levelsoftwareTheabstractimageofacomputingsystemthatisseenbyamachine/assemblylanguageprogrammerIncludinginstructions,registers,memoryaccess,I/O,TheBIGPicture,ISAs,System/360andupwardscompatiblesuccessorsz/ArchitecturePowerArchitecturePDP-11SPARCSuperHTricoreTransputerUNIVAC1100/2200seriesVAXx86IA-32(32-bitx86,firstimplementedintheIntel80386)x86-64(64-bitsupersetofIA-32,firstimplementedintheAMDOpteron)EISC(AE32K),Chapter1ComputerAbstractionsandTechnology27,4004,40406800,6502,6809,68HC11,68HC08.8008,8080,8085,Z80,Z180,eZ80,etc.8048,8051,etc.Z8,eZ8,etc.AlphaARMBurroughsB5000seriesBurroughsB6000/B7000serieseSi-RISCIA-64(Itanium)Mico32MIPSMotorola68kPA-RISCIBM700/7000lines,Chapter1ComputerAbstractionsandTechnology28,Abstractions,ApplicationBinaryInterface(ABI)thelow-levelinterfacebetweenanapplicationprogramandOSABIscoverdetailssuchasdatatype,size,alignment,callingconvention,binaryformatofobjectfiles,etc.Definesastandardforbinaryportabilityacrosscomputers.,Chapter1ComputerAbstractionsandTechnology29,Abstractions,ImplementationHardwarethatobeysthearchitectureabstractionManyimplementationsforthesameISAExample:IntelPentiumvsAMDAthlon,almostidenticalISA,butradicallydifferentinternaldesigns.,PERFORMANCE,Chapter1ComputerAbstractionsandTechnology30,Chapter1ComputerAbstractionsandTechnology31,DefiningPerformance,Whichairplanehasthebestperformance?,1.4Performance,Chapter1ComputerAbstractionsandTechnology32,DefiningPerformance,Whichairplanehasthebestperformance?,1.4Performance,DefiningPerformance,Performance=SpeedE.g.,takingasinglepassengerfromonepointtoanotherWinner:ConcordePerformance=PassengerThroughputPassengersxm.p.hE.g.,transporting450passengersfromonepointtoanotherWinner:Boeing747,Chapter1ComputerAbstractionsandTechnology33,Chapter1ComputerAbstractionsandTechnology34,ResponseTimeandThroughput,ResponsetimeHowlongittakestodoataskThroughputTotalworkdoneperunittimee.g.,tasks/transactions/perhour,Chapter1ComputerAbstractionsandTechnology35,ResponseTimeandThroughput,HowareresponsetimeandthroughputaffectedbyReplacingtheprocessorwithafasterversion?Addingmoreprocessors?Wellfocusonresponsetimefornow,Chapter1ComputerAbstractionsandTechnology36,RelativePerformance,DefinePerformance=1/ExecutionTime“XisntimefasterthanY”,Example:timetakentorunaprogram10sonA,15sonBExecutionTimeB/ExecutionTimeA=15s/10s=1.5SoAis1.5timesfasterthanB,Chapter1ComputerAbstractionsandTechnology37,MeasuringExecutionTime,ElapsedtimeTotalresponsetime,includingallaspectsProcessing,I/O,OSoverhead,idletimeDeterminessystemperformanceCPUtimeTimespentprocessingagivenjobDiscountsI/Otime,otherjobssharesComprisesuserCPUtimeandsystemCPUtimeDifferentprogramsareaffecteddifferentlybyCPUandsystemperformance,Chapter1ComputerAbstractionsandTechnology38,CPUClocking,Operationofdigitalhardwaregovernedbyaconstant-rateclock,Clock(cycles),Datatransferandcomputation,Updatestate,Clockperiod,Clockcycletime:durationofaclockcyclee.g.,250ps=0.25ns=2501012sClockfrequency(clockrate):cyclesperseconde.g.,4.0GHz=4000MHz=4.0109Hz,Chapter1ComputerAbstractionsandTechnology39,CPUTime,PerformanceimprovedbyReducingnumberofclockcyclesIncreasingclockrateHardwaredesignermustoftentradeoffclockrateagainstcyclecount,Chapter1ComputerAbstractionsandTechnology40,CPUTimeExample,ComputerA:2GHzclock,10sCPUtimeDesigningComputerBAimfor6sCPUtimeCandofasterclock,butcauses1.2clockcyclesHowfastmustComputerBsclockratebe?,Chapter1ComputerAbstractionsandTechnology41,InstructionCountandCPI,InstructionCountforaprogramDeterminedbyprogram,ISAandcompilerAveragecyclesperinstructionDeterminedbyCPUhardwareIfdifferentinstructionshavedifferentCPIAverageCPIaffectedbyinstructionmix,CPI,Chapter1ComputerAbstractionsandTechnology42,CPIExample,ComputerA:CycleTime=250ps,CPI=2.0ComputerB:CycleTime=500ps,CPI=1.2SameISAWhichisfaster,andbyhowmuch?,Aisfaster,bythismuch,Chapter1ComputerAbstractionsandTechnology43,CPIinMoreDetail,Ifdifferentinstructionclassestakedifferentnumbersofcycles,WeightedaverageCPI,Relativefrequency,Chapter1ComputerAbstractionsandTechnology44,CPIExample,AlternativecompiledcodesequencesusinginstructionsinclassesA,B,C,Sequence1:IC=5ClockCycles=21+12+23=10Avg.CPI=10/5=2.0,Sequence2:IC=6ClockCycles=41+12+13=9Avg.CPI=9/6=1.5,Whatisavg.CPI?,IC=InstructionCount,Chapter1ComputerAbstractionsandTechnology45,PerformanceSummary,PerformancedependsonAlgorithm:affectsIC,possiblyCPIProgramminglanguage:affectsIC,CPICompiler:affectsIC,CPIInstructionsetarchitecture:affectsIC,CPI,Tc,TheBIGPicture,POWER,Chapter1ComputerAbstractionsandTechnology46,Chapter1ComputerAbstractionsandTechnology47,PowerTrends,InCMOSICtechnology,1.5ThePowerWall,1000,30,5V1V,Chapter1ComputerAbstractionsandTechnology48,ReducingPower,SupposeanewCPUhas85%ofcapacitiveloadofoldCPU15%voltageand15%frequencyreduction,ThepowerwallWecantreducevoltagefurtherWecantremovemoreheatHowelsecanweimproveperformance?,Chapter1ComputerAbstractionsandTechnology49,UniprocessorPerformance,1.6TheSeaChange:TheSwitchtoMultiprocessors,Constrainedbypower,instruction-levelparallelism,memorylatency,Chapter1ComputerAbstractionsandTechnology50,Multiprocessors,MulticoremicroprocessorsMorethanoneprocessorperchipRequiresexplicitlyparallelprogrammingComparewithinstructionlevelparallelismHardwareexecutesmultipleinstructionsatonceHiddenfromtheprogrammerHardtodoProgrammingforperformanceLoadbalancingOptimizingcommunicationandsynchronization,Manufacturing,Chapter1ComputerAbstractionsandTechnology51,Chapter1ComputerAbstractionsandTechnology52,ManufacturingICs,Yield:proportionofworkingdiesperwafer,1.7RealStuff:TheAMDOpteronX4,Chapter1ComputerAbstractionsandTechnology53,AMDOpteronX2Wafer,X2:300mmwafer,117chips,90nmtechnologyX4:45nmtechnology,Chapter1ComputerAbstractionsandTechnology54,IntegratedCircuitCost,NonlinearrelationtoareaanddefectrateWafercostandareaarefixedDefectratedeterminedbymanufacturingprocessDieareadeterminedbyarchitectureandcircuitdesign,Benchmarking,Chapter1ComputerAbstractionsandTechnology55,Chapter1ComputerAbstractionsandTechnology56,SPECCPUBenchmark,ProgramsusedtomeasureperformanceSupposedlytypicalofactualworkloadStandardPerformanceEvaluationCorp(SPEC)DevelopsbenchmarksforCPU,I/O,Web,SPECCPU2006ElapsedtimetoexecuteaselectionofprogramsNegligibleI/O,sofocusesonCPUperformanceNormalizerelativetoreferencemachineSummarizeasgeometricmeanofperformanceratiosCINT2006(integer)andCFP2006(floating-point),Chapter1ComputerAbstractionsandTechnology57,CINT2006forOpteronX42356,Chapter1ComputerAbstractionsandTechnology58,SPECPowerBenchmark,PowerconsumptionofserveratdifferentworkloadlevelsPerformance:ssj_ops/secPo