【文档说明】计算机组成原理英文版第一章Computer-Abstractions-and-Technology--课件.ppt，共(63)页，8.526 MB，由小橙橙上传

转载请保留链接：https://www.ichengzhen.cn/view-76280.html

以下为本文档部分文字说明：

Chapter1ComputerAbstractionsandTechnologyTheComputerRevolutionProgressincomputertechnologyUnderpinnedbyMoor

e’sLawWhatisMoore’sLaw???Moore'slawdescribesalong-termtrendinthehistoryofcomputinghardware.Thequantityoftransistorsthatcanbeplacedinexpensivelyon

anintegratedcircuithasdoubledapproximatelyeverytwoyears.§1.1Introductionppt课件Moore’sLawThetrendhascontinuedformorethanhalfacenturyandisnotexpe

ctedtostopuntil2015orlater.ppt课件TheComputerRevolutionMakesnovelapplicationsfeasibleComputersinautomo

biles§1.1Introductionppt课件TheComputerRevolutionMakesnovelapplicationsfeasibleCellphones§1.1Introductionp

pt课件TheComputerRevolutionMakesnovelapplicationsfeasibleHumangenomeprojectWorldWideWebSearchEngines§1.1Introductionppt课件TheComputerRevolut

ionComputersarepervasive§1.1Introductionppt课件ClassesofComputersQuestion:Howdoyouclassifycomputers?DesktopComputersServe

rComputersEmbeddedComputersppt课件ClassesofComputersDesktopcomputersPCGeneralpurpose,varietyofsoftwareSubjecttocost/performancetradeof

fppt课件ClassesofComputersServercomputersNetworkbasedHighcapacity,performance,reliabilityRangefromsmallserverstobuil

dingsizedWorld’ssmallestwebserverppt课件ClassesofComputersEmbeddedcomputersHiddenascomponentsofsystemsStringentpower/performance/costconstrain

tsppt课件TheProcessorMarketppt课件WhatYouWillLearnHowprogramsaretranslatedintothemachinelanguageAndhowthehardwareexecutesthemThehardware/soft

wareinterfaceWhatdeterminesprogramperformanceAndhowitcanbeimprovedHowhardwaredesignersimproveperformanceWhatisparallel

processingppt课件LevelsofProgramCodeHigh-levellanguageLevelofabstractionclosertoproblemdomainProvidesforproductivityandportabilityAssemblylan

guageTextualrepresentationofinstructionsHardwarerepresentationBinarydigits(bits)Encodedinstructionsanddatappt课件BelowYourProgramApp

licationsoftwareWritteninhigh-levellanguage(HLL)SystemsoftwareCompiler:translatesHLLcodetomachineco

deOperatingSystem:servicecodeHandlinginput/outputManagingmemoryandstorageSchedulingtasks&sharingresourcesHardwareP

rocessor,memory,I/Ocontrollers§1.2BelowYourProgramppt课件UnderstandingPerformanceAlgorithmDeterminesnumberofoperationsexecutedProgramming

language,compiler,architectureDeterminenumberofmachineinstructionsexecutedperoperationProcessorandmemor

ysystemDeterminehowfastinstructionsareexecutedI/Osystem(includingOS)DetermineshowfastI/Ooperationsareexecutedp

pt课件ComponentsofaComputerSamecomponentsforallkindsofcomputerDesktop,server,embeddedInput/outputincludesUser-interfacedevices

Display,keyboard,mouseStoragedevicesHarddisk,CD/DVD,flashNetworkadaptersForcommunicatingwithothercomputers§1.3Underth

eCoversTheBIGPictureppt课件Anatomy(结构)ofaComputerOutputdeviceInputdeviceInputdeviceNetworkcableppt课件AnatomyofaMouse

OpticalmouseLEDilluminatesdesktopSmalllow-rescameraBasicimageprocessorLooksforx,ymovementButtons&wheelSupersedesroller-ballmech

anicalmouseppt课件ThroughtheLookingGlassLCDscreen:pictureelements(pixels)Mirrorscontentofframebuffermemoryppt课件OpeningtheBoxppt课件In

sidetheProcessor(CPU)Datapath:performsoperationsondataControl:sequencesdatapath,memory,...CachememorySmallfastSRAMmemoryforimmediateaccesst

odataSRAM–StaticRandomAccessMemoryppt课件InsidetheProcessorAMDBarcelona:4processorcoresppt课件ASafePlac

eforDataVolatile(易变的)mainmemoryLosesinstructionsanddatawhenpoweroffNon-volatilesecondarymemoryMagneticdiskFlashmemoryOpticald

isk(CDROM,DVD)ppt课件NetworksCommunicationandresourcesharingLocalareanetwork(LAN):EthernetWithinabuildingW

ideareanetwork(WAN):theInternetppt课件AbstractionsInstructionSetArchitecture(ISA)Aninterfacebetweenthehardwareandthelowest-levelsoftw

areTheabstractimageofacomputingsystemthatisseenbyamachine/assemblylanguageprogrammerIncludinginstructions,regis

ters,memoryaccess,I/O,…TheBIGPictureppt课件ISAsSystem/360andupwardscompatiblesuccessorsz/ArchitecturePowerArchitectur

ePDP-11SPARCSuperHTricoreTransputerUNIVAC1100/2200seriesVAXx86IA-32(32-bitx86,firstimplemente

dintheIntel80386)x86-64(64-bitsupersetofIA-32,firstimplementedintheAMDOpteron)EISC(AE32K)4004,40406800,6502,6809,68HC11,68HC08.8008,8080,8085,Z

80,Z180,eZ80,etc.8048,8051,etc.Z8,eZ8,etc.AlphaARMBurroughsB5000seriesBurroughsB6000/B7000serieseSi-RISCIA-64(Itanium)

Mico32MIPSMotorola68kPA-RISCIBM700/7000linesppt课件AbstractionsApplicationBinaryInterface(ABI)thelow-levelinter

facebetweenanapplicationprogramandOSABIscoverdetailssuchasdatatype,size,alignment,callingconvention,binaryformatofob

jectfiles,etc.Definesastandardforbinaryportabilityacrosscomputers.ppt课件AbstractionsImplementationHardwarethatobeysthearchi

tectureabstractionManyimplementationsforthesameISAExample:IntelPentiumvsAMDAthlon,almostidenticalISA,butradica

llydifferentinternaldesigns.ppt课件PERFORMANCEppt课件DefiningPerformanceWhichairplanehasthebestperformance?§1.4Performanceppt课件DefiningPerformance

Whichairplanehasthebestperformance?0100200300400500DouglasDC-8-50BAC/SudConcordeBoeing747Boeing777PassengerCapacity02000400060008000

10000DouglasDC-8-50BAC/SudConcordeBoeing747Boeing777CruisingRange(miles)050010001500DouglasDC-8-50BAC/SudConcordeBoeing74

7Boeing777CruisingSpeed(mph)0100000200000300000400000DouglasDC-8-50BAC/SudConcordeBoeing747Boeing777Passengersxmph§1.4Perfo

rmanceppt课件DefiningPerformancePerformance=SpeedE.g.,takingasinglepassengerfromonepointtoanotherWinner

:ConcordePerformance=PassengerThroughputPassengersxm.p.hE.g.,transporting450passengersfromonepointtoanotherWinner:Boeing

747ppt课件ResponseTimeandThroughputResponsetimeHowlongittakestodoataskThroughputTotalworkdoneperunittimee.g.,tasks/transactions/…perhou

rppt课件ResponseTimeandThroughputHowareresponsetimeandthroughputaffectedbyReplacingtheprocessorwithafasterversion?

Addingmoreprocessors?We’llfocusonresponsetimefornow…ppt课件RelativePerformanceDefinePerformance=1/ExecutionTime“XisntimefasterthanY”nXY

YXtimeExecutiontimeExecutionePerformancePerformancExample:timetakentorunaprogram10sonA,15sonBExecu

tionTimeB/ExecutionTimeA=15s/10s=1.5SoAis1.5timesfasterthanBppt课件MeasuringExecutionTimeElapsedtimeTotalre

sponsetime,includingallaspectsProcessing,I/O,OSoverhead,idletimeDeterminessystemperformanceCPUtimeTimespentprocessingagivenj

obDiscountsI/Otime,otherjobs’sharesComprisesuserCPUtimeandsystemCPUtimeDifferentprogramsareaffecteddifferentlybyCPUandsystemperformanceppt课件CPU

ClockingOperationofdigitalhardwaregovernedbyaconstant-rateclockClock(cycles)DatatransferandcomputationUpdatestateClockperiodC

lockcycletime:durationofaclockcyclee.g.,250ps=0.25ns=250×10–12sClockfrequency(clockrate):cyclesperseconde.g.,4.0GHz=4000MHz=4.0×109Hzppt课件CPUTime

PerformanceimprovedbyReducingnumberofclockcyclesIncreasingclockrateHardwaredesignermustoftentrad

eoffclockrateagainstcyclecountRateClockCyclesClockCPUTimeCycleClockCyclesClockCPUTimeCPUppt课件CPUTimeExampleComputerA:2GHzclock,10sCPUtimeDesign

ingComputerBAimfor6sCPUtimeCandofasterclock,butcauses1.2×clockcyclesHowfastmustComputerB’sclockra

tebe?4GHz6s10246s10201.2RateClock10202GHz10sRateClockTimeCPUCyclesClock6sCyclesClock1.2TimeCPUCyclesClockRateCloc

k99B9AAAABBBppt课件InstructionCountandCPIInstructionCountforaprogramDeterminedbyprogram,ISAandcompilerAveragecyclesper

instructionDeterminedbyCPUhardwareIfdifferentinstructionshavedifferentCPIAverageCPIaffectedbyinstructionmixRat

eClockCPICountnInstructioTimeCycleClockCPICountnInstructioTimeCPUnInstructioperCyclesCountnInstructioCyclesClockCPIppt课件C

PIExampleComputerA:CycleTime=250ps,CPI=2.0ComputerB:CycleTime=500ps,CPI=1.2SameISAWhichisfaster,andbyhowmuch?1.2500psI600psIATimeCPUBTimeC

PU600psI500ps1.2IBTimeCycleBCPICountnInstructioBTimeCPU500psI250ps2.0IATimeCycleACPICountnInstructioATim

eCPUAisfaster……bythismuchppt课件CPIinMoreDetailIfdifferentinstructionclassestakedifferentnumbersofc

yclesn1iii)CountnInstructio(CPICyclesClockWeightedaverageCPIn1iiiCountnInstructioCountnInstructioCPICountnInstructioC

yclesClockCPIRelativefrequencyppt课件CPIExampleAlternativecompiledcodesequencesusinginstructionsinclassesA,B,CClassABCCPIforclass123ICinsequence1

212ICinsequence2411Sequence1:IC=5ClockCycles=2×1+1×2+2×3=10Avg.CPI=10/5=2.0Sequence2:IC=6ClockCycles=4×1+1×2+1×3=

9Avg.CPI=9/6=1.5Whatisavg.CPI?IC=InstructionCountppt课件PerformanceSummaryPerformancedependsonAlgorithm:aff

ectsIC,possiblyCPIProgramminglanguage:affectsIC,CPICompiler:affectsIC,CPIInstructionsetarchitecture:affectsIC,CPI,TcTheBIGPictur

ecycleClockSecondsnInstructiocyclesClockProgramnsInstructioTimeCPUppt课件POWERppt课件PowerTrendsInCMOSICtechnology§1.5ThePowerWallFre

quencyVoltageloadCapacitivePower2×1000×305V→1Vppt课件ReducingPowerSupposeanewCPUhas85%ofcapacitiveloadofoldCPU

15%voltageand15%frequencyreduction0.520.85FVC0.85F0.85)(V0.85CPP4old2oldoldold2oldoldoldnewThepowerwallWecan’treducevoltag

efurtherWecan’tremovemoreheatHowelsecanweimproveperformance?ppt课件UniprocessorPerformance§1.6TheSeaChange:TheSwitchtoMu

ltiprocessorsConstrainedbypower,instruction-levelparallelism,memorylatencyppt课件MultiprocessorsMulticoremicroprocessorsMorethanonepr

ocessorperchipRequiresexplicitlyparallelprogrammingComparewithinstructionlevelparallelismHardwareexecutesmultipleinstruction

satonceHiddenfromtheprogrammerHardtodoProgrammingforperformanceLoadbalancingOptimizingcommunicationandsynchronizationppt课件MANUFACTURINGppt课件Ma

nufacturingICsYield:proportionofworkingdiesperwafer§1.7RealStuff:TheAMDOpteronX4ppt课件AMDOpteronX2WaferX2:300mmwafer,117

chips,90nmtechnologyX4:45nmtechnologyppt课件IntegratedCircuitCostNonlinearrelationtoareaanddefectrateWafercostandareaaref

ixedDefectratedeterminedbymanufacturingprocessDieareadeterminedbyarchitectureandcircuitdesign2area/2))Dieareaper(Defects(11Yie

ldareaDieareaWaferwaferperDiesYieldwaferperDieswaferperCostdieperCostppt课件BENCHMARKINGppt课件SPECCPUBenchmarkProgramsusedtomeasureperformance

SupposedlytypicalofactualworkloadStandardPerformanceEvaluationCorp(SPEC)DevelopsbenchmarksforCPU,I/O,Web,…SPECCPU2006Elapsedtimetoexecuteasel

ectionofprogramsNegligibleI/O,sofocusesonCPUperformanceNormalizerelativetoreferencemachineSummari

zeasgeometricmeanofperformanceratiosCINT2006(integer)andCFP2006(floating-point)nn1iiratiotimeExecutionppt课件CINT2006forOpteronX42356NameDescripti

onIC×109CPITc(ns)ExectimeReftimeSPECratioperlInterpretedstringprocessing2,1180.750.406379,77715.3bzip2Block-sortingcompress

ion2,3890.850.408179,65011.8gccGNUCCompiler1,0501.720.47248,05011.1mcfCombinatorialoptimization33610.000.401,3459,

1206.8goGogame(AI)1,6581.090.4072110,49014.6hmmerSearchgenesequence2,7830.800.408909,33010.5sjengChessgame(AI)2,1760.

960.483712,10014.5libquantumQuantumcomputersimulation1,6231.610.401,04720,72019.8h264avcVideocompression3,1020.800.4099322,13022.3omnetppDiscretee

ventsimulation5872.940.406906,2509.1astarGames/pathfinding1,0821.790.407737,0209.1xalancbmkXMLparsing1,0582.700.401,1436,9006.0Geometricmean11.7pp

t课件SPECPowerBenchmarkPowerconsumptionofserveratdifferentworkloadlevelsPerformance:ssj_ops/secPower:Watts(Joules/sec)

100ii100iipowerssj_opsWattperssj_opsOverallppt课件SPECpower_ssj2008forX4TargetLoad%Performance(ssj_ops/s

ec)AveragePower(Watts)100%231,86729590%211,28228680%185,80327570%163,42726560%140,16025650%118,32424640%92

0,3523330%70,50022220%47,12620610%23,0661800%0141Overallsum1,283,5902,605∑ssj_ops/∑power493ppt课件Pitfall:Amdahl’sLawImproving

anaspectofacomputerandexpectingaproportionalimprovementinoverallperformance§1.8FallaciesandPitfalls208020nCan’tbedone!unaffectedaffectedimprovedTf

actortimprovemenTTExample:multiplyaccountsfor80s/100sHowmuchimprovementinmultiplyperformancetoget5×overall?ppt课件Fallacy:LowPoweratI

dleLookbackatX4powerbenchmarkAt100%load:295WAt50%load:246W(83%)At10%load:180W(61%)GoogledatacenterMostlyopera

tesat10%–50%loadAt100%loadlessthan1%ofthetimeConsiderdesigningprocessorstomakepowerproportionaltoloadppt

课件Pitfall:MIPSasaPerformanceMetricMIPS:MillionsofInstructionsPerSecondDoesn’taccountforDifferencesinISAsbetweencomputersDifferencesincomp

lexitybetweeninstructions66610CPIrateClock10rateClockCPIcountnInstructiocountnInstructio10timeExecutioncountnInstructioMIPSCPIvariesbetw

eenprogramsonagivenCPUppt课件ConcludingRemarksCost/performanceisimprovingDuetounderlyingtechnologydevelopmentHierarchicallayersof

abstractionInbothhardwareandsoftwareInstructionsetarchitectureThehardware/softwareinterfaceExecutiontime:thebestperformancemeasurePowe

risalimitingfactorUseparallelismtoimproveperformance§1.9ConcludingRemarksppt课件