【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UnderpinnedbyMoore’sLawWhatisMoo

re’sLaw???Moore'slawdescribesalong-termtrendinthehistoryofcomputinghardware.Thequantityoftransistorsthatcanbeplacedinexpensivelyonanin

tegratedcircuithasdoubledapproximatelyeverytwoyears.§1.1Introductionppt课件Moore’sLawThetrendhascontinued

formorethanhalfacenturyandisnotexpectedtostopuntil2015orlater.ppt课件TheComputerRevolutionMakesnovelapplication

sfeasibleComputersinautomobiles§1.1Introductionppt课件TheComputerRevolutionMakesnovelapplicationsfeasibleCellphones§1.1Introductio

nppt课件TheComputerRevolutionMakesnovelapplicationsfeasibleHumangenomeprojectWorldWideWebSearchEngines§1.1Introductionppt课件The

ComputerRevolutionComputersarepervasive§1.1Introductionppt课件ClassesofComputersQuestion:Howdoyouclass

ifycomputers?DesktopComputersServerComputersEmbeddedComputersppt课件ClassesofComputersDesktopcomputersPCGeneralpurpose,

varietyofsoftwareSubjecttocost/performancetradeoffppt课件ClassesofComputersServercomputersNetworkbasedHighcapacity,performance,reliability

RangefromsmallserverstobuildingsizedWorld’ssmallestwebserverppt课件ClassesofComputersEmbeddedcomputersHiddenascompon

entsofsystemsStringentpower/performance/costconstraintsppt课件TheProcessorMarketppt课件WhatYouWillLearnHowprogramsaretranslatedint

othemachinelanguageAndhowthehardwareexecutesthemThehardware/softwareinterfaceWhatdeterminesprogramperformanceAndhowitcanbeimprovedHowhardwa

redesignersimproveperformanceWhatisparallelprocessingppt课件LevelsofProgramCodeHigh-levellanguageLevelofabstraction

closertoproblemdomainProvidesforproductivityandportabilityAssemblylanguageTextualrepresentationofinstructionsHar

dwarerepresentationBinarydigits(bits)Encodedinstructionsanddatappt课件BelowYourProgramApplicationsoftwareWritteninhigh-levellan

guage(HLL)SystemsoftwareCompiler:translatesHLLcodetomachinecodeOperatingSystem:servicecodeHandlinginput/outp

utManagingmemoryandstorageSchedulingtasks&sharingresourcesHardwareProcessor,memory,I/Ocontrollers§1.2B

elowYourProgramppt课件UnderstandingPerformanceAlgorithmDeterminesnumberofoperationsexecutedProgramminglanguage,compiler,archite

ctureDeterminenumberofmachineinstructionsexecutedperoperationProcessorandmemorysystemDeterminehowfastinstruc

tionsareexecutedI/Osystem(includingOS)DetermineshowfastI/Ooperationsareexecutedppt课件ComponentsofaComputerSamecomponentsforallkindsofcomputerD

esktop,server,embeddedInput/outputincludesUser-interfacedevicesDisplay,keyboard,mouseStoragedevicesHarddisk,

CD/DVD,flashNetworkadaptersForcommunicatingwithothercomputers§1.3UndertheCoversTheBIGPictureppt课件Anatomy(结构)ofaComputerOutputdeviceInput

deviceInputdeviceNetworkcableppt课件AnatomyofaMouseOpticalmouseLEDilluminatesdesktopSmalllow-rescameraBasicimag

eprocessorLooksforx,ymovementButtons&wheelSupersedesroller-ballmechanicalmouseppt课件ThroughtheLookingGlassLCDscreen:pictureelemen

ts(pixels)Mirrorscontentofframebuffermemoryppt课件OpeningtheBoxppt课件InsidetheProcessor(CPU)Datapath:performsoperati

onsondataControl:sequencesdatapath,memory,...CachememorySmallfastSRAMmemoryforimmediateaccesstodataS

RAM–StaticRandomAccessMemoryppt课件InsidetheProcessorAMDBarcelona:4processorcoresppt课件ASafePlaceforDataVolatile(易变的)mainmemoryLoses

instructionsanddatawhenpoweroffNon-volatilesecondarymemoryMagneticdiskFlashmemoryOpticaldisk(CDROM,DVD)ppt课件NetworksCommunicationand

resourcesharingLocalareanetwork(LAN):EthernetWithinabuildingWideareanetwork(WAN):theInternetppt课件Abst

ractionsInstructionSetArchitecture(ISA)Aninterfacebetweenthehardwareandthelowest-levelsoftwareTheabstractimageofa

computingsystemthatisseenbyamachine/assemblylanguageprogrammerIncludinginstructions,registers,memoryaccess,I/O,…TheBIGPictureppt课件ISAsSystem/360a

ndupwardscompatiblesuccessorsz/ArchitecturePowerArchitecturePDP-11SPARCSuperHTricoreTransputerUNIVAC1100/2200seriesVAXx86IA-32

(32-bitx86,firstimplementedintheIntel80386)x86-64(64-bitsupersetofIA-32,firstimplementedintheAMDOpteron)EISC(AE32K)4004,40406800,6

502,6809,68HC11,68HC08.8008,8080,8085,Z80,Z180,eZ80,etc.8048,8051,etc.Z8,eZ8,etc.AlphaARMBurroughsB5000seriesBurroughsB6000/B7

000serieseSi-RISCIA-64(Itanium)Mico32MIPSMotorola68kPA-RISCIBM700/7000linesppt课件AbstractionsApplicationBinaryInterface(

ABI)thelow-levelinterfacebetweenanapplicationprogramandOSABIscoverdetailssuchasdatatype,size,alignment,callingconvention,binaryformat

ofobjectfiles,etc.Definesastandardforbinaryportabilityacrosscomputers.ppt课件AbstractionsImplementationHardwarethatob

eysthearchitectureabstractionManyimplementationsforthesameISAExample:IntelPentiumvsAMDAthlon,almostidenticalISA,butradi

callydifferentinternaldesigns.ppt课件PERFORMANCEppt课件DefiningPerformanceWhichairplanehasthebestperformance?§1.4

Performanceppt课件DefiningPerformanceWhichairplanehasthebestperformance?0100200300400500DouglasDC-8-50BAC/SudConcordeBoeing747

Boeing777PassengerCapacity0200040006000800010000DouglasDC-8-50BAC/SudConcordeBoeing747Boeing777CruisingRange(miles)0500100015

00DouglasDC-8-50BAC/SudConcordeBoeing747Boeing777CruisingSpeed(mph)0100000200000300000400000DouglasDC-8-50BAC/

SudConcordeBoeing747Boeing777Passengersxmph§1.4Performanceppt课件DefiningPerformancePerformance=SpeedE.g.,takingasinglepassengerfro

monepointtoanotherWinner:ConcordePerformance=PassengerThroughputPassengersxm.p.hE.g.,transporting450

passengersfromonepointtoanotherWinner:Boeing747ppt课件ResponseTimeandThroughputResponsetimeHowlongittakestodoataskThroughputTotalworkdoneperu

nittimee.g.,tasks/transactions/…perhourppt课件ResponseTimeandThroughputHowareresponsetimeandthroughputaffectedbyReplacingtheprocessorwi

thafasterversion?Addingmoreprocessors?We’llfocusonresponsetimefornow…ppt课件RelativePerformanceDefinePe

rformance=1/ExecutionTime“XisntimefasterthanY”nXYYXtimeExecutiontimeExecutionePerformancePerformancExample

:timetakentorunaprogram10sonA,15sonBExecutionTimeB/ExecutionTimeA=15s/10s=1.5SoAis1.5timesfasterthanBppt课件MeasuringExecutionTi

meElapsedtimeTotalresponsetime,includingallaspectsProcessing,I/O,OSoverhead,idletimeDeterminessystemperformanceCPUtimeTimespentp

rocessingagivenjobDiscountsI/Otime,otherjobs’sharesComprisesuserCPUtimeandsystemCPUtimeDifferentprogram

sareaffecteddifferentlybyCPUandsystemperformanceppt课件CPUClockingOperationofdigitalhardwaregovernedbyaconstant-rateclockClock(cycles)Datatransferan

dcomputationUpdatestateClockperiodClockcycletime: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Hardwarede

signermustoftentradeoffclockrateagainstcyclecountRateClockCyclesClockCPUTimeCycleClockCyclesClockCPUTimeCP

Uppt课件CPUTimeExampleComputerA:2GHzclock,10sCPUtimeDesigningComputerBAimfor6sCPUtimeCandofasterclo

ck,butcauses1.2×clockcyclesHowfastmustComputerB’sclockratebe?4GHz6s10246s10201.2RateClock10202GHz10sRateClockTimeCPUCyclesClock6sCycl

esClock1.2TimeCPUCyclesClockRateClock99B9AAAABBBppt课件InstructionCountandCPIInstructionCountforaprogramDeterminedbyp

rogram,ISAandcompilerAveragecyclesperinstructionDeterminedbyCPUhardwareIfdifferentinstructionshavedifferentCPIAvera

geCPIaffectedbyinstructionmixRateClockCPICountnInstructioTimeCycleClockCPICountnInstructioTimeCPUnInstructioperCyclesCountnIns

tructioCyclesClockCPIppt课件CPIExampleComputerA:CycleTime=250ps,CPI=2.0ComputerB:CycleTime=500ps,CPI=1.2

SameISAWhichisfaster,andbyhowmuch?1.2500psI600psIATimeCPUBTimeCPU600psI500ps1.2IBTimeCycleBCPICountnInstructi

oBTimeCPU500psI250ps2.0IATimeCycleACPICountnInstructioATimeCPUAisfaster……bythismuchppt课件CPIinMoreDetailIfdifferentinstructio

nclassestakedifferentnumbersofcyclesn1iii)CountnInstructio(CPICyclesClockWeightedaverageCPIn1iiiCountnInstructioCountn

InstructioCPICountnInstructioCyclesClockCPIRelativefrequencyppt课件CPIExampleAlternativecompiledcodesequencesusinginstruct

ionsinclassesA,B,CClassABCCPIforclass123ICinsequence1212ICinsequence2411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Perfo

rmancedependsonAlgorithm:affectsIC,possiblyCPIProgramminglanguage:affectsIC,CPICompiler:affectsIC,CPIIns

tructionsetarchitecture:affectsIC,CPI,TcTheBIGPicturecycleClockSecondsnInstructiocyclesClockProgramnsInstructioTimeCP

Uppt课件POWERppt课件PowerTrendsInCMOSICtechnology§1.5ThePowerWallFrequencyVoltageloadCapacitivePower2×1000×305V→1Vppt课件Reduci

ngPowerSupposeanewCPUhas85%ofcapacitiveloadofoldCPU15%voltageand15%frequencyreduction0.520.85FVC0.85F0.85)(V0.85CPP4old2oldoldol

d2oldoldoldnewThepowerwallWecan’treducevoltagefurtherWecan’tremovemoreheatHowelsecanweimproveperf

ormance?ppt课件UniprocessorPerformance§1.6TheSeaChange:TheSwitchtoMultiprocessorsConstrainedbypower,instruction-levelpa

rallelism,memorylatencyppt课件MultiprocessorsMulticoremicroprocessorsMorethanoneprocessorperchipRequire

sexplicitlyparallelprogrammingComparewithinstructionlevelparallelismHardwareexecutesmultipleinstructions

atonceHiddenfromtheprogrammerHardtodoProgrammingforperformanceLoadbalancingOptimizingcommunicationandsynchronization

ppt课件MANUFACTURINGppt课件ManufacturingICsYield:proportionofworkingdiesperwafer§1.7RealStuff:TheAMDOpteronX4ppt课件AMDOpteronX2WaferX2:300mmw

afer,117chips,90nmtechnologyX4:45nmtechnologyppt课件IntegratedCircuitCostNonlinearrelationtoareaanddefe

ctrateWafercostandareaarefixedDefectratedeterminedbymanufacturingprocessDieareadeterminedbyarchitectureandcircuitdesign2area/2)

)Dieareaper(Defects(11YieldareaDieareaWaferwaferperDiesYieldwaferperDieswaferperCostdieperCostppt课件BENCHMARKINGppt课件SPECCPUBenchmark

ProgramsusedtomeasureperformanceSupposedlytypicalofactualworkloadStandardPerformanceEvaluationCorp(SPEC)Developsbenchmar

ksforCPU,I/O,Web,…SPECCPU2006ElapsedtimetoexecuteaselectionofprogramsNegligibleI/O,sofocusesonCPUperformanceNormalizerelativ

etoreferencemachineSummarizeasgeometricmeanofperformanceratiosCINT2006(integer)andCFP2006(floating-point)nn1iiratiotimeExecution

ppt课件CINT2006forOpteronX42356NameDescriptionIC×109CPITc(ns)ExectimeReftimeSPECratioperlInterpretedstringprocessing2,1180.750.406379,7

7715.3bzip2Block-sortingcompression2,3890.850.408179,65011.8gccGNUCCompiler1,0501.720.47248,05011.1mc

fCombinatorialoptimization33610.000.401,3459,1206.8goGogame(AI)1,6581.090.4072110,49014.6hmmerSearchgenesequenc

e2,7830.800.408909,33010.5sjengChessgame(AI)2,1760.960.483712,10014.5libquantumQuantumcomputersimulation1,6231.610.401,04720,72019.8h264avcVide

ocompression3,1020.800.4099322,13022.3omnetppDiscreteeventsimulation5872.940.406906,2509.1astarGames/pathfinding1,0821.790.407737,0209.1xa

lancbmkXMLparsing1,0582.700.401,1436,9006.0Geometricmean11.7ppt课件SPECPowerBenchmarkPowerconsumptionofserver

atdifferentworkloadlevelsPerformance:ssj_ops/secPower:Watts(Joules/sec)100ii100iipowerss

j_opsWattperssj_opsOverallppt课件SPECpower_ssj2008forX4TargetLoad%Performance(ssj_ops/sec)AveragePower(Watts)100%231,8672959

0%211,28228680%185,80327570%163,42726560%140,16025650%118,32424640%920,3523330%70,50022220%47,12620610%23,06

61800%0141Overallsum1,283,5902,605∑ssj_ops/∑power493ppt课件Pitfall:Amdahl’sLawImprovinganaspectofacomputerandexpectingaproportionalimprovementinoveral

lperformance§1.8FallaciesandPitfalls208020nCan’tbedone!unaffectedaffectedimprovedTfactortimprovemenTTExample:mu

ltiplyaccountsfor80s/100sHowmuchimprovementinmultiplyperformancetoget5×overall?ppt课件Fallacy:LowPoweratIdleL

ookbackatX4powerbenchmarkAt100%load:295WAt50%load:246W(83%)At10%load:180W(61%)GoogledatacenterMostlyoperatesat10%–

50%loadAt100%loadlessthan1%ofthetimeConsiderdesigningprocessorstomakepowerproportionaltoloadppt课件Pitfall

:MIPSasaPerformanceMetricMIPS:MillionsofInstructionsPerSecondDoesn’taccountforDifferencesinISAsbetweencomputersDifferencesincomple

xitybetweeninstructions66610CPIrateClock10rateClockCPIcountnInstructiocountnInstructio10timeExecutioncountnInstructioMIPSCPIvariesbetweenp

rogramsonagivenCPUppt课件ConcludingRemarksCost/performanceisimprovingDuetounderlyingtechnologydevelopmentHierarchicallayers

ofabstractionInbothhardwareandsoftwareInstructionsetarchitectureThehardware/softwareinterfaceExecutiontime:thebestperformancemeasurePower

isalimitingfactorUseparallelismtoimproveperformance§1.9ConcludingRemarksppt课件