Developmenthistory
CPUappearedintheeraoflarge-scaleintegratedcircuits.TheiterativeupdateofprocessorarchitecturedesignandthecontinuousimprovementofintegratedcircuittechnologypromoteitscontinuousdevelopmentComplete.Frominitiallydedicatedtomathematicalcalculationstowidelyusedingeneralcalculations,from4-bitto8-bit,16-bit,32-bitprocessors,andfinallyto64-bitprocessors,fromtheincompatibilityofvariousmanufacturerstotheemergenceofdifferentinstructionsetarchitecturespecifications,TheCPUhasbeendevelopingrapidlysinceitsinception.
CPUdevelopmenthasahistoryofmorethan40years.Weusuallydivideitintosixstages.
(1)Thefirststage(1971-1973).Thisistheeraof4-bitand8-bitlow-endmicroprocessors,andtherepresentativeproductistheIntel4004processor.
In1971,Intel’s4004microprocessorintegratedarithmeticunitandcontrolleronasinglechip,markingthebirthoftheCPU;in1978,theappearanceofthe8086processorlaidthefoundationfortheX86instructionsetarchitecture.Subsequently,the8086seriesprocessorswerewidelyusedinpersonalcomputerterminals,high-performanceserversandcloudservers.
(2)Thesecondstage(1974-1977).Thisistheeraof8-bitmid-to-high-endmicroprocessors,andtherepresentativeproductisIntel8080.Atthistime,thecommandsystemisrelativelycomplete.
(3)Thethirdstage(1978-1984).Thisistheeraof16-bitmicroprocessors,andtherepresentativeproductisIntel8086.Relativelyspeaking,itisrelativelymature.
(4)Thefourthstage(1985-1992).Thisistheageof32-bitmicroprocessors,andtherepresentativeproductisIntel80386.Alreadycapableofmulti-tasking,multi-useroperations.
The80486processorreleasedin1989implementeda5-levelscalarpipeline,markingtheinitialmaturityoftheCPUandtheendofthedevelopmentphaseofthetraditionalprocessor.
(5)Thefifthstage(1993-2005).ThisistheeraofthePentiumseriesofmicroprocessors.
InNovember1995,IntelreleasedthePentiumprocessor,whichforthefirsttimeadoptedasuperscalarinstructionpipelinestructure,introducedout-of-orderexecutionofinstructionsandbranchpredictiontechnology,whichgreatlyimprovedtheperformanceoftheprocessor.Therefore,thesuperscalarinstructionpipelinestructurehasbeenadoptedbysubsequentmodernprocessors,suchasAMD(AdvancedMicrodevices)RyzenandIntel'sCoreseries.
(6)Thesixthstage(2005-2021).Theprocessorgraduallydevelopstowardsmorecoresandhigherparallelism.TypicalrepresentativesareIntel'sCoreseriesprocessorsandAMD'sRyzenseriesprocessors.
Inordertomeettheupper-levelworkrequirementsoftheoperatingsystem,modernprocessorshavefurtherintroducedfunctionssuchasparallelization,multi-core,virtualization,andremotemanagementsystems,whichcontinuetopromotethedevelopmentofupper-levelinformationsystems.
Workingprinciple
TheVonNeumannarchitectureisthefoundationofmoderncomputers.Underthisarchitecture,programsanddataarestoredinaunifiedmanner.Instructionsanddataneedtobeaccessedfromthesamestoragespace,transmittedviathesamebus,andcannotbeoverlappedforexecution.AccordingtotheVonNeumannsystem,theCPU'sworkisdividedintothefollowingfivestages:instructionfetchingstage,instructiondecodingstage,instructionexecutionstage,accessnumberandresultwriteback.
Instructionfetch(IF,instructionfetch)istheprocessoffetchinganinstructionfromthemainmemorytotheinstructionregister.Thevalueintheprogramcounterisusedtoindicatethepositionofthecurrentinstructioninthemainmemory.Whenaninstructionisfetched,thevalueinthePCwillautomaticallyincreaseaccordingtothelengthoftheinstructionword.
Intheinstructiondecodestage(ID,instructiondecode),afterfetchingtheinstruction,theinstructiondecodersplitsandinterpretstheretrievedinstructionaccordingtothepredeterminedinstructionformat,identifyinganddistinguishingdifferentinstructiontypesAndvariousmethodsofobtainingoperands.ModernCISCprocessorswillsplittoimproveparallelismandefficiency.
Theexecutioninstructionstage(EX,execute),whichspecificallyimplementsthefunctionoftheinstruction.ThedifferentpartsoftheCPUareconnectedtoperformtherequiredoperations.
Accesstotheaccessnumberstage(MEM,memory),accordingtotheinstructionneedstoaccessthemainmemory,readtheoperand,theCPUobtainstheaddressoftheoperandinthemainmemory,andreadstheoperationfromthemainmemoryNumbersareusedforcalculations.Someinstructionsdonotneedtoaccessthemainmemory,youcanskipthisstage.
Resultwritebackstage(WB,writeback),asthelaststage,theresultwritebackstage"writesback"therunningresultdataoftheexecutioninstructionstagetosomestorageform.TheresultdataisgenerallywrittentotheinternalregisteroftheCPUsothatitcanbequicklyaccessedbysubsequentinstructions;manyinstructionswillalsochangethestateoftheflagbitsintheprogramstatuswordregister.TheseflagbitsindicatedifferentoperationresultsandcanbeUsedtoinfluencetheactionsoftheprogram.
Aftertheinstructionisexecutedandtheresultdataiswrittenback,ifnounexpectedevents(suchasresultoverflow,etc.)occur,thecomputerwillobtaintheaddressofthenextinstructionfromtheprogramcounterandstartanewcycle.Oneinstructioncyclewillfetchthenextinstructioninsequence.
Introduction
Thecentralprocessingunit(CPU)isoneofthemainequipmentofelectroniccomputersandthecorecomponentofcomputers.Itsfunctionismainlytointerpretcomputerinstructionsandprocessdataincomputersoftware.TheCPUisthecorecomponentofthecomputerresponsibleforreadinginstructions,decodingandexecutinginstructions.Thecentralprocessingunitmainlyincludestwoparts,namelythecontrollerandthearithmeticunit,whichalsoincludesthehigh-speedbuffermemoryandthebusthatrealizesthedataandcontroloftheconnectionbetweenthem.ThethreecorecomponentsofanelectroniccomputeraretheCPU,internalmemory,andinput/outputdevices.Thefunctionsofthecentralprocessingunitaremainlytoprocessinstructions,performoperations,controltime,andprocessdata.
Inthecomputerarchitecture,theCPUisthecorehardwareunitthatcontrolsallthehardwareresourcesofthecomputer(suchasmemory,inputandoutputunits)andperformsgeneraloperations.TheCPUisthecomputingandcontrolcoreofthecomputer.TheoperationofallsoftwarelayersinthecomputersystemwilleventuallybemappedtotheoperationoftheCPUthroughtheinstructionset.
Performancestructure
Performancemeasurementindicators
FortheCPU,theindicatorsthataffectitsperformancemainlyincludethemainfrequency,thenumberofbitsoftheCPU,andthecacheinstructionsoftheCPU.Sets,numberofCPUcores,andIPC(numberofinstructionspercycle).Theso-calledCPUfrequencyreferstotheclockfrequency,whichdirectlydeterminestheperformanceoftheCPU.YoucanincreasetheCPUfrequencybyoverclockingtoobtainhigherperformance.ThenumberofCPUbitsreferstothenumberoffloating-pointnumbersthattheprocessorcancalculateatonetime.Generally,thehigherthenumberofCPUbits,thefastertheCPUperformsoperations.Afterthe1920s,theCPUsusedinpersonalcomputersaregenerally64-bit.Thisisbecause64-bitprocessorscanprocessalargerrangeofdataandnativelysupporthighermemoryaddressingcapacity,whichimprovespeople'sworkefficiency.ThecacheinstructionsetoftheCPUisstoredinsidetheCPU,whichmainlyreferstothehardprogramthatcanguideandoptimizetheoperationoftheCPU.Generallyspeaking,thecacheofaCPUcanbedividedintoafirst-levelcache,asecond-levelcache,andathird-levelcache.ThecacheperformancedirectlyaffectstheprocessingperformanceoftheCPU.SomeCPUswithspecialfunctionsmaybeequippedwithafour-levelcache.
CPUstructure
Generallyspeaking,thestructureofaCPUcanberoughlydividedintoarithmeticlogiccomponents,registercomponents,andcontrolcomponents.Theso-calledarithmeticlogiccomponentsaremainlycapableofperformingrelatedlogicaloperations,suchas:shiftoperationsandlogicaloperations,inadditiontofixed-pointorfloating-pointarithmeticoperations,andaddressoperationsandconversioncommands.Functionalarithmeticunit.Theregistercomponentisusedtotemporarilystoreinstructions,dataandaddresses.Thecontrolcomponentismainlyusedtoanalyzetheinstructionandcansendoutthecorrespondingcontrolsignal.Thecomputer'smemorycanbedividedintorandomaccessmemory(RAM)andread-onlymemory(ROM).ThedifferencebetweenthetwoisthattherandomaccessmemorycandirectlyexchangedatawiththeCPU,anditcanalsobecalledthemainmemory.RAMcanbereadandwrittenatanytime,andthespeedofthisprocessisveryfast.Therefore,duetothisadvantageofmainmemory,itisoftenusedasatemporarydatastoragemediumforoperatingsystemsorotherrunningprograms;andread-onlymemoryROMisakindofmemorythatcanonlyreadthedatastoredinadvance.Theuserdoesnothavetheauthoritytochangethedatastoredinitsinternalstorageandcannotdeleteit,andthedatawillnotdisappearafterthepoweristurnedoff.Thiskindofmemoryhasalsobeenwidelyused,andhasbeenwellusedinelectronicorcomputersystemswheredatadoesnotneedtobechangedfrequently.
Forthecentralprocessingunit,itcanberegardedasalarge-scaleintegratedcircuitwhosemaintaskistoprocessandprocessvariousdata.Thestoragecapacityofthetraditionalcomputerisrelativelysmall,ithascertaindifficultyintheprocessoflarge-scaledataprocessing,andtheprocessingeffectisrelativelylow.Withtherapiddevelopmentofmycountry'sinformationtechnologylevel,ahigh-configurationprocessorcomputerhasappeared,andthehigh-configurationprocessorisusedasthecontrolcenter,whichplaysanimportantroleinimprovingthestructureandfunctionofthecomputer'sCPU.Thecorepartofthecentralprocessingunitisthecontrollerandarithmeticunit,whichplayanimportantroleinimprovingtheoverallfunctionofthecomputer.Itcanrealizetheproliferationofmultiplefunctionssuchasregistercontrol,logicoperation,signaltransmissionandreception,andlayagoodfoundationforimprovingtheperformanceofthecomputer.Base.
Integratedcircuitsactascontrolsignalsinthecomputerandexecutedifferentcommandtasksaccordingtouseroperationinstructions.Thecentralprocessingunitisaverylarge-scaleintegratedcircuit.Itiscomposedofarithmeticunits,controllers,registers,etc.,asshowninthefigurebelow.Thekeyoperationliesintheprocessingandprocessingofvariousdata.
Traditionalcomputershaverelativelysmallstoragecapacityandlowoperatingefficiencyforlarge-scaledatasets.Thenewgenerationofcomputersusesahigh-configurationprocessorasthecontrolcenter,andtheCPUhasalotofroomforimprovementintermsofstructureandfunction.Thecentralprocessingunitusesarithmeticunitsandcontrollersasthemaindevices,andgraduallyspreadsintomultiplefunctionssuchaslogicoperations,registercontrol,programcoding,andsignaltransmissionandreception.TheseallspeeduptheoptimizationandupgradingofCPUcontrolperformance.
CPUbus
TheCPUbusisthefastestbusinacomputersystem,anditisalsothecoreofthechipsetandmotherboard.PeopleusuallycallthelocalbusdirectlyconnectedtotheCPUCPUbusorinternalbus,andcallthoselocalbusesconnectedtovariousgeneral-purposeexpansionslotsassystembusorexternalbus.InaCPUwitharelativelysingleinternalstructure,thereisoftenonlyasetofdatatransferbuses,thatis,theCPUinternalbus,whichisusedtoconnecttheinternalregistersoftheCPUandthearithmeticlogicoperationcomponents,sothistypeofbuscanalsobecalledItistheALUbus.Thebusinsidethecomponentconnectsthechipstogetherbyusingasetofbuses,soitcanbecalledthebusinsidethecomponent,whichgenerallycontainstwosetsoflines,theaddresslineandthedataline.Thesystembusreferstothelinethatconnectsthevariouscomponentsofthesystemtogether,andisthebasisforconnectingthewholesystemtogether;thebusoutsidethesystemisthebasiclinethatconnectsthecomputerandotherdevicestogether.
Thecorepart
Thearithmeticunit
Thearithmeticunitisthepartofthecomputerthatperformsvariousarithmeticandlogicaloperations,andthearithmeticlogicunitisthecentralprocessingcorepart.
(1)Arithmeticlogicunit(ALU).Arithmeticlogicunitreferstoacombinationallogiccircuitthatcanrealizemultiplesetsofarithmeticoperationsandlogicoperations,anditisanimportantpartofthecentralprocessing.Theoperationofthearithmeticlogicunitismainlytoperformtwo-bitarithmeticoperations,suchasaddition,subtraction,andmultiplication.Intheprocessofcalculation,thearithmeticlogicunitismainlytoperformarithmeticandlogicaloperationsinacentralizedcomputerinstruction.Generallyspeaking,ALUcanplaytheroleofdirectreadingandreading,whichisspecificallyreflectedintheprocessorcontroller,memoryandinputandoutputdevices.,Inputandoutputareimplementedonthebasisofthebus.Theinputinstructioncontainsaninstructionword,includingoperationcode,formatcodeandsoon.
(2)Intermediateregister(IR).Itslengthis128bits,anditsactuallengthisdeterminedbytheoperand.IRplaysanimportantroleinthe"pushintostackandfetch"instruction.Duringtheexecutionofthisinstruction,thecontentsofACCaresenttoIR,andthentheoperandisfetchedtoACC,andthenthecontentsofIRareputonthestack.
(3)Operationalaccumulator(ACC).Currentregistersaregenerallysingleaccumulatorswithalengthof128bits.ForACC,itcanberegardedasavariablelengthaccumulator.Intheprocessofnarratinginstructions,theexpressionofthelengthofACCisgenerallybasedonthevalueofACS,andthelengthofACSisdirectlyrelatedtothelengthofACC,doublingorhalvingthelengthofACScanalsoberegardedasdoublingorhalvingthelengthofACC.
(4)Descriptionwordregister(DR).Itismainlyusedtostoreandmodifydescriptionwords.ThelengthofDRis64bits.Inordertosimplifydatastructureprocessing,theuseofdescriptionwordsplaysanimportantrole.
(5)RegisterB.Itplaysanimportantroleinthemodificationofinstructions.ThelengthoftheBregisteris32bits,andtheaddressmodificationamountcanbesavedintheprocessofmodifyingtheaddress.Themainmemoryaddresscanonlybemodifiedwithadescriptionword.Pointingtothefirstelementinthearrayisthedescriptionword,soaccesstootherelementsinthearrayshouldrequireamodifier.Fornumbercomposition,itiscomposedofdataofthesamesizeorelementsofthesamesize,andisstoredcontinuously.Thecommonaccessmethodisvectordescriptionword.Becausetheaddressinvectordescriptionwordisbyteaddress,Intheconversionprocess,thebasicaddressshouldbeaddedfirst.Fortheconversionwork,itismainlyrealizedautomaticallybythehardware.Inthisprocess,specialattentionshouldbepaidtoalignment,soasnottoexceedtheboundaryofthearray.
Controller
Thecontrollerreferstochangingthewiringofthemaincircuitorthecontrolcircuitandchangingtheresistancevalueinthecircuitaccordingtoapredeterminedsequencetocontrolthestart,speed,brakeandreverseofthemotor.Tothemasterdevice.ThecontrolleriscomposedoftheprogramstatusregisterPSR,thesystemstatusregisterSSR,theprogramcounterPC,andtheinstructionregister.Asa"decision-makingbody",themaintaskistoissuecommandsandplaytheroleofcoordinatingandcommandingtheoperationoftheentirecomputersystem.Theclassificationofcontrolmainlyincludestwokinds,respectivelyisthecombinationallogiccontroller,themicroprogramcontroller,thetwopartshavetheirownadvantagesanddisadvantages.Amongthem,thestructureofthecombinationallogiccontrollerisrelativelycomplex,buttheadvantageisthatitisfaster;thestructureofthemicroprogramcontrollerissimple,butinthemodificationofamachineinstructionfunction,allthemicroprogramsneedtobereprogrammed.
Brandintroduction
"Loongson"seriesofchips
"Loongson"seriesofchipsaredesignedanddevelopedbytheChineseAcademyofSciencesZhongkeTechnologyCo.,Ltd.,usingMIPSarchitecture,Withindependentintellectualpropertyrights,theproductsnowincludethreeseriesofLoongsonNo.1smallCPU,LoongsonNo.2mediumCPUandLoongsonNo.3largeCPU,andalsoincludeLoongson7A1000bridgechips.TheLoongson1series32/64-bitprocessorsarespeciallydesignedfortheembeddedfield,mainlyusedincloudterminals,industrialcontrol,dataacquisition,handheldterminals,networksecurity,consumerelectronicsandotherfields,withlowpowerconsumption,highintegrationandhighFeaturessuchascostperformance.Amongthem,theLoongson1A32-bitprocessorandLoongson1C64-bitprocessorworkstablyat266-300MHz,andtheLoongson1Bprocessorisalightweight32-bitchip.TheLoongson1Dprocessorisadedicatedchipforultrasonicheatmeters,watermetersandgasmeters.In2015,thenewgenerationofBeidounavigationsatellitesisequippedwiththeLoongson1Eand1Fchipsindependentlydevelopedbymycountry.Thesetwochipsaremainlyusedtocompletethedataprocessingtask1oftheinter-satellitelink.
Loongson2seriesare64-bithigh-performancelow-powerprocessorsfordesktopandhigh-endembeddedapplications.Loongson2productsincludeLoongson2E,2F,2Hand2K1000chips.Forthefirsttime,Loongson2Ehasachievedexternalproductionandsalesauthorization.TheaverageperformanceofLoongson2Fismorethan20%higherthanthatofLoongson2E,anditcanbeusedinpersonalcomputers,industrialterminals,industrialcontrol,dataacquisition,networksecurityandotherfields.Loongson2Hlauncheditsofficialproductin2012,whichissuitablefortheneedsofcomputers,cloudterminals,networkequipment,consumerelectronicsandotherfields,andcanbeusedasafull-featuredchipwithHTorPCI-einterfaces.In2018,LoongsonlaunchedtheLoongson2K1000processor,whichismainlyadual-coreprocessingchipforthefieldofnetworksecurityandmobileintelligence.Themainfrequencycanreach1GHz,whichcanmeettheneedsoftherapiddevelopmentoftheindustrialInternetofthingsandanindependentandcontrollableindustrialsecuritysystem.
TheLoongson3seriesaremulti-coreprocessorsforhigh-performancecomputers,serversandhigh-enddesktopapplications,featuringhighbandwidth,highperformance,andlowpowerconsumption.TheLoongson3A3000/3B3000processoradoptsanindependentmicrostructuredesign,andthemainfrequencycanreach1.5GHzormore;theLoongson3A4000,whichisplannedforthemarketin2019,isthefirstquad-corechipofLoongson'sthird-generationproduct.Thechipisbasedona28nmprocessandadoptsnewresearchanddevelopment.GS464V64-bithigh-performanceprocessorcorearchitecture,andimplements256-bitvectorinstructions,whileoptimizingon-chipinterconnectionandmemoryaccesschannels,integrated64-bitDDR3/4memorycontroller,integratedon-chipsecuritymechanism,mainfrequencyandperformanceIthasbeengreatlyimprovedagain.
TheLoongson7A1000bridgeisthefirstdedicatedbridgesetproductofLoongson.ThegoalistoreplacetheAMDRS780+SB710bridgesetandprovidethenorth-southbridgefunctionfortheLoongsonprocessor.ItwasreleasedinFebruary2018andiscurrentlyusedonahigh-performancenetworkplatformwithLoongson3A3000andZiguang4GDDR3memory.Comparedwiththe3A3000+780eplatform,theoverallperformanceoftheprogramhasbeengreatlyimproved,andithasthecharacteristicsofhighdomesticproductionrate,highperformance,andhighreliability.
Intel
AccordingtotheIntelproductlineplan,asof2021,Intel’s11th-generationconsumercorehasfivetypesofproducts:i9/i7/i5/i3/Pentium/Celeron.Inaddition,thereareXeonPlatinum/Gold/Silver/BronzeforserversandXeonWseriesforHEDTplatform.
AMD
AccordingtoAMD'sproductlineplanning,asof2021,AMDRyzen5000seriesprocessorshavefourconsumerproductlines:ryzen9/ryzen7/ryzen5/ryzen3.Inaddition,therearethethird-generationXiaolongEPYCprocessorfortheservermarketandthethreadtearerseriesfortheHEDTplatform.
ShanghaiZhaoxin
ShanghaiZhaoxinIntegratedCircuitCo.,Ltd.isastate-ownedholdingcompanyestablishedin2013.Itsprocessorsadoptthex86architecture.TheproductsmainlyincludethefirstZX-A,ZX-c/ZX-C+,ZX-D,KaixianKX-5000andKX-6000;KaishengZX-C+,ZX-D,KH-20,000,etc.Amongthem,KaixianKX-5000seriesprocessorsadopt28nmprocessandareavailablein4-coreor8-coreversions.Theoverallperformanceisupto140%higherthanthepreviousgenerationproduct,reachingtheinternationalmainstreamgeneral-purposeprocessorperformancelevel,whichcanfullymeetthepartyandgovernmentdesktopOfficeapplications,andavarietyofentertainmentapplicationsincluding4Kultra-high-definitionvideoviewing.KaishengKH-20000seriesprocessorsareCPUproductslaunchedbyZhaoxinforserversandotherequipment.KaixianKX-6000seriesprocessorshaveamainfrequencyofupto3.0GHzandarecompatiblewithafullrangeofWindowsoperatingsystemsanddomesticallyproducedindependentcontrollableoperatingsystemssuchasZhongkeFangde,WinnerKirin,andPricewaterhouseCoopers.TheperformanceisequivalenttoIntel'sseventh-generationCorei5.
ShanghaiShenwei
Shenweiprocessorisreferredtoas"Swprocessor",fromDEC'sAlpha21164,adoptsAlphaarchitecture,hascompletelyindependentintellectualpropertyrights,anditsproductshavesinglecoreSw-1,dual-coreSw-2,quad-coreSw-410,sixteen-coreSW-1600/SW-1610,etc.TheSunwayBlu-raysupercomputeruses8704SW-1600,equippedwithSunwayRuisioperatingsystem,realizingallthelocalizationofsoftwareandhardware.SinceitsreleaseinJune2016,the"Shenwei·LightofTaihu"supercomputerbasedonSw-26010hasrankedfirstintheworld'sTOP500supercomputersforfourconsecutivetimes,andtwoitemsonthe"Shenwei·LightofTaihu"Tensofmillionsofcoremachineapplicationswonthe"GordonBell"award,theworld'shighestawardforhigh-performancecomputingapplicationsin2016and2017.
Classification
Instructionsetmethod
TheclassificationofCPUcanalsobedividedintoreducedinstructionsetcomputer(RISC)andcomplexinstructionaccordingtotheinstructionsetmethodSetComputer(CISC).ThelengthandexecutiontimeofRISCinstructionsareconstant,whilethelengthandexecutiontimeofCISCinstructionsarenotnecessarily.ThedegreeofparallelexecutionofRISCinstructionsisbetter,andtheefficiencyofthecompilerisalsohigher.TheCISCinstructionhasbetteroptimizationfordifferenttasks,atthecostofcomplexcircuitsandmoredifficulttoimproveparallelism.ThetypicalCISCinstructionsethasthex86micro-architecture,andthetypicalRISCinstructionsethastheARMmicro-architecture.However,inmodernprocessorarchitectures,bothRISCandCISCinstructionswillbeconvertedduringthedecodingprocessandsplitintoRISC-likeinstructionsinsidetheCPU
EmbeddedSystemCPU
TraditionalembeddingThetypeofdomainreferstoaverywiderange,anditisthemainapplicationdomainofprocessorsinadditiontotheserverandPCdomains.Theso-called"embedded"meansthatinmanychips,theprocessorcontainedinitisasifitisembeddedinitandisunknown.
Inrecentyears,withthefurtherdevelopmentofvariousnewtechnologiesandnewfields,theembeddedfielditselfhasalsobeendevelopedintoseveraldifferentsub-fieldsandhasdifferentiated.
Firstofall,withthedevelopmentofsmartphones(MobileSmartPhone)andhandhelddevices(MobileDevices),themobilefieldhasgraduallydevelopedintoanindependentfieldthatrivalsorevenexceedsthePCfieldinscale.BecausetheprocessorintheMobilefieldneedstoloadtheLinuxoperatingsystemandinvolvesacomplexsoftwareecosystem,ithasthesameseriousdependenceonthesoftwareecosystemasthePCfield.
Nextisthereal-time(RealTime)embeddedfield.Relativelyspeaking,thisfielddoesnothavesuchserioussoftwaredependence,sothereisnoabsolutemonopoly.However,duetothesuccessfulcommercialpromotionofARMprocessorIP,ARM'sprocessorarchitecturestillaccountsformostofthemarketshare.OtherprocessorarchitecturessuchasSynopsysARC,etc.alsohavegoodmarketresults.
Finallyisthedeepembeddedfield.Thisfieldismorelikethetraditionalembeddedfieldmentionedearlier.Thedemandinthisfieldisverylarge,butittendstofocusonlowpowerconsumption,lowcostandhighenergyefficiency.ThereisnoneedtoloadalargeapplicationoperatingsystemlikeLinux.Mostofthesoftwarerequirescustomizedbaremetalprogramsorsimplereal-timeoperatingsystems.Therefore,Thedependenceonthesoftwareecologyisrelativelylow.
MainframeCPU
Mainframe,ormainframe.Mainframesusededicatedprocessorinstructionsets,operatingsystems,andapplicationsoftware.Thetermmainframeoriginallyreferredtoalargecomputersystemhousedinaverylargeironboxwithaframetodistinguishitfromsmallerminicomputersandmicrocomputers.
ReducingmainframeCPUconsumptionisanimportanttask.SavingeachCPUcyclecannotonlydelayhardwareupgrades,butalsoreducesoftwarelicensingfeesbasedonthescaleofuse.
Themainframearchitecturemainlyincludesthefollowingtwopoints:ahighdegreeofvirtualization,andallsystemresourcesareshared.Themainframecanintegratealargenumberofloadsintooneandmaximizeresourceutilization;asynchronousI/Ooperations.Thatis,whentheI/Ooperationisperformed,theCPUhandstheI/OinstructionstotheI/Osubsystemforcompletion,andtheCPUitselfisreleasedtoexecuteotherinstructions.Therefore,thehostcanperformothertasksatthesametimewhileperformingheavyI/Otasks.
Controltechnologyform
Thepowerfuldataprocessingfunctionofthecentralprocessingunithaseffectivelyimprovedtheworkefficiencyofthecomputer.Inthedataprocessingoperation,itisnotjustasimpleoperation.Theoperationoftheprocessorisbasedontheinstructiontaskissuedbythecomputeruser,andthecontrolinstructioninputbytheusercorrespondstotheCPUduringtheexecutionoftheinstructiontask.Withtherapiddevelopmentofinformationtechnologyinourcountry,computersarewidelyusedinpeople'slives,work,andcorporateofficeautomation.Asamastercontroldevice,theypromotethedevelopmentofe-commercenetworksandupgradetheCPUcontrolperformance.Theprocesshasbeengreatlyimproved.Commandcontrol,actualcontrol,operationcontrol,etc.aretheperformanceoftheapplicationofcomputerCPUtechnology.
(1)Selectioncontrol.Theoperationofthecentralizedprocessingmodeisimplementedonthebasisofspecificprograminstructionstomeettheneedsofcomputerusers.TheCPUcanchooseaccordingtotheactualsituationduringtheoperationtomeettheuser'sdataflowrequirements.Theimportantroleplayedbycommandandcontroltechnology.Accordingtotheneedsofusers,thecalculationmethodisformulated,sothattheorderlyformulationofdatacommandactionsiswellmaintained.DuringtheexecutionoftheCPU,theimplementationoftheinstructionsoftheprogramiscompletedsmoothly.Onlybyfollowingacertainsequencecanthecomputeruseeffectbeguaranteed.TheCPUismainlyusedtoexpandtheautomaticprocessingofdatasets,whichisthekeytorealizecentralizedcontrol,anditscoreisinstructioncontroloperation.
(2)Insertioncontrol.ThegenerationoftheoperationcontrolsignalbytheCPUismainlyrealizedbythefunctionofinstructions,andthepurposeofcontrollingthesecomponentsisachievedbysendingtheinstructionstothecorrespondingcomponents.Therealizationofaninstructionfunctionismainlyaccomplishedbyasequenceofoperationsperformedbythecomponentsinthecomputer.Moresmallcontrolcomponentsarethekeytobuildingacentralizedprocessingmode,thepurposeistobettercompletetheCPUdataprocessingoperations.
(3)Timecontrol.Applyingtimetimingtovariousoperationsistheso-calledtimecontrol.Whenaninstructionisexecuted,itshouldbecompletedwithinthespecifiedtime.TheinstructionoftheCPUisfetchedfromthecacheormemory,andthentheinstructiondecodingoperationisperformed,whichismainlyimplementedintheinstructionregister.Inthisprocess,itisnecessarytoPayattentiontostrictlycontroltheprogramtime.
Securityissues
TheflourishingdevelopmentofCPUsalsobringsmanysecurityissues.TheFDIVbug(Pentiumfloatingpointdivisionerror)thatappearedonthePentiumprocessorin1994cancauseerrorsinthedivisionoffloating-pointnumbers;theF00FabnormalinstructiononthePentiumprocessorin1997cancausetheCPUtocrash;theIntelprocessortrustedexecutiontechnologyin2011(TXT,trustedexecutiontechnology)hasabufferoverflowproblem,whichcanbeusedbyattackersforprivilegeescalation;avulnerabilityintheIntelmanagementengine(ME,managementengine)componentin2017canleadtoremoteunauthorizedarbitrarycodeexecution;in2018,TwoCPUvulnerabilitiesinMeltdownandSpectrehaveaffectedalmosteverycomputingdevicemanufacturedinthepast20years,leavingtheprivateinformationstoredonbillionsofdevicesatriskofbeingleaked.Thesesecurityissueshaveseriouslyendangerednationalnetworksecurity,criticalinfrastructuresecurity,andinformationsecurityinimportantindustries,andhavecausedorwillcausehugelosses.
ComparisonwithGPU
GPU
GPUistheimageprocessor.TheworkflowandphysicalstructureofCPUandGPUareroughlysimilar.ComparedtoCPUInotherwords,theworkoftheGPUismoreunitary.Inmostpersonalcomputers,theGPUisonlyusedtodrawimages.IftheCPUwantstodrawatwo-dimensionalgraph,itonlyneedstosendaninstructiontotheGPU,andtheGPUcanquicklycalculateallthepixelsofthegraphanddrawthecorrespondinggraphatthespecifiedpositiononthedisplay.SincetheGPUgeneratesalotofheat,thereisusuallyaseparateheatsinkonthegraphicscard.
Designstructure
CPUhasapowerfularithmeticoperationunit,whichcancompletearithmeticcalculationsinafewclockcycles.Atthesametime,thereisalargecachethatcanstorealotofdatainit.Inaddition,therearecomplexlogiccontrolunits,whichreducethedelaybyprovidingbranchpredictioncapabilitieswhentheprogramhasmultiplebranches.GPUisbasedonalargethroughputdesign,withmanyarithmeticoperationunitsandveryfewcaches.Atthesametime,GPUsupportsalargenumberofthreadstorunatthesametime.Iftheyneedtoaccessthesamedata,thecachewillmergetheseaccesses,whichwillnaturallycausedelayproblems.Despitethedelay,becauseofthelargenumberofarithmeticoperationunits,itcanachieveaverylargethroughputeffect.
Usagescenarios
Obviously,becausetheCPUhasalargenumberofcachesandcomplexlogiccontrolunits,itisverygoodatlogiccontrolandserialoperations.Incomparison,GPUhasalargenumberofarithmeticoperationunits,soitcanperformalargenumberofcalculationsatthesametime.Itisgoodatlarge-scaleconcurrentcalculations.Thecalculationislargebutthereisnotechnicalcontent,andithastoberepeatedmanytimes.Inthisway,thewayweuseGPUtoincreasethespeedofprogramoperationisobvious.UsingCPUtodocomplexlogiccontrolandGPUtodosimplebutlargearithmeticoperationscangreatlyimprovetherunningspeedoftheprogram.
Futuredevelopment
Thegeneral-purposecentralprocessingunit(CPU)chipisabasiccomponentoftheinformationindustryandacorecomponentofweaponry.mycountry'slackofCPUtechnologyandindustrywithindependentintellectualpropertyrightshasnotonlycausedtheinformationindustrytobecontrolledbyothers,butalsothenationalsecuritycannotbefullyguaranteed.Duringthe"TenthFive-YearPlan"period,thenational"863Program"begantosupportindependentresearchanddevelopmentofCPUs.Duringthe"EleventhFive-YearPlan"period,the"CoreElectronicDevices,High-endGeneral-purposeChipsandBasicSoftwareProducts"("NuclearHighBase")majorprojectintroducedtheCPUachievementsofthe"863Program"intotheindustry.Startingfromthe"TwelfthFive-YearPlan"period,mycountryhascarriedoutindependentresearchanddevelopmentofCPUapplicationsandpilotprojectsinmanyfields,forminganindependenttechnologyandindustrialsystemwithinacertainrange,whichcanmeettheapplicationneedsofweapons,equipment,informationtechnologyandotherfields.However,foreignCPUmonopolizationhasbeenforalongtime,anditwilltakesometimeformycountry'sindependentresearchanddevelopmentofCPUproductsandthematurityofthemarket.