計(jì)算機(jī)專(zhuān)業(yè)外語(yǔ)翻譯

計(jì)算機(jī)專(zhuān)業(yè)外語(yǔ)STYLEREF"標(biāo)題1,章標(biāo)題"Chapter9Multimediachapter1PrinciplesOfComputerOrganizationINTRODUTIONThegenericcomputercontainsinputdevices,acomputationalunit,andoutputdevices.Thecomputationalunitisthecomputer'sheart,andusuallyconsistsofacentralprocessingunit(CPU),amemory,aninput/output(I/O)interfaceandotherimportantcomponents.Togetabetterunderstandingofcomputerorganizaion,abriefoverviewofthecomputersorganizationisgivenintheflowing.一般計(jì)算機(jī)包含輸入設(shè)備，計(jì)算單元，和輸出設(shè)備。計(jì)算單元是計(jì)算機(jī)的心臟，通常包括一個(gè)中央處理單元（CPU），存儲(chǔ)器，輸入/輸出（I/O）接口和其他重要的組成部分。為了更好地了解計(jì)算機(jī)組成，下面簡(jiǎn)要概述了計(jì)算機(jī)的組成。1.1COMPUTERHARDWAREWebuildcomputertosolveproblems.Earlycomputersolvedmathematicalandengineeringproblems,andlatercomputersemphasizedinformationprocessingforbusinessapplications.Today,computersalsocontrolmachinesasdiverseasautomobileengines,robots,andmicrowaveovens.Acomputersystemsolvesaproblemfromanyofthesedomainsbyacceptinginput,processingit,andproducingoutput.我們創(chuàng)造計(jì)算機(jī)是為了解決問(wèn)題。早期的計(jì)算機(jī)解決數(shù)學(xué)和工程問(wèn)題，后來(lái)的計(jì)算機(jī)主要用于商業(yè)應(yīng)用的信息處理方面。今天，電腦還可以控制機(jī)器設(shè)備，如：汽車(chē)發(fā)動(dòng)機(jī)，機(jī)器人，微波爐。計(jì)算機(jī)系統(tǒng)通過(guò)接受輸入，處理，并產(chǎn)生輸出來(lái)解決這些領(lǐng)域中的相關(guān)問(wèn)題。Fig.1.1thethreeactivitiesofacomputersystemComputersystemsconsistofhardwareandsoftware.Hardwareisthephysicalpartofthesystem.Oncedesigned,hardwareisdifficultandexpensivetochange.Softwareisthesetofprogramsthatinstructthehardwareandiseasiertomodifythanhardware.Computersarevaluablebecausetheyaregeneral-purposemachinesthatcansolvemanydifferentkindsofproblems,asopposedtospecial-purposemachinesthatcaneachsolveonlyonekindofproblem[1].Differentproblemscanbesolvedwiththesamehardwarebysupplyingthesystemwithadifferentsetofinstructions,thatis,withdifferentsoftware.計(jì)算機(jī)系統(tǒng)包括硬件和軟件。硬件是計(jì)算機(jī)系統(tǒng)的物理部分，一旦設(shè)計(jì)，改變硬件是一件很難而且昂貴的事情。軟件是指示硬件工作的一套程序，與改變硬件相比更容易修改。由于計(jì)算機(jī)是能夠解決多種不同類(lèi)型問(wèn)題的通用機(jī)器，而不是僅解決某一類(lèi)問(wèn)題的專(zhuān)用機(jī)器，它們具有（重要）價(jià)值[1]。通過(guò)給系統(tǒng)提供不同的指令集，也就是說(shuō)，不同的軟件，可以使相同的硬件解決不同的問(wèn)題。Everycomputerhasfourbasichardwarecomponents.Inputdevices.Outputdevices.Mainmemory.Centralprocessingunit(CPU).每臺(tái)計(jì)算機(jī)都由四個(gè)基本硬件組件。輸入設(shè)備.輸出設(shè)備.主存儲(chǔ)器.中央處理單元(CPU).Fig.1.2showsthesecomponentsinablockdiagram.Thelinesbetweentheblocksrepresenttheflowofinformationflowsfromonecomponenttoanotheronthebus,whichissimplyagroupofwiresconnectingthecomponents.ProcessingoccursintheCPUandmainmemory.TheorganizationinFig.1.2,withthecomponentsconnectedtoeachotherbythebus,iscommon.However,otherconfigurationsarepossibleaswell.圖1.2在結(jié)構(gòu)圖中顯示了這些組件。模塊之間的連線代表信息流通過(guò)總線從一個(gè)組件流入到另一個(gè)組件的流向，總線只不過(guò)是連接組成部分的一組電線。信息處理發(fā)生在CPU和主存儲(chǔ)器。從組織結(jié)構(gòu)圖1.2，組件之間通常是通過(guò)總線彼此相連的。然而，其他配置也是可能的。Computerhardwareisoftenclassifiedbyitsrelativephysicalsize.Smallmicrocomputer;Mediumminicomputer;Largemainframe.Fig.1.2Blockdiagramofthefourcomponentsofacomputersystem計(jì)算機(jī)硬件通常是按相對(duì)物理尺寸來(lái)劃分。微型計(jì)算機(jī);中等小型機(jī);大型主機(jī).JusttheCPUofamainframeoftenoccupiesanentirecabinet.Itsinput/output(I/O)devicesandmemorymightfillanentireroom.Microcomputerscanbesmallenoughtofitonadeskorinabriefcase.Astechnologyadvances,theamountofprocessingpreviouslypossibleonlyonlargemachinesbecomespossibleonsmallermachines.Microcomputersnowcandomuchoftheworkthatonlyminicomputersormainframescoulddointhepast.大型主機(jī)的CPU主機(jī)往往占據(jù)整個(gè)櫥柜。它的輸入/輸出（I/O）設(shè)備和內(nèi)存也可能填滿整個(gè)房間。微型計(jì)算機(jī)可以足夠小，適合于辦公桌或公文包里。隨著技術(shù)的進(jìn)步，以前只能在大型機(jī)上的處理數(shù)據(jù)量現(xiàn)在在小型機(jī)上變?yōu)榭赡?。微機(jī)現(xiàn)在可以做許多以前只有小型機(jī)或大型機(jī)可以做的工作。Theclassificationjustdescribedisbasedonphysicalsizeasopposedtostoragesize.Acomputersystemuserisgenerallymoreconcernedwithstoragesize,becausethatisamoredirectindicationoftheamountofusefulworkthatthehardwarecanperform[2].剛才所描述的分類(lèi)是基于物理尺寸，而不是存儲(chǔ)容量。計(jì)算機(jī)用戶通常更關(guān)心存儲(chǔ)容量的大小，因?yàn)樗怯布軌驁?zhí)行的有用工作量的更直接的指示。Speedofcomputationisanothercharacteristicthatisimportanttotheuser.Generallyspeaking,userswantafastCPUandlargeamountsofstorage,butaphysicallysmallmachinefortheI/Odevicesandmainmemory.計(jì)算速度對(duì)于用戶來(lái)說(shuō)是計(jì)算機(jī)的另一個(gè)特征。一般來(lái)說(shuō)，用戶除了需要一個(gè)小體積機(jī)器的I/O設(shè)備和主存儲(chǔ)器外，還需要一個(gè)快速的CPU和大量的存儲(chǔ)空間。Whencomputerscientistsstudyproblems,therefore,theyareconcernedwithspaceandtime-thespacenecessaryinsideacomputersystemtostoreaproblemandthetimerequiredtosolveit.TheycommonlyusethemetricprefixesofTABLE1.1toexpresslargeorsmallquantitiesofspaceortime.因此，當(dāng)計(jì)算機(jī)科學(xué)家研究問(wèn)題的時(shí)候，他們所關(guān)心的是空間和時(shí)間，也就是在計(jì)算機(jī)系統(tǒng)里面存儲(chǔ)問(wèn)題的必要的空間內(nèi)和解決問(wèn)題所需的時(shí)間。他們通常使用表1.1的公制前綴來(lái)表示，空間或時(shí)間的大小量。Table1.1Prefixesforpowersof10MultiplePrefixAbbrev109giga-G106mega-M103kilo-K10-3milli-m10-6micro-u10-9nano-nExampleSupposeittakes4.5microseconds,alsowritten4.5us,totransfersomeinformationacrossthebusfromonecomponenttoanother.(a)Howmanysecondsarerequiredforthetransfer?(b)Howmanytransferscantakeplaceduringoneminute?例如：假設(shè)要通過(guò)總線從一個(gè)組件傳遞一些信息到另一個(gè)組件需要4.5微秒，也可以寫(xiě)成4.5us。（a）傳遞需要多少秒？（b）在一分鐘內(nèi)可以傳遞多少？(a)Atimeof4.5psis4.5X10-6fromTABLE1.1or0.0000045s.(b)Becausethereare60secondsinoneminute,thenumberoftimesthetransfercanoccuris(60s)/(0.0000045s/transfer)or13300000transfers.Notethatsincetheoriginalvaluewasgivenwithtwosignificantfigures,theresultshouldnotbegiventomorethantwoorthreesignificantfigures.（a）4.5ps從表1.1是4.5×10-6或0.0000045秒（b）由于一分鐘有60秒，可能發(fā)生的傳遞次數(shù)是（60s）/（0.0000045秒/傳遞）或13300000傳遞。請(qǐng)注意，由于原值是帶有兩個(gè)有效數(shù)字，其結(jié)果不應(yīng)該考慮到多于兩個(gè)或三個(gè)有效數(shù)字。Table1.1showsthatinthemetricsystemtheprefixkilo-is1000andmega-is1000000.Butincomputerscience,akilo-is210or1024.Thedifferentbetween1000and1024islessthan3%,soyoucanthinkofacomputersciencekilo-asbeingabout1000eventhoughitisalittlemore.Thesameapplies'tomega-andgiga-,asinTable1.2.Thistime,theapproximationisalittleworse,butformega-itisstillwithin5%.Table1.1表明，在公制系統(tǒng)前綴中，前綴kilo是1000和前綴mega是特大1000000。但是，在計(jì)算機(jī)科學(xué)中，一千，是210或1024。1000和1024之間的差距小于3％，所以你可以認(rèn)為計(jì)算機(jī)科學(xué)kilo約為1000盡管它稍微多一點(diǎn)。在Table1.2里mega（兆）和千兆同樣也適用。這時(shí)這一次，兩者的近似值差距稍微大一些，但對(duì)于兆來(lái)說(shuō)它仍然在5％以?xún)?nèi)。Table1.2ComputersciencevaluesofthelargeprefixesPrefixComputersciencevaluegiga-230=1073741824mega-220=1048576kilo-210=1024KeywordsandAbbreviations:computer計(jì)算機(jī)inputdevice輸入設(shè)備informationprocessing信息處理outputdevice輸出設(shè)備hardware硬件mainmemory主存儲(chǔ)器software軟件bus總線program程序instruction指令microcomputer微型計(jì)算機(jī)minicomputer小型計(jì)算機(jī)mainframe主機(jī)，特大型general-purposemachine通用計(jì)算機(jī)centralprocessingunit(CPU)中央處理器pecial-purposemachine專(zhuān)用計(jì)算機(jī)Notes：[1]Computersarevaluablebecausetheyaregeneral-purposemachinesthatcansolvemanydifferentkindsofproblems,asopposedtospecial-purposemachinesthatcaneachsolveonlyonekindofproblem.由于計(jì)算機(jī)是能夠解決多種不同類(lèi)型問(wèn)題的通用機(jī)器，而不是僅解決某一類(lèi)問(wèn)題的專(zhuān)用機(jī)器，它們具有（重要）價(jià)值。[2]Acomputersystemuserisgenerallymoreconcernedwithstoragesize,becausethatisamoredirectindicationoftheamountofusefulworkthatthehardwarecanperform.計(jì)算機(jī)用戶通常更關(guān)心存儲(chǔ)容量的大小，因?yàn)樗怯布軌驁?zhí)行的有用工作量的更直接的指示。1.2PROCESSORAprocessorisafunctionalunitthatinterpretsandcarriesoutinstructions.EveryprocessorcomeswithauniquesetofoperationssuchasADD,STORE,orLOADthatrepresenttheprocessor'sinstructionset[1].Computerdesignersarefondofcallingtheircomputersmachines,sotheinstructionsetissometimesreferredtoasmachineinstructionsandthebinarylanguageinwhichtheyarewritteniscalledmachinelanguage!Youshouldn'tconfusetheprocessor'sinstructionsetwiththeinstructionsfoundinhigh-levelprogramminglanguages,suchasBASICorPascal.處理器是解釋并執(zhí)行指令的功能部件。每個(gè)處理器都有一個(gè)諸如ADD、STORE或LOAD之類(lèi)的操作集，這個(gè)操作集就是該處理器的指令系統(tǒng)。計(jì)算機(jī)系統(tǒng)設(shè)計(jì)者習(xí)慣將計(jì)算機(jī)稱(chēng)為機(jī)器，所以該指令系統(tǒng)有時(shí)也稱(chēng)作機(jī)器指令系統(tǒng)，而書(shū)寫(xiě)指令的二進(jìn)制語(yǔ)言叫做機(jī)器語(yǔ)言。注意，不要將處理器的指令系統(tǒng)與BASIC或PASCAL這樣的高級(jí)程序設(shè)計(jì)語(yǔ)言中的指令系統(tǒng)相混淆。Aninstructionismadeupofoperationsthatspecifythefunctiontobeperformedandoperandsthatrepresentthedatatobeoperatedon.Forexample,ifaninstructionistoperformtheoperationofaddingtwonumbers,itmustknow(1)whatthetwonumbersareand(2)wherethetwonumbersare.Whenthenumbersarestoredinthecomputer'smemory,theyhavetheiraddressestoindicatewheretheyare.Soifanoperandreferstodatainthecomputer'smemoryitiscalledanaddress.Theprocessor'sjobistoretrieveinstructionsandoperandsfrommemoryandtoperformeachoperation[2].Havingdonethat,itsignalsmemorytosenditthenextinstruction..指令由操作碼和操作數(shù)組成，其中操作碼指明要完成的操作功能，而操作數(shù)則表示所處理的數(shù)據(jù)。例如，一條指令要完成兩數(shù)相加的操作，它就必須知道：（1）這兩個(gè)數(shù)是什么？（2）這兩個(gè)數(shù)在哪兒？當(dāng)參與操作的數(shù)存儲(chǔ)在計(jì)算機(jī)內(nèi)存中時(shí)，應(yīng)有指明該操作數(shù)存放的內(nèi)存地址，所以，如果操作數(shù)表示的是計(jì)算機(jī)內(nèi)存中的數(shù)據(jù)，則該操作數(shù)叫做地址。處理器的工作就是從存儲(chǔ)器中獲得指令和操作數(shù)，并執(zhí)行每個(gè)操作。處理器完成這些工作后就通知存儲(chǔ)器送來(lái)下一條指令。Thisstep-by-stepoperationisrepeatedoverandoveragainatawesomeSpeed.Atimercalledaclockreleasespreciselytimedelectricalsignalsthatprovidearegularpulsefortheprocessor'swork.Thetermthatisusedtomeasurethecomputer'sspeedisborrowedfromthedomainofelectricalengineeringandiscalledamegahertz(MHz),whichmeansmillioncyclespersecond.Forexample,inan8-MHzprocessor,thecomputer'sclockticks8milliontimestoevery1secondtickofanordinaryclock.處理器以驚人的速度一遍又一遍地重復(fù)以上一步步的操作。一個(gè)稱(chēng)作時(shí)鐘的計(jì)時(shí)器準(zhǔn)確地發(fā)出定時(shí)電信號(hào)，該信號(hào)為處理器工作提供有規(guī)律的脈沖。測(cè)量計(jì)算機(jī)速度的術(shù)語(yǔ)是從電工領(lǐng)域借用來(lái)的，稱(chēng)作兆赫（MHz），兆赫意指每秒百萬(wàn)個(gè)周期。例如，普通時(shí)鐘每秒滴答一次，而在8MHz的處理器中，計(jì)算機(jī)的時(shí)鐘則滴答了800萬(wàn)次。Aprocessoriscomposedoftwofunctionalunitsacontrolunitandanarithmetic/logicunit--andasetofspecialworkspacescalledregisters.處理器由兩個(gè)功能部件（控制部件和算術(shù)/邏輯部件）和一組稱(chēng)作寄存器的特殊工作空間組成。1.TheControlUnitThecontrolunitisthefunctionalunitthatisresponsibleforsupervisingtheoperationoftheentirecomputersystem.Insomeways,itisanalogoustoatelephoneswitch-boardwithintelligencebecauseitmakestheconnectionsbetweenvariousfunctionalunitsofthecomputersystemandcallsintooperationeachunitthatisrequiredbytheprogramcurrentlyinoperation.I控制部件控制部件是負(fù)責(zé)監(jiān)控整個(gè)計(jì)算機(jī)系統(tǒng)操作的功能部件。從某些方面看，它類(lèi)似于智能電話交換機(jī)，因?yàn)樗鼘⒂?jì)算機(jī)系統(tǒng)的各功能部件連接起來(lái)，并按正在執(zhí)行中的程序的需要，控制每個(gè)部件完成操作。Thecontrolunitfetchesinstructionsfrommemoryanddeterminestheirtypesordecodesthem.Itthenbreakseachinstructionintoaseriesofsimplesmallstepsoractions.Bydoingthis,itcontrolsthestep-by-stepoperationoftheentirecomputersystem.控制部件從存儲(chǔ)器中取出指令，并確定其類(lèi)型或?qū)ζ溥M(jìn)行譯碼，然后將每條指令分解成一系列簡(jiǎn)單的、很小的步驟或動(dòng)作。這樣，它就控制了整個(gè)計(jì)算機(jī)系統(tǒng)的每步操作。2.TheArithmeticandLogicUnitThearithmeticandlogicunit(ALU)isthefunctionalunitthatprovidesthecomputerwithlogicalandcomputationalcapabilities.DataarebroughtintotheALUbythecontrolunit,andtheALUperformswhateverarithmeticorlogicoperationsarerequiredtohelpcarryouttheinstruction.II算術(shù)/邏輯部件算術(shù)/邏輯部件（ALU）是為計(jì)算機(jī)提供邏輯及計(jì)算功能的部件。控制部件將數(shù)據(jù)送到算術(shù)/邏輯部件中，然后由算術(shù)/邏輯部件完成執(zhí)行指令所需的算術(shù)或邏輯操作。Arithmeticoperationsincludeadding,subtracting,multiplying,anddividing.Logicoperationsmakeacomparisonandtakeactionbasedontheresults.Forexample,twonumbersmightbecomparedtodetermineiftheyareequal.Iftheyareequal,processingwillcontinue;iftheyarenotequal,processingwillstop.算術(shù)運(yùn)算包括加、減、乘、除。邏輯運(yùn)算完成比較，并根據(jù)結(jié)果選擇操作。例如，比較兩個(gè)數(shù)是否相等，如果相等，則繼續(xù)處理；如果不等，則停止處理。3.RegistersAregisterisastoragelocationinsidetheprocessor.Registersinthecontrolunitareusedtokeeptrackoftheoverallstatusoftheprogramthatisrunning.Controlunitregistersstoreinformationsuchasthecurrentinstruction,thelocationofthenextinstructiontobeexecuted,andtheoperandsoftheinstruction.IntheALU,registersstoredataitemsthatareadded,subtracted,multiplied,divided,andcompared.Otherregistersstoretheresultsofarithmeticandlogicoperations.III寄存器寄存器是處理器內(nèi)部的存儲(chǔ)單元?？刂撇考械募拇嫫饔脕?lái)跟蹤正在運(yùn)行的程序的總體狀態(tài)，它存儲(chǔ)如當(dāng)前指令、下一條將要執(zhí)行的指令地址以及當(dāng)前指令的操作數(shù)等信息。在算術(shù)邏輯部件中，寄存器存放要進(jìn)行加、減、乘、除和比較的數(shù)據(jù)項(xiàng)，而其他寄存器則存放算術(shù)運(yùn)算及邏輯運(yùn)算的結(jié)果。Animportantfactorthataffectsthespeedandperformanceofaprocessoristhesizeoftheregisters.Technically,thetermwordsize(alsocalledwordlength)describesthesizeofanoperandregister,butitisalsousedmorelooselytodescribethesizeofthepathwaystoandfromtheprocessor.Currently,wordsizesingeneralpurposecomputersrangefrom8to64bits.Iftheoperandregistersofaprocessorare16bitswide,theprocessorissaidtobea16-bitprocessor.影響處理器速度和性能的一個(gè)重要因素是寄存器的大小。技術(shù)上講術(shù)語(yǔ)“字的大小”（也稱(chēng)字長(zhǎng)）描述了操作數(shù)寄存器的大小，但它也可用來(lái)不那么嚴(yán)格地描述出入處理器的通道的大小?，F(xiàn)在，通用計(jì)算機(jī)的字長(zhǎng)從8位到64位不等。如果處理器的操作數(shù)寄存器是16位的，那么就稱(chēng)該處理器是16位處理器。KeywordsandAbbreviations:clock時(shí)鐘megahertz(MHz)兆赫processor處理器controlunit控制部件operation操作、操作碼指令Wordsize字長(zhǎng)operand操作數(shù)machinelanguage機(jī)器語(yǔ)言register寄存器arithmeticandlogicunit(ALU)算術(shù)/邏輯部件Notes：[1]EveryprocessorcomeswithauniquesetofoperationssuchasADD,STORE,orLOADthatrepresenttheprocessor'sinstructionset.每個(gè)處理器都具有如加、存儲(chǔ)或裝入這樣代表處理器指令集的一組獨(dú)特的操作集。[2]Theprocessor'sjobistoretrieveinstructionsandoperandsfrommemoryandtoperformeachoperation.處理器的工作是從內(nèi)存檢索指令和操作數(shù)并完成每一個(gè)操作。1.3MEMORYSYSTEMS1.MemorySystemDesiderataThememorysystemhasthreedesiderata.(1)Size:Infinitelylarge,noconstraintsonprogramordatasetsize.(2)Speed:Infinitelyfast,latencyequaltothefastestmemorytechnologyavailable.(3)Cost:Theperbitcostshouldapproachthelowest-costtechnologyavailable.Clearlythesespecificationscannotallbeachievedastheyaremutuallyexclusive[1].However,withthesemiconductorandmagneticmemorytechnologyoftoday,thesespecificationsarecloselyapproximated.1.3存儲(chǔ)系統(tǒng)I存儲(chǔ)系統(tǒng)需求存儲(chǔ)器系統(tǒng)有以下三項(xiàng)需求：（1）容量：無(wú)限大，對(duì)程序或者數(shù)據(jù)集大小沒(méi)有任何約束。（2）速度：無(wú)限快，等待時(shí)間在現(xiàn)有存儲(chǔ)器技術(shù)下是最短的。（3）價(jià)格：每位的價(jià)格在可使用的技術(shù)中應(yīng)接近最低。很明顯，由于這三個(gè)需求相互制約，所以很難全部滿足。不過(guò)，隨著半導(dǎo)體和磁存儲(chǔ)器技術(shù)的發(fā)展，這些需求越來(lái)越接近滿足。2.HierarchicalMemoryInthissectionitisshownhowdesignersimplementapracticalmemorythatapproachestheperformanceofanidealmemoryatreasonablecost.Thismemorysystemhasahierarchyoflevels:Thememoryclosesttotheprocessorisfastandrelativelysmall,buthasahighcostperbit.Thisleveliscalledthecache;Therealmemory,sometimesknownasmainmemory,isslower,larger,andhasalowercostperhitthanthecache;Thelowestlevelinthehierarchyisusuallyamagneticdiskthathasthelongestlatencyandthelowestbandwidth;however,itcanbeverylargeandhasaverylowcostperbit.ThishierarchyisillustratedinFig.1.3.Fig.1.3HierarchicalMemoryNotethatFig.1.3doesnotincludetheprocessorregisterfileinthememoryhierarchy.Theregisterfileisaprogram-managedcacheandisgenerallynotincludedinthememorysystem.Also,therecanbemorethanonecacheinthehierarchy.II存儲(chǔ)的層次結(jié)構(gòu)下面說(shuō)明設(shè)計(jì)者如何在合理的價(jià)格下設(shè)計(jì)出性能理想的實(shí)際存儲(chǔ)器。這一存儲(chǔ)系統(tǒng)具有下述層次結(jié)構(gòu)：最靠近處理器的是快速、容量相對(duì)小，但每位價(jià)格高的存儲(chǔ)器，這一級(jí)叫做高速緩沖存儲(chǔ)器；下一級(jí)是實(shí)存儲(chǔ)器，有時(shí)也稱(chēng)為主存儲(chǔ)器，是速度較低、容量較大、每位價(jià)格比高速緩存低的存儲(chǔ)器；層次結(jié)構(gòu)中最低一層通常是等待時(shí)間最長(zhǎng)，帶寬最窄的磁盤(pán)；但它的容量非常大，每位價(jià)格非常低，如圖1.3所示。圖1.3存儲(chǔ)的層次結(jié)構(gòu)請(qǐng)注意，圖1.3中沒(méi)有包括存儲(chǔ)器層次結(jié)構(gòu)中的處理器寄存器組。寄存器組是程序控制的高速緩存，一般不包括在存儲(chǔ)系統(tǒng)中。此外，在存儲(chǔ)器層次中可以有不只一個(gè)高速緩存。3.PagedVirtualMemoryPagedvirtualmemoryprovidesthesolutiontothefirstdesideratumofaverylargememory'sbeingavailabletotheprocessor.Becauseoftheimportanceofthisdesideratum,therelationshipbetweenvirtualandrealmemoryisdiscussedfirst.Withvirtualmemory,theprocessoraddressesthedisk,thelarge,slowendofthehierarchy.Thememoriesbetweentheprocessorandthediskaretheretoincreasetheeffectiveperformance(reducedlatencyandincreasedbandwidth)oftheveryslowdisk.III頁(yè)式虛擬存儲(chǔ)器頁(yè)式虛擬存儲(chǔ)器為處理器可用的超大的存儲(chǔ)器容量的需求提供了解決辦法。由于這一需求很重要，故先討論虛擬存儲(chǔ)器與實(shí)存儲(chǔ)器之間的關(guān)系。處理器以虛擬存儲(chǔ)器的方式訪問(wèn)位于層次結(jié)構(gòu)最底層的大容量、低速磁盤(pán)。在處理器和磁盤(pán)之間放置存儲(chǔ)器是為了提高低速磁盤(pán)的有效性能（減少等待時(shí)間和增加帶寬）。Ifeveryinstructionanddatareferenceweremadetothedisk,theprocessorperformancewouldbeslowindeed.Thenwhyisvirtualmemorysoimportant?Largememoryisneededforlargeprogramsanddatasets.Earlycomputerswithsmallrealmemoryrequiredthatthetransferofdatabetweenrealmemoryanddiskbemanagedexplicitlybytheoperatingsystemortheuser.Virtualmemoryprovidesforautomaticmanagementofthisportionofthememoryhierarchythroughacombinationofhardwareandsoftwareaids[2].如果每條指令和數(shù)據(jù)引用都要訪問(wèn)磁盤(pán)，則處理器的性能必定很差。那么為什么虛擬存儲(chǔ)器如此重要？大型程序和數(shù)據(jù)集需要大容量存儲(chǔ)器。早期的計(jì)算機(jī)實(shí)存儲(chǔ)器容量很小，要求由操作系統(tǒng)或用戶直接管理實(shí)存儲(chǔ)器和磁盤(pán)之間的數(shù)據(jù)傳送。虛擬存儲(chǔ)器通過(guò)硬件和軟件的結(jié)合自動(dòng)地管理存儲(chǔ)器層次結(jié)構(gòu)的這一部分。ThevirtualmemoryinterfaceisshowninFig.1.4.Arealmemoryof16Mbytesandavirtualmemoryof2Gbytesareshownforillustration;manymodernvirtual-memorysystemsaremuchlargerthanthis.Virtual-memoryspaceisdividedintoequal-sizedgroupscalledpages.Apageinamoderncomputeris1K,2K,or4Kbytes.Realmemoryisalsodividedintothesameequal-sizedgroups,calledpageframes.Wheninformationismovedbetweenvirtual-memoryspaceandreal-memoryspace,acompletepageismoved.虛擬存儲(chǔ)器接口如圖1.4所示，圖中畫(huà)出了16MB實(shí)存儲(chǔ)器和2GB虛擬存儲(chǔ)器；現(xiàn)在很多虛擬存儲(chǔ)器系統(tǒng)的容量都比它大很多。虛擬存儲(chǔ)器空間被分成多個(gè)相同容量的組，稱(chēng)為頁(yè)面?，F(xiàn)在計(jì)算機(jī)的頁(yè)面大小是1KB、2KB或4KB。實(shí)存儲(chǔ)器也被分成多個(gè)容量相同的組，叫做頁(yè)幀。信息是整頁(yè)整頁(yè)地在虛擬存儲(chǔ)器和實(shí)存儲(chǔ)器之間傳送的。4.CachesSection3discussedhowvirtualmemoryextendstheaddressspaceofaprocessor.However,thelatencyofrealmemoryistoolongtosupporthigh-performanceprocessors.Evenwiththehigh-speedDRAMsusedtodayforrealmemory,somethingmustbedonetoovercomethislatencyproblem.4高速緩沖存儲(chǔ)器第三部分討論了虛擬存儲(chǔ)器如何擴(kuò)展處理器的地址空間。但是實(shí)存儲(chǔ)器的等待時(shí)間對(duì)高性能處理器來(lái)講仍然太長(zhǎng)。即使在目前很多實(shí)存儲(chǔ)器采用高速動(dòng)態(tài)隨機(jī)存儲(chǔ)器（DRAM）的情況下，為了克服這一等待時(shí)問(wèn)題，還得做一些工作。Thesolutiontothisperformanceproblemistoaddanotherleveltothememoryhierarchy,calledacache,showninFig.1.3.Theallocationofspacesinathree-levelmemoryisshowninFig.1.5.Generally,thevirtual-addressspaceisdividedintoequal-sizedpages.Thesepagesareplacedinreal-memoryframesofthesamesize.Becauseapagecanbeplacedinanyvacantframe,thereisnoparticularordertotheallocationofpagestoframes.Withtheadditionofacache,blocksof16-32bytesaretakenfromthepageframesinrealmemoryandplacedintoablockslotforaccessbytheprocessor.Formodernprocessorsthecacheusuallyhasalatencyofoneprocessorclock,sothatinstructionsanddatamaybefetchedwithoutdelayexceptwhenthereferenceditemisnotinthecache.Fig.1.4PageallocationandaddresstranslationFig.1.5Realmemoryandcacheallocation解決這一性能問(wèn)題的方法是在存儲(chǔ)層次結(jié)構(gòu)中加入另外一層，稱(chēng)為高速緩沖存儲(chǔ)器，如圖1.3所示。圖1.5顯示了3級(jí)存儲(chǔ)空間分配。通常，虛擬地址空間被分為大小相同的頁(yè)，這些頁(yè)放在大小相同的實(shí)存儲(chǔ)器的幀中。因?yàn)橐豁?yè)可以放在任何空幀中，因而對(duì)于幀而言頁(yè)的放置沒(méi)有具體順序的要求。當(dāng)加入了高速緩存之后，實(shí)存儲(chǔ)器中的頁(yè)幀就分成多個(gè)16~32B的數(shù)據(jù)塊，這些塊放到高速緩存中的塊存取窗口中供處理器訪問(wèn)?，F(xiàn)在的處理器中，高速緩存的等待時(shí)間只占一個(gè)處理器時(shí)鐘周期，所以取數(shù)和取指令沒(méi)有時(shí)延，除非要訪問(wèn)的數(shù)據(jù)項(xiàng)不在高速緩存中。圖1.5實(shí)存和高速緩存分配5.MemoryDevices(1)Random-AccessMemoryRandom-accessmemory,orRAM,isthekindofmemoryweusuallyrefertowhenwespeakofcomputermemory.Itisthemostwidelyusedtype,andconsistsofrowsofchipswithlocationsestablishedintablesmaintainedbythecontrolunit.5存儲(chǔ)器設(shè)備（1）隨機(jī)存儲(chǔ)器隨機(jī)存儲(chǔ)器或稱(chēng)RAM，是人們常說(shuō)的計(jì)算機(jī)存儲(chǔ)器。它是使用最廣的一種類(lèi)型，由一組芯片構(gòu)成，其存儲(chǔ)單元由控制部件中的地址表管理。Asthenamesuggests,itemsstoredinRAMcanbegotten(accessed)botheasilyandinanyorder(randomly)ratherthaninsomesequence.RAMreliesonelectriccurrentforallitsoperations;moreover,ifthepoweristurnedofforinterrupted,RAMquicklyemptiesitselfofallyourhardwork.Thus,wesayRAMisvolatile,ornonpermanent.顧名思義，RAM中的內(nèi)容可以很容易地以任意順序（隨機(jī)地）讀出（或稱(chēng)訪問(wèn)），而不用考慮什么先來(lái)后到。RAM始終靠電流進(jìn)行操作；而且，如果電源關(guān)閉或中斷，RAM會(huì)很快丟失人們辛辛苦苦存入的內(nèi)容。因此人們把RAM稱(chēng)為易失性或非永久性的存儲(chǔ)器。(2)Read-OnlyMemoryRead-onlymemory,orROM,typicallyholdsprograms.Theseprogramsaremanufacturedor"hard-wired"inplaceontheROMchips.Forexample,amicrocomputerhasabuilt-inROMchipthatstorescriticalprogramssuchastheonethatstartsup,or"boots",thecomputer.ROMis"slower"thanRAMmemory,andasaresult,itemsinROMaretransferredtoRAMwhenneededforfastprocessing.（2）只讀存儲(chǔ)器只讀存儲(chǔ)器或ROM，主要用來(lái)存儲(chǔ)程序。這些程序是做在ROM芯片中或者說(shuō)通過(guò)硬連線實(shí)現(xiàn)的。例如在微型計(jì)算機(jī)中就有內(nèi)置ROM芯片（有時(shí)稱(chēng)為ROMBIOS，即ROM基本輸入輸出系統(tǒng)），該芯片存儲(chǔ)一些關(guān)鍵的程序，如計(jì)算機(jī)的啟動(dòng)或引導(dǎo)程序。ROM的讀取速度比RAM慢，因此如果需要快速處理，則將ROM中的內(nèi)容傳送到RAM中。ItemsheldinROMcanberead,buttheycannotbechangedorerasedbynormalinputmethods.NewitemscannotbewrittenintoROM.TheonlywaytochangeitemsinmostformsofROMistochangetheactualcircuits.ROM中的內(nèi)容可以讀，但不能用一般的輸入方法更改或擦除。新內(nèi)容也不能寫(xiě)入ROM。在大多數(shù)型號(hào)ROM中改變內(nèi)容的惟一方法是更換實(shí)際電路。(3)MagneticDisksThemagneticdiskisacircularplatterwithasmoothsurfaceandacoatingthatcanbemagnetized.Dataisstoredonitasmagnetizedspots.Thereadingandrecordingdevice,thediskdrive,spinsthediskpastread/writeheadsthatdetectorwritethemagnetizedspotsonthedisk.（3）磁盤(pán)磁盤(pán)是一個(gè)表面光滑且涂敷可磁化材料的圓盤(pán)。數(shù)據(jù)在磁盤(pán)上作為磁化點(diǎn)存入。讀出與記錄設(shè)備即磁盤(pán)驅(qū)動(dòng)器驅(qū)動(dòng)磁盤(pán)轉(zhuǎn)動(dòng)，當(dāng)通過(guò)讀寫(xiě)磁頭時(shí)，磁頭從磁盤(pán)上的磁化點(diǎn)檢測(cè)出或?qū)懭胄畔ⅰ?4)CD-ROMSOpticaldisksneedthinbeamsofconcentratedlighttostoreandreaddata.Itisaformoflaserstorage,calledCD-ROM.Therearetwotypesofopticaldisksthatcanbeuser-recorded:WORManderasableoptical.WORMstandsfor"writeonce,readmany":Datacanbewrittentothisdiskjustonetime,butthedatacanbereadmanytimes.Erasableopticaldiskscanbewrittento,read,anderased.（4）只讀光盤(pán)光盤(pán)用細(xì)聚光光束去存儲(chǔ)和讀出數(shù)據(jù)。只讀光盤(pán)是激光存儲(chǔ)器的一種形式。用戶使用的可讀寫(xiě)光盤(pán)有兩種類(lèi)型：WORM和可擦除的光盤(pán)。WORM代表一次寫(xiě)多次讀：數(shù)據(jù)只能寫(xiě)入一次但可讀出多次?？刹脸獗P(pán)可以讀、寫(xiě)和擦除。(5)MagneticTapeAmagnetictapeisanarrowplasticstripsimilartothetapeusedintaperecorders.Thetapesarereadbytapedrivemovingthetapepastaread/writehead,whichdetectsorwritesmagnetizedspotsontheiron-oxidecoatingofthetape.Eachpatternofspotsmatchesthebytecodeforcharacterbeingstored.（5）磁帶磁帶是一種類(lèi)似于錄音磁帶的窄塑料帶。由磁帶驅(qū)動(dòng)器去讀磁帶的內(nèi)容。驅(qū)動(dòng)器驅(qū)動(dòng)磁帶，由讀寫(xiě)磁頭檢出或?qū)懭胗裳趸F涂敷的磁化點(diǎn)，以各磁化點(diǎn)的不同極性表示所存儲(chǔ)字符的二進(jìn)制代碼。KeywordsandAbbreviations:hierarchicalmemory存儲(chǔ)器層次結(jié)構(gòu)magneticdisk磁盤(pán)diskdrive磁盤(pán)驅(qū)動(dòng)器registerfile寄存器組pageframe頁(yè)幀magnetictape磁帶cache高速緩沖存儲(chǔ)器Mainmemory主存儲(chǔ)器floppydisk(diskette)軟磁盤(pán)latency等待時(shí)間realmemory(storage)實(shí)存儲(chǔ)器pagedvirtualmemory頁(yè)式虛擬存儲(chǔ)器DynamicRAM(DRAM)動(dòng)態(tài)隨機(jī)儲(chǔ)器Read-OnlyMemory(ROM)只讀存儲(chǔ)器CompactDiskROM(CD-ROM)只讀光盤(pán)writeonce,readmany(WORM)一次寫(xiě)，多次讀Random-AccessMemory(RAM)隨機(jī)（訪問(wèn)）存儲(chǔ)Notes：[1]Clearlythesespecificationscannotallbeachievedastheyaremutuallyexclusive.很明顯，在不能共存的情況下，這些技術(shù)指標(biāo)是不能完全達(dá)到的。[2]Virtualmemoryprovidesforautomaticmanagementofthisportionofthememoryhierarchythroughacombinationofhardwareandsoftwareaids.虛擬存儲(chǔ)器在硬件與軟件的共同幫助下實(shí)現(xiàn)了對(duì)內(nèi)存結(jié)構(gòu)的自動(dòng)化管理。1.4Input/OutputSystemProcessorsneedasourceofinputdataandadestinationforoutputdata.Processorsalsorequirestoragespaceforlargevolumesofintermediatedata.ThustheI/Osystemprovidesthesestoragefacilitiesbyconnectingtheprocessor’smemorytotheI/Odevices,thesedevicesvarywidelyinlatencyandbandwidth;thus,amajorfunctionoftheI/Osystemistomatchtheirlatenciesandbandwidthstothatofthememorysystem.I/Odevicesconsistofsuchdiverseitemsaskeyboards,modems,andharddisks.Eachofthesedeviceshasitsownlatencyandbandwidthrequirementsthatmustbesatisfiedforaproperlyfunctioningsystem[1].1.4輸入/輸出系統(tǒng)處理器需要有輸入數(shù)據(jù)的源設(shè)備和用于數(shù)據(jù)輸出的目的設(shè)備。它還要求有存儲(chǔ)大量中間數(shù)據(jù)的存儲(chǔ)空間，I/O系統(tǒng)就是通過(guò)把處理器的存儲(chǔ)器與I/O設(shè)備連接起來(lái)的存儲(chǔ)設(shè)施。這些設(shè)備在等待時(shí)間和帶寬上是大不相同的；因此I/O系統(tǒng)的主要功能是使其在執(zhí)行時(shí)間和帶寬上與存儲(chǔ)系統(tǒng)協(xié)調(diào)起來(lái)。I/O設(shè)備包括鍵盤(pán)、調(diào)制解調(diào)器和硬盤(pán)等多種設(shè)備。對(duì)每種設(shè)備自身的等待時(shí)間和帶寬，必須滿足系統(tǒng)正確操作的要求。ThemajortasksoftheI/OsystemarethefollowingtoestablishtheconnectionbetweenthememoryandtheI/Odevices;tosynchronizeandmanagethecontrolofthedatatransfer;toprovidebufferingwhenthesourceandsinkbandwidthsandlatenciesdiffer;toperformcodeconversionsifrequired;tointerrupttheprocessorwhenrequired;toterminatetheoperationwhenitiscompleted.I/O系統(tǒng)的主要任務(wù)是：（1）建立存儲(chǔ)器和I/O設(shè)備之間的連接；（2）對(duì)數(shù)據(jù)傳輸?shù)目刂七M(jìn)行同步化與管理；（3）當(dāng)信源和信宿的帶寬和等待時(shí)間不同時(shí)，提供緩存功能；（4）需要時(shí)完成代碼的轉(zhuǎn)換；（5）需要時(shí)中斷處理器的工作；（6）任務(wù)完成時(shí)終止操作。Thesetasksareaccomplishedbyhardwareandsoftware,invaryingamountsofeach.Themorehardware-oriented,thehighertheperformanceandthehigherthehardwarecost.Themoresoftware-oriented,thelowertheperformanceandthelowerthehardwarecost.Controlcanbeclassifiedintotwobroadcategories,programmedI/OandcoprocessorI/O.這些任務(wù)都是通過(guò)硬件和軟件完成的，不過(guò)對(duì)硬件或軟件的依賴(lài)程度各不相同。越是依靠硬件，則性能越高，硬件成本越大；越是依靠軟件，則性能越差，硬件費(fèi)用越低。控制方法可以分為兩大類(lèi)：程序控制I/O和協(xié)處理器I/O。1.ProgrammedI/OAlsoknownasdirectI/O,isaccomplishedbyaprogramexecutingontheprocessoritselftocontroltheI/Ooperationsandtransferthedata.WithprogrammedI/O,showninFig.1.6thereareseveralarchitecturalregister(minimallyaninputregisterandanoutputregister)thatareaddressedbyspecialmoveinstructions.Fig.1.6ProgrammedI/OThesemoveinstructionsmoveinformationfromtheaccumulator(orgeneral-purposeregister)to/fromtheI/Oregisters.Theperipheralsareconnecteddirectlytothebitsoftheseregisters.1.程序控制I/O程序控制I/O，又稱(chēng)直接I/O，由在處理器上執(zhí)行的程序去控制I/O操作和傳送數(shù)據(jù)。如圖1.6所示的程序控制I/O，有幾個(gè)體系結(jié)構(gòu)寄存器（至少1個(gè)輸入寄存器和1個(gè)輸出寄存器），它們由專(zhuān)用的傳送指令尋址。圖1.6程序控制I/O這些傳送指令在累加器（或通用寄存器）和I/O寄存器之間傳送信息，而外圍設(shè)備直接與每個(gè)寄存器的位相連。2.Memory-MappedI/OMemory-MappedI/OisanotherformofprogrammedI/Othatmapsthedeviceconnectionstobitsinmemoryaddressspace,asshowninFig.1.7.Anoutputcommandisanormalstoreinstructionthatstoresapatternofbitsintoadesignatedmemorylocation;aninputcommandisanormalloadinstructionthatreadsamemorylocation.Fig.1.7Memory-mappedI/OMemory-mappedI/Orequiresthefullparticipationoftheprocessor,aswithprogrammedI/O.ThusthereisasimilarlossofefficiencythatisduetoexecutingtheI/Ocontrolprogram.2.存儲(chǔ)器映射I/O存儲(chǔ)器映射I/O是另一種程序控制I/O，它把I/O設(shè)備連接映射成存儲(chǔ)器地址空間中的位組，如圖1.7所示。輸出命令是一條普通的存儲(chǔ)指令，把位模式存入指定的存儲(chǔ)單元中。輸入命令是一條普通的從指定存儲(chǔ)單元裝入（讀?。?shù)據(jù)的指令。圖1.7存儲(chǔ)器映射I/O像程序控制I/O一樣，存儲(chǔ)器映射I/O也要求處理器全面參與。由于要執(zhí)行I/O控制程序，故效率同樣降低。TheIntelPentiumProsupportsmemory-mappedI/OinadditiontoprogrammedI/O.Anyoftheprocessor'sinstructionsthatreferencememorycanaddressmemory-mappedI/Olocations.IntelPentiumPro在支持程序控制I/O的同時(shí)，也支持存儲(chǔ)器映射I/O。任何訪問(wèn)內(nèi)存的處理器指令都能訪問(wèn)存儲(chǔ)器映射I/O地址。Themoveinstructioncantransferdatabetweenanyoftheprocessor'sregistersandthespacesinmemoryallocatedforI/O.OtheroperationssuchasADD,AND,OR,andtestcanalsohaveonesourceargumentinI/Ospace.ThelogicaloperationsareusefulforcontrolofbitsinthecontrolandstatusregistersofanI/Odevicethatismappedintooneormoreofthememory-addressspaces.傳送（移動(dòng)）指令可以在任何處理器、寄存器和為I/O分配的內(nèi)存空間之間傳送數(shù)據(jù)。其他操作如“加（ADD）”、“與（AND）”、“或（OR）”以及測(cè)試，也可以有一個(gè)I/O空間的源參量。邏輯操作常用于對(duì)I/O設(shè)備中的控件和狀態(tài)寄存器的位進(jìn)行控制，而每一設(shè)備都可以映射為1個(gè)或多個(gè)存儲(chǔ)器地址空間。3.InterruptI/OBecauseofthelowefficiencyofpureprogrammedormemory-mappedI/Oinhigh-performancesystem,sointerruptisproduced.Earlymainframecomputers,suchastheUnivac1103andtheIBM701wereusersofthistechnique.Theprocessorsendsacommandtothedevice,usuallybymeansofaspecialprogrammedI/Oregister,tostartitsoperation.Aftertherequestismadetothedevice,theprocessorreturnstothemainprogram.Whenthelatencyissatisfiedandthedeviceisreadytotransfer,thedevicesendsaninterruptsignaltotheprocessorandtheprocessorservicestheinterruptandperformsthedatatransfer.VariousstatusreportsaresentfromthedevicetotheprocessorbymeansoftheprogrammedI/Opaths.InterruptI/Ocanbeusedwithbothprogrammedandmemory-mappedI/O；TheIntelPentiumProimplementsbothtechniques.3.中斷I/O由于對(duì)高性能系統(tǒng)來(lái)說(shuō)純程序控制I/O或存儲(chǔ)器映射I/O的效率都很低，所以又出現(xiàn)了中斷方式。早期的大型計(jì)算機(jī)，如Univac1103，IBM701都采用中斷技術(shù)。處理器通常使用專(zhuān)用的程序控制I/O寄存器向設(shè)備發(fā)送一個(gè)啟動(dòng)設(shè)備操作的命令。在向設(shè)備發(fā)出請(qǐng)求之后，處理器返回到主程序上。當(dāng)?shù)却龝r(shí)間