《信息工程专业英语》课件第14章

Unit14IntroductionofSomeTypicalDevices14.1Text14.2ReadingMaterials

14.1Text

IntroducedFPGA

Programmablelogicdeviceisagenericlogiccanuseavarietyofchips,whichistoachieveASIC(ApplicationSpecificIntegratedCircuit)semi-customizeddevice.ItsemergenceanddevelopmentofelectronicsystemsdesignersuseCADtoolstodesigntheirownlaboratoryintheASICdevice.

EspeciallyFPGA(FieldProgrammableGateArray)generatedanddevelopment,asamicroprocessor,memory,thefiguresforelectronicsystemdesignandsetanewindustrystandard(thatisbasedonstandardproductsalescatalogueinthemarkettobuy).Isadigitalsystemformicroprocessors,memories,FPGAorthreestandardbuildingblocksconstitutetheirintegrationdirection.

DigitalcircuitdesignusingFPGAdevices,cannotonlysimplifythedesignprocessandcanreducethesizeandcostoftheentiresystem,increasingsystemreliability.Theydonotneedtospendthetraditionalsensealotoftimeandeffortrequiredtocreateintegratedcircuits,toavoidtheinvestmentriskandbecomethefastest-growingindustriesofelectronicdevicesgroup.DigitalcircuitdesignsystemFPGAdevicesusingthefollowingmainadvantages.

1.Designflexible

UseFPGAdevicesmaynotinthestandardseriesdevicelogicfunctionallimitations.Andchangesinsystemdesignandtheuseoflogicinanyonestageoftheprocess,andonlythroughtheuseofre-programmingtheFPGAdevicecanbecompleted,thesystemdesignprovidesforgreatflexibility.

2.Increasedfunctionaldensity

Functionaldensityinagivenspacereferstothenumberoffunctionalintegrationlogic.Programmablelogicchipcomponentsdoorsseveralhigh,aFPGAcanreplaceseveralfilms,filmscoresorevenhundredsofsmall-scaledigitalICchipillustratedinthefilm.FPGAdevicesusingthechiptousedigitalsystemsinsmallnumbers,thusreducingthenumberofchipsusedtoreducethenumberofprintedsizeandprinted,andwillultimatelyleadtoareductionintheoverallsizeofthesystem.

3.Improvereliability

Printingplatesandreducethenumberofchips,notonlycanreducesystemsize,butitgreatlyenhancedsystemreliability.Ahigherdegreeofintegrationthansystemsinmanylow-standardintegrationcomponentsforthedesignofthesamesystem,withmuchhigherreliability.FPGAdeviceusedtoreducethenumberofchipsrequiredtoachievethesysteminthenumberprintedonthecordandjointsarereduced,thereliabilityofthesystemcanbeimproved.

4.Shorteningthedesigncycle

AsFPGAdevicesandtheprogrammableflexibility,useittodesignasystemforlongerthantraditionalmethodsgreatlyshortened.FPGAdevicemasterdegreeshigh,useprintedcircuitlayoutwiringsimple.Atthesametime,successintheprototypedesign,thedevelopmentofadvancedtools,ahighdegreeofautomation,theirlogicisverysimplechangesquickly.InternalstructureofFPGAisshowninFig14.1.Therefore,theuseofFPGAdevicescansignificantlyshortenthedesigncyclesystem,andspeedupthepaceofproductintothemarket,improvingproductcompetitiveness.

Fig14.1InternalstructureofFPGA

5.Workfast

FPGA/CPLDdevicesworkfast,generallycanreachseveraloriginalHertz,farlargerthantheDSPdevice.Atthesametime,theuseofFPGAdevices,thesystemneededtoachievecircuitclassesandsmall,andthusthepaceofworkoftheentiresystemwillbeimproved.

6.Increasedsystemperformanceconfidentiality

ManyFPGAdeviceshaveencryptionfunctionsinthesystemwidelyusedFPGAdevicescaneffectivelypreventillegalcopyingproductswereothers.

7.Toreducecosts

FPGAdeviceusedtoachievedigitalsystemdesign,ifonlydeviceitselfintotheprice,sometimesyouwouldnotknowitadvantages,buttherearemanyfactorsaffectingthecostofthesystem,takentogether,thecostadvantagesofusingFPGAisobvious.First,theuseofFPGAdevicesdesignedtofacilitatechange,shortendesigncycles,reducedevelopmentcostsforsystemdevelopment;Secondly,thesizeandFPGAdevicesallowautomationneedsplug-ins,reducingthemanufacturingsystemtolowercosts;Again,theuseofFPGAdevicescanenhancesystemreliability,reducedmaintenanceworkload,therebyloweringthecostofmaintenanceservicesforthesystem.Inshort,theuseofFPGAdevicesforsystemdesigntosavecosts.

Technicalwordsandphrases

generic

adj.类的，属性的；一般的

microprocessor n.微处理器

catalogue n.目录

ultimately adv.最后，最终

layout

n.布局，安排，设计

prototype n.原型，雏形，蓝本

confidentiality n.机密性

encryption n.编密码；加密

facilitate vt.促进，助长

plug-in

n.插件程序

semi-customizeddevice 半定制器件

industrystandard

工业标准

functionaldensity

功能密集度

componentsdoor

组件门数

printingplates

印刷板

designcycle

设计周期

speedup

加速

illegalcopying

非法仿制

digitalsystemdesign

数字系统设计

savecosts

节约成本

maintenanceworkload 维修工作量

FPGA(Field-ProgrammableGateArray)

现场可编程门阵列

ASIC(ApplicationSpecificIntegratedCircuit)

专用集成电路

14.1.1Exercises

1.PutthePhrasesintoEnglish

(1)简化设计过程；

(2)投资风险；

(3)可编程逻辑芯片；

(4)自动化程度高；

(5)维修服务费；

(6)产品竞争力。

2.PutthePhrasesintoChinese

(1)productcompetitiveness;

(2)systemreliability;

(3)investmentrisk;

(4)integratedcircuits;

(5)integrationdirection;

(6)standardproductsalescatalogue.

3.Translation

(1)AFPGAcanreplaceseveralfilms,filmscoresorevenhundredsofsmall-scaledigitalICchipillustratedinthefilm.

(2)Ahigherdegreeofintegrationthansystemsinmanylow-standardintegrationcomponentsforthedesignofthesamesystem,withmuchhigherreliability.

(3)AsFPGAdevicesandtheprogrammableflexibility,useittodesignasystemforlongerthantraditionalmethodsgreatlyshortened.

14.1.2参考译文

可编程逻辑器件是一种可以构成各种用途的通用逻辑芯片，它是实现专用集成电路ASIC(ApplicationSpecificIntegratedCircuit)的半定制器件，它的出现和发展使得电子系统设计师借助于CAD手段在实验室里就可以设计自己的ASIC器件。特别是FPGA(FieldProgrammableGateArray)的产生与发展，使其成为继微处理器、存储器之后为电子数字系统设计而确定的又一种新的工业标准(即可以按标准产品目录在销售市场上购到)。数字系统正朝向以微处理器、存储器、FPGA三种标准模块构成的一体化方向发展。

使用FPGA器件设计数字电路，不仅可以简化设计过程，而且可以降低整个系统的体积和成本，增加系统的可靠性。它们无需花费传统意义上制造集成电路所需的大量时间和精力，避免了投资风险，成为电子器件行业中发展最快的一族。使用FPGA器件设计数字系统电路的主要优点如下。

1.设计灵活

使用FPGA器件可不受标准系列器件在逻辑功能上的限制，而且修改逻辑可在系统设计和使用过程的任一阶段中进行，并且只需通过对所用的FPGA器件进行重新编程即可完成，给系统设计提供了很大的灵活性。

2.增大功能密集度

功能密集度是指在给定的空间能集成的逻辑功能数量。可编程逻辑芯片内的组件门数高，一片FPGA可代替几片、几十片乃至上百片中小规模的数字集成电路芯片。用FPGA器件实现数字系统时用的芯片数量少，从而减少了芯片的使用数目，减少了印刷线路板面积和印刷线路板数目，最终实现系统规模的全面缩减。

3.提高可靠性

减少芯片和印刷板数目，不仅能缩小系统规模，而且还极大地提高了系统的可靠性。

具有较高集成度的系统比用许多低集成度的标准组件设计的相同系统具有高得多的可靠性。使用FPGA器件减少了实现系统所需要的芯片数目，印刷线路板上的引线以及焊点数量也随之减少，所以系统的可靠性得以提高。

4．缩短设计周期

由于FPGA器件的可编程性和灵活性，用它来设计一个系统所需时间较传统方法大为缩短。FPGA器件集成度高，使用时印刷线路板电路布局布线简单。同时，在样机设计成功后，由于开发工具先进，自动化程度高，对其进行逻辑修改也十分简便迅速。FPGA的内部结构如图14.1所示。因此，使用FPGA器件可大大缩短系统的设计周期，加快产品投放市场的速度，提高产品的竞争能力。

5.工作速度快

FPGA/CPLD器件的工作速度快，一般可以达到几百兆赫兹，远远大于DSP器件。同时，使用FPGA器件实现系统所需要的电路级数少，因而整个系统的工作速度得到了提高。

6.增加系统的保密性能

很多FPGA器件都具有加密功能，在系统中广泛使用FPGA器件可以有效防止产品被他人非法仿制。

7.降低成本

使用FPGA器件实现数字系统设计时，如果仅从器件本身的价格考虑，有时还看不出它的优势，但是影响系统成本的因素是多方面的，综合考虑，使用FPGA的成本优越性是很明显的。首先，使用FPGA器件修改设计方便，设计周期缩短，使系统的研制开发费用降低；其次，FPGA器件可使印刷线路板面积和需要的插件减少，从而使系统的制造费用降低；再次，使用FPGA器件能使系统的可靠性提高，维修工作量减少，进而使系统的维修服务费用降低。总之，使用FPGA器件进行系统设计能节约成本。

14.2ReadingMaterials

14.2.1ARM7

ARM7(Fig14.2)isagenerationofARMprocessordesigns.ThisgenerationintroducedtheThumb16-bitinstructionsetprovidingimprovedcodedensitycomparedtopreviousdesigns.ThemostwidelyusedARM7designsimplementtheARMv4Tarchitecture,butsomeimplementARMv3orARMv5TEJ.AllthesedesignsuseaVonNeumannarchitecture,thusthefewversionscomprisingacachedonotseparatedataandinstructioncaches.Fig14.2ARM7

SomeARM7coresareobsolete.Onehistoricallysignificantmodel,theARM7DisnotableforhavingintroducedJTAGbasedon-chipdebugging;theprecedingARM6coresdidnotsupportit.The“D”representedaJTAGTAPfordebugging;the“I”denotedanICEBreakerdebugmodulesupportinghardwarebreakpointsandwatchpoints,andlettingthesystembestalledfordebugging.Subsequentcoresincludedandenhancedthissupport.

ARM7-TDMI

TheARM7-TDMI(ARM7-humb+Debug+Multiplier+ICE)processorisa32-bitRISC

CPUdesignedbyARM,andlicensedformanufacturebyanarrayofsemiconductorcompanies.In2009itremainsoneofthemostwidelyusedARMcores,andisfoundinnumerousdeeplyembeddedsystemdesigns.TexasInstrumentslicensedtheARM7-TDMI,whichwasdesignedintotheNokia6110.TheARM7TDMI-Svariantisthesynthesizablecore.

Specifications

Itisaversatileprocessordesignedformobiledevicesandotherlowpowerelectronics.Thisprocessorarchitectureiscapableofupto130MIPSonatypical0.13µmprocess.TheARM7TDMIprocessorcoreimplementsARMArchitecturev4T.Theprocessorsupportsboth32-bitand16-bitinstructionsviatheARMandThumbinstructionsets.

ARM

licensestheprocessortovarioussemiconductorcompanies,whichdesignfullchipsbasedontheARMprocessorarchitecture.

Applications

Perhapsthemostcommonpiecesofelectronicequipmenttohaveusedthisprocessorare:

AudiocontrollerintheSegaDreamcast.

D-LinkDSL-604+WirelessADSLRouter.

iPod

fromApple.

iriver

portabledigitalaudioplayers(theH10usesachipwiththisprocessor).

JuiceBox.

LegoMindstormsNXT.

MostofNokia’smobilephonerange.

NintendoDS(co-processor)andGameBoyAdvancefromNintendo.

PocketStation.

Roomba

500seriesfromiRobot.

SiriusSatelliteRadioreceivers.

ThemainCPUinSternPinballS.A.MSystemgames.

InBuildingAutomation,theAmericanAuto-MatrixBBC-SD(BACnetTouchscreenDisplay)usesanARM7TDMIcore.

Intournamentwaterskiandwakeboardtowboats,PerfectPassspeedcontrol.

ManyautomobilesembedARM7cores.

SamsungmicroSDcardscontainanARM7TDMIcontrollerwith128KBofcode.

14.2.2TMS320C67××SeriesFloating-PointDSPs

TheTMS320C672×isthenextgenerationofTexasInstruments’C67×generationofhigh-performance32-/64-bitfloating-pointdigitalsignalprocessors.TheTMS320C672×includestheTMS320C6727B,TMS320C6726B.TMS320C6722B.andTMS320C6720devices.

EnhancedC67×+CPU.TheC67×+CPUisanenhancedversionoftheC67×CPUusedontheC671×DSPs.ItiscompatiblewiththeC67×CPUbutofferssignificantimprovementsinspeed,codedensity,andfloating-pointperformanceperclockcycle.At350MHz,theCPUiscapableofamaximumperformanceof2800MIPS/2100MFLOPSbyexecutinguptoeightinstructions(sixofwhicharefloating-pointinstructions)inparalleleachcycle.TheCPUnativelysupports32-bitfixed-point,32-bitsingle-precisionfloating-point,and64-bitdouble-precisionfloating-pointarithmetic.

EfficientMemorySystem.Thememorycontrollermapsthelargeon-chip256K-byteRAMand384K-byteROMasunifiedprogram/datamemory.Developmentissimplifiedsincethereisnofixeddivisionbetweenprogramanddatamemorysizeasonsomeotherdevices.

Thememorycontrollersupportssingle-cycledataaccessesfromtheC67×+CPUtotheRAMandROM.UptothreeparallelaccessestotheinternalRAMandROMfromthreeofthefollowingfoursourcesaresupported:

Two64-bitdataaccessesfromtheC67×+CPU

One256-bitprogramfetchfromthecoreandprogramcache

One32-bitdataaccessfromtheperipheralsystem(eitherdMAXorUHPI)

Thelarge(32K-byte)programcachetranslatestoahighhitrateformostapplications.Thispreventsmostprogram/dataaccessconflictstotheon-chipmemory.Italsoenableseffectiveprogramexecutionfromanoff-chipmemorysuchasanSDRAM.

High-PerformanceCrossbarSwitch.Ahigh-performancecrossbarswitchactsasacentralhubbetweenthedifferentbusmasters(CPU，dMAX,UHPI)anddifferenttargets(peripheralsandmemory).Thecrossbarispartiallyconnected;someconnectionsarenotsupported(forexample,UHPI-to-peripheralconnections).

Multipletransfersoccurinparallelthroughthecrossbaraslongasthereisnoconflictbetweenbusmastersforaparticulartarget.Whenaconflictdoesoccur,thearbitrationisasimpleanddeterministicfixed-priorityscheme.

ThedMAXisgivenhighest-prioritysinceitisresponsibleforthemosttime-criticalI/Otransfers,followednextbytheUHPI,andfinallybytheCPU.

dMAXDualDataMovementAccelerator.ThedMAXisamoduledesignedtoperformDataMovementAcceleration.TheDataMovementAccelerator(dMAX)controllerhandlesuser-programmeddatatransfersbetweentheinternaldatamemorycontrollerandthedeviceperipheralsontheC672×DSPs.ThedMAXallowsmovementofdatato/fromanyaddressablememoryspaceincludinginternalmemory,peripherals,andexternalmemory.

ThedMAXcontrollerincludesfeaturessuchasthecapabilitytoperformthree-dimensionaldatatransfersforadvanceddatasorting,andthecapabilitytomanageasectionofthememoryasacircularbuffer/FIFOwithdelay-tapbasedreadingandwritingofdata.ThedMAXcontrolleriscapableofconcurrentlyprocessingtwotransferrequests(providedthattheyareto/fromdifferentsource/destinations).

ExternalMemoryInterface(EMIF)forFlexibilityandExpansion.TheexternalmemoryinterfaceontheC672×supportsasinglebankofSDRAMandasinglebankofasynchronousmemory.TheEMIFdatawidthis16bitswideontheC6726B,C6722B,andC6720and32bitswideontheC6727B.

SDRAMsupportincludesx16andx32SDRAMdeviceswith1，2,or4banks.

TheC6726B,C6722B,andC6720supportSDRAMdevicesupto128Mbits.

Real-TimeInterruptTimer(RTI).Thereal-timeinterrupttimermoduleincludes:

Two32-bitcounter/prescalerpairs.

Twoinputcaptures(tiedtoMcBSPdirectmemoryaccess[DMA]eventsforsampleratemeasurement).

Fourcompareswithautomaticupdatecapability.

DigitalWatchdog(optional)forenhancedsystemrobustness.

ClockGeneration(PLLandOSC).TheC672xDSPincludesanon-chiposcillatorthatsupportscrystalsintherangeof12MHzto25MHz.Alternatively,theclockcanbeprovidedexternallythroughtheCLKINpin.

TheDSPincludesaflexible,software-programmablephase-lockedloop(PLL)clockgenerator.Threedifferentclockdomains(SYSCLKI，SYSCLK2,andSYSCLK3)aregeneratedbydividingdownthePLLoutput.SYSCLKIistheclockusedbytheCPU，memorycontroller,andmemories.SYSCLK2isusedbytheperipheralsubsystemanddMAX.SYSCLK3isusedexclusivelyfortheEMIF.

14.2.3ASurveyofEmbeddedOperatingSystem

AnEmbeddedOperatingSystemisShowninFig14.3.

Fig14.3EmbeddedOperatingSystem

Embeddedsystemisapplication-orientedspecialcomputersystemwhichisscalableonbothsoftwareandhardware.Itcansatisfythestrictrequirementoffunctionality,reliability,cost,volume,andpowerconsumptionoftheparticularapplication.

WithrapiddevelopmentofICdesignandmanufacture,CPUsbecamecheap.LotsofconsumerelectronicshaveembeddedCPUandthusbecameembeddedsystems.Forexample,PDAs,cellphones,point-of-saledevices,VCRs,industrialrobotcontrol,orevenyourtoasterscanbeembeddedsystem.Thereismoreandmoredemandontheembeddedsystemmarket.SomereportexpectsthatthedemandonembeddedCPUsis10timesaslargeasgeneralpurposePCCPUs.

Asapplicationsoftheembeddedsystemsbecomemorecomplex,theoperatingsystemsupportanddevelopmentenvironmentbecamecrucial.Inthispaper,wemainlyanalyzethreemajorembeddedoperatingsystems,QNX4RTOS,WindowsCEandembeddedLinux.WindowsCEandembeddedLinuxaremostwidelyusedembeddedoperationsystems.QNXisarelativelysimpleoneandcanfitinsomesimpleapplications.

1.Hardwarespecification

Therearevarioushardwareplatformsforembeddedsystem.Themostpopularonesincludex86,MIPS,PowerPC，HitachiSH,PowerPCandStrongArmprocessors.Ononehand,theembeddedoperatingsystemsareadaptedtorunonmostoftheprocessoravailabletoembeddedsystem.Ontheotherhand,theyalsohavesomeownrestrictions.

Processor

QNXcanrunonallgenericx86-basedprocessors(386andup).

Linuxiscurrentlyavailabletorunonvirtuallyeverygeneral-purposemicroprocessorandtosupportthemostcommonprocessorsusedintheembeddedworld,includingtheARM,StrongArm,MIPS,HitachiSH,Motorola/IBMPowerPC,andx86-compatiblefamilies.EHOEvendors,suchasRedHat,alsoofferportingservicesthatwillquicklymoveaLinuxEHOEtothenewprocessorarchitecture.

WindowsCErequiretheunderlyingCPUtouseaflat32-bitaddressspaceandsupportkernel-anduser-modeoperation,tosupportvirtualmemoryusinganMMUandtobelittle-endian.Fortunately,NowWindowsCEcantargetvariousplatformslikex86,MIPS,HitachiSH3andSH4,PowerPCandStrongArmprocessors.

Memory

Memoryisalwaysapreciousresourceonembeddedsystem.Theembeddedoperatingsystemsthusmakelargeefforttoreduceitsmemoryoccupationsize.QNXisthesmallestOSamongthethree.Ithasaverysmallkernelofabout12kbytes.Thekernelcaneasilyfitintotheon-chipcache.WindowsCEandLinuxrequiremorememorybecausetheyaremorecomplicated.WindowsCEneeds350KBforaminimalsystemwiththekernelandsomecommunicationsupport.Linuxneeds125-256KBforareasonableconfiguredkernel,andover100KBforothercomponents.BothWindowsCEandLinuxareconfigurable,sothesystemsizevariesfordifferentapplication.

2.Architecture

Embeddedoperatingsystemuseeithermicrokerneloramodulararchitecturetomakethemeasilytailoredtofitindifferentapplicationrequirement.

QNXcontainsaverysmallmicrokernelsurroundedbyateamofcooperatingprocessesthatprovidehigher-levelOSservices.TheQNXmicrokernelimplementfourservices:1)Interprocesscommunication.2)Low-levelnetworkcommunication.3)Processscheduling.4)Interruptdispatching.TheOSserviceprocessesareoptional,andusercanchoosewhichoneisneededfortheirapplications.ThiskindofdesignmakesQNXflexiblefordifferentkindsofapplicationswhichrequiredifferentkindsofOSservices.

TheoperatingsystemarchitectureofWindowsCEisahierarchicalone.AtthebottomliesthedevicedriversandOAL(OEMAbstractionLayer).TheyareimplementedbytheOEMwhenportingWindowsCE.AbovethemlietheGraphics,WindowingandEventsSubsystem(GWES),thekernelitselfandthecommunicationstacks.ThislayerisimplementedbyMicrosoft.TheRemoteAPIcapabilityisbuiltontopofthecommunicationfunctionality.Ontopofthekernelliesthedatabaseandfilesystem.ThisisaccessedbytheRAPIcalls,andismadeavailabletotheapplicationsviatheWin32interface.ApplicationexecuteintheirownaddressspaceandinteractwiththerestofWindowsCEviatheWin32systemcallinterface.

Linuxisalsoalayeringstructureandcomprisedofmodules,suchasLinuxkernel,filesystem,devicedriversandnetworkprotocols.EmbeddedLinuxtakestheLinuxkernelandextractthenecessarymodulesasneeded.Withinthekernellayer,Linuxiscomposedoffivemajorsubsystems:theprocessscheduler(sched),thememorymanager(mm),thevirtualfilesystem(vfs),thenetworkinterface(net),andtheinter-processcommunication(ipc).Conceptually,theclusteringofthecomponentscomposestheLinuxKernelandeachsubsystemisanindependentcomponentoftheLinuxkernel.However,therearestillquitesomeinterdependenceamongthefivesubsystemsintheconcretearchitecture.

3.Processmanagement

QNXprocessmanagerdosenotresideinQNXmicrokernel.Althoughitsharesthesameaddressspaceasthemicrokernel(andistheonlyprocesstodoso),theProcessManagerrunsasotherprocesses,scheduledbymicrokernelandusestheMicrokernal’smessagepassingprimitivestocommunicateswithotherprocessesinthesystem.Processmanagerisresponsesforthecreationofprocessesandmanagementofprocessresource.

Theschedulingofprocessesismanagedbythemicrokernelscheduler.Themicrokernelschedulermakesschedulingdecisionwhen:

(1)Aprocessbecomesunblocked.

(2)Thetimesliceforarunningprocessexpires.

(3)Arunningprocessispreempted.

Everyprocessisassignedapriority.Theschedulerselectsthenextprocesstorunbylookingattheprocesspriority.Whenmorethanoneprocessatthesamepriorityarereadytorun,theschedulerusesthreeschedulingmethods:

(1)FIFOscheduling.

(2)Round-robinscheduling.

(3)Adaptivescheduling.

TheQNXisafullypreemptiblesystemwhichsupportsreal-timeoperatingsystem.

WindowsCEsupportbothprocessesandthreads.Fullmemoryprotectionappliestoapplicationprocesses.Threadschedulingiscarriedoutpreemptively,using8differentprioritylevels.WindowsCEalsouseWaitFunctionsandWaitObjectsforsynchronization.WindowsCEisapreemptivemultitaskingoperatingsystem.Itallowsmultipleapplications,orprocessestorunwithinthesystematthesametime.Itsupportsamaximumof32simultaneousprocesses.Aprocessconsistsofoneormorethreads.

WindowsCEusesapriority-basedtime-slicealgorithmtoscheduletheexecutionofthreads.Itsupports8discreteprioritylevels.Preemptionisbasedsolelyonthethread'spriority.Threadswithahigherpriorityarescheduledtorunfirst.Threadsatthesameprioritylevelruninaround-robinfashionwitheachthreadreceivingasliceofexecutiontime.LikeotherWindowsoperatingsystem,WindowsCEoffersarichsetof“waitobjects”forthreadsynchronization.Theseincludecriticalsection,eventandmutexobject.Thesewaitobjectallowathreadtoblockitsownexecutionuntilthespecifiedobjectchanges.WindowsCEqueuesmutex,criticalsectionandeventrequestsin“FIFO-by-priority”order:adifferentFIFOqueueisdefinedforeachofthe8prioritylevels.

Real-timeapplicationuseinterruptsasawayofensuringthatexternaleventsarequicklynoticedbytheoperatingsystem.WindowsCEbalanceperformanceandeaseofimplementationbysplittinginterruptprocessingintotwosteps:aninterruptserviceroutineandaninterruptservicethread.EachhardwareinterruptrequestlineisassociatedwithoneISR.Whenaninterruptoccurs,thekernelcallstheregisteredISRforthatinterrupt.TheISR,thekernel-modeportionofinterruptprocessingiskeptasshortaspossible.ItreturnsaninterruptIDtothekernel.ThekernelsetstheassociatedeventaccordingtotheID.Theinterruptservicethreadiswaitingonthatevent.ISTsareusuallygiventhehighest2threadprioritylevelstoensureittorunasquicklyaspossible.

WithaSH3referenceplatformrunningat58.98MHzinternaland14.745externalfrequency,thetypicallatencyofISRisbetween1.3and7.5ms,andlatencyofISTisbetween93to275ms.

Linuximplementsthreadsinthekernel.Therefore,schedulinginLINUXusesthreadsasentities,notprocesses,inthekerneldatastructures.LINUXdistinguishesthreeclassesofthreadsforschedulingpurposes,whichare1)Real-timeFIFOthreadsarethehighestpriorityandnotpreemptable.2)Real-timeroundrobinthreadsarethesameasReal-timeFIFOthreadsexceptforitspreemptibilty;3)timesharingthreadshavelowerprioritythantheprevioustwo.

Eachthreadhasschedulingpriorityandaquantumassociatedwithit.LinuxschedulesthreadsviaaGOODNESSalgorithm,whichchoosestorunthethreadwithhighestgoodnessandthethread’squantumisdecrementedbyoneasitruns.DifferentfromQNXandWindowCE,LINUX,bynature,isnotafullypreemptabledesign.WhenLINUXisportedtotheembeddedsystem,itsuffersfromseveralchallengestoreal-timeapplicability:determinismingeneral,andresponseunderloadinspecific.

Real-timeoperatingsystem(RTOS)referstoanoperatingsystemthatcanperforminapredictableandrepeatablemanner,regard-lessofworkload.Therefore,afeweffortshavebeencontributedtomakeLINUXabetterRTOS.OneideaistocreateaRTOSkernelandrunningLinuxasanunscheduledthreadundertheRTOS.InthisduelstructureLinux,high-prioritythreadsthatneedtrueRTOSperformancewouldrununderthehostoperatingenvironmentatahi