GeneralPurpose Registers.docx

GeneralPurpose Registers.docx

《GeneralPurpose Registers.docx》由会员分享，可在线阅读，更多相关《GeneralPurpose Registers.docx（67页珍藏版）》请在冰豆网上搜索。

GeneralPurposeRegisters

General-PurposeRegisters

3.4.1.General-PurposeRegisters

The32-bitgeneral-purposeregistersEAX,EBX,ECX,EDX,ESI,EDI,EBP,andESPareprovidedforholdingthefollowingitems:

•Operandsforlogicalandarithmeticoperations

•Operandsforaddresscalculations

•Memorypointers

Althoughalloftheseregistersareavailableforgeneralstorageofoperands,results,and

pointers,cautionshouldbeusedwhenreferencingtheESPregister.TheESPregisterholdsthe

stackpointerandasageneralruleshouldnotbeusedforanyotherpurpose.

Manyinstructionsassignspecificregisterstoholdoperands.Forexample,stringinstructions

usethecontentsoftheECX,ESI,andEDIregistersasoperands.Whenusingasegmented

memorymodel,someinstructionsassumethatpointersincertainregistersarerelativeto

specificsegments.Forinstance,someinstructionsassumethatapointerintheEBXregister

pointstoamemorylocationintheDSsegment.

Thespecialusesofgeneral-purposeregistersbyinstructionsaredescribedinSection5,

“InstructionSetSummary”inthisvolume;seealsoChapter3,InstructionSetReferenceA-M

andChapter4,InstructionSetReferenceN-ZoftheIA-32IntelArchitectureSoftwareDeveloper’s

Manual,Volumes2A&2B.Thefollowingisasummaryofspecialuses:

•EAX—Accumulatorforoperandsandresultsdata

•EBX—PointertodataintheDSsegment

•ECX—Counterforstringandloopoperations

•EDX—I/Opointer

•ESI—PointertodatainthesegmentpointedtobytheDSregister;sourcepointerforstringoperations

•EDI—Pointertodata（ordestination）inthesegmentpointedtobytheESregister;destinationpointerforstringoperations

•ESP—Stackpointer（intheSSsegment）

•EBP—Pointertodataonthestack（intheSSsegment）

3.4.2.SegmentRegisters

Thesegmentregisters（CS,DS,SS,ES,FS,andGS）hold16-bitsegmentselectors.Asegment

selectorisaspecialpointerthatidentifiesasegmentinmemory.Toaccessaparticularsegment

inmemory,thesegmentselectorforthatsegmentmustbepresentintheappropriatesegment

register.

Whenwritingapplicationcode,programmersgenerallycreatesegmentselectorswithassembler

directivesandsymbols.Theassemblerandothertoolsthencreatetheactualsegmentselector

valuesassociatedwiththesedirectivesandsymbols.Ifwritingsystemcode,programmersmay

needtocreatesegmentselectorsdirectly.Adetaileddescriptionofthesegment-selectordata

structureisgiveninChapter3,Protected-ModeMemoryManagement,oftheIA-32IntelArchitecture

SoftwareDeveloper’sManual,Volume3.

Howsegmentregistersareuseddependsonthetypeofmemorymanagementmodelthatthe

operatingsystemorexecutiveisusing.Whenusingtheflat（unsegmented）memorymodel,the

segmentregistersareloadedwithsegmentselectorsthatpointtooverlappingsegments,eachof

whichbeginsataddress0ofthelinearaddressspace（asshowninFigure3-5）.Theseoverlapping

segmentsthencomprisethelinearaddressspacefortheprogram.Typically,twooverlapping

segmentsaredefined:

oneforcodeandanotherfordataandstacks.TheCSsegment

registerpointstothecodesegmentandalltheothersegmentregisterspointtothedataandstack

segment.

Whenusingthesegmentedmemorymodel,eachsegmentregisterisordinarilyloadedwitha

differentsegmentselectorsothateachsegmentregisterpointstoadifferentsegmentwithinthe

linearaddressspace（asshowninFigure3-6）.Atanytime,aprogramcanthusaccessuptosix

segmentsinthelinearaddressspace.Toaccessasegmentnotpointedtobyoneofthesegment

registers,aprogrammustfirstloadthesegmentselectorforthesegmenttobeaccessedintoa

segmentregister.

Figure3-4.AlternateGeneral-PurposeRegisterNames

311615870

AHAL

BHBL

CHCL

DHDL

BP

SI

DI

SP

16-bit

AX

DX

CX

BX

32-bit

EAX

EBX

ECX

EDX

EBP

ESI

ESP

General-PurposeRegisters

EDI

Eachofthesegmentregistersisassociatedwithoneofthreetypesofstorage:

code,data,or

stack.Forexample,theCSregistercontainsthesegmentselectorforthecodesegment,where

theinstructionsbeingexecutedarestored.Theprocessorfetchesinstructionsfromthecode

segment,usingalogicaladdressthatconsistsofthesegmentselectorintheCSregisterandthe

contentsoftheEIPregister.TheEIPregistercontainstheoffsetwithinthecodesegmentofthe

nextinstructiontobeexecuted.TheCSregistercannotbeloadedexplicitlybyanapplication

program.Instead,itisloadedimplicitlybyinstructionsorinternalprocessoroperationsthat

changeprogramcontrol（suchas,procedurecalls,interrupthandling,ortaskswitching）.

TheDS,ES,FS,andGSregisterspointtofourdatasegments.Theavailabilityoffourdata

segmentspermitsefficientandsecureaccesstodifferenttypesofdatastructures.Forexample,

fourseparatedatasegmentsmightbecreated:

oneforthedatastructuresofthecurrentmodule,

anotherforthedataexportedfromahigher-levelmodule,athirdforadynamicallycreateddata

structure,andafourthfordatasharedwithanotherprogram.Toaccessadditionaldatasegments,

theapplicationprogrammustloadsegmentselectorsforthesesegmentsintotheDS,ES,FS,and

GSregisters,asneeded.

TheSSregistercontainsthesegmentselectorforthestacksegment,wheretheprocedurestack

isstoredfortheprogram,task,orhandlercurrentlybeingexecuted.Allstackoperationsusethe

SSregistertofindthestacksegment.UnliketheCSregister,theSSregistercanbeloaded

explicitly,whichpermitsapplicationprogramstosetupmultiplestacksandswitchamongthem.

SeeSection3.3.,“MemoryOrganization”,foranoverviewofhowthesegmentregistersare

usedinreal-addressmode.

ThefoursegmentregistersCS,DS,SS,andESarethesameasthesegmentregistersfoundin

theIntel8086andIntel286processorsandtheFSandGSregisterswereintroducedintothe

IA-32ArchitecturewiththeIntel386™familyofprocessors.

3.4.3.EFLAGSRegister

The32-bitEFLAGSregistercontainsagroupofstatusflags,acontrolflag,andagroupofsystemflags.Figure3-7definestheflagswithinthisregister.Followinginitializationoftheprocessor（eitherbyassertingtheRESETpinortheINITpin）,thestateoftheEFLAGSregisteris00000002H.Bits1,3,5,15,and22through31ofthisregisterarereserved.Softwareshouldnotuseordependonthestatesofanyofthesebits.SomeoftheflagsintheEFLAGSregistercanbemodifieddirectly,usingspecial-purposeinstructions（describedinthefollowingsections）.Therearenoinstructionsthatallowthewholeregistertobeexaminedormodifieddirectly.ThefollowinginstructionscanbeusedtomovegroupsofflagstoandfromtheprocedurestackortheEAXregister:

LAHF,SAHF,PUSHF,PUSHFD,POPF,andPOPFD.AfterthecontentsoftheEFLAGSregisterhavebeentransferredtotheprocedurestackorEAXregister,theflagscanbeexaminedandmodifiedusingtheprocessor’sbitmanipulationinstructions（BT,BTS,BTR,andBTC）.Whensuspendingatask（usingtheprocessor’smultitaskingfacilities）,theprocessorautomatically

savesthestateoftheEFLAGSregisterinthetaskstatesegment（TSS）forthetaskbeingsuspended.Whenbindingitselftoanewtask,theprocessorloadstheEFLAGSregisterwithdatafromthenewtask’sTSS.Whenacallismadetoaninterruptorexceptionhandlerprocedure,theprocessorautomaticallysavesthestateoftheEFLAGSregistersontheprocedurestack.Whenaninterruptorexceptionishandledwithataskswitch,thestateoftheEFLAGSregisterissavedintheTSSforthetaskbeingsuspended.

AstheIA-32Architecturehasevolved,flagshavebeenaddedtotheEFLAGSregister,butthefunctionandplacementofexistingflagshaveremainedthesamefromonefamilyoftheIA-32processorstothenext.Asaresult,codethataccessesormodifiestheseflagsforonefamilyofIA-32processorsworksasexpectedwhenrunonlaterfamiliesofprocessors.

3.4.3.1.STATUSFLAGS

Thestatusflags（bits0,2,4,6,7,and11）oftheEFLAGSregisterindicatetheresultsofarithmetic

instructions,suchastheADD,SUB,MUL,andDIVinstructions.Thestatusflagfunctions

are:

CF（bit0）Carryflag.Setifanarithmeticoperationgeneratesacarryora

borrowoutofthemost-significantbitoftheresult;clearedotherwise.

Thisflagindicatesanoverflowconditionforunsigned-integerarithmetic.

Itisalsousedinmultiple-precisionarithmetic.

PF（bit2）Parityflag.Setiftheleast-significantbyteoftheresultcontainsan

evennumberof1bits;clearedotherwise.

AF（bit4）Adjustflag.Setifanarithmeticoperationgeneratesacarryora

borrowoutofbit3oftheresult;clearedotherwise.Thisflagisused

inbinary-codeddecimal（BCD）arithmetic.

ZF（bit6）Zeroflag.Setiftheresultiszero;clearedotherwise.

SF（bit7）Signflag.Setequaltothemost-significantbitoftheresult,whichis

thesignbitofasignedinteger.（0indicatesapositivevalueand1

indicatesanegativevalue.）

OF（bit11）Overflowflag.Setiftheintegerresultistoolargeapositivenumber

ortoosmallanegativenumber（excludingthesign-bit）tofitinthe

destinationoperand;clearedotherwise.Thisflagindicatesanoverflow

conditionforsigned-integer（two’scomplement）arithmetic.

Ofthesestatusflags,onlytheCFflagcanbemodifieddirectly,usingtheSTC,CLC,andCMC

instructions.Alsothebitinstructions（BT,BTS,BTR,andBTC）copyaspecifiedbitintotheCF

flag.

Thestatusflagsallowasinglearithmeticoperationtoproduceresultsforthreedifferentdata

types:

unsignedintegers,signedintegers,andBCDinteg