3_3 C6000指令系统(3)-SOP(前一部分)

第3章汇编语言程序设计主讲：谢秀娟Email：[e-mail],BristolUniversity,(c)TexasInstruments20022第一节C6000体系结构第二节C6000指令集第三节实例:乘积和（SOP）LearningObjectives第三节点积（SOP）ImplementationofSumofProducts(SOP)Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20024Twobasicoperationsarerequiredforthisalgorithm.（1）Multiplication（2）AdditionThereforetwobasicinstructionsarerequiredSOPisthekeyelementformostDSPalgorithms.Solet’swritethecodeforthisalgorithmandatthesametimelearntheC6000architecture.NnnnxaY1NNxaxaxa2211Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20025（1）Multiply----MPYThemultiplicationofa1byx1isdoneinassemblybythefollowinginstruction:MPYa1,x1,prodThisinstructionisperformedbyamultiplierunitthatiscalled“.M”NnnnxaY1NNxaxaxa2211Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20026（1）Multiply----.Munit.MThe.MunitperformsmultiplicationsinhardwareMPY.Ma1,x1,prodNote:16-bitby16-bitmultiplierprovidesa32-bitresult.32-bitby32-bitmultiplierprovidesa64-bitresult.401nnnxaYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20027（2）Addition----.?unit.M.?MPY.Ma1,x1,prodADD.?Y,prod,Y401nnnxaYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20028（2）Add----.LunitRISCprocessorssuchastheC6000useregisterstoholdtheoperands,soletschangethiscode..M.?MPY.Ma1,x1,prodADD.LY,prod,Y401nnnxaYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments20029（3）RegisterFile-A.M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsYLetuscorrectthisbyreplacinga,x,prodandYbytheregistersasshownabove.MPY.Ma1,x1,prodADD.LY,prod,Y401nnnxaYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200210（3）SpecifyingRegisterNames.M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsYTheregistersA0,A1,A3andA4containthevaluestobeusedbytheinstructions.MPY.MA0,A1,A3ADD.LA4,A3,A4401nnnxaYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200211（4）DataloadingQ:Howdoweloadtheoperandsintotheregisters?RegisterFileAcontains16registers(A0-A15)whichare32-bitswide..M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsYDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200212（4）LoadUnit“.D”A:Theoperandsareloadedintotheregistersbyloadingthemfromthememoryusingthe.Dunit.Q:Howdoweloadtheoperandsintotheregisters?DataMemory.M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsY.DDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200213（4）LoadUnit“.D”Itisworthnotingatthisstagethattheonlywaytoaccessmemoryisthroughthe.Dunit.DataMemory.M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsY.DDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200214（5）LoadInstructionQ:Whichinstruction(s)canbeusedforloadingoperandsfromthememorytotheregisters?DataMemory.M.LA0A1A2A3A4A15RegisterFileA...a1x1prod32-bitsY.DA:Theloadinstructions:LDB,LDH,LDW,LDDWDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200215（6）UsingtheLoadInstructionsBeforeusingtheloadunityouhavetobeawarethatthisprocessorisbyteaddressable,whichmeansthateachbyteisrepresentedbyauniqueaddress.0000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFFAlsotheaddressesare32-bitwide.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200216Thesyntaxfortheloadinstructionis:（6）UsingtheLoadInstructionsLD*Rn,Rm0000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFFWhere:RnisaregisterthatcontainstheaddressoftheoperandtobeloadedandRmisthedestinationregister.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200217Thequestionnowishowmanybytesaregoingtobeloadedintothedestinationregister?（6）UsingtheLoadInstructions0000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFFThesyntaxfortheloadinstructionis:LD*Rn,RmDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200218（6）UsingtheLoadInstructionsTheanswer,isthatitdependsontheinstructionyouchoose:•LDB:loadsonebyte(8-bit)•LDH:loadshalfword(16-bit)•LDW:loadsaword(32-bit)•LDDW:loadsadoubleword(64-bit)Thesyntaxfortheloadinstructionis:LD*Rn,RmNote:LDonitsowndoesnotexist.0000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFFDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200219（6）UsingtheLoadInstructionsExample:IfweassumethatA5=0x4then:(1)LDB*A5,A7;givesA7=0x00000001(2)LDH*A5,A7;givesA7=0x00000201(3)LDW*A5,A7;givesA7=0x04030201(4)LDDW*A5,A7:A6;givesA7:A6=0x08070605040302010000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFF0xB0xA0xD0xC0x10x20x30x40x50x60x70x801Thesyntaxfortheloadinstructionis:LD*Rn,RmDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200220（6）UsingtheLoadInstructionsQuestion:Ifdatacanonlybeaccessedbytheloadinstructionandthe.Dunit,howcanweloadtheregisterpointerRninthefirstplace?Thesyntaxfortheloadinstructionis:LD*Rn,Rm0000000000000002000000040000000600000008Data16-bitsaddressFFFFFFFF0xB0xA0xD0xC0x10x20x30x40x50x60x70x801Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments200221TheinstructionMVKLwillallowamoveofa16-bitconstantintoaregisterasshownbelow:MVKL.?a,A5(‘a’isaconstantorlabel)Howmanybitsrepresentafulladdress?32bitsSowhydoestheinstructionnotallowa32-bitmove?Allinstructionsare32-bitwide(seeinst