欢迎来到天天文库
浏览记录
ID:52266918
大小:308.00 KB
页数:32页
时间:2020-04-03
《DSP28335教程附录5浮点运算单元.ppt》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、Floating-PointUnitCopyright©2009TexasInstruments.Allrightsreserved.TechnicalTrainingOrganizationTTOAppendixETMS320C28x™MCUWorkshopLearningObjectivesArchitecture–floating-pointformat,registers,andpipelineInstructions–instructiontypes,delayslots,parallelinstructions,RPTB
2、,floating-pointflagsInstructionSummaryIEEESingle-PrecisionFloating-PointFormat*NormalPositiveandNegativeValuesareCalculatedas:(-1)sx2(E-127)x1.M+/-~1.7x10-38to+/-~3.4x10+38TheNormalizedIEEEnumbershaveahidden1;thustheequivalentsignedintegerresolutionisthenumberofmantis
3、sabits+sign+123-bitMantissa(ImplicitLeadingBit+FractionBits)8-bitExponent(Biased)1SignBit(0=Positive,1=Negative)SEMSEMValue0100PositiveorNegativeZero010Non-ZeroDenormalizedNumber011-2540-0x7FFFFPositiveorNegativeValues*01255(max)0PositiveorNegativeInfinity01255(max)Non
4、-ZeroNotaNumber(NaN)IEEESingle-PrecisionFloating-PointFormat(IEEE754)Mostwidelyusedstandardforfloating-pointStandardnumberformats,Specialvalues(NaN,Infinity)Roundingmodes&floating-pointoperationsUsedonmanyCPUsSimplificationsfortheC28xfloating-pointunitFlags&CompareOper
5、ations:NegativezeroistreatedaspositivezeroDenormalizedvaluesaretreatedaszeroNot-a-number(NaN)istreatedasinfinityRound-to-ZeroModeSupported(truncate)Round-to-NearestModeSupported(even)TheseformatsarecommonlyhandledthiswayonembeddedprocessorsC28xCPUandFPURegistersACCPXTX
6、AR0XAR1XAR2XAR3XAR4XAR5XAR6XAR7PCRPCDPSPST0ST1IERIFRDBGIERC28RegisterSetR0HR1HR2HR3HR5HR6HR7HR4HSTFFPURegistersRB32-bit8FPUResultRegistersFPUStatusRepeatBlock32-bitAccumulatorProductTemporary8Auxiliary22-bitProgramCounterReturnPC16-bitDataPagePointerStackPointer2S
7、tatusInterruptEnableInterruptFlagR0H–R7HandSTFareshadowedforfastcontextsaveandrestoreExtendedforFPUStandardC28xCPUFloating-PointUnitPipelineF1F2D1D2R1R2EWC28xPipelineFetchDecodeReadExeWriteDRE1E2/WFPUInstructionLoadStoreNorequireddelayslotinstruction1requireddelaysloti
8、nstructionCMP/MIN/MAX/NEG/ABSMPY/ADD/SUB/MACF32Floating-pointoperationsarenotpipelineprotectedSomeinstructionsrequire
此文档下载收益归作者所有