Texas Instruments TMS320C67X/C67X+ DSP specs

TMS320C67x DSP Features and Options

The VelociTI architecture of the C6000 platform of devices make them the first off-the-shelf DSPs to use advanced VLIW to achieve high performance through increased instruction-level parallelism. A traditional VLIW architecture consists of multiple execution units running in parallel, performing multiple instructions during a single clock cycle. Parallelism is the key to extremely high performance, taking these DSPs well beyond the performance capabilities of traditional superscalar designs. VelociTI is a highly deterministic architecture, having few restrictions on how or when instructions are fetched, executed, or stored. It is this architectural flexibility that is key to the breakthrough efficiency levels of the TMS320C6000 Optimizing C compiler. VelociTI’s advanced features include:

-Instruction packing: reduced code size

-All instructions can operate conditionally: flexibility of code

-Variable-width instructions: flexibility of data types

-Fully pipelined branches: zero-overhead branching.

1-6

Introduction

SPRU733

Image 23

Texas Instruments TMS320C67X/C67X+ DSP manual TMS320C67x DSP Features and Options

Contents

TMS320C67x/C67x+ DSP CPU and Instruction Set Reference Guide Copyright 2005, Texas Instruments Incorporated Read This First Trademarks Contents Instruction Set Contents Vii Mvkl Move Signed Constant Into Register Pipeline Interrupts SPRU733 Figures −18 Tables 131 Tables Examples Introduction TMS320C6000 DSP Family Overview TMS320 DSP Family Overview Instrumentation Medical Military −1. Typical Applications for the TMS320 DSPsAutomotive Consumer Control General-Purpose Graphics/Imaging Industrial TMS320C67x DSP Features and Options SPRU733 TMS320C67x DSP Features and Options TMS320C67x DSP Architecture −1. TMS320C67x DSP Block Diagram Memory and Peripheral Options Central Processing Unit CPUInternal Memory SPRU733 CPU Data Paths and Control General-Purpose Register Files Introduction −1. TMS320C67x CPU Data Paths Register Files Devices −1 -Bit/64-Bit Register Pairs −2. Functional Units and Operations Performed Functional Units Memory, Load, and Store Paths Register File Cross Paths Acronym Register Name Section Data Address PathsControl Register File −3. Control Registers Acronym Register Name Address Read/ Write Register Addresses for Accessing the Control Registers−4. Register Addresses for Accessing the Control Registers Pipeline Stage Pipeline/Timing of Control Register Accesses Bit Field Value Description Addressing Mode Register AMR−5. Addressing Mode Register AMR Field Descriptions B6 Mode BK n Value Block Size −6. Block Size Calculations −4. Control Status Register CSR Control Status Register CSR Bit Field −7. Control Status Register CSR Field Descriptions DCC −8. Interrupt Clear Register ICR Field Descriptions Interrupt Clear Register ICR −9. Interrupt Enable Register IER Field Descriptions Interrupt Enable Register IER −10. Interrupt Flag Register IFR Field Descriptions Interrupt Flag Register IFR −9. Interrupt Return Pointer Register IRP Interrupt Return Pointer Register IRP −11. Interrupt Set Register ISR Field Descriptions Interrupt Set Register ISR −11.Interrupt Service Table Pointer Register Istp Interrupt Service Table Pointer Register Istp Nonmaskable Interrupt NMI Return Pointer Register NRP 12 E1 Phase Program Counter PCE1 −13. Control Register File Extensions Floating-Point Adder Configuration Register FadcrControl Register File Extensions −14. Floating-Point Adder Configuration Register Fadcr NAN2 Inexact results status for .L1 −15. Floating-Point Auxiliary Configuration Register Faucr Floating-Point Auxiliary Configuration Register Faucr UND NaN select for .S2 src2 Signed infinity for .S1 −16. Floating-Point Multiplier Configuration Register Fmcr Floating-Point Multiplier Configuration Register Fmcr Inexact results status for .M2 Rounding mode select for .M1 Denormalized number select for .M1 src1 Topic Instruction Set −1. Instruction Operation and Execution Notations Symbol MeaningInstruction Operation and Execution Notations Rotl Occurs Extu l,r Greater than −2. Instruction Syntax and Opcode Notations Instruction Syntax and Opcode Notations Ucstn Bit unsigned constant field Ucst n SPRU733 Sdfpn −3. Ieee Floating-Point Notations −4. Special Single-Precision Values Symbol Sign s Exponent e Fraction f −2. Double-Precision Floating-Point Fields Symbol Hex Value Decimal Value −6. Special Double-Precision Values Delay Slots Instruction Type Slots Unit Latency Read Cycles† −8. Delay Slot and Functional Unit Latency −3. Basic Format of a Fetch Packet Parallel Operations Instructions Example 3−1. Fully Serial p-Bit Pattern in a Fetch PacketExample 3−2. Fully Parallel p-Bit Pattern in a Fetch Packet Cycle/Execute Cycle/Execute Packet Instructions Example Parallel CodeBranching Into the Middle of an Execute Packet Conditional Bit Register Conditional Operations−9. Registers That Can Be Tested by Conditional Operations Specified Resource Constraints Constraints on Instructions Using the Same Functional Unit Constraints on Cross Paths 1X Following execute packets are valid Constraints on Loads and Stores Following code sequence is invalid Constraints on Long 40-Bit Data Constraints on Register Reads However, this code sequence is valid Constraints on Register Writes MPYSP2DP Constraints on Floating-Point Instructions Intdp Tional unit on cycle i + 4, i + 5, or i + Addsp Subsp Spint Sptrunc Intsp Mpysp Addressing Modes Linear Addressing Mode Cycle after LDW Cycles after LDW Circular Addressing ModeExample 3−4. LDW Instruction in Circular Mode Before LDW Cycle after Addah Example 3−5. Addah Instruction in Circular ModeSyntax for Load/Store Address Generation Before Addah Addressing Type Address Register Mode Field Syntax Modification Performed−10. Indirect Address Generation for Load/Store −11. Address Generator Options for Load/Store Instruction Descriptions Instruction Compatibility Example Way each instruction is described Opcode map field used For operand type Unit Opfield Execution for .L1, .L2 and .S1, .S2 Opcodes Pipeline StageE1 Read src2 Written dst Unit in use Absolute Value With SaturationABS Before instruction Cycle after instruction ABSDP, AbsspBefore instruction Cycle after instruction Absdp Absolute Value, Double-Precision Floating-Point Before instruction Cycles after instruction ABS, AbsspPipeline Stage Read Written Abssp Absolute Value, Single-Precision Floating-Point ABS, Absdp ADD Add Two Signed Integers Without Saturation ADD Unit in use Or .D ADDDP, ADDK, ADDSP, ADDU, ADD2, SADD, SUB Add Two Signed Integers Without Saturation ADD ADD, ADDAD, ADDAH, Addaw Add Using Byte Addressing ModeAddab Add Using Byte Addressing Mode Addab Addad Add Using Doubleword Addressing Mode ADD, ADDAB, ADDAH, Addaw ADD, ADDAB, ADDAD, Addaw Add Using Halfword Addressing ModeAddah Add Using Halfword Addressing Mode Addah ADD, ADDAB, ADDAD, Addah Add Using Word Addressing ModeAddaw Add Using Word Addressing Mode Addaw Adddp Add Two Double-Precision Floating-Point Values Add Two Double-Precision Floating-Point Values Adddp Unit in use Or .S ADD, ADDSP, ADDU, Subdp Pipeline StageE1 Read cst16 Written dst Unit in use Add Signed 16-Bit Constant to RegisterAddk Addsp Add Two Single-Precision Floating-Point Values Add Two Single-Precision Floating-Point Values Addsp ADD, ADDDP, ADDU, Subsp ADD, SADD, Subu Add Two Unsigned Integers Without SaturationAddu Addu Add Two Unsigned Integers Without Saturation ADD2 Add Two 16-Bit Integers on Upper and Lower Register Halves ADD, ADDU, SUB2 Bitwise OR, XOR S1, .S2 Branch Using a DisplacementOpcode map field used For operand type Unit Cycle Program Counter Value Action Delay Slots Example Xuint Branch Using a Register Written Branch Taken Unit in use Target Instruction Pipeline Stage Read Xsint Branch Using an Interrupt Return PointerIRP −15. Program Counter Values for B IRP Instruction NRP Branch Using NMI Return Pointer −16. Program Counter Values for B NRP Instruction CLR Clear a Bit Field SET Execution Pipeline Clear a Bit Field CLR Cmpeq Compare for Equality, Signed Integers CMPEQDP, CMPEQSP, CMPGT, Cmplt Cmpeqdp Compare for Equality, Double-Precision Floating-Point Values Cmpeqdp .S1 Cmpeqsp Compare for Equality, Single-Precision Floating-Point Values CMPEQ, CMPEQDP, CMPGTSP, Cmpltsp Cmpgt Compare for Greater Than, Signed Integers CMPEQ, CMPGTDP, CMPGTSP, CMPGTU, Cmplt Cmpgt .L1X Cmpgtdp Delay Slots Functional Unit Latency See Also Example Cmpgtsp CMPEQSP, CMPGT, CMPGTDP, CMPGTU, Cmpltsp Cmpgtu Compare for Greater Than, Unsigned Integers CMPGT, CMPGTDP, CMPGTSP, Cmpltu Cmplt Compare for Less Than, Signed Integers Cmplt Compare for Less Than, Signed Integers Compare for Less Than, Signed Integers Cmplt Wise, 0 is written to dst CMPEQDP, CMPGTDP, CMPLT, CMPLTSP, Cmpltu Cmpltsp Cmpltsp .S1 A1,A2,A3 Cmpltu Compare for Less Than, Unsigned Integers Instruction Type Single-cycle Delay Slots See Also Example Dpint Convert Double-Precision Floating-Point Value to Integer Example Dpint Floating-Point Value Dpsp DPINT, DPTRUNC, INTSP, Spdp With Truncation Dptrunc DPINT, DPSP, Sptrunc Dptrunc EXT Extract and Sign-Extend a Bit Field If cond src2 ext csta, cstb → dst else nop Extu Extu Extract and Zero-Extend a Bit Field If cond src2 extu csta, cstb → dst else nop EXT Idle Multicycle NOP With No Termination Until InterruptIdle DPINT, INTDPU, INTSP, Intspu Intdp Intdp INTDP, INTSP, Intspu Intdpu Intdpu INTDP, INTDPU, Intspu Intsp INTDP, INTDPU, Intsp Intspu Left Shift LdbuRegister Offset −17. Data Types Supported by Ldbu Instruction LDH, LDW Cycles after LDB Before LDBCycle after LDB −18. Data Types Supported by Ldbu Instruction 15-Bit Offset Load Byte From Memory With a 15-Bit Unsigned Constant Offset Before LDB Cycle after LDB Pipeline Stage Read B14 / B15 Written Or Register Offset Lddw Execution Pipeline Instruction Type LDB, LDH, LDW −19. Data Types Supported by Ldhu Instruction Ldhu LDB, LDW Cycles after LDH Before LDHCycle after LDH Tion operates only on the .D2 unit −20. Data Types Supported by Ldhu Instruction 15-Bit Offset LDW LDB, LDDW, LDH Cycle after LDW LDW LDB, LDH Leftmost Bit Detection Lmbd → dst Pipeline StageE1 E2 Read Multiply Signed 16 LSB y Signed 16 LSBMPY MPYU, MPYSU, MPYUS, Smpy MPY Multiply Signed 16 LSB x Signed 16 LSB Mpydp Multiply Two Double-Precision Floating-Point Values Pipeline E10 Stage Read MPY, Mpysp MPYHU, MPYHSU, MPYHUS, Smpyh Multiply Signed 16 MSB y Signed 16 MSB Mpyh Multiply Signed 16 MSB x Signed 16 MSB MPYHLU, MPYHSLU, MPYHULS, Smpyhl Multiply Signed 16 MSB y Signed 16 LSBMpyhl Mpyhl Multiply Signed 16 MSB x Signed 16 LSB MPYHL, MPYHSLU, Mpyhuls Multiply Unsigned 16 MSB y Unsigned 16 LSB MPYHL, MPYHLU, Mpyhuls MpyhsluMultiply Signed 16 MSB y Unsigned 16 LSB MPYH, MPYHU, Mpyhus Multiply Signed 16 MSB y Unsigned 16 MSB MPYH, MPYHSU, Mpyhus MpyhuMultiply Unsigned 16 MSB y Unsigned 16 MSB MPYHL, MPYHLU, Mpyhslu Multiply Unsigned 16 MSB y Signed 16 LSBMpyhuls MPYH, MPYHU, Mpyhsu Multiply Unsigned 16 MSB y Signed 16 MSB Mpyi Multiply 32-Bit y 32-Bit Into 32-Bit ResultMpyi Mpyid Mpyid Multiply 32-Bit y 32-Bit Into 64-Bit Result Mpyid Multiply 32-Bit x 32-Bit Into 64-Bit Result MPYLHU, MPYLSHU, MPYLUHS, Smpylh Multiply Signed 16 LSB y Signed 16 MSBMpylh Mpylh Multiply Signed 16 LSB x Signed 16 MSB MPYLH, MPYLSHU, Mpyluhs Multiply Unsigned 16 LSB y Unsigned 16 MSB MPYLH, MPYLHU, Mpyluhs MpylshuMultiply Signed 16 LSB y Unsigned 16 MSB MPYLH, MPYLHU, Mpylshu Multiply Unsigned 16 LSB y Signed 16 MSBMpyluhs Mpysp Multiply Two Single-Precision Floating-Point Values MPY, MPYDP, MPYSP2DP Mpysp Mpyspdp MPY, MPYDP, MPYSP, MPYSP2DP Mpyspdp Double-Precision Result Multiply Two Single-Precision Floating-Point Values forMPYSP2DP MPYSP2DP Multiply Signed 16 LSB y Unsigned 16 LSB Mpysu MPY, MPYU, Mpyus MPY, MPYSU, Mpyus Multiply Unsigned 16 LSB y Unsigned 16 LSB Multiply Unsigned 16 LSB x Unsigned 16 LSB Mpyu MPY, MPYU, Mpysu MpyusMultiply Unsigned 16 LSB y Signed 16 LSB Multiply Unsigned 16 LSB x Signed 16 LSB Mpyus Move From Register to Register If cond 0 + src2 → dst MVC Move Between Control File and Register File Src2 → dst −21. Register Addresses for Accessing the Control Registers Pipeline StageE1 Read Written dst Unit in use Move Signed Constant Into Register and Sign ExtendMVK MVKH, MVKL, Mvklh MVKH/MVKLH Move 16-Bit Constant Into Upper Bits of Register If you are loading the address of a label, use Mvkl Pipeline Stage Read Written MVK, MVKH, Mvklh NEG Negate Ucst4 None No OperationNOP Cycle after ADD Before NOPCycle after NOP No operation Executes Norm Normalize Integer Execution If cond Norm src → dst Else nop Pipeline Not Bitwise not Bitwise or AND, XOR Src1 or src2 → dst Rcpdp Double-Precision Floating-Point Reciprocal Approximation RCPSP, Rsqrdp Rcpsp Single-Precision Floating-Point Reciprocal Approximation RCPDP, Rsqrsp Rsqrdp If src2 is positive infinity, positive 0 is placed in dst Rsqrsp Rsqrsp .S1 Sadd Add Two Signed Integers With Saturation ADD, Ssub Add Two Signed Integers With Saturation Sadd SAT Saturate a 40-Bit Integer to a 32-Bit Integer SAT .L2 SET Set a Bit Field CLR If cond src2 SET csta, cstb → dst else nop SET .S1 SHL Arithmetic Shift Left SHR, Sshl SHR Arithmetic Shift Right SHL, Shru Shru Logical Shift Right SHL, SHR MPY, SMPYH, SMPYHL, Smpylh Smpy CSR MPYH, SMPY, SMPYHL, Smpylh Smpyh MPYHL, SMPY, SMPYH, Smpylh Smpyhl Instruction Set 223 MPYLH, SMPY, SMPYH, Smpyhl Smpylh Instruction Set 225 Spdp DPSP, INTDP, SPINT, Sptrunc Spint Convert Single-Precision Floating-Point Value to Integer DPINT, INTSP, SPDP, Sptrunc Spint Sptrunc DPTRUNC, SPDP, Spint Sptrunc Sshl Shift Left With Saturation Sshl .S1 Ssub Subtract Two Signed Integers With Saturation SUB STB Before Cycle after Instruction STH, STW Store Byte to Memory With a 15-Bit Unsigned Constant Offset Pipeline Stage Read B14 /B15 , src Written Unit in use STH Before STB, STW Instruction Cycles after STH Instruction Type Store Delay Slots See Also STW STB, STH Store Word to Memory With a 15-Bit Unsigned Constant Offset 248 SUB Subtract Two Signed Integers Without Saturation Src1 − src2 → dst else nop Src2 − src1 → dst ADD, SSUB, SUBC, SUBDP, SUBSP, SUBU, SUB2 SUB, SUBAH, Subaw Subtract Using Byte Addressing ModeSubab BK0 = 3 → size = A5 in circular addressing mode using BK0 SUB, SUBAB, Subaw Subtract Using Halfword Addressing ModeSubah SUB, SUBAB, Subah Subtract Using Word Addressing ModeSubaw Subtract Using Word Addressing Mode Subaw ADD, SSUB, SUB, SUBDP, SUBSP, SUBU, SUB2 Subtract Conditionally and Shift-Used for DivisionSubc Subtract Conditionally and Shift−Used for Division Subc Subdp Subtract Two Double-Precision Floating-Point Values Subtract Two Double-Precision Floating-Point Values Subdp ADDDP, SUB, SUBSP, Subu Subsp Subtract Two Single-Precision Floating-Point Values Subsp Subtract Two Single-Precision Floating-Point Values ADDSP, SUB, SUBDP, Subu ADDU, SSUB, SUB, SUBC, SUBDP, SUBSP, SUB2 Subtract Two Unsigned Integers Without SaturationSubu Subtract Two Unsigned Integers Without Saturation Subu Src1 and placed in the lower-half of dst SUB2 ADD2, SSUB, SUB, SUBC, Subu XOR Bitwise Exclusive or AND, or Zero Zero a Register Pipeline Fetch Pipeline Operation Overview Decode −2. Fetch Phases of the Pipeline −3. Decode Phases of the Pipeline Execute −4. Execute Phases of the Pipeline Clock cycle Fetch Pipeline Operation Summary −1. Operations Occurring During Pipeline Phases Stage Phase Symbol During This Phase Completed Adddp −7. Pipeline Phases Block Diagram Example 4−1. Execute Packet in −7 Pipeline Execution of Instruction Types Delay slots Functional Unit latency DP Compare Instruction Type Execution Phases Cycle DP ADDDP/SUBDP Mpyi Mpyid Mpydp Instruction Type Execution Phases Instruction Type Execution Phases Unit in use M, or .D Single-Cycle Instructions−3. Single-Cycle Instruction Execution −4 y 16-Bit Multiply Instruction Execution 2 16 y 16-Bit Multiply Instructions −5. Store Instruction Execution Store Instructions −13. Store Instruction Execution Block Diagram −6. Load Instruction Execution Load Instructions −15. Load Instruction Execution Block Diagram −7. Branch Instruction Execution Branch Instructions −17. Branch Instruction Execution Block Diagram −8. Two-Cycle DP Instruction Execution Two-Cycle DP Instructions Unit in use Or .M Four-Cycle Instructions−9. Four-Cycle Instruction Execution −10. Intdp Instruction Execution Intdp Instruction −11. DP Compare Instruction Execution DP Compare Instructions −12. ADDDP/SUBDP Instruction Execution ADDDP/SUBDP Instructions −13. Mpyi Instruction Execution Mpyi Instruction Pipeline Stage E4 E5 E6 E7 E8 E9 E10 Read Mpyid Instruction−14. Mpyid Instruction Execution −15. Mpydp Instruction Execution Mpydp Instruction −16. Mpyspdp Instruction Execution Mpyspdp Instruction −17. MPYSP2DP Instruction Execution Functional Unit ConstraintsMPYSP2DP Instruction Instruction Execution Unit Constraints−18. Single-Cycle .S-Unit Instruction Constraints −19. DP Compare .S-Unit Instruction Constraints −20 -Cycle DP .S-Unit Instruction Constraints −21. ADDSP/SUBSP .S-Unit Instruction Constraints −22. ADDDP/SUBDP .S-Unit Instruction Constraints −23. Branch .S-Unit Instruction Constraints −24 y 16 Multiply .M-Unit Instruction Constraints −25 -Cycle .M-Unit Instruction Constraints −26. Mpyi .M-Unit Instruction Constraints −27. Mpyid .M-Unit Instruction Constraints −28. Mpydp .M-Unit Instruction Constraints −29. Mpysp .M-Unit Instruction Constraints −30. Mpyspdp .M-Unit Instruction Constraints −31. MPYSP2DP .M-Unit Instruction Constraints −32. Single-Cycle .L-Unit Instruction Constraints −33 -Cycle .L-Unit Instruction Constraints −34. Intdp .L-Unit Instruction Constraints −35. ADDDP/SUBDP .L-Unit Instruction Constraints −36. Load .D-Unit Instruction Constraints Unit Instruction Constraints −37. Store .D-Unit Instruction Constraints −38. Single-Cycle .D-Unit Instruction Constraints Lddw Performance Considerations Pipeline Stall −29. Multicycle NOP in an Execute Packet Multicycle NOPs Cycle # Pipeline Phase Branch Target Operation −40. Program Memory Accesses Versus Data Load AccessesMemory Considerations Data Load +10 −32. Program and Data Memory Stalls Example 4−2. Load From Memory Banks −33 -Bank Interleaved Memory −34 -Bank Interleaved Memory With Two Memory Spaces −41. Loads in Pipeline from Example 4−2 Interrupts Overview Types of Interrupts and Signals Used Priority Interrupt Name Interrupt Type −1. Interrupt Priorities Nonmaskable Interrupt NMI Interrupt Acknowledgment Iack and Interrupt Number INUMn −1. Interrupt Service Table Interrupt Service Table IST −2. Interrupt Service Fetch Packet 1238h Interrupt Service Table Pointer Istp Example 5−1. Relocation of Interrupt Service Table Acronym Register Name Description Summary of Interrupt Control Registers−2. Interrupt Control Registers Globally Enabling and Disabling Interrupts MVC CSR,B0 Enabling and Disabling Interrupts Individual Interrupt Control Status of Interrupts Setting and Clearing Interrupts Example 5−9. Code to Return from a Maskable Interrupt Returning From Interrupt ServicingExample 5−8. Code to Return From NMI Interrupt Detection and Processing Setting the Nonreset Interrupt FlagConditions for Processing a Nonreset Interrupt Isfp Actions Taken During Nonreset Interrupt Processing PS PW PR DP DC Setting the Reset Interrupt Flag Actions Taken During Reset Interrupt Processing Pipeline Interaction General Performance Single Assignment Programming Programming Considerations Nested Interrupts Example 5−11. Code Using Single Assignment STW Example 5−14. Manual Interrupt Processing Manual Interrupt Processing Example 5−16. Code Sequence for Trap Return TrapsExample 5−15. Code Sequence to Invoke a Trap Instruction C62x DSP C64x DSP C67x DSP C67x+ DSP Instruction Compatibility Cmpltu Intsp Mpylh Rsqrdp Subab Functional Unit Instruction Table B−1. Functional Unit to Instruction Mapping Displacement Register Intdp Intdpu Intsp Intspu N n n n n n n n n n n n n STB memory SUB Subab Subah Subaw Subc Subdp Subsp Subu SUB2 XOR Zero Unit Instructions and Opcode Maps Table C−1. Instructions Executing in the .D Functional Unit Instructions Executing in the .D Functional Unit Table C−2. .D Unit Opcode Map Symbol Definitions Opcode Map Symbols and Meanings Syntax Modification Performed Table C−3. Address Generator Options for Load/Store Figure C−1 or 2 Sources Instruction Format 32-Bit Opcode Maps Appendix D Table D−1. Instructions Executing in the .L Functional Unit Instructions Executing in the .L Functional Unit Table D−2. .L Unit Opcode Map Symbol Definitions Figure D−1 or 2 Sources Instruction Format Appendix E Table E−1. Instructions Executing in the .M Functional Unit Instructions Executing in the .M Functional Unit Table E−2. .M Unit Opcode Map Symbol Definitions Figure E−1. Extended M-Unit with Compound Operations Appendix F Table F−1. Instructions Executing in the .S Functional Unit Instructions Executing in the .S Functional Unit Table F−2. .S Unit Opcode Map Symbol Definitions Figure F−1 or 2 Sources Instruction Format Figure F−7. Branch with NOP Constant Instruction Format No Unit Specified Instructions and Opcode Maps Table G−1. Instructions Executing With No Unit Specified Instructions Executing With No Unit Specified Figure G−1. Loop Buffer Instruction Format Index Cmpeqdp Dpsp Index-4 Reset Index-6 Mpysu Unit No unit instructions Rsqrdp To memory with a 15-bit unsigned constant offset STB 260 Single-precision Subsp 263