1. Chap. 3 ARM CPU Architecture

2. 2 Outline 3.1 Registers 3.2 Memory 3.3 Exceptions

3. 3 3.1 Registers  Introduction  ARM Processor Core  Processor Modes  Register Organization  Accessing Registers  The Program Status Registers (CPSR and SPSRs)  Condition Flags  Conditional Execution

4. 4 Introduction  ARM has 37 registers in total, all of which are 32-bits long.  1 dedicated program counter  1 dedicated current program status register  5 dedicated saved program status registers  30 general purpose registers

5. 5 Introduction (cont’d)  However these are arranged into several banks, with the accessible bank being governed by the processor mode. Each mode can access  a particular set of r0-r12 registers  a particular r13 (the stack pointer) and r14 (link register)  r15 (the program counter)  cpsr (the current program status register) and privileged modes can also access  a particular spsr (saved program status register)

6. 6 ARM Processor Core  Architecture  Versions 1 and 2 – Acorn RISC, 26-bit address  Version 3 – 32-bit address, CPSR, and SPSR  Version 4 – half-word, Thumb  Version 5 – BLX, CLZ and BRK instructions  Processor cores  ARM7TDMI (Thumb, debug, multiplier, ICE) – version 4T, low-end ARM core, 3-stage pipeline, 50-100MHz  ARM9TDMI – 5-stage pipeline, 130MHz or 200MHz  ARM10TDMI – version 5, 300MHz  CPU Core: co-processor, MMU, AMBA  ARM 710, 720, 740  ARM 920, 940

7. 7 Processor Modes  The ARM has six operating modes:  User  User (unprivileged mode under which most tasks run)  FIQ  FIQ (entered when a high priority (fast) interrupt is raised)  IRQ  IRQ (entered when a low priority (normal) interrupt is raised)  Supervisor  Supervisor (entered on reset and when a Software Interrupt instruction is executed)  Abort  Abort (used to handle memory access violations)  Undef  Undef (used to handle undefined instructions)  ARM Architecture Version 4 adds a seventh mode:  System  System (privileged mode using the same registers as user mode)

8. 8 Register Organization General registers and Program Counter Program Status Registers r15 (pc) r14 (lr) r13 (sp) r14_svc r13_svc r14_irq r13_irq r14_abt r13_abt r14_undef r13_undef User32 / System FIQ32Supervisor32Abort32IRQ32Undefined32 cpsr sprsr_fiq spsr_abtspsr_svcsprsr_fiq spsr_fiqsprsr_fiq spsr_irq r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r7 r4 r5 r2 r1 r0 r3 r6 r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r15 (pc) cpsr r14_fiq r13_fiq r12_fiq r10_fiq r11_fiq r9_fiq r8_fiq sprsr_fiq spsr_undef Ref. [8]

9. 9 Register Example: User to FIQ Mode spsr_fiq cpsr r7 r4 r5 r2 r1 r0 r3 r6 r15 (pc) r14_fiq r13_fiq r12_fiq r10_fiq r11_fiq r9_fiq r8_fiq r14 (lr) r13 (sp) r12 r10 r11 r9 r8 User mode CPSR copied to FIQ mode SPSR cpsr r15 (pc) r14 (lr) r13 (sp) r12 r10 r11 r9 r8 r7 r4 r5 r2 r1 r0 r3 r6 r14_fiq r13_fiq r12_fiq r10_fiq r11_fiq r9_fiq r8_fiq Return address calculated from User mode PC value and stored in FIQ mode LR Registers in use EXCEPTION User ModeFIQ Mode spsr_fiq Ref. [8]

10. 10 Accessing Registers  No breakdown of currently accessible registers.  All instructions can access r0-r14 directly.  Most instructions also allow use of the PC.  Specific instructions to allow access to CPSR and SPSR.  Note : When in a privileged mode, it is also possible to load / store the (banked out) user mode registers to or from memory.

11. 11 The Program Status Registers (CPSR and SPSRs)  Condition Code Flags N = Negative result from ALU flag. Z = Zero result from ALU flag. C = ALU operation Carried out V = ALU operation oVerflowed  Interrupt Disable bits. I = 1, disables the IRQ. F = 1, disables the FIQ.  T Bit: Processor in ARM (0) or Thumb (1)  Mode Bits: processor mode Copies of the ALU status flags (latched if the instruction has the "S" bit set). N Mode ZCV 28 318 40 I F T

12. 12 Condition Flags FlagLogical Instruction Arithmetic Instruction NegativeNo meaningBit 31 of the result has been set (N=‘1’)Indicates a negative number in signed operations ZeroResult is all zeroesResult of operation was zero (Z=‘1’) CarryAfter Shift operationResult was greater than 32 bits (C=‘1’)‘1’ was left in carry flag oVerflowNo meaningResult was greater than 31 bits (V=‘1’)Indicates a possible corruption of the sign bit in signed numbers

13. 13 The Condition Field of Instruction Set 28 31 242016 128 40 Cond Not equal Unsigned higher or same Unsigned lower Minus Equal Overflow No overflow Unsigned higher Unsigned lower or same Positive or Zero Less than Greater than Less than or equal Always Greater or equal EQ NE CS/HS CC/LO PL VS HI LS GE LT GT LE AL MI VC SuffixDescription Z=0 C=1 C=0 Z=1 Flags N=1 N=0 V=1 V=0 C=1 & Z=0 C=0 or Z=1 N=V N!=V Z=0 & N=V Z=1 or N=!V none 0000 0001 0010 0011 0101 0110 1000 1001 1010 1011 1100 1101 1110 0100 0111 Code

14. 14 Conditional Execution  Most instruction sets only allow branches to be executed conditionally.  However by reusing the condition evaluation hardware, ARM effectively increases number of instructions.  All instructions contain a condition field which determines whether the CPU will execute them.  Non-executed instructions soak up 1 cycle. Still have to complete cycle so as to allow fetching and decoding of following instructions.

15. 15 Conditional Execution (cont’d)  This removes the need for many branches, which stall the pipeline (3 cycles to refill).  Allows very dense in-line code, without branches.  The time penalty of not executing several conditional instructions is frequently less than overhead of the branch or subroutine call that would otherwise be needed.

16. 16 Using and updating the Condition Field  To execute an instruction conditionally, simply postfix it with the appropriate condition:  For example an add instruction takes the form: ADD r0,r1,r2; r0 = r1 + r2 (ADDAL)  To execute this only if the zero flag is set: ADDEQ r0,r1,r2; If zero flag set then… ;... r0 = r1 + r2

17. 17 Using and updating the Condition Field (cont’d)  By default, data processing operations do not affect the condition flags (apart from the comparisons where this is the only effect).  To cause the condition flags to be updated, the S bit of the instruction needs to be set by postfixing the instruction (and any condition code) with an “ S ”.  For example to add two numbers and set the condition flags: ADDS r0,r1,r2; r0 = r1 + r2 ;... and set flags

18. 18 Conditional Execution Example  Greatest Common Divisor ( 最大公因數 )  Normal Assembler gcd cmp r0, r1 ;reached the end? beq stop blt less ;if r0 > r1 sub r0, r0, r1 ;subtract r1 from r0 bal gcd less sub r1, r1, r0 ;subtract r0 from r1 bal gcd stop  ARM Conditional Assembler gcd cmp r0, r1 ;if r0 > r1 subgt r0, r0, r1 ;subtract r1 from r0 sublt r1, r1, r0 ;else subtract r0 from r1 bne gcd ;reached the end?

19. 19 The Barrel Shifter  ARM has a barrel shifter which provides a mechanism to carry out shifts as part of other instructions.  Immediate value  8 bit number  can be rotated right through an even number of positions.  assembler will calculate rotate for you from constant.  Register, optionally with shift operation applied.  Shift value can be either be:  5 bit unsigned integer  specified in bottom byte of another register. Operand 1 Result ALU Barrel Shifter Operand 2

20. 20 Second Operand: Shifted Register  Using a multiplication instruction to multiply by a constant  load the constant into a register  wait for a number of internal cycles for the multiplication to complete  A more optimum solution can be often found by using some combination of MOVs, ADDs, SUBs and RSBs with shifts.  Multiplications by a constant equal to a ((power of 2) ± 1) can be done in one cycle.  Example: r0 = r1 * 5  r0= r1 + (r1 * 4) ADD r0, r1, r1, LSL #2  Example: r2 = r3 * 105  r2 = r3 * 15 * 7 r2 = r3 * (16 - 1) * (8 - 1) RSB r2, r3, r3, LSL #4; r2 = r3 * 15 RSB r2, r2, r2, LSL #3; r2 = r2 * 7

21. 21 Outline 3.1 Registers 3.2 Memory 3.3 Exceptions

22. 22 3.2 Memory  Introduction  Memory Organization  Pipeline  Memory Access  ARM Memory Interface

23. 23 Introduction  Word, half-word alignment (xxxx00 or xxxxx0)  ARM can be set up to access data in either little- endian or big-endian format, through they default to little-endian.  The ARM uses a pipeline in order to increase the speed of the flow of instructions to the processor.  Allows several operations to be undertaken simultaneously, rather than serially.  Rather than pointing to the instruction being executed, the PC points to the instruction being fetched.

24. 24 Memory Organization

25. 25 Pipeline  3 stages (ARM7) and 5 stages (ARM9TDMI) fetch memory executedecode fetchwrite-backexecutedecode PC PC-4 PC-8 access memory write result Load an instruction if needed to register from memory

26. 26 Memory Access  The ARM7/9 is a Von Neumann, load/store architecture, i.e.,  Only 32 bit data bus for both instr. and data.  Only the load/store instr. (and SWP) access memory.  Memory is addressed as a 32 bit address space.  Data type can be  8 bit bytes,  16 bit half-words,  or 32 bit words,  and may be seen as a byte line folded into 4-byte words.  Words must be aligned to 4 byte boundaries, and half-words to 2 byte boundaries.  Always ensure that memory controller supports all three access sizes.

27. 27 Outline 3.1 Registers 3.2 Memory 3.3 Exceptions

28. 28 3.3 Exceptions  Introduction  Types of ARM exceptions  Exception and Interrupt  Entering an Exception  Returning from an Exception  Exception Entry/Exit  Exceptions and the Vector Table Address

29. 29 Introduction  Exceptions and interrupts break the sequential flow of a program, jumping to architecturally defined memory locations.  In ARM, Software Interrupt (SWI) is the “ system call ” exception.

30. 30 Types of ARM exceptions  reset  when CPU reset pin is asserted  undefined instruction  when CPU tries to execute an undefined op-code  software interrupt  when CPU executes the SWI instruction  prefetch abort  when CPU tries to execute an instruction pre-fetched from an illegal addr  data abort  when data transfer instruction tries to read or write at an illegal address  IRQ  when CPU's external interrupt request pin is asserted  FIQ  when CPU's external fast interrupt request pin is asserted

31. 31 Exception and Interrupt  The terms exception and interrupt are often confused.  Exception usually refers to an internal CPU event such as  floating point overflow  MMU fault (e.g., page fault)  trap (SWI)  Interrupt usually refers to an external I/O event such as  I/O device request  Timer interrupt  In the ARM architecture manuals, the two terms are mixed together.

32. 32 Entering an Exception  When an exception is generated, the processor takes the following actions:  Copy the CPSR to SPSR for the mode in which the exception is to be handled.  Change the appropriate CPSR mode bits in order to Change to the appropriate mode, and map in the appropriate banked registers for that mode. Disable interrupts. IRQs are disabled when any exception occurs. FIQs are disabled when a FIQ occurs, and on reset. Set lr_mode to the return address. Set PC to the vector address for the exception.

33. 33 Returning from an Exception  The actions taken by the processor  Restore the CPSR from SPSR_mode  Restore the PC using return address stored in lr_mode  The way to return depends on whether a stack is used during entering the subroutine  Without a stack Performing a data processing instruction with S flag set and the PC as the destination register.  With a stack Restoring the saved registers by performing LDMFDsp!, {r0-r12, pc}^ ^ indicates that the CPSR is restored from the SPSR

34. 34 Returning from SWI and Undefined Instruction Handlers  SWI and UI exceptions are generated by the instruction itself, so the PC is not updated when the exception is taken. Thus, storing (PC-4) in lr_mode makes lr_mode points to the next instruction be executed. SWI xxx  being executed  PC-8 INST-1  being decoded  PC-4 INST-2  being fetched  PC  Restoring the PC from lr  Without a stack MOVSpc, lr  With a stack STMFDsp!, {reglist,lr} … LDMFDsp!, {reglist, pc}^

35. 35 Returning from FIQ and IRQ  Check IRQ and FIQ at the end of executing each instruction. INST-1  being executed  PC-12, IRQ or FIQ checked INST-2  being decoded  PC-8 INST-3  being fetched  PC-4 INST-4  PC  Restoring PC from lr  Without a stack SUBSpc, lr, #4  With a stack SUBlr, lr, #4 STMFDsp!, {reglist, lr} … LDMFDsp!, {reglist, pc}^

36. 36 Returning from Prefetch Abort  Prefetch abort is generated when it reaches the execution stage. INST-1  being executed  PC-8, Aborted INST-2  being decoded  PC-4 INST-3  being fetched  PC  Restoring PC from lr  Without a stack SUBSpc, lr, #4  With a stack SUBlr, lr, #4 STMFDsp!, {reglist, lr} … LDMFDsp!, {reglist, pc}^

37. 37 Returning from Data Abort  When a data abort occurs, the program counter has been updated. INST-1  being executed  PC-12, Aborted INST-2  being decoded  PC-8 INST-3  being fetched  PC-4 INST-4  PC  Restoring PC from lr  Without a stack SUBSpc, lr, #8  With a stack SUBlr, lr, #8 STMFDsp!, {reglist, lr} … LDMFDsp!, {reglist, pc}^

38. 38 Exception Entry/Exit Ref. [4]

39. 39 Exceptions and the Vector Table Address Exception Vectors Priority Address Exception type Exception Mode (1=high,6=low) 0x00000000 Reset Supervisor1 0x00000004 Undefined instruction Undefined6 0x00000008 Software Interrupt Supervisor6 0x0000000C Abort (prefetch) Abort5 0x00000010 Abort (data) Abort2 0x00000014 Reserved ReservedNot applicable 0x00000018 IRQ IRQ4 0x0000001C FIQ FIQ3

40. 40 References [1] Andrew Sloss, Dominic Symes and Chris Wright, “ARM System Developer's Guide“, published by MORGAN KAUFFMAN, 2004 [2] David Seal, “ARM Architecture Reference Manual “, published by Addison- Wesley, 2000 [3] ARM DUI 0021A “Programming Techniques“, 1995 [4] http://www.samsung.com/Products/Semiconductor/SystemLSI/Networks /PersonalNTASSP/CommunicationProcessor/S3C4510B/um_s3c4510b_rev1.pdf [5] www.arm.com [6] http://www.arm.com/pdfs/DUI0056D_ADS1_2_Dev.pdf [7] http://nthucad.cs.nthu.edu.tw/~wcyao/ [8] www-courses.cs.uiuc.edu/ ~cs433/Processors/ARM/ARMInstV1.0.ppt

41. 41 Exercise 1. How many registers does ARM have? And what purpose do they use for? 2. Describe the processor modes of ARM in detail. 3. What is the meaning of ARM condition flags in logical instructions and in Arithmetic Instructions? 4. What is the benefit of Conditional Execution? 5. What is the difference of little-endian and big-endian format? 6. Please describe ARM memory cycle types and how to decide which type. 7. List all type of ARM exceptions and describe their addresses and priorities.