ARM Introduction

Editor's Notes

• #2: v10: 28/04/03, Chris Shore slide 4: added China to text (already on graphic). Updated employee count and geographical distribution. slide 19: added V6 slides 32: Imported from RV Overview to replace original ADS slide. slide 33: New general debug architecture diagram slide 34: new product montage (crib in notes) slide 35: New question about embedded trace. v09: 19/11/02, Chris Shore slides 6-8: New slides showing IP deployment (imported from 926 core module) v08: 08/02, Rob Levy - Style update, black & white view amended v07: 12/01, CJS Main changes: - ARM Development Boards slide removed (now in Debug Solutions module) - Register set slides re-ordered so that the animated graphic comes first - slide 12: Q bit in v5TEJ as well as v5TE - slide 14: CPSR changes rephrased slightly - slide 16: reference to v5T removed. - slide 27: EASY/Micropack replaced with ADK/ACT - slide 30: Trace slide updated
• #3: Introduction to ARM Background to who ARM Ltd are, what we do, and how our business model works. Programmers Model The structure of the ARM architecture How it has developed Register set, modes and exceptions The endian issue Instruction Sets Overview of the features of the ARM instruction set The coprocessor mechanism Overview of Thumb - Why it was designed and the benefits it gives. System Design Overview of some of the hardware and software technologies that ARM has to support the design in of the ARM core into real products. Also looks at some of the issues involved with memory maps in ARM based systems.
• #4: The ARM processor core originates within a British computer company called Acorn. In the mid-1980s they were looking for replacement for the 6502 processor used in their BBC computer range, which were widely used in UK schools. None of the 16-bit architectures becoming available at that time met their requirements, so they designed their own 32-bit processor. Other companies became interested in this processor, including Apple who were looking for a processor for their PDA project (which became the Newton). After much discussion this led to Acorn’s processor design team splitting off from Acorn at the end of 1990 to become Advanced RISC Machines Ltd, now just ARM Ltd. Thus ARM Ltd now designs the ARM family of RISC processor cores, together with a range of other supporting technologies. One important point about ARM is that it does not fabricate silicon itself, but instead just produces the design - we are an Intellectual Property (or IP) company. Instead silicon is produced by companies who license the ARM processor design.
• #5: ARM’s business model centres around the principle of partnership. At the centre of this are ARM’s semiconductor partners who design, manufacture and market ARM-compliant products. Having so many partner companies producing silicon executing the same instruction set is a very important part of ARM’s strength in the market place. However each of our semiconductor partners bring their own unique strengths to the partnership - each having their own technologies, applications knowledge, product focus, culture, geography, and key customers. In addition to our partnering with semiconductor companies, we also partner with a large number of other third parties to ensure that operating systems, EDA and software development tools, application software and design services are available for doing ARM based designs. “ATAP” stands for ARM Technology Access Program. Creates a network of independent design service companies and equips them to deliver ARM-powered designs. Members get access to ARM technology, expertise and support. Members sometimes referred to as “Approved Design Centers”.
• #7: Programmers Model The structure of the ARM architecture How it has developed Register set, modes and exceptions The endian issue
• #8: The cause of confusion here is the term “word” which will mean 16-bits to people with a 16-bit background. In the ARM world 16-bits is a “halfword” as the architecture is a 32-bit one, whereas “word” means 32-bits. Java bytecodes are 8-bit instructions designed to be architecture independent. Jazelle transparently executes most bytecodes in hardware and some in highly optimized ARM code. This is due to a tradeoff between hardware complexity (power consumption & silicon area) and speed.
• #9: The Programmers Model can be split into two elements - first of all, the processor modes and secondly, the processor registers. So let’s start by looking at the modes. Now the typical application will run in an unprivileged mode know as “User” mode, whereas the various exception types will be dealt with in one of the privileged modes : Fast Interrupt, Supervisor, Abort, Normal Interrupt and Undefined (and we will look at what causes each of the exceptions later on). NB - spell out the word FIQ, otherwise you are saying something rude in German! One question here is what is the difference between the privileged and unprivileged modes? Well in reality very little really - the ARM core has an output signal (nTRANS on ARM7TDMI, InTRANS, DnTRANS on 9, or encoded as part of HPROT or BPROT in AMBA) which indicates whether the current mode is privileged or unprivileged, and this can be used, for instance, by a memory controller to only allow IO access in a privileged mode. In addition some operations are only permitted in a privileged mode, such as directly changing the mode and enabling of interrupts. All current ARM cores implement system mode (added in architecture v4). This is simply a privileged version of user mode. Important for re-entrant exceptions because no exceptions can cause system mode to be entered.
• #10: This animated slide shows the way that the banking of registers works. On the left the currently visible set of registers are shown for a particular mode. On the right are the registers that are banked out whilst in that mode. Each key press will switch mode: user -> FIQ ->user -> IRQ -> user ->SVC -> User -> Undef -> User -> Abort and then back to user. The following slide then shows this in a more static way that is more useful for reference
• #11: The ARM architecture provides a total of 37 registers, all of which are 32-bits long. However these are arranged into several banks, with the accessible bank being governed by the current processor mode. We will see this in more detail in a couple of slides. In summary though, in each mode, the core can access: a particular set of 13 general purpose registers (r0 - r12). a particular r13 - which is typically used as a stack pointer. This will be a different r13 for each mode, so allowing each exception type to have its own stack. a particular r14 - which is used as a link (or return address) register. Again this will be a different r14 for each mode. r15 - whose only use is as the Program counter. The CPSR (Current Program Status Register) - this stores additional information about the state of the processor: And finally in privileged modes, a particular SPSR (Saved Program Status Register). This stores a copy of the previous CPSR value when an exception occurs. This combined with the link register allows exceptions to return without corrupting processor state.
• #12: Green psr bits are only in certain versions of the ARM architecture ALU status flags (set if "S" bit set, implied in Thumb state). Sticky overflow flag (Q flag) is set either when saturation occurs during QADD, QDADD, QSUB or QDSUB, or the result of SMLAxy or SMLAWx overflows 32-bits Once flag has been set can not be modified by one of the above instructions and must write to CPSR using MSR instruction to cleared PSRs split into four 8-bit fields that can be individually written: Control(c)bits 0-7 Extension(x)bits 8-15Reserved for future use Status(s)bits 16-23Reserved for future use Flags(f)bits 24-31 Bits that are reserved for future use should not be modified by current software. Typically, a read-modify-write strategy should be used to update the value of a status register to ensure future compatibility. Note that the T/J bits in the CPSR should never be changed directly by writing to the PSR (use the BX/BXJ instruction to change state instead). However, in cases where the processor state is known in advance (e.g. on reset, following an interrupt, or some other exception), an immediate value may be written directly into the status registers, to change only specific bits (e.g. to change mode). New ARM V6 bits now shown.
• #13: ARM is designed to efficiently access memory using a single memory access cycle. So word accesses must be on a word address boundary, halfword accesses must be on a halfword address boundary. This includes instruction fetches. Point out that strictly, the bottom bits of the PC simply do not exist within the ARM core - hence they are ‘undefined’. Memory system must ignore these for instruction fetches. In Jazelle state, the processor doesn’t perform 8-bit fetches from memory. Instead it does aligned 32-bit fetches (4-byte prefetching) which is more efficient. Note we don’t mention the PC in Jazelle state because the ‘Jazelle PC’ is actually stored in r14 - this is technical detail that is not relevant as it is completely hidden by the Jazelle support code.
• #14: Exception handling on the ARM is controlled through the use of an area of memory called the vector table. This lives (normally) at the bottom of the memory map from 0x0 to 0x1c. Within this table one word is allocated to each of the various exception types. This word will contain some form of ARM instruction that should perform a branch. It does not contain an address. Reset - executed on power on Undef - when an invalid instruction reaches the execute stage of the pipeline SWI - when a software interrupt instruction is executed Prefetch - when an instruction is fetched from memory that is invalid for some reason, if it reaches the execute stage then this exception is taken Data - if a load/store instruction tries to access an invalid memory location, then this exception is taken IRQ - normal interrupt FIQ - fast interrupt When one of these exceptions is taken, the ARM goes through a low-overhead sequence of actions in order to invoke the appropriate exception handler. The current instruction is always allowed to complete (except in case of Reset). IRQ is disabled on entry to all exceptions; FIQ is also disabled on entry to Reset and FIQ.
• #15: Instruction Sets Overview of the features of the ARM instruction set The coprocessor mechanism Overview of Thumb - Why it was designed and the benefits it gives.
• #16: Unusual but powerful feature of the ARM instruction set. Other architectures normally only have conditional branches. Some recently-added ARM instructions (in v5T and v5TE) are not conditional (e.g. v5T BLX offset) Core compares condition field in instruction against NZCV flags to determine if instruction should be executed.
• #17: Sequence of conditional instructions: - no instruction must reset cond code flags - BL corrupts flags so must be last - limit sequence to max 3 or so instrs Can use different condition codes. Give if then else example. Note GCD practical coming later. Conditional compare - resets condition code when executed - compiler will make use of this - can be difficult for a human to understand! Not just for compare, using data processing with condition code and S bit is useful in some circumstances. LDM/LDR instruction cannot set flags due to datapath issues (data comes back only at the very end of the cycle, so there is no opportunity to perform a comparison and set the status flags).
• #18: PC-relative to allow position independent code, and allows restricted branch range to jump to nearby addresses. How to access full 32-bit address space? Can set up LR manually if needed, then load into PC MOV lr, pc LDR pc, =dest ADS linker will automatically generate long branch veneers for branches beyond 32Mb range.
• #19: BICbit clear ORRbit set ANDbit mask EORbit invert Comparisons produce no results - just set condition codes. CMPlike SUB CMNlike ADD (subtract of a negative number is the same as add) TSTlike AND TEQlike EOR (eor of identical numbers gives result of zero) Generally single-cycle execution (except write to PC and register-controlled shift). Mention ARM NOP & Thumb NOP. Explain RSB and RSC which do subtract in other order (e.g. y-x not x-y) Does not include multiply (separate instr format). No divide - compiler uses run-time library or barrel shifter to perform division. Can combine “S” bit with conditional execution, e.g. ADDEQS r0, r1, r2
• #20: Rotate left can be implemented as rotate right (32-number), e.g. rotate left of 10 is performed using rotate right of 22. RRX shifts by 1 bit position, of a 33 bit amount (includes carry flag). Very specialized application (e.g. encryption algorithms). Cannot be generated by C compiler. We have used it for 64/64 bit divide. RRX allows you to shift multiprecision values right by one efficiently. Also used in ARM’s MPEG code in a very tricky piece of code. ANSI C does not have a rotate operation (it only has “<<“ and “>>” which are the equivalent of LSL, LSR and ASR). However the ARM compiler recognizes rotate type expresssions and optimizes these to use ROR, e.g. int f(unsigned int a) { return (a << 10) | (a >>22) ; } => MOV a1,a1,ROR #22 Carry flag set out of the shifter for *logical* data processing operations
• #21: Mention A bus and B bus on 7TDMI core. Give examples: ADDr0, r1, r2 ADDr0, r1, r2, LSL#7 ADDr0, r1, r2, LSL r3 ADDr0, r1, #0x4E
• #22: Could have used 12 bits directly for immediate value - this would allow 0-4095. But this does not allow any large numbers, which are useful for: base address of memory devices in target system large, but simple hex constants (0x10000) Research has shown there is a need for a large range of small numbers (frequently needed) but also some large numbers. 50% of all constants lie between the range -15 and +15 and 90% lie in the range -511 and +511. Will vary depending on the application. ROR #n is confusing… but can be considered as ROL #32-n Opcode 0xe3a004ff = MOV r0, #0xff, 8 Core rotates 0xff right by 4 pairs of bits => MOV r0, #0xff000000
• #23: Literal pools These are constant data areas embedded in the code at the end of assembler modules, and at other locations if specified by the user using LTORG. Data value must not be executed (will probably be an undefined instruction), assembly programmer must ensure this by placing LTORG at an appropriate location. ARM C compilers will handle placement of literal pools automatically.
• #24: Variable number of cycles for some processors which implement ‘early termination’. The multiply is faster for smaller values in Rs. ARM7TDMI and ARM9TDMI use 8-bit Booth’s algorithm which takes 1 cycle for each byte in Rs. Terminates when rest of Rs is all zeros or all ones. MUL/MLA don’t need signed/unsigned specified - because they return the low 32-bit of the result which is the same whatever the sign of the arguments. Cycle information is general and specific cores have some specific variations from this, specifically with respect to result delays where accumulation is involved. Refer to TRM for exact details if required. XScale and StrongARM have a split pipeline with multiple execution units - so can issue multiplies in 1 or 2 cycles and continue with following instructions, assuming no resource or result dependencies. XScale can issue MUL/MLA/MULL in one cycle (MLAL requires 2 cycles), providing multiplier is not already in use. Cycle timing is dependent on result latency - the core will stall if an instruction tries to use the result before multiplier has completed. Note that there is no form of the multiply instruction which has an immediate constant operand - registers only. For the interested student - C flag is unpredictable if S is set in architectures prior to V5. MULS/MLAS always take 4 cycles; MULLS, MLALS always take 5.
• #25: Point out destination (reg) first for LDR, but destination (mem) last for STR. Different to Motorola, but it keeps the instruction mnemonic format consistent. Always have register loaded/stored first, then address accessed second Size specifier comes out on MAS (memory access size) signal. Important that memory supports full range of accesses - especially important for writes where only the specified size should be written. Special types of sign extended load - this is needed because ARM registers only hold 32-bit values. Draw diagram. No need for special store instructions though. Instruction cycle timing: STRLDR 7TDMI2 cycles3 cycles 9TDMI1 cycle1 cycle - interlock if used in next cycle StrongARM11 cycle1 cycle - interlock if used in next cycle Xscale1 cycle1 cycle - interlock if used in next 2 cycles Note size specifier comes after condition code. Link: <address> explained on next slide. Note that load/store instructions never set condition codes.
• #26: Halfword access and signed halfword/byte accesses were added to the architecture in v4T, this is the reason the offset field is not as flexible as the normal word/byte load/store - not a problem because these accesses are less common. Link: diagram on next slide
• #27: “!” indicates “writeback” i.e. the base register is to be updated after the instruction. No “!” for post-indexed because post-increment of base register always happens (otherwise the offset field would not be used at all). Give C example: int *ptr; x = *ptr++; Compiles to a single instruction: LDR r0, [r1], #4
• #28: In effect, a SWI is a user-defined instruction. Used for calling the operating system (switches to privileged mode). SWI number field can be used to specify the operation code, e.g. SWI 1 start a new task, SWI 2 allocate memory, etc. Using a number has the advantage that the O.S. can have different revisions, and the same application code will work on each O.S. rev.
• #29: The status registers are split into four 8-bit fields that can be individually written: bits 31 to 24 : the flags field (NZCV flags and 4 unused bits) bits 23 to 16 : the status field (unused in Arch 3, 4 & 4T) bits 15 to 8 : the extension field (unused in Arch 3, 4 & 4T) bits 7 to 0 : the control field (I & F interrupt disable bits, 5 processor mode bits, and the T bit on ARMv4T.) Immediate form of MSR can actually be used with any of the field masks, but care must be taken that a read-modify-write strategy is followed so that currently unallocated bits are not affected. Otherwise the code could have distinctly different effect on future cores where such bits are allocated. When used with the flag bits, the immediate form is shielded from this as bits 27-24 can be considered to be read only. For MSR operations, we recommend that only the minimum number of fields are written, because future ARM implementations may need to take extra cycles to write specific fields; not writing fields you don't want to change reduces any such extra cycles to a minimum. For example, an MRS/BIC/ORR/MSR sequence whose purpose is to change processor mode (only) is best written with the last instruction being MSR CPSR_c,Rm, though any other set of fields that includes "c" will also work.
• #30: This slide shows the way that ARM branch instructions work It also shows the need to stack the LR (using STM/LDM instructions) when making subroutine calls within subroutines.
• #31: The Thumb instruction set was designed by looking at the instructions produced by the ARM C compiler from real application code to see which instructions were most often used. This subset of instructions was then compressed into 16-bit opcodes to give better code density and better performance from narrow memory A Thumb compatible processor is still a 32-bit processor, but it has the ability to execute either sections of ARM code or sections of Thumb code. The two instruction sets cannot be interleaved though, a special form of branch has to be used to change “state”. The diagram then shows the way that a typical 32-bit ARM instruction might be “compressed” into a 16-bit Thumb one.
• #32: System Design Overview of some of the hardware and software technologies that ARM has to support the design in of the ARM core into real products. Also looks at some of the issues involved with memory maps in ARM based systems.
• #33: AMBA is ARM’s on-chip bus specification. The aims of AMBA are to: Make life easier for Systems designers Standardise the bus interface Reduce the support required from ARM and between internal design teams Allows increased re-use of IP in designs Enable the creation of upgrades and families of devices Why use AMBA not the original ARM Bus Improved Tools support Modularity Upgrading to other ARM cores ADK is ARM’s AMBA design kit. A generic, stand-alone development environment enabling rapid creation of AMBD-based components and designs. ACT is a complete environment for testing compliance to the AMBA spec. The PrimeCell peripherals are a set of AMBA-compliant peripherals that are available for licensing from ARM. They include a UART, real time clock, keyboard & mouse interface, GPIO and a generic IR interface. Consult the ARM website for more details - a selection of reference manuals are also on the technical documentation CD.
• #34: System Design Overview of some of the hardware and software technologies that ARM has to support the design in of the ARM core into real products. Also looks at some of the issues involved with memory maps in ARM based systems.
• #35: Debugger trace tools Have copy of the code image Configure ETM trace via JTAG Receive compressed trace from ETM Decompress ETM trace using code image