Embedded Systems/ARM Microprocessors

ARM Architecture The ARM architecture is a widely used 32-bit RISC processor architecture. In fact, the ARM family accounts for about 75% of all 32-bit CPUs, and about 90% of all embedded 32-bit CPUs. ARM Limited licenses several popular microprocessor cores to many vendors (ARM does not sell physical microprocessors). Originally ARM stood for Advanced RISC Machines. Some cores offered by ARM: ARM7TDMI ARM9 ARM11 Some examples of ARM based processors: Intel X-Scale (PXA-255 and PXA-270), used in Palm PDAs Philips LPC2000 family (ARM7TDMI-S core), LPC3000 family (ARM9 core) Atmel AT91SAM7 (ARM7TDMI core) ST Microelectronics STR710 (ARM7TDMI core) Freescale MCIMX27 series (ARM9 core) The lowest-cost ARM processors (in the LPC2000 series) have dropped below US$ 5 in ones, which is less than the cost of many 16-bit and 8-bit microprocessors. Thumb calling convention In ARM Thumb code, the 16 registers r0 - r15 typically have the same roles they have in all ARM code: r0 - r3, called a1 - a4: argument/scratch/result registers. r4 - r9, called v1 - v6: variables r10, called sl: stack limit r11, called fp: frame pointer (usually not used in Thumb code) r12, called ip r13, called sp: stack pointer r14, called lr: link register r15, called pc: the program counter The standard C calling convention for ARM Thumb is:ARM. ARM Software Development Toolkit. 1997. Chapter 9: ARM Procedure Call Standard. Chapter 10: Thumb Procedure Call Standard. Subroutine-preserved registers When the return address is placed in pc (r15), returning from the subroutine, the sp, fp, sl, and v1-v6 registers must contain the same values they did when the subroutine was called. The stack Every execution environment has a limit to how low in memory the stack can grow -- the "minimum sp". In order to give interrupts (which may occur at any time) room to work, at every instant the memory between sp and the "minimum sp" must contain nothing of value to the executing program. Systems where the application and its library support code is responsible for detecting and handling stack overflow are called "explicit stack limit". In such systems, the sl register must always point at least 256 bytes higher address than the "minimum sp". Caller-preserved registers A subroutine is free to clobber a1-a4, ip, and lr. Return values If the subroutine returns a simple value no bigger than one word, the value must be in a1 (r0). If the subroutine returns a simple floating-point value, the value is encoded in a1; or {a1, a2}; or {a1, a2, a3}, whichever is sufficient to hold the full precision. A typical subroutine The simplest entry and exit sequence for Thumb functions is: an_example_Thumb_subroutine: PUSH {save-registers, lr} ; one-line entry sequence ; ... first part of function ... BL subroutine_name ;Must be in a space of +/- 4 MB ; ... rest of function goes here, perhaps including other function calls ; ... POP {save-registers, pc} ; one-line exit sequence ARM calling convention The standard C calling convention for ARM is specified in detail by ARM PLC.The "Procedure Call Standard for the ARM Architecture" The simplest entry and exit sequence for 32-bit ARM functions is very similar to Thumb functions:RealView Compilation Tools Developer Guide "Calling between C, C++, and ARM assembly language"section "Stacks and Subroutines" p. 59"Stacking registers for nested subroutines" an_example_ARM32_subroutine: PUSH {r4-r11, lr} ; one-line function prologue ; ... first part of function ... BL subroutine_name ;Must be in a space of +/- 4 MB ; ... rest of function goes here, perhaps including other function calls ; ... POP {r4-r11, pc} ; one-line exit sequence (function epilogue) Using alternate mnemonics for the same instructions, an_example_ARM32_subroutine: ; Push the return address (in LR) and the work registers ; "store multiple registers, full descending" STMFD sp!,{r4-r11, lr} ; aka PUSH {r4-r11, lr} ; (A "sp" alone would leave the stack pointer unchanged. ; We must use "sp!" to update the stack pointer appropriately.) ; ... first part of function ... BL subroutine_name ;Must be in a space of +/- 4 MB ; ... rest of function goes here, perhaps including other function calls ; ... ; Pop the return address (into PC) and the work registers ; and return automatically. ; "load multiple registers, full descending" LDMFD sp!,{r4-r11, pc} ; aka POP {r4-r11, pc} The BL (branch-and-link) instruction stores the return address in the link register LR (r14) and loads the program counter PC (r15) with the subroutine address. Typical subroutines (as shown above) immediately push that return address onto the stack. That frees up r14 so that the subroutine can call sub-subroutines of its own. Subroutine-preserved registers Typically r4-r11 are used to hold local variables of the currently-executing routine. The registers r4-r11 are "subroutine-preserved registers" -- When the subroutine places the return address in pc (r15), returning from the subroutine, the registers r4-r11 and the stack pointer sp (r13) must contain the same values they did when the subroutine was called. Typical subroutines (as shown above) immediately push the values of those registers onto the stack. That frees up r4-r11 to hold local variables of the currently-executing subroutine. Optimizing ARM compilers save and restore the precise subset of r4-r11 and r14 (if any) actually modified by that subroutine, since it is a little slower (but otherwise harmless) to save and restore registers that are unused by that subroutine. scratch registers A subroutine is free to clobber r0-r3, r12, and the link register lr (r14). The first four registers r0-r3 are used to pass argument values into a subroutine and to return a result value from a function. Mixed ARM32 and Thumb calls Normal function calls are easy with the BL instruction. A person types BL destination_subroutine and the assembler and linker will automatically Do the Right Thing -- inserting the appropriate (32-bit-long) ARM32 BL instruction for ARM32-to-ARM32 or ARM32-to-Thumb call or the appropriate (32-bit-long)[infocenter.arm.com/help/topic/com.arm.doc.ddi0234b/i107462.html "Thumb Branch with Link (BL)"] Thumb BL instruction for Thumb-to-Thumb or Thumb-to-ARM32 instruction. (Some mixed calls and some long branches require the linker to insert code that overwrites scratch register r12 with a temporary value. Exactly how the linker does that can be confusing, especially when you mix in using the BX and BLX instructions."Arm/Thumb: using BX in Thumb code, to call a Thumb function, or to jump to a Thumb instruction in another function""Arm / Thumb Interworking") For further reading ARM Limited Embedded Systems/Assembly Language Embedded_Systems/Mixed_C_and_Assembly_Programming#ARM the ARM microcontroller wiki "ARM Overview" at the OS Dev wiki Whirlwind Tour of ARM Assembly GCC ARM Improvement Project at the University of Szeged The ARM Linux Project: Linux for all ARM based machines ARM ARM developers discussion forums ARM Cortex-M3 Technical Reference Manual ARM Assembler by Richard Murray Assembly languages