ARM Specific Information

4. ARM Specific Information#

This chapter discusses the ARM architecture (http://en.wikipedia.org/wiki/ARM_architecture) dependencies in this port of RTEMS. The ARMv4T (and compatible), ARMv7-A, ARMv7-R and ARMv7-M architecture versions are supported by RTEMS. Processors with a MMU use a static configuration which is set up during system start. SMP is supported.

Architecture Documents

For information on the ARM architecture refer to the ARM Infocenter (http://infocenter.arm.com/).

4.1. CPU Model Dependent Features#

This section presents the set of features which vary across ARM implementations and are of importance to RTEMS. The set of CPU model feature macros are defined in the file cpukit/score/cpu/arm/rtems/score/arm.h based upon the particular CPU model flags specified on the compilation command line.

4.1.1. CPU Model Name#

The macro CPU_MODEL_NAME is a string which designates the architectural level of this CPU model. See in cpukit/score/cpu/arm/rtems/score/arm.h for the values.

4.1.2. Count Leading Zeroes Instruction#

The ARMv5 and later instruction sets have the count leading zeroes clz instruction which could be used to speed up the find first bit operation. The use of this instruction should significantly speed up the scheduling associated with a thread blocking. This is currently not used.

4.1.3. Floating Point Unit#

The following floating point units are supported:

  • VFPv2 (for example available on ARM926EJ-S processors)

  • VFPv3-D32/NEON (for example available on Cortex-A processors)

  • VFPv3-D16 (for example available on Cortex-R processors)

  • FPv4-SP-D16 (for example available on Cortex-M processors)

  • FPv5-D16 (for example available on Cortex-M7 processors)

4.2. Multilibs#

The following multilibs are available:

  1. .: ARMv4T, ARM instruction set

  2. vfp/hard: ARMv4T, ARM instruction set with hard-float ABI and VFPv2 support

  3. thumb: ARMv4T, Thumb-1 instruction set

  4. thumb/armv6-m: ARMv6M, subset of Thumb-2 instruction set

  5. thumb/armv7-a: ARMv7-A, Thumb-2 instruction set

  6. thumb/armv7-a/neon/hard: ARMv7-A, Thumb-2 instruction set with hard-float ABI Neon and VFP-D32 support

  7. thumb/armv7-r: ARMv7-R, Thumb-2 instruction set

  8. thumb/armv7-r/vfpv3-d16/hard: ARMv7-R, Thumb-2 instruction set with hard-float ABI VFP-D16 support

  9. thumb/cortex-m3: Cortex-M3, Thumb-2 instruction set with hardware integer division (SDIV/UDIV) and a fix for Cortex-M3 Errata 602117.

  10. thumb/cortex-m4: Cortex-M4, Thumb-2 instruction set with hardware integer division (SDIV/UDIV) and DSP instructions

  11. thumb/cortex-m4/fpv4-sp-d16: Cortex-M4, Thumb-2 instruction set with hardware integer division (SDIV/UDIV), DSP instructions and hard-float ABI FPv4-SP support

  12. thumb/cortex-m7/fpv5-d16: Cortex-M7, Thumb-2 instruction set with hard-float ABI and FPv5-D16 support

  13. eb/thumb/armv7-r: ARMv7-R, Big-endian Thumb-2 instruction set

  14. eb/thumb/armv7-r/vfpv3-d16/hard: ARMv7-R, Big-endian Thumb-2 instruction set with hard-float ABI VFP-D16 support

Multilib 1., 2. and 3. support the legacy ARM7TDMI and ARM926EJ-S processors.

Multilib 4. supports the Cortex-M0 and Cortex-M1 cores.

Multilib 5. and 6. support the Cortex-A processors.

Multilib 7., 8., 13. and 14. support the Cortex-R processors. Here also big-endian variants are available.

Use for example the following GCC options:

-mthumb -march=armv7-a -mfpu=neon -mfloat-abi=hard -mtune=cortex-a9

to build an application or BSP for the ARMv7-A architecture and tune the code for a Cortex-A9 processor. It is important to select the options used for the multilibs. For example:

-mthumb -mcpu=cortex-a9

alone will not select the ARMv7-A multilib.

4.3. Calling Conventions#

Please refer to the Procedure Call Standard for the ARM Architecture (http://infocenter.arm.com/help/topic/com.arm.doc.ihi0042c/IHI0042C_aapcs.pdf).

4.4. Memory Model#

A flat 32-bit memory model is supported. The board support package must take care of initializing the MMU if necessary.

Note that architecture variants which support unaligned accesses must not use memcpy() or memset() on device memory as those functions are hand-optimized and will take advantage of unaligned accesses where available. As per ARM (https://developer.arm.com/documentation/ddi0406/c/Application-Level-Architecture/Application-Level-Memory-Model/Alignment-support/Unaligned-data-access-restrictions-in-ARMv7-and-ARMv6), unaligned accesses are not permitted for device memory.

4.5. Interrupt Processing#

The ARMv4T (and compatible) architecture has seven exception types:

  • Reset

  • Undefined

  • Software Interrupt (SWI)

  • Prefetch Abort

  • Data Abort

  • Interrupt (IRQ)

  • Fast Interrupt (FIQ)

Of these types only the IRQ has explicit operating system support. It is intentional that the FIQ is not supported by the operating system. Without operating system support for the FIQ it is not necessary to disable them during critical sections of the system.

The ARMv7-M architecture has a completely different exception model. Here interrupts are disabled with a write of 0x80 to the basepri_max register. This means that all exceptions and interrupts with a priority value of greater than or equal to 0x80 are disabled. Thus exceptions and interrupts with a priority value of less than 0x80 are non-maskable with respect to the operating system and therefore must not use operating system services. Several support libraries of chip vendors implicitly shift the priority value somehow before the value is written to the NVIC IPR register. This can easily lead to confusion.

4.5.1. Interrupt Levels#

There are exactly two interrupt levels on ARM with respect to RTEMS. Level zero corresponds to interrupts enabled. Level one corresponds to interrupts disabled.

4.5.2. Interrupt Stack#

The board support package must initialize the interrupt stack. The memory for the stacks is usually reserved in the linker script.

4.6. Symmetric Multiprocessing#

SMP is supported on ARMv7-A. Available platforms are:

  • Altera Cyclone V

  • NXP i.MX 7

  • Xilinx Zynq

4.7. Thread-Local Storage#

Thread-local storage is supported.