2.6.2. TI Arm-Specific Pre-Defined Macro Symbols

2.6.2.1. TI Arm C Language Extensions (ACLE) Predefined Macro Symbols

The tiarmclang compiler supports ACLE pre-defined symbols that are relevant for Arm Cortex-M and Cortex-R series processors. The following table summarizes the ACLE pre-defined macros that are supported in the tiarmclang compiler.

Table 2.5 ACLE Predefined Macro Symbols

Symbol

Value Kind

Value / Description

__ARM_ARCH

<version>

Identifies the Arm architecture version being compiled for.

One of the following values:

  • 6 - for Cortex-M0/M0+

  • 7 - for Cortex-M3/M4/R4/R5

  • 8 - for Cortex-M33

__ARM_ACLE

<version>

Identifies the version of ACLE specification that the compiler adheres to, where <version> = <major>*100 + <minor> The tiarmclang compiler adheres to version 2.00 of the ACLE specification (<version> = 200).

__ARM_BIG_ENDIAN

<constant>

Defined to 1 if the -mbig-endian option is specified on the tiarmclang command-line. By default, __ARM_BIG_ENDIAN is not defined.

__ARM_ARCH_ISA_ARM

<constant>

Defined to 1 to indicate Arm processor being compiled for supports ARM mode instructions (Cortex- R4/R5).

__ARM_ARCH_ISA_THUMB

<constant>

Identifies the set of THUMB mode instructions supported by the Arm processor being compiled for. Value is:

  • 1: THUMB1(Cortex-M0/M0+)

  • 2: THUMB2(Cortex-M3/M4/M33, Cortex-R4/R5)

__ARM_32BIT_STATE

<constant>

Defined to 1 if the Arm architecture being compiled for is an AArch32 type.

__ARM_ARCH_PROFILE

<char>

Identifies the Arm architecture profile being compiled for:

  • ‘M’ - for Cortex-M series

  • ‘R’ - for Cortex-R series

  • ‘A’ - for Cortex-A series (not supported by tiarmclang).

__ARM_FEATURE_UNALIGNED

<constant>

Defined to 1 if Arm processor being compiled for supports unaligned memory accesses (i.e. if -munaligned-access is on).

Defined to 1 by default for Cortex-M3/M4/M33/R4/ R5, but can be overridden for the processors using the -mno-unaligned-access compiler option.

__ARM_FEATURE_LDREX

<bitmap>

Identifies which access-widths are supported for LDREX/STREX instructions on the Arm processor being compiled for.

<bitmap> values/width: * bit 0 - byte * bit 1 - half-word * bit 2 - word * bit 3 - double-word

<bitmap> values supported: * 0x7 - Cortex-M3/M4 * 0xF - Cortex-M33/R4/R5

__ARM_FEATURE_CLZ

<constant>

Defined to 1 if the CLZ instruction is available on the Arm processor being compiled for (including Cortex-M3/M4/M33/R4/R5).

__ARM_FEATURE_QBIT

<constant>

Defined to 1 if the Q saturation flag exists and relevant Q saturation arithmetic intrinsics are available on the Arm processor being compiled for (Cortex-M3/M4/M33/R4/ R5).

__ARM_FEATURE_DSP

<constant>

Defined to 1 if DSP instructions are supported and relevant intrinsics are available on the Arm processor being compiled for (Cortex-M4/M33/R4/R5).

__ARN_FEATURE_SAT

<constant>

Defined to 1 if the SSAT and USAT instructions are supported and the relevant intrinsics are available on the Arm processor being compiled for (Cortex-M3/ M4/M33/R4/R5).

__ARM_FEATURE_SIMD32

<constant>

Defined to 1 if the 32-bit SIMD instructions are supported and the relevant intrinsics are available on the Arm processor being compiled for (Cortex-M4/ M33/R4/R5).

__ARM_FEATURE_IDIV

<constant>

Defined to 1 if hardware integer divide instructions are available on the Arm processor being compiled for (Cortex-M3/ M4/M33).

__ARM_FP

<bitmap>

Defined if floating-point hardware is available. The value assigned to __ARM_FP is a bitmap indicating what floating-point precisions are supported:

  • bit 1 - half (16-bit)

  • bit 2 - single (32-bit)

  • bit 3 - double (64-bit)

<bitmap> values supported: * 0x6 - Cortex-M4/M33 * 0xC - Cortex-R4/R5

__ARM_FEATURE_FMA

<constant>

Defined to 1 if floating- point hardware that supports fused multiply- and-accumulate operations is available on the Arm processor being compiled for (Cortex-M4/ M33).

__ARM_FEATURE_DIRECTED _ROUNDING

<constant>

Defined to 1 if the directed rounding and conversion vector instructions are supported and the relevant intrinsics are available on the Arm processor being compiled for (Cortex-M33).

__ARM_FEATURE_NUMERIC _MAXMIN

<constant>

Defined to 1 if the IEEE 754-2008 compliant floating-point maximum and minimum vector instructions are supported and the relevant intrinsics are available on the Arm processor being compiled for (Cortex-M33).

__ARM_PCS

<constant>

Defined to 1 if the compiler conforms to the default procedure calling standard as defined in the Procedure Call Standard for the Arm Architecture (Cortex-M0/M0+/M3/M4/M33/ R4/R5).

__ARM_PCS_VFP

<constant>

Defined to 1 if the compiler will pass floating-point arguments and return values in FPU registers according to the Procedure Call Standard for the Arm Architecture when floating-point hardware is available during compilation (Cortex-M4/M33/R4/R5).

__ARM_SIZEOF_WCHAR_T

<bytes>

Defined to <bytes>, where <bytes> is the minimum number of bytes used by the compiler to represent the value of a wchar_t type object. On Cortex-M0/ M0+M3/M4/M33/R4/R5, <bytes> = 4.

__ARM_SIZEOF_MINIMAL_ENUM

<bytes>

Defined to <bytes>, where <bytes> is the minimum number of bytes used by the compiler to represent the value of an enum object. Cortex-M0/M0+/M3/ M4/M33/R4/R5, <bytes> = 1.

More details about ACLE pre-defined macro symbols can be found in the Arm C Language Extensions Documentation page.

2.6.2.2. Other Arm-Specific Predefined Macro Symbols

Table 2.6 Arm-Specific Architecture/Processor/Instruction Set Macro Symbols

Symbol

Value Kind

Value / Description

__arm__

<constant>

Defined to 1 if target of compile is an Arm processor.

__arm

<constant>

Defined to 1 if target of compile is an Arm processor.

__thumb__

<constant>

Defined to 1 if compiling for an Arm processor in Thumb mode.

__thumb2__

<constant>

Defined to 1 if T32 (Thumb2) instructions are available when compiling for an Arm processor.

__ARM_ARCH_6M__

<constant>

Defined to 1 if compiling for an Arm processor based on the version 6M Arm architecture (Cortex-M0/M0+).

__ARM_ARCH_7M__

<constant>

Defined to 1 if compiling for an Arm processor based on the version 7M Arm architecture (Cortex-M3).

__ARM_ARCH_7EM__

<constant>

Defined to 1 if compiling for an Arm processor based on the version 7EM Arm architecture (Cortex-M4).

__ARM_ARCH_8M_MAIN__

<constant>

Defined to 1 if compiling for an Arm processor based on the version 8M- MAIN Arm architecture (Cortex-M33).

__ARM_ARCH_7R__

<constant>

Defined to 1 if compiling for an Arm processor based on the version 7R Arm architecture (Cortex-R4/R5).

__VFP_FP__

<constant>

Defined to 1 if floating-point hardware use is enabled for a tiarmclang compilation.

__ARM_VFPV2__

<constant>

Defined to 1 if use of the VFPV2 floating-point hardware is enabled for a tiarmclang compilation.

__ARM_VFPV3__

<constant>

Defined to 1 if use of the VFPV3 floating-point hardware is enabled for a tiarmclang compilation.

__ARM_VFPV4__

<constant>

Defined to 1 if use of the VFPV4 floating-point hardware is enabled for a tiarmclang compilation.

__ARM_FPV5__

<constant>

Defined to 1 if use of the FPV5 floating-point hardware is enabled for a tiarmclang compilation.

__ARMEB__

<constant>

Defined to 1 if compiler has been instructed to generate big-endian object code.

__ARMEL__

<constant>

Defined to 1 if compiler has been instructed to generate little-endian object code (default).

__THUMBEB__

<constant>

Defined to 1 if compiling for a Cortex-M series Arm processor in big-endian mode.

__THUMBEL__

<constant>

Defined to 1 if compiling for a Cortex-M series Arm processor in little-endian mode.
Table 2.7 Arm-Specific Feature Test Predefined Macro Symbols

Symbol

Value Kind

Value / Description

__APCS_32__

<constant>

Defined to 1 unless a different procedure calling convention is in effect for a given compilation.

__ARM_ARCH_EXT_IDIV__

<constant>

Defined to 1 if hardware integer divide instructions are available on the Arm processor being compiled for (Cortex-M3/M4/M33/R5).

__ARM_FEATURE_CMSE

<constant>

Defined to 1 if Cortex-M Security Extensions (CMSE) are supported on the Arm processor being compiled for (Cortex-M33).

__ARM_FP_FAST

<constant>

Defined to 1 if the -ffast-math or -ffp-mode=fast option is enabled (-ffast-math is implied when the deprecated -Ofast optimization level is specified on the compiler command-line).