The compiler supports the predefined macros listed in the table below. The compiler also includes predefined macros specified by the ISO/ANSI standard.
Unless otherwise stated, the macros are supported on systems based on IA-32 and Intel® 64 architectures. IA-32 is not available on macOS*.
The Intel® C++ Compiler defines the same target-architecture macros that GCC does. For -m feature, GCC defines __FEATURE__.
You can target specific processor architectures by using the -x, -m, and -march compiler options. Each of these options enables feature-specific macros in the compiler. These macros are used to guard a section of application code that uses target-specific feature. The following command emits the list of predefined macros enabled by targeting a specific processor architecture:
icpc -dM -E helloworld.cc -xarch
For example, you could do the following to determine which feature macros would help identify whether this is ICELAKE-SERVER:
icpc -dM -E helloworld.cc -xSKYLAKE-AVX512 > avx512.txt 2>&1 icpc -dM -E helloworld.cc -xICELAKE-SERVER > icelake.txt 2>&1 diff avx512.txt icelake.txt 317a318 > #define AVX512IFMA 1 320a322,329 > #define AVX512VBMI 1 > #define AVX512VPOPCNTDQ 1 > #define AVX512BITALG 1 > #define AVX512VBMI2 1 > #define GFNI 1 > #define VAES 1 > #define VPCLMUL 1 > #define AVX512VNNI 1 321a331,334 > #define RDPID 1 > #define SGX 1 > #define WBNOINVD 1 > #define PCONFIG 1
The result of the diff command is the list of feature macros that can be used to differentiate icelake-server from skylake-avx512.
| Macro | Description |
|---|---|
__APPLE__ (macOS*) |
Defined as '1'. |
__APPLE_CC__ (macOS*) |
The gcc* build number |
__ARRAY_OPERATORS (Linux*) |
Defined as '1'. |
__AVX__ (Windows*, Linux, macOS*) |
On Windows*, defined as '1' when option /arch:AVX, /QxAVX, or higher processor targeting options are specified. On Linux*, defined as '1' when option -march=corei7-avx, -mavx, -xAVX, or higher processor targeting options are specified. NoteAvailable only for compilations targeting Intel® 64 architecture. |
__AVX2__ (Windows, Linux, macOS*) |
On Windows, defined as '1' when option /arch:CORE-AVX2, /QxCORE-AVX2, or higher processor targeting options are specified. On Linux, defined as '1' when option -march=core-avx2, -xCORE-AVX2, or higher processor targeting options are specified. NoteAvailable only for compilations targeting Intel® 64 architecture. |
__AVX512BW__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Byte and Word instructions. It is also defined as '1' when option [Q]xCORE-AVX512 or higher processor-targeting options are specified. |
__AVX512CD__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Conflict Detection instructions. It is also defined as '1' when option [Q]xCORE-AVX512, [Q]xCOMMON-AVX512, or higher processor-targeting options are specified. |
__AVX512DQ__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Doubleword and Quadword instructions. It is also defined as '1' when option [Q]xCORE-AVX512 or higher processor-targeting options are specified. |
__AVX512ER__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Exponential and Reciprocal instructions. |
__AVX512F__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Foundation instructions. It is also defined as '1' when option [Q]xCORE-AVX512, [Q]xCOMMON-AVX512, or higher processor-targeting options are specified. |
__AVX512PF__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Prefetch instructions. |
__AVX512VL__ (Windows*, Linux, macOS*) |
Defined as '1' for processors that support Intel® Advanced Vector Extensions 512 (Intel® AVX-512) Vector Length extensions. It is also defined as '1' when option [Q]xCORE-AVX512 or higher processor-targeting options are specified. |
__BASE_FILE__ (Linux) |
Name of source file |
_BOOL (Linux) |
Defined as '1'. |
__COUNTER__ (Windows) |
Defined as '0'. |
__cplusplus (Linux) |
Defined as '1' (for the Intel® C++ Compiler). |
__DEPRECATED (Linux) |
Defined as '1'. |
__DYNAMIC__ (macOS*) |
Defined as '1'. |
__EDG__ (Windows, Linux, macOS*) |
Defined as '1'. |
__EDG_VERSION__ (Windows, Linux, macOS*) |
EDG version |
__ELF__ (Linux) |
Defined as '1' at the start of compilation. |
__EXCEPTIONS (Linux) |
Defined as '1' when option fno-exceptions is not used. |
__gnu_linux__ (Linux) |
Defined as '1' at the start of compilation. |
__GNUC__ (Linux) |
The major version number of gcc* installed on the system or explicitly specified via –gcc-name/ –gxx-name. |
__GNUC_MINOR__ (Linux) |
The minor version number of gcc* or g++* installed on the system or explicitly specified via –gcc-name/ –gxx-name. |
__GNUC_PATCHLEVEL__ (Linux) |
The patch level version number of gcc* or g++* installed on the system or explicitly specified via –gcc-name/ –gxx-name. |
__GNUG__ (Linux) |
The major version number of g++* installed on the system or explicitly specified via –gcc-name/ –gxx-name. |
__GXX_ABI_VERSION (Linux) |
The value of this is dependent on the -fabi-version option in effect. 102: -fabi-version=1 || gcc version < 3.4 1008: gcc version >= 5.0 999999: -fabi-version=0 100? Where ? matches the -fabi-version passed: -fabi-version=2,3,4,5,6,7,8,9 1002: Gcc version > 3.5 and < 5.0 |
__HONOR_STD (Linux, macOS*) |
Defined as '1'. |
__i386__ __i386 i386 (Linux, macOS*) |
Defined as '1' for compilations targeting IA-32 architecture. IA-32 is not available on macOS*. |
__ICC (Linux, macOS*) |
The version of the compiler. NoteThis macro may be affected by compiler options, such as -no-icc. |
__ICL (Windows) |
The version of the compiler. NoteThis macro may be affected by compiler options, such as /Qicl-. |
_INC_STDIO (Windows) |
Defined, no value. |
_INTEGRAL_MAX_BITS (Windows) |
64 |
__INTEL_COMPILER (Windows*, Linux, macOS*) |
The version of the compiler. NoteThis macro may be affected by compiler options, such as -no-icc. |
__INTEL_COMPILER_BUILD_DATE (Windows*, Linux, macOS*) |
The compiler build date. It takes the form YYYYMMDD, where YYYY is the year, MM is the month, and DD is the day. |
__INTEL_COMPILER_UPDATE (Windows, Linux, macOS*) |
Returns the current minor update number of the compiler, starting at 0. You can use this macro to differentiate between compiler updates when you have multiple updates of the Intel® C++ Compiler installed concurrently. Example: For Intel® C++ Compiler version XX.0.2, the macro would preprocess to "2". |
__INTEL_CXX11_MODE__ (Windows, Linux) |
Enables C++11 experimental support for C++ programs. Defined as '1' when option [Q]std=c++11 is specified. |
__INTEL_MS_COMPAT_LEVEL (Windows) |
Defined as '1'. Equal to the same value n as specified by option [Q]msn. |
__INTEL_RTTI__ (Linux, macOS*) |
Defined as '1' when option -fno-rtti is not specified. |
__INTEL_STRICT_ANSI__ (Linux, macOS*) |
Defined as '1' when option -strict-ansi is specified. |
__linux__ __linux linux (Linux) |
Defined as '1' at the start of compilation. |
__LITTLE_ENDIAN__ (macOS*) |
Defined as '1'. |
__LONG_DOUBLE_SIZE__ (Windows*, Linux, macOS*) |
On Linux and macOS*, defined as 80. On Windows, defined as 64; defined as 80 when option /Qlong-double is specified. |
__LONG_DOUBLE_64__ (Linux) |
When this macro is defined, the long double type is 64-bits. It is defined when you specify option -mlong-double-64. |
__LONG_MAX__ (Linux) |
9223372036854775807L NoteAvailable only for compilations targeting Intel® 64 architecture. |
__LP64__ (Linux) __LP64 (Linux) |
Defined as '1'. NoteAvailable only for compilations targeting Intel® 64 architecture. |
_M_AMD64 (Windows) |
Defined as '1' while building code targeting Intel® 64 architecture. |
_M_IX86 (Windows) |
700 |
_M_X64 (Windows) |
Defined as '1' while building code targeting Intel® 64 architecture. |
__MACH__ (macOS*) |
Defined as '1'. |
__MMX__ (Linux, macOS*) |
Defined as '1'. On Linux, it is available only on systems based on Intel® 64 architecture. |
_MSC_EXTENSIONS (Windows) |
Defined as '1'. This macro is defined when Microsoft extensions are enabled. |
_MSC_FULL_VER (Windows) |
The Visual C++* version being used. 190022609 for Visual C++* 2015 1800210051 for Visual C++* 2013 |
_MSC_VER (Windows) |
The Visual C++* version being used. 1900 for Visual C++* 2015 1800 for Visual C++* 2013 |
_MT (Windows) |
On Windows, defined as '1' when a multithreaded DLL or library is used (when option /MD[d] or /MT[d] is specified). |
__NO_INLINE__ __NO_MATH_INLINES __NO_STRING_INLINES (Linux, macOS*) |
Defined as '1'. |
_OPENMP (Windows, Linux, macOS*) |
201611 when you specify option [Q]openmp. |
__OPTIMIZE__ (Linux, macOS*) |
Defined as '1'. |
__pentium4 __pentium4__ (Linux, macOS*) |
Defined as '1'. |
_PGO_INSTRUMENT (Windows, Linux) |
Defined as '1' when option [Q]cov-gen or [Q]prof-gen is specified. |
__PIC__ __pic__ (Linux, macOS*) |
On Linux, defined as '1' when option fPIC is specified. On macOS*, defined as '1'. Only __PIC__ is allowed on macOS*. |
_PLACEMENT_DELETE (Linux) |
Defined as '1'. |
__PTRDIFF_TYPE__ (Linux, macOS*) |
On Linux, defined as int on IA-32 architecture; defined as long on Intel® 64 architecture. On macOS*, defined as int/long. |
__QMSPP_ (Windows, macOS*) |
Defined as '1'. |
__REGISTER_PREFIX__ (Linux, macOS*) |
|
__SIGNED_CHARS__ (Windows, Linux, macOS*) |
Defined as '1'. |
_SIZE_T_DEFINED (Windows) |
Defined, no value. |
__SIZE_TYPE__ (Linux, macOS*) |
On Linux, defined as unsigned on IA-32 architecture; defined as unsigned long on Intel® 64 architecture. On macOS*, defined as unsigned long. |
__SSE__ (Windows, Linux, macOS*) |
On Linux and macOS*, defined as '1' for processors that support SSE instructions. On Windows, defined as '1'. It is undefined when option /arch:IA32 is specified. |
__SSE2__ (Windows, Linux, macOS*) |
On Linux and macOS*, defined as '1' for processors that support Intel® SSE2 instructions. On Windows, defined as '1' by default or when option /arch:SSE2, /QxSSE2, /QaxSSE2, or higher processor targeting options are specified. |
__SSE3__ (Windows, Linux, macOS*) |
On Linux and macOS*, defined as '1' for processors that support Intel® SSE3 instructions. On Windows, defined as '1' when option /arch:SSE3, /QxSSE3, or higher processor targeting options are specified. |
__SSE4_1__ (Windows, Linux) |
On Linux, defined as '1' for processors that support Intel® SSE4 instructions. On Windows, defined as '1' when option /arch:SSE4.1, /QxSSE4.1, or higher processor targeting options are specified. |
__SSE4_2__ (Windows, Linux) |
On Linux, defined as '1' for processors that support SSSE4 instructions. On Windows, defined as '1' when option /arch:SSE4.2, /QxSSE4.2, or higher processor targeting options are specified. |
__SSSE3__ (Windows, Linux, macOS*) |
On Linux and macOS*, defined as '1' for processors that support SSSE3 instructions. On Windows, defined as '1' when option arch:SSSE3, QxSSSE3, or higher processor targeting options are specified. |
__STDC__ (macOS*) |
Defined as '1'. |
__STDC_HOSTED__ (macOS*) |
Defined as '1'. |
unix __unix __unix__ (Linux) |
Defined as '1'. |
__USER_LABEL_PREFIX__ (Linux, macOS*) |
|
_VA_LIST_DEFINED (Windows) |
Defined, no value. |
__VERSION__ (Linux, macOS*) |
The compiler version string |
__w64 (Windows) |
Defined, no value. |
__WCHAR_MAX__ (macOS*) |
2147483647 |
__WCHAR_T (Linux) |
Defined as '1'. |
_WCHAR_T_DEFINED (Windows) |
Defined when option /Zc:wchar_t is specified or "wctype_t" is defined in the header file. |
__WCHAR_TYPE__ (Linux, macOS*) |
On Linux, defined as long int on IA-32 architecture; defined as int on Intel® 64 architecture. On macOS*, defined as long int. |
_WCTYPE_T_DEFINED (Windows) |
Defined when "wctype_t" is defined in the header file. |
_WIN32 (Windows) |
Defined as '1' while building code targeting IA-32 or Intel® 64 architecture. IA-32 is not available on macOS*. |
_WIN64 (Windows) |
Defined as '1' while building code targeting Intel® 64 architecture. |
__WINT_TYPE__ (Linux, macOS*) |
Defined as unsigned int. |
__x86_64 __x86_64__ (Linux, macOS*) |
Defined as '1' while building code targeting Intel® 64 architecture. |