Skip to main content
DirectXMath Library Compiler Directives

Compiler directives tune the functionality that the DirectXMath library uses.

DirectiveDescription
_XM_NO_INTRINSICS_

When _XM_NO_INTRINSICS_ is defined, DirectXMath operations are implemented without using any platform-specific intrinsics. Instead, DirectXMath uses only standard floating point operations.

By default, _XM_NO_INTRINSICS_ is not defined.

This directive overrides all other directives. Therefore, we recommend that you check for _XM_NO_INTRINSICS_ before you determine the status of _XM_ARM_NEON_INTRINSICS_ or _XM_SSE_INTRINSICS_.

_XM_SSE_INTRINSICS_

When _XM_SSE_INTRINSICS_ is defined, code is compiled to use supporting SSE and SSE2 on platforms that support these instruction sets.

The Windows versions providing SSE intrinsics support both SSE and SSE2.

_XM_SSE_INTRINSICS_ has no effect on systems that do not support SSE and SSE2.

By default, _XM_SSE_INTRINSICS_ is defined when users compile for a Windows platform.

DirectXMath uses the standard compiler defines (_M_IX86 / _M_AMD64) to determine Windows x86 or Windows x64 targets.

_XM_ARM_NEON_INTRINSICS_

This indicates the DirectXMath implementation uses of the ARM-NEON intrinsics. By defualt, _XM_ARM_NEON_INTRINSICS_ is defined when users compile for a Windows RT platform. DirectXMath uses the standard compiler define (_M_ARM, _M_ARM64) to determine this binary target.

ARM-NEON intrinsics support is new to DirectXMath and is not supported by XNAMath 2.x.

_XM_SSE4_INTRINSICS_

New for Windows 10 Anniversary SDK.

When you specify this compiler directive, DirectXMath functions are implemented to make use of SSE3 and SSE4.1 instriniscs where applicable; otherwise it uses to SSE/SSE2.

_XM_AVX_INTRINSICS_

New for Windows 10 Anniversary SDK.

Use of /arch:AVX will enable this directive.

When you specify this compiler directive, DirectXMath functions are implemented to make use of SSE3, SSE4.1, and AVX 128-bit intrinsics where applicable; otherwise DirectXMath uses SSE/SSE2. When you specify this compiler directive, it also implies _XM_SSE4_INTRINSICS_ and _XM_SSE_INTRINSICS_, and includes a small subset of SSE3 instructions.

_XM_F16C_INTRINSICS_

New for Windows 10 Anniversary SDK.

Use of /arch:AVX2 will enable this directive.

When you specify this compiler directive, DirectXMath functions are implemented to make use of SSE3, SSE4.1, AVX 128-bit, and F16C/CVT16 intrinsics where applicable; otherwise DirectXMath uses SSE/SSE2. When you use _XM_F16C_INTRINSICS_, it implies _XM_AVX_INTRINSICS_, _XM_SSE4_INTRINSICS_, and _XM_SSE_INTRINSICS_.

_XM_VECTORCALL_

This enables the use of the new __vectorcall calling convention for Windows x86 or Windows x64 targets. It defaults to _XM_VECTORCALL_=1 when the compiler supports this feature. You can manually set it to _XM_VECTORCALL_=0 to always disable it.

__vectorcall is not supported for the Windows RT platform or Managed C++.

 

Related topics

DirectXMath Programming Reference