FeatureDetection.h File Reference

CPU feature detection for SIMD instruction sets. More...

#include "nfx/detail/cpu/FeatureDetection.inl"

Include dependency graph for FeatureDetection.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions
bool	nfx::cpu::hasSse42Support () noexcept
	Gets the cached SSE4.2 support status.
bool	nfx::cpu::hasAvxSupport () noexcept
	Gets the cached AVX support status.
bool	nfx::cpu::hasAvx2Support () noexcept
	Gets the cached AVX2 support status.
bool	nfx::cpu::verifySse42Support () noexcept
	Verifies that compile-time flags match runtime capabilities for SSE4.2.
bool	nfx::cpu::verifyAvxSupport () noexcept
	Verifies that compile-time flags match runtime capabilities for AVX.
bool	nfx::cpu::verifyAvx2Support () noexcept
	Verifies that compile-time flags match runtime capabilities for AVX2.

Detailed Description

CPU feature detection for SIMD instruction sets.

Runtime detection of SSE4.2, AVX, and AVX2 support with static caching for zero-overhead repeated queries

Definition in file FeatureDetection.h.

Function Documentation

hasAvx2Support()

bool nfx::cpu::hasAvx2Support ( )

inlinenodiscardnoexcept

Gets the cached AVX2 support status.

Checks CPU capabilities for AVX2 (Advanced Vector Extensions 2) instructions, which provide 256-bit SIMD operations for vectorized processing. AVX2 enables:

• 256-bit integer operations (vs 128-bit SSE)
• Vectorized string processing and comparison
• Parallel integer computation for multiple data elements
• SIMD-accelerated mathematical operations Result is cached via static initialization for zero runtime overhead.

  Returns

  true if AVX2 is supported, false otherwise.

  Note

  This function is marked [[nodiscard]] - the return value should not be ignored

  Requires CPUID leaf 7, subfunction 0, EBX bit 5

hasAvxSupport()

bool nfx::cpu::hasAvxSupport ( )

inlinenodiscardnoexcept

Gets the cached AVX support status.

Checks CPU capabilities for AVX (Advanced Vector Extensions) instructions, which provide 256-bit SIMD operations for floating-point processing. AVX enables:

• 256-bit floating-point operations (vs 128-bit SSE)
• Vectorized mathematical computations
• SIMD-accelerated floating-point algorithms Result is cached via static initialization for zero runtime overhead.

  Returns

  true if AVX is supported, false otherwise.

  Note

  This function is marked [[nodiscard]] - the return value should not be ignored

  Requires CPUID leaf 1, ECX bit 28

hasSse42Support()

bool nfx::cpu::hasSse42Support ( )

inlinenodiscardnoexcept

Gets the cached SSE4.2 support status.

Checks CPU capabilities for SSE4.2 CRC32 instructions. Result is cached via static initialization for zero runtime overhead.

Returns

true if SSE4.2 is supported, false otherwise.

Note

This function is marked [[nodiscard]] - the return value should not be ignored

Requires CPUID leaf 1, ECX bit 20

verifyAvx2Support()

bool nfx::cpu::verifyAvx2Support ( )

inlinenodiscardnoexcept

Verifies that compile-time flags match runtime capabilities for AVX2.

Checks if AVX2 was compiled in and is available at runtime. Issues warnings or assertions if there's a mismatch.

Returns

true if AVX2 is both compiled and available, false otherwise.

Note

In debug builds, this may assert on compile/runtime mismatches

verifyAvxSupport()

bool nfx::cpu::verifyAvxSupport ( )

inlinenodiscardnoexcept

Verifies that compile-time flags match runtime capabilities for AVX.

Checks if AVX was compiled in and is available at runtime. Issues warnings or assertions if there's a mismatch.

Returns

true if AVX is both compiled and available, false otherwise.

Note

In debug builds, this may assert on compile/runtime mismatches

verifySse42Support()

bool nfx::cpu::verifySse42Support ( )

inlinenodiscardnoexcept

Verifies that compile-time flags match runtime capabilities for SSE4.2.

Checks if SSE4.2 was compiled in and is available at runtime. Issues warnings or assertions if there's a mismatch.

Returns

true if SSE4.2 is both compiled and available, false otherwise.

Note

In debug builds, this may assert on compile/runtime mismatches