Struct SimdCapabilityInfo 

pub struct SimdCapabilityInfo {
    pub has_sse42: bool,
    pub has_avx: bool,
    pub has_avx2: bool,
    pub has_fma: bool,
    pub has_avx512f: bool,
    pub has_avx512bw: bool,
    pub has_avx512vl: bool,
    pub has_neon: bool,
    pub has_sve: bool,
    pub cache_line_bytes: usize,
    pub vector_width_bytes: usize,
}

Extended CPU SIMD capability description with the field naming convention required by the multi-versioning layer.

On std-enabled builds, SimdCapabilityInfo::detect returns a &'static Self reference that is initialised exactly once per process (via OnceLock). On no_std builds it returns a fresh Self value computed from compile-time cfg! constants.

Fields

has_sse42: bool

SSE4.2 support (x86-64 only).

has_avx: bool

AVX support (x86-64 only).

has_avx2: bool

AVX2 support (x86-64 only).

has_fma: bool

FMA (Fused Multiply-Add) support (x86-64 only).

has_avx512f: bool

AVX-512 Foundation support (x86-64 only).

has_avx512bw: bool

AVX-512 Byte & Word instructions (x86-64 only).

has_avx512vl: bool

AVX-512 Vector Length extensions (x86-64 only).

has_neon: bool

NEON support. Always true on AArch64.

has_sve: bool

SVE (Scalable Vector Extension) support.

cache_line_bytes: usize

Bytes in a single cache line (typically 64 on modern CPUs).

vector_width_bytes: usize

Width of the widest supported SIMD vector register in bytes.

Implementations

impl SimdCapabilityInfo

pub fn detect() -> &'static Self

Detect (or derive) capabilities for the current CPU and return a reference to the process-wide cached value.

The first call performs runtime detection and stores the result. Subsequent calls return the same &'static reference.

pub fn has_avx512_full(&self) -> bool

Returns true when AVX-512F, BW, and VL are all present.

pub fn has_avx2_fma(&self) -> bool

Returns true when both AVX2 and FMA are present.

pub fn f64_simd_width(&self) -> usize

Number of f64 elements that fit in the widest supported SIMD register.

For AVX-512 this is 8; for AVX2 / NEON-128 this is 4 / 2; for scalar 1.

pub fn f32_simd_width(&self) -> usize

Number of f32 elements that fit in the widest supported SIMD register.

Always twice f64_simd_width().

pub fn optimal_level(&self) -> LegacyLevel

Returns the LegacyLevel that best summarises these capabilities.

Trait Implementations

impl Clone for SimdCapabilityInfo

fn clone(&self) -> SimdCapabilityInfo

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SimdCapabilityInfo

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for SimdCapabilityInfo

fn eq(&self, other: &SimdCapabilityInfo) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for SimdCapabilityInfo

impl Eq for SimdCapabilityInfo

impl StructuralPartialEq for SimdCapabilityInfo