Enum DenseF32Kernel 

pub enum DenseF32Kernel {
    Avx512,
    Avx2Fma,
    Sse,
    Scalar,
}

f32 dot / fused dot+norm kernel resolved once for a whole batch.

The batch scorers (batch_*_precomp) call a kernel once per stored vector; resolving it up front keeps runtime feature detection out of that loop (mirrors HammingKernel). Every variant has the scalar fallback.

Variants

Avx512

Avx2Fma

Sse

Scalar

Implementations

impl DenseF32Kernel

pub fn resolve() -> Self

Detect the widest kernel this CPU supports.

pub fn dot(self, a: &[f32], b: &[f32], count: usize) -> f32

dot(a[..count], b[..count]). Callers guarantee count is in bounds.

pub fn fused_dot_norm(self, a: &[f32], b: &[f32], count: usize) -> (f32, f32)

(dot(a, b), dot(b, b)) over the first count elements.

Trait Implementations

impl Clone for DenseF32Kernel

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for DenseF32Kernel

impl Debug for DenseF32Kernel

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

Formats the value using the given formatter.

impl Eq for DenseF32Kernel

impl PartialEq for DenseF32Kernel

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for DenseF32Kernel