Enum ReductionPattern 

pub enum ReductionPattern {
    LayerNorm,
    Softmax,
    Generic,
}

The classified shape of a fused reduction region.

Classification is a descriptive label derived from the member kernel function names; it never affects whether a region is fusible (that decision is purely structural). It is used to tag the fused kernel and to let the codegen layer pick a specialised template.

Variants

LayerNorm

mean → subtract → variance → normalize → scale/shift (two reductions).

Softmax

max → subtract → exp → sum → divide (two reductions, one exp).

Generic

A reduction-broadcast region that does not match a known template.

Implementations

impl ReductionPattern

pub fn name(self) -> &'static str

Returns a short, stable identifier for the pattern.

Trait Implementations

impl Clone for ReductionPattern

fn clone(&self) -> ReductionPattern

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for ReductionPattern

impl Debug for ReductionPattern

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

Formats the value using the given formatter.

impl Display for ReductionPattern

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

Formats the value using the given formatter.

impl Eq for ReductionPattern

impl PartialEq for ReductionPattern

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for ReductionPattern