Enum ReduceOp 

#[non_exhaustive]
pub enum ReduceOp {
    Sum,
    Product,
    StableProduct,
    SumSquares,
    Max,
    Min,
}

Runtime reduction operation for dtype-erased full reductions.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Sum

Product

StableProduct

Reassociated, overflow/underflow-safe product.

Tracks a signed mantissa and an exact integer exponent per lane, so factors outside the dtype range do not overflow or underflow a partial. The reduction is reassociated and threaded like Self::Sum.

A zero, subnormal, infinite or NaN input cannot be represented by the exponent-tracked accumulator; the affected output is then recomputed in source order, keeping IEEE propagation (0 * inf = NaN, a zero or an infinity result) identical to the sequential fold. For integers this is the reassociated wrapping product and for complex dtypes the source-order product.

SumSquares

Sum of same-dtype rounded squares.

Each input is first multiplied by itself without FMA contraction, then accumulated under the same association policy as Self::Sum.

Max

NaN-propagating maximum for f32/f64, ordered maximum for i32/i64.

Any NaN operand makes the result the canonical NaN of the dtype. The identity written for an empty reduction is -inf for floats and the dtype minimum for integers. Complex and bool dtypes are rejected. When the reduced values contain both -0.0 and +0.0, which zero is returned is unspecified.

Min

NaN-propagating minimum for f32/f64, ordered minimum for i32/i64.

Any NaN operand makes the result the canonical NaN of the dtype. The identity written for an empty reduction is +inf for floats and the dtype maximum for integers. Complex and bool dtypes are rejected. When the reduced values contain both -0.0 and +0.0, which zero is returned is unspecified.

Trait Implementations

impl Clone for ReduceOp

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 ReduceOp

impl Debug for ReduceOp

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

Formats the value using the given formatter.

impl Eq for ReduceOp

impl PartialEq for ReduceOp

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for ReduceOp