Enum ReduceKind 

#[non_exhaustive]
#[repr(u16)]
pub enum ReduceKind {
    Sum = 0,
    Mean = 1,
    Max = 2,
    Min = 3,
    Prod = 4,
    Var = 5,
    Std = 6,
    Norm2 = 7,
    Argmax = 8,
    Argmin = 9,
    Any = 10,
    All = 11,
    LogSumExp = 12,
    Trace = 13,
    CountNonzero = 14,
}

Reduction op discriminant — Phase 4 (Category E).

Output shape differs from input: the reduced axis collapses to size 1 (keepdim convention). Other variants are reserved for fanout.

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 = 0

Sum along the reduced axis. Phase 4 trailblazer.

Mean = 1

Arithmetic mean along the reduced axis.

Max = 2

Maximum value along the reduced axis.

Min = 3

Minimum value along the reduced axis.

Prod = 4

Product along the reduced axis.

Var = 5

Sample variance (Bessel-corrected) along the reduced axis.

Std = 6

Sample standard deviation along the reduced axis.

Norm2 = 7

||x||_2 along the reduced axis.

Argmax = 8

argmax along the reduced axis — returns indices (different output dtype). Will need a separate plan shape, reserved here.

Argmin = 9

argmin along the reduced axis. Will need a separate plan shape.

Any = 10

any (logical OR) along the reduced axis.

All = 11

all (logical AND) along the reduced axis.

LogSumExp = 12

logsumexp(x) = log(sum(exp(x - max))), numerically stable.

Trace = 13

trace(M) = sum(diag(M)) — sum of the diagonal of a 2-D square matrix. Reduces both axes via the i == i constraint rather than a single reduce-axis, so dispatch goes through a dedicated TracePlan (separate from ReducePlan); the discriminant lives here for telemetry / SKU-tagging consistency with the rest of the reduction family.

CountNonzero = 14

count_nonzero(x) along the reduced axis — output is i64 (PyTorch torch.count_nonzero returns int64). Heterogeneous output dtype (always i64 regardless of input), so dispatch goes through a dedicated CountReducePlan (separate from ReducePlan); the discriminant lives here for telemetry / SKU-tagging consistency with the rest of the reduction family.

Trait Implementations

impl Clone for ReduceKind

fn clone(&self) -> ReduceKind

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for ReduceKind

impl Debug for ReduceKind

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

Formats the value using the given formatter.

impl Eq for ReduceKind

impl Hash for ReduceKind

fn hash<__H>(&self, state: &mut __H)

where __H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for ReduceKind

fn eq(&self, other: &ReduceKind) -> 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 StructuralPartialEq for ReduceKind