Enum GemmLayout 

pub enum GemmLayout {
    Nn,
    Nt,
    Tn,
    Tt,
}

GEMM operand layout (BLAS transpose convention). The int8 tensor cores are fastest in Tn.

Variants

Nn

A row-major, B row-major — the layout most callers hand you. Universal; tensor cores must rearrange.

Nt

A row-major, B col-major.

Tn

A col-major (= A^T), B row-major — the mma.sync.s8 NATIVE layout. Fastest on the tensor cores.

Tt

A col-major, B col-major.

Implementations

impl GemmLayout

pub fn is_tensor_core_native(&self) -> bool

Whether this layout is the tensor cores’ native contraction-contiguous layout (zero rearrange cost).

pub fn all() -> [GemmLayout; 4]

All four layouts — the layout axis of the frontier.

Trait Implementations

impl Clone for GemmLayout

fn clone(&self) -> GemmLayout

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GemmLayout

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

Formats the value using the given formatter.

impl PartialEq for GemmLayout

fn eq(&self, other: &GemmLayout) -> 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 GemmLayout

impl Eq for GemmLayout

impl StructuralPartialEq for GemmLayout