Enum WgmmaShape 

pub enum WgmmaShape {
    M64N64K16,
    M64N128K16,
    M64N256K16,
    M64N64K32,
    M64N128K32,
    M64N256K32,
}

Subset of WGMMA matmul shapes commonly exercised by attention / matmul kernels. The numeric tuple is (M, N, K) (row × col × inner).

Variants

M64N64K16

m64n64k16 — most-common fp16 variant.

M64N128K16

m64n128k16 — wider tile, same fp16.

M64N256K16

m64n256k16 — full warpgroup output tile.

M64N64K32

m64n64k32 — fp8 (e4m3/e5m2) variant.

M64N128K32

m64n128k32 — fp8 wider.

M64N256K32

m64n256k32 — fp8 full.

Implementations

impl WgmmaShape

pub fn dims(self) -> (u32, u32, u32)

(M, N, K) decomposition.

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

Macro name (matches atomr_hopper.cuh).

Trait Implementations

impl Clone for WgmmaShape

fn clone(&self) -> WgmmaShape

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WgmmaShape

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

Formats the value using the given formatter.

impl PartialEq for WgmmaShape

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

impl Eq for WgmmaShape

impl StructuralPartialEq for WgmmaShape