Enum AmdWmma 

pub enum AmdWmma {
    Rdna3,
    Rdna4,
}

Which generation of AMD’s WMMA instructions to emit. RDNA4 renamed the builtins and dropped RDNA3’s duplication of the A/B fragments across lane halves, halving them to 8 elements, so the two aren’t interchangeable.

Variants

Rdna3

Rdna4

Implementations

impl AmdWmma

pub fn frag_ab_elems(&self, k: usize) -> usize

Elements of A/B each lane holds. RDNA3 gives every lane the whole k range and duplicates it across lanes 0-15 / 16-31; RDNA4 splits k between the two halves instead.

Trait Implementations

impl Clone for AmdWmma

fn clone(&self) -> AmdWmma

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for AmdWmma

impl Debug for AmdWmma

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

Formats the value using the given formatter.

impl Eq for AmdWmma

impl PartialEq for AmdWmma

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for AmdWmma