Struct XmxTileConfig 

pub struct XmxTileConfig {
    pub m: u32,
    pub n: u32,
    pub k: u32,
}

Tile dimensions for XMX GEMM.

Intel’s XMX engines support the following sizes on Xe-HPC (Ponte Vecchio):

• FP16 / BF16 input, FP32 accumulation: 8 × 16, 8 × 32
• INT8 / INT4 input, INT32 accumulation: 8 × 32

On Arc (Alchemist) and later, additional sizes are available.

Fields

m: u32

Rows of the A/C matrix tile.

n: u32

Columns of the B/C matrix tile.

k: u32

Inner dimension (columns of A, rows of B).

Implementations

impl XmxTileConfig

pub const XE_HPC_FP16: Self

Default XMX tile size for FP16 input targeting Xe-HPC.

pub const XE_DEFAULT: Self

XMX tile size for FP32 GEMM fallback (regular compute path).

pub fn accum_elements(&self) -> u32

Returns the number of accumulator elements per tile.

Trait Implementations

impl Clone for XmxTileConfig

fn clone(&self) -> XmxTileConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for XmxTileConfig

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

Formats the value using the given formatter.

impl Default for XmxTileConfig

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for XmxTileConfig

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

impl Eq for XmxTileConfig

impl StructuralPartialEq for XmxTileConfig