Struct OrderedTilingOrder

pub struct OrderedTilingOrder {}

Tiles are laid out in column-major order across a fixed number of rows, with all tiles from those rows placed contiguously side by side.

The grouping should match the set of tiles processed by a warp, so warp-local tile memory remains contiguous.

This layout ensures that for Lhs data, all tiles needed for a given k iteration are stored contiguously, before moving to the next iteration.

Note: use only for Lhs

Example tile indices for 4 rows grouped 2 at a time:

┌───┬───┐
│ 0 │ 2 │
├───┼───┤
│ 1 │ 3 │
├───┼───┤
│ 4 │ 6 │
├───┼───┤
│ 5 │ 7 │
└───┴───┘

Trait Implementations

impl Clone for OrderedTilingOrder

fn clone(&self) -> OrderedTilingOrder

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl CubeType for OrderedTilingOrder

type ExpandType = OrderedTilingOrderExpand

fn into_mut(scope: &mut Scope, expand: Self::ExpandType) -> Self::ExpandType

Wrapper around the init method, necessary to type inference.

impl TilingOrder for OrderedTilingOrder

fn to_row_col<C: StageConfig>( nth: u32, tile_count_rows: u32, tile_count_cols: u32, ident: Ident, config: C, ) -> (u32, u32)

Returns the coordinates (row, col) of the tile

fn to_nth_tile<C: StageConfig>( row: u32, col: u32, tile_count_rows: u32, tile_count_cols: u32, ident: Ident, config: C, ) -> u32

Given the coordinates (row, col) of the tile, returns its index in shared memory

fn to_enum() -> TilingOrderEnum

Return the trait value as enum

fn __expand_to_row_col<C: StageConfig>( scope: &mut Scope, nth: <u32 as CubeType>::ExpandType, tile_count_rows: u32, tile_count_cols: u32, ident: Ident, config: C, ) -> <(u32, u32) as CubeType>::ExpandType

fn __expand_to_nth_tile<C: StageConfig>( scope: &mut Scope, row: <u32 as CubeType>::ExpandType, col: <u32 as CubeType>::ExpandType, tile_count_rows: u32, tile_count_cols: u32, ident: Ident, config: C, ) -> <u32 as CubeType>::ExpandType

fn __expand_to_enum(scope: &mut Scope) -> TilingOrderEnum

impl Copy for OrderedTilingOrder