Struct AsyncPartialMaximizeSliceLengthLoading

pub struct AsyncPartialMaximizeSliceLengthLoading {}

Executes one memcpy_async call per contiguous slice. The goal is to reduce the total number of memcpy_async calls, though it may result in idle threads.

Trait Implementations

impl AsyncPartialLoadingStrategy for AsyncPartialMaximizeSliceLengthLoading

type TilingLayout = StridedTilingLayout

The layout describing how data is tiled across the stage.

type Job<MP: MatmulPrecision> = AsyncPartialMaximizeSliceLengthJob

The [LoadingJob] for this strategy.

fn new_job<MP: MatmulPrecision, G: GlobalConfig>( stage_index: u32, input_ident: InputIdent, config: G, ) -> AsyncPartialMaximizeSliceLengthJob

Returns the job with preliminary calculations done.

fn barrier_level() -> BarrierLevel

The barrier level at which the copy mechanism works

fn __expand_new_job<MP: MatmulPrecision, G: GlobalConfig>( scope: &mut Scope, stage_index: u32, input_ident: InputIdent, config: G, ) -> <AsyncPartialMaximizeSliceLengthJob as CubeType>::ExpandType

fn __expand_barrier_level( scope: &mut Scope, ) -> <BarrierLevel as CubeType>::ExpandType

impl Clone for AsyncPartialMaximizeSliceLengthLoading

fn clone(&self) -> AsyncPartialMaximizeSliceLengthLoading

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl CubeType for AsyncPartialMaximizeSliceLengthLoading

type ExpandType = AsyncPartialMaximizeSliceLengthLoadingExpand

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

Wrapper around the init method, necessary to type inference.

impl LoadingValidation for AsyncPartialMaximizeSliceLengthLoading

fn check<C: GlobalConfig>( _config: &C, _ident: Ident, ) -> Result<(), InvalidConfigError>

Verify that configs are valid for a loader, otherwise return an error stating why

impl Copy for AsyncPartialMaximizeSliceLengthLoading