Struct SimpleBarrierAlgorithm 

pub struct SimpleBarrierAlgorithm<TMM, L: AsyncFullLoadingStrategy> {
    pub _tmm: PhantomData<TMM>,
    pub _l: PhantomData<L>,
}

Plane accelerated single stage matmul with async barrier loading

Fields

_tmm: PhantomData<TMM>

_l: PhantomData<L>

Trait Implementations

impl<TMM, L> Algorithm for SimpleBarrierAlgorithm<TMM, L>

where TMM: TileMatmulFamily<LhsTile = Strided, RhsTile = Strided, AccTile = Filled, OutTile = Strided>, L: AsyncFullLoadingStrategy,

type SelectionArgs = ()

type TileMatmul = TMM

type StageMatmul = PlaneMatmulFamily<<SimpleBarrierAlgorithm<TMM, L> as Algorithm>::TileMatmul, StridedStageFamily, StridedStageFamily, FilledStageFamily>

type GlobalMatmul = SimpleBarrierMatmulFamily<<SimpleBarrierAlgorithm<TMM, L> as Algorithm>::StageMatmul, L, L, PlaneWriterFamily>

type BatchMatmul = PartitionedBatchMatmulFamily<<SimpleBarrierAlgorithm<TMM, L> as Algorithm>::GlobalMatmul, RowMajorGlobalPartitionMatmul>

fn selection<R: Runtime>( client: &ComputeClient<R::Server>, problem: &MatmulProblem, plane_dim: u32, _line_sizes: &MatmulLineSizes, elems: MatmulElems, _args: &Self::SelectionArgs, ) -> Result<MatmulSelection, MatmulSetupError>

fn setup<MP: MatmulPrecision, R: Runtime>( client: &ComputeClient<R::Server>, problem: &MatmulProblem, selection: &MatmulSelection, line_sizes: &MatmulLineSizes, ) -> Result<<Self::BatchMatmul as BatchMatmulFamily>::Config, MatmulSetupError>

fn filter_line_sizes( available_line_sizes: AvailableLineSizes, ) -> AvailableLineSizes

fn select_plane_dim<R: Runtime>(client: &ComputeClient<R::Server>) -> u32