Struct ContiguousTilingLayout

pub struct ContiguousTilingLayout<T: TilingOrder> { /* private fields */ }

Each tile is stored contiguously in shared memory. Global memory loads may require remapping to match this layout.

Implementations

impl<T: TilingOrder> ContiguousTilingLayout<T>

pub fn to_x_y<S: StageConfig>(nth: u32, ident: Ident, config: S) -> (u32, u32)

Converts a tile index in the stage to its (x,y) position

pub fn __expand_to_x_y<S: StageConfig>( scope: &mut Scope, nth: <u32 as CubeType>::ExpandType, ident: Ident, config: S, ) -> <(u32, u32) as CubeType>::ExpandType

Trait Implementations

impl<MP: MatmulPrecision, TO: TilingOrder> AsyncLoadingJob<MP, ContiguousTilingLayout<TO>> for AsyncFullCyclicJob

fn execute_task<CM: CopyMechanism<MP::ES>, G: GlobalConfig>( this: &mut Self, task_id: u32, tensor_reader: &TensorReader<MP::EI>, stage: &mut StageMemory<MP::ES, ContiguousTilingLayout<TO>>, mechanism: &CM, config: G, )

Execute the task_idth loading task

fn task_count(this: &Self) -> u32

Get the number of tasks

fn __expand_execute_task<CM: CopyMechanism<MP::ES>, G: GlobalConfig>( scope: &mut Scope, this: <Self as CubeType>::ExpandType, task_id: <u32 as CubeType>::ExpandType, tensor_reader: <TensorReader<MP::EI> as CubeType>::ExpandType, stage: <StageMemory<MP::ES, ContiguousTilingLayout<TO>> as CubeType>::ExpandType, mechanism: <CM as CubeType>::ExpandType, config: G, ) -> <() as CubeType>::ExpandType

fn __expand_task_count( scope: &mut Scope, this: <Self as CubeType>::ExpandType, ) -> u32

impl<T: Clone + TilingOrder> Clone for ContiguousTilingLayout<T>

fn clone(&self) -> ContiguousTilingLayout<T>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<MP: MatmulPrecision, TO: TilingOrder> LoadingJob<MP, ContiguousTilingLayout<TO>> for SyncFullCyclicJob

fn execute_task<G: GlobalConfig>( this: &mut Self, task_id: u32, tensor_reader: &TensorReader<MP::EI>, stage: &mut StageMemory<MP::ES, ContiguousTilingLayout<TO>>, quantization: &CubeOption<Quantization<MP>>, config: G, )

Execute the task_idth loading task

fn task_count(this: &Self) -> u32

Get the number of tasks

fn __expand_execute_task<G: GlobalConfig>( scope: &mut Scope, this: <Self as CubeType>::ExpandType, task_id: u32, tensor_reader: <TensorReader<MP::EI> as CubeType>::ExpandType, stage: <StageMemory<MP::ES, ContiguousTilingLayout<TO>> as CubeType>::ExpandType, quantization: <CubeOption<Quantization<MP>> as CubeType>::ExpandType, config: G, ) -> <() as CubeType>::ExpandType

fn __expand_task_count( scope: &mut Scope, this: <Self as CubeType>::ExpandType, ) -> u32

impl<MP: MatmulPrecision, TO: TilingOrder> LoadingJob<MP, ContiguousTilingLayout<TO>> for SyncFullTilewiseJob

fn execute_task<G: GlobalConfig>( this: &mut Self, task_id: u32, tensor_reader: &TensorReader<MP::EI>, stage: &mut StageMemory<MP::ES, ContiguousTilingLayout<TO>>, quantization: &CubeOption<Quantization<MP>>, config: G, )

Execute the task_idth loading task

fn task_count(this: &Self) -> u32

Get the number of tasks

fn __expand_execute_task<G: GlobalConfig>( scope: &mut Scope, this: <Self as CubeType>::ExpandType, task_id: u32, tensor_reader: <TensorReader<MP::EI> as CubeType>::ExpandType, stage: <StageMemory<MP::ES, ContiguousTilingLayout<TO>> as CubeType>::ExpandType, quantization: <CubeOption<Quantization<MP>> as CubeType>::ExpandType, config: G, ) -> <() as CubeType>::ExpandType

fn __expand_task_count( scope: &mut Scope, this: <Self as CubeType>::ExpandType, ) -> u32

impl<MP: MatmulPrecision, TO: TilingOrder> LoadingJob<MP, ContiguousTilingLayout<TO>> for SyncPartialCyclicJob

fn execute_task<G: GlobalConfig>( this: &mut Self, task_id: u32, tensor_reader: &TensorReader<MP::EI>, stage: &mut StageMemory<MP::ES, ContiguousTilingLayout<TO>>, quantization: &CubeOption<Quantization<MP>>, config: G, )

Execute the task_idth loading task

fn task_count(this: &Self) -> u32

Get the number of tasks

fn __expand_execute_task<G: GlobalConfig>( scope: &mut Scope, this: <Self as CubeType>::ExpandType, task_id: u32, tensor_reader: <TensorReader<MP::EI> as CubeType>::ExpandType, stage: <StageMemory<MP::ES, ContiguousTilingLayout<TO>> as CubeType>::ExpandType, quantization: <CubeOption<Quantization<MP>> as CubeType>::ExpandType, config: G, ) -> <() as CubeType>::ExpandType

fn __expand_task_count( scope: &mut Scope, this: <Self as CubeType>::ExpandType, ) -> u32

impl<MP: MatmulPrecision, TO: TilingOrder> LoadingJob<MP, ContiguousTilingLayout<TO>> for SyncPartialTilewiseJob

fn execute_task<G: GlobalConfig>( this: &mut Self, task_id: u32, tensor_reader: &TensorReader<MP::EI>, stage: &mut StageMemory<MP::ES, ContiguousTilingLayout<TO>>, quantization: &CubeOption<Quantization<MP>>, config: G, )

Execute the task_idth loading task

fn task_count(this: &Self) -> u32

Get the number of tasks

fn __expand_execute_task<G: GlobalConfig>( scope: &mut Scope, this: <Self as CubeType>::ExpandType, task_id: u32, tensor_reader: <TensorReader<MP::EI> as CubeType>::ExpandType, stage: <StageMemory<MP::ES, ContiguousTilingLayout<TO>> as CubeType>::ExpandType, quantization: <CubeOption<Quantization<MP>> as CubeType>::ExpandType, config: G, ) -> <() as CubeType>::ExpandType

fn __expand_task_count( scope: &mut Scope, this: <Self as CubeType>::ExpandType, ) -> u32

impl<TO: TilingOrder> TilingLayout for ContiguousTilingLayout<TO>

fn get_tile<ES: Numeric, S: StageConfig>( stage_memory: &StageMemory<ES, Self>, row: u32, col: u32, buffer_index: u32, ident: Ident, config: S, ) -> Tile<ES>

Returns the tile at shared memory coordinates

fn __expand_get_tile<ES: Numeric, S: StageConfig>( scope: &mut Scope, stage_memory: <StageMemory<ES, Self> as CubeType>::ExpandType, row: <u32 as CubeType>::ExpandType, col: <u32 as CubeType>::ExpandType, buffer_index: u32, ident: Ident, config: S, ) -> <Tile<ES> as CubeType>::ExpandType

impl<T: Copy + TilingOrder> Copy for ContiguousTilingLayout<T>