IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /get-started.md). For the complete documentation index, see llms.txt.
Skip to main content
For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Mojo struct

BlockScaledTilePayload

struct BlockScaledTilePayload[a_type: DType, b_type: DType, sfa_type: DType, sfb_type: DType, a_shape: IndexList[Int(2)], b_shape: IndexList[Int(2)], sfa_shape: IndexList[Int(2)], sfb_shape: IndexList[Int(2)], num_pipeline_stages: Int]

Tile payload for block-scaled matmul (A, B, SFA, SFB tiles).

Parameters

  • a_type (DType): Element type of the A operand tiles.
  • b_type (DType): Element type of the B operand tiles.
  • sfa_type (DType): Element type of the A scale factor (SFA) tiles.
  • sfb_type (DType): Element type of the B scale factor (SFB) tiles.
  • a_shape (IndexList[Int(2)]): A tile dimensions as (BM, BK).
  • b_shape (IndexList[Int(2)]): B tile dimensions as (BN, BK).
  • sfa_shape (IndexList[Int(2)]): Dimensions of the A scale factor (SFA) tile.
  • sfb_shape (IndexList[Int(2)]): Dimensions of the B scale factor (SFB) tile.
  • num_pipeline_stages (Int): Number of pipeline buffer stages.

Fields

  • a_tiles (BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].ATileArray):
  • b_tiles (BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].BTileArray):
  • sfa_tiles (BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].SFATileArray):
  • sfb_tiles (BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].SFBTileArray):

Implemented traits

AnyType, Copyable, Deinitable, ImplicitlyCopyable, Movable, RegisterPassable, TilePayload, TrivialRegisterPassable

comptime members

ATile

comptime ATile = BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].ATileArray.Tile

ATileArray

comptime ATileArray = SMemTileArray2D[a_type, a_shape[Int(0)], a_shape[Int(1)], num_pipeline_stages]

BTile

comptime BTile = BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].BTileArray.Tile

BTileArray

comptime BTileArray = SMemTileArray2D[b_type, b_shape[Int(0)], b_shape[Int(1)], num_pipeline_stages]

sfa_layout

comptime sfa_layout = Layout(Coord(Coord(ComptimeInt(), ComptimeInt()), Coord(Coord(ComptimeInt(), ComptimeInt()), ComptimeInt())), Coord(Coord(ComptimeInt(), ComptimeInt()), Coord(Coord(ComptimeInt(), ComptimeInt()), ComptimeInt())))

SFATile

comptime SFATile = BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].SFATileArray.Tile

SFATileArray

comptime SFATileArray = SMemTileArrayWithLayout[sfa_type, BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].sfa_layout, num_pipeline_stages]

sfb_layout

comptime sfb_layout = Layout(Coord(Coord(ComptimeInt(), ComptimeInt()), Coord(Coord(ComptimeInt(), ComptimeInt()), ComptimeInt())), Coord(Coord(ComptimeInt(), ComptimeInt()), Coord(Coord(ComptimeInt(), ComptimeInt()), ComptimeInt())))

SFBTile

comptime SFBTile = BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].SFBTileArray.Tile

SFBTileArray

comptime SFBTileArray = SMemTileArrayWithLayout[sfb_type, BlockScaledTilePayload[a_type, b_type, sfa_type, sfb_type, a_shape, b_shape, sfa_shape, sfb_shape, num_pipeline_stages].sfb_layout, num_pipeline_stages]

Methods

__init__

def __init__(a_tiles: SMemTileArray2D[a_type, a_shape[Int(0)], a_shape[Int(1)], num_pipeline_stages], b_tiles: SMemTileArray2D[b_type, b_shape[Int(0)], b_shape[Int(1)], num_pipeline_stages], sfa_tiles: SMemTileArrayWithLayout[sfa_type, Self.sfa_layout, num_pipeline_stages], sfb_tiles: SMemTileArrayWithLayout[sfb_type, Self.sfb_layout, num_pipeline_stages]) -> Self

get_tile

def get_tile[k_group_size: Int](self, stage: UInt32, k_idx: Int) -> Tuple[TileTensor[a_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[b_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[sfa_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[sfb_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED]]

Get A, B, SFA, SFB tiles at the specified stage and k-group index.

Parameters:

  • k_group_size (Int): Number of K-group tiles stored per pipeline stage.

Args:

  • stage (UInt32): Pipeline stage index into the circular buffer.
  • k_idx (Int): K-group index within the current pipeline stage.

Returns:

Tuple[TileTensor[a_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[b_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[sfa_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED], TileTensor[sfb_type, Layout[*?, *?], MutAnyOrigin, address_space=AddressSpace.SHARED]]

get_a_tile

def get_a_tile[k_group_size: Int](self, stage: UInt32, k_idx: Int) -> Self.ATile

Get A tile at the specified stage and k-group index.

Returns:

Self.ATile

get_b_tile

def get_b_tile[k_group_size: Int](self, stage: UInt32, k_idx: Int) -> Self.BTile

Get B tile at the specified stage and k-group index.

Returns:

Self.BTile

get_sfa_tile

def get_sfa_tile[k_group_size: Int](self, stage: UInt32, k_idx: Int) -> Self.SFATile

Get SFA tile at the specified stage and k-group index.

Returns:

Self.SFATile

get_sfb_tile

def get_sfb_tile[k_group_size: Int](self, stage: UInt32, k_idx: Int) -> Self.SFBTile

Get SFB tile at the specified stage and k-group index.

Returns:

Self.SFBTile