Mojo struct
TiledMatmul
struct TiledMatmul[config: KernelConfig, transpose_b: Bool, b_packed: Bool, elementwise_epilogue_enabled: Bool, kernel_id: InnerKernelID, a_type: DType, a_layout: TensorLayout, a_origin: ImmutOrigin, b_type: DType, b_layout: TensorLayout, b_origin: ImmutOrigin, c_type: DType, c_layout: TensorLayout, c_origin: MutOrigin, algorithm: InnerMatmulKernel, ElementwiseEpilogueFnType: ImplicitlyCopyable & def(GemmShape, GemmShape) -> None]
Tiled matmul implementation integrating packing, inner loop and tile partitions.
TODO: add tag based implementation dispatch. TODO: add fusion hooks.
Fields
- alg (
algorithm): - c (
TileTensor[c_type, c_layout, c_origin]): - a (
TileTensor[a_type, a_layout, a_origin]): - b (
TileTensor[b_type, b_layout, b_origin]): - tile_n_k (
IndexList[2]): - global_tile_offset (
GemmShape): - global_tile_shape (
GemmShape): - b_tile_generator (
BTileGenerator[config, a_type, b_type, c_type, b_layout, transpose_b, b_packed, b_origin]): - elementwise_epilogue_fn (
ElementwiseEpilogueFnType):
Implemented traits
AnyType,
Copyable,
ImplicitlyCopyable,
ImplicitlyDestructible,
Movable
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!