Modules§
Structs§
- ColMajor
Span Matmul - Iterates on global matmuls in a col major fashion
- ColMajor
Span Matmul Expand - Natural
Dispatch - Operates on data further along the m dimension as
cube_pos_xincreases, and further along the n dimension ascube_pos_yincreases. - RowMajor
Span Matmul - Iterates on global matmuls in a row major fashion
- RowMajor
Span Matmul Expand - Span
- Area of a tensor a cube is responsible of performing matmul Similar to the concept of tensor slice, but specialized for matmul constraints
- SpanDim
- Span information in one dimension
- Span
DimExpand - Span
Expand - Swizzle
Natural Dispatch - Processes data in a swizzled pattern, prioritizing cubes along the x-axis first.
- Swizzle
Span Matmul - Iterates on global matmuls following the swizzle algorithm
- Swizzle
Span Matmul Expand - Swizzle
Transposed Dispatch - Processes data in a swizzled pattern, prioritizing cubes along the y-axis first.
- Transposed
Dispatch - Operates on data further along the m dimension as
cube_pos_xincreases, and further along the n dimension ascube_pos_yincreases.
Traits§
- Batch
Config - Configuration for the batch matmul level.
- Batch
Matmul - Provides matrix multiplication operations at the batch level.
- Batch
Matmul Family - A family of matmuls working with any precision.
- Cube
Count Dispatch - Cube
Dispatch - Distributes cube instances across the tensor, assigning each to compute data in distinct regions.
- Span
Matmul - Iterates on several global matmul across a span