For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).
Mojo function
smem_tile_layout
def smem_tile_layout[k_tile_size: Int, block_rows: Int, block_cols: Int]() -> Layout
Builds the shared memory layout for a matrix tile used by the warp-specialized kernel.
The layout tiles block_rows x block_cols shared memory into horizontally
repeated block_rows x k_tile_size blocks, where block_cols must be a
multiple of k_tile_size.
Parameters:
- โk_tile_size (
Int): Width of a single K tile that forms the base block layout column extent. - โblock_rows (
Int): Number of rows in the shared memory tile. - โblock_cols (
Int): Number of columns in the shared memory tile, must be a multiple ofk_tile_size.
Returns:
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!