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
get_producer_warp_thread_layout
def get_producer_warp_thread_layout[k_tile_size: Int, simd_width: Int, block_rows: Int, block_cols: Int]() -> Layout
Computes the thread-to-element layout used by a producer warp when loading a tile.
Arranges the warp's threads into inner blocks that each cover one
k_tile_size-wide row, then tiles those inner blocks across the warp to
cover the full block_rows x block_cols tile.
Parameters:
- k_tile_size (
Int): Width of a single K tile loaded by one row of inner blocks. - simd_width (
Int): SIMD vector width of the matrix element type, used to size inner-block columns. - block_rows (
Int): Number of rows in the block tile to load. - block_cols (
Int): Number of columns in the block tile to load.
Returns:
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!