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

find_K_alignment_upto_16B

def find_K_alignment_upto_16B(row_bytes_arg: Int) -> Int

Find alignment among 1B, 2B, 4B, 16B based on the row's bytes.

This function determines the largest power-of-2 alignment (up to 16 bytes) that evenly divides the given row size. This is used to determine the optimal vector size for cp.async operations when K dimension alignment doesn't meet TMA requirements.

Args:

row_bytes_arg (Int): Number of bytes in a row (K * sizeof(element)).

Returns:

Int: Alignment in bytes (1, 2, 4, 8, or 16).