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
test_alignment_sm100_conv2d
def test_alignment_sm100_conv2d[input_type: DType, output_type: DType](in_channels: Int, out_channels: Int) -> Bool
Checks whether the input and output channel counts meet SM100 conv2d alignment requirements.
Returns True when the input activation row is 64-byte aligned and the output row is 4-byte aligned, otherwise False.
Parameters:
- input_type (
DType): Element type of the input activation tensor, used to compute the per-row byte width for the 64-byte alignment check. - output_type (
DType): Element type of the output tensor, used to compute the per-row byte width for the 4-byte alignment check.
Args:
- in_channels (
Int): Number of input channels per activation row. - out_channels (
Int): Number of output channels per output row.
Returns:
Bool: True if both alignment constraints are satisfied, False otherwise.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!