Mojo function

dispatch_sm100_conv2d

dispatch_sm100_conv2d[input_layout: Layout, filter_layout: Layout, output_layout: Layout, input_type: DType, filter_type: DType, output_type: DType, filter_is_fcrs: Bool](input: LayoutTensor[input_type, input_layout, origin, address_space=address_space, element_layout=element_layout, layout_int_type=layout_int_type, linear_idx_type=linear_idx_type, masked=masked, alignment=alignment], filter: LayoutTensor[filter_type, filter_layout, origin, address_space=address_space, element_layout=element_layout, layout_int_type=layout_int_type, linear_idx_type=linear_idx_type, masked=masked, alignment=alignment], output: LayoutTensor[output_type, output_layout, origin, address_space=address_space, element_layout=element_layout, layout_int_type=layout_int_type, linear_idx_type=linear_idx_type, masked=masked, alignment=alignment], symmetric_padding: IndexList[2], ctx: DeviceContext)

Dispatch to SM100 structured conv2d with filter transpose.

This function gates the SM100 kernel import behind @parameter if on dtype, so the kernel is never compiled for unsupported dtypes.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit