[QST] Universal convolution supports for sm70/80 using Cute?

What is your question?

I've read the example 59, it seems there is a easy and elegant way to assemble a conv kernel by using Cute, but the conv params are assumed to be known at complie time, and if I made these params determined during runtime, the tiling cannot work correctly when shape<n, p, q> cannot divisable by the tiler(E.g., npq are Shape<128, 14, 14>, while tiler is Shape<_128>(Actually, it should be qpn Shape<14, 14, 128> since the tiler always start tiling from the first dim)).

I would like to assemble a conv kernel using Cute to handle universal padding/stride, or in other words, R/S/P/Q is determined in runtime period. It seems only sm90 conv features can handle these universal cases, but I can't apply it on my workstation. I try to migrate it to sm70/80 but the TileCopy for im2col is complicated, I can't understand the parts that linearize the npq shape and strides, and seems it's related to sm90 hardware intrinsics.

I'm greatly appreciate if there's a approach to write a universal conv kernel using Cute while I can avoid using sm90 im2col intrinsics.

Thank you so much!

Start with example 59 and the Cute convolution code, then inspect the sm90 im2col TileCopy path and the corresponding sm70/80 support. Trace how runtime R/S/P/Q values and the npq shape are linearized when dimensions are not divisible by the tiler. Done means establishing a universal-padding/stride convolution approach that works on sm70/80 without sm90 im2col intrinsics.