Where can I get the source code for the optimized fold method?

Question

C++ / C#

Where can I see a quick implementation of the convolution / unfolding methods (two-dimensional and multilayer) under c++/c? I found a version with intrinsics, but I'm not sure how much their addition really speeds up the work with a naive implementation. If there are no specific sources, then maybe there is a library containing these functions? (preferably not very large and without using cuda, because the final program needs to be able to run on someone else's PC.)