Basically, i have a set of mxnxk vector that i’ll combine over the n and k directions to shrink into a m long array. I can do this in a kernel with no reduction (dimensions n and k are relatively small, k=3, n~=20) but i was wondering if there was a good way to something similar with thrust. I looked at a zip_iterator, but being that n is both variable and pretty big, this doesn’t seem to be the way to go.