I'm sorry. I guess I thought his requirements didn't preclude a destructive routine, since he opens with $r = [1,2,3,4,5,6] and wants to replace that with $r = [[1,2],[3,4],[5,6]] and both the shifter and splicer techniques in the benchmark are destructive. My money's on the c-loop for non-destructive, and if I had large tuples or frequently changing tuple lengths, I'd use splicer in spite of the speed hit shown here.