So in this video you claim making a separate reference to the transform is somehow an optimization, even an important one as far as iPhone is concerned. My question is: how is this doing anything but wasting 4 bytes of memory to hold the reference? What am I not getting here?