Yeah the optimization is similar to the realization that array-of-structs can be the faster memory layout if you use all/most of the members of the struct. So you organize memory by archetype rather than keeping the components separately.