Buddy allocation is generally pretty efficient but suffers from internal fragmentation. Can slab allocation have similar code efficiency? 