An experimental single-threaded skip list prototype, which is a unrolled-skip-list, optimized for ordered batched operations (affect how we search the next item).
Won't do any other low level optimizations, will just see how good this optimization is
Given
- MAX_HEIGHT = 32 or 64
- NODE_SIZE = 256
- 1024 key/value per inserts, ordered per batch
It is reaching a rate of 1.6-1.8M/s