There are stones still unturned. For typical unicode table lookup's you'd need batches for all characters of a word. It would be much faster to lookup eg 16 cha

There are stones still unturned. For typical unicode table lookup's you'd need batches for all characters of a word. It would be much faster to lookup eg 16 chars at once, to optimize cache misses. But when I tested this even Eytzinger was too slow.