Agreed on multiplying the top 64 bits rather than dividing: if the hash is calculated via prime multiplications, the top bits are "more random" than the bottom

Agreed on multiplying the top 64 bits rather than dividing: if the hash is calculated via prime multiplications, the top bits are "more random" than the bottom bits. But we don't really know the hash table type. It could be doing something like hash table -> linked list -> items, in which case they might still be utilising the full 128-bit hash (I'd usually just store the lower 64 bits TBH).