> A simple way to improve this by 2x is in `lptabz` which is to store some of > the hash code along with the index in the "hash-table-like part". Then if the > hash table part uses linear probing this reduces to just 2 loads. This is 2x > more than the 1 load "unordered" variant, but you get compactness & > insertion-order in exchange (as well as optimal iteration over elements > performance).
Is there any chance that `lptabz` becomes `Table` implementation is the std lib?
