this greatly improves backward mapping performance, with improvements
ranging from 50% to 400%. this also allows for the exact accounting of
the binary size overhead.