Unfortunately it seems to be slightly more than twice as slow. On the plus side, that is some justification for the existing optimizations.