Memory based == is not unreasonable for floats, once in exact representation domain. Of course, if operating in approximated domain, especially involving any of existing math libraries, the float compare should be done with corresp. functions/macro and be epsilon based.