Mostly about effectivity vs. efficiency in terms of game programming.
Just fuck it
Yeah, just fuck it. I'll simply convert to float every time I need to and that's it. Will perhaps try some other variant later, but I don't want to sacrifice all the comfort of floating point operations (not so mention the mess I'd need to do for physics calculation). Sure, this is expensive, but used seldomly and at only the necessary positions, it bears more flexibility for everything else. There are more alghorithmic performance holes fixable compared to the base speed required for iteration.