Comment by aarongeisler

I've had to wrestle with float precision issues writing physics code for games. Single precision floats do not behave well when you are doing calculations far from the origin (especially when doing subtractions). I ended up having to adopt double precision floats to track the camera and entity positions in my game with kilometer-scale environments. I've also had issues with floats when doing repeat transformations on a rotation basis without periodic orthonormalization.