The code reveals floating point numbers for what they truly are: just fancy structs with a sign, exponent, and mantissa wearing a trench coat. The programmer manually constructs a float by setting each field, then casts it back to a float with that sketchy pointer manipulation. And of course, there's the mandatory comment warning you to never actually do this in production because bitfield layout will betray you faster than a coworker who "fixed" your code. Typical C behavior - giving you enough rope to not only hang yourself but the entire dev team.