Return the float as a 2s compliment int. Just to be used to compare floats
to each other or against positive float-bit-constants (like 0). This does
not return the int equivalent of the float, just something cheaper for
compares-only.
Convert a sign-bit int (i.e. float interpreted as int) into a 2s compliement
int. This also converts -0 (0x80000000) to 0. Doing this to a float allows
it to be compared using normal C operators (<, <=, etc.)