Conversions Between Floats, Doubles, and Long Doubles

When a long double is converted to a double or float, or when a double is converted to a float, the original value is rounded to the nearest representable value as the new type. If the original value is equally close to two distinct representable values, then the value chosen is the one with the least significant bit equal to zero.

Conversions between floating-point types involve a change in the exponent, as well as the mantissa. It is possible for such a conversion to overflow.