IEEE 754 half-precision format (library)

 x448/float16 package provides IEEE 754 half-precision
 floating-point format (binary16) with IEEE 754 default
 rounding for conversions. IEEE 754-2008 refers to this
 16-bit floating-point format as binary16.
 .
 IEEE 754 default rounding ("Round-to-Nearest
 RoundTiesToEven") is considered the most accurate and
 statistically unbiased estimate of the true result.