Sometimes when working with C# you discover some hidden gems. Some of them very useful, other ones a little bit harder to find a good way to benefit from their functionality. One of those hidden gems that I discovered some days ago is the
The Half type got introduced in .NET 5 as a new floating point type. It extends support for the
IEEE 754 specification. In this specification many floating point types are defined, including:
binary128. You probably are already familiar with
binary32 (equivalent to
float in C#) and
binary64 (equivalent to
double in C#). With the new
binary16 is added to the list of supported types.
From the documentation:
The Half value type represents a half-precision 16-bit number with values ranging from negative 65,504 to positive 65,504, as well as positive or negative zero, PositiveInfinity, NegativeInfinity, and not a number (NaN).
The Half type is useful in cases when less precision is required. Be aware that you cannot do any arithmetic operations on the Half type. Its primary use case is to use it as an interchange type in for example Machine Learning workloads .
If you want to use it in calculations, you have to convert it first to a float or double:
Remark: Any Half value, because Half uses only 16 bits, can be represented as a float/double without loss of precision. However, the inverse is not true. Some precision may be lost when going from float/double to Half.