The IEEE 754 specification defines many floating point types, including: binary16, binary32, binary64 and binary128. Most developers are familiar with binary32 (equivalent to float in C#) and binary64 (equivalent to double in C#). They provide a standard format to represent a wide range of values with a precision acceptable for many applications. .NET has always had float and double and with .NET 5 Preview 7, we’ve added a new Half type (equivalent to binary16)!
A Half is a binary floating-point number that occupies 16 bits. With half the number of bits as float, a Half number can represent values in the range ±65504. More formally, the Half type is defined as a base-2 16-bit interchange format meant to support the exchange of floating-point data between implementations. One of the primary use cases of the Half type is to save on storage space where the computed result does not need to be stored with full precision. Many computation workloads already take advantage of the Half type: machine learning, graphics cards, the latest processors, native SIMD libraries etc. With the new Half type, we expect to unlock many applications in these workloads.
Let’s explore the Half type:The 16 bits in the Half type are split into:
Sign bit: 1 bitExponent bits: 5 bitsSignificand bits: 10 bits (with 1 implicit bit that is not stored)Despite that fact that the significand is made up of 10 bits, the total precision is really 11 bits. The format is assumed to have an implicit leading bit of value 1 (unless the exponent field is all zeros, in which case the leading bit has a value 0). To represent the number 1 in the Half format, we’d use the bits:
0 01111 0000000000 = 1The leading bit (our sign bit) is 0, indicating a positive number. The exponent bits are 01111, or 15 in decimal. However, the exponent bits don’t represent the exponent directly. Instead, an exponent bias is defined that lets the format represent both positive and negative exponents. For the Half type, that exponent bias is 15. The true exponent is derived by subtracting 15 from the stored exponent. Therefore, 01111 represents the exponent e = 01111 (in binary) - 15 (the exponent bias) = 0. The significand is 0000000000, which can be interpreted as the number .significand(in base 2) in base 2, 0 in our case. If, for example, the significand was 0000011010 (26 in decimal), we can divide its decimal value 26 by the number of values representable in 10 bits (1