New posts in Half Precision Float

Half-precision floating-point arithmetic on Intel chips

Why is there no 2-byte float and does an implementation already exist?