Precise representations of constants specified in hexadecimal notation to ensure an accurate constant is generated across compilers and on different hosts. The compiler generates a warning message when the mantissa is more precise than the host floating point format. The compiler generates an error message if the exponent is too wide for the host float format.
Examples:
0x2f.3a2p3
0xEp1f
0x1.8p0L
The standard library supports printing values of type float in this format using the " %a" and " %A" specifiers.