Floating-point formats
XL C supports
the following binary floating-point formats:
- 32-bit single precision, with an approximate absolute normalized range of 0 and 10-38 to 1038 and with a precision of about 7 decimal digits
- 64-bit double precision, with an approximate absolute normalized range of 0 and 10-308 to 10308 and with a precision of about 16 decimal digits
- 128-bit extended precision, with slightly greater range than double-precision values, and with a precision of about 32 decimal digits
Note that the long double type might represent either double-precision or extended-precision values, depending on the setting of the -qldbl128 compiler option.
Beginning in V9.0, on selected hardware and operating
system levels, the compiler also supports the following decimal floating-point
formats:
- 32-bit single precision, with an approximate range of 10-101 to 1090 and precision of 7 decimal digits
- 64-bit double precision, with an approximate range of 10-398 to 10369 and precision of 16 decimal digits
- 128-bit extended precision, with an approximate range of 10-6176 to 106111, and with a precision of 34 decimal digits
The 128-bit extended precision format of XL C is different from the binary128 formats that are suggested by the IEEE standard. The IEEE standard suggests that extended formats use more bits in the exponent for greater range and the fraction for higher precision.
It is possible that special numbers, such as NaN, infinity, and negative zero, cannot be represented by the 128-bit extended precision values. Arithmetic operations do not necessarily propagate these numbers in extended precision.
