ten_byte_long_double

Changes the size of the long double to 80-bits (10 bytes) allowing the full precision and range of extended precision or temporary real format used in x86 floating point registers.

Syntax

#pragma ten_byte_long_double on | off | reset

Targets
Intel x86
Remarks

The range is approximately 3.632^10-4932 to 1.190^10+4932.

Note there is no indication in the object code about what long double size was used to create it, thus the linker cannot warn on mismatches. To ensure a match, all libraries and programs exchanging long double must be rebuilt.

The default setting is off, where both long double and double are both 64-bits.