FE_DOWNWARD, FE_TONEAREST, FE_TOWARDZERO, FE_UPWARD - cppreference.com (original) (raw)
| Defined in header | | | | ------------------------------------------------------------------- | | ------------- | | #define FE_DOWNWARD /*implementation defined*/ | | (since C++11) | | #define FE_TONEAREST /*implementation defined*/ | | (since C++11) | | #define FE_TOWARDZERO /*implementation defined*/ | | (since C++11) | | #define FE_UPWARD /*implementation defined*/ | | (since C++11) |
Each of these macro constants expands to a nonnegative integer constant expression, which can be used with std::fesetround and std::fegetround to indicate one of the supported floating-point rounding modes. The implementation may define additional rounding mode constants in , which should all begin with FE_ followed by at least one uppercase letter. Each macro is only defined if it is supported.
| Constant | Explanation |
|---|---|
| FE_DOWNWARD | rounding towards negative infinity |
| FE_TONEAREST | rounding towards nearest representable value |
| FE_TOWARDZERO | rounding towards zero |
| FE_UPWARD | rounding towards positive infinity |
Additional rounding modes may be supported by an implementation.
The current rounding mode affects the following:
- results of floating-point arithmetic operators outside of constant expressions
double x = 1; x / 10; // 0.09999999999999999167332731531132594682276248931884765625 // or 0.1000000000000000055511151231257827021181583404541015625
- results of standard library mathematical functions
std::sqrt(2); // 1.41421356237309492343001693370752036571502685546875 // or 1.4142135623730951454746218587388284504413604736328125
- floating-point to floating-point implicit conversion and casts
double d = 1 + std::numeric_limits::epsilon(); float f = d; // 1.00000000000000000000000 // or 1.00000011920928955078125
- string conversions such as std::strtod or std::printf
std::stof("0.1"); // 0.0999999940395355224609375 // or 0.100000001490116119384765625
- the library rounding functions std::nearbyint, std::rint, std::lrint
The current rounding mode does NOT affect the following:
- floating-point to integer implicit conversion and casts (always towards zero),
- results of floating-point arithmetic operators in expressions executed at compile time (always to nearest),
- the library functions std::round, std::lround, std::llround, std::ceil, std::floor, std::trunc.
As with any floating-point environment functionality, rounding is only guaranteed if #pragma STDC FENV_ACCESS ON is set.
Compilers that do not support the pragma may offer their own ways to support current rounding mode. For example Clang and GCC have the option -frounding-math intended to disable optimizations that would change the meaning of rounding-sensitive code.
[edit] Example
Output:
rounding down: pi = 3.141592502593994140625 stof("1.1") = 1.099999904632568359375 rint(2.1) = 2 rounding up: pi = 3.1415927410125732421875 stof("1.1") = 1.10000002384185791015625 rint(2.1) = 3