std::isnan
<cmath>
bool isnan( double num );
(until C++23)
<simd>
constexpr typename /*deduced-simd-t*/<V>::mask_type
<cmath>
bool isnan( Integer num );
(constexpr since C++23)
std::isnan on v_num.- (See math-floating-point and deduced-simd-t for their definitions.)
Contents
[edit] Parameters
[edit] Return value
[0, v_num.size()).[edit] Notes
There are many different NaN values with different sign bits and payloads, see std::nan and std::numeric_limits::quiet_NaN .
NaN values never compare equal to themselves or to other NaN values. Copying a NaN is not required, by IEEE-754, to preserve its bit representation (sign and payload), though most implementation do.
Another way to test if a floating-point value is NaN is to compare it with itself: bool is_nan(double x) { return x != x; }.
GCC and Clang support a -ffinite-math option (additionally implied by -ffast-math), which allows the respective compiler to assume the nonexistence of special IEEE-754 floating point values such as NaN, infinity, or negative zero. In other words, std::isnan is assumed to always return false under this option.
The additional overloads are not required to be provided exactly as (A). They only need to be sufficient to ensure that for their argument num of integer type, std::isnan(num) has the same effect as std::isnan(static_cast<double>(num)).
[edit] Example
#include <cfloat> #include <cmath> #include <iostream> int main() { std::cout << std::boolalpha << "isnan(NaN) = " << std::isnan(NAN ) << '\n' << "isnan(Inf) = " << std::isnan(INFINITY ) << '\n' << "isnan(0.0) = " << std::isnan(0.0) << '\n' << "isnan(DBL_MIN/2.0) = " << std::isnan(DBL_MIN / 2.0) << '\n' << "isnan(0.0 / 0.0) = " << std::isnan(0.0 / 0.0) << '\n' << "isnan(Inf - Inf) = " << std::isnan(INFINITY - INFINITY ) << '\n'; }
Output:
isnan(NaN) = true isnan(Inf) = false isnan(0.0) = false isnan(DBL_MIN/2.0) = false isnan(0.0 / 0.0) = true isnan(Inf - Inf) = true