# V3529\. AUTOSAR\. Floating\-point values should not be tested for equality or inequality\.

This diagnostic rule is based on the software development guidelines developed by [AUTOSAR](https://www.autosar.org/) \(AUTomotive Open System ARchitecture\)\.

When comparing values of real types for equality or non\-equality, the results may vary depending on the processor being used and compiler settings\.

Example of non\-compliant code:

```cpp
const double PI_div_2 = 1.0;
const double sinValue = sin(M_PI / 2.0);

if (sinValue == PI_div_2) { .... }
```

To compare values of real types correctly, either use the predefined constant 'std::numeric\_limits<float\>::epsilon\(\)' or 'std::numeric\_limits<double\>::epsilon\(\)' or create your own constant 'Epsilon' of custom precision\.

Fixed code:

```cpp
const double PI_div_2 = 1.0;
const double sinValue = sin(M_PI / 2.0);

// equality
if (fabs(a - b) <= std::numeric_limits<double>::epsilon()) { .... };

// inequality
if (fabs(a - b) > std::numeric_limits<double>::epsilon()) { .... };
```

In some cases, it is allowed to compare two real numbers using the '\=\=' or '\!\=' operator, for example, when checking a variable for a known value:

```cpp
bool foo();
double bar();

double val = foo() ? bar() : 0.0;
if (val == 0.0) { .... }
```

The analyzer does not issue the warning if a value is compared with itself\. Such a comparison is useful to check a variable for [NaN](https://en.wikipedia.org/wiki/NaN):

```cpp
bool isnan(double value) { return value != value; }
```

However, a better style is to implement this check through the '[std::isnan](https://en.cppreference.com/w/cpp/numeric/math/isnan)' function\.