# std::logb, std::logbf, std::logbl

> C++ Atlas: `cppatlas.com/w/cpp/numeric/math/logb` | Source: https://en.cppreference.com/w/cpp/numeric/math/logb | Domain: Numerics | Standards: C++11, C++23, C++26

Formally, the unbiased exponent is the signed integral part of logr|num| (returned by this function as a floating-point value), for non-zero num, where r is std::numeric_limits<T>::radix and T is the floating-point type of num.

## Official Specification Summary (cppreference)

```cpp
// Defined in header <cmath>

float       logb ( float num );
double      logb ( double num );
long double logb ( long double num ); // (1) (until C++23)

constexpr /*floating-point-type*/
            logb ( /*floating-point-type*/ num ); // (1) (since C++23)

float       logbf( float num ); // (2) (since C++11) (constexpr since C++23)

long double logbl( long double num ); // (3) (since C++11) (constexpr since C++23)

// SIMD overload (since C++26)

// Defined in header <simd>

template< /*math-floating-point*/ V >
constexpr /*deduced-simd-t*/<V>
            logb ( const V& v_num ); // (S) (since C++26)

// Additional overloads (since C++11)

// Defined in header <cmath>

template< class Integer >
double      logb ( Integer num ); // (A) (constexpr since C++23)
```

1-3) Extracts the value of the unbiased radix-independent exponent from the floating-point argument `num`, and returns it as a floating-point value. The library provides overloads of `std::logb` for all cv-unqualified floating-point types as the type of the parameter. (since C++23)

(since C++26):

S) The SIMD overload performs an element-wise `std::logb` on `v_num`.

  (See `math-floating-point` and `deduced-simd-t` for their definitions.)

A) Additional overloads are provided for all integer types, which are treated as `double`. (since C++11)

Formally, the unbiased exponent is the signed integral part of log^(r)|num| (returned by this function as a floating-point value), for non-zero `num`, where `r` is `std::numeric_limits<T>::radix` and `T` is the floating-point type of `num`. If `num` is subnormal, it is treated as though it was normalized.

### Parameters

- `num` - floating-point or integer value

### Return value

If no errors occur, the unbiased exponent of `num` is returned as a signed floating-point value.

If a domain error occurs, an implementation-defined value is returned.

If a pole error occurs, `-HUGE_VAL`, `-HUGE_VALF`, or `-HUGE_VALL` is returned.

### Error handling

Errors are reported as specified in `math_errhandling`.

Domain or range error may occur if `num` is zero.

If the implementation supports IEEE floating-point arithmetic (IEC 60559),

- If `num` is ±0, -∞ is returned and `FE_DIVBYZERO` is raised.
- If `num` is ±∞, +∞ is returned.
- If `num` is NaN, NaN is returned.
- In all other cases, the result is exact (`FE_INEXACT` is never raised) and the current rounding mode is ignored.

### Notes

POSIX requires that a pole error occurs if `num` is ±0.

The value of the exponent returned by `std::logb` is always 1 less than the exponent returned by `std::frexp` because of the different normalization requirements: for the exponent `e` returned by `std::logb`, |num*r^(-e)| is between `1` and `r` (typically between `1` and `2`), but for the exponent `e` returned by `std::frexp`, |num*2^(-e)| is between `0.5` and `1`.

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::logb(num)` has the same effect as `std::logb(static_cast<double>(num))`.

### Example

Compares different floating-point decomposition functions:

```cpp
#include <cfenv>
#include <cmath>
#include <iostream>
#include <limits>
// #pragma STDC FENV_ACCESS ON

int main()
{
    double f = 123.45;
    std::cout << "Given the number " << f << " or " << std::hexfloat
              << f << std::defaultfloat << " in hex,\n";

    double f3;
    double f2 = std::modf(f, &f3);
    std::cout << "modf() makes " << f3 << " + " << f2 << '\n';

    int i;
    f2 = std::frexp(f, &i);
    std::cout << "frexp() makes " << f2 << " * 2^" << i << '\n';

    i = std::ilogb(f);
    std::cout << "logb()/ilogb() make " << f / std::scalbn(1.0, i) << " * "
              << std::numeric_limits<double>::radix
              << "^" << std::ilogb(f) << '\n';

    // error handling
    std::feclearexcept(FE_ALL_EXCEPT);

    std::cout << "logb(0) = " << std::logb(0) << '\n';
    if (std::fetestexcept(FE_DIVBYZERO))
        std::cout << "    FE_DIVBYZERO raised\n";
}
```

Possible output:

```text
Given the number 123.45 or 0x1.edccccccccccdp+6 in hex,
modf() makes 123 + 0.45
frexp() makes 0.964453 * 2^7
logb()/ilogb() make 1.92891 * 2^6
logb(0) = -Inf
    FE_DIVBYZERO raised
```

### See also

- [`frexp`](https://en.cppreference.com/w/cpp/numeric/math/frexp), [`frexpf`](https://en.cppreference.com/w/cpp/numeric/math/frexp) (C++11), [`frexpl`](https://en.cppreference.com/w/cpp/numeric/math/frexp) (C++11): decomposes a number into significand and base-2 exponent (function)
- [`ilogb`](https://en.cppreference.com/w/cpp/numeric/math/ilogb) (C++11), [`ilogbf`](https://en.cppreference.com/w/cpp/numeric/math/ilogb) (C++11), [`ilogbl`](https://en.cppreference.com/w/cpp/numeric/math/ilogb) (C++11): extracts exponent of the number (function)
- [`scalbn`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11), [`scalbnf`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11), [`scalbnl`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11), [`scalbln`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11), [`scalblnf`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11), [`scalblnl`](https://en.cppreference.com/w/cpp/numeric/math/scalbn) (C++11): multiplies a number by `FLT_RADIX` raised to a power (function)
- [C documentation for `logb`](https://en.cppreference.com/w/c/numeric/math/logb)

---
Attribution: content from https://en.cppreference.com/w/cpp/numeric/math/logb (cppreference.com), licensed under CC-BY-SA 3.0 and GFDL. Modified: converted to markdown, study notes added. Offline snapshot of 9 February 2025.