Legacy Document

logb

You can use the logb function to determine the value in the exponent field of a floating-point number.

double_t logb (double_t x);

x

Any floating-point number.

DESCRIPTION

The logb function returns the signed exponent of its argument x as a signed integer value.

$logb(x)\; =\; y$ such that $x\; =\; f\times 2y$

When the argument is a denormalized number, the exponent is determined as if the input argument had first been normalized.

Note that for a nonzero finite x, $1$ fabs(scalb(x,-logb(x)))2 .

That is, for a nonzero finite x, the magnitude of x taken to the power of its inverse exponent is between 1 and 2.

This function conforms to IEEE Standard 854, which differs from IEEE Standard 754 on the treatment of a denormalized argument x.

EXCEPTIONS

If x is finite and nonzero, the result of $logb(x)$ is exact.

SPECIAL CASES

Table 10-18 shows the results when the argument to the logb function is a zero, a NaN, or an Infinity.

Special cases for the logb function

Operation	Result	Exceptions raised
$logb(+0)$	-	Divide-by-zero
$logb(-0)$	-	Divide-by-zero
$logb(NaN)$	NaN	None[39]
$logb(+$ )	+	None
$logb(-$ )	+	None

EXAMPLES

z = logb(789.9);  /* z = 9.0 because 789.9  1.54  29 */
z = logb(21456789);/* z = 24.0 because 21456789  1.28  224 */

---

[39] If the NaN is a signaling NaN, the invalid exception is raised.