HP-UX Reference (11i v2 04/09) - 3 Library Functions N-Z (vol 7)

ManualsBrandsHP ManualsSoftwareHP-UX Reference Manuals

341

342

343

344

345

346

347

348

349

350

rsqrt(3M) rsqrt(3M)

(Itanium(R)-based System Only)

NAME

rsqrt( ), rsqrtf( ), rsqrtl( ), rsqrtw( ), rsqrtq( ) - reciprocal square root functions

SYNOPSIS

#include <math.h>

double rsqrt(double x);

float rsqrtf(float x);

long double rsqrtl(long double x);

extended rsqrtw(extended x);

quad rsqrtq(quad x);

DESCRIPTION

These functions are available only for Itanium-based systems.

rsqrt() returns the reciprocal of the non-negative square root of x.

The expression a*

rsqrt(b) is intended as a higher performance alternative to a/

sqrt(b). The two

expressions are comparably accurate, but do not compute exactly the same value in every case. For exam-

ple, a*

rsqrt(a*a + b*b) can be just slightly greater than 1, in rare cases.

rsqrtf() is a float version of rsqrt(); it takes a float argument and returns a float

result.

Itanium-based System Only

rsqrtl() is a long double version of rsqrt(); it takes a long double argument and returns a

long double result.

rsqrtw() is an extended version of rsqrt(); it takes an extended argument and returns an

extended result.

rsqrtq() is equivalent to rsqrtl() on HP-UX systems.

USAGE

To use these functions compile either with the default

-Ae option or with the -Aa

and the

-D_HPUX_SOURCE options.

To use (for Itanium-based systems)

rsqrtw() or rsqrtq(), compile also with the -fpwidetypes

option.

To use any of these functions, make sure your program includes

<math.h>, and link in the math library

by specifying -lm on the compiler or linker command line.

RETURN VALUE

If x is ±0,

rsqrt() returns ±INFINITY and raises the divide-by-zero exception.

If x is +INFINITY,

rsqrt() returns +0.

If x is NaN,

rsqrt() returns NaN.

If x is less than zero,

rsqrt() returns NaN and raises the invalid ﬂoating-point exception.

rsqrt() raises the inexact exception whenever a rounded result does not equal the mathematical result.

ERRORS

No errors are deﬁned.