glibc/math/s_catanhl.c
Joseph Myers a84e78c8b3 Fix catan, catanh, __ieee754_logf in round-downward mode (bug 16799, bug 16800).
This patch fixes incorrect results from catan and catanh of certain
special inputs in round-downward mode (bug 16799), and incorrect
results of __ieee754_logf (+/-0) in round-downward mode (bug 16800)
that show up through catan/catanh when tested in all rounding modes,
but not directly in the testing for logf because the bug gets hidden
by the wrappers.

Both bugs involve a zero that should be +0 being -0 instead: one
computed as (1-x)*(1+x) in the catan/catanh case, and one as (x-x) in
the logf case.  The fixes ensure positive zero is used.  Testing of
catan and catanh in all rounding modes is duly enabled.

I expect there are various other bugs in special cases in __ieee754_*
functions that are normally hidden by the wrappers but would show up
for testing with -lieee (or in future with -fno-math-errno if we
replace -lieee and _LIB_VERSION with compile-time redirection to new
*_noerrno symbol names).

Tested x86_64 and x86 and ulps updated accordingly.

	[BZ #16799]
	[BZ #16800]
	* math/s_catan.c (__catan): Avoid passing -0 denominator to atan2
	with 0 numerator.
	* math/s_catanf.c (__catanf): Likewise.
	* math/s_catanh.c (__catanh): Likewise.
	* math/s_catanhf.c (__catanhf): Likewise.
	* math/s_catanhl.c (__catanhl): Likewise.
	* math/s_catanl.c (__catanl): Likewise.
	* sysdeps/ieee754/flt-32/e_logf.c (__ieee754_logf): Always divide
	by positive zero when computing -Inf result.
	* math/libm-test.inc (catan_test): Use ALL_RM_TEST.
	(catanh_test): Likewise.
	* sysdeps/i386/fpu/libm-test-ulps: Update.
	* sysdeps/x86_64/fpu/libm-test-ulps: Likewise.
2014-04-02 17:41:02 +00:00

151 lines
4.1 KiB
C

/* Return arc hyperbole tangent for long double value.
Copyright (C) 1997-2014 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
#include <complex.h>
#include <math.h>
#include <math_private.h>
#include <float.h>
/* To avoid spurious overflows, use this definition to treat IBM long
double as approximating an IEEE-style format. */
#if LDBL_MANT_DIG == 106
# undef LDBL_EPSILON
# define LDBL_EPSILON 0x1p-106L
#endif
__complex__ long double
__catanhl (__complex__ long double x)
{
__complex__ long double res;
int rcls = fpclassify (__real__ x);
int icls = fpclassify (__imag__ x);
if (__glibc_unlikely (rcls <= FP_INFINITE || icls <= FP_INFINITE))
{
if (icls == FP_INFINITE)
{
__real__ res = __copysignl (0.0, __real__ x);
__imag__ res = __copysignl (M_PI_2l, __imag__ x);
}
else if (rcls == FP_INFINITE || rcls == FP_ZERO)
{
__real__ res = __copysignl (0.0, __real__ x);
if (icls >= FP_ZERO)
__imag__ res = __copysignl (M_PI_2l, __imag__ x);
else
__imag__ res = __nanl ("");
}
else
{
__real__ res = __nanl ("");
__imag__ res = __nanl ("");
}
}
else if (__glibc_unlikely (rcls == FP_ZERO && icls == FP_ZERO))
{
res = x;
}
else
{
if (fabsl (__real__ x) >= 16.0L / LDBL_EPSILON
|| fabsl (__imag__ x) >= 16.0L / LDBL_EPSILON)
{
__imag__ res = __copysignl (M_PI_2l, __imag__ x);
if (fabsl (__imag__ x) <= 1.0L)
__real__ res = 1.0L / __real__ x;
else if (fabsl (__real__ x) <= 1.0L)
__real__ res = __real__ x / __imag__ x / __imag__ x;
else
{
long double h = __ieee754_hypotl (__real__ x / 2.0L,
__imag__ x / 2.0L);
__real__ res = __real__ x / h / h / 4.0L;
}
}
else
{
if (fabsl (__real__ x) == 1.0L
&& fabsl (__imag__ x) < LDBL_EPSILON * LDBL_EPSILON)
__real__ res = (__copysignl (0.5L, __real__ x)
* (M_LN2l - __ieee754_logl (fabsl (__imag__ x))));
else
{
long double i2 = 0.0;
if (fabsl (__imag__ x) >= LDBL_EPSILON * LDBL_EPSILON)
i2 = __imag__ x * __imag__ x;
long double num = 1.0L + __real__ x;
num = i2 + num * num;
long double den = 1.0L - __real__ x;
den = i2 + den * den;
long double f = num / den;
if (f < 0.5L)
__real__ res = 0.25L * __ieee754_logl (f);
else
{
num = 4.0L * __real__ x;
__real__ res = 0.25L * __log1pl (num / den);
}
}
long double absx, absy, den;
absx = fabsl (__real__ x);
absy = fabsl (__imag__ x);
if (absx < absy)
{
long double t = absx;
absx = absy;
absy = t;
}
if (absy < LDBL_EPSILON / 2.0L)
{
den = (1.0L - absx) * (1.0L + absx);
if (den == -0.0L)
den = 0.0L;
}
else if (absx >= 1.0L)
den = (1.0L - absx) * (1.0L + absx) - absy * absy;
else if (absx >= 0.75L || absy >= 0.5L)
den = -__x2y2m1l (absx, absy);
else
den = (1.0L - absx) * (1.0L + absx) - absy * absy;
__imag__ res = 0.5L * __ieee754_atan2l (2.0L * __imag__ x, den);
}
if (fabsl (__real__ res) < LDBL_MIN)
{
volatile long double force_underflow = __real__ res * __real__ res;
(void) force_underflow;
}
if (fabsl (__imag__ res) < LDBL_MIN)
{
volatile long double force_underflow = __imag__ res * __imag__ res;
(void) force_underflow;
}
}
return res;
}
weak_alias (__catanhl, catanhl)