mirror of
https://sourceware.org/git/glibc.git
synced 2024-11-30 08:40:07 +00:00
e02920bc02
In non-default rounding modes, tgamma can be slightly less accurate than permitted by glibc's accuracy goals. Part of the problem is error accumulation, addressed in this patch by setting round-to-nearest for internal computations. However, there was also a bug in the code dealing with computing pow (x + n, x + n) where x + n is not exactly representable, providing another source of error even in round-to-nearest mode; it was necessary to address both bugs to get errors for all testcases within glibc's accuracy goals. Given this second fix, accuracy in round-to-nearest mode is also improved (hence regeneration of ulps for tgamma should be from scratch - truncate libm-test-ulps or at least remove existing tgamma entries - so that the expected ulps can be reduced). Some additional complications also arose. Certain tgamma tests should strictly, according to IEEE semantics, overflow or not depending on the rounding mode; this is beyond the scope of glibc's accuracy goals for any function without exactly-determined results, but gen-auto-libm-tests doesn't handle being lax there as it does for underflow. (libm-test.inc also doesn't handle being lax about whether the result in cases very close to the overflow threshold is infinity or a finite value close to overflow, but that doesn't cause problems in this case though I've seen it cause problems with random test generation for some functions.) Thus, spurious-overflow markings, with a comment, are added to auto-libm-test-in (no bug in Bugzilla because the issue is with the testsuite, not a user-visible bug in glibc). And on x86, after the patch I saw ERANGE issues as previously reported by Carlos (see my commentary in <https://sourceware.org/ml/libc-alpha/2015-01/msg00485.html>), which needed addressing by ensuring excess range and precision were eliminated at various points if FLT_EVAL_METHOD != 0. I also noticed and fixed a cosmetic issue where 1.0f was used in long double functions and should have been 1.0L. This completes the move of all functions to testing in all rounding modes with ALL_RM_TEST, so gen-libm-have-vector-test.sh is updated to remove the workaround for some functions not using ALL_RM_TEST. Tested for x86_64, x86, mips64 and powerpc. [BZ #18613] * sysdeps/ieee754/dbl-64/e_gamma_r.c (gamma_positive): Take log of X_ADJ not X when adjusting exponent. (__ieee754_gamma_r): Do intermediate computations in round-to-nearest then adjust overflowing and underflowing results as needed. * sysdeps/ieee754/flt-32/e_gammaf_r.c (gammaf_positive): Take log of X_ADJ not X when adjusting exponent. (__ieee754_gammaf_r): Do intermediate computations in round-to-nearest then adjust overflowing and underflowing results as needed. * sysdeps/ieee754/ldbl-128/e_gammal_r.c (gammal_positive): Take log of X_ADJ not X when adjusting exponent. (__ieee754_gammal_r): Do intermediate computations in round-to-nearest then adjust overflowing and underflowing results as needed. Use 1.0L not 1.0f as numerator of division. * sysdeps/ieee754/ldbl-128ibm/e_gammal_r.c (gammal_positive): Take log of X_ADJ not X when adjusting exponent. (__ieee754_gammal_r): Do intermediate computations in round-to-nearest then adjust overflowing and underflowing results as needed. Use 1.0L not 1.0f as numerator of division. * sysdeps/ieee754/ldbl-96/e_gammal_r.c (gammal_positive): Take log of X_ADJ not X when adjusting exponent. (__ieee754_gammal_r): Do intermediate computations in round-to-nearest then adjust overflowing and underflowing results as needed. Use 1.0L not 1.0f as numerator of division. * math/libm-test.inc (tgamma_test_data): Remove one test. Moved to auto-libm-test-in. (tgamma_test): Use ALL_RM_TEST. * math/auto-libm-test-in: Add one test of tgamma. Mark some other tests of tgamma with spurious-overflow. * math/auto-libm-test-out: Regenerated. * math/gen-libm-have-vector-test.sh: Do not check for START. * sysdeps/i386/fpu/libm-test-ulps: Update. * sysdeps/x86_64/fpu/libm-test-ulps: Likewise.
210 lines
6.0 KiB
C
210 lines
6.0 KiB
C
/* Implementation of gamma function according to ISO C.
|
|
Copyright (C) 1997-2015 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 <math.h>
|
|
#include <math_private.h>
|
|
#include <float.h>
|
|
|
|
/* Coefficients B_2k / 2k(2k-1) of x^-(2k-1) inside exp in Stirling's
|
|
approximation to gamma function. */
|
|
|
|
static const long double gamma_coeff[] =
|
|
{
|
|
0x1.5555555555555556p-4L,
|
|
-0xb.60b60b60b60b60bp-12L,
|
|
0x3.4034034034034034p-12L,
|
|
-0x2.7027027027027028p-12L,
|
|
0x3.72a3c5631fe46aep-12L,
|
|
-0x7.daac36664f1f208p-12L,
|
|
0x1.a41a41a41a41a41ap-8L,
|
|
-0x7.90a1b2c3d4e5f708p-8L,
|
|
};
|
|
|
|
#define NCOEFF (sizeof (gamma_coeff) / sizeof (gamma_coeff[0]))
|
|
|
|
/* Return gamma (X), for positive X less than 1766, in the form R *
|
|
2^(*EXP2_ADJ), where R is the return value and *EXP2_ADJ is set to
|
|
avoid overflow or underflow in intermediate calculations. */
|
|
|
|
static long double
|
|
gammal_positive (long double x, int *exp2_adj)
|
|
{
|
|
int local_signgam;
|
|
if (x < 0.5L)
|
|
{
|
|
*exp2_adj = 0;
|
|
return __ieee754_expl (__ieee754_lgammal_r (x + 1, &local_signgam)) / x;
|
|
}
|
|
else if (x <= 1.5L)
|
|
{
|
|
*exp2_adj = 0;
|
|
return __ieee754_expl (__ieee754_lgammal_r (x, &local_signgam));
|
|
}
|
|
else if (x < 7.5L)
|
|
{
|
|
/* Adjust into the range for using exp (lgamma). */
|
|
*exp2_adj = 0;
|
|
long double n = __ceill (x - 1.5L);
|
|
long double x_adj = x - n;
|
|
long double eps;
|
|
long double prod = __gamma_productl (x_adj, 0, n, &eps);
|
|
return (__ieee754_expl (__ieee754_lgammal_r (x_adj, &local_signgam))
|
|
* prod * (1.0L + eps));
|
|
}
|
|
else
|
|
{
|
|
long double eps = 0;
|
|
long double x_eps = 0;
|
|
long double x_adj = x;
|
|
long double prod = 1;
|
|
if (x < 13.0L)
|
|
{
|
|
/* Adjust into the range for applying Stirling's
|
|
approximation. */
|
|
long double n = __ceill (13.0L - x);
|
|
x_adj = x + n;
|
|
x_eps = (x - (x_adj - n));
|
|
prod = __gamma_productl (x_adj - n, x_eps, n, &eps);
|
|
}
|
|
/* The result is now gamma (X_ADJ + X_EPS) / (PROD * (1 + EPS)).
|
|
Compute gamma (X_ADJ + X_EPS) using Stirling's approximation,
|
|
starting by computing pow (X_ADJ, X_ADJ) with a power of 2
|
|
factored out. */
|
|
long double exp_adj = -eps;
|
|
long double x_adj_int = __roundl (x_adj);
|
|
long double x_adj_frac = x_adj - x_adj_int;
|
|
int x_adj_log2;
|
|
long double x_adj_mant = __frexpl (x_adj, &x_adj_log2);
|
|
if (x_adj_mant < M_SQRT1_2l)
|
|
{
|
|
x_adj_log2--;
|
|
x_adj_mant *= 2.0L;
|
|
}
|
|
*exp2_adj = x_adj_log2 * (int) x_adj_int;
|
|
long double ret = (__ieee754_powl (x_adj_mant, x_adj)
|
|
* __ieee754_exp2l (x_adj_log2 * x_adj_frac)
|
|
* __ieee754_expl (-x_adj)
|
|
* __ieee754_sqrtl (2 * M_PIl / x_adj)
|
|
/ prod);
|
|
exp_adj += x_eps * __ieee754_logl (x_adj);
|
|
long double bsum = gamma_coeff[NCOEFF - 1];
|
|
long double x_adj2 = x_adj * x_adj;
|
|
for (size_t i = 1; i <= NCOEFF - 1; i++)
|
|
bsum = bsum / x_adj2 + gamma_coeff[NCOEFF - 1 - i];
|
|
exp_adj += bsum / x_adj;
|
|
return ret + ret * __expm1l (exp_adj);
|
|
}
|
|
}
|
|
|
|
long double
|
|
__ieee754_gammal_r (long double x, int *signgamp)
|
|
{
|
|
u_int32_t es, hx, lx;
|
|
long double ret;
|
|
|
|
GET_LDOUBLE_WORDS (es, hx, lx, x);
|
|
|
|
if (__glibc_unlikely (((es & 0x7fff) | hx | lx) == 0))
|
|
{
|
|
/* Return value for x == 0 is Inf with divide by zero exception. */
|
|
*signgamp = 0;
|
|
return 1.0 / x;
|
|
}
|
|
if (__glibc_unlikely (es == 0xffffffff && ((hx & 0x7fffffff) | lx) == 0))
|
|
{
|
|
/* x == -Inf. According to ISO this is NaN. */
|
|
*signgamp = 0;
|
|
return x - x;
|
|
}
|
|
if (__glibc_unlikely ((es & 0x7fff) == 0x7fff))
|
|
{
|
|
/* Positive infinity (return positive infinity) or NaN (return
|
|
NaN). */
|
|
*signgamp = 0;
|
|
return x + x;
|
|
}
|
|
if (__builtin_expect ((es & 0x8000) != 0, 0) && __rintl (x) == x)
|
|
{
|
|
/* Return value for integer x < 0 is NaN with invalid exception. */
|
|
*signgamp = 0;
|
|
return (x - x) / (x - x);
|
|
}
|
|
|
|
if (x >= 1756.0L)
|
|
{
|
|
/* Overflow. */
|
|
*signgamp = 0;
|
|
return LDBL_MAX * LDBL_MAX;
|
|
}
|
|
else
|
|
{
|
|
SET_RESTORE_ROUNDL (FE_TONEAREST);
|
|
if (x > 0.0L)
|
|
{
|
|
*signgamp = 0;
|
|
int exp2_adj;
|
|
ret = gammal_positive (x, &exp2_adj);
|
|
ret = __scalbnl (ret, exp2_adj);
|
|
}
|
|
else if (x >= -LDBL_EPSILON / 4.0L)
|
|
{
|
|
*signgamp = 0;
|
|
ret = 1.0L / x;
|
|
}
|
|
else
|
|
{
|
|
long double tx = __truncl (x);
|
|
*signgamp = (tx == 2.0L * __truncl (tx / 2.0L)) ? -1 : 1;
|
|
if (x <= -1766.0L)
|
|
/* Underflow. */
|
|
ret = LDBL_MIN * LDBL_MIN;
|
|
else
|
|
{
|
|
long double frac = tx - x;
|
|
if (frac > 0.5L)
|
|
frac = 1.0L - frac;
|
|
long double sinpix = (frac <= 0.25L
|
|
? __sinl (M_PIl * frac)
|
|
: __cosl (M_PIl * (0.5L - frac)));
|
|
int exp2_adj;
|
|
ret = M_PIl / (-x * sinpix
|
|
* gammal_positive (-x, &exp2_adj));
|
|
ret = __scalbnl (ret, -exp2_adj);
|
|
}
|
|
}
|
|
}
|
|
if (isinf (ret) && x != 0)
|
|
{
|
|
if (*signgamp < 0)
|
|
return -(-__copysignl (LDBL_MAX, ret) * LDBL_MAX);
|
|
else
|
|
return __copysignl (LDBL_MAX, ret) * LDBL_MAX;
|
|
}
|
|
else if (ret == 0)
|
|
{
|
|
if (*signgamp < 0)
|
|
return -(-__copysignl (LDBL_MIN, ret) * LDBL_MIN);
|
|
else
|
|
return __copysignl (LDBL_MIN, ret) * LDBL_MIN;
|
|
}
|
|
else
|
|
return ret;
|
|
}
|
|
strong_alias (__ieee754_gammal_r, __gammal_r_finite)
|