mirror/glibc
2
0
Fork 0
mirror of git://sourceware.org/git/glibc.git synced 2025-01-18 12:16:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

math: Use tanhf from CORE-MATH

Browse Source 
The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows slight better performance to the generic tanhf.

The code was adapted to glibc style and to use the definition of
math_config.h (to handle errno, overflow, and underflow).

Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (Neoverse-N1,
gcc 13.3.1), and powerpc (POWER10, gcc 13.2.1):

Latency                      master        patched   improvement
x86_64                      51.5273        41.0951        20.25%
x86_64v2                    47.7021        39.1526        17.92%
x86_64v3                    45.0373        34.2737        23.90%
i686                       133.9970        83.8596        37.42%
aarch64 (Neoverse)          21.5439        14.7961        31.32%
power10                     13.3301         8.4406        36.68%

reciprocal-throughput        master        patched   improvement
x86_64                      24.9493        12.8547        48.48%
x86_64v2                    20.7051        12.7761        38.29%
x86_64v3                    19.2492        11.0851        42.41%
i686                        78.6498        29.8211        62.08%
aarch64 (Neoverse)          11.6026        7.11487        38.68%
power10                      6.3328         2.8746        54.61%

Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: DJ Delorie <dj@redhat.com>
This commit is contained in:
Adhemerval Zanella 2024-11-19 16:30:16 -03:00
parent 1751c0519a
commit 0e0be3ed80
27 changed files with 84 additions and 144 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
SHARED-FILES
View File

@ -330,3 +330,7 @@ sysdeps/ieee754/flt-32/e_sinhf.c:
(src/binary32/sinh/sinhf.c in CORE-MATH)
- the code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_tanhf.c:
(src/binary32/tanh/tanhf.c in CORE-MATH)
- the code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.

4
sysdeps/aarch64/libm-test-ulps
View File

@ -1665,7 +1665,6 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_advsimd":
@ -1674,7 +1673,6 @@ float: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_sve":
@ -1683,12 +1681,10 @@ float: 2
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tanpi":

4
sysdeps/alpha/fpu/libm-test-ulps
View File

@ -1318,22 +1318,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tgamma":

4
sysdeps/arc/fpu/libm-test-ulps
View File

@ -1053,19 +1053,15 @@ double: 1
Function: "tanh":
double: 3
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 3
float: 3
Function: "tanh_upward":
double: 3
float: 3
Function: "tgamma":
double: 9

1
sysdeps/arc/nofpu/libm-test-ulps
View File

@ -251,7 +251,6 @@ double: 2
Function: "tanh":
double: 2
float: 2
Function: "tgamma":
double: 9

4
sysdeps/arm/libm-test-ulps
View File

@ -1143,19 +1143,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tanpi":
double: 2

4
sysdeps/csky/fpu/libm-test-ulps
View File

@ -971,19 +971,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tgamma":
double: 9

4
sysdeps/csky/nofpu/libm-test-ulps
View File

@ -1002,19 +1002,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tgamma":
double: 9

4
sysdeps/hppa/fpu/libm-test-ulps
View File

@ -1176,19 +1176,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tanpi":
double: 2

4
sysdeps/i386/fpu/libm-test-ulps
View File

@ -1754,25 +1754,21 @@ ldouble: 2
Function: "tanh":
double: 2
float: 2
float128: 2
ldouble: 3
Function: "tanh_downward":
double: 3
float: 3
float128: 4
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
float128: 3
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
float128: 3
ldouble: 4

4
sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
View File

@ -1759,25 +1759,21 @@ ldouble: 2
Function: "tanh":
double: 2
float: 2
float128: 2
ldouble: 3
Function: "tanh_downward":
double: 3
float: 3
float128: 4
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
float128: 3
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
float128: 3
ldouble: 4

133
sysdeps/ieee754/flt-32/s_tanhf.c
View File

@ -1,63 +1,90 @@
/* s_tanhf.c -- float version of s_tanh.c.
*/
/* Correctly-rounded hyperbolic tangent function for binary32 value.
/*
* ====================================================
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
*
* Developed at SunPro, a Sun Microsystems, Inc. business.
* Permission to use, copy, modify, and distribute this
* software is freely granted, provided that this notice
* is preserved.
* ====================================================
*/
Copyright (c) 2022-2024 Alexei Sibidanov.
#if defined(LIBM_SCCS) && !defined(lint)
static char rcsid[] = "$NetBSD: s_tanhf.c,v 1.4 1995/05/10 20:48:24 jtc Exp $";
#endif
The original version of this file was copied from the CORE-MATH
project (file src/binary32/tanh/tanhf.c, revision b1ecd83).
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
*/
#include <float.h>
#include <math.h>
#include <math_private.h>
#include <math-underflow.h>
#include <stdint.h>
#include <libm-alias-float.h>
#include "math_config.h"
static const float one=1.0, two=2.0, tiny = 1.0e-30;
float __tanhf(float x)
float
__tanhf (float x)
{
float t,z;
int32_t jx,ix;
GET_FLOAT_WORD(jx,x);
ix = jx&0x7fffffff;
/* x is INF or NaN */
if(ix>=0x7f800000) {
if (jx>=0) return one/x+one; /* tanh(+-inf)=+-1 */
else return one/x-one; /* tanh(NaN) = NaN */
double z = x;
uint32_t ux = asuint (x);
int e = (ux >> 23) & 0xff;
if (__glibc_unlikely (e == 0xff))
{
if (ux << 9)
return x + x; /* x = nan */
static const float ir[] = { 1.0f, -1.0f };
return ir[ux >> 31]; /* x = +-inf */
}
if (__glibc_unlikely (e < 115))
{
if (__glibc_unlikely (e < 102))
{
if (__glibc_unlikely ((ux << 1) == 0))
return x;
return fmaf (-x, fabsf (x), x);
}
/* |x| < 22 */
if (ix < 0x41b00000) { /* |x|<22 */
if (ix == 0)
return x; /* x == +-0 */
if (ix<0x24000000) /* |x|<2**-55 */
{
math_check_force_underflow (x);
return x*(one+x); /* tanh(small) = small */
}
if (ix>=0x3f800000) { /* |x|>=1 */
t = __expm1f(two*fabsf(x));
z = one - two/(t+two);
} else {
t = __expm1f(-two*fabsf(x));
z= -t/(t+two);
}
/* |x| > 22, return +-1 */
} else {
z = one - tiny; /* raised inexact flag */
}
return (jx>=0)? z: -z;
float x2 = x * x;
return fmaf (x, -0x1.555556p-2f * x2, x);
}
if ((ux << 1) > (0x41102cb3u << 1))
return copysignf (1.0f, x) - copysignf (0x1p-25f, x);
double z2 = z * z;
double z4 = z2 * z2;
double z8 = z4 * z4;
static const double cn[] =
{
0x1p+0, 0x1.30877b8b72d33p-3, 0x1.694aa09ae9e5ep-8,
0x1.4101377abb729p-14, 0x1.e0392b1db0018p-22, 0x1.2533756e546f7p-30,
0x1.d62e5abe6ae8ap-41, 0x1.b06be534182dep-54
};
static const double cd[] =
{
0x1p+0, 0x1.ed99131b0ebeap-2, 0x1.0d27ed6c95a69p-5,
0x1.7cbdaca0e9fccp-11, 0x1.b4e60b892578ep-18, 0x1.a6f707c5c71abp-26,
0x1.35a8b6e2cd94cp-35, 0x1.ca8230677aa01p-47
};
double n0 = cn[0] + z2 * cn[1];
double n2 = cn[2] + z2 * cn[3];
double n4 = cn[4] + z2 * cn[5];
double n6 = cn[6] + z2 * cn[7];
n0 += z4 * n2;
n4 += z4 * n6;
n0 += z8 * n4;
double d0 = cd[0] + z2 * cd[1];
double d2 = cd[2] + z2 * cd[3];
double d4 = cd[4] + z2 * cd[5];
double d6 = cd[6] + z2 * cd[7];
d0 += z4 * d2;
d4 += z4 * d6;
d0 += z8 * d4;
double r = z * n0 / d0;
return r;
}
libm_alias_float (__tanh, tanh)

4
sysdeps/loongarch/lp64/libm-test-ulps
View File

@ -1325,22 +1325,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tgamma":

3
sysdeps/m68k/m680x0/fpu/libm-test-ulps
View File

@ -1152,15 +1152,12 @@ double: 1
Function: "tanh_downward":
double: 1
float: 1
Function: "tanh_towardzero":
double: 1
float: 1
Function: "tanh_upward":
double: 1
float: 1
Function: "tgamma":
double: 3

1
sysdeps/microblaze/libm-test-ulps
View File

@ -233,7 +233,6 @@ double: 2
Function: "tanh":
double: 2
float: 2
Function: "tgamma":
double: 5

4
sysdeps/mips/mips32/libm-test-ulps
View File

@ -1050,19 +1050,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tgamma":
double: 9

4
sysdeps/mips/mips64/libm-test-ulps
View File

@ -1336,22 +1336,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tgamma":

4
sysdeps/or1k/fpu/libm-test-ulps
View File

@ -984,19 +984,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tgamma":
double: 9

4
sysdeps/or1k/nofpu/libm-test-ulps
View File

@ -1013,19 +1013,15 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_downward":
double: 3
float: 3
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tanh_upward":
double: 3
float: 3
Function: "tanpi":
double: 2

4
sysdeps/powerpc/fpu/libm-test-ulps
View File

@ -1861,25 +1861,21 @@ ldouble: 3
Function: "tanh":
double: 2
float: 2
float128: 2
ldouble: 1
Function: "tanh_downward":
double: 3
float: 3
float128: 4
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
float128: 3
ldouble: 4
Function: "tanh_upward":
double: 3
float: 3
float128: 3
ldouble: 6

4
sysdeps/powerpc/nofpu/libm-test-ulps
View File

@ -1452,22 +1452,18 @@ ldouble: 3
Function: "tanh":
double: 2
float: 2
ldouble: 1
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 4
Function: "tanh_upward":
double: 3
float: 3
ldouble: 6
Function: "tgamma":

4
sysdeps/riscv/nofpu/libm-test-ulps
View File

@ -1265,22 +1265,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tgamma":

4
sysdeps/riscv/rvd/libm-test-ulps
View File

@ -1323,22 +1323,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tgamma":

4
sysdeps/s390/fpu/libm-test-ulps
View File

@ -1362,22 +1362,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tanpi":

2
sysdeps/sh/libm-test-ulps
View File

@ -486,11 +486,9 @@ double: 1
Function: "tanh":
double: 2
float: 2
Function: "tanh_towardzero":
double: 2
float: 2
Function: "tgamma":
double: 9

4
sysdeps/sparc/fpu/libm-test-ulps
View File

@ -1377,22 +1377,18 @@ ldouble: 1
Function: "tanh":
double: 2
float: 2
ldouble: 2
Function: "tanh_downward":
double: 3
float: 3
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
ldouble: 3
Function: "tanpi":

4
sysdeps/x86_64/fpu/libm-test-ulps
View File

@ -2280,25 +2280,21 @@ float: 2
Function: "tanh":
double: 2
float: 2
float128: 2
ldouble: 3
Function: "tanh_downward":
double: 3
float: 3
float128: 4
ldouble: 4
Function: "tanh_towardzero":
double: 2
float: 2
float128: 3
ldouble: 3
Function: "tanh_upward":
double: 3
float: 3
float128: 3
ldouble: 4