Add kernel functions for 80-bit long doubles. Many thanks to Steve and
Bruce for putting lots of effort into these; getting them right isn't easy, and they went through many iterations. Submitted by: Steve Kargl <sgk@apl.washington.edu> with revisions from bde
This commit is contained in:
David Schultz
parent
079299f710
commit
de336b0c5e
Notes:
svn2git
2020-12-20 02:59:44 +00:00
svn path=/head/; revision=176357
78
lib/msun/ld80/k_cosl.c
Normal file
78
lib/msun/ld80/k_cosl.c
Normal file
@ -0,0 +1,78 @@
|
||||
/* From: @(#)k_cos.c 1.3 95/01/18 */
|
||||
/*
|
||||
* ====================================================
|
||||
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
|
||||
* Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
|
||||
*
|
||||
* Developed at SunSoft, a Sun Microsystems, Inc. business.
|
||||
* Permission to use, copy, modify, and distribute this
|
||||
* software is freely granted, provided that this notice
|
||||
* is preserved.
|
||||
* ====================================================
|
||||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__FBSDID("$FreeBSD$");
|
||||
|
||||
/*
|
||||
* ld80 version of k_cos.c. See ../src/k_cos.c for most comments.
|
||||
*/
|
||||
|
||||
#include "math_private.h"
|
||||
|
||||
/*
|
||||
* Domain [-0.7854, 0.7854], range ~[-2.43e-23, 2.425e-23]:
|
||||
* |cos(x) - c(x)| < 2**-75.1
|
||||
*
|
||||
* The coefficients of c(x) were generated by a pari-gp script using
|
||||
* a Remez algorithm that searches for the best higher coefficients
|
||||
* after rounding leading coefficients to a specified precision.
|
||||
*
|
||||
* Simpler methods like Chebyshev or basic Remez barely suffice for
|
||||
* cos() in 64-bit precision, because we want the coefficient of x^2
|
||||
* to be precisely -0.5 so that multiplying by it is exact, and plain
|
||||
* rounding of the coefficients of a good polynomial approximation only
|
||||
* gives this up to about 64-bit precision. Plain rounding also gives
|
||||
* a mediocre approximation for the coefficient of x^4, but a rounding
|
||||
* error of 0.5 ulps for this coefficient would only contribute ~0.01
|
||||
* ulps to the final error, so this is unimportant. Rounding errors in
|
||||
* higher coefficients are even less important.
|
||||
*
|
||||
* In fact, coefficients above the x^4 one only need to have 53-bit
|
||||
* precision, and this is more efficient. We get this optimization
|
||||
* almost for free from the complications needed to search for the best
|
||||
* higher coefficients.
|
||||
*/
|
||||
static const double
|
||||
one = 1.0;
|
||||
|
||||
#if defined(__amd64__) || defined(__i386__)
|
||||
/* Long double constants are slow on these arches, and broken on i386. */
|
||||
static const volatile double
|
||||
C1hi = 0.041666666666666664, /* 0x15555555555555.0p-57 */
|
||||
C1lo = 2.2598839032744733e-18; /* 0x14d80000000000.0p-111 */
|
||||
#define C1 ((long double)C1hi + C1lo)
|
||||
#else
|
||||
static const long double
|
||||
C1 = 0.0416666666666666666136L; /* 0xaaaaaaaaaaaaaa9b.0p-68 */
|
||||
#endif
|
||||
|
||||
static const double
|
||||
C2 = -0.0013888888888888874, /* -0x16c16c16c16c10.0p-62 */
|
||||
C3 = 0.000024801587301571716, /* 0x1a01a01a018e22.0p-68 */
|
||||
C4 = -0.00000027557319215507120, /* -0x127e4fb7602f22.0p-74 */
|
||||
C5 = 0.0000000020876754400407278, /* 0x11eed8caaeccf1.0p-81 */
|
||||
C6 = -1.1470297442401303e-11, /* -0x19393412bd1529.0p-89 */
|
||||
C7 = 4.7383039476436467e-14; /* 0x1aac9d9af5c43e.0p-97 */
|
||||
|
||||
long double
|
||||
__kernel_cosl(long double x, long double y)
|
||||
{
|
||||
long double hz,z,r,w;
|
||||
|
||||
z = x*x;
|
||||
r = z*(C1+z*(C2+z*(C3+z*(C4+z*(C5+z*(C6+z*C7))))));
|
||||
hz = 0.5*z;
|
||||
w = one-hz;
|
||||
return w + (((one-w)-hz) + (z*r-x*y));
|
||||
}
|
||||
62
lib/msun/ld80/k_sinl.c
Normal file
62
lib/msun/ld80/k_sinl.c
Normal file
@ -0,0 +1,62 @@
|
||||
/* From: @(#)k_sin.c 1.3 95/01/18 */
|
||||
/*
|
||||
* ====================================================
|
||||
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
|
||||
* Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
|
||||
*
|
||||
* Developed at SunSoft, a Sun Microsystems, Inc. business.
|
||||
* Permission to use, copy, modify, and distribute this
|
||||
* software is freely granted, provided that this notice
|
||||
* is preserved.
|
||||
* ====================================================
|
||||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__FBSDID("$FreeBSD$");
|
||||
|
||||
/*
|
||||
* ld80 version of k_sin.c. See ../src/k_sin.c for most comments.
|
||||
*/
|
||||
|
||||
#include "math_private.h"
|
||||
|
||||
static const double
|
||||
half = 0.5;
|
||||
|
||||
/*
|
||||
* Domain [-0.7854, 0.7854], range ~[-1.89e-22, 1.915e-22]
|
||||
* |sin(x)/x - s(x)| < 2**-72.1
|
||||
*
|
||||
* See ../ld80/k_cosl.c for more details about the polynomial.
|
||||
*/
|
||||
#if defined(__amd64__) || defined(__i386__)
|
||||
/* Long double constants are slow on these arches, and broken on i386. */
|
||||
static const volatile double
|
||||
S1hi = -0.16666666666666666, /* -0x15555555555555.0p-55 */
|
||||
S1lo = -9.2563760475949941e-18; /* -0x15580000000000.0p-109 */
|
||||
#define S1 ((long double)S1hi + S1lo)
|
||||
#else
|
||||
static const long double
|
||||
S1 = -0.166666666666666666671L; /* -0xaaaaaaaaaaaaaaab.0p-66 */
|
||||
#endif
|
||||
|
||||
static const double
|
||||
S2 = 0.0083333333333333332, /* 0x11111111111111.0p-59 */
|
||||
S3 = -0.00019841269841269427, /* -0x1a01a01a019f81.0p-65 */
|
||||
S4 = 0.0000027557319223597490, /* 0x171de3a55560f7.0p-71 */
|
||||
S5 = -0.000000025052108218074604, /* -0x1ae64564f16cad.0p-78 */
|
||||
S6 = 1.6059006598854211e-10, /* 0x161242b90243b5.0p-85 */
|
||||
S7 = -7.6429779983024564e-13, /* -0x1ae42ebd1b2e00.0p-93 */
|
||||
S8 = 2.6174587166648325e-15; /* 0x179372ea0b3f64.0p-101 */
|
||||
|
||||
long double
|
||||
__kernel_sinl(long double x, long double y, int iy)
|
||||
{
|
||||
long double z,r,v;
|
||||
|
||||
z = x*x;
|
||||
v = z*x;
|
||||
r = S2+z*(S3+z*(S4+z*(S5+z*(S6+z*(S7+z*S8)))));
|
||||
if(iy==0) return x+v*(S1+z*r);
|
||||
else return x-((z*(half*y-v*r)-y)-v*S1);
|
||||
}
|
||||
124
lib/msun/ld80/k_tanl.c
Normal file
124
lib/msun/ld80/k_tanl.c
Normal file
@ -0,0 +1,124 @@
|
||||
/* From: @(#)k_tan.c 1.5 04/04/22 SMI */
|
||||
|
||||
/*
|
||||
* ====================================================
|
||||
* Copyright 2004 Sun Microsystems, Inc. All Rights Reserved.
|
||||
* Copyright (c) 2008 Steven G. Kargl, David Schultz, Bruce D. Evans.
|
||||
*
|
||||
* Permission to use, copy, modify, and distribute this
|
||||
* software is freely granted, provided that this notice
|
||||
* is preserved.
|
||||
* ====================================================
|
||||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__FBSDID("$FreeBSD$");
|
||||
|
||||
/*
|
||||
* ld80 version of k_tan.c. See ../src/k_tan.c for most comments.
|
||||
*/
|
||||
|
||||
#include "math.h"
|
||||
#include "math_private.h"
|
||||
|
||||
/*
|
||||
* Domain [-0.67434, 0.67434], range ~[-2.25e-22, 1.921e-22]
|
||||
* |tan(x)/x - t(x)| < 2**-71.9
|
||||
*
|
||||
* See k_cosl.c for more details about the polynomial.
|
||||
*/
|
||||
#if defined(__amd64__) || defined(__i386__)
|
||||
/* Long double constants are slow on these arches, and broken on i386. */
|
||||
static const volatile double
|
||||
T3hi = 0.33333333333333331, /* 0x15555555555555.0p-54 */
|
||||
T3lo = 1.8350121769317163e-17, /* 0x15280000000000.0p-108 */
|
||||
T5hi = 0.13333333333333336, /* 0x11111111111112.0p-55 */
|
||||
T5lo = 1.3051083651294260e-17, /* 0x1e180000000000.0p-109 */
|
||||
T7hi = 0.053968253968250494, /* 0x1ba1ba1ba1b827.0p-57 */
|
||||
T7lo = 3.1509625637859973e-18, /* 0x1d100000000000.0p-111 */
|
||||
pio4_hi = 0.78539816339744828, /* 0x1921fb54442d18.0p-53 */
|
||||
pio4_lo = 3.0628711372715500e-17, /* 0x11a80000000000.0p-107 */
|
||||
pio4lo_hi = -1.2541394031670831e-20, /* -0x1d9cceba3f91f2.0p-119 */
|
||||
pio4lo_lo = 6.1493048227390915e-37; /* 0x1a280000000000.0p-173 */
|
||||
#define T3 ((long double)T3hi + T3lo)
|
||||
#define T5 ((long double)T5hi + T5lo)
|
||||
#define T7 ((long double)T7hi + T7lo)
|
||||
#define pio4 ((long double)pio4_hi + pio4_lo)
|
||||
#define pio4lo ((long double)pio4lo_hi + pio4lo_lo)
|
||||
#else
|
||||
static const long double
|
||||
T3 = 0.333333333333333333180L, /* 0xaaaaaaaaaaaaaaa5.0p-65 */
|
||||
T5 = 0.133333333333333372290L, /* 0x88888888888893c3.0p-66 */
|
||||
T7 = 0.0539682539682504975744L; /* 0xdd0dd0dd0dc13ba2.0p-68 */
|
||||
pio4 = 0.785398163397448309628, /* 0xc90fdaa22168c235.0p-64 */
|
||||
pio4lo = -1.25413940316708300586e-20; /* -0xece675d1fc8f8cbb.0p-130 */
|
||||
#endif
|
||||
|
||||
static const double
|
||||
T9 = 0.021869488536312216, /* 0x1664f4882cc1c2.0p-58 */
|
||||
T11 = 0.0088632355256619590, /* 0x1226e355c17612.0p-59 */
|
||||
T13 = 0.0035921281113786528, /* 0x1d6d3d185d7ff8.0p-61 */
|
||||
T15 = 0.0014558334756312418, /* 0x17da354aa3f96b.0p-62 */
|
||||
T17 = 0.00059003538700862256, /* 0x13559358685b83.0p-63 */
|
||||
T19 = 0.00023907843576635544, /* 0x1f56242026b5be.0p-65 */
|
||||
T21 = 0.000097154625656538905, /* 0x1977efc26806f4.0p-66 */
|
||||
T23 = 0.000038440165747303162, /* 0x14275a09b3ceac.0p-67 */
|
||||
T25 = 0.000018082171885432524, /* 0x12f5e563e5487e.0p-68 */
|
||||
T27 = 0.0000024196006108814377, /* 0x144c0d80cc6896.0p-71 */
|
||||
T29 = 0.0000078293456938132840, /* 0x106b59141a6cb3.0p-69 */
|
||||
T31 = -0.0000032609076735050182, /* -0x1b5abef3ba4b59.0p-71 */
|
||||
T33 = 0.0000023261313142559411; /* 0x13835436c0c87f.0p-71 */
|
||||
|
||||
long double
|
||||
__kernel_tanl(long double x, long double y, int iy) {
|
||||
long double z, r, v, w, s;
|
||||
long double osign;
|
||||
int i;
|
||||
|
||||
iy = (iy == 1 ? -1 : 1); /* XXX recover original interface */
|
||||
osign = (x >= 0 ? 1.0 : -1.0); /* XXX slow, probably wrong for -0 */
|
||||
if (fabsl(x) >= 0.67434) {
|
||||
if (x < 0) {
|
||||
x = -x;
|
||||
y = -y;
|
||||
}
|
||||
z = pio4 - x;
|
||||
w = pio4lo - y;
|
||||
x = z + w;
|
||||
y = 0.0;
|
||||
i = 1;
|
||||
} else
|
||||
i = 0;
|
||||
z = x * x;
|
||||
w = z * z;
|
||||
r = T5 + w * (T9 + w * (T13 + w * (T17 + w * (T21 +
|
||||
w * (T25 + w * (T29 + w * T33))))));
|
||||
v = z * (T7 + w * (T11 + w * (T15 + w * (T19 + w * (T23 +
|
||||
w * (T27 + w * T31))))));
|
||||
s = z * x;
|
||||
r = y + z * (s * (r + v) + y);
|
||||
r += T3 * s;
|
||||
w = x + r;
|
||||
if (i == 1) {
|
||||
v = (long double) iy;
|
||||
return osign *
|
||||
(v - 2.0 * (x - (w * w / (w + v) - r)));
|
||||
}
|
||||
if (iy == 1)
|
||||
return w;
|
||||
else {
|
||||
/*
|
||||
* if allow error up to 2 ulp, simply return
|
||||
* -1.0 / (x+r) here
|
||||
*/
|
||||
/* compute -1.0 / (x+r) accurately */
|
||||
long double a, t;
|
||||
z = w;
|
||||
z = z + 0x1p32 - 0x1p32;
|
||||
v = r - (z - x); /* z+v = r+x */
|
||||
t = a = -1.0 / w; /* a = -1.0/w */
|
||||
t = t + 0x1p32 - 0x1p32;
|
||||
s = 1.0 + t * z;
|
||||
return t + a * (s + t * v);
|
||||
}
|
||||
}
|
||||
Loading…
Reference in New Issue
Block a user