pow.c revision 25c28e83beb90e7c80452a7c818c5e6f73a07dc8
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
*/
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#if defined(ELFOBJ)
#pragma weak pow = __pow
#endif
/*
* pow(x,y) return x**y
* n
* Method: Let x = 2 * (1+f)
* 1. Compute and return log2(x) in two pieces:
* log2(x) = w1 + w2,
* where w1 has 24 bits trailing zero.
* 2. Perform y*log2(x) by simulating muti-precision arithmetic
* 3. Return x**y = exp2(y*log(x))
*
* Special cases:
* 1. (anything) ** +-0 is 1
* 1'. 1 ** (anything) is 1 (C99; 1 ** +-INF/NAN used to be NAN)
* 2. (anything) ** 1 is itself
* 3. (anything except 1) ** NAN is NAN ("except 1" is C99)
* 4. NAN ** (anything except 0) is NAN
* 5. +-(|x| > 1) ** +INF is +INF
* 6. +-(|x| > 1) ** -INF is +0
* 7. +-(|x| < 1) ** +INF is +0
* 8. +-(|x| < 1) ** -INF is +INF
* 9. -1 ** +-INF is 1 (C99; -1 ** +-INF used to be NAN)
* 10. +0 ** (+anything except 0, NAN) is +0
* 11. -0 ** (+anything except 0, NAN, odd integer) is +0
* 12. +0 ** (-anything except 0, NAN) is +INF
* 13. -0 ** (-anything except 0, NAN, odd integer) is +INF
* 14. -0 ** (odd integer) = -( +0 ** (odd integer) )
* 15. +INF ** (+anything except 0,NAN) is +INF
* 16. +INF ** (-anything except 0,NAN) is +0
* 17. -INF ** (anything) = -0 ** (-anything)
* 18. (-anything) ** (integer) is (-1)**(integer)*(+anything**integer)
* 19. (-anything except 0 and inf) ** (non-integer) is NAN
*
* Accuracy:
* pow(x,y) returns x**y nearly rounded. In particular
* pow(integer,integer)
* always returns the correct integer provided it is representable.
*/
#include "libm.h"
#include "xpg6.h" /* __xpg6 */
#define _C99SUSv3_pow _C99SUSv3_pow_treats_Inf_as_an_even_int
static const double zero = 0.0, one = 1.0, two = 2.0;
extern const double _TBL_log2_hi[], _TBL_log2_lo[];
static const double
two53 = 9007199254740992.0,
A1_hi = 2.8853900432586669921875,
A1_lo = 3.8519259825035041963606002e-8,
A1 = 2.885390081777926817222541963606002026086e+0000,
A2 = 9.617966939207270828380543979852286255862e-0001,
A3 = 5.770807680887875964868853124873696201995e-0001,
B0_hi = 2.8853900432586669921875,
B0_lo = 3.8519259822532793056374320585e-8,
B0 = 2.885390081777926814720293056374320585689e+0000,
B1 = 9.617966939259755138949202350396200257632e-0001,
B2 = 5.770780163585687000782112776448797953382e-0001,
B3 = 4.121985488948771523290174512461778354953e-0001,
B4 = 3.207590534812432970433641789022666850193e-0001;
static double
log2_x(double x, double *w) {
double f, s, z, qn, h, t;
int *px = (int *) &x;
int *pz = (int *) &z;
int i, j, ix, n;
n = 0;
ix = px[HIWORD];
if (ix >= 0x3fef03f1 && ix < 0x3ff08208) { /* 65/63 > x > 63/65 */
double f1, v;
f = x - one;
if (((ix - 0x3ff00000) | px[LOWORD]) == 0) {
*w = zero;
return (zero); /* log2(1)= +0 */
}
qn = one / (two + f);
s = f * qn; /* |s|<2**-6 */
v = s * s;
h = (double) ((float) s);
f1 = (double) ((float) f);
t = qn * (((f - two * h) - h * f1) - h * (f - f1));
/* s = h+t */
f1 = h * B0_lo + s * (v * (B1 + v * (B2 + v * (B3 + v * B4))));
t = f1 + t * B0;
h *= B0_hi;
s = (double) ((float) (h + t));
*w = t - (s - h);
return (s);
}
if (ix < 0x00100000) { /* subnormal x */
x *= two53;
n = -53;
ix = px[HIWORD];
}
/* LARGE N */
n += ((ix + 0x1000) >> 20) - 0x3ff;
ix = (ix & 0x000fffff) | 0x3ff00000; /* scale x to [1,2] */
px[HIWORD] = ix;
i = ix + 0x1000;
pz[HIWORD] = i & 0xffffe000;
pz[LOWORD] = 0;
qn = one / (x + z);
f = x - z;
s = f * qn;
h = (double) ((float) s);
t = qn * ((f - (h + h) * z) - h * f);
j = (i >> 13) & 0x7f;
f = s * s;
t = t * A1 + h * A1_lo;
t += (s * f) * (A2 + f * A3);
qn = h * A1_hi;
s = n + _TBL_log2_hi[j];
h = qn + s;
t += _TBL_log2_lo[j] - ((h - s) - qn);
f = (double) ((float) (h + t));
*w = t - (f - h);
return (f);
}
extern const double _TBL_exp2_hi[], _TBL_exp2_lo[];
static const double /* poly app of 2^x-1 on [-1e-10,2^-7+1e-10] */
E1 = 6.931471805599453100674958533810346197328e-0001,
E2 = 2.402265069587779347846769151717493815979e-0001,
E3 = 5.550410866475410512631124892773937864699e-0002,
E4 = 9.618143209991026824853712740162451423355e-0003,
E5 = 1.333357676549940345096774122231849082991e-0003;
double
pow(double x, double y) {
double z, ax;
double y1, y2, w1, w2;
int sbx, sby, j, k, yisint;
int hx, hy, ahx, ahy;
unsigned lx, ly;
int *pz = (int *) &z;
hx = ((int *) &x)[HIWORD];
lx = ((unsigned *) &x)[LOWORD];
hy = ((int *) &y)[HIWORD];
ly = ((unsigned *) &y)[LOWORD];
ahx = hx & ~0x80000000;
ahy = hy & ~0x80000000;
if ((ahy | ly) == 0) { /* y==zero */
if ((ahx | lx) == 0)
z = _SVID_libm_err(x, y, 20); /* +-0**+-0 */
else if ((ahx | (((lx | -lx) >> 31) & 1)) > 0x7ff00000)
z = _SVID_libm_err(x, y, 42); /* NaN**+-0 */
else
z = one; /* x**+-0 = 1 */
return (z);
} else if (hx == 0x3ff00000 && lx == 0 &&
(__xpg6 & _C99SUSv3_pow) != 0)
return (one); /* C99: 1**anything = 1 */
else if (ahx > 0x7ff00000 || (ahx == 0x7ff00000 && lx != 0) ||
ahy > 0x7ff00000 || (ahy == 0x7ff00000 && ly != 0))
return (x * y); /* +-NaN return x*y; + -> * for Cheetah */
/* includes Sun: 1**NaN = NaN */
sbx = (unsigned) hx >> 31;
sby = (unsigned) hy >> 31;
ax = fabs(x);
/*
* determine if y is an odd int when x < 0
* yisint = 0 ... y is not an integer
* yisint = 1 ... y is an odd int
* yisint = 2 ... y is an even int
*/
yisint = 0;
if (sbx) {
if (ahy >= 0x43400000)
yisint = 2; /* even integer y */
else if (ahy >= 0x3ff00000) {
k = (ahy >> 20) - 0x3ff; /* exponent */
if (k > 20) {
j = ly >> (52 - k);
if ((j << (52 - k)) == ly)
yisint = 2 - (j & 1);
} else if (ly == 0) {
j = ahy >> (20 - k);
if ((j << (20 - k)) == ahy)
yisint = 2 - (j & 1);
}
}
}
/* special value of y */
if (ly == 0) {
if (ahy == 0x7ff00000) { /* y is +-inf */
if (((ahx - 0x3ff00000) | lx) == 0) {
if ((__xpg6 & _C99SUSv3_pow) != 0)
return (one);
/* C99: (-1)**+-inf = 1 */
else
return (y - y);
/* Sun: (+-1)**+-inf = NaN */
} else if (ahx >= 0x3ff00000)
/* (|x|>1)**+,-inf = inf,0 */
return (sby == 0 ? y : zero);
else /* (|x|<1)**-,+inf = inf,0 */
return (sby != 0 ? -y : zero);
}
if (ahy == 0x3ff00000) { /* y is +-1 */
if (sby != 0) { /* y is -1 */
if (x == zero) /* divided by zero */
return (_SVID_libm_err(x, y, 23));
else if (ahx < 0x40000 || ((ahx - 0x40000) |
lx) == 0) /* overflow */
return (_SVID_libm_err(x, y, 21));
else
return (one / x);
} else
return (x);
}
if (hy == 0x40000000) { /* y is 2 */
if (ahx >= 0x5ff00000 && ahx < 0x7ff00000)
return (_SVID_libm_err(x, y, 21));
/* x*x overflow */
else if ((ahx < 0x1e56a09e && (ahx | lx) != 0) ||
(ahx == 0x1e56a09e && lx < 0x667f3bcd))
return (_SVID_libm_err(x, y, 22));
/* x*x underflow */
else
return (x * x);
}
if (hy == 0x3fe00000) {
if (!((ahx | lx) == 0 || ((ahx - 0x7ff00000) | lx) ==
0 || sbx == 1))
return (sqrt(x)); /* y is 0.5 and x > 0 */
}
}
/* special value of x */
if (lx == 0) {
if (ahx == 0x7ff00000 || ahx == 0 || ahx == 0x3ff00000) {
/* x is +-0,+-inf,-1 */
z = ax;
if (sby == 1) {
z = one / z; /* z = |x|**y */
if (ahx == 0)
return (_SVID_libm_err(x, y, 23));
}
if (sbx == 1) {
if (ahx == 0x3ff00000 && yisint == 0)
z = _SVID_libm_err(x, y, 24);
/* neg**non-integral is NaN + invalid */
else if (yisint == 1)
z = -z; /* (x<0)**odd = -(|x|**odd) */
}
return (z);
}
}
/* (x<0)**(non-int) is NaN */
if (sbx == 1 && yisint == 0)
return (_SVID_libm_err(x, y, 24));
/* Now ax is finite, y is finite */
/* first compute log2(ax) = w1+w2, with 24 bits w1 */
w1 = log2_x(ax, &w2);
/* split up y into y1+y2 and compute (y1+y2)*(w1+w2) */
if (((ly & 0x07ffffff) == 0) || ahy >= 0x47e00000 ||
ahy <= 0x38100000) {
/* no need to split if y is short or too large or too small */
y1 = y * w1;
y2 = y * w2;
} else {
y1 = (double) ((float) y);
y2 = (y - y1) * w1 + y * w2;
y1 *= w1;
}
z = y1 + y2;
j = pz[HIWORD];
if (j >= 0x40900000) { /* z >= 1024 */
if (!(j == 0x40900000 && pz[LOWORD] == 0)) /* z > 1024 */
return (_SVID_libm_err(x, y, 21)); /* overflow */
else {
w2 = y1 - z;
w2 += y2;
/* rounded to inf */
if (w2 >= -8.008566259537296567160e-17)
return (_SVID_libm_err(x, y, 21));
/* overflow */
}
} else if ((j & ~0x80000000) >= 0x4090cc00) { /* z <= -1075 */
if (!(j == 0xc090cc00 && pz[LOWORD] == 0)) /* z < -1075 */
return (_SVID_libm_err(x, y, 22)); /* underflow */
else {
w2 = y1 - z;
w2 += y2;
if (w2 <= zero) /* underflow */
return (_SVID_libm_err(x, y, 22));
}
}
/*
* compute 2**(k+f[j]+g)
*/
k = (int) (z * 64.0 + (((hy ^ (ahx - 0x3ff00000)) > 0) ? 0.5 : -0.5));
j = k & 63;
w1 = y2 - ((double) k * 0.015625 - y1);
w2 = _TBL_exp2_hi[j];
z = _TBL_exp2_lo[j] + (w2 * w1) * (E1 + w1 * (E2 + w1 * (E3 + w1 *
(E4 + w1 * E5))));
z += w2;
k >>= 6;
if (k < -1021)
z = scalbn(z, k);
else /* subnormal output */
pz[HIWORD] += k << 20;
if (sbx == 1 && yisint == 1)
z = -z; /* (-ve)**(odd int) */
return (z);
}