mirror of
https://github.com/jart/cosmopolitan.git
synced 2025-06-27 14:58:30 +00:00
Make quality improvements
- Write some more unit tests - memcpy() on ARM is now faster - Address the Musl complex math FIXME comments - Some libm funcs like pow() now support setting errno - Import the latest and greatest math functions from ARM - Use more accurate atan2f() and log1pf() implementations - atoi() and atol() will no longer saturate or clobber errno
This commit is contained in:
Justine Tunney
parent
af8f2bd19f
commit
592f6ebc20
122 changed files with 6305 additions and 3859 deletions
|
|
@ -1,9 +1,9 @@
|
|||
/*-*- mode:c;indent-tabs-mode:t;c-basic-offset:8;tab-width:8;coding:utf-8 -*-│
|
||||
│ vi: set noet ft=c ts=8 sw=8 fenc=utf-8 :vi │
|
||||
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
|
||||
│ vi: set et ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi │
|
||||
╚──────────────────────────────────────────────────────────────────────────────╝
|
||||
│ │
|
||||
│ Musl Libc │
|
||||
│ Copyright © 2005-2014 Rich Felker, et al. │
|
||||
│ Optimized Routines │
|
||||
│ Copyright (c) 2018-2024, Arm Limited. │
|
||||
│ │
|
||||
│ Permission is hereby granted, free of charge, to any person obtaining │
|
||||
│ a copy of this software and associated documentation files (the │
|
||||
|
|
@ -25,310 +25,247 @@
|
|||
│ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. │
|
||||
│ │
|
||||
╚─────────────────────────────────────────────────────────────────────────────*/
|
||||
#include "libc/math.h"
|
||||
__static_yoink("musl_libc_notice");
|
||||
__static_yoink("fdlibm_notice");
|
||||
#include "libc/tinymath/arm.internal.h"
|
||||
__static_yoink("arm_optimized_routines_notice");
|
||||
|
||||
/* origin: FreeBSD /usr/src/lib/msun/src/s_erf.c */
|
||||
/*
|
||||
* ====================================================
|
||||
* 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.
|
||||
* ====================================================
|
||||
*/
|
||||
/* double erf(double x)
|
||||
* double erfc(double x)
|
||||
* x
|
||||
* 2 |\
|
||||
* erf(x) = --------- | exp(-t*t)dt
|
||||
* sqrt(pi) \|
|
||||
* 0
|
||||
*
|
||||
* erfc(x) = 1-erf(x)
|
||||
* Note that
|
||||
* erf(-x) = -erf(x)
|
||||
* erfc(-x) = 2 - erfc(x)
|
||||
*
|
||||
* Method:
|
||||
* 1. For |x| in [0, 0.84375]
|
||||
* erf(x) = x + x*R(x^2)
|
||||
* erfc(x) = 1 - erf(x) if x in [-.84375,0.25]
|
||||
* = 0.5 + ((0.5-x)-x*R) if x in [0.25,0.84375]
|
||||
* where R = P/Q where P is an odd poly of degree 8 and
|
||||
* Q is an odd poly of degree 10.
|
||||
* -57.90
|
||||
* | R - (erf(x)-x)/x | <= 2
|
||||
*
|
||||
*
|
||||
* Remark. The formula is derived by noting
|
||||
* erf(x) = (2/sqrt(pi))*(x - x^3/3 + x^5/10 - x^7/42 + ....)
|
||||
* and that
|
||||
* 2/sqrt(pi) = 1.128379167095512573896158903121545171688
|
||||
* is close to one. The interval is chosen because the fix
|
||||
* point of erf(x) is near 0.6174 (i.e., erf(x)=x when x is
|
||||
* near 0.6174), and by some experiment, 0.84375 is chosen to
|
||||
* guarantee the error is less than one ulp for erf.
|
||||
*
|
||||
* 2. For |x| in [0.84375,1.25], let s = |x| - 1, and
|
||||
* c = 0.84506291151 rounded to single (24 bits)
|
||||
* erf(x) = sign(x) * (c + P1(s)/Q1(s))
|
||||
* erfc(x) = (1-c) - P1(s)/Q1(s) if x > 0
|
||||
* 1+(c+P1(s)/Q1(s)) if x < 0
|
||||
* |P1/Q1 - (erf(|x|)-c)| <= 2**-59.06
|
||||
* Remark: here we use the taylor series expansion at x=1.
|
||||
* erf(1+s) = erf(1) + s*Poly(s)
|
||||
* = 0.845.. + P1(s)/Q1(s)
|
||||
* That is, we use rational approximation to approximate
|
||||
* erf(1+s) - (c = (single)0.84506291151)
|
||||
* Note that |P1/Q1|< 0.078 for x in [0.84375,1.25]
|
||||
* where
|
||||
* P1(s) = degree 6 poly in s
|
||||
* Q1(s) = degree 6 poly in s
|
||||
*
|
||||
* 3. For x in [1.25,1/0.35(~2.857143)],
|
||||
* erfc(x) = (1/x)*exp(-x*x-0.5625+R1/S1)
|
||||
* erf(x) = 1 - erfc(x)
|
||||
* where
|
||||
* R1(z) = degree 7 poly in z, (z=1/x^2)
|
||||
* S1(z) = degree 8 poly in z
|
||||
*
|
||||
* 4. For x in [1/0.35,28]
|
||||
* erfc(x) = (1/x)*exp(-x*x-0.5625+R2/S2) if x > 0
|
||||
* = 2.0 - (1/x)*exp(-x*x-0.5625+R2/S2) if -6<x<0
|
||||
* = 2.0 - tiny (if x <= -6)
|
||||
* erf(x) = sign(x)*(1.0 - erfc(x)) if x < 6, else
|
||||
* erf(x) = sign(x)*(1.0 - tiny)
|
||||
* where
|
||||
* R2(z) = degree 6 poly in z, (z=1/x^2)
|
||||
* S2(z) = degree 7 poly in z
|
||||
*
|
||||
* Note1:
|
||||
* To compute exp(-x*x-0.5625+R/S), let s be a single
|
||||
* precision number and s := x; then
|
||||
* -x*x = -s*s + (s-x)*(s+x)
|
||||
* exp(-x*x-0.5626+R/S) =
|
||||
* exp(-s*s-0.5625)*exp((s-x)*(s+x)+R/S);
|
||||
* Note2:
|
||||
* Here 4 and 5 make use of the asymptotic series
|
||||
* exp(-x*x)
|
||||
* erfc(x) ~ ---------- * ( 1 + Poly(1/x^2) )
|
||||
* x*sqrt(pi)
|
||||
* We use rational approximation to approximate
|
||||
* g(s)=f(1/x^2) = log(erfc(x)*x) - x*x + 0.5625
|
||||
* Here is the error bound for R1/S1 and R2/S2
|
||||
* |R1/S1 - f(x)| < 2**(-62.57)
|
||||
* |R2/S2 - f(x)| < 2**(-61.52)
|
||||
*
|
||||
* 5. For inf > x >= 28
|
||||
* erf(x) = sign(x) *(1 - tiny) (raise inexact)
|
||||
* erfc(x) = tiny*tiny (raise underflow) if x > 0
|
||||
* = 2 - tiny if x<0
|
||||
*
|
||||
* 7. Special case:
|
||||
* erf(0) = 0, erf(inf) = 1, erf(-inf) = -1,
|
||||
* erfc(0) = 1, erfc(inf) = 0, erfc(-inf) = 2,
|
||||
* erfc/erf(NaN) is NaN
|
||||
*/
|
||||
#define TwoOverSqrtPiMinusOne 0x1.06eba8214db69p-3
|
||||
#define C 0x1.b0ac16p-1
|
||||
#define PA __erf_data.erf_poly_A
|
||||
#define NA __erf_data.erf_ratio_N_A
|
||||
#define DA __erf_data.erf_ratio_D_A
|
||||
#define NB __erf_data.erf_ratio_N_B
|
||||
#define DB __erf_data.erf_ratio_D_B
|
||||
#define PC __erf_data.erfc_poly_C
|
||||
#define PD __erf_data.erfc_poly_D
|
||||
#define PE __erf_data.erfc_poly_E
|
||||
#define PF __erf_data.erfc_poly_F
|
||||
|
||||
static const double
|
||||
erx = 8.45062911510467529297e-01, /* 0x3FEB0AC1, 0x60000000 */
|
||||
/*
|
||||
* Coefficients for approximation to erf on [0,0.84375]
|
||||
*/
|
||||
efx8 = 1.02703333676410069053e+00, /* 0x3FF06EBA, 0x8214DB69 */
|
||||
pp0 = 1.28379167095512558561e-01, /* 0x3FC06EBA, 0x8214DB68 */
|
||||
pp1 = -3.25042107247001499370e-01, /* 0xBFD4CD7D, 0x691CB913 */
|
||||
pp2 = -2.84817495755985104766e-02, /* 0xBF9D2A51, 0xDBD7194F */
|
||||
pp3 = -5.77027029648944159157e-03, /* 0xBF77A291, 0x236668E4 */
|
||||
pp4 = -2.37630166566501626084e-05, /* 0xBEF8EAD6, 0x120016AC */
|
||||
qq1 = 3.97917223959155352819e-01, /* 0x3FD97779, 0xCDDADC09 */
|
||||
qq2 = 6.50222499887672944485e-02, /* 0x3FB0A54C, 0x5536CEBA */
|
||||
qq3 = 5.08130628187576562776e-03, /* 0x3F74D022, 0xC4D36B0F */
|
||||
qq4 = 1.32494738004321644526e-04, /* 0x3F215DC9, 0x221C1A10 */
|
||||
qq5 = -3.96022827877536812320e-06, /* 0xBED09C43, 0x42A26120 */
|
||||
/*
|
||||
* Coefficients for approximation to erf in [0.84375,1.25]
|
||||
*/
|
||||
pa0 = -2.36211856075265944077e-03, /* 0xBF6359B8, 0xBEF77538 */
|
||||
pa1 = 4.14856118683748331666e-01, /* 0x3FDA8D00, 0xAD92B34D */
|
||||
pa2 = -3.72207876035701323847e-01, /* 0xBFD7D240, 0xFBB8C3F1 */
|
||||
pa3 = 3.18346619901161753674e-01, /* 0x3FD45FCA, 0x805120E4 */
|
||||
pa4 = -1.10894694282396677476e-01, /* 0xBFBC6398, 0x3D3E28EC */
|
||||
pa5 = 3.54783043256182359371e-02, /* 0x3FA22A36, 0x599795EB */
|
||||
pa6 = -2.16637559486879084300e-03, /* 0xBF61BF38, 0x0A96073F */
|
||||
qa1 = 1.06420880400844228286e-01, /* 0x3FBB3E66, 0x18EEE323 */
|
||||
qa2 = 5.40397917702171048937e-01, /* 0x3FE14AF0, 0x92EB6F33 */
|
||||
qa3 = 7.18286544141962662868e-02, /* 0x3FB2635C, 0xD99FE9A7 */
|
||||
qa4 = 1.26171219808761642112e-01, /* 0x3FC02660, 0xE763351F */
|
||||
qa5 = 1.36370839120290507362e-02, /* 0x3F8BEDC2, 0x6B51DD1C */
|
||||
qa6 = 1.19844998467991074170e-02, /* 0x3F888B54, 0x5735151D */
|
||||
/*
|
||||
* Coefficients for approximation to erfc in [1.25,1/0.35]
|
||||
*/
|
||||
ra0 = -9.86494403484714822705e-03, /* 0xBF843412, 0x600D6435 */
|
||||
ra1 = -6.93858572707181764372e-01, /* 0xBFE63416, 0xE4BA7360 */
|
||||
ra2 = -1.05586262253232909814e+01, /* 0xC0251E04, 0x41B0E726 */
|
||||
ra3 = -6.23753324503260060396e+01, /* 0xC04F300A, 0xE4CBA38D */
|
||||
ra4 = -1.62396669462573470355e+02, /* 0xC0644CB1, 0x84282266 */
|
||||
ra5 = -1.84605092906711035994e+02, /* 0xC067135C, 0xEBCCABB2 */
|
||||
ra6 = -8.12874355063065934246e+01, /* 0xC0545265, 0x57E4D2F2 */
|
||||
ra7 = -9.81432934416914548592e+00, /* 0xC023A0EF, 0xC69AC25C */
|
||||
sa1 = 1.96512716674392571292e+01, /* 0x4033A6B9, 0xBD707687 */
|
||||
sa2 = 1.37657754143519042600e+02, /* 0x4061350C, 0x526AE721 */
|
||||
sa3 = 4.34565877475229228821e+02, /* 0x407B290D, 0xD58A1A71 */
|
||||
sa4 = 6.45387271733267880336e+02, /* 0x40842B19, 0x21EC2868 */
|
||||
sa5 = 4.29008140027567833386e+02, /* 0x407AD021, 0x57700314 */
|
||||
sa6 = 1.08635005541779435134e+02, /* 0x405B28A3, 0xEE48AE2C */
|
||||
sa7 = 6.57024977031928170135e+00, /* 0x401A47EF, 0x8E484A93 */
|
||||
sa8 = -6.04244152148580987438e-02, /* 0xBFAEEFF2, 0xEE749A62 */
|
||||
/*
|
||||
* Coefficients for approximation to erfc in [1/.35,28]
|
||||
*/
|
||||
rb0 = -9.86494292470009928597e-03, /* 0xBF843412, 0x39E86F4A */
|
||||
rb1 = -7.99283237680523006574e-01, /* 0xBFE993BA, 0x70C285DE */
|
||||
rb2 = -1.77579549177547519889e+01, /* 0xC031C209, 0x555F995A */
|
||||
rb3 = -1.60636384855821916062e+02, /* 0xC064145D, 0x43C5ED98 */
|
||||
rb4 = -6.37566443368389627722e+02, /* 0xC083EC88, 0x1375F228 */
|
||||
rb5 = -1.02509513161107724954e+03, /* 0xC0900461, 0x6A2E5992 */
|
||||
rb6 = -4.83519191608651397019e+02, /* 0xC07E384E, 0x9BDC383F */
|
||||
sb1 = 3.03380607434824582924e+01, /* 0x403E568B, 0x261D5190 */
|
||||
sb2 = 3.25792512996573918826e+02, /* 0x40745CAE, 0x221B9F0A */
|
||||
sb3 = 1.53672958608443695994e+03, /* 0x409802EB, 0x189D5118 */
|
||||
sb4 = 3.19985821950859553908e+03, /* 0x40A8FFB7, 0x688C246A */
|
||||
sb5 = 2.55305040643316442583e+03, /* 0x40A3F219, 0xCEDF3BE6 */
|
||||
sb6 = 4.74528541206955367215e+02, /* 0x407DA874, 0xE79FE763 */
|
||||
sb7 = -2.24409524465858183362e+01; /* 0xC03670E2, 0x42712D62 */
|
||||
|
||||
#define asuint(f) ((union{float _f; uint32_t _i;}){f})._i
|
||||
#define asfloat(i) ((union{uint32_t _i; float _f;}){i})._f
|
||||
#define asuint64(f) ((union{double _f; uint64_t _i;}){f})._i
|
||||
#define asdouble(i) ((union{uint64_t _i; double _f;}){i})._f
|
||||
#define INSERT_WORDS(d,hi,lo) \
|
||||
do { \
|
||||
(d) = asdouble(((uint64_t)(hi)<<32) | (uint32_t)(lo)); \
|
||||
} while (0)
|
||||
#define GET_HIGH_WORD(hi,d) \
|
||||
do { \
|
||||
(hi) = asuint64(d) >> 32; \
|
||||
} while (0)
|
||||
#define GET_LOW_WORD(lo,d) \
|
||||
do { \
|
||||
(lo) = (uint32_t)asuint64(d); \
|
||||
} while (0)
|
||||
#define SET_HIGH_WORD(d,hi) \
|
||||
INSERT_WORDS(d, hi, (uint32_t)asuint64(d))
|
||||
#define SET_LOW_WORD(d,lo) \
|
||||
INSERT_WORDS(d, asuint64(d)>>32, lo)
|
||||
|
||||
static double erfc1(double x)
|
||||
/* Top 32 bits of a double. */
|
||||
static inline uint32_t
|
||||
top32 (double x)
|
||||
{
|
||||
double_t s,P,Q;
|
||||
|
||||
s = fabs(x) - 1;
|
||||
P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
|
||||
Q = 1+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
|
||||
return 1 - erx - P/Q;
|
||||
}
|
||||
|
||||
static double erfc2(uint32_t ix, double x)
|
||||
{
|
||||
double_t s,R,S;
|
||||
double z;
|
||||
|
||||
if (ix < 0x3ff40000) /* |x| < 1.25 */
|
||||
return erfc1(x);
|
||||
|
||||
x = fabs(x);
|
||||
s = 1/(x*x);
|
||||
if (ix < 0x4006db6d) { /* |x| < 1/.35 ~ 2.85714 */
|
||||
R = ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(
|
||||
ra5+s*(ra6+s*ra7))))));
|
||||
S = 1.0+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(
|
||||
sa5+s*(sa6+s*(sa7+s*sa8)))))));
|
||||
} else { /* |x| > 1/.35 */
|
||||
R = rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(
|
||||
rb5+s*rb6)))));
|
||||
S = 1.0+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(
|
||||
sb5+s*(sb6+s*sb7))))));
|
||||
}
|
||||
z = x;
|
||||
SET_LOW_WORD(z,0);
|
||||
return exp(-z*z-0.5625)*exp((z-x)*(z+x)+R/S)/x;
|
||||
return asuint64 (x) >> 32;
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns error function of 𝑥.
|
||||
* Returns error function of x.
|
||||
*
|
||||
* Highest measured error is 1.01 ULPs at 0x1.39956ac43382fp+0.
|
||||
*
|
||||
* @raise ERANGE on underflow
|
||||
*/
|
||||
double erf(double x)
|
||||
double
|
||||
erf (double x)
|
||||
{
|
||||
double r,s,z,y;
|
||||
uint32_t ix;
|
||||
int sign;
|
||||
/* Get top word and sign. */
|
||||
uint32_t ix = top32 (x);
|
||||
uint32_t ia = ix & 0x7fffffff;
|
||||
uint32_t sign = ix >> 31;
|
||||
|
||||
GET_HIGH_WORD(ix, x);
|
||||
sign = ix>>31;
|
||||
ix &= 0x7fffffff;
|
||||
if (ix >= 0x7ff00000) {
|
||||
/* erf(nan)=nan, erf(+-inf)=+-1 */
|
||||
return 1-2*sign + 1/x;
|
||||
/* Normalized and subnormal cases */
|
||||
if (ia < 0x3feb0000)
|
||||
{ /* a = |x| < 0.84375. */
|
||||
|
||||
if (ia < 0x3e300000)
|
||||
{ /* a < 2^(-28). */
|
||||
if (ia < 0x00800000)
|
||||
{ /* a < 2^(-1015). */
|
||||
double y = fma (TwoOverSqrtPiMinusOne, x, x);
|
||||
return check_uflow (y);
|
||||
}
|
||||
return x + TwoOverSqrtPiMinusOne * x;
|
||||
}
|
||||
if (ix < 0x3feb0000) { /* |x| < 0.84375 */
|
||||
if (ix < 0x3e300000) { /* |x| < 2**-28 */
|
||||
/* avoid underflow */
|
||||
return 0.125*(8*x + efx8*x);
|
||||
}
|
||||
z = x*x;
|
||||
r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
|
||||
s = 1.0+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
|
||||
y = r/s;
|
||||
return x + x*y;
|
||||
|
||||
double x2 = x * x;
|
||||
|
||||
if (ia < 0x3fe00000)
|
||||
{ /* a < 0.5 - Use polynomial approximation. */
|
||||
double r1 = fma (x2, PA[1], PA[0]);
|
||||
double r2 = fma (x2, PA[3], PA[2]);
|
||||
double r3 = fma (x2, PA[5], PA[4]);
|
||||
double r4 = fma (x2, PA[7], PA[6]);
|
||||
double r5 = fma (x2, PA[9], PA[8]);
|
||||
double x4 = x2 * x2;
|
||||
double r = r5;
|
||||
r = fma (x4, r, r4);
|
||||
r = fma (x4, r, r3);
|
||||
r = fma (x4, r, r2);
|
||||
r = fma (x4, r, r1);
|
||||
return fma (r, x, x); /* This fma is crucial for accuracy. */
|
||||
}
|
||||
if (ix < 0x40180000) /* 0.84375 <= |x| < 6 */
|
||||
y = 1 - erfc2(ix,x);
|
||||
else
|
||||
y = 1 - 0x1p-1022;
|
||||
return sign ? -y : y;
|
||||
else
|
||||
{ /* 0.5 <= a < 0.84375 - Use rational approximation. */
|
||||
double x4, x8, r1n, r2n, r1d, r2d, r3d;
|
||||
|
||||
r1n = fma (x2, NA[1], NA[0]);
|
||||
x4 = x2 * x2;
|
||||
r2n = fma (x2, NA[3], NA[2]);
|
||||
x8 = x4 * x4;
|
||||
r1d = fma (x2, DA[0], 1.0);
|
||||
r2d = fma (x2, DA[2], DA[1]);
|
||||
r3d = fma (x2, DA[4], DA[3]);
|
||||
double P = r1n + x4 * r2n + x8 * NA[4];
|
||||
double Q = r1d + x4 * r2d + x8 * r3d;
|
||||
return fma (P / Q, x, x);
|
||||
}
|
||||
}
|
||||
else if (ia < 0x3ff40000)
|
||||
{ /* 0.84375 <= |x| < 1.25. */
|
||||
double a2, a4, a6, r1n, r2n, r3n, r4n, r1d, r2d, r3d, r4d;
|
||||
double a = fabs (x) - 1.0;
|
||||
r1n = fma (a, NB[1], NB[0]);
|
||||
a2 = a * a;
|
||||
r1d = fma (a, DB[0], 1.0);
|
||||
a4 = a2 * a2;
|
||||
r2n = fma (a, NB[3], NB[2]);
|
||||
a6 = a4 * a2;
|
||||
r2d = fma (a, DB[2], DB[1]);
|
||||
r3n = fma (a, NB[5], NB[4]);
|
||||
r3d = fma (a, DB[4], DB[3]);
|
||||
r4n = NB[6];
|
||||
r4d = DB[5];
|
||||
double P = r1n + a2 * r2n + a4 * r3n + a6 * r4n;
|
||||
double Q = r1d + a2 * r2d + a4 * r3d + a6 * r4d;
|
||||
if (sign)
|
||||
return -C - P / Q;
|
||||
else
|
||||
return C + P / Q;
|
||||
}
|
||||
else if (ia < 0x40000000)
|
||||
{ /* 1.25 <= |x| < 2.0. */
|
||||
double a = fabs (x);
|
||||
a = a - 1.25;
|
||||
|
||||
double r1 = fma (a, PC[1], PC[0]);
|
||||
double r2 = fma (a, PC[3], PC[2]);
|
||||
double r3 = fma (a, PC[5], PC[4]);
|
||||
double r4 = fma (a, PC[7], PC[6]);
|
||||
double r5 = fma (a, PC[9], PC[8]);
|
||||
double r6 = fma (a, PC[11], PC[10]);
|
||||
double r7 = fma (a, PC[13], PC[12]);
|
||||
double r8 = fma (a, PC[15], PC[14]);
|
||||
|
||||
double a2 = a * a;
|
||||
|
||||
double r = r8;
|
||||
r = fma (a2, r, r7);
|
||||
r = fma (a2, r, r6);
|
||||
r = fma (a2, r, r5);
|
||||
r = fma (a2, r, r4);
|
||||
r = fma (a2, r, r3);
|
||||
r = fma (a2, r, r2);
|
||||
r = fma (a2, r, r1);
|
||||
|
||||
if (sign)
|
||||
return -1.0 + r;
|
||||
else
|
||||
return 1.0 - r;
|
||||
}
|
||||
else if (ia < 0x400a0000)
|
||||
{ /* 2 <= |x| < 3.25. */
|
||||
double a = fabs (x);
|
||||
a = fma (0.5, a, -1.0);
|
||||
|
||||
double r1 = fma (a, PD[1], PD[0]);
|
||||
double r2 = fma (a, PD[3], PD[2]);
|
||||
double r3 = fma (a, PD[5], PD[4]);
|
||||
double r4 = fma (a, PD[7], PD[6]);
|
||||
double r5 = fma (a, PD[9], PD[8]);
|
||||
double r6 = fma (a, PD[11], PD[10]);
|
||||
double r7 = fma (a, PD[13], PD[12]);
|
||||
double r8 = fma (a, PD[15], PD[14]);
|
||||
double r9 = fma (a, PD[17], PD[16]);
|
||||
|
||||
double a2 = a * a;
|
||||
|
||||
double r = r9;
|
||||
r = fma (a2, r, r8);
|
||||
r = fma (a2, r, r7);
|
||||
r = fma (a2, r, r6);
|
||||
r = fma (a2, r, r5);
|
||||
r = fma (a2, r, r4);
|
||||
r = fma (a2, r, r3);
|
||||
r = fma (a2, r, r2);
|
||||
r = fma (a2, r, r1);
|
||||
|
||||
if (sign)
|
||||
return -1.0 + r;
|
||||
else
|
||||
return 1.0 - r;
|
||||
}
|
||||
else if (ia < 0x40100000)
|
||||
{ /* 3.25 <= |x| < 4.0. */
|
||||
double a = fabs (x);
|
||||
a = a - 3.25;
|
||||
|
||||
double r1 = fma (a, PE[1], PE[0]);
|
||||
double r2 = fma (a, PE[3], PE[2]);
|
||||
double r3 = fma (a, PE[5], PE[4]);
|
||||
double r4 = fma (a, PE[7], PE[6]);
|
||||
double r5 = fma (a, PE[9], PE[8]);
|
||||
double r6 = fma (a, PE[11], PE[10]);
|
||||
double r7 = fma (a, PE[13], PE[12]);
|
||||
|
||||
double a2 = a * a;
|
||||
|
||||
double r = r7;
|
||||
r = fma (a2, r, r6);
|
||||
r = fma (a2, r, r5);
|
||||
r = fma (a2, r, r4);
|
||||
r = fma (a2, r, r3);
|
||||
r = fma (a2, r, r2);
|
||||
r = fma (a2, r, r1);
|
||||
|
||||
if (sign)
|
||||
return -1.0 + r;
|
||||
else
|
||||
return 1.0 - r;
|
||||
}
|
||||
else if (ia < 0x4017a000)
|
||||
{ /* 4 <= |x| < 5.90625. */
|
||||
double a = fabs (x);
|
||||
a = fma (0.5, a, -2.0);
|
||||
|
||||
double r1 = fma (a, PF[1], PF[0]);
|
||||
double r2 = fma (a, PF[3], PF[2]);
|
||||
double r3 = fma (a, PF[5], PF[4]);
|
||||
double r4 = fma (a, PF[7], PF[6]);
|
||||
double r5 = fma (a, PF[9], PF[8]);
|
||||
double r6 = fma (a, PF[11], PF[10]);
|
||||
double r7 = fma (a, PF[13], PF[12]);
|
||||
double r8 = fma (a, PF[15], PF[14]);
|
||||
double r9 = PF[16];
|
||||
|
||||
double a2 = a * a;
|
||||
|
||||
double r = r9;
|
||||
r = fma (a2, r, r8);
|
||||
r = fma (a2, r, r7);
|
||||
r = fma (a2, r, r6);
|
||||
r = fma (a2, r, r5);
|
||||
r = fma (a2, r, r4);
|
||||
r = fma (a2, r, r3);
|
||||
r = fma (a2, r, r2);
|
||||
r = fma (a2, r, r1);
|
||||
|
||||
if (sign)
|
||||
return -1.0 + r;
|
||||
else
|
||||
return 1.0 - r;
|
||||
}
|
||||
else
|
||||
{
|
||||
/* Special cases : erf(nan)=nan, erf(+inf)=+1 and erf(-inf)=-1. */
|
||||
if (unlikely (ia >= 0x7ff00000))
|
||||
return (double) (1.0 - (sign << 1)) + 1.0 / x;
|
||||
|
||||
if (sign)
|
||||
return -1.0;
|
||||
else
|
||||
return 1.0;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns complementary error function of 𝑥.
|
||||
*/
|
||||
double erfc(double x)
|
||||
{
|
||||
double r,s,z,y;
|
||||
uint32_t ix;
|
||||
int sign;
|
||||
|
||||
GET_HIGH_WORD(ix, x);
|
||||
sign = ix>>31;
|
||||
ix &= 0x7fffffff;
|
||||
if (ix >= 0x7ff00000) {
|
||||
/* erfc(nan)=nan, erfc(+-inf)=0,2 */
|
||||
return 2*sign + 1/x;
|
||||
}
|
||||
if (ix < 0x3feb0000) { /* |x| < 0.84375 */
|
||||
if (ix < 0x3c700000) /* |x| < 2**-56 */
|
||||
return 1.0 - x;
|
||||
z = x*x;
|
||||
r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
|
||||
s = 1.0+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
|
||||
y = r/s;
|
||||
if (sign || ix < 0x3fd00000) { /* x < 1/4 */
|
||||
return 1.0 - (x+x*y);
|
||||
}
|
||||
return 0.5 - (x - 0.5 + x*y);
|
||||
}
|
||||
if (ix < 0x403c0000) { /* 0.84375 <= |x| < 28 */
|
||||
return sign ? 2 - erfc2(ix,x) : erfc2(ix,x);
|
||||
}
|
||||
return sign ? 2 - 0x1p-1022 : 0x1p-1022*0x1p-1022;
|
||||
}
|
||||
|
||||
#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
|
||||
#if LDBL_MANT_DIG == 53
|
||||
__weak_reference(erf, erfl);
|
||||
__weak_reference(erfc, erfcl);
|
||||
#endif
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue