diff options
Diffstat (limited to 'lib/libm')
-rw-r--r-- | lib/libm/atan2f.c | 89 | ||||
-rw-r--r-- | lib/libm/atanf.c | 100 | ||||
-rw-r--r-- | lib/libm/libm.h | 52 | ||||
-rw-r--r-- | lib/libm/math.c | 886 | ||||
-rw-r--r-- | lib/libm/mathsincos.c | 627 |
5 files changed, 1754 insertions, 0 deletions
diff --git a/lib/libm/atan2f.c b/lib/libm/atan2f.c new file mode 100644 index 0000000000..03d000c9e1 --- /dev/null +++ b/lib/libm/atan2f.c @@ -0,0 +1,89 @@ +/*****************************************************************************/ +/*****************************************************************************/ +// atan2f from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/e_atan2f.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +#include "libm.h" + +static const float +pi = 3.1415927410e+00, /* 0x40490fdb */ +pi_lo = -8.7422776573e-08; /* 0xb3bbbd2e */ + +float atan2f(float y, float x) +{ + float z; + uint32_t m,ix,iy; + + if (isnan(x) || isnan(y)) + return x+y; + GET_FLOAT_WORD(ix, x); + GET_FLOAT_WORD(iy, y); + if (ix == 0x3f800000) /* x=1.0 */ + return atanf(y); + m = ((iy>>31)&1) | ((ix>>30)&2); /* 2*sign(x)+sign(y) */ + ix &= 0x7fffffff; + iy &= 0x7fffffff; + + /* when y = 0 */ + if (iy == 0) { + switch (m) { + case 0: + case 1: return y; /* atan(+-0,+anything)=+-0 */ + case 2: return pi; /* atan(+0,-anything) = pi */ + case 3: return -pi; /* atan(-0,-anything) =-pi */ + } + } + /* when x = 0 */ + if (ix == 0) + return m&1 ? -pi/2 : pi/2; + /* when x is INF */ + if (ix == 0x7f800000) { + if (iy == 0x7f800000) { + switch (m) { + case 0: return pi/4; /* atan(+INF,+INF) */ + case 1: return -pi/4; /* atan(-INF,+INF) */ + case 2: return 3*pi/4; /*atan(+INF,-INF)*/ + case 3: return -3*pi/4; /*atan(-INF,-INF)*/ + } + } else { + switch (m) { + case 0: return 0.0f; /* atan(+...,+INF) */ + case 1: return -0.0f; /* atan(-...,+INF) */ + case 2: return pi; /* atan(+...,-INF) */ + case 3: return -pi; /* atan(-...,-INF) */ + } + } + } + /* |y/x| > 0x1p26 */ + if (ix+(26<<23) < iy || iy == 0x7f800000) + return m&1 ? -pi/2 : pi/2; + + /* z = atan(|y/x|) with correct underflow */ + if ((m&2) && iy+(26<<23) < ix) /*|y/x| < 0x1p-26, x < 0 */ + z = 0.0; + else + z = atanf(fabsf(y/x)); + switch (m) { + case 0: return z; /* atan(+,+) */ + case 1: return -z; /* atan(-,+) */ + case 2: return pi - (z-pi_lo); /* atan(+,-) */ + default: /* case 3 */ + return (z-pi_lo) - pi; /* atan(-,-) */ + } +} diff --git a/lib/libm/atanf.c b/lib/libm/atanf.c new file mode 100644 index 0000000000..053fc1b6d6 --- /dev/null +++ b/lib/libm/atanf.c @@ -0,0 +1,100 @@ +/*****************************************************************************/ +/*****************************************************************************/ +// atanf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/s_atanf.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + + +#include "libm.h" + +static const float atanhi[] = { + 4.6364760399e-01, /* atan(0.5)hi 0x3eed6338 */ + 7.8539812565e-01, /* atan(1.0)hi 0x3f490fda */ + 9.8279368877e-01, /* atan(1.5)hi 0x3f7b985e */ + 1.5707962513e+00, /* atan(inf)hi 0x3fc90fda */ +}; + +static const float atanlo[] = { + 5.0121582440e-09, /* atan(0.5)lo 0x31ac3769 */ + 3.7748947079e-08, /* atan(1.0)lo 0x33222168 */ + 3.4473217170e-08, /* atan(1.5)lo 0x33140fb4 */ + 7.5497894159e-08, /* atan(inf)lo 0x33a22168 */ +}; + +static const float aT[] = { + 3.3333328366e-01, + -1.9999158382e-01, + 1.4253635705e-01, + -1.0648017377e-01, + 6.1687607318e-02, +}; + +float atanf(float x) +{ + float_t w,s1,s2,z; + uint32_t ix,sign; + int id; + + GET_FLOAT_WORD(ix, x); + sign = ix>>31; + ix &= 0x7fffffff; + if (ix >= 0x4c800000) { /* if |x| >= 2**26 */ + if (isnan(x)) + return x; + z = atanhi[3] + 0x1p-120f; + return sign ? -z : z; + } + if (ix < 0x3ee00000) { /* |x| < 0.4375 */ + if (ix < 0x39800000) { /* |x| < 2**-12 */ + if (ix < 0x00800000) + /* raise underflow for subnormal x */ + FORCE_EVAL(x*x); + return x; + } + id = -1; + } else { + x = fabsf(x); + if (ix < 0x3f980000) { /* |x| < 1.1875 */ + if (ix < 0x3f300000) { /* 7/16 <= |x| < 11/16 */ + id = 0; + x = (2.0f*x - 1.0f)/(2.0f + x); + } else { /* 11/16 <= |x| < 19/16 */ + id = 1; + x = (x - 1.0f)/(x + 1.0f); + } + } else { + if (ix < 0x401c0000) { /* |x| < 2.4375 */ + id = 2; + x = (x - 1.5f)/(1.0f + 1.5f*x); + } else { /* 2.4375 <= |x| < 2**26 */ + id = 3; + x = -1.0f/x; + } + } + } + /* end of argument reduction */ + z = x*x; + w = z*z; + /* break sum from i=0 to 10 aT[i]z**(i+1) into odd and even poly */ + s1 = z*(aT[0]+w*(aT[2]+w*aT[4])); + s2 = w*(aT[1]+w*aT[3]); + if (id < 0) + return x - x*(s1+s2); + z = atanhi[id] - ((x*(s1+s2) - atanlo[id]) - x); + return sign ? -z : z; +} diff --git a/lib/libm/libm.h b/lib/libm/libm.h new file mode 100644 index 0000000000..c87c558219 --- /dev/null +++ b/lib/libm/libm.h @@ -0,0 +1,52 @@ +/*****************************************************************************/ +/*****************************************************************************/ +// portions extracted from musl-0.9.15 libm.h +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/math_private.h */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +#include <stdint.h> +#include <math.h> + +#define FORCE_EVAL(x) do { \ + if (sizeof(x) == sizeof(float)) { \ + volatile float __x; \ + __x = (x); \ + (void)__x; \ + } else if (sizeof(x) == sizeof(double)) { \ + volatile double __x; \ + __x = (x); \ + (void)__x; \ + } else { \ + volatile long double __x; \ + __x = (x); \ + (void)__x; \ + } \ +} while(0) + +/* Get a 32 bit int from a float. */ +#define GET_FLOAT_WORD(w,d) \ +do { \ + union {float f; uint32_t i;} __u; \ + __u.f = (d); \ + (w) = __u.i; \ +} while (0) + +/* Set a float from a 32 bit int. */ +#define SET_FLOAT_WORD(d,w) \ +do { \ + union {float f; uint32_t i;} __u; \ + __u.i = (w); \ + (d) = __u.f; \ +} while (0) diff --git a/lib/libm/math.c b/lib/libm/math.c new file mode 100644 index 0000000000..8433081b97 --- /dev/null +++ b/lib/libm/math.c @@ -0,0 +1,886 @@ +/* + * This file is part of the Micro Python project, http://micropython.org/ + * + * The MIT License (MIT) + * + * Copyright (c) 2013, 2014 Damien P. George + * + * 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 "libm.h" + +typedef float float_t; +typedef union { + float f; + struct { + uint64_t m : 23; + uint64_t e : 8; + uint64_t s : 1; + }; +} float_s_t; + +typedef union { + double d; + struct { + uint64_t m : 52; + uint64_t e : 11; + uint64_t s : 1; + }; +} double_s_t; + +double __attribute__((pcs("aapcs"))) __aeabi_i2d(int32_t x) { + return (float)x; +} + +double __attribute__((pcs("aapcs"))) __aeabi_f2d(float x) { + float_s_t fx={0}; + double_s_t dx={0}; + + fx.f = x; + dx.s = (fx.s); + dx.e = (fx.e-127+1023) & 0x7FF; + dx.m = fx.m; + dx.m <<=(52-23); // left justify + return dx.d; +} + +float __attribute__((pcs("aapcs"))) __aeabi_d2f(double x) { + float_s_t fx={0}; + double_s_t dx={0}; + + dx.d = x; + fx.s = (dx.s); + fx.e = (dx.e-1023+127) & 0xFF; + fx.m = (dx.m>>(52-23)); // right justify + return fx.f; +} + +double __aeabi_dmul(double x , double y) { + return 0.0; + +} + +float sqrtf(float x) { + asm volatile ( + "vsqrt.f32 %[r], %[x]\n" + : [r] "=t" (x) + : [x] "t" (x)); + return x; +} + +#ifndef NDEBUG +float copysignf(float x, float y) { + float_s_t fx={.f = x}; + float_s_t fy={.f = y}; + + // copy sign bit; + fx.s = fy.s; + + return fx.f; +} +#endif + +// some compilers define log2f in terms of logf +#ifdef log2f +#undef log2f +#endif +// some compilers have _M_LN2 defined in math.h, some don't +#ifndef _M_LN2 +#define _M_LN2 (0.69314718055994530942) +#endif +float log2f(float x) { return logf(x) / (float)_M_LN2; } + +static const float _M_LN10 = 2.30258509299404; // 0x40135d8e +float log10f(float x) { return logf(x) / (float)_M_LN10; } + +float tanhf(float x) { return sinhf(x) / coshf(x); } + +// TODO we need import these functions from some library (eg musl or newlib) +float acoshf(float x) { return 0.0; } +float asinhf(float x) { return 0.0; } +float atanhf(float x) { return 0.0; } +float acosf(float x) { return 0.0; } +float asinf(float x) { return 0.0; } +float fmodf(float x, float y) { return 0.0; } +float tgammaf(float x) { return 0.0; } +float lgammaf(float x) { return 0.0; } +float erff(float x) { return 0.0; } +float erfcf(float x) { return 0.0; } +float modff(float x, float *y) { return 0.0; } +float frexpf(float x, int *exp) { return 0.0; } +float ldexpf(float x, int exp) { return 0.0; } + +/*****************************************************************************/ +/*****************************************************************************/ +// __fpclassifyf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +int __fpclassifyf(float x) +{ + union {float f; uint32_t i;} u = {x}; + int e = u.i>>23 & 0xff; + if (!e) return u.i<<1 ? FP_SUBNORMAL : FP_ZERO; + if (e==0xff) return u.i<<9 ? FP_NAN : FP_INFINITE; + return FP_NORMAL; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// scalbnf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +float scalbnf(float x, int n) +{ + union {float f; uint32_t i;} u; + float_t y = x; + + if (n > 127) { + y *= 0x1p127f; + n -= 127; + if (n > 127) { + y *= 0x1p127f; + n -= 127; + if (n > 127) + n = 127; + } + } else if (n < -126) { + y *= 0x1p-126f; + n += 126; + if (n < -126) { + y *= 0x1p-126f; + n += 126; + if (n < -126) + n = -126; + } + } + u.i = (uint32_t)(0x7f+n)<<23; + x = y * u.f; + return x; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// powf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/e_powf.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +static const float +bp[] = {1.0, 1.5,}, +dp_h[] = { 0.0, 5.84960938e-01,}, /* 0x3f15c000 */ +dp_l[] = { 0.0, 1.56322085e-06,}, /* 0x35d1cfdc */ +two24 = 16777216.0, /* 0x4b800000 */ +huge = 1.0e30, +tiny = 1.0e-30, +/* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */ +L1 = 6.0000002384e-01, /* 0x3f19999a */ +L2 = 4.2857143283e-01, /* 0x3edb6db7 */ +L3 = 3.3333334327e-01, /* 0x3eaaaaab */ +L4 = 2.7272811532e-01, /* 0x3e8ba305 */ +L5 = 2.3066075146e-01, /* 0x3e6c3255 */ +L6 = 2.0697501302e-01, /* 0x3e53f142 */ +P1 = 1.6666667163e-01, /* 0x3e2aaaab */ +P2 = -2.7777778450e-03, /* 0xbb360b61 */ +P3 = 6.6137559770e-05, /* 0x388ab355 */ +P4 = -1.6533901999e-06, /* 0xb5ddea0e */ +P5 = 4.1381369442e-08, /* 0x3331bb4c */ +lg2 = 6.9314718246e-01, /* 0x3f317218 */ +lg2_h = 6.93145752e-01, /* 0x3f317200 */ +lg2_l = 1.42860654e-06, /* 0x35bfbe8c */ +ovt = 4.2995665694e-08, /* -(128-log2(ovfl+.5ulp)) */ +cp = 9.6179670095e-01, /* 0x3f76384f =2/(3ln2) */ +cp_h = 9.6191406250e-01, /* 0x3f764000 =12b cp */ +cp_l = -1.1736857402e-04, /* 0xb8f623c6 =tail of cp_h */ +ivln2 = 1.4426950216e+00, /* 0x3fb8aa3b =1/ln2 */ +ivln2_h = 1.4426879883e+00, /* 0x3fb8aa00 =16b 1/ln2*/ +ivln2_l = 7.0526075433e-06; /* 0x36eca570 =1/ln2 tail*/ + +float powf(float x, float y) +{ + float z,ax,z_h,z_l,p_h,p_l; + float y1,t1,t2,r,s,sn,t,u,v,w; + int32_t i,j,k,yisint,n; + int32_t hx,hy,ix,iy,is; + + GET_FLOAT_WORD(hx, x); + GET_FLOAT_WORD(hy, y); + ix = hx & 0x7fffffff; + iy = hy & 0x7fffffff; + + /* x**0 = 1, even if x is NaN */ + if (iy == 0) + return 1.0f; + /* 1**y = 1, even if y is NaN */ + if (hx == 0x3f800000) + return 1.0f; + /* NaN if either arg is NaN */ + if (ix > 0x7f800000 || iy > 0x7f800000) + return x + y; + + /* 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 (hx < 0) { + if (iy >= 0x4b800000) + yisint = 2; /* even integer y */ + else if (iy >= 0x3f800000) { + k = (iy>>23) - 0x7f; /* exponent */ + j = iy>>(23-k); + if ((j<<(23-k)) == iy) + yisint = 2 - (j & 1); + } + } + + /* special value of y */ + if (iy == 0x7f800000) { /* y is +-inf */ + if (ix == 0x3f800000) /* (-1)**+-inf is 1 */ + return 1.0f; + else if (ix > 0x3f800000) /* (|x|>1)**+-inf = inf,0 */ + return hy >= 0 ? y : 0.0f; + else if (ix != 0) /* (|x|<1)**+-inf = 0,inf if x!=0 */ + return hy >= 0 ? 0.0f: -y; + } + if (iy == 0x3f800000) /* y is +-1 */ + return hy >= 0 ? x : 1.0f/x; + if (hy == 0x40000000) /* y is 2 */ + return x*x; + if (hy == 0x3f000000) { /* y is 0.5 */ + if (hx >= 0) /* x >= +0 */ + return sqrtf(x); + } + + ax = fabsf(x); + /* special value of x */ + if (ix == 0x7f800000 || ix == 0 || ix == 0x3f800000) { /* x is +-0,+-inf,+-1 */ + z = ax; + if (hy < 0) /* z = (1/|x|) */ + z = 1.0f/z; + if (hx < 0) { + if (((ix-0x3f800000)|yisint) == 0) { + z = (z-z)/(z-z); /* (-1)**non-int is NaN */ + } else if (yisint == 1) + z = -z; /* (x<0)**odd = -(|x|**odd) */ + } + return z; + } + + sn = 1.0f; /* sign of result */ + if (hx < 0) { + if (yisint == 0) /* (x<0)**(non-int) is NaN */ + return (x-x)/(x-x); + if (yisint == 1) /* (x<0)**(odd int) */ + sn = -1.0f; + } + + /* |y| is huge */ + if (iy > 0x4d000000) { /* if |y| > 2**27 */ + /* over/underflow if x is not close to one */ + if (ix < 0x3f7ffff8) + return hy < 0 ? sn*huge*huge : sn*tiny*tiny; + if (ix > 0x3f800007) + return hy > 0 ? sn*huge*huge : sn*tiny*tiny; + /* now |1-x| is tiny <= 2**-20, suffice to compute + log(x) by x-x^2/2+x^3/3-x^4/4 */ + t = ax - 1; /* t has 20 trailing zeros */ + w = (t*t)*(0.5f - t*(0.333333333333f - t*0.25f)); + u = ivln2_h*t; /* ivln2_h has 16 sig. bits */ + v = t*ivln2_l - w*ivln2; + t1 = u + v; + GET_FLOAT_WORD(is, t1); + SET_FLOAT_WORD(t1, is & 0xfffff000); + t2 = v - (t1-u); + } else { + float s2,s_h,s_l,t_h,t_l; + n = 0; + /* take care subnormal number */ + if (ix < 0x00800000) { + ax *= two24; + n -= 24; + GET_FLOAT_WORD(ix, ax); + } + n += ((ix)>>23) - 0x7f; + j = ix & 0x007fffff; + /* determine interval */ + ix = j | 0x3f800000; /* normalize ix */ + if (j <= 0x1cc471) /* |x|<sqrt(3/2) */ + k = 0; + else if (j < 0x5db3d7) /* |x|<sqrt(3) */ + k = 1; + else { + k = 0; + n += 1; + ix -= 0x00800000; + } + SET_FLOAT_WORD(ax, ix); + + /* compute s = s_h+s_l = (x-1)/(x+1) or (x-1.5)/(x+1.5) */ + u = ax - bp[k]; /* bp[0]=1.0, bp[1]=1.5 */ + v = 1.0f/(ax+bp[k]); + s = u*v; + s_h = s; + GET_FLOAT_WORD(is, s_h); + SET_FLOAT_WORD(s_h, is & 0xfffff000); + /* t_h=ax+bp[k] High */ + is = ((ix>>1) & 0xfffff000) | 0x20000000; + SET_FLOAT_WORD(t_h, is + 0x00400000 + (k<<21)); + t_l = ax - (t_h - bp[k]); + s_l = v*((u - s_h*t_h) - s_h*t_l); + /* compute log(ax) */ + s2 = s*s; + r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6))))); + r += s_l*(s_h+s); + s2 = s_h*s_h; + t_h = 3.0f + s2 + r; + GET_FLOAT_WORD(is, t_h); + SET_FLOAT_WORD(t_h, is & 0xfffff000); + t_l = r - ((t_h - 3.0f) - s2); + /* u+v = s*(1+...) */ + u = s_h*t_h; + v = s_l*t_h + t_l*s; + /* 2/(3log2)*(s+...) */ + p_h = u + v; + GET_FLOAT_WORD(is, p_h); + SET_FLOAT_WORD(p_h, is & 0xfffff000); + p_l = v - (p_h - u); + z_h = cp_h*p_h; /* cp_h+cp_l = 2/(3*log2) */ + z_l = cp_l*p_h + p_l*cp+dp_l[k]; + /* log2(ax) = (s+..)*2/(3*log2) = n + dp_h + z_h + z_l */ + t = (float)n; + t1 = (((z_h + z_l) + dp_h[k]) + t); + GET_FLOAT_WORD(is, t1); + SET_FLOAT_WORD(t1, is & 0xfffff000); + t2 = z_l - (((t1 - t) - dp_h[k]) - z_h); + } + + /* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */ + GET_FLOAT_WORD(is, y); + SET_FLOAT_WORD(y1, is & 0xfffff000); + p_l = (y-y1)*t1 + y*t2; + p_h = y1*t1; + z = p_l + p_h; + GET_FLOAT_WORD(j, z); + if (j > 0x43000000) /* if z > 128 */ + return sn*huge*huge; /* overflow */ + else if (j == 0x43000000) { /* if z == 128 */ + if (p_l + ovt > z - p_h) + return sn*huge*huge; /* overflow */ + } else if ((j&0x7fffffff) > 0x43160000) /* z < -150 */ // FIXME: check should be (uint32_t)j > 0xc3160000 + return sn*tiny*tiny; /* underflow */ + else if (j == 0xc3160000) { /* z == -150 */ + if (p_l <= z-p_h) + return sn*tiny*tiny; /* underflow */ + } + /* + * compute 2**(p_h+p_l) + */ + i = j & 0x7fffffff; + k = (i>>23) - 0x7f; + n = 0; + if (i > 0x3f000000) { /* if |z| > 0.5, set n = [z+0.5] */ + n = j + (0x00800000>>(k+1)); + k = ((n&0x7fffffff)>>23) - 0x7f; /* new k for n */ + SET_FLOAT_WORD(t, n & ~(0x007fffff>>k)); + n = ((n&0x007fffff)|0x00800000)>>(23-k); + if (j < 0) + n = -n; + p_h -= t; + } + t = p_l + p_h; + GET_FLOAT_WORD(is, t); + SET_FLOAT_WORD(t, is & 0xffff8000); + u = t*lg2_h; + v = (p_l-(t-p_h))*lg2 + t*lg2_l; + z = u + v; + w = v - (z - u); + t = z*z; + t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5)))); + r = (z*t1)/(t1-2.0f) - (w+z*w); + z = 1.0f - (r - z); + GET_FLOAT_WORD(j, z); + j += n<<23; + if ((j>>23) <= 0) /* subnormal output */ + z = scalbnf(z, n); + else + SET_FLOAT_WORD(z, j); + return sn*z; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// expf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/e_expf.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +static const float +half[2] = {0.5,-0.5}, +ln2hi = 6.9314575195e-1f, /* 0x3f317200 */ +ln2lo = 1.4286067653e-6f, /* 0x35bfbe8e */ +invln2 = 1.4426950216e+0f, /* 0x3fb8aa3b */ +/* + * Domain [-0.34568, 0.34568], range ~[-4.278e-9, 4.447e-9]: + * |x*(exp(x)+1)/(exp(x)-1) - p(x)| < 2**-27.74 + */ +expf_P1 = 1.6666625440e-1f, /* 0xaaaa8f.0p-26 */ +expf_P2 = -2.7667332906e-3f; /* -0xb55215.0p-32 */ + +float expf(float x) +{ + float_t hi, lo, c, xx, y; + int k, sign; + uint32_t hx; + + GET_FLOAT_WORD(hx, x); + sign = hx >> 31; /* sign bit of x */ + hx &= 0x7fffffff; /* high word of |x| */ + + /* special cases */ + if (hx >= 0x42aeac50) { /* if |x| >= -87.33655f or NaN */ + if (hx >= 0x42b17218 && !sign) { /* x >= 88.722839f */ + /* overflow */ + x *= 0x1p127f; + return x; + } + if (sign) { + /* underflow */ + FORCE_EVAL(-0x1p-149f/x); + if (hx >= 0x42cff1b5) /* x <= -103.972084f */ + return 0; + } + } + + /* argument reduction */ + if (hx > 0x3eb17218) { /* if |x| > 0.5 ln2 */ + if (hx > 0x3f851592) /* if |x| > 1.5 ln2 */ + k = invln2*x + half[sign]; + else + k = 1 - sign - sign; + hi = x - k*ln2hi; /* k*ln2hi is exact here */ + lo = k*ln2lo; + x = hi - lo; + } else if (hx > 0x39000000) { /* |x| > 2**-14 */ + k = 0; + hi = x; + lo = 0; + } else { + /* raise inexact */ + FORCE_EVAL(0x1p127f + x); + return 1 + x; + } + + /* x is now in primary range */ + xx = x*x; + c = x - xx*(expf_P1+xx*expf_P2); + y = 1 + (x*c/(2-c) - lo + hi); + if (k == 0) + return y; + return scalbnf(y, k); +} + +/*****************************************************************************/ +/*****************************************************************************/ +// expm1f from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/s_expm1f.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +static const float +o_threshold = 8.8721679688e+01, /* 0x42b17180 */ +ln2_hi = 6.9313812256e-01, /* 0x3f317180 */ +ln2_lo = 9.0580006145e-06, /* 0x3717f7d1 */ +//invln2 = 1.4426950216e+00, /* 0x3fb8aa3b */ +/* + * Domain [-0.34568, 0.34568], range ~[-6.694e-10, 6.696e-10]: + * |6 / x * (1 + 2 * (1 / (exp(x) - 1) - 1 / x)) - q(x)| < 2**-30.04 + * Scaled coefficients: Qn_here = 2**n * Qn_for_q (see s_expm1.c): + */ +Q1 = -3.3333212137e-2, /* -0x888868.0p-28 */ +Q2 = 1.5807170421e-3; /* 0xcf3010.0p-33 */ + +float expm1f(float x) +{ + float_t y,hi,lo,c,t,e,hxs,hfx,r1,twopk; + union {float f; uint32_t i;} u = {x}; + uint32_t hx = u.i & 0x7fffffff; + int k, sign = u.i >> 31; + + /* filter out huge and non-finite argument */ + if (hx >= 0x4195b844) { /* if |x|>=27*ln2 */ + if (hx > 0x7f800000) /* NaN */ + return x; + if (sign) + return -1; + if (x > o_threshold) { + x *= 0x1p127f; + return x; + } + } + + /* argument reduction */ + if (hx > 0x3eb17218) { /* if |x| > 0.5 ln2 */ + if (hx < 0x3F851592) { /* and |x| < 1.5 ln2 */ + if (!sign) { + hi = x - ln2_hi; + lo = ln2_lo; + k = 1; + } else { + hi = x + ln2_hi; + lo = -ln2_lo; + k = -1; + } + } else { + k = invln2*x + (sign ? -0.5f : 0.5f); + t = k; + hi = x - t*ln2_hi; /* t*ln2_hi is exact here */ + lo = t*ln2_lo; + } + x = hi-lo; + c = (hi-x)-lo; + } else if (hx < 0x33000000) { /* when |x|<2**-25, return x */ + if (hx < 0x00800000) + FORCE_EVAL(x*x); + return x; + } else + k = 0; + + /* x is now in primary range */ + hfx = 0.5f*x; + hxs = x*hfx; + r1 = 1.0f+hxs*(Q1+hxs*Q2); + t = 3.0f - r1*hfx; + e = hxs*((r1-t)/(6.0f - x*t)); + if (k == 0) /* c is 0 */ + return x - (x*e-hxs); + e = x*(e-c) - c; + e -= hxs; + /* exp(x) ~ 2^k (x_reduced - e + 1) */ + if (k == -1) + return 0.5f*(x-e) - 0.5f; + if (k == 1) { + if (x < -0.25f) + return -2.0f*(e-(x+0.5f)); + return 1.0f + 2.0f*(x-e); + } + u.i = (0x7f+k)<<23; /* 2^k */ + twopk = u.f; + if (k < 0 || k > 56) { /* suffice to return exp(x)-1 */ + y = x - e + 1.0f; + if (k == 128) + y = y*2.0f*0x1p127f; + else + y = y*twopk; + return y - 1.0f; + } + u.i = (0x7f-k)<<23; /* 2^-k */ + if (k < 23) + y = (x-e+(1-u.f))*twopk; + else + y = (x-(e+u.f)+1)*twopk; + return y; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// __expo2f from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* k is such that k*ln2 has minimal relative error and x - kln2 > log(FLT_MIN) */ +static const int k = 235; +static const float kln2 = 0x1.45c778p+7f; + +/* expf(x)/2 for x >= log(FLT_MAX), slightly better than 0.5f*expf(x/2)*expf(x/2) */ +float __expo2f(float x) +{ + float scale; + + /* note that k is odd and scale*scale overflows */ + SET_FLOAT_WORD(scale, (uint32_t)(0x7f + k/2) << 23); + /* exp(x - k ln2) * 2**(k-1) */ + return expf(x - kln2) * scale * scale; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// logf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +/* origin: FreeBSD /usr/src/lib/msun/src/e_logf.c */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +static const float +/* |(log(1+s)-log(1-s))/s - Lg(s)| < 2**-34.24 (~[-4.95e-11, 4.97e-11]). */ +Lg1 = 0xaaaaaa.0p-24, /* 0.66666662693 */ +Lg2 = 0xccce13.0p-25, /* 0.40000972152 */ +Lg3 = 0x91e9ee.0p-25, /* 0.28498786688 */ +Lg4 = 0xf89e26.0p-26; /* 0.24279078841 */ + +float logf(float x) +{ + union {float f; uint32_t i;} u = {x}; + float_t hfsq,f,s,z,R,w,t1,t2,dk; + uint32_t ix; + int k; + + ix = u.i; + k = 0; + if (ix < 0x00800000 || ix>>31) { /* x < 2**-126 */ + if (ix<<1 == 0) + return -1/(x*x); /* log(+-0)=-inf */ + if (ix>>31) + return (x-x)/0.0f; /* log(-#) = NaN */ + /* subnormal number, scale up x */ + k -= 25; + x *= 0x1p25f; + u.f = x; + ix = u.i; + } else if (ix >= 0x7f800000) { + return x; + } else if (ix == 0x3f800000) + return 0; + + /* reduce x into [sqrt(2)/2, sqrt(2)] */ + ix += 0x3f800000 - 0x3f3504f3; + k += (int)(ix>>23) - 0x7f; + ix = (ix&0x007fffff) + 0x3f3504f3; + u.i = ix; + x = u.f; + + f = x - 1.0f; + s = f/(2.0f + f); + z = s*s; + w = z*z; + t1= w*(Lg2+w*Lg4); + t2= z*(Lg1+w*Lg3); + R = t2 + t1; + hfsq = 0.5f*f*f; + dk = k; + return s*(hfsq+R) + dk*ln2_lo - hfsq + f + dk*ln2_hi; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// coshf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +float coshf(float x) +{ + union {float f; uint32_t i;} u = {.f = x}; + uint32_t w; + float t; + + /* |x| */ + u.i &= 0x7fffffff; + x = u.f; + w = u.i; + + /* |x| < log(2) */ + if (w < 0x3f317217) { + if (w < 0x3f800000 - (12<<23)) { + FORCE_EVAL(x + 0x1p120f); + return 1; + } + t = expm1f(x); + return 1 + t*t/(2*(1+t)); + } + + /* |x| < log(FLT_MAX) */ + if (w < 0x42b17217) { + t = expf(x); + return 0.5f*(t + 1/t); + } + + /* |x| > log(FLT_MAX) or nan */ + t = __expo2f(x); + return t; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// sinhf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +float sinhf(float x) +{ + union {float f; uint32_t i;} u = {.f = x}; + uint32_t w; + float t, h, absx; + + h = 0.5; + if (u.i >> 31) + h = -h; + /* |x| */ + u.i &= 0x7fffffff; + absx = u.f; + w = u.i; + + /* |x| < log(FLT_MAX) */ + if (w < 0x42b17217) { + t = expm1f(absx); + if (w < 0x3f800000) { + if (w < 0x3f800000 - (12<<23)) + return x; + return h*(2*t - t*t/(t+1)); + } + return h*(t + t/(t+1)); + } + + /* |x| > logf(FLT_MAX) or nan */ + t = 2*h*__expo2f(absx); + return t; +} + +/*****************************************************************************/ +/*****************************************************************************/ +// ceilf, floorf and truncf from musl-0.9.15 +/*****************************************************************************/ +/*****************************************************************************/ + +float ceilf(float x) +{ + union {float f; uint32_t i;} u = {x}; + int e = (int)(u.i >> 23 & 0xff) - 0x7f; + uint32_t m; + + if (e >= 23) + return x; + if (e >= 0) { + m = 0x007fffff >> e; + if ((u.i & m) == 0) + return x; + FORCE_EVAL(x + 0x1p120f); + if (u.i >> 31 == 0) + u.i += m; + u.i &= ~m; + } else { + FORCE_EVAL(x + 0x1p120f); + if (u.i >> 31) + u.f = -0.0; + else if (u.i << 1) + u.f = 1.0; + } + return u.f; +} + +float floorf(float x) +{ + union {float f; uint32_t i;} u = {x}; + int e = (int)(u.i >> 23 & 0xff) - 0x7f; + uint32_t m; + + if (e >= 23) + return x; + if (e >= 0) { + m = 0x007fffff >> e; + if ((u.i & m) == 0) + return x; + FORCE_EVAL(x + 0x1p120f); + if (u.i >> 31) + u.i += m; + u.i &= ~m; + } else { + FORCE_EVAL(x + 0x1p120f); + if (u.i >> 31 == 0) + u.i = 0; + else if (u.i << 1) + u.f = -1.0; + } + return u.f; +} + +float truncf(float x) +{ + union {float f; uint32_t i;} u = {x}; + int e = (int)(u.i >> 23 & 0xff) - 0x7f + 9; + uint32_t m; + + if (e >= 23 + 9) + return x; + if (e < 9) + e = 1; + m = -1U >> e; + if ((u.i & m) == 0) + return x; + FORCE_EVAL(x + 0x1p120f); + u.i &= ~m; + return u.f; +} diff --git a/lib/libm/mathsincos.c b/lib/libm/mathsincos.c new file mode 100644 index 0000000000..9ea08e6c87 --- /dev/null +++ b/lib/libm/mathsincos.c @@ -0,0 +1,627 @@ +/* + * This file is part of the Micro Python project, http://micropython.org/ + * + * These math functions are taken from newlib 2.1.0, the newlib/libm directory. + * Appropriate copyright headers are reproduced below. + * + * The MIT License (MIT) + * + * Copyright (c) 2013, 2014 Damien P. George + * + * 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 <stdint.h> +#include <math.h> + +/* @(#)fdlibm.h 5.1 93/09/24 */ +/* + * ==================================================== + * 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. + * ==================================================== + */ + +#define FLT_UWORD_IS_FINITE(x) ((x)<0x7f800000L) + +/* A union which permits us to convert between a float and a 32 bit + int. */ + +typedef union +{ + float value; + __uint32_t word; +} ieee_float_shape_type; + + +/* Get a 32 bit int from a float. */ + +#define GET_FLOAT_WORD(i,d) \ +do { \ + ieee_float_shape_type gf_u; \ + gf_u.value = (d); \ + (i) = gf_u.word; \ +} while (0) + +/* Set a float from a 32 bit int. */ + +#define SET_FLOAT_WORD(d,i) \ +do { \ + ieee_float_shape_type sf_u; \ + sf_u.word = (i); \ + (d) = sf_u.value; \ +} while (0) + + + +/* kf_rem_pio2.c -- float version of k_rem_pio2.c + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + */ + +/* In the float version, the input parameter x contains 8 bit + integers, not 24 bit integers. 113 bit precision is not supported. */ + +#ifdef __STDC__ +static const int init_jk[] = {4,7,9}; /* initial value for jk */ +#else +static int init_jk[] = {4,7,9}; +#endif + +#ifdef __STDC__ +static const float PIo2[] = { +#else +static float PIo2[] = { +#endif + 1.5703125000e+00, /* 0x3fc90000 */ + 4.5776367188e-04, /* 0x39f00000 */ + 2.5987625122e-05, /* 0x37da0000 */ + 7.5437128544e-08, /* 0x33a20000 */ + 6.0026650317e-11, /* 0x2e840000 */ + 7.3896444519e-13, /* 0x2b500000 */ + 5.3845816694e-15, /* 0x27c20000 */ + 5.6378512969e-18, /* 0x22d00000 */ + 8.3009228831e-20, /* 0x1fc40000 */ + 3.2756352257e-22, /* 0x1bc60000 */ + 6.3331015649e-25, /* 0x17440000 */ +}; + +static const float +zero = 0.0000000000e+00, /* 0x00000000 */ +one = 1.0000000000e+00, /* 0x3f800000 */ +two8 = 2.5600000000e+02, /* 0x43800000 */ +twon8 = 3.9062500000e-03; /* 0x3b800000 */ + +int __kernel_rem_pio2f(float *x, float *y, int e0, int nx, int prec, const __int32_t *ipio2) +{ + __int32_t jz,jx,jv,jp,jk,carry,n,iq[20],i,j,k,m,q0,ih; + float z,fw,f[20],fq[20],q[20]; + + /* initialize jk*/ + jk = init_jk[prec]; + jp = jk; + + /* determine jx,jv,q0, note that 3>q0 */ + jx = nx-1; + jv = (e0-3)/8; if(jv<0) jv=0; + q0 = e0-8*(jv+1); + + /* set up f[0] to f[jx+jk] where f[jx+jk] = ipio2[jv+jk] */ + j = jv-jx; m = jx+jk; + for(i=0;i<=m;i++,j++) f[i] = (j<0)? zero : (float) ipio2[j]; + + /* compute q[0],q[1],...q[jk] */ + for (i=0;i<=jk;i++) { + for(j=0,fw=0.0;j<=jx;j++) fw += x[j]*f[jx+i-j]; q[i] = fw; + } + + jz = jk; +recompute: + /* distill q[] into iq[] reversingly */ + for(i=0,j=jz,z=q[jz];j>0;i++,j--) { + fw = (float)((__int32_t)(twon8* z)); + iq[i] = (__int32_t)(z-two8*fw); + z = q[j-1]+fw; + } + + /* compute n */ + z = scalbnf(z,(int)q0); /* actual value of z */ + z -= (float)8.0*floorf(z*(float)0.125); /* trim off integer >= 8 */ + n = (__int32_t) z; + z -= (float)n; + ih = 0; + if(q0>0) { /* need iq[jz-1] to determine n */ + i = (iq[jz-1]>>(8-q0)); n += i; + iq[jz-1] -= i<<(8-q0); + ih = iq[jz-1]>>(7-q0); + } + else if(q0==0) ih = iq[jz-1]>>8; + else if(z>=(float)0.5) ih=2; + + if(ih>0) { /* q > 0.5 */ + n += 1; carry = 0; + for(i=0;i<jz ;i++) { /* compute 1-q */ + j = iq[i]; + if(carry==0) { + if(j!=0) { + carry = 1; iq[i] = 0x100- j; + } + } else iq[i] = 0xff - j; + } + if(q0>0) { /* rare case: chance is 1 in 12 */ + switch(q0) { + case 1: + iq[jz-1] &= 0x7f; break; + case 2: + iq[jz-1] &= 0x3f; break; + } + } + if(ih==2) { + z = one - z; + if(carry!=0) z -= scalbnf(one,(int)q0); + } + } + + /* check if recomputation is needed */ + if(z==zero) { + j = 0; + for (i=jz-1;i>=jk;i--) j |= iq[i]; + if(j==0) { /* need recomputation */ + for(k=1;iq[jk-k]==0;k++); /* k = no. of terms needed */ + + for(i=jz+1;i<=jz+k;i++) { /* add q[jz+1] to q[jz+k] */ + f[jx+i] = (float) ipio2[jv+i]; + for(j=0,fw=0.0;j<=jx;j++) fw += x[j]*f[jx+i-j]; + q[i] = fw; + } + jz += k; + goto recompute; + } + } + + /* chop off zero terms */ + if(z==(float)0.0) { + jz -= 1; q0 -= 8; + while(iq[jz]==0) { jz--; q0-=8;} + } else { /* break z into 8-bit if necessary */ + z = scalbnf(z,-(int)q0); + if(z>=two8) { + fw = (float)((__int32_t)(twon8*z)); + iq[jz] = (__int32_t)(z-two8*fw); + jz += 1; q0 += 8; + iq[jz] = (__int32_t) fw; + } else iq[jz] = (__int32_t) z ; + } + + /* convert integer "bit" chunk to floating-point value */ + fw = scalbnf(one,(int)q0); + for(i=jz;i>=0;i--) { + q[i] = fw*(float)iq[i]; fw*=twon8; + } + + /* compute PIo2[0,...,jp]*q[jz,...,0] */ + for(i=jz;i>=0;i--) { + for(fw=0.0,k=0;k<=jp&&k<=jz-i;k++) fw += PIo2[k]*q[i+k]; + fq[jz-i] = fw; + } + + /* compress fq[] into y[] */ + switch(prec) { + case 0: + fw = 0.0; + for (i=jz;i>=0;i--) fw += fq[i]; + y[0] = (ih==0)? fw: -fw; + break; + case 1: + case 2: + fw = 0.0; + for (i=jz;i>=0;i--) fw += fq[i]; + y[0] = (ih==0)? fw: -fw; + fw = fq[0]-fw; + for (i=1;i<=jz;i++) fw += fq[i]; + y[1] = (ih==0)? fw: -fw; + break; + case 3: /* painful */ + for (i=jz;i>0;i--) { + fw = fq[i-1]+fq[i]; + fq[i] += fq[i-1]-fw; + fq[i-1] = fw; + } + for (i=jz;i>1;i--) { + fw = fq[i-1]+fq[i]; + fq[i] += fq[i-1]-fw; + fq[i-1] = fw; + } + for (fw=0.0,i=jz;i>=2;i--) fw += fq[i]; + if(ih==0) { + y[0] = fq[0]; y[1] = fq[1]; y[2] = fw; + } else { + y[0] = -fq[0]; y[1] = -fq[1]; y[2] = -fw; + } + } + return n&7; +} + + +/* ef_rem_pio2.c -- float version of e_rem_pio2.c + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + * + */ + +/* __ieee754_rem_pio2f(x,y) + * + * return the remainder of x rem pi/2 in y[0]+y[1] + * use __kernel_rem_pio2f() + */ + + +/* + * Table of constants for 2/pi, 396 Hex digits (476 decimal) of 2/pi + */ +static const __int32_t two_over_pi[] = { +0xA2, 0xF9, 0x83, 0x6E, 0x4E, 0x44, 0x15, 0x29, 0xFC, +0x27, 0x57, 0xD1, 0xF5, 0x34, 0xDD, 0xC0, 0xDB, 0x62, +0x95, 0x99, 0x3C, 0x43, 0x90, 0x41, 0xFE, 0x51, 0x63, +0xAB, 0xDE, 0xBB, 0xC5, 0x61, 0xB7, 0x24, 0x6E, 0x3A, +0x42, 0x4D, 0xD2, 0xE0, 0x06, 0x49, 0x2E, 0xEA, 0x09, +0xD1, 0x92, 0x1C, 0xFE, 0x1D, 0xEB, 0x1C, 0xB1, 0x29, +0xA7, 0x3E, 0xE8, 0x82, 0x35, 0xF5, 0x2E, 0xBB, 0x44, +0x84, 0xE9, 0x9C, 0x70, 0x26, 0xB4, 0x5F, 0x7E, 0x41, +0x39, 0x91, 0xD6, 0x39, 0x83, 0x53, 0x39, 0xF4, 0x9C, +0x84, 0x5F, 0x8B, 0xBD, 0xF9, 0x28, 0x3B, 0x1F, 0xF8, +0x97, 0xFF, 0xDE, 0x05, 0x98, 0x0F, 0xEF, 0x2F, 0x11, +0x8B, 0x5A, 0x0A, 0x6D, 0x1F, 0x6D, 0x36, 0x7E, 0xCF, +0x27, 0xCB, 0x09, 0xB7, 0x4F, 0x46, 0x3F, 0x66, 0x9E, +0x5F, 0xEA, 0x2D, 0x75, 0x27, 0xBA, 0xC7, 0xEB, 0xE5, +0xF1, 0x7B, 0x3D, 0x07, 0x39, 0xF7, 0x8A, 0x52, 0x92, +0xEA, 0x6B, 0xFB, 0x5F, 0xB1, 0x1F, 0x8D, 0x5D, 0x08, +0x56, 0x03, 0x30, 0x46, 0xFC, 0x7B, 0x6B, 0xAB, 0xF0, +0xCF, 0xBC, 0x20, 0x9A, 0xF4, 0x36, 0x1D, 0xA9, 0xE3, +0x91, 0x61, 0x5E, 0xE6, 0x1B, 0x08, 0x65, 0x99, 0x85, +0x5F, 0x14, 0xA0, 0x68, 0x40, 0x8D, 0xFF, 0xD8, 0x80, +0x4D, 0x73, 0x27, 0x31, 0x06, 0x06, 0x15, 0x56, 0xCA, +0x73, 0xA8, 0xC9, 0x60, 0xE2, 0x7B, 0xC0, 0x8C, 0x6B, +}; + +/* This array is like the one in e_rem_pio2.c, but the numbers are + single precision and the last 8 bits are forced to 0. */ +static const __int32_t npio2_hw[] = { +0x3fc90f00, 0x40490f00, 0x4096cb00, 0x40c90f00, 0x40fb5300, 0x4116cb00, +0x412fed00, 0x41490f00, 0x41623100, 0x417b5300, 0x418a3a00, 0x4196cb00, +0x41a35c00, 0x41afed00, 0x41bc7e00, 0x41c90f00, 0x41d5a000, 0x41e23100, +0x41eec200, 0x41fb5300, 0x4203f200, 0x420a3a00, 0x42108300, 0x4216cb00, +0x421d1400, 0x42235c00, 0x4229a500, 0x422fed00, 0x42363600, 0x423c7e00, +0x4242c700, 0x42490f00 +}; + +/* + * invpio2: 24 bits of 2/pi + * pio2_1: first 17 bit of pi/2 + * pio2_1t: pi/2 - pio2_1 + * pio2_2: second 17 bit of pi/2 + * pio2_2t: pi/2 - (pio2_1+pio2_2) + * pio2_3: third 17 bit of pi/2 + * pio2_3t: pi/2 - (pio2_1+pio2_2+pio2_3) + */ + +static const float +half = 5.0000000000e-01, /* 0x3f000000 */ +invpio2 = 6.3661980629e-01, /* 0x3f22f984 */ +pio2_1 = 1.5707855225e+00, /* 0x3fc90f80 */ +pio2_1t = 1.0804334124e-05, /* 0x37354443 */ +pio2_2 = 1.0804273188e-05, /* 0x37354400 */ +pio2_2t = 6.0770999344e-11, /* 0x2e85a308 */ +pio2_3 = 6.0770943833e-11, /* 0x2e85a300 */ +pio2_3t = 6.1232342629e-17; /* 0x248d3132 */ + +__int32_t __ieee754_rem_pio2f(float x, float *y) +{ + float z,w,t,r,fn; + float tx[3]; + __int32_t i,j,n,ix,hx; + int e0,nx; + + GET_FLOAT_WORD(hx,x); + ix = hx&0x7fffffff; + if(ix<=0x3f490fd8) /* |x| ~<= pi/4 , no need for reduction */ + {y[0] = x; y[1] = 0; return 0;} + if(ix<0x4016cbe4) { /* |x| < 3pi/4, special case with n=+-1 */ + if(hx>0) { + z = x - pio2_1; + if((ix&0xfffffff0)!=0x3fc90fd0) { /* 24+24 bit pi OK */ + y[0] = z - pio2_1t; + y[1] = (z-y[0])-pio2_1t; + } else { /* near pi/2, use 24+24+24 bit pi */ + z -= pio2_2; + y[0] = z - pio2_2t; + y[1] = (z-y[0])-pio2_2t; + } + return 1; + } else { /* negative x */ + z = x + pio2_1; + if((ix&0xfffffff0)!=0x3fc90fd0) { /* 24+24 bit pi OK */ + y[0] = z + pio2_1t; + y[1] = (z-y[0])+pio2_1t; + } else { /* near pi/2, use 24+24+24 bit pi */ + z += pio2_2; + y[0] = z + pio2_2t; + y[1] = (z-y[0])+pio2_2t; + } + return -1; + } + } + if(ix<=0x43490f80) { /* |x| ~<= 2^7*(pi/2), medium size */ + t = fabsf(x); + n = (__int32_t) (t*invpio2+half); + fn = (float)n; + r = t-fn*pio2_1; + w = fn*pio2_1t; /* 1st round good to 40 bit */ + if(n<32&&(ix&0xffffff00)!=npio2_hw[n-1]) { + y[0] = r-w; /* quick check no cancellation */ + } else { + __uint32_t high; + j = ix>>23; + y[0] = r-w; + GET_FLOAT_WORD(high,y[0]); + i = j-((high>>23)&0xff); + if(i>8) { /* 2nd iteration needed, good to 57 */ + t = r; + w = fn*pio2_2; + r = t-w; + w = fn*pio2_2t-((t-r)-w); + y[0] = r-w; + GET_FLOAT_WORD(high,y[0]); + i = j-((high>>23)&0xff); + if(i>25) { /* 3rd iteration need, 74 bits acc */ + t = r; /* will cover all possible cases */ + w = fn*pio2_3; + r = t-w; + w = fn*pio2_3t-((t-r)-w); + y[0] = r-w; + } + } + } + y[1] = (r-y[0])-w; + if(hx<0) {y[0] = -y[0]; y[1] = -y[1]; return -n;} + else return n; + } + /* + * all other (large) arguments + */ + if(!FLT_UWORD_IS_FINITE(ix)) { + y[0]=y[1]=x-x; return 0; + } + /* set z = scalbn(|x|,ilogb(x)-7) */ + e0 = (int)((ix>>23)-134); /* e0 = ilogb(z)-7; */ + SET_FLOAT_WORD(z, ix - ((__int32_t)e0<<23)); + for(i=0;i<2;i++) { + tx[i] = (float)((__int32_t)(z)); + z = (z-tx[i])*two8; + } + tx[2] = z; + nx = 3; + while(tx[nx-1]==zero) nx--; /* skip zero term */ + n = __kernel_rem_pio2f(tx,y,e0,nx,2,two_over_pi); + if(hx<0) {y[0] = -y[0]; y[1] = -y[1]; return -n;} + return n; +} + + +/* kf_sin.c -- float version of k_sin.c + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + */ + + +static const float +S1 = -1.6666667163e-01, /* 0xbe2aaaab */ +S2 = 8.3333337680e-03, /* 0x3c088889 */ +S3 = -1.9841270114e-04, /* 0xb9500d01 */ +S4 = 2.7557314297e-06, /* 0x3638ef1b */ +S5 = -2.5050759689e-08, /* 0xb2d72f34 */ +S6 = 1.5896910177e-10; /* 0x2f2ec9d3 */ + +float __kernel_sinf(float x, float y, int iy) /* iy=0 if y is zero */ +{ + float z,r,v; + __int32_t ix; + GET_FLOAT_WORD(ix,x); + ix &= 0x7fffffff; /* high word of x */ + if(ix<0x32000000) /* |x| < 2**-27 */ + {if((int)x==0) return x;} /* generate inexact */ + z = x*x; + v = z*x; + r = S2+z*(S3+z*(S4+z*(S5+z*S6))); + if(iy==0) return x+v*(S1+z*r); + else return x-((z*(half*y-v*r)-y)-v*S1); +} + + + +/* kf_cos.c -- float version of k_cos.c + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + */ + + +static const float +C1 = 4.1666667908e-02, /* 0x3d2aaaab */ +C2 = -1.3888889225e-03, /* 0xbab60b61 */ +C3 = 2.4801587642e-05, /* 0x37d00d01 */ +C4 = -2.7557314297e-07, /* 0xb493f27c */ +C5 = 2.0875723372e-09, /* 0x310f74f6 */ +C6 = -1.1359647598e-11; /* 0xad47d74e */ + +float __kernel_cosf(float x, float y) +{ + float a,hz,z,r,qx; + __int32_t ix; + GET_FLOAT_WORD(ix,x); + ix &= 0x7fffffff; /* ix = |x|'s high word*/ + if(ix<0x32000000) { /* if x < 2**27 */ + if(((int)x)==0) return one; /* generate inexact */ + } + z = x*x; + r = z*(C1+z*(C2+z*(C3+z*(C4+z*(C5+z*C6))))); + if(ix < 0x3e99999a) /* if |x| < 0.3 */ + return one - ((float)0.5*z - (z*r - x*y)); + else { + if(ix > 0x3f480000) { /* x > 0.78125 */ + qx = (float)0.28125; + } else { + SET_FLOAT_WORD(qx,ix-0x01000000); /* x/4 */ + } + hz = (float)0.5*z-qx; + a = one-qx; + return a - (hz - (z*r-x*y)); + } +} + + +/* sf_sin.c -- float version of s_sin.c. + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + */ + + +float sinf(float x) +{ + float y[2],z=0.0; + __int32_t n,ix; + + GET_FLOAT_WORD(ix,x); + + /* |x| ~< pi/4 */ + ix &= 0x7fffffff; + if(ix <= 0x3f490fd8) return __kernel_sinf(x,z,0); + + /* sin(Inf or NaN) is NaN */ + else if (!FLT_UWORD_IS_FINITE(ix)) return x-x; + + /* argument reduction needed */ + else { + n = __ieee754_rem_pio2f(x,y); + switch(n&3) { + case 0: return __kernel_sinf(y[0],y[1],1); + case 1: return __kernel_cosf(y[0],y[1]); + case 2: return -__kernel_sinf(y[0],y[1],1); + default: + return -__kernel_cosf(y[0],y[1]); + } + } +} + + +/* sf_cos.c -- float version of s_cos.c. + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ + +/* + * ==================================================== + * 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. + * ==================================================== + */ + + +float cosf(float x) +{ + float y[2],z=0.0; + __int32_t n,ix; + + GET_FLOAT_WORD(ix,x); + + /* |x| ~< pi/4 */ + ix &= 0x7fffffff; + if(ix <= 0x3f490fd8) return __kernel_cosf(x,z); + + /* cos(Inf or NaN) is NaN */ + else if (!FLT_UWORD_IS_FINITE(ix)) return x-x; + + /* argument reduction needed */ + else { + n = __ieee754_rem_pio2f(x,y); + switch(n&3) { + case 0: return __kernel_cosf(y[0],y[1]); + case 1: return -__kernel_sinf(y[0],y[1],1); + case 2: return -__kernel_cosf(y[0],y[1]); + default: + return __kernel_sinf(y[0],y[1],1); + } + } +} |