lh | 9ed821d | 2023-04-07 01:36:19 -0700 | [diff] [blame^] | 1 | /* Return arc hyperbole sine for float value, with the imaginary part |
| 2 | of the result possibly adjusted for use in computing other |
| 3 | functions. |
| 4 | Copyright (C) 1997-2015 Free Software Foundation, Inc. |
| 5 | This file is part of the GNU C Library. |
| 6 | |
| 7 | The GNU C Library is free software; you can redistribute it and/or |
| 8 | modify it under the terms of the GNU Lesser General Public |
| 9 | License as published by the Free Software Foundation; either |
| 10 | version 2.1 of the License, or (at your option) any later version. |
| 11 | |
| 12 | The GNU C Library is distributed in the hope that it will be useful, |
| 13 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | Lesser General Public License for more details. |
| 16 | |
| 17 | You should have received a copy of the GNU Lesser General Public |
| 18 | License along with the GNU C Library; if not, see |
| 19 | <http://www.gnu.org/licenses/>. */ |
| 20 | |
| 21 | #include <complex.h> |
| 22 | #include <math.h> |
| 23 | #include <math_private.h> |
| 24 | #include <float.h> |
| 25 | |
| 26 | /* Return the complex inverse hyperbolic sine of finite nonzero Z, |
| 27 | with the imaginary part of the result subtracted from pi/2 if ADJ |
| 28 | is nonzero. */ |
| 29 | |
| 30 | __complex__ float |
| 31 | __kernel_casinhf (__complex__ float x, int adj) |
| 32 | { |
| 33 | __complex__ float res; |
| 34 | float rx, ix; |
| 35 | __complex__ float y; |
| 36 | |
| 37 | /* Avoid cancellation by reducing to the first quadrant. */ |
| 38 | rx = fabsf (__real__ x); |
| 39 | ix = fabsf (__imag__ x); |
| 40 | |
| 41 | if (rx >= 1.0f / FLT_EPSILON || ix >= 1.0f / FLT_EPSILON) |
| 42 | { |
| 43 | /* For large x in the first quadrant, x + csqrt (1 + x * x) |
| 44 | is sufficiently close to 2 * x to make no significant |
| 45 | difference to the result; avoid possible overflow from |
| 46 | the squaring and addition. */ |
| 47 | __real__ y = rx; |
| 48 | __imag__ y = ix; |
| 49 | |
| 50 | if (adj) |
| 51 | { |
| 52 | float t = __real__ y; |
| 53 | __real__ y = __copysignf (__imag__ y, __imag__ x); |
| 54 | __imag__ y = t; |
| 55 | } |
| 56 | |
| 57 | res = __clogf (y); |
| 58 | __real__ res += (float) M_LN2; |
| 59 | } |
| 60 | else if (rx >= 0.5f && ix < FLT_EPSILON / 8.0f) |
| 61 | { |
| 62 | float s = __ieee754_hypotf (1.0f, rx); |
| 63 | |
| 64 | __real__ res = __ieee754_logf (rx + s); |
| 65 | if (adj) |
| 66 | __imag__ res = __ieee754_atan2f (s, __imag__ x); |
| 67 | else |
| 68 | __imag__ res = __ieee754_atan2f (ix, s); |
| 69 | } |
| 70 | else if (rx < FLT_EPSILON / 8.0f && ix >= 1.5f) |
| 71 | { |
| 72 | float s = __ieee754_sqrtf ((ix + 1.0f) * (ix - 1.0f)); |
| 73 | |
| 74 | __real__ res = __ieee754_logf (ix + s); |
| 75 | if (adj) |
| 76 | __imag__ res = __ieee754_atan2f (rx, __copysignf (s, __imag__ x)); |
| 77 | else |
| 78 | __imag__ res = __ieee754_atan2f (s, rx); |
| 79 | } |
| 80 | else if (ix > 1.0f && ix < 1.5f && rx < 0.5f) |
| 81 | { |
| 82 | if (rx < FLT_EPSILON * FLT_EPSILON) |
| 83 | { |
| 84 | float ix2m1 = (ix + 1.0f) * (ix - 1.0f); |
| 85 | float s = __ieee754_sqrtf (ix2m1); |
| 86 | |
| 87 | __real__ res = __log1pf (2.0f * (ix2m1 + ix * s)) / 2.0f; |
| 88 | if (adj) |
| 89 | __imag__ res = __ieee754_atan2f (rx, __copysignf (s, __imag__ x)); |
| 90 | else |
| 91 | __imag__ res = __ieee754_atan2f (s, rx); |
| 92 | } |
| 93 | else |
| 94 | { |
| 95 | float ix2m1 = (ix + 1.0f) * (ix - 1.0f); |
| 96 | float rx2 = rx * rx; |
| 97 | float f = rx2 * (2.0f + rx2 + 2.0f * ix * ix); |
| 98 | float d = __ieee754_sqrtf (ix2m1 * ix2m1 + f); |
| 99 | float dp = d + ix2m1; |
| 100 | float dm = f / dp; |
| 101 | float r1 = __ieee754_sqrtf ((dm + rx2) / 2.0f); |
| 102 | float r2 = rx * ix / r1; |
| 103 | |
| 104 | __real__ res |
| 105 | = __log1pf (rx2 + dp + 2.0f * (rx * r1 + ix * r2)) / 2.0f; |
| 106 | if (adj) |
| 107 | __imag__ res = __ieee754_atan2f (rx + r1, __copysignf (ix + r2, |
| 108 | __imag__ x)); |
| 109 | else |
| 110 | __imag__ res = __ieee754_atan2f (ix + r2, rx + r1); |
| 111 | } |
| 112 | } |
| 113 | else if (ix == 1.0f && rx < 0.5f) |
| 114 | { |
| 115 | if (rx < FLT_EPSILON / 8.0f) |
| 116 | { |
| 117 | __real__ res = __log1pf (2.0f * (rx + __ieee754_sqrtf (rx))) / 2.0f; |
| 118 | if (adj) |
| 119 | __imag__ res = __ieee754_atan2f (__ieee754_sqrtf (rx), |
| 120 | __copysignf (1.0f, __imag__ x)); |
| 121 | else |
| 122 | __imag__ res = __ieee754_atan2f (1.0f, __ieee754_sqrtf (rx)); |
| 123 | } |
| 124 | else |
| 125 | { |
| 126 | float d = rx * __ieee754_sqrtf (4.0f + rx * rx); |
| 127 | float s1 = __ieee754_sqrtf ((d + rx * rx) / 2.0f); |
| 128 | float s2 = __ieee754_sqrtf ((d - rx * rx) / 2.0f); |
| 129 | |
| 130 | __real__ res = __log1pf (rx * rx + d + 2.0f * (rx * s1 + s2)) / 2.0f; |
| 131 | if (adj) |
| 132 | __imag__ res = __ieee754_atan2f (rx + s1, |
| 133 | __copysignf (1.0f + s2, |
| 134 | __imag__ x)); |
| 135 | else |
| 136 | __imag__ res = __ieee754_atan2f (1.0f + s2, rx + s1); |
| 137 | } |
| 138 | } |
| 139 | else if (ix < 1.0f && rx < 0.5f) |
| 140 | { |
| 141 | if (ix >= FLT_EPSILON) |
| 142 | { |
| 143 | if (rx < FLT_EPSILON * FLT_EPSILON) |
| 144 | { |
| 145 | float onemix2 = (1.0f + ix) * (1.0f - ix); |
| 146 | float s = __ieee754_sqrtf (onemix2); |
| 147 | |
| 148 | __real__ res = __log1pf (2.0f * rx / s) / 2.0f; |
| 149 | if (adj) |
| 150 | __imag__ res = __ieee754_atan2f (s, __imag__ x); |
| 151 | else |
| 152 | __imag__ res = __ieee754_atan2f (ix, s); |
| 153 | } |
| 154 | else |
| 155 | { |
| 156 | float onemix2 = (1.0f + ix) * (1.0f - ix); |
| 157 | float rx2 = rx * rx; |
| 158 | float f = rx2 * (2.0f + rx2 + 2.0f * ix * ix); |
| 159 | float d = __ieee754_sqrtf (onemix2 * onemix2 + f); |
| 160 | float dp = d + onemix2; |
| 161 | float dm = f / dp; |
| 162 | float r1 = __ieee754_sqrtf ((dp + rx2) / 2.0f); |
| 163 | float r2 = rx * ix / r1; |
| 164 | |
| 165 | __real__ res |
| 166 | = __log1pf (rx2 + dm + 2.0f * (rx * r1 + ix * r2)) / 2.0f; |
| 167 | if (adj) |
| 168 | __imag__ res = __ieee754_atan2f (rx + r1, |
| 169 | __copysignf (ix + r2, |
| 170 | __imag__ x)); |
| 171 | else |
| 172 | __imag__ res = __ieee754_atan2f (ix + r2, rx + r1); |
| 173 | } |
| 174 | } |
| 175 | else |
| 176 | { |
| 177 | float s = __ieee754_hypotf (1.0f, rx); |
| 178 | |
| 179 | __real__ res = __log1pf (2.0f * rx * (rx + s)) / 2.0f; |
| 180 | if (adj) |
| 181 | __imag__ res = __ieee754_atan2f (s, __imag__ x); |
| 182 | else |
| 183 | __imag__ res = __ieee754_atan2f (ix, s); |
| 184 | } |
| 185 | if (__real__ res < FLT_MIN) |
| 186 | { |
| 187 | volatile float force_underflow = __real__ res * __real__ res; |
| 188 | (void) force_underflow; |
| 189 | } |
| 190 | } |
| 191 | else |
| 192 | { |
| 193 | __real__ y = (rx - ix) * (rx + ix) + 1.0f; |
| 194 | __imag__ y = 2.0f * rx * ix; |
| 195 | |
| 196 | y = __csqrtf (y); |
| 197 | |
| 198 | __real__ y += rx; |
| 199 | __imag__ y += ix; |
| 200 | |
| 201 | if (adj) |
| 202 | { |
| 203 | float t = __real__ y; |
| 204 | __real__ y = __copysignf (__imag__ y, __imag__ x); |
| 205 | __imag__ y = t; |
| 206 | } |
| 207 | |
| 208 | res = __clogf (y); |
| 209 | } |
| 210 | |
| 211 | /* Give results the correct sign for the original argument. */ |
| 212 | __real__ res = __copysignf (__real__ res, __real__ x); |
| 213 | __imag__ res = __copysignf (__imag__ res, (adj ? 1.0f : __imag__ x)); |
| 214 | |
| 215 | return res; |
| 216 | } |