| 1 |
// Start of scalar.h. |
| 2 |
|
| 3 |
// Implementation of the primitive scalar operations. Very |
| 4 |
// repetitive. This code is inserted directly into both CUDA and |
| 5 |
// OpenCL programs, as well as the CPU code, so it has some #ifdefs to |
| 6 |
// work everywhere. Some operations are defined as macros because |
| 7 |
// this allows us to use them as constant expressions in things like |
| 8 |
// array sizes and static initialisers. |
| 9 |
|
| 10 |
// Some of the #ifdefs are because OpenCL uses type-generic functions |
| 11 |
// for some operations (e.g. sqrt), while C and CUDA sensibly use |
| 12 |
// distinct functions for different precisions (e.g. sqrtf() and |
| 13 |
// sqrt()). This is quite annoying. Due to C's unfortunate casting |
| 14 |
// rules, it is also really easy to accidentally implement |
| 15 |
// floating-point functions in the wrong precision, so be careful. |
| 16 |
|
| 17 |
// Double-precision definitions are only included if the preprocessor |
| 18 |
// macro FUTHARK_F64_ENABLED is set. |
| 19 |
|
| 20 |
#ifndef M_PI |
| 21 |
#define M_PI 3.141592653589793 |
| 22 |
#endif |
| 23 |
|
| 24 |
SCALAR_FUN_ATTR int32_t fptobits_f32_i32(float x); |
| 25 |
SCALAR_FUN_ATTR float bitstofp_i32_f32(int32_t x); |
| 26 |
|
| 27 |
SCALAR_FUN_ATTR uint8_t add8(uint8_t x, uint8_t y) { return x + y; } |
| 28 |
SCALAR_FUN_ATTR uint16_t add16(uint16_t x, uint16_t y) { return x + y; } |
| 29 |
SCALAR_FUN_ATTR uint32_t add32(uint32_t x, uint32_t y) { return x + y; } |
| 30 |
SCALAR_FUN_ATTR uint64_t add64(uint64_t x, uint64_t y) { return x + y; } |
| 31 |
|
| 32 |
SCALAR_FUN_ATTR uint8_t sub8(uint8_t x, uint8_t y) { return x - y; } |
| 33 |
SCALAR_FUN_ATTR uint16_t sub16(uint16_t x, uint16_t y) { return x - y; } |
| 34 |
SCALAR_FUN_ATTR uint32_t sub32(uint32_t x, uint32_t y) { return x - y; } |
| 35 |
SCALAR_FUN_ATTR uint64_t sub64(uint64_t x, uint64_t y) { return x - y; } |
| 36 |
|
| 37 |
SCALAR_FUN_ATTR uint8_t mul8(uint8_t x, uint8_t y) { return x * y; } |
| 38 |
SCALAR_FUN_ATTR uint16_t mul16(uint16_t x, uint16_t y) { return x * y; } |
| 39 |
SCALAR_FUN_ATTR uint32_t mul32(uint32_t x, uint32_t y) { return x * y; } |
| 40 |
SCALAR_FUN_ATTR uint64_t mul64(uint64_t x, uint64_t y) { return x * y; } |
| 41 |
|
| 42 |
#if defined(ISPC) |
| 43 |
|
| 44 |
SCALAR_FUN_ATTR uint8_t udiv8(uint8_t x, uint8_t y) { |
| 45 |
// This strange pattern is used to prevent the ISPC compiler from |
| 46 |
// causing SIGFPEs and bogus results on divisions where inactive lanes |
| 47 |
// have 0-valued divisors. It ensures that any inactive lane instead |
| 48 |
// has a divisor of 1. https://github.com/ispc/ispc/issues/2292 |
| 49 |
uint8_t ys = 1; |
| 50 |
foreach_active(i) { ys = y; } |
| 51 |
return x / ys; |
| 52 |
} |
| 53 |
|
| 54 |
SCALAR_FUN_ATTR uint16_t udiv16(uint16_t x, uint16_t y) { |
| 55 |
uint16_t ys = 1; |
| 56 |
foreach_active(i) { ys = y; } |
| 57 |
return x / ys; |
| 58 |
} |
| 59 |
|
| 60 |
SCALAR_FUN_ATTR uint32_t udiv32(uint32_t x, uint32_t y) { |
| 61 |
uint32_t ys = 1; |
| 62 |
foreach_active(i) { ys = y; } |
| 63 |
return x / ys; |
| 64 |
} |
| 65 |
|
| 66 |
SCALAR_FUN_ATTR uint64_t udiv64(uint64_t x, uint64_t y) { |
| 67 |
uint64_t ys = 1; |
| 68 |
foreach_active(i) { ys = y; } |
| 69 |
return x / ys; |
| 70 |
} |
| 71 |
|
| 72 |
SCALAR_FUN_ATTR uint8_t udiv_up8(uint8_t x, uint8_t y) { |
| 73 |
uint8_t ys = 1; |
| 74 |
foreach_active(i) { ys = y; } |
| 75 |
return (x + y - 1) / ys; |
| 76 |
} |
| 77 |
|
| 78 |
SCALAR_FUN_ATTR uint16_t udiv_up16(uint16_t x, uint16_t y) { |
| 79 |
uint16_t ys = 1; |
| 80 |
foreach_active(i) { ys = y; } |
| 81 |
return (x + y - 1) / ys; |
| 82 |
} |
| 83 |
|
| 84 |
SCALAR_FUN_ATTR uint32_t udiv_up32(uint32_t x, uint32_t y) { |
| 85 |
uint32_t ys = 1; |
| 86 |
foreach_active(i) { ys = y; } |
| 87 |
return (x + y - 1) / ys; |
| 88 |
} |
| 89 |
|
| 90 |
SCALAR_FUN_ATTR uint64_t udiv_up64(uint64_t x, uint64_t y) { |
| 91 |
uint64_t ys = 1; |
| 92 |
foreach_active(i) { ys = y; } |
| 93 |
return (x + y - 1) / ys; |
| 94 |
} |
| 95 |
|
| 96 |
SCALAR_FUN_ATTR uint8_t umod8(uint8_t x, uint8_t y) { |
| 97 |
uint8_t ys = 1; |
| 98 |
foreach_active(i) { ys = y; } |
| 99 |
return x % ys; |
| 100 |
} |
| 101 |
|
| 102 |
SCALAR_FUN_ATTR uint16_t umod16(uint16_t x, uint16_t y) { |
| 103 |
uint16_t ys = 1; |
| 104 |
foreach_active(i) { ys = y; } |
| 105 |
return x % ys; |
| 106 |
} |
| 107 |
|
| 108 |
SCALAR_FUN_ATTR uint32_t umod32(uint32_t x, uint32_t y) { |
| 109 |
uint32_t ys = 1; |
| 110 |
foreach_active(i) { ys = y; } |
| 111 |
return x % ys; |
| 112 |
} |
| 113 |
|
| 114 |
SCALAR_FUN_ATTR uint64_t umod64(uint64_t x, uint64_t y) { |
| 115 |
uint64_t ys = 1; |
| 116 |
foreach_active(i) { ys = y; } |
| 117 |
return x % ys; |
| 118 |
} |
| 119 |
|
| 120 |
SCALAR_FUN_ATTR uint8_t udiv_safe8(uint8_t x, uint8_t y) { |
| 121 |
uint8_t ys = 1; |
| 122 |
foreach_active(i) { ys = y; } |
| 123 |
return y == 0 ? 0 : x / ys; |
| 124 |
} |
| 125 |
|
| 126 |
SCALAR_FUN_ATTR uint16_t udiv_safe16(uint16_t x, uint16_t y) { |
| 127 |
uint16_t ys = 1; |
| 128 |
foreach_active(i) { ys = y; } |
| 129 |
return y == 0 ? 0 : x / ys; |
| 130 |
} |
| 131 |
|
| 132 |
SCALAR_FUN_ATTR uint32_t udiv_safe32(uint32_t x, uint32_t y) { |
| 133 |
uint32_t ys = 1; |
| 134 |
foreach_active(i) { ys = y; } |
| 135 |
return y == 0 ? 0 : x / ys; |
| 136 |
} |
| 137 |
|
| 138 |
SCALAR_FUN_ATTR uint64_t udiv_safe64(uint64_t x, uint64_t y) { |
| 139 |
uint64_t ys = 1; |
| 140 |
foreach_active(i) { ys = y; } |
| 141 |
return y == 0 ? 0 : x / ys; |
| 142 |
} |
| 143 |
|
| 144 |
SCALAR_FUN_ATTR uint8_t udiv_up_safe8(uint8_t x, uint8_t y) { |
| 145 |
uint8_t ys = 1; |
| 146 |
foreach_active(i) { ys = y; } |
| 147 |
return y == 0 ? 0 : (x + y - 1) / ys; |
| 148 |
} |
| 149 |
|
| 150 |
SCALAR_FUN_ATTR uint16_t udiv_up_safe16(uint16_t x, uint16_t y) { |
| 151 |
uint16_t ys = 1; |
| 152 |
foreach_active(i) { ys = y; } |
| 153 |
return y == 0 ? 0 : (x + y - 1) / ys; |
| 154 |
} |
| 155 |
|
| 156 |
SCALAR_FUN_ATTR uint32_t udiv_up_safe32(uint32_t x, uint32_t y) { |
| 157 |
uint32_t ys = 1; |
| 158 |
foreach_active(i) { ys = y; } |
| 159 |
return y == 0 ? 0 : (x + y - 1) / ys; |
| 160 |
} |
| 161 |
|
| 162 |
SCALAR_FUN_ATTR uint64_t udiv_up_safe64(uint64_t x, uint64_t y) { |
| 163 |
uint64_t ys = 1; |
| 164 |
foreach_active(i) { ys = y; } |
| 165 |
return y == 0 ? 0 : (x + y - 1) / ys; |
| 166 |
} |
| 167 |
|
| 168 |
SCALAR_FUN_ATTR uint8_t umod_safe8(uint8_t x, uint8_t y) { |
| 169 |
uint8_t ys = 1; |
| 170 |
foreach_active(i) { ys = y; } |
| 171 |
return y == 0 ? 0 : x % ys; |
| 172 |
} |
| 173 |
|
| 174 |
SCALAR_FUN_ATTR uint16_t umod_safe16(uint16_t x, uint16_t y) { |
| 175 |
uint16_t ys = 1; |
| 176 |
foreach_active(i) { ys = y; } |
| 177 |
return y == 0 ? 0 : x % ys; |
| 178 |
} |
| 179 |
|
| 180 |
SCALAR_FUN_ATTR uint32_t umod_safe32(uint32_t x, uint32_t y) { |
| 181 |
uint32_t ys = 1; |
| 182 |
foreach_active(i) { ys = y; } |
| 183 |
return y == 0 ? 0 : x % ys; |
| 184 |
} |
| 185 |
|
| 186 |
SCALAR_FUN_ATTR uint64_t umod_safe64(uint64_t x, uint64_t y) { |
| 187 |
uint64_t ys = 1; |
| 188 |
foreach_active(i) { ys = y; } |
| 189 |
return y == 0 ? 0 : x % ys; |
| 190 |
} |
| 191 |
|
| 192 |
SCALAR_FUN_ATTR int8_t sdiv8(int8_t x, int8_t y) { |
| 193 |
int8_t ys = 1; |
| 194 |
foreach_active(i) { ys = y; } |
| 195 |
int8_t q = x / ys; |
| 196 |
int8_t r = x % ys; |
| 197 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 198 |
} |
| 199 |
|
| 200 |
SCALAR_FUN_ATTR int16_t sdiv16(int16_t x, int16_t y) { |
| 201 |
int16_t ys = 1; |
| 202 |
foreach_active(i) { ys = y; } |
| 203 |
int16_t q = x / ys; |
| 204 |
int16_t r = x % ys; |
| 205 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 206 |
} |
| 207 |
|
| 208 |
SCALAR_FUN_ATTR int32_t sdiv32(int32_t x, int32_t y) { |
| 209 |
int32_t ys = 1; |
| 210 |
foreach_active(i) { ys = y; } |
| 211 |
int32_t q = x / ys; |
| 212 |
int32_t r = x % ys; |
| 213 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 214 |
} |
| 215 |
|
| 216 |
SCALAR_FUN_ATTR int64_t sdiv64(int64_t x, int64_t y) { |
| 217 |
int64_t ys = 1; |
| 218 |
foreach_active(i) { ys = y; } |
| 219 |
int64_t q = x / ys; |
| 220 |
int64_t r = x % ys; |
| 221 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 222 |
} |
| 223 |
|
| 224 |
SCALAR_FUN_ATTR int8_t sdiv_up8(int8_t x, int8_t y) { return sdiv8(x + y - 1, y); } |
| 225 |
SCALAR_FUN_ATTR int16_t sdiv_up16(int16_t x, int16_t y) { return sdiv16(x + y - 1, y); } |
| 226 |
SCALAR_FUN_ATTR int32_t sdiv_up32(int32_t x, int32_t y) { return sdiv32(x + y - 1, y); } |
| 227 |
SCALAR_FUN_ATTR int64_t sdiv_up64(int64_t x, int64_t y) { return sdiv64(x + y - 1, y); } |
| 228 |
|
| 229 |
SCALAR_FUN_ATTR int8_t smod8(int8_t x, int8_t y) { |
| 230 |
int8_t ys = 1; |
| 231 |
foreach_active(i) { ys = y; } |
| 232 |
int8_t r = x % ys; |
| 233 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 234 |
} |
| 235 |
|
| 236 |
SCALAR_FUN_ATTR int16_t smod16(int16_t x, int16_t y) { |
| 237 |
int16_t ys = 1; |
| 238 |
foreach_active(i) { ys = y; } |
| 239 |
int16_t r = x % ys; |
| 240 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 241 |
} |
| 242 |
|
| 243 |
SCALAR_FUN_ATTR int32_t smod32(int32_t x, int32_t y) { |
| 244 |
int32_t ys = 1; |
| 245 |
foreach_active(i) { ys = y; } |
| 246 |
int32_t r = x % ys; |
| 247 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 248 |
} |
| 249 |
|
| 250 |
SCALAR_FUN_ATTR int64_t smod64(int64_t x, int64_t y) { |
| 251 |
int64_t ys = 1; |
| 252 |
foreach_active(i) { ys = y; } |
| 253 |
int64_t r = x % ys; |
| 254 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 255 |
} |
| 256 |
|
| 257 |
SCALAR_FUN_ATTR int8_t sdiv_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : sdiv8(x, y); } |
| 258 |
SCALAR_FUN_ATTR int16_t sdiv_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : sdiv16(x, y); } |
| 259 |
SCALAR_FUN_ATTR int32_t sdiv_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : sdiv32(x, y); } |
| 260 |
SCALAR_FUN_ATTR int64_t sdiv_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : sdiv64(x, y); } |
| 261 |
|
| 262 |
SCALAR_FUN_ATTR int8_t sdiv_up_safe8(int8_t x, int8_t y) { return sdiv_safe8(x + y - 1, y); } |
| 263 |
SCALAR_FUN_ATTR int16_t sdiv_up_safe16(int16_t x, int16_t y) { return sdiv_safe16(x + y - 1, y); } |
| 264 |
SCALAR_FUN_ATTR int32_t sdiv_up_safe32(int32_t x, int32_t y) { return sdiv_safe32(x + y - 1, y); } |
| 265 |
SCALAR_FUN_ATTR int64_t sdiv_up_safe64(int64_t x, int64_t y) { return sdiv_safe64(x + y - 1, y); } |
| 266 |
|
| 267 |
SCALAR_FUN_ATTR int8_t smod_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : smod8(x, y); } |
| 268 |
SCALAR_FUN_ATTR int16_t smod_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : smod16(x, y); } |
| 269 |
SCALAR_FUN_ATTR int32_t smod_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : smod32(x, y); } |
| 270 |
SCALAR_FUN_ATTR int64_t smod_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : smod64(x, y); } |
| 271 |
|
| 272 |
SCALAR_FUN_ATTR int8_t squot8(int8_t x, int8_t y) { |
| 273 |
int8_t ys = 1; |
| 274 |
foreach_active(i) { ys = y; } |
| 275 |
return x / ys; |
| 276 |
} |
| 277 |
|
| 278 |
SCALAR_FUN_ATTR int16_t squot16(int16_t x, int16_t y) { |
| 279 |
int16_t ys = 1; |
| 280 |
foreach_active(i) { ys = y; } |
| 281 |
return x / ys; |
| 282 |
} |
| 283 |
|
| 284 |
SCALAR_FUN_ATTR int32_t squot32(int32_t x, int32_t y) { |
| 285 |
int32_t ys = 1; |
| 286 |
foreach_active(i) { ys = y; } |
| 287 |
return x / ys; |
| 288 |
} |
| 289 |
|
| 290 |
SCALAR_FUN_ATTR int64_t squot64(int64_t x, int64_t y) { |
| 291 |
int64_t ys = 1; |
| 292 |
foreach_active(i) { ys = y; } |
| 293 |
return x / ys; |
| 294 |
} |
| 295 |
|
| 296 |
SCALAR_FUN_ATTR int8_t srem8(int8_t x, int8_t y) { |
| 297 |
int8_t ys = 1; |
| 298 |
foreach_active(i) { ys = y; } |
| 299 |
return x % ys; |
| 300 |
} |
| 301 |
|
| 302 |
SCALAR_FUN_ATTR int16_t srem16(int16_t x, int16_t y) { |
| 303 |
int16_t ys = 1; |
| 304 |
foreach_active(i) { ys = y; } |
| 305 |
return x % ys; |
| 306 |
} |
| 307 |
|
| 308 |
SCALAR_FUN_ATTR int32_t srem32(int32_t x, int32_t y) { |
| 309 |
int32_t ys = 1; |
| 310 |
foreach_active(i) { ys = y; } |
| 311 |
return x % ys; |
| 312 |
} |
| 313 |
|
| 314 |
SCALAR_FUN_ATTR int64_t srem64(int64_t x, int64_t y) { |
| 315 |
int8_t ys = 1; |
| 316 |
foreach_active(i) { ys = y; } |
| 317 |
return x % ys; |
| 318 |
} |
| 319 |
|
| 320 |
SCALAR_FUN_ATTR int8_t squot_safe8(int8_t x, int8_t y) { |
| 321 |
int8_t ys = 1; |
| 322 |
foreach_active(i) { ys = y; } |
| 323 |
return y == 0 ? 0 : x / ys; |
| 324 |
} |
| 325 |
|
| 326 |
SCALAR_FUN_ATTR int16_t squot_safe16(int16_t x, int16_t y) { |
| 327 |
int16_t ys = 1; |
| 328 |
foreach_active(i) { ys = y; } |
| 329 |
return y == 0 ? 0 : x / ys; |
| 330 |
} |
| 331 |
|
| 332 |
SCALAR_FUN_ATTR int32_t squot_safe32(int32_t x, int32_t y) { |
| 333 |
int32_t ys = 1; |
| 334 |
foreach_active(i) { ys = y; } |
| 335 |
return y == 0 ? 0 : x / ys; |
| 336 |
} |
| 337 |
|
| 338 |
SCALAR_FUN_ATTR int64_t squot_safe64(int64_t x, int64_t y) { |
| 339 |
int64_t ys = 1; |
| 340 |
foreach_active(i) { ys = y; } |
| 341 |
return y == 0 ? 0 : x / ys; |
| 342 |
} |
| 343 |
|
| 344 |
SCALAR_FUN_ATTR int8_t srem_safe8(int8_t x, int8_t y) { |
| 345 |
int8_t ys = 1; |
| 346 |
foreach_active(i) { ys = y; } |
| 347 |
return y == 0 ? 0 : x % ys; |
| 348 |
} |
| 349 |
|
| 350 |
SCALAR_FUN_ATTR int16_t srem_safe16(int16_t x, int16_t y) { |
| 351 |
int16_t ys = 1; |
| 352 |
foreach_active(i) { ys = y; } |
| 353 |
return y == 0 ? 0 : x % ys; |
| 354 |
} |
| 355 |
|
| 356 |
SCALAR_FUN_ATTR int32_t srem_safe32(int32_t x, int32_t y) { |
| 357 |
int32_t ys = 1; |
| 358 |
foreach_active(i) { ys = y; } |
| 359 |
return y == 0 ? 0 : x % ys; |
| 360 |
} |
| 361 |
|
| 362 |
SCALAR_FUN_ATTR int64_t srem_safe64(int64_t x, int64_t y) { |
| 363 |
int64_t ys = 1; |
| 364 |
foreach_active(i) { ys = y; } |
| 365 |
return y == 0 ? 0 : x % ys; |
| 366 |
} |
| 367 |
|
| 368 |
#else |
| 369 |
|
| 370 |
SCALAR_FUN_ATTR uint8_t udiv8(uint8_t x, uint8_t y) { return x / y; } |
| 371 |
SCALAR_FUN_ATTR uint16_t udiv16(uint16_t x, uint16_t y) { return x / y; } |
| 372 |
SCALAR_FUN_ATTR uint32_t udiv32(uint32_t x, uint32_t y) { return x / y; } |
| 373 |
SCALAR_FUN_ATTR uint64_t udiv64(uint64_t x, uint64_t y) { return x / y; } |
| 374 |
|
| 375 |
SCALAR_FUN_ATTR uint8_t udiv_up8(uint8_t x, uint8_t y) { return (x + y - 1) / y; } |
| 376 |
SCALAR_FUN_ATTR uint16_t udiv_up16(uint16_t x, uint16_t y) { return (x + y - 1) / y; } |
| 377 |
SCALAR_FUN_ATTR uint32_t udiv_up32(uint32_t x, uint32_t y) { return (x + y - 1) / y; } |
| 378 |
SCALAR_FUN_ATTR uint64_t udiv_up64(uint64_t x, uint64_t y) { return (x + y - 1) / y; } |
| 379 |
|
| 380 |
SCALAR_FUN_ATTR uint8_t umod8(uint8_t x, uint8_t y) { return x % y; } |
| 381 |
SCALAR_FUN_ATTR uint16_t umod16(uint16_t x, uint16_t y) { return x % y; } |
| 382 |
SCALAR_FUN_ATTR uint32_t umod32(uint32_t x, uint32_t y) { return x % y; } |
| 383 |
SCALAR_FUN_ATTR uint64_t umod64(uint64_t x, uint64_t y) { return x % y; } |
| 384 |
|
| 385 |
SCALAR_FUN_ATTR uint8_t udiv_safe8(uint8_t x, uint8_t y) { return y == 0 ? 0 : x / y; } |
| 386 |
SCALAR_FUN_ATTR uint16_t udiv_safe16(uint16_t x, uint16_t y) { return y == 0 ? 0 : x / y; } |
| 387 |
SCALAR_FUN_ATTR uint32_t udiv_safe32(uint32_t x, uint32_t y) { return y == 0 ? 0 : x / y; } |
| 388 |
SCALAR_FUN_ATTR uint64_t udiv_safe64(uint64_t x, uint64_t y) { return y == 0 ? 0 : x / y; } |
| 389 |
|
| 390 |
SCALAR_FUN_ATTR uint8_t udiv_up_safe8(uint8_t x, uint8_t y) { return y == 0 ? 0 : (x + y - 1) / y; } |
| 391 |
SCALAR_FUN_ATTR uint16_t udiv_up_safe16(uint16_t x, uint16_t y) { return y == 0 ? 0 : (x + y - 1) / y; } |
| 392 |
SCALAR_FUN_ATTR uint32_t udiv_up_safe32(uint32_t x, uint32_t y) { return y == 0 ? 0 : (x + y - 1) / y; } |
| 393 |
SCALAR_FUN_ATTR uint64_t udiv_up_safe64(uint64_t x, uint64_t y) { return y == 0 ? 0 : (x + y - 1) / y; } |
| 394 |
|
| 395 |
SCALAR_FUN_ATTR uint8_t umod_safe8(uint8_t x, uint8_t y) { return y == 0 ? 0 : x % y; } |
| 396 |
SCALAR_FUN_ATTR uint16_t umod_safe16(uint16_t x, uint16_t y) { return y == 0 ? 0 : x % y; } |
| 397 |
SCALAR_FUN_ATTR uint32_t umod_safe32(uint32_t x, uint32_t y) { return y == 0 ? 0 : x % y; } |
| 398 |
SCALAR_FUN_ATTR uint64_t umod_safe64(uint64_t x, uint64_t y) { return y == 0 ? 0 : x % y; } |
| 399 |
|
| 400 |
SCALAR_FUN_ATTR int8_t sdiv8(int8_t x, int8_t y) { |
| 401 |
int8_t q = x / y; |
| 402 |
int8_t r = x % y; |
| 403 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 404 |
} |
| 405 |
|
| 406 |
SCALAR_FUN_ATTR int16_t sdiv16(int16_t x, int16_t y) { |
| 407 |
int16_t q = x / y; |
| 408 |
int16_t r = x % y; |
| 409 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 410 |
} |
| 411 |
|
| 412 |
SCALAR_FUN_ATTR int32_t sdiv32(int32_t x, int32_t y) { |
| 413 |
int32_t q = x / y; |
| 414 |
int32_t r = x % y; |
| 415 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 416 |
} |
| 417 |
|
| 418 |
SCALAR_FUN_ATTR int64_t sdiv64(int64_t x, int64_t y) { |
| 419 |
int64_t q = x / y; |
| 420 |
int64_t r = x % y; |
| 421 |
return q - ((r != 0 && r < 0 != y < 0) ? 1 : 0); |
| 422 |
} |
| 423 |
|
| 424 |
SCALAR_FUN_ATTR int8_t sdiv_up8(int8_t x, int8_t y) { return sdiv8(x + y - 1, y); } |
| 425 |
SCALAR_FUN_ATTR int16_t sdiv_up16(int16_t x, int16_t y) { return sdiv16(x + y - 1, y); } |
| 426 |
SCALAR_FUN_ATTR int32_t sdiv_up32(int32_t x, int32_t y) { return sdiv32(x + y - 1, y); } |
| 427 |
SCALAR_FUN_ATTR int64_t sdiv_up64(int64_t x, int64_t y) { return sdiv64(x + y - 1, y); } |
| 428 |
|
| 429 |
SCALAR_FUN_ATTR int8_t smod8(int8_t x, int8_t y) { |
| 430 |
int8_t r = x % y; |
| 431 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 432 |
} |
| 433 |
|
| 434 |
SCALAR_FUN_ATTR int16_t smod16(int16_t x, int16_t y) { |
| 435 |
int16_t r = x % y; |
| 436 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 437 |
} |
| 438 |
|
| 439 |
SCALAR_FUN_ATTR int32_t smod32(int32_t x, int32_t y) { |
| 440 |
int32_t r = x % y; |
| 441 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 442 |
} |
| 443 |
|
| 444 |
SCALAR_FUN_ATTR int64_t smod64(int64_t x, int64_t y) { |
| 445 |
int64_t r = x % y; |
| 446 |
return r + (r == 0 || (x > 0 && y > 0) || (x < 0 && y < 0) ? 0 : y); |
| 447 |
} |
| 448 |
|
| 449 |
SCALAR_FUN_ATTR int8_t sdiv_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : sdiv8(x, y); } |
| 450 |
SCALAR_FUN_ATTR int16_t sdiv_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : sdiv16(x, y); } |
| 451 |
SCALAR_FUN_ATTR int32_t sdiv_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : sdiv32(x, y); } |
| 452 |
SCALAR_FUN_ATTR int64_t sdiv_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : sdiv64(x, y); } |
| 453 |
|
| 454 |
SCALAR_FUN_ATTR int8_t sdiv_up_safe8(int8_t x, int8_t y) { return sdiv_safe8(x + y - 1, y);} |
| 455 |
SCALAR_FUN_ATTR int16_t sdiv_up_safe16(int16_t x, int16_t y) { return sdiv_safe16(x + y - 1, y); } |
| 456 |
SCALAR_FUN_ATTR int32_t sdiv_up_safe32(int32_t x, int32_t y) { return sdiv_safe32(x + y - 1, y); } |
| 457 |
SCALAR_FUN_ATTR int64_t sdiv_up_safe64(int64_t x, int64_t y) { return sdiv_safe64(x + y - 1, y); } |
| 458 |
|
| 459 |
SCALAR_FUN_ATTR int8_t smod_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : smod8(x, y); } |
| 460 |
SCALAR_FUN_ATTR int16_t smod_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : smod16(x, y); } |
| 461 |
SCALAR_FUN_ATTR int32_t smod_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : smod32(x, y); } |
| 462 |
SCALAR_FUN_ATTR int64_t smod_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : smod64(x, y); } |
| 463 |
|
| 464 |
SCALAR_FUN_ATTR int8_t squot8(int8_t x, int8_t y) { return x / y; } |
| 465 |
SCALAR_FUN_ATTR int16_t squot16(int16_t x, int16_t y) { return x / y; } |
| 466 |
SCALAR_FUN_ATTR int32_t squot32(int32_t x, int32_t y) { return x / y; } |
| 467 |
SCALAR_FUN_ATTR int64_t squot64(int64_t x, int64_t y) { return x / y; } |
| 468 |
|
| 469 |
SCALAR_FUN_ATTR int8_t srem8(int8_t x, int8_t y) { return x % y; } |
| 470 |
SCALAR_FUN_ATTR int16_t srem16(int16_t x, int16_t y) { return x % y; } |
| 471 |
SCALAR_FUN_ATTR int32_t srem32(int32_t x, int32_t y) { return x % y; } |
| 472 |
SCALAR_FUN_ATTR int64_t srem64(int64_t x, int64_t y) { return x % y; } |
| 473 |
|
| 474 |
SCALAR_FUN_ATTR int8_t squot_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : x / y; } |
| 475 |
SCALAR_FUN_ATTR int16_t squot_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : x / y; } |
| 476 |
SCALAR_FUN_ATTR int32_t squot_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : x / y; } |
| 477 |
SCALAR_FUN_ATTR int64_t squot_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : x / y; } |
| 478 |
|
| 479 |
SCALAR_FUN_ATTR int8_t srem_safe8(int8_t x, int8_t y) { return y == 0 ? 0 : x % y; } |
| 480 |
SCALAR_FUN_ATTR int16_t srem_safe16(int16_t x, int16_t y) { return y == 0 ? 0 : x % y; } |
| 481 |
SCALAR_FUN_ATTR int32_t srem_safe32(int32_t x, int32_t y) { return y == 0 ? 0 : x % y; } |
| 482 |
SCALAR_FUN_ATTR int64_t srem_safe64(int64_t x, int64_t y) { return y == 0 ? 0 : x % y; } |
| 483 |
|
| 484 |
#endif |
| 485 |
|
| 486 |
SCALAR_FUN_ATTR int8_t smin8(int8_t x, int8_t y) { return x < y ? x : y; } |
| 487 |
SCALAR_FUN_ATTR int16_t smin16(int16_t x, int16_t y) { return x < y ? x : y; } |
| 488 |
SCALAR_FUN_ATTR int32_t smin32(int32_t x, int32_t y) { return x < y ? x : y; } |
| 489 |
SCALAR_FUN_ATTR int64_t smin64(int64_t x, int64_t y) { return x < y ? x : y; } |
| 490 |
|
| 491 |
SCALAR_FUN_ATTR uint8_t umin8(uint8_t x, uint8_t y) { return x < y ? x : y; } |
| 492 |
SCALAR_FUN_ATTR uint16_t umin16(uint16_t x, uint16_t y) { return x < y ? x : y; } |
| 493 |
SCALAR_FUN_ATTR uint32_t umin32(uint32_t x, uint32_t y) { return x < y ? x : y; } |
| 494 |
SCALAR_FUN_ATTR uint64_t umin64(uint64_t x, uint64_t y) { return x < y ? x : y; } |
| 495 |
|
| 496 |
SCALAR_FUN_ATTR int8_t smax8(int8_t x, int8_t y) { return x < y ? y : x; } |
| 497 |
SCALAR_FUN_ATTR int16_t smax16(int16_t x, int16_t y) { return x < y ? y : x; } |
| 498 |
SCALAR_FUN_ATTR int32_t smax32(int32_t x, int32_t y) { return x < y ? y : x; } |
| 499 |
SCALAR_FUN_ATTR int64_t smax64(int64_t x, int64_t y) { return x < y ? y : x; } |
| 500 |
|
| 501 |
SCALAR_FUN_ATTR uint8_t umax8(uint8_t x, uint8_t y) { return x < y ? y : x; } |
| 502 |
SCALAR_FUN_ATTR uint16_t umax16(uint16_t x, uint16_t y) { return x < y ? y : x; } |
| 503 |
SCALAR_FUN_ATTR uint32_t umax32(uint32_t x, uint32_t y) { return x < y ? y : x; } |
| 504 |
SCALAR_FUN_ATTR uint64_t umax64(uint64_t x, uint64_t y) { return x < y ? y : x; } |
| 505 |
|
| 506 |
SCALAR_FUN_ATTR uint8_t shl8(uint8_t x, uint8_t y) { return (uint8_t)(x << y); } |
| 507 |
SCALAR_FUN_ATTR uint16_t shl16(uint16_t x, uint16_t y) { return (uint16_t)(x << y); } |
| 508 |
SCALAR_FUN_ATTR uint32_t shl32(uint32_t x, uint32_t y) { return x << y; } |
| 509 |
SCALAR_FUN_ATTR uint64_t shl64(uint64_t x, uint64_t y) { return x << y; } |
| 510 |
|
| 511 |
SCALAR_FUN_ATTR uint8_t lshr8(uint8_t x, uint8_t y) { return x >> y; } |
| 512 |
SCALAR_FUN_ATTR uint16_t lshr16(uint16_t x, uint16_t y) { return x >> y; } |
| 513 |
SCALAR_FUN_ATTR uint32_t lshr32(uint32_t x, uint32_t y) { return x >> y; } |
| 514 |
SCALAR_FUN_ATTR uint64_t lshr64(uint64_t x, uint64_t y) { return x >> y; } |
| 515 |
|
| 516 |
SCALAR_FUN_ATTR int8_t ashr8(int8_t x, int8_t y) { return x >> y; } |
| 517 |
SCALAR_FUN_ATTR int16_t ashr16(int16_t x, int16_t y) { return x >> y; } |
| 518 |
SCALAR_FUN_ATTR int32_t ashr32(int32_t x, int32_t y) { return x >> y; } |
| 519 |
SCALAR_FUN_ATTR int64_t ashr64(int64_t x, int64_t y) { return x >> y; } |
| 520 |
|
| 521 |
SCALAR_FUN_ATTR uint8_t and8(uint8_t x, uint8_t y) { return x & y; } |
| 522 |
SCALAR_FUN_ATTR uint16_t and16(uint16_t x, uint16_t y) { return x & y; } |
| 523 |
SCALAR_FUN_ATTR uint32_t and32(uint32_t x, uint32_t y) { return x & y; } |
| 524 |
SCALAR_FUN_ATTR uint64_t and64(uint64_t x, uint64_t y) { return x & y; } |
| 525 |
|
| 526 |
SCALAR_FUN_ATTR uint8_t or8(uint8_t x, uint8_t y) { return x | y; } |
| 527 |
SCALAR_FUN_ATTR uint16_t or16(uint16_t x, uint16_t y) { return x | y; } |
| 528 |
SCALAR_FUN_ATTR uint32_t or32(uint32_t x, uint32_t y) { return x | y; } |
| 529 |
SCALAR_FUN_ATTR uint64_t or64(uint64_t x, uint64_t y) { return x | y; } |
| 530 |
|
| 531 |
SCALAR_FUN_ATTR uint8_t xor8(uint8_t x, uint8_t y) { return x ^ y; } |
| 532 |
SCALAR_FUN_ATTR uint16_t xor16(uint16_t x, uint16_t y) { return x ^ y; } |
| 533 |
SCALAR_FUN_ATTR uint32_t xor32(uint32_t x, uint32_t y) { return x ^ y; } |
| 534 |
SCALAR_FUN_ATTR uint64_t xor64(uint64_t x, uint64_t y) { return x ^ y; } |
| 535 |
|
| 536 |
SCALAR_FUN_ATTR bool ult8(uint8_t x, uint8_t y) { return x < y; } |
| 537 |
SCALAR_FUN_ATTR bool ult16(uint16_t x, uint16_t y) { return x < y; } |
| 538 |
SCALAR_FUN_ATTR bool ult32(uint32_t x, uint32_t y) { return x < y; } |
| 539 |
SCALAR_FUN_ATTR bool ult64(uint64_t x, uint64_t y) { return x < y; } |
| 540 |
|
| 541 |
SCALAR_FUN_ATTR bool ule8(uint8_t x, uint8_t y) { return x <= y; } |
| 542 |
SCALAR_FUN_ATTR bool ule16(uint16_t x, uint16_t y) { return x <= y; } |
| 543 |
SCALAR_FUN_ATTR bool ule32(uint32_t x, uint32_t y) { return x <= y; } |
| 544 |
SCALAR_FUN_ATTR bool ule64(uint64_t x, uint64_t y) { return x <= y; } |
| 545 |
|
| 546 |
SCALAR_FUN_ATTR bool slt8(int8_t x, int8_t y) { return x < y; } |
| 547 |
SCALAR_FUN_ATTR bool slt16(int16_t x, int16_t y) { return x < y; } |
| 548 |
SCALAR_FUN_ATTR bool slt32(int32_t x, int32_t y) { return x < y; } |
| 549 |
SCALAR_FUN_ATTR bool slt64(int64_t x, int64_t y) { return x < y; } |
| 550 |
|
| 551 |
SCALAR_FUN_ATTR bool sle8(int8_t x, int8_t y) { return x <= y; } |
| 552 |
SCALAR_FUN_ATTR bool sle16(int16_t x, int16_t y) { return x <= y; } |
| 553 |
SCALAR_FUN_ATTR bool sle32(int32_t x, int32_t y) { return x <= y; } |
| 554 |
SCALAR_FUN_ATTR bool sle64(int64_t x, int64_t y) { return x <= y; } |
| 555 |
|
| 556 |
SCALAR_FUN_ATTR uint8_t pow8(uint8_t x, uint8_t y) { |
| 557 |
uint8_t res = 1, rem = y; |
| 558 |
while (rem != 0) { |
| 559 |
if (rem & 1) |
| 560 |
res *= x; |
| 561 |
rem >>= 1; |
| 562 |
x *= x; |
| 563 |
} |
| 564 |
return res; |
| 565 |
} |
| 566 |
|
| 567 |
SCALAR_FUN_ATTR uint16_t pow16(uint16_t x, uint16_t y) { |
| 568 |
uint16_t res = 1, rem = y; |
| 569 |
while (rem != 0) { |
| 570 |
if (rem & 1) |
| 571 |
res *= x; |
| 572 |
rem >>= 1; |
| 573 |
x *= x; |
| 574 |
} |
| 575 |
return res; |
| 576 |
} |
| 577 |
|
| 578 |
SCALAR_FUN_ATTR uint32_t pow32(uint32_t x, uint32_t y) { |
| 579 |
uint32_t res = 1, rem = y; |
| 580 |
while (rem != 0) { |
| 581 |
if (rem & 1) |
| 582 |
res *= x; |
| 583 |
rem >>= 1; |
| 584 |
x *= x; |
| 585 |
} |
| 586 |
return res; |
| 587 |
} |
| 588 |
|
| 589 |
SCALAR_FUN_ATTR uint64_t pow64(uint64_t x, uint64_t y) { |
| 590 |
uint64_t res = 1, rem = y; |
| 591 |
while (rem != 0) { |
| 592 |
if (rem & 1) |
| 593 |
res *= x; |
| 594 |
rem >>= 1; |
| 595 |
x *= x; |
| 596 |
} |
| 597 |
return res; |
| 598 |
} |
| 599 |
|
| 600 |
SCALAR_FUN_ATTR bool itob_i8_bool(int8_t x) { return x != 0; } |
| 601 |
SCALAR_FUN_ATTR bool itob_i16_bool(int16_t x) { return x != 0; } |
| 602 |
SCALAR_FUN_ATTR bool itob_i32_bool(int32_t x) { return x != 0; } |
| 603 |
SCALAR_FUN_ATTR bool itob_i64_bool(int64_t x) { return x != 0; } |
| 604 |
|
| 605 |
SCALAR_FUN_ATTR int8_t btoi_bool_i8(bool x) { return x; } |
| 606 |
SCALAR_FUN_ATTR int16_t btoi_bool_i16(bool x) { return x; } |
| 607 |
SCALAR_FUN_ATTR int32_t btoi_bool_i32(bool x) { return x; } |
| 608 |
SCALAR_FUN_ATTR int64_t btoi_bool_i64(bool x) { return x; } |
| 609 |
|
| 610 |
#define sext_i8_i8(x) ((int8_t) (int8_t) (x)) |
| 611 |
#define sext_i8_i16(x) ((int16_t) (int8_t) (x)) |
| 612 |
#define sext_i8_i32(x) ((int32_t) (int8_t) (x)) |
| 613 |
#define sext_i8_i64(x) ((int64_t) (int8_t) (x)) |
| 614 |
#define sext_i16_i8(x) ((int8_t) (int16_t) (x)) |
| 615 |
#define sext_i16_i16(x) ((int16_t) (int16_t) (x)) |
| 616 |
#define sext_i16_i32(x) ((int32_t) (int16_t) (x)) |
| 617 |
#define sext_i16_i64(x) ((int64_t) (int16_t) (x)) |
| 618 |
#define sext_i32_i8(x) ((int8_t) (int32_t) (x)) |
| 619 |
#define sext_i32_i16(x) ((int16_t) (int32_t) (x)) |
| 620 |
#define sext_i32_i32(x) ((int32_t) (int32_t) (x)) |
| 621 |
#define sext_i32_i64(x) ((int64_t) (int32_t) (x)) |
| 622 |
#define sext_i64_i8(x) ((int8_t) (int64_t) (x)) |
| 623 |
#define sext_i64_i16(x) ((int16_t) (int64_t) (x)) |
| 624 |
#define sext_i64_i32(x) ((int32_t) (int64_t) (x)) |
| 625 |
#define sext_i64_i64(x) ((int64_t) (int64_t) (x)) |
| 626 |
#define zext_i8_i8(x) ((int8_t) (uint8_t) (x)) |
| 627 |
#define zext_i8_i16(x) ((int16_t) (uint8_t) (x)) |
| 628 |
#define zext_i8_i32(x) ((int32_t) (uint8_t) (x)) |
| 629 |
#define zext_i8_i64(x) ((int64_t) (uint8_t) (x)) |
| 630 |
#define zext_i16_i8(x) ((int8_t) (uint16_t) (x)) |
| 631 |
#define zext_i16_i16(x) ((int16_t) (uint16_t) (x)) |
| 632 |
#define zext_i16_i32(x) ((int32_t) (uint16_t) (x)) |
| 633 |
#define zext_i16_i64(x) ((int64_t) (uint16_t) (x)) |
| 634 |
#define zext_i32_i8(x) ((int8_t) (uint32_t) (x)) |
| 635 |
#define zext_i32_i16(x) ((int16_t) (uint32_t) (x)) |
| 636 |
#define zext_i32_i32(x) ((int32_t) (uint32_t) (x)) |
| 637 |
#define zext_i32_i64(x) ((int64_t) (uint32_t) (x)) |
| 638 |
#define zext_i64_i8(x) ((int8_t) (uint64_t) (x)) |
| 639 |
#define zext_i64_i16(x) ((int16_t) (uint64_t) (x)) |
| 640 |
#define zext_i64_i32(x) ((int32_t) (uint64_t) (x)) |
| 641 |
#define zext_i64_i64(x) ((int64_t) (uint64_t) (x)) |
| 642 |
|
| 643 |
SCALAR_FUN_ATTR int8_t abs8(int8_t x) { return (int8_t)abs(x); } |
| 644 |
SCALAR_FUN_ATTR int16_t abs16(int16_t x) { return (int16_t)abs(x); } |
| 645 |
SCALAR_FUN_ATTR int32_t abs32(int32_t x) { return abs(x); } |
| 646 |
SCALAR_FUN_ATTR int64_t abs64(int64_t x) { |
| 647 |
#if defined(__OPENCL_VERSION__) || defined(ISPC) |
| 648 |
return abs(x); |
| 649 |
#else |
| 650 |
return llabs(x); |
| 651 |
#endif |
| 652 |
} |
| 653 |
|
| 654 |
#if defined(__OPENCL_VERSION__) |
| 655 |
|
| 656 |
SCALAR_FUN_ATTR int32_t futrts_popc8(int8_t x) { return popcount(x); } |
| 657 |
SCALAR_FUN_ATTR int32_t futrts_popc16(int16_t x) { return popcount(x); } |
| 658 |
SCALAR_FUN_ATTR int32_t futrts_popc32(int32_t x) { return popcount(x); } |
| 659 |
SCALAR_FUN_ATTR int32_t futrts_popc64(int64_t x) { return popcount(x); } |
| 660 |
|
| 661 |
#elif defined(__CUDA_ARCH__) |
| 662 |
|
| 663 |
SCALAR_FUN_ATTR int32_t futrts_popc8(int8_t x) { return __popc(zext_i8_i32(x)); } |
| 664 |
SCALAR_FUN_ATTR int32_t futrts_popc16(int16_t x) { return __popc(zext_i16_i32(x)); } |
| 665 |
SCALAR_FUN_ATTR int32_t futrts_popc32(int32_t x) { return __popc(x); } |
| 666 |
SCALAR_FUN_ATTR int32_t futrts_popc64(int64_t x) { return __popcll(x); } |
| 667 |
|
| 668 |
#else // Not OpenCL or CUDA, but plain C. |
| 669 |
|
| 670 |
SCALAR_FUN_ATTR int32_t futrts_popc8(uint8_t x) { |
| 671 |
int c = 0; |
| 672 |
for (; x; ++c) { x &= x - 1; } |
| 673 |
return c; |
| 674 |
} |
| 675 |
|
| 676 |
SCALAR_FUN_ATTR int32_t futrts_popc16(uint16_t x) { |
| 677 |
int c = 0; |
| 678 |
for (; x; ++c) { x &= x - 1; } |
| 679 |
return c; |
| 680 |
} |
| 681 |
|
| 682 |
SCALAR_FUN_ATTR int32_t futrts_popc32(uint32_t x) { |
| 683 |
int c = 0; |
| 684 |
for (; x; ++c) { x &= x - 1; } |
| 685 |
return c; |
| 686 |
} |
| 687 |
|
| 688 |
SCALAR_FUN_ATTR int32_t futrts_popc64(uint64_t x) { |
| 689 |
int c = 0; |
| 690 |
for (; x; ++c) { x &= x - 1; } |
| 691 |
return c; |
| 692 |
} |
| 693 |
#endif |
| 694 |
|
| 695 |
#if defined(__OPENCL_VERSION__) |
| 696 |
SCALAR_FUN_ATTR uint8_t futrts_umul_hi8 ( uint8_t a, uint8_t b) { return mul_hi(a, b); } |
| 697 |
SCALAR_FUN_ATTR uint16_t futrts_umul_hi16(uint16_t a, uint16_t b) { return mul_hi(a, b); } |
| 698 |
SCALAR_FUN_ATTR uint32_t futrts_umul_hi32(uint32_t a, uint32_t b) { return mul_hi(a, b); } |
| 699 |
SCALAR_FUN_ATTR uint64_t futrts_umul_hi64(uint64_t a, uint64_t b) { return mul_hi(a, b); } |
| 700 |
SCALAR_FUN_ATTR uint8_t futrts_smul_hi8 ( int8_t a, int8_t b) { return mul_hi(a, b); } |
| 701 |
SCALAR_FUN_ATTR uint16_t futrts_smul_hi16(int16_t a, int16_t b) { return mul_hi(a, b); } |
| 702 |
SCALAR_FUN_ATTR uint32_t futrts_smul_hi32(int32_t a, int32_t b) { return mul_hi(a, b); } |
| 703 |
SCALAR_FUN_ATTR uint64_t futrts_smul_hi64(int64_t a, int64_t b) { return mul_hi(a, b); } |
| 704 |
#elif defined(__CUDA_ARCH__) |
| 705 |
SCALAR_FUN_ATTR uint8_t futrts_umul_hi8(uint8_t a, uint8_t b) { return ((uint16_t)a) * ((uint16_t)b) >> 8; } |
| 706 |
SCALAR_FUN_ATTR uint16_t futrts_umul_hi16(uint16_t a, uint16_t b) { return ((uint32_t)a) * ((uint32_t)b) >> 16; } |
| 707 |
SCALAR_FUN_ATTR uint32_t futrts_umul_hi32(uint32_t a, uint32_t b) { return __umulhi(a, b); } |
| 708 |
SCALAR_FUN_ATTR uint64_t futrts_umul_hi64(uint64_t a, uint64_t b) { return __umul64hi(a, b); } |
| 709 |
SCALAR_FUN_ATTR uint8_t futrts_smul_hi8 ( int8_t a, int8_t b) { return ((int16_t)a) * ((int16_t)b) >> 8; } |
| 710 |
SCALAR_FUN_ATTR uint16_t futrts_smul_hi16(int16_t a, int16_t b) { return ((int32_t)a) * ((int32_t)b) >> 16; } |
| 711 |
SCALAR_FUN_ATTR uint32_t futrts_smul_hi32(int32_t a, int32_t b) { return __mulhi(a, b); } |
| 712 |
SCALAR_FUN_ATTR uint64_t futrts_smul_hi64(int64_t a, int64_t b) { return __mul64hi(a, b); } |
| 713 |
#elif defined(ISPC) |
| 714 |
SCALAR_FUN_ATTR uint8_t futrts_umul_hi8(uint8_t a, uint8_t b) { return ((uint16_t)a) * ((uint16_t)b) >> 8; } |
| 715 |
SCALAR_FUN_ATTR uint16_t futrts_umul_hi16(uint16_t a, uint16_t b) { return ((uint32_t)a) * ((uint32_t)b) >> 16; } |
| 716 |
SCALAR_FUN_ATTR uint32_t futrts_umul_hi32(uint32_t a, uint32_t b) { return ((uint64_t)a) * ((uint64_t)b) >> 32; } |
| 717 |
SCALAR_FUN_ATTR uint64_t futrts_umul_hi64(uint64_t a, uint64_t b) { |
| 718 |
uint64_t ah = a >> 32; |
| 719 |
uint64_t al = a & 0xffffffff; |
| 720 |
uint64_t bh = b >> 32; |
| 721 |
uint64_t bl = b & 0xffffffff; |
| 722 |
|
| 723 |
uint64_t p1 = al * bl; |
| 724 |
uint64_t p2 = al * bh; |
| 725 |
uint64_t p3 = ah * bl; |
| 726 |
uint64_t p4 = ah * bh; |
| 727 |
|
| 728 |
uint64_t p1h = p1 >> 32; |
| 729 |
uint64_t p2h = p2 >> 32; |
| 730 |
uint64_t p3h = p3 >> 32; |
| 731 |
uint64_t p2l = p2 & 0xffffffff; |
| 732 |
uint64_t p3l = p3 & 0xffffffff; |
| 733 |
|
| 734 |
uint64_t l = p1h + p2l + p3l; |
| 735 |
uint64_t m = (p2 >> 32) + (p3 >> 32); |
| 736 |
uint64_t h = (l >> 32) + m + p4; |
| 737 |
|
| 738 |
return h; |
| 739 |
} |
| 740 |
SCALAR_FUN_ATTR int8_t futrts_smul_hi8 ( int8_t a, int8_t b) { return ((uint16_t)a) * ((uint16_t)b) >> 8; } |
| 741 |
SCALAR_FUN_ATTR int16_t futrts_smul_hi16(int16_t a, int16_t b) { return ((uint32_t)a) * ((uint32_t)b) >> 16; } |
| 742 |
SCALAR_FUN_ATTR int32_t futrts_smul_hi32(int32_t a, int32_t b) { return ((uint64_t)a) * ((uint64_t)b) >> 32; } |
| 743 |
SCALAR_FUN_ATTR int64_t futrts_smul_hi64(int64_t a, int64_t b) { |
| 744 |
uint64_t ah = a >> 32; |
| 745 |
uint64_t al = a & 0xffffffff; |
| 746 |
uint64_t bh = b >> 32; |
| 747 |
uint64_t bl = b & 0xffffffff; |
| 748 |
|
| 749 |
uint64_t p1 = al * bl; |
| 750 |
int64_t p2 = al * bh; |
| 751 |
int64_t p3 = ah * bl; |
| 752 |
uint64_t p4 = ah * bh; |
| 753 |
|
| 754 |
uint64_t p1h = p1 >> 32; |
| 755 |
uint64_t p2h = p2 >> 32; |
| 756 |
uint64_t p3h = p3 >> 32; |
| 757 |
uint64_t p2l = p2 & 0xffffffff; |
| 758 |
uint64_t p3l = p3 & 0xffffffff; |
| 759 |
|
| 760 |
uint64_t l = p1h + p2l + p3l; |
| 761 |
uint64_t m = (p2 >> 32) + (p3 >> 32); |
| 762 |
uint64_t h = (l >> 32) + m + p4; |
| 763 |
|
| 764 |
return h; |
| 765 |
} |
| 766 |
|
| 767 |
#else // Not OpenCL, ISPC, or CUDA, but plain C. |
| 768 |
SCALAR_FUN_ATTR uint8_t futrts_umul_hi8(uint8_t a, uint8_t b) { return ((uint16_t)a) * ((uint16_t)b) >> 8; } |
| 769 |
SCALAR_FUN_ATTR uint16_t futrts_umul_hi16(uint16_t a, uint16_t b) { return ((uint32_t)a) * ((uint32_t)b) >> 16; } |
| 770 |
SCALAR_FUN_ATTR uint32_t futrts_umul_hi32(uint32_t a, uint32_t b) { return ((uint64_t)a) * ((uint64_t)b) >> 32; } |
| 771 |
SCALAR_FUN_ATTR uint64_t futrts_umul_hi64(uint64_t a, uint64_t b) { return ((__uint128_t)a) * ((__uint128_t)b) >> 64; } |
| 772 |
SCALAR_FUN_ATTR int8_t futrts_smul_hi8(int8_t a, int8_t b) { return ((int16_t)a) * ((int16_t)b) >> 8; } |
| 773 |
SCALAR_FUN_ATTR int16_t futrts_smul_hi16(int16_t a, int16_t b) { return ((int32_t)a) * ((int32_t)b) >> 16; } |
| 774 |
SCALAR_FUN_ATTR int32_t futrts_smul_hi32(int32_t a, int32_t b) { return ((int64_t)a) * ((int64_t)b) >> 32; } |
| 775 |
SCALAR_FUN_ATTR int64_t futrts_smul_hi64(int64_t a, int64_t b) { return ((__int128_t)a) * ((__int128_t)b) >> 64; } |
| 776 |
#endif |
| 777 |
|
| 778 |
#if defined(__OPENCL_VERSION__) |
| 779 |
SCALAR_FUN_ATTR uint8_t futrts_umad_hi8 ( uint8_t a, uint8_t b, uint8_t c) { return mad_hi(a, b, c); } |
| 780 |
SCALAR_FUN_ATTR uint16_t futrts_umad_hi16(uint16_t a, uint16_t b, uint16_t c) { return mad_hi(a, b, c); } |
| 781 |
SCALAR_FUN_ATTR uint32_t futrts_umad_hi32(uint32_t a, uint32_t b, uint32_t c) { return mad_hi(a, b, c); } |
| 782 |
SCALAR_FUN_ATTR uint64_t futrts_umad_hi64(uint64_t a, uint64_t b, uint64_t c) { return mad_hi(a, b, c); } |
| 783 |
SCALAR_FUN_ATTR uint8_t futrts_smad_hi8( int8_t a, int8_t b, int8_t c) { return mad_hi(a, b, c); } |
| 784 |
SCALAR_FUN_ATTR uint16_t futrts_smad_hi16(int16_t a, int16_t b, int16_t c) { return mad_hi(a, b, c); } |
| 785 |
SCALAR_FUN_ATTR uint32_t futrts_smad_hi32(int32_t a, int32_t b, int32_t c) { return mad_hi(a, b, c); } |
| 786 |
SCALAR_FUN_ATTR uint64_t futrts_smad_hi64(int64_t a, int64_t b, int64_t c) { return mad_hi(a, b, c); } |
| 787 |
#else // Not OpenCL |
| 788 |
|
| 789 |
SCALAR_FUN_ATTR uint8_t futrts_umad_hi8( uint8_t a, uint8_t b, uint8_t c) { return futrts_umul_hi8(a, b) + c; } |
| 790 |
SCALAR_FUN_ATTR uint16_t futrts_umad_hi16(uint16_t a, uint16_t b, uint16_t c) { return futrts_umul_hi16(a, b) + c; } |
| 791 |
SCALAR_FUN_ATTR uint32_t futrts_umad_hi32(uint32_t a, uint32_t b, uint32_t c) { return futrts_umul_hi32(a, b) + c; } |
| 792 |
SCALAR_FUN_ATTR uint64_t futrts_umad_hi64(uint64_t a, uint64_t b, uint64_t c) { return futrts_umul_hi64(a, b) + c; } |
| 793 |
SCALAR_FUN_ATTR uint8_t futrts_smad_hi8 ( int8_t a, int8_t b, int8_t c) { return futrts_smul_hi8(a, b) + c; } |
| 794 |
SCALAR_FUN_ATTR uint16_t futrts_smad_hi16(int16_t a, int16_t b, int16_t c) { return futrts_smul_hi16(a, b) + c; } |
| 795 |
SCALAR_FUN_ATTR uint32_t futrts_smad_hi32(int32_t a, int32_t b, int32_t c) { return futrts_smul_hi32(a, b) + c; } |
| 796 |
SCALAR_FUN_ATTR uint64_t futrts_smad_hi64(int64_t a, int64_t b, int64_t c) { return futrts_smul_hi64(a, b) + c; } |
| 797 |
#endif |
| 798 |
|
| 799 |
#if defined(__OPENCL_VERSION__) |
| 800 |
SCALAR_FUN_ATTR int32_t futrts_clzz8(int8_t x) { return clz(x); } |
| 801 |
SCALAR_FUN_ATTR int32_t futrts_clzz16(int16_t x) { return clz(x); } |
| 802 |
SCALAR_FUN_ATTR int32_t futrts_clzz32(int32_t x) { return clz(x); } |
| 803 |
SCALAR_FUN_ATTR int32_t futrts_clzz64(int64_t x) { return clz(x); } |
| 804 |
|
| 805 |
#elif defined(__CUDA_ARCH__) |
| 806 |
|
| 807 |
SCALAR_FUN_ATTR int32_t futrts_clzz8(int8_t x) { return __clz(zext_i8_i32(x)) - 24; } |
| 808 |
SCALAR_FUN_ATTR int32_t futrts_clzz16(int16_t x) { return __clz(zext_i16_i32(x)) - 16; } |
| 809 |
SCALAR_FUN_ATTR int32_t futrts_clzz32(int32_t x) { return __clz(x); } |
| 810 |
SCALAR_FUN_ATTR int32_t futrts_clzz64(int64_t x) { return __clzll(x); } |
| 811 |
|
| 812 |
#elif defined(ISPC) |
| 813 |
|
| 814 |
SCALAR_FUN_ATTR int32_t futrts_clzz8(int8_t x) { return count_leading_zeros((int32_t)(uint8_t)x)-24; } |
| 815 |
SCALAR_FUN_ATTR int32_t futrts_clzz16(int16_t x) { return count_leading_zeros((int32_t)(uint16_t)x)-16; } |
| 816 |
SCALAR_FUN_ATTR int32_t futrts_clzz32(int32_t x) { return count_leading_zeros(x); } |
| 817 |
SCALAR_FUN_ATTR int32_t futrts_clzz64(int64_t x) { return count_leading_zeros(x); } |
| 818 |
|
| 819 |
#else // Not OpenCL, ISPC or CUDA, but plain C. |
| 820 |
|
| 821 |
SCALAR_FUN_ATTR int32_t futrts_clzz8(int8_t x) |
| 822 |
{ return x == 0 ? 8 : __builtin_clz((uint32_t)zext_i8_i32(x)) - 24; } |
| 823 |
SCALAR_FUN_ATTR int32_t futrts_clzz16(int16_t x) |
| 824 |
{ return x == 0 ? 16 : __builtin_clz((uint32_t)zext_i16_i32(x)) - 16; } |
| 825 |
SCALAR_FUN_ATTR int32_t futrts_clzz32(int32_t x) |
| 826 |
{ return x == 0 ? 32 : __builtin_clz((uint32_t)x); } |
| 827 |
SCALAR_FUN_ATTR int32_t futrts_clzz64(int64_t x) |
| 828 |
{ return x == 0 ? 64 : __builtin_clzll((uint64_t)x); } |
| 829 |
#endif |
| 830 |
|
| 831 |
#if defined(__OPENCL_VERSION__) |
| 832 |
SCALAR_FUN_ATTR int32_t futrts_ctzz8(int8_t x) { |
| 833 |
int i = 0; |
| 834 |
for (; i < 8 && (x & 1) == 0; i++, x >>= 1) ; |
| 835 |
return i; |
| 836 |
} |
| 837 |
|
| 838 |
SCALAR_FUN_ATTR int32_t futrts_ctzz16(int16_t x) { |
| 839 |
int i = 0; |
| 840 |
for (; i < 16 && (x & 1) == 0; i++, x >>= 1) ; |
| 841 |
return i; |
| 842 |
} |
| 843 |
|
| 844 |
SCALAR_FUN_ATTR int32_t futrts_ctzz32(int32_t x) { |
| 845 |
int i = 0; |
| 846 |
for (; i < 32 && (x & 1) == 0; i++, x >>= 1) ; |
| 847 |
return i; |
| 848 |
} |
| 849 |
|
| 850 |
SCALAR_FUN_ATTR int32_t futrts_ctzz64(int64_t x) { |
| 851 |
int i = 0; |
| 852 |
for (; i < 64 && (x & 1) == 0; i++, x >>= 1) ; |
| 853 |
return i; |
| 854 |
} |
| 855 |
|
| 856 |
#elif defined(__CUDA_ARCH__) |
| 857 |
|
| 858 |
SCALAR_FUN_ATTR int32_t futrts_ctzz8(int8_t x) { |
| 859 |
int y = __ffs(x); |
| 860 |
return y == 0 ? 8 : y - 1; |
| 861 |
} |
| 862 |
|
| 863 |
SCALAR_FUN_ATTR int32_t futrts_ctzz16(int16_t x) { |
| 864 |
int y = __ffs(x); |
| 865 |
return y == 0 ? 16 : y - 1; |
| 866 |
} |
| 867 |
|
| 868 |
SCALAR_FUN_ATTR int32_t futrts_ctzz32(int32_t x) { |
| 869 |
int y = __ffs(x); |
| 870 |
return y == 0 ? 32 : y - 1; |
| 871 |
} |
| 872 |
|
| 873 |
SCALAR_FUN_ATTR int32_t futrts_ctzz64(int64_t x) { |
| 874 |
int y = __ffsll(x); |
| 875 |
return y == 0 ? 64 : y - 1; |
| 876 |
} |
| 877 |
|
| 878 |
#elif defined(ISPC) |
| 879 |
|
| 880 |
SCALAR_FUN_ATTR int32_t futrts_ctzz8(int8_t x) { return x == 0 ? 8 : count_trailing_zeros((int32_t)x); } |
| 881 |
SCALAR_FUN_ATTR int32_t futrts_ctzz16(int16_t x) { return x == 0 ? 16 : count_trailing_zeros((int32_t)x); } |
| 882 |
SCALAR_FUN_ATTR int32_t futrts_ctzz32(int32_t x) { return count_trailing_zeros(x); } |
| 883 |
SCALAR_FUN_ATTR int32_t futrts_ctzz64(int64_t x) { return count_trailing_zeros(x); } |
| 884 |
|
| 885 |
#else // Not OpenCL or CUDA, but plain C. |
| 886 |
|
| 887 |
SCALAR_FUN_ATTR int32_t futrts_ctzz8(int8_t x) { return x == 0 ? 8 : __builtin_ctz((uint32_t)x); } |
| 888 |
SCALAR_FUN_ATTR int32_t futrts_ctzz16(int16_t x) { return x == 0 ? 16 : __builtin_ctz((uint32_t)x); } |
| 889 |
SCALAR_FUN_ATTR int32_t futrts_ctzz32(int32_t x) { return x == 0 ? 32 : __builtin_ctz((uint32_t)x); } |
| 890 |
SCALAR_FUN_ATTR int32_t futrts_ctzz64(int64_t x) { return x == 0 ? 64 : __builtin_ctzll((uint64_t)x); } |
| 891 |
#endif |
| 892 |
|
| 893 |
SCALAR_FUN_ATTR float fdiv32(float x, float y) { return x / y; } |
| 894 |
SCALAR_FUN_ATTR float fadd32(float x, float y) { return x + y; } |
| 895 |
SCALAR_FUN_ATTR float fsub32(float x, float y) { return x - y; } |
| 896 |
SCALAR_FUN_ATTR float fmul32(float x, float y) { return x * y; } |
| 897 |
SCALAR_FUN_ATTR bool cmplt32(float x, float y) { return x < y; } |
| 898 |
SCALAR_FUN_ATTR bool cmple32(float x, float y) { return x <= y; } |
| 899 |
SCALAR_FUN_ATTR float sitofp_i8_f32(int8_t x) { return (float) x; } |
| 900 |
|
| 901 |
SCALAR_FUN_ATTR float sitofp_i16_f32(int16_t x) { return (float) x; } |
| 902 |
SCALAR_FUN_ATTR float sitofp_i32_f32(int32_t x) { return (float) x; } |
| 903 |
SCALAR_FUN_ATTR float sitofp_i64_f32(int64_t x) { return (float) x; } |
| 904 |
SCALAR_FUN_ATTR float uitofp_i8_f32(uint8_t x) { return (float) x; } |
| 905 |
SCALAR_FUN_ATTR float uitofp_i16_f32(uint16_t x) { return (float) x; } |
| 906 |
SCALAR_FUN_ATTR float uitofp_i32_f32(uint32_t x) { return (float) x; } |
| 907 |
SCALAR_FUN_ATTR float uitofp_i64_f32(uint64_t x) { return (float) x; } |
| 908 |
|
| 909 |
#ifdef __OPENCL_VERSION__ |
| 910 |
SCALAR_FUN_ATTR float fabs32(float x) { return fabs(x); } |
| 911 |
SCALAR_FUN_ATTR float fmax32(float x, float y) { return fmax(x, y); } |
| 912 |
SCALAR_FUN_ATTR float fmin32(float x, float y) { return fmin(x, y); } |
| 913 |
SCALAR_FUN_ATTR float fpow32(float x, float y) { return pow(x, y); } |
| 914 |
|
| 915 |
#elif defined(ISPC) |
| 916 |
|
| 917 |
SCALAR_FUN_ATTR float fabs32(float x) { return abs(x); } |
| 918 |
SCALAR_FUN_ATTR float fmax32(float x, float y) { return isnan(x) ? y : isnan(y) ? x : max(x, y); } |
| 919 |
SCALAR_FUN_ATTR float fmin32(float x, float y) { return isnan(x) ? y : isnan(y) ? x : min(x, y); } |
| 920 |
SCALAR_FUN_ATTR float fpow32(float a, float b) { |
| 921 |
float ret; |
| 922 |
foreach_active (i) { |
| 923 |
uniform float r = pow(extract(a, i), extract(b, i)); |
| 924 |
ret = insert(ret, i, r); |
| 925 |
} |
| 926 |
return ret; |
| 927 |
} |
| 928 |
|
| 929 |
#else // Not OpenCL, but CUDA or plain C. |
| 930 |
|
| 931 |
SCALAR_FUN_ATTR float fabs32(float x) { return fabsf(x); } |
| 932 |
SCALAR_FUN_ATTR float fmax32(float x, float y) { return fmaxf(x, y); } |
| 933 |
SCALAR_FUN_ATTR float fmin32(float x, float y) { return fminf(x, y); } |
| 934 |
SCALAR_FUN_ATTR float fpow32(float x, float y) { return powf(x, y); } |
| 935 |
#endif |
| 936 |
|
| 937 |
SCALAR_FUN_ATTR bool futrts_isnan32(float x) { return isnan(x); } |
| 938 |
|
| 939 |
#if defined(ISPC) |
| 940 |
|
| 941 |
SCALAR_FUN_ATTR bool futrts_isinf32(float x) { return !isnan(x) && isnan(x - x); } |
| 942 |
|
| 943 |
SCALAR_FUN_ATTR bool futrts_isfinite32(float x) { return !isnan(x) && !futrts_isinf32(x); } |
| 944 |
|
| 945 |
#else |
| 946 |
|
| 947 |
SCALAR_FUN_ATTR bool futrts_isinf32(float x) { return isinf(x); } |
| 948 |
|
| 949 |
#endif |
| 950 |
|
| 951 |
SCALAR_FUN_ATTR int8_t fptosi_f32_i8(float x) { |
| 952 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 953 |
return 0; |
| 954 |
} else { |
| 955 |
return (int8_t) x; |
| 956 |
} |
| 957 |
} |
| 958 |
|
| 959 |
SCALAR_FUN_ATTR int16_t fptosi_f32_i16(float x) { |
| 960 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 961 |
return 0; |
| 962 |
} else { |
| 963 |
return (int16_t) x; |
| 964 |
} |
| 965 |
} |
| 966 |
|
| 967 |
SCALAR_FUN_ATTR int32_t fptosi_f32_i32(float x) { |
| 968 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 969 |
return 0; |
| 970 |
} else { |
| 971 |
return (int32_t) x; |
| 972 |
} |
| 973 |
} |
| 974 |
|
| 975 |
SCALAR_FUN_ATTR int64_t fptosi_f32_i64(float x) { |
| 976 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 977 |
return 0; |
| 978 |
} else { |
| 979 |
return (int64_t) x; |
| 980 |
}; |
| 981 |
} |
| 982 |
|
| 983 |
SCALAR_FUN_ATTR uint8_t fptoui_f32_i8(float x) { |
| 984 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 985 |
return 0; |
| 986 |
} else { |
| 987 |
return (uint8_t) (int8_t) x; |
| 988 |
} |
| 989 |
} |
| 990 |
|
| 991 |
SCALAR_FUN_ATTR uint16_t fptoui_f32_i16(float x) { |
| 992 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 993 |
return 0; |
| 994 |
} else { |
| 995 |
return (uint16_t) (int16_t) x; |
| 996 |
} |
| 997 |
} |
| 998 |
|
| 999 |
SCALAR_FUN_ATTR uint32_t fptoui_f32_i32(float x) { |
| 1000 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 1001 |
return 0; |
| 1002 |
} else { |
| 1003 |
return (uint32_t) (int32_t) x; |
| 1004 |
} |
| 1005 |
} |
| 1006 |
|
| 1007 |
SCALAR_FUN_ATTR uint64_t fptoui_f32_i64(float x) { |
| 1008 |
if (futrts_isnan32(x) || futrts_isinf32(x)) { |
| 1009 |
return 0; |
| 1010 |
} else { |
| 1011 |
return (uint64_t) (int64_t) x; |
| 1012 |
} |
| 1013 |
} |
| 1014 |
|
| 1015 |
SCALAR_FUN_ATTR bool ftob_f32_bool(float x) { return x != 0; } |
| 1016 |
SCALAR_FUN_ATTR float btof_bool_f32(bool x) { return x ? 1 : 0; } |
| 1017 |
|
| 1018 |
#ifdef __OPENCL_VERSION__ |
| 1019 |
SCALAR_FUN_ATTR float futrts_log32(float x) { return log(x); } |
| 1020 |
SCALAR_FUN_ATTR float futrts_log2_32(float x) { return log2(x); } |
| 1021 |
SCALAR_FUN_ATTR float futrts_log10_32(float x) { return log10(x); } |
| 1022 |
SCALAR_FUN_ATTR float futrts_log1p_32(float x) { return log1p(x); } |
| 1023 |
SCALAR_FUN_ATTR float futrts_sqrt32(float x) { return sqrt(x); } |
| 1024 |
SCALAR_FUN_ATTR float futrts_rsqrt32(float x) { return rsqrt(x); } |
| 1025 |
SCALAR_FUN_ATTR float futrts_cbrt32(float x) { return cbrt(x); } |
| 1026 |
SCALAR_FUN_ATTR float futrts_exp32(float x) { return exp(x); } |
| 1027 |
SCALAR_FUN_ATTR float futrts_cos32(float x) { return cos(x); } |
| 1028 |
SCALAR_FUN_ATTR float futrts_cospi32(float x) { return cospi(x); } |
| 1029 |
SCALAR_FUN_ATTR float futrts_sin32(float x) { return sin(x); } |
| 1030 |
SCALAR_FUN_ATTR float futrts_sinpi32(float x) { return sinpi(x); } |
| 1031 |
SCALAR_FUN_ATTR float futrts_tan32(float x) { return tan(x); } |
| 1032 |
SCALAR_FUN_ATTR float futrts_tanpi32(float x) { return tanpi(x); } |
| 1033 |
SCALAR_FUN_ATTR float futrts_acos32(float x) { return acos(x); } |
| 1034 |
SCALAR_FUN_ATTR float futrts_acospi32(float x) { return acospi(x); } |
| 1035 |
SCALAR_FUN_ATTR float futrts_asin32(float x) { return asin(x); } |
| 1036 |
SCALAR_FUN_ATTR float futrts_asinpi32(float x) { return asinpi(x); } |
| 1037 |
SCALAR_FUN_ATTR float futrts_atan32(float x) { return atan(x); } |
| 1038 |
SCALAR_FUN_ATTR float futrts_atanpi32(float x) { return atanpi(x); } |
| 1039 |
SCALAR_FUN_ATTR float futrts_cosh32(float x) { return cosh(x); } |
| 1040 |
SCALAR_FUN_ATTR float futrts_sinh32(float x) { return sinh(x); } |
| 1041 |
SCALAR_FUN_ATTR float futrts_tanh32(float x) { return tanh(x); } |
| 1042 |
SCALAR_FUN_ATTR float futrts_acosh32(float x) { return acosh(x); } |
| 1043 |
SCALAR_FUN_ATTR float futrts_asinh32(float x) { return asinh(x); } |
| 1044 |
SCALAR_FUN_ATTR float futrts_atanh32(float x) { return atanh(x); } |
| 1045 |
SCALAR_FUN_ATTR float futrts_atan2_32(float x, float y) { return atan2(x, y); } |
| 1046 |
SCALAR_FUN_ATTR float futrts_atan2pi_32(float x, float y) { return atan2pi(x, y); } |
| 1047 |
SCALAR_FUN_ATTR float futrts_hypot32(float x, float y) { return hypot(x, y); } |
| 1048 |
SCALAR_FUN_ATTR float futrts_gamma32(float x) { return tgamma(x); } |
| 1049 |
SCALAR_FUN_ATTR float futrts_lgamma32(float x) { return lgamma(x); } |
| 1050 |
SCALAR_FUN_ATTR float futrts_erf32(float x) { return erf(x); } |
| 1051 |
SCALAR_FUN_ATTR float futrts_erfc32(float x) { return erfc(x); } |
| 1052 |
SCALAR_FUN_ATTR float fmod32(float x, float y) { return fmod(x, y); } |
| 1053 |
SCALAR_FUN_ATTR float futrts_round32(float x) { return rint(x); } |
| 1054 |
SCALAR_FUN_ATTR float futrts_floor32(float x) { return floor(x); } |
| 1055 |
SCALAR_FUN_ATTR float futrts_ceil32(float x) { return ceil(x); } |
| 1056 |
SCALAR_FUN_ATTR float futrts_nextafter32(float x, float y) { return nextafter(x, y); } |
| 1057 |
SCALAR_FUN_ATTR float futrts_lerp32(float v0, float v1, float t) { return mix(v0, v1, t); } |
| 1058 |
SCALAR_FUN_ATTR float futrts_ldexp32(float x, int32_t y) { return ldexp(x, y); } |
| 1059 |
SCALAR_FUN_ATTR float futrts_copysign32(float x, float y) { return copysign(x, y); } |
| 1060 |
SCALAR_FUN_ATTR float futrts_mad32(float a, float b, float c) { return mad(a, b, c); } |
| 1061 |
SCALAR_FUN_ATTR float futrts_fma32(float a, float b, float c) { return fma(a, b, c); } |
| 1062 |
|
| 1063 |
#elif defined(ISPC) |
| 1064 |
|
| 1065 |
SCALAR_FUN_ATTR float futrts_log32(float x) { return futrts_isfinite32(x) || (futrts_isinf32(x) && x < 0)? log(x) : x; } |
| 1066 |
SCALAR_FUN_ATTR float futrts_log2_32(float x) { return futrts_log32(x) / log(2.0f); } |
| 1067 |
SCALAR_FUN_ATTR float futrts_log10_32(float x) { return futrts_log32(x) / log(10.0f); } |
| 1068 |
|
| 1069 |
SCALAR_FUN_ATTR float futrts_log1p_32(float x) { |
| 1070 |
if(x == -1.0f || (futrts_isinf32(x) && x > 0.0f)) return x / 0.0f; |
| 1071 |
float y = 1.0f + x; |
| 1072 |
float z = y - 1.0f; |
| 1073 |
return log(y) - (z-x)/y; |
| 1074 |
} |
| 1075 |
|
| 1076 |
SCALAR_FUN_ATTR float futrts_sqrt32(float x) { return sqrt(x); } |
| 1077 |
SCALAR_FUN_ATTR float futrts_rsqrt32(float x) { return 1/sqrt(x); } |
| 1078 |
|
| 1079 |
extern "C" unmasked uniform float cbrtf(uniform float); |
| 1080 |
SCALAR_FUN_ATTR float futrts_cbrt32(float x) { |
| 1081 |
float res; |
| 1082 |
foreach_active (i) { |
| 1083 |
uniform float r = cbrtf(extract(x, i)); |
| 1084 |
res = insert(res, i, r); |
| 1085 |
} |
| 1086 |
return res; |
| 1087 |
} |
| 1088 |
|
| 1089 |
SCALAR_FUN_ATTR float futrts_exp32(float x) { return exp(x); } |
| 1090 |
SCALAR_FUN_ATTR float futrts_cos32(float x) { return cos(x); } |
| 1091 |
SCALAR_FUN_ATTR float futrts_cospi32(float x) { return cos((float)M_PI*x); } |
| 1092 |
SCALAR_FUN_ATTR float futrts_sin32(float x) { return sin(x); } |
| 1093 |
SCALAR_FUN_ATTR float futrts_sinpi32(float x) { return sin(M_PI*x); } |
| 1094 |
SCALAR_FUN_ATTR float futrts_tan32(float x) { return tan(x); } |
| 1095 |
SCALAR_FUN_ATTR float futrts_tanpi32(float x) { return tan((float)M_PI*x); } |
| 1096 |
SCALAR_FUN_ATTR float futrts_acos32(float x) { return acos(x); } |
| 1097 |
SCALAR_FUN_ATTR float futrts_acospi32(float x) { return acos(x)/(float)M_PI; } |
| 1098 |
SCALAR_FUN_ATTR float futrts_asin32(float x) { return asin(x); } |
| 1099 |
SCALAR_FUN_ATTR float futrts_asinpi32(float x) { return asin(x)/(float)M_PI; } |
| 1100 |
SCALAR_FUN_ATTR float futrts_atan32(float x) { return atan(x); } |
| 1101 |
SCALAR_FUN_ATTR float futrts_atanpi32(float x) { return atan(x)/(float)M_PI; } |
| 1102 |
SCALAR_FUN_ATTR float futrts_cosh32(float x) { return (exp(x)+exp(-x)) / 2.0f; } |
| 1103 |
SCALAR_FUN_ATTR float futrts_sinh32(float x) { return (exp(x)-exp(-x)) / 2.0f; } |
| 1104 |
SCALAR_FUN_ATTR float futrts_tanh32(float x) { return futrts_sinh32(x)/futrts_cosh32(x); } |
| 1105 |
|
| 1106 |
SCALAR_FUN_ATTR float futrts_acosh32(float x) { |
| 1107 |
float f = x+sqrt(x*x-1); |
| 1108 |
if (futrts_isfinite32(f)) return log(f); |
| 1109 |
return f; |
| 1110 |
} |
| 1111 |
|
| 1112 |
SCALAR_FUN_ATTR float futrts_asinh32(float x) { |
| 1113 |
float f = x+sqrt(x*x+1); |
| 1114 |
if (futrts_isfinite32(f)) return log(f); |
| 1115 |
return f; |
| 1116 |
} |
| 1117 |
|
| 1118 |
SCALAR_FUN_ATTR float futrts_atanh32(float x) { |
| 1119 |
float f = (1+x)/(1-x); |
| 1120 |
if (futrts_isfinite32(f)) return log(f)/2.0f; |
| 1121 |
return f; |
| 1122 |
} |
| 1123 |
|
| 1124 |
SCALAR_FUN_ATTR float futrts_atan2_32(float x, float y) |
| 1125 |
{ return (x == 0.0f && y == 0.0f) ? 0.0f : atan2(x, y); } |
| 1126 |
SCALAR_FUN_ATTR float futrts_atan2pi_32(float x, float y) |
| 1127 |
{ return (x == 0.0f && y == 0.0f) ? 0.0f : atan2(x, y) / (float)M_PI; } |
| 1128 |
|
| 1129 |
SCALAR_FUN_ATTR float futrts_hypot32(float x, float y) { |
| 1130 |
if (futrts_isfinite32(x) && futrts_isfinite32(y)) { |
| 1131 |
x = abs(x); |
| 1132 |
y = abs(y); |
| 1133 |
float a; |
| 1134 |
float b; |
| 1135 |
if (x >= y){ |
| 1136 |
a = x; |
| 1137 |
b = y; |
| 1138 |
} else { |
| 1139 |
a = y; |
| 1140 |
b = x; |
| 1141 |
} |
| 1142 |
if(b == 0){ |
| 1143 |
return a; |
| 1144 |
} |
| 1145 |
|
| 1146 |
int e; |
| 1147 |
float an; |
| 1148 |
float bn; |
| 1149 |
an = frexp (a, &e); |
| 1150 |
bn = ldexp (b, - e); |
| 1151 |
float cn; |
| 1152 |
cn = sqrt (an * an + bn * bn); |
| 1153 |
return ldexp (cn, e); |
| 1154 |
} else { |
| 1155 |
if (futrts_isinf32(x) || futrts_isinf32(y)) return INFINITY; |
| 1156 |
else return x + y; |
| 1157 |
} |
| 1158 |
|
| 1159 |
} |
| 1160 |
|
| 1161 |
extern "C" unmasked uniform float tgammaf(uniform float x); |
| 1162 |
SCALAR_FUN_ATTR float futrts_gamma32(float x) { |
| 1163 |
float res; |
| 1164 |
foreach_active (i) { |
| 1165 |
uniform float r = tgammaf(extract(x, i)); |
| 1166 |
res = insert(res, i, r); |
| 1167 |
} |
| 1168 |
return res; |
| 1169 |
} |
| 1170 |
|
| 1171 |
extern "C" unmasked uniform float lgammaf(uniform float x); |
| 1172 |
SCALAR_FUN_ATTR float futrts_lgamma32(float x) { |
| 1173 |
float res; |
| 1174 |
foreach_active (i) { |
| 1175 |
uniform float r = lgammaf(extract(x, i)); |
| 1176 |
res = insert(res, i, r); |
| 1177 |
} |
| 1178 |
return res; |
| 1179 |
} |
| 1180 |
|
| 1181 |
extern "C" unmasked uniform float erff(uniform float x); |
| 1182 |
SCALAR_FUN_ATTR float futrts_erf32(float x) { |
| 1183 |
float res; |
| 1184 |
foreach_active (i) { |
| 1185 |
uniform float r = erff(extract(x, i)); |
| 1186 |
res = insert(res, i, r); |
| 1187 |
} |
| 1188 |
return res; |
| 1189 |
} |
| 1190 |
|
| 1191 |
extern "C" unmasked uniform float erfcf(uniform float x); |
| 1192 |
SCALAR_FUN_ATTR float futrts_erfc32(float x) { |
| 1193 |
float res; |
| 1194 |
foreach_active (i) { |
| 1195 |
uniform float r = erfcf(extract(x, i)); |
| 1196 |
res = insert(res, i, r); |
| 1197 |
} |
| 1198 |
return res; |
| 1199 |
} |
| 1200 |
|
| 1201 |
SCALAR_FUN_ATTR float fmod32(float x, float y) { return x - y * trunc(x/y); } |
| 1202 |
SCALAR_FUN_ATTR float futrts_round32(float x) { return round(x); } |
| 1203 |
SCALAR_FUN_ATTR float futrts_floor32(float x) { return floor(x); } |
| 1204 |
SCALAR_FUN_ATTR float futrts_ceil32(float x) { return ceil(x); } |
| 1205 |
|
| 1206 |
extern "C" unmasked uniform float nextafterf(uniform float x, uniform float y); |
| 1207 |
SCALAR_FUN_ATTR float futrts_nextafter32(float x, float y) { |
| 1208 |
float res; |
| 1209 |
foreach_active (i) { |
| 1210 |
uniform float r = nextafterf(extract(x, i), extract(y, i)); |
| 1211 |
res = insert(res, i, r); |
| 1212 |
} |
| 1213 |
return res; |
| 1214 |
} |
| 1215 |
|
| 1216 |
SCALAR_FUN_ATTR float futrts_lerp32(float v0, float v1, float t) { |
| 1217 |
return v0 + (v1 - v0) * t; |
| 1218 |
} |
| 1219 |
|
| 1220 |
SCALAR_FUN_ATTR float futrts_ldexp32(float x, int32_t y) { |
| 1221 |
return x * pow((uniform float)2.0, (float)y); |
| 1222 |
} |
| 1223 |
|
| 1224 |
SCALAR_FUN_ATTR float futrts_copysign32(float x, float y) { |
| 1225 |
int32_t xb = fptobits_f32_i32(x); |
| 1226 |
int32_t yb = fptobits_f32_i32(y); |
| 1227 |
return bitstofp_i32_f32((xb & ~(1<<31)) | (yb & (1<<31))); |
| 1228 |
} |
| 1229 |
|
| 1230 |
SCALAR_FUN_ATTR float futrts_mad32(float a, float b, float c) { |
| 1231 |
return a * b + c; |
| 1232 |
} |
| 1233 |
|
| 1234 |
SCALAR_FUN_ATTR float futrts_fma32(float a, float b, float c) { |
| 1235 |
return a * b + c; |
| 1236 |
} |
| 1237 |
|
| 1238 |
#else // Not OpenCL or ISPC, but CUDA or plain C. |
| 1239 |
|
| 1240 |
SCALAR_FUN_ATTR float futrts_log32(float x) { return logf(x); } |
| 1241 |
SCALAR_FUN_ATTR float futrts_log2_32(float x) { return log2f(x); } |
| 1242 |
SCALAR_FUN_ATTR float futrts_log10_32(float x) { return log10f(x); } |
| 1243 |
SCALAR_FUN_ATTR float futrts_log1p_32(float x) { return log1pf(x); } |
| 1244 |
SCALAR_FUN_ATTR float futrts_sqrt32(float x) { return sqrtf(x); } |
| 1245 |
SCALAR_FUN_ATTR float futrts_rsqrt32(float x) { return 1/sqrtf(x); } |
| 1246 |
SCALAR_FUN_ATTR float futrts_cbrt32(float x) { return cbrtf(x); } |
| 1247 |
SCALAR_FUN_ATTR float futrts_exp32(float x) { return expf(x); } |
| 1248 |
SCALAR_FUN_ATTR float futrts_cos32(float x) { return cosf(x); } |
| 1249 |
|
| 1250 |
SCALAR_FUN_ATTR float futrts_cospi32(float x) { |
| 1251 |
#if defined(__CUDA_ARCH__) |
| 1252 |
return cospif(x); |
| 1253 |
#else |
| 1254 |
return cosf(((float)M_PI)*x); |
| 1255 |
#endif |
| 1256 |
} |
| 1257 |
SCALAR_FUN_ATTR float futrts_sin32(float x) { return sinf(x); } |
| 1258 |
|
| 1259 |
SCALAR_FUN_ATTR float futrts_sinpi32(float x) { |
| 1260 |
#if defined(__CUDA_ARCH__) |
| 1261 |
return sinpif(x); |
| 1262 |
#else |
| 1263 |
return sinf((float)M_PI*x); |
| 1264 |
#endif |
| 1265 |
} |
| 1266 |
|
| 1267 |
SCALAR_FUN_ATTR float futrts_tan32(float x) { return tanf(x); } |
| 1268 |
SCALAR_FUN_ATTR float futrts_tanpi32(float x) { return tanf((float)M_PI*x); } |
| 1269 |
SCALAR_FUN_ATTR float futrts_acos32(float x) { return acosf(x); } |
| 1270 |
SCALAR_FUN_ATTR float futrts_acospi32(float x) { return acosf(x)/(float)M_PI; } |
| 1271 |
SCALAR_FUN_ATTR float futrts_asin32(float x) { return asinf(x); } |
| 1272 |
SCALAR_FUN_ATTR float futrts_asinpi32(float x) { return asinf(x)/(float)M_PI; } |
| 1273 |
SCALAR_FUN_ATTR float futrts_atan32(float x) { return atanf(x); } |
| 1274 |
SCALAR_FUN_ATTR float futrts_atanpi32(float x) { return atanf(x)/(float)M_PI; } |
| 1275 |
SCALAR_FUN_ATTR float futrts_cosh32(float x) { return coshf(x); } |
| 1276 |
SCALAR_FUN_ATTR float futrts_sinh32(float x) { return sinhf(x); } |
| 1277 |
SCALAR_FUN_ATTR float futrts_tanh32(float x) { return tanhf(x); } |
| 1278 |
SCALAR_FUN_ATTR float futrts_acosh32(float x) { return acoshf(x); } |
| 1279 |
SCALAR_FUN_ATTR float futrts_asinh32(float x) { return asinhf(x); } |
| 1280 |
SCALAR_FUN_ATTR float futrts_atanh32(float x) { return atanhf(x); } |
| 1281 |
SCALAR_FUN_ATTR float futrts_atan2_32(float x, float y) { return atan2f(x, y); } |
| 1282 |
SCALAR_FUN_ATTR float futrts_atan2pi_32(float x, float y) { return atan2f(x, y) / (float)M_PI; } |
| 1283 |
SCALAR_FUN_ATTR float futrts_hypot32(float x, float y) { return hypotf(x, y); } |
| 1284 |
SCALAR_FUN_ATTR float futrts_gamma32(float x) { return tgammaf(x); } |
| 1285 |
SCALAR_FUN_ATTR float futrts_lgamma32(float x) { return lgammaf(x); } |
| 1286 |
SCALAR_FUN_ATTR float futrts_erf32(float x) { return erff(x); } |
| 1287 |
SCALAR_FUN_ATTR float futrts_erfc32(float x) { return erfcf(x); } |
| 1288 |
SCALAR_FUN_ATTR float fmod32(float x, float y) { return fmodf(x, y); } |
| 1289 |
SCALAR_FUN_ATTR float futrts_round32(float x) { return rintf(x); } |
| 1290 |
SCALAR_FUN_ATTR float futrts_floor32(float x) { return floorf(x); } |
| 1291 |
SCALAR_FUN_ATTR float futrts_ceil32(float x) { return ceilf(x); } |
| 1292 |
SCALAR_FUN_ATTR float futrts_nextafter32(float x, float y) { return nextafterf(x, y); } |
| 1293 |
SCALAR_FUN_ATTR float futrts_lerp32(float v0, float v1, float t) { return v0 + (v1 - v0) * t; } |
| 1294 |
SCALAR_FUN_ATTR float futrts_ldexp32(float x, int32_t y) { return ldexpf(x, y); } |
| 1295 |
SCALAR_FUN_ATTR float futrts_copysign32(float x, float y) { return copysignf(x, y); } |
| 1296 |
SCALAR_FUN_ATTR float futrts_mad32(float a, float b, float c) { return a * b + c; } |
| 1297 |
SCALAR_FUN_ATTR float futrts_fma32(float a, float b, float c) { return fmaf(a, b, c); } |
| 1298 |
|
| 1299 |
#endif |
| 1300 |
|
| 1301 |
#if defined(ISPC) |
| 1302 |
|
| 1303 |
SCALAR_FUN_ATTR int32_t fptobits_f32_i32(float x) { return intbits(x); } |
| 1304 |
SCALAR_FUN_ATTR float bitstofp_i32_f32(int32_t x) { return floatbits(x); } |
| 1305 |
SCALAR_FUN_ATTR uniform int32_t fptobits_f32_i32(uniform float x) { return intbits(x); } |
| 1306 |
SCALAR_FUN_ATTR uniform float bitstofp_i32_f32(uniform int32_t x) { return floatbits(x); } |
| 1307 |
|
| 1308 |
#else |
| 1309 |
|
| 1310 |
SCALAR_FUN_ATTR int32_t fptobits_f32_i32(float x) { |
| 1311 |
union { |
| 1312 |
float f; |
| 1313 |
int32_t t; |
| 1314 |
} p; |
| 1315 |
|
| 1316 |
p.f = x; |
| 1317 |
return p.t; |
| 1318 |
} |
| 1319 |
|
| 1320 |
SCALAR_FUN_ATTR float bitstofp_i32_f32(int32_t x) { |
| 1321 |
union { |
| 1322 |
int32_t f; |
| 1323 |
float t; |
| 1324 |
} p; |
| 1325 |
|
| 1326 |
p.f = x; |
| 1327 |
return p.t; |
| 1328 |
} |
| 1329 |
#endif |
| 1330 |
|
| 1331 |
SCALAR_FUN_ATTR float fsignum32(float x) { |
| 1332 |
return futrts_isnan32(x) ? x : (x > 0 ? 1 : 0) - (x < 0 ? 1 : 0); |
| 1333 |
} |
| 1334 |
|
| 1335 |
#ifdef FUTHARK_F64_ENABLED |
| 1336 |
|
| 1337 |
SCALAR_FUN_ATTR double bitstofp_i64_f64(int64_t x); |
| 1338 |
SCALAR_FUN_ATTR int64_t fptobits_f64_i64(double x); |
| 1339 |
|
| 1340 |
#if defined(ISPC) |
| 1341 |
|
| 1342 |
SCALAR_FUN_ATTR bool futrts_isinf64(float x) { return !isnan(x) && isnan(x - x); } |
| 1343 |
SCALAR_FUN_ATTR bool futrts_isfinite64(float x) { return !isnan(x) && !futrts_isinf64(x); } |
| 1344 |
SCALAR_FUN_ATTR double fdiv64(double x, double y) { return x / y; } |
| 1345 |
SCALAR_FUN_ATTR double fadd64(double x, double y) { return x + y; } |
| 1346 |
SCALAR_FUN_ATTR double fsub64(double x, double y) { return x - y; } |
| 1347 |
SCALAR_FUN_ATTR double fmul64(double x, double y) { return x * y; } |
| 1348 |
SCALAR_FUN_ATTR bool cmplt64(double x, double y) { return x < y; } |
| 1349 |
SCALAR_FUN_ATTR bool cmple64(double x, double y) { return x <= y; } |
| 1350 |
SCALAR_FUN_ATTR double sitofp_i8_f64(int8_t x) { return (double) x; } |
| 1351 |
SCALAR_FUN_ATTR double sitofp_i16_f64(int16_t x) { return (double) x; } |
| 1352 |
SCALAR_FUN_ATTR double sitofp_i32_f64(int32_t x) { return (double) x; } |
| 1353 |
SCALAR_FUN_ATTR double sitofp_i64_f64(int64_t x) { return (double) x; } |
| 1354 |
SCALAR_FUN_ATTR double uitofp_i8_f64(uint8_t x) { return (double) x; } |
| 1355 |
SCALAR_FUN_ATTR double uitofp_i16_f64(uint16_t x) { return (double) x; } |
| 1356 |
SCALAR_FUN_ATTR double uitofp_i32_f64(uint32_t x) { return (double) x; } |
| 1357 |
SCALAR_FUN_ATTR double uitofp_i64_f64(uint64_t x) { return (double) x; } |
| 1358 |
SCALAR_FUN_ATTR double fabs64(double x) { return abs(x); } |
| 1359 |
SCALAR_FUN_ATTR double fmax64(double x, double y) { return isnan(x) ? y : isnan(y) ? x : max(x, y); } |
| 1360 |
SCALAR_FUN_ATTR double fmin64(double x, double y) { return isnan(x) ? y : isnan(y) ? x : min(x, y); } |
| 1361 |
|
| 1362 |
SCALAR_FUN_ATTR double fpow64(double a, double b) { |
| 1363 |
float ret; |
| 1364 |
foreach_active (i) { |
| 1365 |
uniform float r = pow(extract(a, i), extract(b, i)); |
| 1366 |
ret = insert(ret, i, r); |
| 1367 |
} |
| 1368 |
return ret; |
| 1369 |
} |
| 1370 |
SCALAR_FUN_ATTR double futrts_log64(double x) { return futrts_isfinite64(x) || (futrts_isinf64(x) && x < 0)? log(x) : x; } |
| 1371 |
SCALAR_FUN_ATTR double futrts_log2_64(double x) { return futrts_log64(x)/log(2.0d); } |
| 1372 |
SCALAR_FUN_ATTR double futrts_log10_64(double x) { return futrts_log64(x)/log(10.0d); } |
| 1373 |
|
| 1374 |
SCALAR_FUN_ATTR double futrts_log1p_64(double x) { |
| 1375 |
if(x == -1.0d || (futrts_isinf64(x) && x > 0.0d)) return x / 0.0d; |
| 1376 |
double y = 1.0d + x; |
| 1377 |
double z = y - 1.0d; |
| 1378 |
return log(y) - (z-x)/y; |
| 1379 |
} |
| 1380 |
|
| 1381 |
SCALAR_FUN_ATTR double futrts_sqrt64(double x) { return sqrt(x); } |
| 1382 |
SCALAR_FUN_ATTR double futrts_rsqrt64(double x) { return 1/sqrt(x); } |
| 1383 |
|
| 1384 |
SCALAR_FUN_ATTR double futrts_cbrt64(double x) { |
| 1385 |
double res; |
| 1386 |
foreach_active (i) { |
| 1387 |
uniform double r = cbrtf(extract(x, i)); |
| 1388 |
res = insert(res, i, r); |
| 1389 |
} |
| 1390 |
return res; |
| 1391 |
} |
| 1392 |
SCALAR_FUN_ATTR double futrts_exp64(double x) { return exp(x); } |
| 1393 |
SCALAR_FUN_ATTR double futrts_cos64(double x) { return cos(x); } |
| 1394 |
SCALAR_FUN_ATTR double futrts_cospi64(double x) { return cos(M_PI*x); } |
| 1395 |
SCALAR_FUN_ATTR double futrts_sin64(double x) { return sin(x); } |
| 1396 |
SCALAR_FUN_ATTR double futrts_sinpi64(double x) { return sin(M_PI*x); } |
| 1397 |
SCALAR_FUN_ATTR double futrts_tan64(double x) { return tan(x); } |
| 1398 |
SCALAR_FUN_ATTR double futrts_tanpi64(double x) { return tan(M_PI*x); } |
| 1399 |
SCALAR_FUN_ATTR double futrts_acos64(double x) { return acos(x); } |
| 1400 |
SCALAR_FUN_ATTR double futrts_acospi64(double x) { return acos(x)/M_PI; } |
| 1401 |
SCALAR_FUN_ATTR double futrts_asin64(double x) { return asin(x); } |
| 1402 |
SCALAR_FUN_ATTR double futrts_asinpi64(double x) { return asin(x)/M_PI; } |
| 1403 |
SCALAR_FUN_ATTR double futrts_atan64(double x) { return atan(x); } |
| 1404 |
SCALAR_FUN_ATTR double futrts_atanpi64(double x) { return atan(x)/M_PI; } |
| 1405 |
SCALAR_FUN_ATTR double futrts_cosh64(double x) { return (exp(x)+exp(-x)) / 2.0d; } |
| 1406 |
SCALAR_FUN_ATTR double futrts_sinh64(double x) { return (exp(x)-exp(-x)) / 2.0d; } |
| 1407 |
SCALAR_FUN_ATTR double futrts_tanh64(double x) { return futrts_sinh64(x)/futrts_cosh64(x); } |
| 1408 |
|
| 1409 |
SCALAR_FUN_ATTR double futrts_acosh64(double x) { |
| 1410 |
double f = x+sqrt(x*x-1.0d); |
| 1411 |
if(futrts_isfinite64(f)) return log(f); |
| 1412 |
return f; |
| 1413 |
} |
| 1414 |
|
| 1415 |
SCALAR_FUN_ATTR double futrts_asinh64(double x) { |
| 1416 |
double f = x+sqrt(x*x+1.0d); |
| 1417 |
if(futrts_isfinite64(f)) return log(f); |
| 1418 |
return f; |
| 1419 |
} |
| 1420 |
|
| 1421 |
SCALAR_FUN_ATTR double futrts_atanh64(double x) { |
| 1422 |
double f = (1.0d+x)/(1.0d-x); |
| 1423 |
if(futrts_isfinite64(f)) return log(f)/2.0d; |
| 1424 |
return f; |
| 1425 |
} |
| 1426 |
SCALAR_FUN_ATTR double futrts_atan2_64(double x, double y) { return atan2(x, y); } |
| 1427 |
|
| 1428 |
SCALAR_FUN_ATTR double futrts_atan2pi_64(double x, double y) { return atan2(x, y) / M_PI; } |
| 1429 |
|
| 1430 |
extern "C" unmasked uniform double hypot(uniform double x, uniform double y); |
| 1431 |
SCALAR_FUN_ATTR double futrts_hypot64(double x, double y) { |
| 1432 |
double res; |
| 1433 |
foreach_active (i) { |
| 1434 |
uniform double r = hypot(extract(x, i), extract(y, i)); |
| 1435 |
res = insert(res, i, r); |
| 1436 |
} |
| 1437 |
return res; |
| 1438 |
} |
| 1439 |
|
| 1440 |
extern "C" unmasked uniform double tgamma(uniform double x); |
| 1441 |
SCALAR_FUN_ATTR double futrts_gamma64(double x) { |
| 1442 |
double res; |
| 1443 |
foreach_active (i) { |
| 1444 |
uniform double r = tgamma(extract(x, i)); |
| 1445 |
res = insert(res, i, r); |
| 1446 |
} |
| 1447 |
return res; |
| 1448 |
} |
| 1449 |
|
| 1450 |
extern "C" unmasked uniform double lgamma(uniform double x); |
| 1451 |
SCALAR_FUN_ATTR double futrts_lgamma64(double x) { |
| 1452 |
double res; |
| 1453 |
foreach_active (i) { |
| 1454 |
uniform double r = lgamma(extract(x, i)); |
| 1455 |
res = insert(res, i, r); |
| 1456 |
} |
| 1457 |
return res; |
| 1458 |
} |
| 1459 |
|
| 1460 |
extern "C" unmasked uniform double erf(uniform double x); |
| 1461 |
SCALAR_FUN_ATTR double futrts_erf64(double x) { |
| 1462 |
double res; |
| 1463 |
foreach_active (i) { |
| 1464 |
uniform double r = erf(extract(x, i)); |
| 1465 |
res = insert(res, i, r); |
| 1466 |
} |
| 1467 |
return res; |
| 1468 |
} |
| 1469 |
|
| 1470 |
extern "C" unmasked uniform double erfc(uniform double x); |
| 1471 |
SCALAR_FUN_ATTR double futrts_erfc64(double x) { |
| 1472 |
double res; |
| 1473 |
foreach_active (i) { |
| 1474 |
uniform double r = erfc(extract(x, i)); |
| 1475 |
res = insert(res, i, r); |
| 1476 |
} |
| 1477 |
return res; |
| 1478 |
} |
| 1479 |
|
| 1480 |
SCALAR_FUN_ATTR double futrts_fma64(double a, double b, double c) { return a * b + c; } |
| 1481 |
SCALAR_FUN_ATTR double futrts_round64(double x) { return round(x); } |
| 1482 |
SCALAR_FUN_ATTR double futrts_ceil64(double x) { return ceil(x); } |
| 1483 |
|
| 1484 |
extern "C" unmasked uniform double nextafter(uniform float x, uniform double y); |
| 1485 |
SCALAR_FUN_ATTR float futrts_nextafter64(double x, double y) { |
| 1486 |
double res; |
| 1487 |
foreach_active (i) { |
| 1488 |
uniform double r = nextafter(extract(x, i), extract(y, i)); |
| 1489 |
res = insert(res, i, r); |
| 1490 |
} |
| 1491 |
return res; |
| 1492 |
} |
| 1493 |
|
| 1494 |
SCALAR_FUN_ATTR double futrts_floor64(double x) { return floor(x); } |
| 1495 |
SCALAR_FUN_ATTR bool futrts_isnan64(double x) { return isnan(x); } |
| 1496 |
|
| 1497 |
SCALAR_FUN_ATTR int8_t fptosi_f64_i8(double x) { |
| 1498 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1499 |
return 0; |
| 1500 |
} else { |
| 1501 |
return (int8_t) x; |
| 1502 |
} |
| 1503 |
} |
| 1504 |
|
| 1505 |
SCALAR_FUN_ATTR int16_t fptosi_f64_i16(double x) { |
| 1506 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1507 |
return 0; |
| 1508 |
} else { |
| 1509 |
return (int16_t) x; |
| 1510 |
} |
| 1511 |
} |
| 1512 |
|
| 1513 |
SCALAR_FUN_ATTR int32_t fptosi_f64_i32(double x) { |
| 1514 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1515 |
return 0; |
| 1516 |
} else { |
| 1517 |
return (int32_t) x; |
| 1518 |
} |
| 1519 |
} |
| 1520 |
|
| 1521 |
SCALAR_FUN_ATTR int64_t fptosi_f64_i64(double x) { |
| 1522 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1523 |
return 0; |
| 1524 |
} else { |
| 1525 |
return (int64_t) x; |
| 1526 |
} |
| 1527 |
} |
| 1528 |
|
| 1529 |
SCALAR_FUN_ATTR uint8_t fptoui_f64_i8(double x) { |
| 1530 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1531 |
return 0; |
| 1532 |
} else { |
| 1533 |
return (uint8_t) (int8_t) x; |
| 1534 |
} |
| 1535 |
} |
| 1536 |
|
| 1537 |
SCALAR_FUN_ATTR uint16_t fptoui_f64_i16(double x) { |
| 1538 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1539 |
return 0; |
| 1540 |
} else { |
| 1541 |
return (uint16_t) (int16_t) x; |
| 1542 |
} |
| 1543 |
} |
| 1544 |
|
| 1545 |
SCALAR_FUN_ATTR uint32_t fptoui_f64_i32(double x) { |
| 1546 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1547 |
return 0; |
| 1548 |
} else { |
| 1549 |
return (uint32_t) (int32_t) x; |
| 1550 |
} |
| 1551 |
} |
| 1552 |
|
| 1553 |
SCALAR_FUN_ATTR uint64_t fptoui_f64_i64(double x) { |
| 1554 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1555 |
return 0; |
| 1556 |
} else { |
| 1557 |
return (uint64_t) (int64_t) x; |
| 1558 |
} |
| 1559 |
} |
| 1560 |
|
| 1561 |
SCALAR_FUN_ATTR bool ftob_f64_bool(double x) { return x != 0.0; } |
| 1562 |
SCALAR_FUN_ATTR double btof_bool_f64(bool x) { return x ? 1.0 : 0.0; } |
| 1563 |
|
| 1564 |
SCALAR_FUN_ATTR int64_t fptobits_f64_i64(double x) { |
| 1565 |
int64_t res; |
| 1566 |
foreach_active (i) { |
| 1567 |
uniform double tmp = extract(x, i); |
| 1568 |
uniform int64_t r = *((uniform int64_t* uniform)&tmp); |
| 1569 |
res = insert(res, i, r); |
| 1570 |
} |
| 1571 |
return res; |
| 1572 |
} |
| 1573 |
|
| 1574 |
SCALAR_FUN_ATTR double bitstofp_i64_f64(int64_t x) { |
| 1575 |
double res; |
| 1576 |
foreach_active (i) { |
| 1577 |
uniform int64_t tmp = extract(x, i); |
| 1578 |
uniform double r = *((uniform double* uniform)&tmp); |
| 1579 |
res = insert(res, i, r); |
| 1580 |
} |
| 1581 |
return res; |
| 1582 |
} |
| 1583 |
|
| 1584 |
SCALAR_FUN_ATTR uniform int64_t fptobits_f64_i64(uniform double x) { |
| 1585 |
return intbits(x); |
| 1586 |
} |
| 1587 |
|
| 1588 |
SCALAR_FUN_ATTR uniform double bitstofp_i64_f64(uniform int64_t x) { |
| 1589 |
return doublebits(x); |
| 1590 |
} |
| 1591 |
|
| 1592 |
SCALAR_FUN_ATTR double fmod64(double x, double y) { |
| 1593 |
return x - y * trunc(x/y); |
| 1594 |
} |
| 1595 |
|
| 1596 |
SCALAR_FUN_ATTR double fsignum64(double x) { |
| 1597 |
return futrts_isnan64(x) ? x : (x > 0 ? 1.0d : 0.0d) - (x < 0 ? 1.0d : 0.0d); |
| 1598 |
} |
| 1599 |
|
| 1600 |
SCALAR_FUN_ATTR double futrts_lerp64(double v0, double v1, double t) { |
| 1601 |
return v0 + (v1 - v0) * t; |
| 1602 |
} |
| 1603 |
|
| 1604 |
SCALAR_FUN_ATTR double futrts_ldexp64(double x, int32_t y) { |
| 1605 |
return x * pow((uniform double)2.0, (double)y); |
| 1606 |
} |
| 1607 |
|
| 1608 |
SCALAR_FUN_ATTR double futrts_copysign64(double x, double y) { |
| 1609 |
int64_t xb = fptobits_f64_i64(x); |
| 1610 |
int64_t yb = fptobits_f64_i64(y); |
| 1611 |
return bitstofp_i64_f64((xb & ~(((int64_t)1)<<63)) | (yb & (((int64_t)1)<<63))); |
| 1612 |
} |
| 1613 |
|
| 1614 |
SCALAR_FUN_ATTR double futrts_mad64(double a, double b, double c) { return a * b + c; } |
| 1615 |
SCALAR_FUN_ATTR float fpconv_f32_f32(float x) { return (float) x; } |
| 1616 |
SCALAR_FUN_ATTR double fpconv_f32_f64(float x) { return (double) x; } |
| 1617 |
SCALAR_FUN_ATTR float fpconv_f64_f32(double x) { return (float) x; } |
| 1618 |
SCALAR_FUN_ATTR double fpconv_f64_f64(double x) { return (double) x; } |
| 1619 |
|
| 1620 |
#else |
| 1621 |
|
| 1622 |
SCALAR_FUN_ATTR double fdiv64(double x, double y) { return x / y; } |
| 1623 |
SCALAR_FUN_ATTR double fadd64(double x, double y) { return x + y; } |
| 1624 |
SCALAR_FUN_ATTR double fsub64(double x, double y) { return x - y; } |
| 1625 |
SCALAR_FUN_ATTR double fmul64(double x, double y) { return x * y; } |
| 1626 |
SCALAR_FUN_ATTR bool cmplt64(double x, double y) { return x < y; } |
| 1627 |
SCALAR_FUN_ATTR bool cmple64(double x, double y) { return x <= y; } |
| 1628 |
SCALAR_FUN_ATTR double sitofp_i8_f64(int8_t x) { return (double) x; } |
| 1629 |
SCALAR_FUN_ATTR double sitofp_i16_f64(int16_t x) { return (double) x; } |
| 1630 |
SCALAR_FUN_ATTR double sitofp_i32_f64(int32_t x) { return (double) x; } |
| 1631 |
SCALAR_FUN_ATTR double sitofp_i64_f64(int64_t x) { return (double) x; } |
| 1632 |
SCALAR_FUN_ATTR double uitofp_i8_f64(uint8_t x) { return (double) x; } |
| 1633 |
SCALAR_FUN_ATTR double uitofp_i16_f64(uint16_t x) { return (double) x; } |
| 1634 |
SCALAR_FUN_ATTR double uitofp_i32_f64(uint32_t x) { return (double) x; } |
| 1635 |
SCALAR_FUN_ATTR double uitofp_i64_f64(uint64_t x) { return (double) x; } |
| 1636 |
SCALAR_FUN_ATTR double fabs64(double x) { return fabs(x); } |
| 1637 |
SCALAR_FUN_ATTR double fmax64(double x, double y) { return fmax(x, y); } |
| 1638 |
SCALAR_FUN_ATTR double fmin64(double x, double y) { return fmin(x, y); } |
| 1639 |
SCALAR_FUN_ATTR double fpow64(double x, double y) { return pow(x, y); } |
| 1640 |
SCALAR_FUN_ATTR double futrts_log64(double x) { return log(x); } |
| 1641 |
SCALAR_FUN_ATTR double futrts_log2_64(double x) { return log2(x); } |
| 1642 |
SCALAR_FUN_ATTR double futrts_log10_64(double x) { return log10(x); } |
| 1643 |
SCALAR_FUN_ATTR double futrts_log1p_64(double x) { return log1p(x); } |
| 1644 |
SCALAR_FUN_ATTR double futrts_sqrt64(double x) { return sqrt(x); } |
| 1645 |
SCALAR_FUN_ATTR double futrts_rsqrt64(double x) { return 1/sqrt(x); } |
| 1646 |
SCALAR_FUN_ATTR double futrts_cbrt64(double x) { return cbrt(x); } |
| 1647 |
SCALAR_FUN_ATTR double futrts_exp64(double x) { return exp(x); } |
| 1648 |
SCALAR_FUN_ATTR double futrts_cos64(double x) { return cos(x); } |
| 1649 |
|
| 1650 |
SCALAR_FUN_ATTR double futrts_cospi64(double x) { |
| 1651 |
#ifdef __OPENCL_VERSION__ |
| 1652 |
return cospi(x); |
| 1653 |
#elif defined(__CUDA_ARCH__) |
| 1654 |
return cospi(x); |
| 1655 |
#else |
| 1656 |
return cos(M_PI*x); |
| 1657 |
#endif |
| 1658 |
} |
| 1659 |
|
| 1660 |
SCALAR_FUN_ATTR double futrts_sin64(double x) { |
| 1661 |
return sin(x); |
| 1662 |
} |
| 1663 |
|
| 1664 |
SCALAR_FUN_ATTR double futrts_sinpi64(double x) { |
| 1665 |
#ifdef __OPENCL_VERSION__ |
| 1666 |
return sinpi(x); |
| 1667 |
#elif defined(__CUDA_ARCH__) |
| 1668 |
return sinpi(x); |
| 1669 |
#else |
| 1670 |
return sin(M_PI*x); |
| 1671 |
#endif |
| 1672 |
} |
| 1673 |
|
| 1674 |
SCALAR_FUN_ATTR double futrts_tan64(double x) { |
| 1675 |
return tan(x); |
| 1676 |
} |
| 1677 |
|
| 1678 |
SCALAR_FUN_ATTR double futrts_tanpi64(double x) { |
| 1679 |
#ifdef __OPENCL_VERSION__ |
| 1680 |
return tanpi(x); |
| 1681 |
#else |
| 1682 |
return tan(M_PI*x); |
| 1683 |
#endif |
| 1684 |
} |
| 1685 |
|
| 1686 |
SCALAR_FUN_ATTR double futrts_acos64(double x) { |
| 1687 |
return acos(x); |
| 1688 |
} |
| 1689 |
|
| 1690 |
SCALAR_FUN_ATTR double futrts_acospi64(double x) { |
| 1691 |
#ifdef __OPENCL_VERSION__ |
| 1692 |
return acospi(x); |
| 1693 |
#else |
| 1694 |
return acos(x) / M_PI; |
| 1695 |
#endif |
| 1696 |
} |
| 1697 |
|
| 1698 |
SCALAR_FUN_ATTR double futrts_asin64(double x) { |
| 1699 |
return asin(x); |
| 1700 |
} |
| 1701 |
|
| 1702 |
SCALAR_FUN_ATTR double futrts_asinpi64(double x) { |
| 1703 |
#ifdef __OPENCL_VERSION__ |
| 1704 |
return asinpi(x); |
| 1705 |
#else |
| 1706 |
return asin(x) / M_PI; |
| 1707 |
#endif |
| 1708 |
} |
| 1709 |
|
| 1710 |
SCALAR_FUN_ATTR double futrts_atan64(double x) { |
| 1711 |
return atan(x); |
| 1712 |
} |
| 1713 |
|
| 1714 |
SCALAR_FUN_ATTR double futrts_atanpi64(double x) { |
| 1715 |
#ifdef __OPENCL_VERSION__ |
| 1716 |
return atanpi(x); |
| 1717 |
#else |
| 1718 |
return atan(x) / M_PI; |
| 1719 |
#endif |
| 1720 |
} |
| 1721 |
|
| 1722 |
SCALAR_FUN_ATTR double futrts_cosh64(double x) { return cosh(x); } |
| 1723 |
SCALAR_FUN_ATTR double futrts_sinh64(double x) { return sinh(x); } |
| 1724 |
SCALAR_FUN_ATTR double futrts_tanh64(double x) { return tanh(x); } |
| 1725 |
SCALAR_FUN_ATTR double futrts_acosh64(double x) { return acosh(x); } |
| 1726 |
SCALAR_FUN_ATTR double futrts_asinh64(double x) { return asinh(x); } |
| 1727 |
SCALAR_FUN_ATTR double futrts_atanh64(double x) { return atanh(x); } |
| 1728 |
SCALAR_FUN_ATTR double futrts_atan2_64(double x, double y) { return atan2(x, y); } |
| 1729 |
|
| 1730 |
SCALAR_FUN_ATTR double futrts_atan2pi_64(double x, double y) { |
| 1731 |
#ifdef __OPENCL_VERSION__ |
| 1732 |
return atan2pi(x, y); |
| 1733 |
#else |
| 1734 |
return atan2(x, y) / M_PI; |
| 1735 |
#endif |
| 1736 |
} |
| 1737 |
|
| 1738 |
SCALAR_FUN_ATTR double futrts_hypot64(double x, double y) { return hypot(x, y); } |
| 1739 |
SCALAR_FUN_ATTR double futrts_gamma64(double x) { return tgamma(x); } |
| 1740 |
SCALAR_FUN_ATTR double futrts_lgamma64(double x) { return lgamma(x); } |
| 1741 |
SCALAR_FUN_ATTR double futrts_erf64(double x) { return erf(x); } |
| 1742 |
SCALAR_FUN_ATTR double futrts_erfc64(double x) { return erfc(x); } |
| 1743 |
SCALAR_FUN_ATTR double futrts_fma64(double a, double b, double c) { return fma(a, b, c); } |
| 1744 |
SCALAR_FUN_ATTR double futrts_round64(double x) { return rint(x); } |
| 1745 |
SCALAR_FUN_ATTR double futrts_ceil64(double x) { return ceil(x); } |
| 1746 |
SCALAR_FUN_ATTR float futrts_nextafter64(float x, float y) { return nextafter(x, y); } |
| 1747 |
SCALAR_FUN_ATTR double futrts_floor64(double x) { return floor(x); } |
| 1748 |
SCALAR_FUN_ATTR bool futrts_isnan64(double x) { return isnan(x); } |
| 1749 |
SCALAR_FUN_ATTR bool futrts_isinf64(double x) { return isinf(x); } |
| 1750 |
|
| 1751 |
SCALAR_FUN_ATTR int8_t fptosi_f64_i8(double x) { |
| 1752 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1753 |
return 0; |
| 1754 |
} else { |
| 1755 |
return (int8_t) x; |
| 1756 |
} |
| 1757 |
} |
| 1758 |
|
| 1759 |
SCALAR_FUN_ATTR int16_t fptosi_f64_i16(double x) { |
| 1760 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1761 |
return 0; |
| 1762 |
} else { |
| 1763 |
return (int16_t) x; |
| 1764 |
} |
| 1765 |
} |
| 1766 |
|
| 1767 |
SCALAR_FUN_ATTR int32_t fptosi_f64_i32(double x) { |
| 1768 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1769 |
return 0; |
| 1770 |
} else { |
| 1771 |
return (int32_t) x; |
| 1772 |
} |
| 1773 |
} |
| 1774 |
|
| 1775 |
SCALAR_FUN_ATTR int64_t fptosi_f64_i64(double x) { |
| 1776 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1777 |
return 0; |
| 1778 |
} else { |
| 1779 |
return (int64_t) x; |
| 1780 |
} |
| 1781 |
} |
| 1782 |
|
| 1783 |
SCALAR_FUN_ATTR uint8_t fptoui_f64_i8(double x) { |
| 1784 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1785 |
return 0; |
| 1786 |
} else { |
| 1787 |
return (uint8_t) (int8_t) x; |
| 1788 |
} |
| 1789 |
} |
| 1790 |
|
| 1791 |
SCALAR_FUN_ATTR uint16_t fptoui_f64_i16(double x) { |
| 1792 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1793 |
return 0; |
| 1794 |
} else { |
| 1795 |
return (uint16_t) (int16_t) x; |
| 1796 |
} |
| 1797 |
} |
| 1798 |
|
| 1799 |
SCALAR_FUN_ATTR uint32_t fptoui_f64_i32(double x) { |
| 1800 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1801 |
return 0; |
| 1802 |
} else { |
| 1803 |
return (uint32_t) (int32_t) x; |
| 1804 |
} |
| 1805 |
} |
| 1806 |
|
| 1807 |
SCALAR_FUN_ATTR uint64_t fptoui_f64_i64(double x) { |
| 1808 |
if (futrts_isnan64(x) || futrts_isinf64(x)) { |
| 1809 |
return 0; |
| 1810 |
} else { |
| 1811 |
return (uint64_t) (int64_t) x; |
| 1812 |
} |
| 1813 |
} |
| 1814 |
|
| 1815 |
SCALAR_FUN_ATTR bool ftob_f64_bool(double x) { return x != 0; } |
| 1816 |
SCALAR_FUN_ATTR double btof_bool_f64(bool x) { return x ? 1 : 0; } |
| 1817 |
|
| 1818 |
SCALAR_FUN_ATTR int64_t fptobits_f64_i64(double x) { |
| 1819 |
union { |
| 1820 |
double f; |
| 1821 |
int64_t t; |
| 1822 |
} p; |
| 1823 |
|
| 1824 |
p.f = x; |
| 1825 |
return p.t; |
| 1826 |
} |
| 1827 |
|
| 1828 |
SCALAR_FUN_ATTR double bitstofp_i64_f64(int64_t x) { |
| 1829 |
union { |
| 1830 |
int64_t f; |
| 1831 |
double t; |
| 1832 |
} p; |
| 1833 |
|
| 1834 |
p.f = x; |
| 1835 |
return p.t; |
| 1836 |
} |
| 1837 |
|
| 1838 |
SCALAR_FUN_ATTR double fmod64(double x, double y) { |
| 1839 |
return fmod(x, y); |
| 1840 |
} |
| 1841 |
|
| 1842 |
SCALAR_FUN_ATTR double fsignum64(double x) { |
| 1843 |
return futrts_isnan64(x) ? x : (x > 0) - (x < 0); |
| 1844 |
} |
| 1845 |
|
| 1846 |
SCALAR_FUN_ATTR double futrts_lerp64(double v0, double v1, double t) { |
| 1847 |
#ifdef __OPENCL_VERSION__ |
| 1848 |
return mix(v0, v1, t); |
| 1849 |
#else |
| 1850 |
return v0 + (v1 - v0) * t; |
| 1851 |
#endif |
| 1852 |
} |
| 1853 |
|
| 1854 |
SCALAR_FUN_ATTR double futrts_ldexp64(double x, int32_t y) { |
| 1855 |
return ldexp(x, y); |
| 1856 |
} |
| 1857 |
|
| 1858 |
SCALAR_FUN_ATTR float futrts_copysign64(double x, double y) { |
| 1859 |
return copysign(x, y); |
| 1860 |
} |
| 1861 |
|
| 1862 |
SCALAR_FUN_ATTR double futrts_mad64(double a, double b, double c) { |
| 1863 |
#ifdef __OPENCL_VERSION__ |
| 1864 |
return mad(a, b, c); |
| 1865 |
#else |
| 1866 |
return a * b + c; |
| 1867 |
#endif |
| 1868 |
} |
| 1869 |
|
| 1870 |
SCALAR_FUN_ATTR float fpconv_f32_f32(float x) { return (float) x; } |
| 1871 |
SCALAR_FUN_ATTR double fpconv_f32_f64(float x) { return (double) x; } |
| 1872 |
SCALAR_FUN_ATTR float fpconv_f64_f32(double x) { return (float) x; } |
| 1873 |
SCALAR_FUN_ATTR double fpconv_f64_f64(double x) { return (double) x; } |
| 1874 |
|
| 1875 |
#endif |
| 1876 |
|
| 1877 |
#endif |
| 1878 |
|
| 1879 |
#define futrts_cond_f16(x,y,z) ((x) ? (y) : (z)) |
| 1880 |
#define futrts_cond_f32(x,y,z) ((x) ? (y) : (z)) |
| 1881 |
#define futrts_cond_f64(x,y,z) ((x) ? (y) : (z)) |
| 1882 |
|
| 1883 |
#define futrts_cond_i8(x,y,z) ((x) ? (y) : (z)) |
| 1884 |
#define futrts_cond_i16(x,y,z) ((x) ? (y) : (z)) |
| 1885 |
#define futrts_cond_i32(x,y,z) ((x) ? (y) : (z)) |
| 1886 |
#define futrts_cond_i64(x,y,z) ((x) ? (y) : (z)) |
| 1887 |
|
| 1888 |
#define futrts_cond_bool(x,y,z) ((x) ? (y) : (z)) |
| 1889 |
#define futrts_cond_unit(x,y,z) ((x) ? (y) : (z)) |
| 1890 |
|
| 1891 |
// End of scalar.h. |