| 1 | /*************************************************************************** |
| 2 | * Copyright (c) Johan Mabille, Sylvain Corlay, Wolf Vollprecht and * |
| 3 | * Martin Renou * |
| 4 | * Copyright (c) QuantStack * |
| 5 | * Copyright (c) Serge Guelton * |
| 6 | * * |
| 7 | * Distributed under the terms of the BSD 3-Clause License. * |
| 8 | * * |
| 9 | * The full license is in the file LICENSE, distributed with this software. * |
| 10 | ****************************************************************************/ |
| 11 | |
| 12 | #ifndef XSIMD_SCALAR_HPP |
| 13 | #define XSIMD_SCALAR_HPP |
| 14 | |
| 15 | #include <cassert> |
| 16 | #include <cmath> |
| 17 | #include <complex> |
| 18 | #include <cstdint> |
| 19 | #include <cstring> |
| 20 | #include <limits> |
| 21 | #include <type_traits> |
| 22 | |
| 23 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 24 | #include "xtl/xcomplex.hpp" |
| 25 | #endif |
| 26 | |
| 27 | namespace xsimd |
| 28 | { |
| 29 | template <class T, class A> |
| 30 | class batch; |
| 31 | template <class T, class A> |
| 32 | class batch_bool; |
| 33 | |
| 34 | using std::abs; |
| 35 | |
| 36 | using std::acos; |
| 37 | using std::acosh; |
| 38 | using std::arg; |
| 39 | using std::asin; |
| 40 | using std::asinh; |
| 41 | using std::atan; |
| 42 | using std::atan2; |
| 43 | using std::atanh; |
| 44 | using std::cbrt; |
| 45 | using std::ceil; |
| 46 | using std::conj; |
| 47 | using std::copysign; |
| 48 | using std::cos; |
| 49 | using std::cosh; |
| 50 | using std::erf; |
| 51 | using std::erfc; |
| 52 | using std::exp; |
| 53 | using std::exp2; |
| 54 | using std::expm1; |
| 55 | using std::fabs; |
| 56 | using std::fdim; |
| 57 | using std::floor; |
| 58 | using std::fmax; |
| 59 | using std::fmin; |
| 60 | using std::fmod; |
| 61 | using std::hypot; |
| 62 | using std::ldexp; |
| 63 | using std::lgamma; |
| 64 | using std::log; |
| 65 | using std::log10; |
| 66 | using std::log1p; |
| 67 | using std::log2; |
| 68 | using std::modf; |
| 69 | using std::nearbyint; |
| 70 | using std::nextafter; |
| 71 | using std::norm; |
| 72 | using std::polar; |
| 73 | using std::proj; |
| 74 | using std::remainder; |
| 75 | using std::rint; |
| 76 | using std::round; |
| 77 | using std::sin; |
| 78 | using std::sinh; |
| 79 | using std::sqrt; |
| 80 | using std::tan; |
| 81 | using std::tanh; |
| 82 | using std::tgamma; |
| 83 | using std::trunc; |
| 84 | |
| 85 | #ifndef _WIN32 |
| 86 | using std::isfinite; |
| 87 | using std::isinf; |
| 88 | using std::isnan; |
| 89 | #else |
| 90 | |
| 91 | // Windows defines catch all templates |
| 92 | template <class T> |
| 93 | inline typename std::enable_if<std::is_floating_point<T>::value, bool>::type |
| 94 | isfinite(T var) noexcept |
| 95 | { |
| 96 | return std::isfinite(var); |
| 97 | } |
| 98 | |
| 99 | template <class T> |
| 100 | inline typename std::enable_if<std::is_integral<T>::value, bool>::type |
| 101 | isfinite(T var) noexcept |
| 102 | { |
| 103 | return isfinite(double(var)); |
| 104 | } |
| 105 | |
| 106 | template <class T> |
| 107 | inline typename std::enable_if<std::is_floating_point<T>::value, bool>::type |
| 108 | isinf(T var) noexcept |
| 109 | { |
| 110 | return std::isinf(var); |
| 111 | } |
| 112 | |
| 113 | template <class T> |
| 114 | inline typename std::enable_if<std::is_integral<T>::value, bool>::type |
| 115 | isinf(T var) noexcept |
| 116 | { |
| 117 | return isinf(double(var)); |
| 118 | } |
| 119 | |
| 120 | template <class T> |
| 121 | inline typename std::enable_if<std::is_floating_point<T>::value, bool>::type |
| 122 | isnan(T var) noexcept |
| 123 | { |
| 124 | return std::isnan(var); |
| 125 | } |
| 126 | |
| 127 | template <class T> |
| 128 | inline typename std::enable_if<std::is_integral<T>::value, bool>::type |
| 129 | isnan(T var) noexcept |
| 130 | { |
| 131 | return isnan(double(var)); |
| 132 | } |
| 133 | #endif |
| 134 | |
| 135 | template <class T, class Tp> |
| 136 | inline auto add(T const& x, Tp const& y) noexcept -> decltype(x + y) |
| 137 | { |
| 138 | return x + y; |
| 139 | } |
| 140 | |
| 141 | template <class T> |
| 142 | inline typename std::enable_if<std::is_integral<T>::value, T>::type |
| 143 | bitwise_and(T x, T y) noexcept |
| 144 | { |
| 145 | return x & y; |
| 146 | } |
| 147 | |
| 148 | inline float bitwise_and(float x, float y) noexcept |
| 149 | { |
| 150 | uint32_t ix, iy; |
| 151 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(float)); |
| 152 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(float)); |
| 153 | uint32_t ir = bitwise_and(x: ix, y: iy); |
| 154 | float r; |
| 155 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(float)); |
| 156 | return r; |
| 157 | } |
| 158 | |
| 159 | inline double bitwise_and(double x, double y) noexcept |
| 160 | { |
| 161 | uint64_t ix, iy; |
| 162 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(double)); |
| 163 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(double)); |
| 164 | uint64_t ir = bitwise_and(x: ix, y: iy); |
| 165 | double r; |
| 166 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(double)); |
| 167 | return r; |
| 168 | } |
| 169 | |
| 170 | template <class T> |
| 171 | inline typename std::enable_if<std::is_integral<T>::value, T>::type |
| 172 | bitwise_andnot(T x, T y) noexcept |
| 173 | { |
| 174 | return x & ~y; |
| 175 | } |
| 176 | |
| 177 | inline float bitwise_andnot(float x, float y) noexcept |
| 178 | { |
| 179 | uint32_t ix, iy; |
| 180 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(float)); |
| 181 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(float)); |
| 182 | uint32_t ir = bitwise_andnot(x: ix, y: iy); |
| 183 | float r; |
| 184 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(float)); |
| 185 | return r; |
| 186 | } |
| 187 | |
| 188 | inline double bitwise_andnot(double x, double y) noexcept |
| 189 | { |
| 190 | uint64_t ix, iy; |
| 191 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(double)); |
| 192 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(double)); |
| 193 | uint64_t ir = bitwise_andnot(x: ix, y: iy); |
| 194 | double r; |
| 195 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(double)); |
| 196 | return r; |
| 197 | } |
| 198 | |
| 199 | template <class T> |
| 200 | inline typename std::enable_if<std::is_integral<T>::value, T>::type |
| 201 | bitwise_not(T x) noexcept |
| 202 | { |
| 203 | return ~x; |
| 204 | } |
| 205 | |
| 206 | inline float bitwise_not(float x) noexcept |
| 207 | { |
| 208 | uint32_t ix; |
| 209 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(float)); |
| 210 | uint32_t ir = bitwise_not(x: ix); |
| 211 | float r; |
| 212 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(float)); |
| 213 | return r; |
| 214 | } |
| 215 | |
| 216 | inline double bitwise_not(double x) noexcept |
| 217 | { |
| 218 | uint64_t ix; |
| 219 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(double)); |
| 220 | uint64_t ir = bitwise_not(x: ix); |
| 221 | double r; |
| 222 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(double)); |
| 223 | return r; |
| 224 | } |
| 225 | |
| 226 | template <class T> |
| 227 | inline typename std::enable_if<std::is_integral<T>::value, T>::type |
| 228 | bitwise_or(T x, T y) noexcept |
| 229 | { |
| 230 | return x | y; |
| 231 | } |
| 232 | |
| 233 | inline float bitwise_or(float x, float y) noexcept |
| 234 | { |
| 235 | uint32_t ix, iy; |
| 236 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(float)); |
| 237 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(float)); |
| 238 | uint32_t ir = bitwise_or(x: ix, y: iy); |
| 239 | float r; |
| 240 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(float)); |
| 241 | return r; |
| 242 | } |
| 243 | |
| 244 | inline double bitwise_or(double x, double y) noexcept |
| 245 | { |
| 246 | uint64_t ix, iy; |
| 247 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(double)); |
| 248 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(double)); |
| 249 | uint64_t ir = bitwise_or(x: ix, y: iy); |
| 250 | double r; |
| 251 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(double)); |
| 252 | return r; |
| 253 | } |
| 254 | |
| 255 | template <class T> |
| 256 | inline typename std::enable_if<std::is_integral<T>::value, T>::type |
| 257 | bitwise_xor(T x, T y) noexcept |
| 258 | { |
| 259 | return x ^ y; |
| 260 | } |
| 261 | |
| 262 | inline float bitwise_xor(float x, float y) noexcept |
| 263 | { |
| 264 | uint32_t ix, iy; |
| 265 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(float)); |
| 266 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(float)); |
| 267 | uint32_t ir = bitwise_xor(x: ix, y: iy); |
| 268 | float r; |
| 269 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(float)); |
| 270 | return r; |
| 271 | } |
| 272 | |
| 273 | inline double bitwise_xor(double x, double y) noexcept |
| 274 | { |
| 275 | uint64_t ix, iy; |
| 276 | std::memcpy(dest: (void*)&ix, src: (void*)&x, n: sizeof(double)); |
| 277 | std::memcpy(dest: (void*)&iy, src: (void*)&y, n: sizeof(double)); |
| 278 | uint64_t ir = bitwise_xor(x: ix, y: iy); |
| 279 | double r; |
| 280 | std::memcpy(dest: (void*)&r, src: (void*)&ir, n: sizeof(double)); |
| 281 | return r; |
| 282 | } |
| 283 | |
| 284 | template <class T, class Tp> |
| 285 | inline auto div(T const& x, Tp const& y) noexcept -> decltype(x / y) |
| 286 | { |
| 287 | return x / y; |
| 288 | } |
| 289 | |
| 290 | template <class T, class Tp> |
| 291 | inline auto mod(T const& x, Tp const& y) noexcept -> decltype(x % y) |
| 292 | { |
| 293 | return x % y; |
| 294 | } |
| 295 | |
| 296 | template <class T, class Tp> |
| 297 | inline auto mul(T const& x, Tp const& y) noexcept -> decltype(x * y) |
| 298 | { |
| 299 | return x * y; |
| 300 | } |
| 301 | |
| 302 | template <class T> |
| 303 | inline auto neg(T const& x) noexcept -> decltype(-x) |
| 304 | { |
| 305 | return -x; |
| 306 | } |
| 307 | |
| 308 | template <class T> |
| 309 | inline auto pos(T const& x) noexcept -> decltype(+x) |
| 310 | { |
| 311 | return +x; |
| 312 | } |
| 313 | |
| 314 | inline float reciprocal(float const& x) noexcept |
| 315 | { |
| 316 | return 1.f / x; |
| 317 | } |
| 318 | |
| 319 | inline double reciprocal(double const& x) noexcept |
| 320 | { |
| 321 | return 1. / x; |
| 322 | } |
| 323 | |
| 324 | #ifdef XSIMD_ENABLE_NUMPY_COMPLEX |
| 325 | template <class T> |
| 326 | inline bool isnan(std::complex<T> var) noexcept |
| 327 | { |
| 328 | return std::isnan(std::real(var)) || std::isnan(std::imag(var)); |
| 329 | } |
| 330 | |
| 331 | template <class T> |
| 332 | inline bool isinf(std::complex<T> var) noexcept |
| 333 | { |
| 334 | return std::isinf(std::real(var)) || std::isinf(std::imag(var)); |
| 335 | } |
| 336 | #endif |
| 337 | |
| 338 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 339 | using xtl::abs; |
| 340 | using xtl::acos; |
| 341 | using xtl::acosh; |
| 342 | using xtl::asin; |
| 343 | using xtl::asinh; |
| 344 | using xtl::atan; |
| 345 | using xtl::atanh; |
| 346 | using xtl::cos; |
| 347 | using xtl::cosh; |
| 348 | using xtl::exp; |
| 349 | using xtl::log; |
| 350 | using xtl::log10; |
| 351 | using xtl::norm; |
| 352 | using xtl::pow; |
| 353 | using xtl::proj; |
| 354 | using xtl::sin; |
| 355 | using xtl::sinh; |
| 356 | using xtl::sqrt; |
| 357 | using xtl::tan; |
| 358 | using xtl::tanh; |
| 359 | #endif |
| 360 | |
| 361 | template <typename T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 362 | inline T clip(const T& val, const T& low, const T& hi) noexcept |
| 363 | { |
| 364 | assert(low <= hi && "ordered clipping bounds" ); |
| 365 | return low > val ? low : (hi < val ? hi : val); |
| 366 | } |
| 367 | |
| 368 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 369 | inline bool is_flint(const T& x) noexcept |
| 370 | { |
| 371 | return std::isnan(x - x) ? false : (x - std::trunc(x)) == T(0); |
| 372 | } |
| 373 | |
| 374 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 375 | inline bool is_even(const T& x) noexcept |
| 376 | { |
| 377 | return is_flint(x * T(0.5)); |
| 378 | } |
| 379 | |
| 380 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 381 | inline bool is_odd(const T& x) noexcept |
| 382 | { |
| 383 | return is_even(x - 1.); |
| 384 | } |
| 385 | |
| 386 | inline int32_t nearbyint_as_int(float var) noexcept |
| 387 | { |
| 388 | return static_cast<int32_t>(std::nearbyint(x: var)); |
| 389 | } |
| 390 | |
| 391 | inline int64_t nearbyint_as_int(double var) noexcept |
| 392 | { |
| 393 | return static_cast<int64_t>(std::nearbyint(x: var)); |
| 394 | } |
| 395 | |
| 396 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 397 | inline bool eq(const T& x0, const T& x1) noexcept |
| 398 | { |
| 399 | return x0 == x1; |
| 400 | } |
| 401 | |
| 402 | template <class T> |
| 403 | inline bool eq(const std::complex<T>& x0, const std::complex<T>& x1) noexcept |
| 404 | { |
| 405 | return x0 == x1; |
| 406 | } |
| 407 | |
| 408 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 409 | inline bool ge(const T& x0, const T& x1) noexcept |
| 410 | { |
| 411 | return x0 >= x1; |
| 412 | } |
| 413 | |
| 414 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 415 | inline bool gt(const T& x0, const T& x1) noexcept |
| 416 | { |
| 417 | return x0 > x1; |
| 418 | } |
| 419 | |
| 420 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 421 | inline bool le(const T& x0, const T& x1) noexcept |
| 422 | { |
| 423 | return x0 <= x1; |
| 424 | } |
| 425 | |
| 426 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 427 | inline bool lt(const T& x0, const T& x1) noexcept |
| 428 | { |
| 429 | return x0 < x1; |
| 430 | } |
| 431 | |
| 432 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 433 | inline bool neq(const T& x0, const T& x1) noexcept |
| 434 | { |
| 435 | return x0 != x1; |
| 436 | } |
| 437 | |
| 438 | template <class T> |
| 439 | inline bool neq(const std::complex<T>& x0, const std::complex<T>& x1) noexcept |
| 440 | { |
| 441 | return !(x0 == x1); |
| 442 | } |
| 443 | |
| 444 | #if defined(_GNU_SOURCE) && !defined(__APPLE__) && !defined(__MINGW32__) && !defined(__ANDROID__) |
| 445 | inline float exp10(const float& x) noexcept |
| 446 | { |
| 447 | return ::exp10f(x: x); |
| 448 | } |
| 449 | inline double exp10(const double& x) noexcept |
| 450 | { |
| 451 | return ::exp10(x: x); |
| 452 | } |
| 453 | #endif |
| 454 | |
| 455 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 456 | inline T exp10(const T& x) noexcept |
| 457 | { |
| 458 | // FIXME: very inefficient |
| 459 | return std::pow(T(10), x); |
| 460 | } |
| 461 | |
| 462 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 463 | inline auto rsqrt(const T& x) noexcept -> decltype(std::sqrt(x)) |
| 464 | { |
| 465 | using float_type = decltype(std::sqrt(x)); |
| 466 | return static_cast<float_type>(1) / std::sqrt(x); |
| 467 | } |
| 468 | |
| 469 | namespace detail |
| 470 | { |
| 471 | template <class C> |
| 472 | inline C expm1_complex_scalar_impl(const C& val) noexcept |
| 473 | { |
| 474 | using T = typename C::value_type; |
| 475 | T isin = std::sin(val.imag()); |
| 476 | T rem1 = std::expm1(val.real()); |
| 477 | T re = rem1 + T(1.); |
| 478 | T si = std::sin(val.imag() * T(0.5)); |
| 479 | return std::complex<T>(rem1 - T(2.) * re * si * si, re * isin); |
| 480 | } |
| 481 | } |
| 482 | |
| 483 | template <class T> |
| 484 | inline std::complex<T> expm1(const std::complex<T>& val) noexcept |
| 485 | { |
| 486 | return detail::expm1_complex_scalar_impl(val); |
| 487 | } |
| 488 | |
| 489 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 490 | template <class T, bool i3ec> |
| 491 | inline xtl::xcomplex<T, T, i3ec> expm1(const xtl::xcomplex<T, T, i3ec>& val) noexcept |
| 492 | { |
| 493 | return detail::expm1_complex_scalar_impl(val); |
| 494 | } |
| 495 | #endif |
| 496 | |
| 497 | namespace detail |
| 498 | { |
| 499 | template <class C> |
| 500 | inline C log1p_complex_scalar_impl(const C& val) noexcept |
| 501 | { |
| 502 | using T = typename C::value_type; |
| 503 | C u = C(1.) + val; |
| 504 | return u == C(1.) ? val : (u.real() <= T(0.) ? log(u) : log(u) * val / (u - C(1.))); |
| 505 | } |
| 506 | } |
| 507 | |
| 508 | template <class T> |
| 509 | inline std::complex<T> log1p(const std::complex<T>& val) noexcept |
| 510 | { |
| 511 | return detail::log1p_complex_scalar_impl(val); |
| 512 | } |
| 513 | |
| 514 | template <class T> |
| 515 | inline std::complex<T> log2(const std::complex<T>& val) noexcept |
| 516 | { |
| 517 | return log(val) / std::log(T(2)); |
| 518 | } |
| 519 | |
| 520 | template <typename T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 521 | inline T sadd(const T& lhs, const T& rhs) noexcept |
| 522 | { |
| 523 | if (std::numeric_limits<T>::is_signed) |
| 524 | { |
| 525 | if ((lhs > 0) && (rhs > std::numeric_limits<T>::max() - lhs)) |
| 526 | { |
| 527 | return std::numeric_limits<T>::max(); |
| 528 | } |
| 529 | else if ((lhs < 0) && (rhs < std::numeric_limits<T>::lowest() - lhs)) |
| 530 | { |
| 531 | return std::numeric_limits<T>::lowest(); |
| 532 | } |
| 533 | else |
| 534 | { |
| 535 | return lhs + rhs; |
| 536 | } |
| 537 | } |
| 538 | else |
| 539 | { |
| 540 | if (rhs > std::numeric_limits<T>::max() - lhs) |
| 541 | { |
| 542 | return std::numeric_limits<T>::max(); |
| 543 | } |
| 544 | else |
| 545 | { |
| 546 | return lhs + rhs; |
| 547 | } |
| 548 | } |
| 549 | } |
| 550 | |
| 551 | template <typename T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 552 | inline T ssub(const T& lhs, const T& rhs) noexcept |
| 553 | { |
| 554 | if (std::numeric_limits<T>::is_signed) |
| 555 | { |
| 556 | return sadd(lhs, (T)-rhs); |
| 557 | } |
| 558 | else |
| 559 | { |
| 560 | if (lhs < rhs) |
| 561 | { |
| 562 | return std::numeric_limits<T>::lowest(); |
| 563 | } |
| 564 | else |
| 565 | { |
| 566 | return lhs - rhs; |
| 567 | } |
| 568 | } |
| 569 | } |
| 570 | |
| 571 | namespace detail |
| 572 | { |
| 573 | template <class T> |
| 574 | struct value_type_or_type_helper |
| 575 | { |
| 576 | using type = T; |
| 577 | }; |
| 578 | template <class T, class A> |
| 579 | struct value_type_or_type_helper<batch<T, A>> |
| 580 | { |
| 581 | using type = T; |
| 582 | }; |
| 583 | |
| 584 | template <class T> |
| 585 | using value_type_or_type = typename value_type_or_type_helper<T>::type; |
| 586 | |
| 587 | template <class T0, class T1> |
| 588 | inline typename std::enable_if<std::is_integral<T1>::value, T0>::type |
| 589 | ipow(const T0& x, const T1& n) noexcept |
| 590 | { |
| 591 | static_assert(std::is_integral<T1>::value, "second argument must be an integer" ); |
| 592 | T0 a = x; |
| 593 | T1 b = n; |
| 594 | bool const recip = b < 0; |
| 595 | T0 r(static_cast<value_type_or_type<T0>>(1)); |
| 596 | while (1) |
| 597 | { |
| 598 | if (b & 1) |
| 599 | { |
| 600 | r *= a; |
| 601 | } |
| 602 | b /= 2; |
| 603 | if (b == 0) |
| 604 | { |
| 605 | break; |
| 606 | } |
| 607 | a *= a; |
| 608 | } |
| 609 | return recip ? static_cast<T0>(1) / r : r; |
| 610 | } |
| 611 | } |
| 612 | |
| 613 | template <class T0, class T1> |
| 614 | inline typename std::enable_if<std::is_integral<T1>::value, T0>::type |
| 615 | pow(const T0& x, const T1& n) noexcept |
| 616 | { |
| 617 | return detail::ipow(x, n); |
| 618 | } |
| 619 | |
| 620 | template <class T0, class T1> |
| 621 | inline auto |
| 622 | pow(const T0& t0, const T1& t1) noexcept |
| 623 | -> typename std::enable_if<std::is_scalar<T0>::value && std::is_floating_point<T1>::value, decltype(std::pow(t0, t1))>::type |
| 624 | { |
| 625 | return std::pow(t0, t1); |
| 626 | } |
| 627 | |
| 628 | template <class T0, class T1> |
| 629 | inline typename std::enable_if<std::is_integral<T1>::value, std::complex<T0>>::type |
| 630 | pow(const std::complex<T0>& t0, const T1& t1) noexcept |
| 631 | { |
| 632 | return detail::ipow(t0, t1); |
| 633 | } |
| 634 | |
| 635 | template <class T0, class T1> |
| 636 | inline typename std::enable_if<!std::is_integral<T1>::value, std::complex<T0>>::type |
| 637 | pow(const std::complex<T0>& t0, const T1& t1) noexcept |
| 638 | { |
| 639 | return std::pow(t0, t1); |
| 640 | } |
| 641 | |
| 642 | template <class T0, class T1> |
| 643 | inline auto |
| 644 | pow(const T0& t0, const std::complex<T1>& t1) noexcept |
| 645 | -> typename std::enable_if<std::is_scalar<T0>::value, decltype(std::pow(t0, t1))>::type |
| 646 | { |
| 647 | return std::pow(t0, t1); |
| 648 | } |
| 649 | |
| 650 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 651 | inline bool bitofsign(T const& x) noexcept |
| 652 | { |
| 653 | return x < T(0); |
| 654 | } |
| 655 | |
| 656 | template <class T> |
| 657 | inline auto signbit(T const& v) noexcept -> decltype(bitofsign(v)) |
| 658 | { |
| 659 | return bitofsign(v); |
| 660 | } |
| 661 | |
| 662 | inline double sign(bool const& v) noexcept |
| 663 | { |
| 664 | return v; |
| 665 | } |
| 666 | |
| 667 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 668 | inline T sign(const T& v) noexcept |
| 669 | { |
| 670 | return v < T(0) ? T(-1.) : v == T(0) ? T(0.) |
| 671 | : T(1.); |
| 672 | } |
| 673 | |
| 674 | namespace detail |
| 675 | { |
| 676 | template <class C> |
| 677 | inline C sign_complex_scalar_impl(const C& v) noexcept |
| 678 | { |
| 679 | using value_type = typename C::value_type; |
| 680 | if (v.real()) |
| 681 | { |
| 682 | return C(sign(v.real()), value_type(0)); |
| 683 | } |
| 684 | else |
| 685 | { |
| 686 | return C(sign(v.imag()), value_type(0)); |
| 687 | } |
| 688 | } |
| 689 | } |
| 690 | |
| 691 | template <class T> |
| 692 | inline std::complex<T> sign(const std::complex<T>& v) noexcept |
| 693 | { |
| 694 | return detail::sign_complex_scalar_impl(v); |
| 695 | } |
| 696 | |
| 697 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 698 | template <class T, bool i3ec> |
| 699 | inline xtl::xcomplex<T, T, i3ec> sign(const xtl::xcomplex<T, T, i3ec>& v) noexcept |
| 700 | { |
| 701 | return detail::sign_complex_scalar_impl(v); |
| 702 | } |
| 703 | #endif |
| 704 | |
| 705 | inline double signnz(bool const&) noexcept |
| 706 | { |
| 707 | return 1; |
| 708 | } |
| 709 | |
| 710 | template <class T, class = typename std::enable_if<std::is_scalar<T>::value>::type> |
| 711 | inline T signnz(const T& v) noexcept |
| 712 | { |
| 713 | return v < T(0) ? T(-1.) : T(1.); |
| 714 | } |
| 715 | |
| 716 | template <class T, class Tp> |
| 717 | inline auto sub(T const& x, Tp const& y) noexcept -> decltype(x - y) |
| 718 | { |
| 719 | return x - y; |
| 720 | } |
| 721 | |
| 722 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 723 | template <class T, bool i3ec> |
| 724 | inline xtl::xcomplex<T, T, i3ec> log2(const xtl::xcomplex<T, T, i3ec>& val) noexcept |
| 725 | { |
| 726 | return log(val) / log(T(2)); |
| 727 | } |
| 728 | #endif |
| 729 | |
| 730 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 731 | template <class T, bool i3ec> |
| 732 | inline xtl::xcomplex<T, T, i3ec> log1p(const xtl::xcomplex<T, T, i3ec>& val) noexcept |
| 733 | { |
| 734 | return detail::log1p_complex_scalar_impl(val); |
| 735 | } |
| 736 | #endif |
| 737 | |
| 738 | template <class T0, class T1> |
| 739 | inline auto min(T0 const& self, T1 const& other) noexcept |
| 740 | -> typename std::enable_if<std::is_scalar<T0>::value && std::is_scalar<T1>::value, |
| 741 | typename std::decay<decltype(self > other ? other : self)>::type>::type |
| 742 | { |
| 743 | return self > other ? other : self; |
| 744 | } |
| 745 | |
| 746 | // numpy defines minimum operator on complex using lexical comparison |
| 747 | template <class T0, class T1> |
| 748 | inline std::complex<typename std::common_type<T0, T1>::type> |
| 749 | min(std::complex<T0> const& self, std::complex<T1> const& other) noexcept |
| 750 | { |
| 751 | return (self.real() < other.real()) ? (self) : (self.real() == other.real() ? (self.imag() < other.imag() ? self : other) : other); |
| 752 | } |
| 753 | |
| 754 | template <class T0, class T1> |
| 755 | inline auto max(T0 const& self, T1 const& other) noexcept |
| 756 | -> typename std::enable_if<std::is_scalar<T0>::value && std::is_scalar<T1>::value, |
| 757 | typename std::decay<decltype(self > other ? other : self)>::type>::type |
| 758 | { |
| 759 | return self < other ? other : self; |
| 760 | } |
| 761 | |
| 762 | // numpy defines maximum operator on complex using lexical comparison |
| 763 | template <class T0, class T1> |
| 764 | inline std::complex<typename std::common_type<T0, T1>::type> |
| 765 | max(std::complex<T0> const& self, std::complex<T1> const& other) noexcept |
| 766 | { |
| 767 | return (self.real() > other.real()) ? (self) : (self.real() == other.real() ? (self.imag() > other.imag() ? self : other) : other); |
| 768 | } |
| 769 | |
| 770 | template <class T> |
| 771 | inline typename std::enable_if<std::is_integral<T>::value, T>::type fma(const T& a, const T& b, const T& c) noexcept |
| 772 | { |
| 773 | return a * b + c; |
| 774 | } |
| 775 | |
| 776 | template <class T> |
| 777 | inline typename std::enable_if<std::is_floating_point<T>::value, T>::type fma(const T& a, const T& b, const T& c) noexcept |
| 778 | { |
| 779 | return std::fma(a, b, c); |
| 780 | } |
| 781 | |
| 782 | template <class T> |
| 783 | inline typename std::enable_if<std::is_scalar<T>::value, T>::type fms(const T& a, const T& b, const T& c) noexcept |
| 784 | { |
| 785 | return a * b - c; |
| 786 | } |
| 787 | |
| 788 | namespace detail |
| 789 | { |
| 790 | template <class C> |
| 791 | inline C fma_complex_scalar_impl(const C& a, const C& b, const C& c) noexcept |
| 792 | { |
| 793 | return { fms(a.real(), b.real(), fms(a.imag(), b.imag(), c.real())), |
| 794 | fma(a.real(), b.imag(), fma(a.imag(), b.real(), c.imag())) }; |
| 795 | } |
| 796 | } |
| 797 | |
| 798 | template <class T> |
| 799 | inline std::complex<T> fma(const std::complex<T>& a, const std::complex<T>& b, const std::complex<T>& c) noexcept |
| 800 | { |
| 801 | return detail::fma_complex_scalar_impl(a, b, c); |
| 802 | } |
| 803 | |
| 804 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 805 | template <class T, bool i3ec> |
| 806 | inline xtl::xcomplex<T, T, i3ec> fma(const xtl::xcomplex<T, T, i3ec>& a, const xtl::xcomplex<T, T, i3ec>& b, const xtl::xcomplex<T, T, i3ec>& c) noexcept |
| 807 | { |
| 808 | return detail::fma_complex_scalar_impl(a, b, c); |
| 809 | } |
| 810 | #endif |
| 811 | |
| 812 | namespace detail |
| 813 | { |
| 814 | template <class C> |
| 815 | inline C fms_complex_scalar_impl(const C& a, const C& b, const C& c) noexcept |
| 816 | { |
| 817 | return { fms(a.real(), b.real(), fma(a.imag(), b.imag(), c.real())), |
| 818 | fma(a.real(), b.imag(), fms(a.imag(), b.real(), c.imag())) }; |
| 819 | } |
| 820 | } |
| 821 | |
| 822 | template <class T> |
| 823 | inline std::complex<T> fms(const std::complex<T>& a, const std::complex<T>& b, const std::complex<T>& c) noexcept |
| 824 | { |
| 825 | return detail::fms_complex_scalar_impl(a, b, c); |
| 826 | } |
| 827 | |
| 828 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 829 | template <class T, bool i3ec> |
| 830 | inline xtl::xcomplex<T, T, i3ec> fms(const xtl::xcomplex<T, T, i3ec>& a, const xtl::xcomplex<T, T, i3ec>& b, const xtl::xcomplex<T, T, i3ec>& c) noexcept |
| 831 | { |
| 832 | return detail::fms_complex_scalar_impl(a, b, c); |
| 833 | } |
| 834 | #endif |
| 835 | |
| 836 | template <class T> |
| 837 | inline typename std::enable_if<std::is_integral<T>::value, T>::type fnma(const T& a, const T& b, const T& c) noexcept |
| 838 | { |
| 839 | return -(a * b) + c; |
| 840 | } |
| 841 | |
| 842 | template <class T> |
| 843 | inline typename std::enable_if<std::is_floating_point<T>::value, T>::type fnma(const T& a, const T& b, const T& c) noexcept |
| 844 | { |
| 845 | return std::fma(-a, b, c); |
| 846 | } |
| 847 | |
| 848 | namespace detail |
| 849 | { |
| 850 | template <class C> |
| 851 | inline C fnma_complex_scalar_impl(const C& a, const C& b, const C& c) noexcept |
| 852 | { |
| 853 | return { fms(a.imag(), b.imag(), fms(a.real(), b.real(), c.real())), |
| 854 | -fma(a.real(), b.imag(), fms(a.imag(), b.real(), c.imag())) }; |
| 855 | } |
| 856 | } |
| 857 | |
| 858 | template <class T> |
| 859 | inline std::complex<T> fnma(const std::complex<T>& a, const std::complex<T>& b, const std::complex<T>& c) noexcept |
| 860 | { |
| 861 | return detail::fnma_complex_scalar_impl(a, b, c); |
| 862 | } |
| 863 | |
| 864 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 865 | template <class T, bool i3ec> |
| 866 | inline xtl::xcomplex<T, T, i3ec> fnma(const xtl::xcomplex<T, T, i3ec>& a, const xtl::xcomplex<T, T, i3ec>& b, const xtl::xcomplex<T, T, i3ec>& c) noexcept |
| 867 | { |
| 868 | return detail::fnma_complex_scalar_impl(a, b, c); |
| 869 | } |
| 870 | #endif |
| 871 | |
| 872 | template <class T> |
| 873 | inline typename std::enable_if<std::is_integral<T>::value, T>::type fnms(const T& a, const T& b, const T& c) noexcept |
| 874 | { |
| 875 | return -(a * b) - c; |
| 876 | } |
| 877 | |
| 878 | template <class T> |
| 879 | inline typename std::enable_if<std::is_floating_point<T>::value, T>::type fnms(const T& a, const T& b, const T& c) noexcept |
| 880 | { |
| 881 | return -std::fma(a, b, c); |
| 882 | } |
| 883 | |
| 884 | namespace detail |
| 885 | { |
| 886 | template <class C> |
| 887 | inline C fnms_complex_scalar_impl(const C& a, const C& b, const C& c) noexcept |
| 888 | { |
| 889 | return { fms(a.imag(), b.imag(), fma(a.real(), b.real(), c.real())), |
| 890 | -fma(a.real(), b.imag(), fma(a.imag(), b.real(), c.imag())) }; |
| 891 | } |
| 892 | } |
| 893 | |
| 894 | template <class T> |
| 895 | inline std::complex<T> fnms(const std::complex<T>& a, const std::complex<T>& b, const std::complex<T>& c) noexcept |
| 896 | { |
| 897 | return detail::fnms_complex_scalar_impl(a, b, c); |
| 898 | } |
| 899 | |
| 900 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 901 | template <class T, bool i3ec> |
| 902 | inline xtl::xcomplex<T, T, i3ec> fnms(const xtl::xcomplex<T, T, i3ec>& a, const xtl::xcomplex<T, T, i3ec>& b, const xtl::xcomplex<T, T, i3ec>& c) noexcept |
| 903 | { |
| 904 | return detail::fnms_complex_scalar_impl(a, b, c); |
| 905 | } |
| 906 | #endif |
| 907 | |
| 908 | namespace detail |
| 909 | { |
| 910 | #define XSIMD_HASSINCOS_TRAIT(func) \ |
| 911 | template <class S> \ |
| 912 | struct has##func \ |
| 913 | { \ |
| 914 | template <class T> \ |
| 915 | static auto get(T* ptr) -> decltype(func(std::declval<T>(), std::declval<T*>(), std::declval<T*>()), std::true_type {}); \ |
| 916 | static std::false_type get(...); \ |
| 917 | static constexpr bool value = decltype(get((S*)nullptr))::value; \ |
| 918 | } |
| 919 | |
| 920 | #define XSIMD_HASSINCOS(func, T) has##func<T>::value |
| 921 | |
| 922 | XSIMD_HASSINCOS_TRAIT(sincos); |
| 923 | XSIMD_HASSINCOS_TRAIT(sincosf); |
| 924 | XSIMD_HASSINCOS_TRAIT(__sincos); |
| 925 | XSIMD_HASSINCOS_TRAIT(__sincosf); |
| 926 | |
| 927 | struct generic_sincosf |
| 928 | { |
| 929 | template <class T> |
| 930 | typename std::enable_if<XSIMD_HASSINCOS(sincosf, T), void>::type |
| 931 | operator()(float val, T& s, T& c) |
| 932 | { |
| 933 | sincosf(val, &s, &c); |
| 934 | } |
| 935 | |
| 936 | template <class T> |
| 937 | typename std::enable_if<!XSIMD_HASSINCOS(sincosf, T) && XSIMD_HASSINCOS(__sincosf, T), void>::type |
| 938 | operator()(float val, T& s, T& c) |
| 939 | { |
| 940 | __sincosf(val, &s, &c); |
| 941 | } |
| 942 | |
| 943 | template <class T> |
| 944 | typename std::enable_if<!XSIMD_HASSINCOS(sincosf, T) && !XSIMD_HASSINCOS(__sincosf, T), void>::type |
| 945 | operator()(float val, T& s, T& c) |
| 946 | { |
| 947 | s = std::sin(x: val); |
| 948 | c = std::cos(x: val); |
| 949 | } |
| 950 | }; |
| 951 | |
| 952 | struct generic_sincos |
| 953 | { |
| 954 | template <class T> |
| 955 | typename std::enable_if<XSIMD_HASSINCOS(sincos, T), void>::type |
| 956 | operator()(double val, T& s, T& c) |
| 957 | { |
| 958 | sincos(val, &s, &c); |
| 959 | } |
| 960 | |
| 961 | template <class T> |
| 962 | typename std::enable_if<!XSIMD_HASSINCOS(sincos, T) && XSIMD_HASSINCOS(__sincos, T), void>::type |
| 963 | operator()(double val, T& s, T& c) |
| 964 | { |
| 965 | __sincos(val, &s, &c); |
| 966 | } |
| 967 | |
| 968 | template <class T> |
| 969 | typename std::enable_if<!XSIMD_HASSINCOS(sincos, T) && !XSIMD_HASSINCOS(__sincos, T), void>::type |
| 970 | operator()(double val, T& s, T& c) |
| 971 | { |
| 972 | s = std::sin(x: val); |
| 973 | c = std::cos(x: val); |
| 974 | } |
| 975 | }; |
| 976 | |
| 977 | #undef XSIMD_HASSINCOS_TRAIT |
| 978 | #undef XSIMD_HASSINCOS |
| 979 | } |
| 980 | |
| 981 | inline std::pair<float, float> sincos(float val) noexcept |
| 982 | { |
| 983 | float s, c; |
| 984 | detail::generic_sincosf {}(val, s, c); |
| 985 | return std::make_pair(x&: s, y&: c); |
| 986 | } |
| 987 | |
| 988 | inline std::pair<double, double> sincos(double val) noexcept |
| 989 | { |
| 990 | double s, c; |
| 991 | detail::generic_sincos {}(val, s, c); |
| 992 | return std::make_pair(x&: s, y&: c); |
| 993 | } |
| 994 | |
| 995 | template <class T> |
| 996 | inline std::pair<std::complex<T>, std::complex<T>> |
| 997 | sincos(const std::complex<T>& val) noexcept |
| 998 | { |
| 999 | return std::make_pair(std::sin(val), std::cos(val)); |
| 1000 | } |
| 1001 | |
| 1002 | #ifdef XSIMD_ENABLE_XTL_COMPLEX |
| 1003 | template <class T> |
| 1004 | inline std::pair<xtl::xcomplex<T>, xtl::xcomplex<T>> sincos(const xtl::xcomplex<T>& val) noexcept |
| 1005 | { |
| 1006 | return std::make_pair(sin(val), cos(val)); |
| 1007 | } |
| 1008 | #endif |
| 1009 | |
| 1010 | template <class T, class _ = typename std::enable_if<std::is_floating_point<T>::value, void>::type> |
| 1011 | inline T frexp(T const& val, int& exp) noexcept |
| 1012 | { |
| 1013 | return std::frexp(val, &exp); |
| 1014 | } |
| 1015 | |
| 1016 | template <class T> |
| 1017 | inline T select(bool cond, T const& true_br, T const& false_br) noexcept |
| 1018 | { |
| 1019 | return cond ? true_br : false_br; |
| 1020 | } |
| 1021 | |
| 1022 | } |
| 1023 | |
| 1024 | #endif |
| 1025 | |