| 1 | /* |
| 2 | * kmp_atomic.h - ATOMIC header file |
| 3 | */ |
| 4 | |
| 5 | //===----------------------------------------------------------------------===// |
| 6 | // |
| 7 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 8 | // See https://llvm.org/LICENSE.txt for license information. |
| 9 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 10 | // |
| 11 | //===----------------------------------------------------------------------===// |
| 12 | |
| 13 | #ifndef KMP_ATOMIC_H |
| 14 | #define KMP_ATOMIC_H |
| 15 | |
| 16 | #include "kmp_lock.h" |
| 17 | #include "kmp_os.h" |
| 18 | |
| 19 | #if OMPT_SUPPORT |
| 20 | #include "ompt-specific.h" |
| 21 | #endif |
| 22 | |
| 23 | // C++ build port. |
| 24 | // Intel compiler does not support _Complex datatype on win. |
| 25 | // Intel compiler supports _Complex datatype on lin and mac. |
| 26 | // On the other side, there is a problem of stack alignment on lin_32 and mac_32 |
| 27 | // if the rhs is cmplx80 or cmplx128 typedef'ed datatype. |
| 28 | // The decision is: to use compiler supported _Complex type on lin and mac, |
| 29 | // to use typedef'ed types on win. |
| 30 | // Condition for WIN64 was modified in anticipation of 10.1 build compiler. |
| 31 | |
| 32 | #if defined(__cplusplus) && (KMP_OS_WINDOWS) |
| 33 | // create shortcuts for c99 complex types |
| 34 | |
| 35 | // Visual Studio cannot have function parameters that have the |
| 36 | // align __declspec attribute, so we must remove it. (Compiler Error C2719) |
| 37 | #if KMP_COMPILER_MSVC |
| 38 | #undef KMP_DO_ALIGN |
| 39 | #define KMP_DO_ALIGN(alignment) /* Nothing */ |
| 40 | #endif |
| 41 | |
| 42 | #if defined(_MSC_VER) && (_MSC_VER < 1600) && defined(_DEBUG) |
| 43 | // Workaround for the problem of _DebugHeapTag unresolved external. |
| 44 | // This problem prevented to use our static debug library for C tests |
| 45 | // compiled with /MDd option (the library itself built with /MTd), |
| 46 | #undef _DEBUG |
| 47 | #define _DEBUG_TEMPORARILY_UNSET_ |
| 48 | #endif |
| 49 | |
| 50 | #include <complex> |
| 51 | |
| 52 | template <typename type_lhs, typename type_rhs> |
| 53 | std::complex<type_lhs> __kmp_lhs_div_rhs(const std::complex<type_lhs> &lhs, |
| 54 | const std::complex<type_rhs> &rhs) { |
| 55 | type_lhs a = lhs.real(); |
| 56 | type_lhs b = lhs.imag(); |
| 57 | type_rhs c = rhs.real(); |
| 58 | type_rhs d = rhs.imag(); |
| 59 | type_rhs den = c * c + d * d; |
| 60 | type_rhs r = (a * c + b * d); |
| 61 | type_rhs i = (b * c - a * d); |
| 62 | std::complex<type_lhs> ret(r / den, i / den); |
| 63 | return ret; |
| 64 | } |
| 65 | |
| 66 | // complex8 |
| 67 | struct __kmp_cmplx64_t : std::complex<double> { |
| 68 | |
| 69 | __kmp_cmplx64_t() : std::complex<double>() {} |
| 70 | |
| 71 | __kmp_cmplx64_t(const std::complex<double> &cd) : std::complex<double>(cd) {} |
| 72 | |
| 73 | void operator/=(const __kmp_cmplx64_t &rhs) { |
| 74 | std::complex<double> lhs = *this; |
| 75 | *this = __kmp_lhs_div_rhs(lhs, rhs); |
| 76 | } |
| 77 | |
| 78 | __kmp_cmplx64_t operator/(const __kmp_cmplx64_t &rhs) { |
| 79 | std::complex<double> lhs = *this; |
| 80 | return __kmp_lhs_div_rhs(lhs, rhs); |
| 81 | } |
| 82 | }; |
| 83 | typedef struct __kmp_cmplx64_t kmp_cmplx64; |
| 84 | |
| 85 | // complex4 |
| 86 | struct __kmp_cmplx32_t : std::complex<float> { |
| 87 | |
| 88 | __kmp_cmplx32_t() : std::complex<float>() {} |
| 89 | |
| 90 | __kmp_cmplx32_t(const std::complex<float> &cf) : std::complex<float>(cf) {} |
| 91 | |
| 92 | __kmp_cmplx32_t operator+(const __kmp_cmplx32_t &b) { |
| 93 | std::complex<float> lhs = *this; |
| 94 | std::complex<float> rhs = b; |
| 95 | return (lhs + rhs); |
| 96 | } |
| 97 | __kmp_cmplx32_t operator-(const __kmp_cmplx32_t &b) { |
| 98 | std::complex<float> lhs = *this; |
| 99 | std::complex<float> rhs = b; |
| 100 | return (lhs - rhs); |
| 101 | } |
| 102 | __kmp_cmplx32_t operator*(const __kmp_cmplx32_t &b) { |
| 103 | std::complex<float> lhs = *this; |
| 104 | std::complex<float> rhs = b; |
| 105 | return (lhs * rhs); |
| 106 | } |
| 107 | |
| 108 | __kmp_cmplx32_t operator+(const kmp_cmplx64 &b) { |
| 109 | kmp_cmplx64 t = kmp_cmplx64(*this) + b; |
| 110 | std::complex<double> d(t); |
| 111 | std::complex<float> f(d); |
| 112 | __kmp_cmplx32_t r(f); |
| 113 | return r; |
| 114 | } |
| 115 | __kmp_cmplx32_t operator-(const kmp_cmplx64 &b) { |
| 116 | kmp_cmplx64 t = kmp_cmplx64(*this) - b; |
| 117 | std::complex<double> d(t); |
| 118 | std::complex<float> f(d); |
| 119 | __kmp_cmplx32_t r(f); |
| 120 | return r; |
| 121 | } |
| 122 | __kmp_cmplx32_t operator*(const kmp_cmplx64 &b) { |
| 123 | kmp_cmplx64 t = kmp_cmplx64(*this) * b; |
| 124 | std::complex<double> d(t); |
| 125 | std::complex<float> f(d); |
| 126 | __kmp_cmplx32_t r(f); |
| 127 | return r; |
| 128 | } |
| 129 | |
| 130 | void operator/=(const __kmp_cmplx32_t &rhs) { |
| 131 | std::complex<float> lhs = *this; |
| 132 | *this = __kmp_lhs_div_rhs(lhs, rhs); |
| 133 | } |
| 134 | |
| 135 | __kmp_cmplx32_t operator/(const __kmp_cmplx32_t &rhs) { |
| 136 | std::complex<float> lhs = *this; |
| 137 | return __kmp_lhs_div_rhs(lhs, rhs); |
| 138 | } |
| 139 | |
| 140 | void operator/=(const kmp_cmplx64 &rhs) { |
| 141 | std::complex<float> lhs = *this; |
| 142 | *this = __kmp_lhs_div_rhs(lhs, rhs); |
| 143 | } |
| 144 | |
| 145 | __kmp_cmplx32_t operator/(const kmp_cmplx64 &rhs) { |
| 146 | std::complex<float> lhs = *this; |
| 147 | return __kmp_lhs_div_rhs(lhs, rhs); |
| 148 | } |
| 149 | }; |
| 150 | typedef struct __kmp_cmplx32_t kmp_cmplx32; |
| 151 | |
| 152 | // complex10 |
| 153 | struct KMP_DO_ALIGN(16) __kmp_cmplx80_t : std::complex<long double> { |
| 154 | |
| 155 | __kmp_cmplx80_t() : std::complex<long double>() {} |
| 156 | |
| 157 | __kmp_cmplx80_t(const std::complex<long double> &cld) |
| 158 | : std::complex<long double>(cld) {} |
| 159 | |
| 160 | void operator/=(const __kmp_cmplx80_t &rhs) { |
| 161 | std::complex<long double> lhs = *this; |
| 162 | *this = __kmp_lhs_div_rhs(lhs, rhs); |
| 163 | } |
| 164 | |
| 165 | __kmp_cmplx80_t operator/(const __kmp_cmplx80_t &rhs) { |
| 166 | std::complex<long double> lhs = *this; |
| 167 | return __kmp_lhs_div_rhs(lhs, rhs); |
| 168 | } |
| 169 | }; |
| 170 | typedef KMP_DO_ALIGN(16) struct __kmp_cmplx80_t kmp_cmplx80; |
| 171 | |
| 172 | // complex16 |
| 173 | #if KMP_HAVE_QUAD |
| 174 | struct __kmp_cmplx128_t : std::complex<_Quad> { |
| 175 | |
| 176 | __kmp_cmplx128_t() : std::complex<_Quad>() {} |
| 177 | |
| 178 | __kmp_cmplx128_t(const std::complex<_Quad> &cq) : std::complex<_Quad>(cq) {} |
| 179 | |
| 180 | void operator/=(const __kmp_cmplx128_t &rhs) { |
| 181 | std::complex<_Quad> lhs = *this; |
| 182 | *this = __kmp_lhs_div_rhs(lhs, rhs); |
| 183 | } |
| 184 | |
| 185 | __kmp_cmplx128_t operator/(const __kmp_cmplx128_t &rhs) { |
| 186 | std::complex<_Quad> lhs = *this; |
| 187 | return __kmp_lhs_div_rhs(lhs, rhs); |
| 188 | } |
| 189 | }; |
| 190 | typedef struct __kmp_cmplx128_t kmp_cmplx128; |
| 191 | #endif /* KMP_HAVE_QUAD */ |
| 192 | |
| 193 | #ifdef _DEBUG_TEMPORARILY_UNSET_ |
| 194 | #undef _DEBUG_TEMPORARILY_UNSET_ |
| 195 | // Set it back now |
| 196 | #define _DEBUG 1 |
| 197 | #endif |
| 198 | |
| 199 | #else |
| 200 | // create shortcuts for c99 complex types |
| 201 | typedef float _Complex kmp_cmplx32; |
| 202 | typedef double _Complex kmp_cmplx64; |
| 203 | typedef long double _Complex kmp_cmplx80; |
| 204 | #if KMP_HAVE_QUAD |
| 205 | typedef _Quad _Complex kmp_cmplx128; |
| 206 | #endif |
| 207 | #endif |
| 208 | |
| 209 | // Compiler 12.0 changed alignment of 16 and 32-byte arguments (like _Quad |
| 210 | // and kmp_cmplx128) on IA-32 architecture. The following aligned structures |
| 211 | // are implemented to support the old alignment in 10.1, 11.0, 11.1 and |
| 212 | // introduce the new alignment in 12.0. See CQ88405. |
| 213 | #if KMP_ARCH_X86 && KMP_HAVE_QUAD |
| 214 | |
| 215 | // 4-byte aligned structures for backward compatibility. |
| 216 | |
| 217 | #pragma pack(push, 4) |
| 218 | |
| 219 | struct KMP_DO_ALIGN(4) Quad_a4_t { |
| 220 | _Quad q; |
| 221 | |
| 222 | Quad_a4_t() : q() {} |
| 223 | Quad_a4_t(const _Quad &cq) : q(cq) {} |
| 224 | |
| 225 | Quad_a4_t operator+(const Quad_a4_t &b) { |
| 226 | _Quad lhs = (*this).q; |
| 227 | _Quad rhs = b.q; |
| 228 | return (Quad_a4_t)(lhs + rhs); |
| 229 | } |
| 230 | |
| 231 | Quad_a4_t operator-(const Quad_a4_t &b) { |
| 232 | _Quad lhs = (*this).q; |
| 233 | _Quad rhs = b.q; |
| 234 | return (Quad_a4_t)(lhs - rhs); |
| 235 | } |
| 236 | Quad_a4_t operator*(const Quad_a4_t &b) { |
| 237 | _Quad lhs = (*this).q; |
| 238 | _Quad rhs = b.q; |
| 239 | return (Quad_a4_t)(lhs * rhs); |
| 240 | } |
| 241 | |
| 242 | Quad_a4_t operator/(const Quad_a4_t &b) { |
| 243 | _Quad lhs = (*this).q; |
| 244 | _Quad rhs = b.q; |
| 245 | return (Quad_a4_t)(lhs / rhs); |
| 246 | } |
| 247 | }; |
| 248 | |
| 249 | struct KMP_DO_ALIGN(4) kmp_cmplx128_a4_t { |
| 250 | kmp_cmplx128 q; |
| 251 | |
| 252 | kmp_cmplx128_a4_t() : q() {} |
| 253 | |
| 254 | #if defined(__cplusplus) && (KMP_OS_WINDOWS) |
| 255 | kmp_cmplx128_a4_t(const std::complex<_Quad> &c128) : q(c128) {} |
| 256 | #endif |
| 257 | kmp_cmplx128_a4_t(const kmp_cmplx128 &c128) : q(c128) {} |
| 258 | |
| 259 | kmp_cmplx128_a4_t operator+(const kmp_cmplx128_a4_t &b) { |
| 260 | kmp_cmplx128 lhs = (*this).q; |
| 261 | kmp_cmplx128 rhs = b.q; |
| 262 | return (kmp_cmplx128_a4_t)(lhs + rhs); |
| 263 | } |
| 264 | kmp_cmplx128_a4_t operator-(const kmp_cmplx128_a4_t &b) { |
| 265 | kmp_cmplx128 lhs = (*this).q; |
| 266 | kmp_cmplx128 rhs = b.q; |
| 267 | return (kmp_cmplx128_a4_t)(lhs - rhs); |
| 268 | } |
| 269 | kmp_cmplx128_a4_t operator*(const kmp_cmplx128_a4_t &b) { |
| 270 | kmp_cmplx128 lhs = (*this).q; |
| 271 | kmp_cmplx128 rhs = b.q; |
| 272 | return (kmp_cmplx128_a4_t)(lhs * rhs); |
| 273 | } |
| 274 | |
| 275 | kmp_cmplx128_a4_t operator/(const kmp_cmplx128_a4_t &b) { |
| 276 | kmp_cmplx128 lhs = (*this).q; |
| 277 | kmp_cmplx128 rhs = b.q; |
| 278 | return (kmp_cmplx128_a4_t)(lhs / rhs); |
| 279 | } |
| 280 | }; |
| 281 | |
| 282 | #pragma pack(pop) |
| 283 | |
| 284 | // New 16-byte aligned structures for 12.0 compiler. |
| 285 | struct KMP_DO_ALIGN(16) Quad_a16_t { |
| 286 | _Quad q; |
| 287 | |
| 288 | Quad_a16_t() : q() {} |
| 289 | Quad_a16_t(const _Quad &cq) : q(cq) {} |
| 290 | |
| 291 | Quad_a16_t operator+(const Quad_a16_t &b) { |
| 292 | _Quad lhs = (*this).q; |
| 293 | _Quad rhs = b.q; |
| 294 | return (Quad_a16_t)(lhs + rhs); |
| 295 | } |
| 296 | |
| 297 | Quad_a16_t operator-(const Quad_a16_t &b) { |
| 298 | _Quad lhs = (*this).q; |
| 299 | _Quad rhs = b.q; |
| 300 | return (Quad_a16_t)(lhs - rhs); |
| 301 | } |
| 302 | Quad_a16_t operator*(const Quad_a16_t &b) { |
| 303 | _Quad lhs = (*this).q; |
| 304 | _Quad rhs = b.q; |
| 305 | return (Quad_a16_t)(lhs * rhs); |
| 306 | } |
| 307 | |
| 308 | Quad_a16_t operator/(const Quad_a16_t &b) { |
| 309 | _Quad lhs = (*this).q; |
| 310 | _Quad rhs = b.q; |
| 311 | return (Quad_a16_t)(lhs / rhs); |
| 312 | } |
| 313 | }; |
| 314 | |
| 315 | struct KMP_DO_ALIGN(16) kmp_cmplx128_a16_t { |
| 316 | kmp_cmplx128 q; |
| 317 | |
| 318 | kmp_cmplx128_a16_t() : q() {} |
| 319 | |
| 320 | #if defined(__cplusplus) && (KMP_OS_WINDOWS) |
| 321 | kmp_cmplx128_a16_t(const std::complex<_Quad> &c128) : q(c128) {} |
| 322 | #endif |
| 323 | kmp_cmplx128_a16_t(const kmp_cmplx128 &c128) : q(c128) {} |
| 324 | |
| 325 | kmp_cmplx128_a16_t operator+(const kmp_cmplx128_a16_t &b) { |
| 326 | kmp_cmplx128 lhs = (*this).q; |
| 327 | kmp_cmplx128 rhs = b.q; |
| 328 | return (kmp_cmplx128_a16_t)(lhs + rhs); |
| 329 | } |
| 330 | kmp_cmplx128_a16_t operator-(const kmp_cmplx128_a16_t &b) { |
| 331 | kmp_cmplx128 lhs = (*this).q; |
| 332 | kmp_cmplx128 rhs = b.q; |
| 333 | return (kmp_cmplx128_a16_t)(lhs - rhs); |
| 334 | } |
| 335 | kmp_cmplx128_a16_t operator*(const kmp_cmplx128_a16_t &b) { |
| 336 | kmp_cmplx128 lhs = (*this).q; |
| 337 | kmp_cmplx128 rhs = b.q; |
| 338 | return (kmp_cmplx128_a16_t)(lhs * rhs); |
| 339 | } |
| 340 | |
| 341 | kmp_cmplx128_a16_t operator/(const kmp_cmplx128_a16_t &b) { |
| 342 | kmp_cmplx128 lhs = (*this).q; |
| 343 | kmp_cmplx128 rhs = b.q; |
| 344 | return (kmp_cmplx128_a16_t)(lhs / rhs); |
| 345 | } |
| 346 | }; |
| 347 | |
| 348 | #endif |
| 349 | |
| 350 | #if (KMP_ARCH_X86) |
| 351 | #define QUAD_LEGACY Quad_a4_t |
| 352 | #define CPLX128_LEG kmp_cmplx128_a4_t |
| 353 | #else |
| 354 | #define QUAD_LEGACY _Quad |
| 355 | #define CPLX128_LEG kmp_cmplx128 |
| 356 | #endif |
| 357 | |
| 358 | #ifdef __cplusplus |
| 359 | extern "C" { |
| 360 | #endif |
| 361 | |
| 362 | extern int __kmp_atomic_mode; |
| 363 | |
| 364 | // Atomic locks can easily become contended, so we use queuing locks for them. |
| 365 | typedef kmp_queuing_lock_t kmp_atomic_lock_t; |
| 366 | |
| 367 | static inline void __kmp_acquire_atomic_lock(kmp_atomic_lock_t *lck, |
| 368 | kmp_int32 gtid) { |
| 369 | #if OMPT_SUPPORT && OMPT_OPTIONAL |
| 370 | if (ompt_enabled.ompt_callback_mutex_acquire) { |
| 371 | ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)( |
| 372 | ompt_mutex_atomic, 0, kmp_mutex_impl_queuing, |
| 373 | (ompt_wait_id_t)(uintptr_t)lck, OMPT_GET_RETURN_ADDRESS(0)); |
| 374 | } |
| 375 | #endif |
| 376 | |
| 377 | __kmp_acquire_queuing_lock(lck, gtid); |
| 378 | |
| 379 | #if OMPT_SUPPORT && OMPT_OPTIONAL |
| 380 | if (ompt_enabled.ompt_callback_mutex_acquired) { |
| 381 | ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)( |
| 382 | ompt_mutex_atomic, (ompt_wait_id_t)(uintptr_t)lck, |
| 383 | OMPT_GET_RETURN_ADDRESS(0)); |
| 384 | } |
| 385 | #endif |
| 386 | } |
| 387 | |
| 388 | static inline int __kmp_test_atomic_lock(kmp_atomic_lock_t *lck, |
| 389 | kmp_int32 gtid) { |
| 390 | return __kmp_test_queuing_lock(lck, gtid); |
| 391 | } |
| 392 | |
| 393 | static inline void __kmp_release_atomic_lock(kmp_atomic_lock_t *lck, |
| 394 | kmp_int32 gtid) { |
| 395 | __kmp_release_queuing_lock(lck, gtid); |
| 396 | #if OMPT_SUPPORT && OMPT_OPTIONAL |
| 397 | if (ompt_enabled.ompt_callback_mutex_released) { |
| 398 | ompt_callbacks.ompt_callback(ompt_callback_mutex_released)( |
| 399 | ompt_mutex_atomic, (ompt_wait_id_t)(uintptr_t)lck, |
| 400 | OMPT_GET_RETURN_ADDRESS(0)); |
| 401 | } |
| 402 | #endif |
| 403 | } |
| 404 | |
| 405 | static inline void __kmp_init_atomic_lock(kmp_atomic_lock_t *lck) { |
| 406 | __kmp_init_queuing_lock(lck); |
| 407 | } |
| 408 | |
| 409 | static inline void __kmp_destroy_atomic_lock(kmp_atomic_lock_t *lck) { |
| 410 | __kmp_destroy_queuing_lock(lck); |
| 411 | } |
| 412 | |
| 413 | // Global Locks |
| 414 | extern kmp_atomic_lock_t __kmp_atomic_lock; /* Control access to all user coded |
| 415 | atomics in Gnu compat mode */ |
| 416 | extern kmp_atomic_lock_t __kmp_atomic_lock_1i; /* Control access to all user |
| 417 | coded atomics for 1-byte fixed |
| 418 | data types */ |
| 419 | extern kmp_atomic_lock_t __kmp_atomic_lock_2i; /* Control access to all user |
| 420 | coded atomics for 2-byte fixed |
| 421 | data types */ |
| 422 | extern kmp_atomic_lock_t __kmp_atomic_lock_4i; /* Control access to all user |
| 423 | coded atomics for 4-byte fixed |
| 424 | data types */ |
| 425 | extern kmp_atomic_lock_t __kmp_atomic_lock_4r; /* Control access to all user |
| 426 | coded atomics for kmp_real32 |
| 427 | data type */ |
| 428 | extern kmp_atomic_lock_t __kmp_atomic_lock_8i; /* Control access to all user |
| 429 | coded atomics for 8-byte fixed |
| 430 | data types */ |
| 431 | extern kmp_atomic_lock_t __kmp_atomic_lock_8r; /* Control access to all user |
| 432 | coded atomics for kmp_real64 |
| 433 | data type */ |
| 434 | extern kmp_atomic_lock_t |
| 435 | __kmp_atomic_lock_8c; /* Control access to all user coded atomics for |
| 436 | complex byte data type */ |
| 437 | extern kmp_atomic_lock_t |
| 438 | __kmp_atomic_lock_10r; /* Control access to all user coded atomics for long |
| 439 | double data type */ |
| 440 | extern kmp_atomic_lock_t __kmp_atomic_lock_16r; /* Control access to all user |
| 441 | coded atomics for _Quad data |
| 442 | type */ |
| 443 | extern kmp_atomic_lock_t __kmp_atomic_lock_16c; /* Control access to all user |
| 444 | coded atomics for double |
| 445 | complex data type*/ |
| 446 | extern kmp_atomic_lock_t |
| 447 | __kmp_atomic_lock_20c; /* Control access to all user coded atomics for long |
| 448 | double complex type*/ |
| 449 | extern kmp_atomic_lock_t __kmp_atomic_lock_32c; /* Control access to all user |
| 450 | coded atomics for _Quad |
| 451 | complex data type */ |
| 452 | |
| 453 | // Below routines for atomic UPDATE are listed |
| 454 | |
| 455 | // 1-byte |
| 456 | void __kmpc_atomic_fixed1_add(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 457 | void __kmpc_atomic_fixed1_andb(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 458 | void __kmpc_atomic_fixed1_div(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 459 | void __kmpc_atomic_fixed1u_div(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 460 | unsigned char rhs); |
| 461 | void __kmpc_atomic_fixed1_mul(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 462 | void __kmpc_atomic_fixed1_orb(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 463 | void __kmpc_atomic_fixed1_shl(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 464 | void __kmpc_atomic_fixed1_shr(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 465 | void __kmpc_atomic_fixed1u_shr(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 466 | unsigned char rhs); |
| 467 | void __kmpc_atomic_fixed1_sub(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 468 | void __kmpc_atomic_fixed1_xor(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 469 | // 2-byte |
| 470 | void __kmpc_atomic_fixed2_add(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 471 | void __kmpc_atomic_fixed2_andb(ident_t *id_ref, int gtid, short *lhs, |
| 472 | short rhs); |
| 473 | void __kmpc_atomic_fixed2_div(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 474 | void __kmpc_atomic_fixed2u_div(ident_t *id_ref, int gtid, unsigned short *lhs, |
| 475 | unsigned short rhs); |
| 476 | void __kmpc_atomic_fixed2_mul(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 477 | void __kmpc_atomic_fixed2_orb(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 478 | void __kmpc_atomic_fixed2_shl(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 479 | void __kmpc_atomic_fixed2_shr(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 480 | void __kmpc_atomic_fixed2u_shr(ident_t *id_ref, int gtid, unsigned short *lhs, |
| 481 | unsigned short rhs); |
| 482 | void __kmpc_atomic_fixed2_sub(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 483 | void __kmpc_atomic_fixed2_xor(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 484 | // 4-byte add / sub fixed |
| 485 | void __kmpc_atomic_fixed4_add(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 486 | kmp_int32 rhs); |
| 487 | void __kmpc_atomic_fixed4_sub(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 488 | kmp_int32 rhs); |
| 489 | // 4-byte add / sub float |
| 490 | void __kmpc_atomic_float4_add(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 491 | kmp_real32 rhs); |
| 492 | void __kmpc_atomic_float4_sub(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 493 | kmp_real32 rhs); |
| 494 | // 8-byte add / sub fixed |
| 495 | void __kmpc_atomic_fixed8_add(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 496 | kmp_int64 rhs); |
| 497 | void __kmpc_atomic_fixed8_sub(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 498 | kmp_int64 rhs); |
| 499 | // 8-byte add / sub float |
| 500 | void __kmpc_atomic_float8_add(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 501 | kmp_real64 rhs); |
| 502 | void __kmpc_atomic_float8_sub(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 503 | kmp_real64 rhs); |
| 504 | // 4-byte fixed |
| 505 | void __kmpc_atomic_fixed4_andb(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 506 | kmp_int32 rhs); |
| 507 | void __kmpc_atomic_fixed4_div(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 508 | kmp_int32 rhs); |
| 509 | void __kmpc_atomic_fixed4u_div(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 510 | kmp_uint32 rhs); |
| 511 | void __kmpc_atomic_fixed4_mul(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 512 | kmp_int32 rhs); |
| 513 | void __kmpc_atomic_fixed4_orb(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 514 | kmp_int32 rhs); |
| 515 | void __kmpc_atomic_fixed4_shl(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 516 | kmp_int32 rhs); |
| 517 | void __kmpc_atomic_fixed4_shr(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 518 | kmp_int32 rhs); |
| 519 | void __kmpc_atomic_fixed4u_shr(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 520 | kmp_uint32 rhs); |
| 521 | void __kmpc_atomic_fixed4_xor(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 522 | kmp_int32 rhs); |
| 523 | // 8-byte fixed |
| 524 | void __kmpc_atomic_fixed8_andb(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 525 | kmp_int64 rhs); |
| 526 | void __kmpc_atomic_fixed8_div(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 527 | kmp_int64 rhs); |
| 528 | void __kmpc_atomic_fixed8u_div(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 529 | kmp_uint64 rhs); |
| 530 | void __kmpc_atomic_fixed8_mul(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 531 | kmp_int64 rhs); |
| 532 | void __kmpc_atomic_fixed8_orb(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 533 | kmp_int64 rhs); |
| 534 | void __kmpc_atomic_fixed8_shl(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 535 | kmp_int64 rhs); |
| 536 | void __kmpc_atomic_fixed8_shr(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 537 | kmp_int64 rhs); |
| 538 | void __kmpc_atomic_fixed8u_shr(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 539 | kmp_uint64 rhs); |
| 540 | void __kmpc_atomic_fixed8_xor(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 541 | kmp_int64 rhs); |
| 542 | // 4-byte float |
| 543 | void __kmpc_atomic_float4_div(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 544 | kmp_real32 rhs); |
| 545 | void __kmpc_atomic_float4_mul(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 546 | kmp_real32 rhs); |
| 547 | // 8-byte float |
| 548 | void __kmpc_atomic_float8_div(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 549 | kmp_real64 rhs); |
| 550 | void __kmpc_atomic_float8_mul(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 551 | kmp_real64 rhs); |
| 552 | // 1-, 2-, 4-, 8-byte logical (&&, ||) |
| 553 | void __kmpc_atomic_fixed1_andl(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 554 | void __kmpc_atomic_fixed1_orl(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 555 | void __kmpc_atomic_fixed2_andl(ident_t *id_ref, int gtid, short *lhs, |
| 556 | short rhs); |
| 557 | void __kmpc_atomic_fixed2_orl(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 558 | void __kmpc_atomic_fixed4_andl(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 559 | kmp_int32 rhs); |
| 560 | void __kmpc_atomic_fixed4_orl(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 561 | kmp_int32 rhs); |
| 562 | void __kmpc_atomic_fixed8_andl(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 563 | kmp_int64 rhs); |
| 564 | void __kmpc_atomic_fixed8_orl(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 565 | kmp_int64 rhs); |
| 566 | // MIN / MAX |
| 567 | void __kmpc_atomic_fixed1_max(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 568 | void __kmpc_atomic_fixed1_min(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 569 | void __kmpc_atomic_fixed2_max(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 570 | void __kmpc_atomic_fixed2_min(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 571 | void __kmpc_atomic_fixed4_max(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 572 | kmp_int32 rhs); |
| 573 | void __kmpc_atomic_fixed4_min(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 574 | kmp_int32 rhs); |
| 575 | void __kmpc_atomic_fixed8_max(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 576 | kmp_int64 rhs); |
| 577 | void __kmpc_atomic_fixed8_min(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 578 | kmp_int64 rhs); |
| 579 | void __kmpc_atomic_float4_max(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 580 | kmp_real32 rhs); |
| 581 | void __kmpc_atomic_float4_min(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 582 | kmp_real32 rhs); |
| 583 | void __kmpc_atomic_float8_max(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 584 | kmp_real64 rhs); |
| 585 | void __kmpc_atomic_float8_min(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 586 | kmp_real64 rhs); |
| 587 | void __kmpc_atomic_float10_max(ident_t *id_ref, int gtid, long double *lhs, |
| 588 | long double rhs); |
| 589 | void __kmpc_atomic_float10_min(ident_t *id_ref, int gtid, long double *lhs, |
| 590 | long double rhs); |
| 591 | #if KMP_HAVE_QUAD |
| 592 | void __kmpc_atomic_float16_max(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 593 | QUAD_LEGACY rhs); |
| 594 | void __kmpc_atomic_float16_min(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 595 | QUAD_LEGACY rhs); |
| 596 | #if (KMP_ARCH_X86) |
| 597 | // Routines with 16-byte arguments aligned to 16-byte boundary; IA-32 |
| 598 | // architecture only |
| 599 | void __kmpc_atomic_float16_max_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 600 | Quad_a16_t rhs); |
| 601 | void __kmpc_atomic_float16_min_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 602 | Quad_a16_t rhs); |
| 603 | #endif |
| 604 | #endif |
| 605 | // .NEQV. (same as xor) |
| 606 | void __kmpc_atomic_fixed1_neqv(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 607 | void __kmpc_atomic_fixed2_neqv(ident_t *id_ref, int gtid, short *lhs, |
| 608 | short rhs); |
| 609 | void __kmpc_atomic_fixed4_neqv(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 610 | kmp_int32 rhs); |
| 611 | void __kmpc_atomic_fixed8_neqv(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 612 | kmp_int64 rhs); |
| 613 | // .EQV. (same as ~xor) |
| 614 | void __kmpc_atomic_fixed1_eqv(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 615 | void __kmpc_atomic_fixed2_eqv(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 616 | void __kmpc_atomic_fixed4_eqv(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 617 | kmp_int32 rhs); |
| 618 | void __kmpc_atomic_fixed8_eqv(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 619 | kmp_int64 rhs); |
| 620 | // long double type |
| 621 | void __kmpc_atomic_float10_add(ident_t *id_ref, int gtid, long double *lhs, |
| 622 | long double rhs); |
| 623 | void __kmpc_atomic_float10_sub(ident_t *id_ref, int gtid, long double *lhs, |
| 624 | long double rhs); |
| 625 | void __kmpc_atomic_float10_mul(ident_t *id_ref, int gtid, long double *lhs, |
| 626 | long double rhs); |
| 627 | void __kmpc_atomic_float10_div(ident_t *id_ref, int gtid, long double *lhs, |
| 628 | long double rhs); |
| 629 | // _Quad type |
| 630 | #if KMP_HAVE_QUAD |
| 631 | void __kmpc_atomic_float16_add(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 632 | QUAD_LEGACY rhs); |
| 633 | void __kmpc_atomic_float16_sub(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 634 | QUAD_LEGACY rhs); |
| 635 | void __kmpc_atomic_float16_mul(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 636 | QUAD_LEGACY rhs); |
| 637 | void __kmpc_atomic_float16_div(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 638 | QUAD_LEGACY rhs); |
| 639 | #if (KMP_ARCH_X86) |
| 640 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 641 | void __kmpc_atomic_float16_add_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 642 | Quad_a16_t rhs); |
| 643 | void __kmpc_atomic_float16_sub_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 644 | Quad_a16_t rhs); |
| 645 | void __kmpc_atomic_float16_mul_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 646 | Quad_a16_t rhs); |
| 647 | void __kmpc_atomic_float16_div_a16(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 648 | Quad_a16_t rhs); |
| 649 | #endif |
| 650 | #endif |
| 651 | // routines for complex types |
| 652 | void __kmpc_atomic_cmplx4_add(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 653 | kmp_cmplx32 rhs); |
| 654 | void __kmpc_atomic_cmplx4_sub(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 655 | kmp_cmplx32 rhs); |
| 656 | void __kmpc_atomic_cmplx4_mul(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 657 | kmp_cmplx32 rhs); |
| 658 | void __kmpc_atomic_cmplx4_div(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 659 | kmp_cmplx32 rhs); |
| 660 | void __kmpc_atomic_cmplx8_add(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 661 | kmp_cmplx64 rhs); |
| 662 | void __kmpc_atomic_cmplx8_sub(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 663 | kmp_cmplx64 rhs); |
| 664 | void __kmpc_atomic_cmplx8_mul(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 665 | kmp_cmplx64 rhs); |
| 666 | void __kmpc_atomic_cmplx8_div(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 667 | kmp_cmplx64 rhs); |
| 668 | void __kmpc_atomic_cmplx10_add(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 669 | kmp_cmplx80 rhs); |
| 670 | void __kmpc_atomic_cmplx10_sub(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 671 | kmp_cmplx80 rhs); |
| 672 | void __kmpc_atomic_cmplx10_mul(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 673 | kmp_cmplx80 rhs); |
| 674 | void __kmpc_atomic_cmplx10_div(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 675 | kmp_cmplx80 rhs); |
| 676 | #if KMP_HAVE_QUAD |
| 677 | void __kmpc_atomic_cmplx16_add(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 678 | CPLX128_LEG rhs); |
| 679 | void __kmpc_atomic_cmplx16_sub(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 680 | CPLX128_LEG rhs); |
| 681 | void __kmpc_atomic_cmplx16_mul(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 682 | CPLX128_LEG rhs); |
| 683 | void __kmpc_atomic_cmplx16_div(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 684 | CPLX128_LEG rhs); |
| 685 | #if (KMP_ARCH_X86) |
| 686 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 687 | void __kmpc_atomic_cmplx16_add_a16(ident_t *id_ref, int gtid, |
| 688 | kmp_cmplx128_a16_t *lhs, |
| 689 | kmp_cmplx128_a16_t rhs); |
| 690 | void __kmpc_atomic_cmplx16_sub_a16(ident_t *id_ref, int gtid, |
| 691 | kmp_cmplx128_a16_t *lhs, |
| 692 | kmp_cmplx128_a16_t rhs); |
| 693 | void __kmpc_atomic_cmplx16_mul_a16(ident_t *id_ref, int gtid, |
| 694 | kmp_cmplx128_a16_t *lhs, |
| 695 | kmp_cmplx128_a16_t rhs); |
| 696 | void __kmpc_atomic_cmplx16_div_a16(ident_t *id_ref, int gtid, |
| 697 | kmp_cmplx128_a16_t *lhs, |
| 698 | kmp_cmplx128_a16_t rhs); |
| 699 | #endif |
| 700 | #endif |
| 701 | |
| 702 | // OpenMP 4.0: x = expr binop x for non-commutative operations. |
| 703 | // Supported only on IA-32 architecture and Intel(R) 64 |
| 704 | #if KMP_ARCH_X86 || KMP_ARCH_X86_64 |
| 705 | |
| 706 | void __kmpc_atomic_fixed1_sub_rev(ident_t *id_ref, int gtid, char *lhs, |
| 707 | char rhs); |
| 708 | void __kmpc_atomic_fixed1_div_rev(ident_t *id_ref, int gtid, char *lhs, |
| 709 | char rhs); |
| 710 | void __kmpc_atomic_fixed1u_div_rev(ident_t *id_ref, int gtid, |
| 711 | unsigned char *lhs, unsigned char rhs); |
| 712 | void __kmpc_atomic_fixed1_shl_rev(ident_t *id_ref, int gtid, char *lhs, |
| 713 | char rhs); |
| 714 | void __kmpc_atomic_fixed1_shr_rev(ident_t *id_ref, int gtid, char *lhs, |
| 715 | char rhs); |
| 716 | void __kmpc_atomic_fixed1u_shr_rev(ident_t *id_ref, int gtid, |
| 717 | unsigned char *lhs, unsigned char rhs); |
| 718 | void __kmpc_atomic_fixed2_sub_rev(ident_t *id_ref, int gtid, short *lhs, |
| 719 | short rhs); |
| 720 | void __kmpc_atomic_fixed2_div_rev(ident_t *id_ref, int gtid, short *lhs, |
| 721 | short rhs); |
| 722 | void __kmpc_atomic_fixed2u_div_rev(ident_t *id_ref, int gtid, |
| 723 | unsigned short *lhs, unsigned short rhs); |
| 724 | void __kmpc_atomic_fixed2_shl_rev(ident_t *id_ref, int gtid, short *lhs, |
| 725 | short rhs); |
| 726 | void __kmpc_atomic_fixed2_shr_rev(ident_t *id_ref, int gtid, short *lhs, |
| 727 | short rhs); |
| 728 | void __kmpc_atomic_fixed2u_shr_rev(ident_t *id_ref, int gtid, |
| 729 | unsigned short *lhs, unsigned short rhs); |
| 730 | void __kmpc_atomic_fixed4_sub_rev(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 731 | kmp_int32 rhs); |
| 732 | void __kmpc_atomic_fixed4_div_rev(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 733 | kmp_int32 rhs); |
| 734 | void __kmpc_atomic_fixed4u_div_rev(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 735 | kmp_uint32 rhs); |
| 736 | void __kmpc_atomic_fixed4_shl_rev(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 737 | kmp_int32 rhs); |
| 738 | void __kmpc_atomic_fixed4_shr_rev(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 739 | kmp_int32 rhs); |
| 740 | void __kmpc_atomic_fixed4u_shr_rev(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 741 | kmp_uint32 rhs); |
| 742 | void __kmpc_atomic_fixed8_sub_rev(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 743 | kmp_int64 rhs); |
| 744 | void __kmpc_atomic_fixed8_div_rev(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 745 | kmp_int64 rhs); |
| 746 | void __kmpc_atomic_fixed8u_div_rev(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 747 | kmp_uint64 rhs); |
| 748 | void __kmpc_atomic_fixed8_shl_rev(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 749 | kmp_int64 rhs); |
| 750 | void __kmpc_atomic_fixed8_shr_rev(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 751 | kmp_int64 rhs); |
| 752 | void __kmpc_atomic_fixed8u_shr_rev(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 753 | kmp_uint64 rhs); |
| 754 | void __kmpc_atomic_float4_sub_rev(ident_t *id_ref, int gtid, float *lhs, |
| 755 | float rhs); |
| 756 | void __kmpc_atomic_float4_div_rev(ident_t *id_ref, int gtid, float *lhs, |
| 757 | float rhs); |
| 758 | void __kmpc_atomic_float8_sub_rev(ident_t *id_ref, int gtid, double *lhs, |
| 759 | double rhs); |
| 760 | void __kmpc_atomic_float8_div_rev(ident_t *id_ref, int gtid, double *lhs, |
| 761 | double rhs); |
| 762 | void __kmpc_atomic_float10_sub_rev(ident_t *id_ref, int gtid, long double *lhs, |
| 763 | long double rhs); |
| 764 | void __kmpc_atomic_float10_div_rev(ident_t *id_ref, int gtid, long double *lhs, |
| 765 | long double rhs); |
| 766 | #if KMP_HAVE_QUAD |
| 767 | void __kmpc_atomic_float16_sub_rev(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 768 | QUAD_LEGACY rhs); |
| 769 | void __kmpc_atomic_float16_div_rev(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 770 | QUAD_LEGACY rhs); |
| 771 | #endif |
| 772 | void __kmpc_atomic_cmplx4_sub_rev(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 773 | kmp_cmplx32 rhs); |
| 774 | void __kmpc_atomic_cmplx4_div_rev(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 775 | kmp_cmplx32 rhs); |
| 776 | void __kmpc_atomic_cmplx8_sub_rev(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 777 | kmp_cmplx64 rhs); |
| 778 | void __kmpc_atomic_cmplx8_div_rev(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 779 | kmp_cmplx64 rhs); |
| 780 | void __kmpc_atomic_cmplx10_sub_rev(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 781 | kmp_cmplx80 rhs); |
| 782 | void __kmpc_atomic_cmplx10_div_rev(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 783 | kmp_cmplx80 rhs); |
| 784 | #if KMP_HAVE_QUAD |
| 785 | void __kmpc_atomic_cmplx16_sub_rev(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 786 | CPLX128_LEG rhs); |
| 787 | void __kmpc_atomic_cmplx16_div_rev(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 788 | CPLX128_LEG rhs); |
| 789 | #if (KMP_ARCH_X86) |
| 790 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 791 | void __kmpc_atomic_float16_sub_a16_rev(ident_t *id_ref, int gtid, |
| 792 | Quad_a16_t *lhs, Quad_a16_t rhs); |
| 793 | void __kmpc_atomic_float16_div_a16_rev(ident_t *id_ref, int gtid, |
| 794 | Quad_a16_t *lhs, Quad_a16_t rhs); |
| 795 | void __kmpc_atomic_cmplx16_sub_a16_rev(ident_t *id_ref, int gtid, |
| 796 | kmp_cmplx128_a16_t *lhs, |
| 797 | kmp_cmplx128_a16_t rhs); |
| 798 | void __kmpc_atomic_cmplx16_div_a16_rev(ident_t *id_ref, int gtid, |
| 799 | kmp_cmplx128_a16_t *lhs, |
| 800 | kmp_cmplx128_a16_t rhs); |
| 801 | #endif |
| 802 | #endif // KMP_HAVE_QUAD |
| 803 | |
| 804 | #endif // KMP_ARCH_X86 || KMP_ARCH_X86_64 |
| 805 | |
| 806 | // routines for mixed types |
| 807 | |
| 808 | // RHS=float8 |
| 809 | void __kmpc_atomic_fixed1_mul_float8(ident_t *id_ref, int gtid, char *lhs, |
| 810 | kmp_real64 rhs); |
| 811 | void __kmpc_atomic_fixed1_div_float8(ident_t *id_ref, int gtid, char *lhs, |
| 812 | kmp_real64 rhs); |
| 813 | void __kmpc_atomic_fixed2_mul_float8(ident_t *id_ref, int gtid, short *lhs, |
| 814 | kmp_real64 rhs); |
| 815 | void __kmpc_atomic_fixed2_div_float8(ident_t *id_ref, int gtid, short *lhs, |
| 816 | kmp_real64 rhs); |
| 817 | void __kmpc_atomic_fixed4_mul_float8(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 818 | kmp_real64 rhs); |
| 819 | void __kmpc_atomic_fixed4_div_float8(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 820 | kmp_real64 rhs); |
| 821 | void __kmpc_atomic_fixed8_mul_float8(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 822 | kmp_real64 rhs); |
| 823 | void __kmpc_atomic_fixed8_div_float8(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 824 | kmp_real64 rhs); |
| 825 | void __kmpc_atomic_float4_add_float8(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 826 | kmp_real64 rhs); |
| 827 | void __kmpc_atomic_float4_sub_float8(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 828 | kmp_real64 rhs); |
| 829 | void __kmpc_atomic_float4_mul_float8(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 830 | kmp_real64 rhs); |
| 831 | void __kmpc_atomic_float4_div_float8(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 832 | kmp_real64 rhs); |
| 833 | |
| 834 | // RHS=float16 (deprecated, to be removed when we are sure the compiler does not |
| 835 | // use them) |
| 836 | #if KMP_HAVE_QUAD |
| 837 | void __kmpc_atomic_fixed1_add_fp(ident_t *id_ref, int gtid, char *lhs, |
| 838 | _Quad rhs); |
| 839 | void __kmpc_atomic_fixed1u_add_fp(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 840 | _Quad rhs); |
| 841 | void __kmpc_atomic_fixed1_sub_fp(ident_t *id_ref, int gtid, char *lhs, |
| 842 | _Quad rhs); |
| 843 | void __kmpc_atomic_fixed1u_sub_fp(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 844 | _Quad rhs); |
| 845 | void __kmpc_atomic_fixed1_mul_fp(ident_t *id_ref, int gtid, char *lhs, |
| 846 | _Quad rhs); |
| 847 | void __kmpc_atomic_fixed1u_mul_fp(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 848 | _Quad rhs); |
| 849 | void __kmpc_atomic_fixed1_div_fp(ident_t *id_ref, int gtid, char *lhs, |
| 850 | _Quad rhs); |
| 851 | void __kmpc_atomic_fixed1u_div_fp(ident_t *id_ref, int gtid, unsigned char *lhs, |
| 852 | _Quad rhs); |
| 853 | |
| 854 | void __kmpc_atomic_fixed2_add_fp(ident_t *id_ref, int gtid, short *lhs, |
| 855 | _Quad rhs); |
| 856 | void __kmpc_atomic_fixed2u_add_fp(ident_t *id_ref, int gtid, |
| 857 | unsigned short *lhs, _Quad rhs); |
| 858 | void __kmpc_atomic_fixed2_sub_fp(ident_t *id_ref, int gtid, short *lhs, |
| 859 | _Quad rhs); |
| 860 | void __kmpc_atomic_fixed2u_sub_fp(ident_t *id_ref, int gtid, |
| 861 | unsigned short *lhs, _Quad rhs); |
| 862 | void __kmpc_atomic_fixed2_mul_fp(ident_t *id_ref, int gtid, short *lhs, |
| 863 | _Quad rhs); |
| 864 | void __kmpc_atomic_fixed2u_mul_fp(ident_t *id_ref, int gtid, |
| 865 | unsigned short *lhs, _Quad rhs); |
| 866 | void __kmpc_atomic_fixed2_div_fp(ident_t *id_ref, int gtid, short *lhs, |
| 867 | _Quad rhs); |
| 868 | void __kmpc_atomic_fixed2u_div_fp(ident_t *id_ref, int gtid, |
| 869 | unsigned short *lhs, _Quad rhs); |
| 870 | |
| 871 | void __kmpc_atomic_fixed4_add_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 872 | _Quad rhs); |
| 873 | void __kmpc_atomic_fixed4u_add_fp(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 874 | _Quad rhs); |
| 875 | void __kmpc_atomic_fixed4_sub_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 876 | _Quad rhs); |
| 877 | void __kmpc_atomic_fixed4u_sub_fp(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 878 | _Quad rhs); |
| 879 | void __kmpc_atomic_fixed4_mul_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 880 | _Quad rhs); |
| 881 | void __kmpc_atomic_fixed4u_mul_fp(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 882 | _Quad rhs); |
| 883 | void __kmpc_atomic_fixed4_div_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 884 | _Quad rhs); |
| 885 | void __kmpc_atomic_fixed4u_div_fp(ident_t *id_ref, int gtid, kmp_uint32 *lhs, |
| 886 | _Quad rhs); |
| 887 | |
| 888 | void __kmpc_atomic_fixed8_add_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 889 | _Quad rhs); |
| 890 | void __kmpc_atomic_fixed8u_add_fp(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 891 | _Quad rhs); |
| 892 | void __kmpc_atomic_fixed8_sub_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 893 | _Quad rhs); |
| 894 | void __kmpc_atomic_fixed8u_sub_fp(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 895 | _Quad rhs); |
| 896 | void __kmpc_atomic_fixed8_mul_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 897 | _Quad rhs); |
| 898 | void __kmpc_atomic_fixed8u_mul_fp(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 899 | _Quad rhs); |
| 900 | void __kmpc_atomic_fixed8_div_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 901 | _Quad rhs); |
| 902 | void __kmpc_atomic_fixed8u_div_fp(ident_t *id_ref, int gtid, kmp_uint64 *lhs, |
| 903 | _Quad rhs); |
| 904 | |
| 905 | void __kmpc_atomic_float4_add_fp(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 906 | _Quad rhs); |
| 907 | void __kmpc_atomic_float4_sub_fp(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 908 | _Quad rhs); |
| 909 | void __kmpc_atomic_float4_mul_fp(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 910 | _Quad rhs); |
| 911 | void __kmpc_atomic_float4_div_fp(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 912 | _Quad rhs); |
| 913 | |
| 914 | void __kmpc_atomic_float8_add_fp(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 915 | _Quad rhs); |
| 916 | void __kmpc_atomic_float8_sub_fp(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 917 | _Quad rhs); |
| 918 | void __kmpc_atomic_float8_mul_fp(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 919 | _Quad rhs); |
| 920 | void __kmpc_atomic_float8_div_fp(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 921 | _Quad rhs); |
| 922 | |
| 923 | void __kmpc_atomic_float10_add_fp(ident_t *id_ref, int gtid, long double *lhs, |
| 924 | _Quad rhs); |
| 925 | void __kmpc_atomic_float10_sub_fp(ident_t *id_ref, int gtid, long double *lhs, |
| 926 | _Quad rhs); |
| 927 | void __kmpc_atomic_float10_mul_fp(ident_t *id_ref, int gtid, long double *lhs, |
| 928 | _Quad rhs); |
| 929 | void __kmpc_atomic_float10_div_fp(ident_t *id_ref, int gtid, long double *lhs, |
| 930 | _Quad rhs); |
| 931 | |
| 932 | // Reverse operations |
| 933 | void __kmpc_atomic_fixed1_sub_rev_fp(ident_t *id_ref, int gtid, char *lhs, |
| 934 | _Quad rhs); |
| 935 | void __kmpc_atomic_fixed1u_sub_rev_fp(ident_t *id_ref, int gtid, |
| 936 | unsigned char *lhs, _Quad rhs); |
| 937 | void __kmpc_atomic_fixed1_div_rev_fp(ident_t *id_ref, int gtid, char *lhs, |
| 938 | _Quad rhs); |
| 939 | void __kmpc_atomic_fixed1u_div_rev_fp(ident_t *id_ref, int gtid, |
| 940 | unsigned char *lhs, _Quad rhs); |
| 941 | void __kmpc_atomic_fixed2_sub_rev_fp(ident_t *id_ref, int gtid, short *lhs, |
| 942 | _Quad rhs); |
| 943 | void __kmpc_atomic_fixed2u_sub_rev_fp(ident_t *id_ref, int gtid, |
| 944 | unsigned short *lhs, _Quad rhs); |
| 945 | void __kmpc_atomic_fixed2_div_rev_fp(ident_t *id_ref, int gtid, short *lhs, |
| 946 | _Quad rhs); |
| 947 | void __kmpc_atomic_fixed2u_div_rev_fp(ident_t *id_ref, int gtid, |
| 948 | unsigned short *lhs, _Quad rhs); |
| 949 | void __kmpc_atomic_fixed4_sub_rev_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 950 | _Quad rhs); |
| 951 | void __kmpc_atomic_fixed4u_sub_rev_fp(ident_t *id_ref, int gtid, |
| 952 | kmp_uint32 *lhs, _Quad rhs); |
| 953 | void __kmpc_atomic_fixed4_div_rev_fp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 954 | _Quad rhs); |
| 955 | void __kmpc_atomic_fixed4u_div_rev_fp(ident_t *id_ref, int gtid, |
| 956 | kmp_uint32 *lhs, _Quad rhs); |
| 957 | void __kmpc_atomic_fixed8_sub_rev_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 958 | _Quad rhs); |
| 959 | void __kmpc_atomic_fixed8u_sub_rev_fp(ident_t *id_ref, int gtid, |
| 960 | kmp_uint64 *lhs, _Quad rhs); |
| 961 | void __kmpc_atomic_fixed8_div_rev_fp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 962 | _Quad rhs); |
| 963 | void __kmpc_atomic_fixed8u_div_rev_fp(ident_t *id_ref, int gtid, |
| 964 | kmp_uint64 *lhs, _Quad rhs); |
| 965 | void __kmpc_atomic_float4_sub_rev_fp(ident_t *id_ref, int gtid, float *lhs, |
| 966 | _Quad rhs); |
| 967 | void __kmpc_atomic_float4_div_rev_fp(ident_t *id_ref, int gtid, float *lhs, |
| 968 | _Quad rhs); |
| 969 | void __kmpc_atomic_float8_sub_rev_fp(ident_t *id_ref, int gtid, double *lhs, |
| 970 | _Quad rhs); |
| 971 | void __kmpc_atomic_float8_div_rev_fp(ident_t *id_ref, int gtid, double *lhs, |
| 972 | _Quad rhs); |
| 973 | void __kmpc_atomic_float10_sub_rev_fp(ident_t *id_ref, int gtid, |
| 974 | long double *lhs, _Quad rhs); |
| 975 | void __kmpc_atomic_float10_div_rev_fp(ident_t *id_ref, int gtid, |
| 976 | long double *lhs, _Quad rhs); |
| 977 | |
| 978 | #endif // KMP_HAVE_QUAD |
| 979 | |
| 980 | // RHS=cmplx8 |
| 981 | void __kmpc_atomic_cmplx4_add_cmplx8(ident_t *id_ref, int gtid, |
| 982 | kmp_cmplx32 *lhs, kmp_cmplx64 rhs); |
| 983 | void __kmpc_atomic_cmplx4_sub_cmplx8(ident_t *id_ref, int gtid, |
| 984 | kmp_cmplx32 *lhs, kmp_cmplx64 rhs); |
| 985 | void __kmpc_atomic_cmplx4_mul_cmplx8(ident_t *id_ref, int gtid, |
| 986 | kmp_cmplx32 *lhs, kmp_cmplx64 rhs); |
| 987 | void __kmpc_atomic_cmplx4_div_cmplx8(ident_t *id_ref, int gtid, |
| 988 | kmp_cmplx32 *lhs, kmp_cmplx64 rhs); |
| 989 | |
| 990 | // generic atomic routines |
| 991 | void __kmpc_atomic_1(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 992 | void (*f)(void *, void *, void *)); |
| 993 | void __kmpc_atomic_2(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 994 | void (*f)(void *, void *, void *)); |
| 995 | void __kmpc_atomic_4(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 996 | void (*f)(void *, void *, void *)); |
| 997 | void __kmpc_atomic_8(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 998 | void (*f)(void *, void *, void *)); |
| 999 | void __kmpc_atomic_10(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 1000 | void (*f)(void *, void *, void *)); |
| 1001 | void __kmpc_atomic_16(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 1002 | void (*f)(void *, void *, void *)); |
| 1003 | void __kmpc_atomic_20(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 1004 | void (*f)(void *, void *, void *)); |
| 1005 | void __kmpc_atomic_32(ident_t *id_ref, int gtid, void *lhs, void *rhs, |
| 1006 | void (*f)(void *, void *, void *)); |
| 1007 | |
| 1008 | // READ, WRITE, CAPTURE |
| 1009 | |
| 1010 | // Below routines for atomic READ are listed |
| 1011 | char __kmpc_atomic_fixed1_rd(ident_t *id_ref, int gtid, char *loc); |
| 1012 | short __kmpc_atomic_fixed2_rd(ident_t *id_ref, int gtid, short *loc); |
| 1013 | kmp_int32 __kmpc_atomic_fixed4_rd(ident_t *id_ref, int gtid, kmp_int32 *loc); |
| 1014 | kmp_int64 __kmpc_atomic_fixed8_rd(ident_t *id_ref, int gtid, kmp_int64 *loc); |
| 1015 | kmp_real32 __kmpc_atomic_float4_rd(ident_t *id_ref, int gtid, kmp_real32 *loc); |
| 1016 | kmp_real64 __kmpc_atomic_float8_rd(ident_t *id_ref, int gtid, kmp_real64 *loc); |
| 1017 | long double __kmpc_atomic_float10_rd(ident_t *id_ref, int gtid, |
| 1018 | long double *loc); |
| 1019 | #if KMP_HAVE_QUAD |
| 1020 | QUAD_LEGACY __kmpc_atomic_float16_rd(ident_t *id_ref, int gtid, |
| 1021 | QUAD_LEGACY *loc); |
| 1022 | #endif |
| 1023 | // Fix for CQ220361: cmplx4 READ will return void on Windows* OS; read value |
| 1024 | // will be returned through an additional parameter |
| 1025 | #if (KMP_OS_WINDOWS) |
| 1026 | void __kmpc_atomic_cmplx4_rd(kmp_cmplx32 *out, ident_t *id_ref, int gtid, |
| 1027 | kmp_cmplx32 *loc); |
| 1028 | #else |
| 1029 | kmp_cmplx32 __kmpc_atomic_cmplx4_rd(ident_t *id_ref, int gtid, |
| 1030 | kmp_cmplx32 *loc); |
| 1031 | #endif |
| 1032 | kmp_cmplx64 __kmpc_atomic_cmplx8_rd(ident_t *id_ref, int gtid, |
| 1033 | kmp_cmplx64 *loc); |
| 1034 | kmp_cmplx80 __kmpc_atomic_cmplx10_rd(ident_t *id_ref, int gtid, |
| 1035 | kmp_cmplx80 *loc); |
| 1036 | #if KMP_HAVE_QUAD |
| 1037 | CPLX128_LEG __kmpc_atomic_cmplx16_rd(ident_t *id_ref, int gtid, |
| 1038 | CPLX128_LEG *loc); |
| 1039 | #if (KMP_ARCH_X86) |
| 1040 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 1041 | Quad_a16_t __kmpc_atomic_float16_a16_rd(ident_t *id_ref, int gtid, |
| 1042 | Quad_a16_t *loc); |
| 1043 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_a16_rd(ident_t *id_ref, int gtid, |
| 1044 | kmp_cmplx128_a16_t *loc); |
| 1045 | #endif |
| 1046 | #endif |
| 1047 | |
| 1048 | // Below routines for atomic WRITE are listed |
| 1049 | void __kmpc_atomic_fixed1_wr(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 1050 | void __kmpc_atomic_fixed2_wr(ident_t *id_ref, int gtid, short *lhs, short rhs); |
| 1051 | void __kmpc_atomic_fixed4_wr(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 1052 | kmp_int32 rhs); |
| 1053 | void __kmpc_atomic_fixed8_wr(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 1054 | kmp_int64 rhs); |
| 1055 | void __kmpc_atomic_float4_wr(ident_t *id_ref, int gtid, kmp_real32 *lhs, |
| 1056 | kmp_real32 rhs); |
| 1057 | void __kmpc_atomic_float8_wr(ident_t *id_ref, int gtid, kmp_real64 *lhs, |
| 1058 | kmp_real64 rhs); |
| 1059 | void __kmpc_atomic_float10_wr(ident_t *id_ref, int gtid, long double *lhs, |
| 1060 | long double rhs); |
| 1061 | #if KMP_HAVE_QUAD |
| 1062 | void __kmpc_atomic_float16_wr(ident_t *id_ref, int gtid, QUAD_LEGACY *lhs, |
| 1063 | QUAD_LEGACY rhs); |
| 1064 | #endif |
| 1065 | void __kmpc_atomic_cmplx4_wr(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1066 | kmp_cmplx32 rhs); |
| 1067 | void __kmpc_atomic_cmplx8_wr(ident_t *id_ref, int gtid, kmp_cmplx64 *lhs, |
| 1068 | kmp_cmplx64 rhs); |
| 1069 | void __kmpc_atomic_cmplx10_wr(ident_t *id_ref, int gtid, kmp_cmplx80 *lhs, |
| 1070 | kmp_cmplx80 rhs); |
| 1071 | #if KMP_HAVE_QUAD |
| 1072 | void __kmpc_atomic_cmplx16_wr(ident_t *id_ref, int gtid, CPLX128_LEG *lhs, |
| 1073 | CPLX128_LEG rhs); |
| 1074 | #if (KMP_ARCH_X86) |
| 1075 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 1076 | void __kmpc_atomic_float16_a16_wr(ident_t *id_ref, int gtid, Quad_a16_t *lhs, |
| 1077 | Quad_a16_t rhs); |
| 1078 | void __kmpc_atomic_cmplx16_a16_wr(ident_t *id_ref, int gtid, |
| 1079 | kmp_cmplx128_a16_t *lhs, |
| 1080 | kmp_cmplx128_a16_t rhs); |
| 1081 | #endif |
| 1082 | #endif |
| 1083 | |
| 1084 | // Below routines for atomic CAPTURE are listed |
| 1085 | |
| 1086 | // 1-byte |
| 1087 | char __kmpc_atomic_fixed1_add_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1088 | char rhs, int flag); |
| 1089 | char __kmpc_atomic_fixed1_andb_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1090 | char rhs, int flag); |
| 1091 | char __kmpc_atomic_fixed1_div_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1092 | char rhs, int flag); |
| 1093 | unsigned char __kmpc_atomic_fixed1u_div_cpt(ident_t *id_ref, int gtid, |
| 1094 | unsigned char *lhs, |
| 1095 | unsigned char rhs, int flag); |
| 1096 | char __kmpc_atomic_fixed1_mul_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1097 | char rhs, int flag); |
| 1098 | char __kmpc_atomic_fixed1_orb_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1099 | char rhs, int flag); |
| 1100 | char __kmpc_atomic_fixed1_shl_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1101 | char rhs, int flag); |
| 1102 | char __kmpc_atomic_fixed1_shr_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1103 | char rhs, int flag); |
| 1104 | unsigned char __kmpc_atomic_fixed1u_shr_cpt(ident_t *id_ref, int gtid, |
| 1105 | unsigned char *lhs, |
| 1106 | unsigned char rhs, int flag); |
| 1107 | char __kmpc_atomic_fixed1_sub_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1108 | char rhs, int flag); |
| 1109 | char __kmpc_atomic_fixed1_xor_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1110 | char rhs, int flag); |
| 1111 | // 2-byte |
| 1112 | short __kmpc_atomic_fixed2_add_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1113 | short rhs, int flag); |
| 1114 | short __kmpc_atomic_fixed2_andb_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1115 | short rhs, int flag); |
| 1116 | short __kmpc_atomic_fixed2_div_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1117 | short rhs, int flag); |
| 1118 | unsigned short __kmpc_atomic_fixed2u_div_cpt(ident_t *id_ref, int gtid, |
| 1119 | unsigned short *lhs, |
| 1120 | unsigned short rhs, int flag); |
| 1121 | short __kmpc_atomic_fixed2_mul_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1122 | short rhs, int flag); |
| 1123 | short __kmpc_atomic_fixed2_orb_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1124 | short rhs, int flag); |
| 1125 | short __kmpc_atomic_fixed2_shl_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1126 | short rhs, int flag); |
| 1127 | short __kmpc_atomic_fixed2_shr_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1128 | short rhs, int flag); |
| 1129 | unsigned short __kmpc_atomic_fixed2u_shr_cpt(ident_t *id_ref, int gtid, |
| 1130 | unsigned short *lhs, |
| 1131 | unsigned short rhs, int flag); |
| 1132 | short __kmpc_atomic_fixed2_sub_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1133 | short rhs, int flag); |
| 1134 | short __kmpc_atomic_fixed2_xor_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1135 | short rhs, int flag); |
| 1136 | // 4-byte add / sub fixed |
| 1137 | kmp_int32 __kmpc_atomic_fixed4_add_cpt(ident_t *id_ref, int gtid, |
| 1138 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1139 | kmp_int32 __kmpc_atomic_fixed4_sub_cpt(ident_t *id_ref, int gtid, |
| 1140 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1141 | // 4-byte add / sub float |
| 1142 | kmp_real32 __kmpc_atomic_float4_add_cpt(ident_t *id_ref, int gtid, |
| 1143 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1144 | int flag); |
| 1145 | kmp_real32 __kmpc_atomic_float4_sub_cpt(ident_t *id_ref, int gtid, |
| 1146 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1147 | int flag); |
| 1148 | // 8-byte add / sub fixed |
| 1149 | kmp_int64 __kmpc_atomic_fixed8_add_cpt(ident_t *id_ref, int gtid, |
| 1150 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1151 | kmp_int64 __kmpc_atomic_fixed8_sub_cpt(ident_t *id_ref, int gtid, |
| 1152 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1153 | // 8-byte add / sub float |
| 1154 | kmp_real64 __kmpc_atomic_float8_add_cpt(ident_t *id_ref, int gtid, |
| 1155 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1156 | int flag); |
| 1157 | kmp_real64 __kmpc_atomic_float8_sub_cpt(ident_t *id_ref, int gtid, |
| 1158 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1159 | int flag); |
| 1160 | // 4-byte fixed |
| 1161 | kmp_int32 __kmpc_atomic_fixed4_andb_cpt(ident_t *id_ref, int gtid, |
| 1162 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1163 | int flag); |
| 1164 | kmp_int32 __kmpc_atomic_fixed4_div_cpt(ident_t *id_ref, int gtid, |
| 1165 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1166 | kmp_uint32 __kmpc_atomic_fixed4u_div_cpt(ident_t *id_ref, int gtid, |
| 1167 | kmp_uint32 *lhs, kmp_uint32 rhs, |
| 1168 | int flag); |
| 1169 | kmp_int32 __kmpc_atomic_fixed4_mul_cpt(ident_t *id_ref, int gtid, |
| 1170 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1171 | kmp_int32 __kmpc_atomic_fixed4_orb_cpt(ident_t *id_ref, int gtid, |
| 1172 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1173 | kmp_int32 __kmpc_atomic_fixed4_shl_cpt(ident_t *id_ref, int gtid, |
| 1174 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1175 | kmp_int32 __kmpc_atomic_fixed4_shr_cpt(ident_t *id_ref, int gtid, |
| 1176 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1177 | kmp_uint32 __kmpc_atomic_fixed4u_shr_cpt(ident_t *id_ref, int gtid, |
| 1178 | kmp_uint32 *lhs, kmp_uint32 rhs, |
| 1179 | int flag); |
| 1180 | kmp_int32 __kmpc_atomic_fixed4_xor_cpt(ident_t *id_ref, int gtid, |
| 1181 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1182 | // 8-byte fixed |
| 1183 | kmp_int64 __kmpc_atomic_fixed8_andb_cpt(ident_t *id_ref, int gtid, |
| 1184 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1185 | int flag); |
| 1186 | kmp_int64 __kmpc_atomic_fixed8_div_cpt(ident_t *id_ref, int gtid, |
| 1187 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1188 | kmp_uint64 __kmpc_atomic_fixed8u_div_cpt(ident_t *id_ref, int gtid, |
| 1189 | kmp_uint64 *lhs, kmp_uint64 rhs, |
| 1190 | int flag); |
| 1191 | kmp_int64 __kmpc_atomic_fixed8_mul_cpt(ident_t *id_ref, int gtid, |
| 1192 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1193 | kmp_int64 __kmpc_atomic_fixed8_orb_cpt(ident_t *id_ref, int gtid, |
| 1194 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1195 | kmp_int64 __kmpc_atomic_fixed8_shl_cpt(ident_t *id_ref, int gtid, |
| 1196 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1197 | kmp_int64 __kmpc_atomic_fixed8_shr_cpt(ident_t *id_ref, int gtid, |
| 1198 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1199 | kmp_uint64 __kmpc_atomic_fixed8u_shr_cpt(ident_t *id_ref, int gtid, |
| 1200 | kmp_uint64 *lhs, kmp_uint64 rhs, |
| 1201 | int flag); |
| 1202 | kmp_int64 __kmpc_atomic_fixed8_xor_cpt(ident_t *id_ref, int gtid, |
| 1203 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1204 | // 4-byte float |
| 1205 | kmp_real32 __kmpc_atomic_float4_div_cpt(ident_t *id_ref, int gtid, |
| 1206 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1207 | int flag); |
| 1208 | kmp_real32 __kmpc_atomic_float4_mul_cpt(ident_t *id_ref, int gtid, |
| 1209 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1210 | int flag); |
| 1211 | // 8-byte float |
| 1212 | kmp_real64 __kmpc_atomic_float8_div_cpt(ident_t *id_ref, int gtid, |
| 1213 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1214 | int flag); |
| 1215 | kmp_real64 __kmpc_atomic_float8_mul_cpt(ident_t *id_ref, int gtid, |
| 1216 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1217 | int flag); |
| 1218 | // 1-, 2-, 4-, 8-byte logical (&&, ||) |
| 1219 | char __kmpc_atomic_fixed1_andl_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1220 | char rhs, int flag); |
| 1221 | char __kmpc_atomic_fixed1_orl_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1222 | char rhs, int flag); |
| 1223 | short __kmpc_atomic_fixed2_andl_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1224 | short rhs, int flag); |
| 1225 | short __kmpc_atomic_fixed2_orl_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1226 | short rhs, int flag); |
| 1227 | kmp_int32 __kmpc_atomic_fixed4_andl_cpt(ident_t *id_ref, int gtid, |
| 1228 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1229 | int flag); |
| 1230 | kmp_int32 __kmpc_atomic_fixed4_orl_cpt(ident_t *id_ref, int gtid, |
| 1231 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1232 | kmp_int64 __kmpc_atomic_fixed8_andl_cpt(ident_t *id_ref, int gtid, |
| 1233 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1234 | int flag); |
| 1235 | kmp_int64 __kmpc_atomic_fixed8_orl_cpt(ident_t *id_ref, int gtid, |
| 1236 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1237 | // MIN / MAX |
| 1238 | char __kmpc_atomic_fixed1_max_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1239 | char rhs, int flag); |
| 1240 | char __kmpc_atomic_fixed1_min_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1241 | char rhs, int flag); |
| 1242 | short __kmpc_atomic_fixed2_max_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1243 | short rhs, int flag); |
| 1244 | short __kmpc_atomic_fixed2_min_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1245 | short rhs, int flag); |
| 1246 | kmp_int32 __kmpc_atomic_fixed4_max_cpt(ident_t *id_ref, int gtid, |
| 1247 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1248 | kmp_int32 __kmpc_atomic_fixed4_min_cpt(ident_t *id_ref, int gtid, |
| 1249 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1250 | kmp_int64 __kmpc_atomic_fixed8_max_cpt(ident_t *id_ref, int gtid, |
| 1251 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1252 | kmp_int64 __kmpc_atomic_fixed8_min_cpt(ident_t *id_ref, int gtid, |
| 1253 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1254 | kmp_real32 __kmpc_atomic_float4_max_cpt(ident_t *id_ref, int gtid, |
| 1255 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1256 | int flag); |
| 1257 | kmp_real32 __kmpc_atomic_float4_min_cpt(ident_t *id_ref, int gtid, |
| 1258 | kmp_real32 *lhs, kmp_real32 rhs, |
| 1259 | int flag); |
| 1260 | kmp_real64 __kmpc_atomic_float8_max_cpt(ident_t *id_ref, int gtid, |
| 1261 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1262 | int flag); |
| 1263 | kmp_real64 __kmpc_atomic_float8_min_cpt(ident_t *id_ref, int gtid, |
| 1264 | kmp_real64 *lhs, kmp_real64 rhs, |
| 1265 | int flag); |
| 1266 | long double __kmpc_atomic_float10_max_cpt(ident_t *id_ref, int gtid, |
| 1267 | long double *lhs, long double rhs, |
| 1268 | int flag); |
| 1269 | long double __kmpc_atomic_float10_min_cpt(ident_t *id_ref, int gtid, |
| 1270 | long double *lhs, long double rhs, |
| 1271 | int flag); |
| 1272 | #if KMP_HAVE_QUAD |
| 1273 | QUAD_LEGACY __kmpc_atomic_float16_max_cpt(ident_t *id_ref, int gtid, |
| 1274 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1275 | int flag); |
| 1276 | QUAD_LEGACY __kmpc_atomic_float16_min_cpt(ident_t *id_ref, int gtid, |
| 1277 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1278 | int flag); |
| 1279 | #endif |
| 1280 | // .NEQV. (same as xor) |
| 1281 | char __kmpc_atomic_fixed1_neqv_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1282 | char rhs, int flag); |
| 1283 | short __kmpc_atomic_fixed2_neqv_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1284 | short rhs, int flag); |
| 1285 | kmp_int32 __kmpc_atomic_fixed4_neqv_cpt(ident_t *id_ref, int gtid, |
| 1286 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1287 | int flag); |
| 1288 | kmp_int64 __kmpc_atomic_fixed8_neqv_cpt(ident_t *id_ref, int gtid, |
| 1289 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1290 | int flag); |
| 1291 | // .EQV. (same as ~xor) |
| 1292 | char __kmpc_atomic_fixed1_eqv_cpt(ident_t *id_ref, int gtid, char *lhs, |
| 1293 | char rhs, int flag); |
| 1294 | short __kmpc_atomic_fixed2_eqv_cpt(ident_t *id_ref, int gtid, short *lhs, |
| 1295 | short rhs, int flag); |
| 1296 | kmp_int32 __kmpc_atomic_fixed4_eqv_cpt(ident_t *id_ref, int gtid, |
| 1297 | kmp_int32 *lhs, kmp_int32 rhs, int flag); |
| 1298 | kmp_int64 __kmpc_atomic_fixed8_eqv_cpt(ident_t *id_ref, int gtid, |
| 1299 | kmp_int64 *lhs, kmp_int64 rhs, int flag); |
| 1300 | // long double type |
| 1301 | long double __kmpc_atomic_float10_add_cpt(ident_t *id_ref, int gtid, |
| 1302 | long double *lhs, long double rhs, |
| 1303 | int flag); |
| 1304 | long double __kmpc_atomic_float10_sub_cpt(ident_t *id_ref, int gtid, |
| 1305 | long double *lhs, long double rhs, |
| 1306 | int flag); |
| 1307 | long double __kmpc_atomic_float10_mul_cpt(ident_t *id_ref, int gtid, |
| 1308 | long double *lhs, long double rhs, |
| 1309 | int flag); |
| 1310 | long double __kmpc_atomic_float10_div_cpt(ident_t *id_ref, int gtid, |
| 1311 | long double *lhs, long double rhs, |
| 1312 | int flag); |
| 1313 | #if KMP_HAVE_QUAD |
| 1314 | // _Quad type |
| 1315 | QUAD_LEGACY __kmpc_atomic_float16_add_cpt(ident_t *id_ref, int gtid, |
| 1316 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1317 | int flag); |
| 1318 | QUAD_LEGACY __kmpc_atomic_float16_sub_cpt(ident_t *id_ref, int gtid, |
| 1319 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1320 | int flag); |
| 1321 | QUAD_LEGACY __kmpc_atomic_float16_mul_cpt(ident_t *id_ref, int gtid, |
| 1322 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1323 | int flag); |
| 1324 | QUAD_LEGACY __kmpc_atomic_float16_div_cpt(ident_t *id_ref, int gtid, |
| 1325 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1326 | int flag); |
| 1327 | #endif |
| 1328 | // routines for complex types |
| 1329 | // Workaround for cmplx4 routines - return void; captured value is returned via |
| 1330 | // the argument |
| 1331 | void __kmpc_atomic_cmplx4_add_cpt(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1332 | kmp_cmplx32 rhs, kmp_cmplx32 *out, int flag); |
| 1333 | void __kmpc_atomic_cmplx4_sub_cpt(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1334 | kmp_cmplx32 rhs, kmp_cmplx32 *out, int flag); |
| 1335 | void __kmpc_atomic_cmplx4_mul_cpt(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1336 | kmp_cmplx32 rhs, kmp_cmplx32 *out, int flag); |
| 1337 | void __kmpc_atomic_cmplx4_div_cpt(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1338 | kmp_cmplx32 rhs, kmp_cmplx32 *out, int flag); |
| 1339 | kmp_cmplx64 __kmpc_atomic_cmplx8_add_cpt(ident_t *id_ref, int gtid, |
| 1340 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1341 | int flag); |
| 1342 | kmp_cmplx64 __kmpc_atomic_cmplx8_sub_cpt(ident_t *id_ref, int gtid, |
| 1343 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1344 | int flag); |
| 1345 | kmp_cmplx64 __kmpc_atomic_cmplx8_mul_cpt(ident_t *id_ref, int gtid, |
| 1346 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1347 | int flag); |
| 1348 | kmp_cmplx64 __kmpc_atomic_cmplx8_div_cpt(ident_t *id_ref, int gtid, |
| 1349 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1350 | int flag); |
| 1351 | kmp_cmplx80 __kmpc_atomic_cmplx10_add_cpt(ident_t *id_ref, int gtid, |
| 1352 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1353 | int flag); |
| 1354 | kmp_cmplx80 __kmpc_atomic_cmplx10_sub_cpt(ident_t *id_ref, int gtid, |
| 1355 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1356 | int flag); |
| 1357 | kmp_cmplx80 __kmpc_atomic_cmplx10_mul_cpt(ident_t *id_ref, int gtid, |
| 1358 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1359 | int flag); |
| 1360 | kmp_cmplx80 __kmpc_atomic_cmplx10_div_cpt(ident_t *id_ref, int gtid, |
| 1361 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1362 | int flag); |
| 1363 | #if KMP_HAVE_QUAD |
| 1364 | CPLX128_LEG __kmpc_atomic_cmplx16_add_cpt(ident_t *id_ref, int gtid, |
| 1365 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1366 | int flag); |
| 1367 | CPLX128_LEG __kmpc_atomic_cmplx16_sub_cpt(ident_t *id_ref, int gtid, |
| 1368 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1369 | int flag); |
| 1370 | CPLX128_LEG __kmpc_atomic_cmplx16_mul_cpt(ident_t *id_ref, int gtid, |
| 1371 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1372 | int flag); |
| 1373 | CPLX128_LEG __kmpc_atomic_cmplx16_div_cpt(ident_t *id_ref, int gtid, |
| 1374 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1375 | int flag); |
| 1376 | #if (KMP_ARCH_X86) |
| 1377 | // Routines with 16-byte arguments aligned to 16-byte boundary |
| 1378 | Quad_a16_t __kmpc_atomic_float16_add_a16_cpt(ident_t *id_ref, int gtid, |
| 1379 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1380 | int flag); |
| 1381 | Quad_a16_t __kmpc_atomic_float16_sub_a16_cpt(ident_t *id_ref, int gtid, |
| 1382 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1383 | int flag); |
| 1384 | Quad_a16_t __kmpc_atomic_float16_mul_a16_cpt(ident_t *id_ref, int gtid, |
| 1385 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1386 | int flag); |
| 1387 | Quad_a16_t __kmpc_atomic_float16_div_a16_cpt(ident_t *id_ref, int gtid, |
| 1388 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1389 | int flag); |
| 1390 | Quad_a16_t __kmpc_atomic_float16_max_a16_cpt(ident_t *id_ref, int gtid, |
| 1391 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1392 | int flag); |
| 1393 | Quad_a16_t __kmpc_atomic_float16_min_a16_cpt(ident_t *id_ref, int gtid, |
| 1394 | Quad_a16_t *lhs, Quad_a16_t rhs, |
| 1395 | int flag); |
| 1396 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_add_a16_cpt(ident_t *id_ref, int gtid, |
| 1397 | kmp_cmplx128_a16_t *lhs, |
| 1398 | kmp_cmplx128_a16_t rhs, |
| 1399 | int flag); |
| 1400 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_sub_a16_cpt(ident_t *id_ref, int gtid, |
| 1401 | kmp_cmplx128_a16_t *lhs, |
| 1402 | kmp_cmplx128_a16_t rhs, |
| 1403 | int flag); |
| 1404 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_mul_a16_cpt(ident_t *id_ref, int gtid, |
| 1405 | kmp_cmplx128_a16_t *lhs, |
| 1406 | kmp_cmplx128_a16_t rhs, |
| 1407 | int flag); |
| 1408 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_div_a16_cpt(ident_t *id_ref, int gtid, |
| 1409 | kmp_cmplx128_a16_t *lhs, |
| 1410 | kmp_cmplx128_a16_t rhs, |
| 1411 | int flag); |
| 1412 | #endif |
| 1413 | #endif |
| 1414 | |
| 1415 | void __kmpc_atomic_start(void); |
| 1416 | void __kmpc_atomic_end(void); |
| 1417 | |
| 1418 | // OpenMP 4.0: v = x = expr binop x; { v = x; x = expr binop x; } { x = expr |
| 1419 | // binop x; v = x; } for non-commutative operations. |
| 1420 | #if KMP_ARCH_X86 || KMP_ARCH_X86_64 |
| 1421 | char __kmpc_atomic_fixed1_sub_cpt_rev(ident_t *id_ref, int gtid, char *lhs, |
| 1422 | char rhs, int flag); |
| 1423 | char __kmpc_atomic_fixed1_div_cpt_rev(ident_t *id_ref, int gtid, char *lhs, |
| 1424 | char rhs, int flag); |
| 1425 | unsigned char __kmpc_atomic_fixed1u_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1426 | unsigned char *lhs, |
| 1427 | unsigned char rhs, int flag); |
| 1428 | char __kmpc_atomic_fixed1_shl_cpt_rev(ident_t *id_ref, int gtid, char *lhs, |
| 1429 | char rhs, int flag); |
| 1430 | char __kmpc_atomic_fixed1_shr_cpt_rev(ident_t *id_ref, int gtid, char *lhs, |
| 1431 | char rhs, int flag); |
| 1432 | unsigned char __kmpc_atomic_fixed1u_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1433 | unsigned char *lhs, |
| 1434 | unsigned char rhs, int flag); |
| 1435 | short __kmpc_atomic_fixed2_sub_cpt_rev(ident_t *id_ref, int gtid, short *lhs, |
| 1436 | short rhs, int flag); |
| 1437 | short __kmpc_atomic_fixed2_div_cpt_rev(ident_t *id_ref, int gtid, short *lhs, |
| 1438 | short rhs, int flag); |
| 1439 | unsigned short __kmpc_atomic_fixed2u_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1440 | unsigned short *lhs, |
| 1441 | unsigned short rhs, int flag); |
| 1442 | short __kmpc_atomic_fixed2_shl_cpt_rev(ident_t *id_ref, int gtid, short *lhs, |
| 1443 | short rhs, int flag); |
| 1444 | short __kmpc_atomic_fixed2_shr_cpt_rev(ident_t *id_ref, int gtid, short *lhs, |
| 1445 | short rhs, int flag); |
| 1446 | unsigned short __kmpc_atomic_fixed2u_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1447 | unsigned short *lhs, |
| 1448 | unsigned short rhs, int flag); |
| 1449 | kmp_int32 __kmpc_atomic_fixed4_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1450 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1451 | int flag); |
| 1452 | kmp_int32 __kmpc_atomic_fixed4_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1453 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1454 | int flag); |
| 1455 | kmp_uint32 __kmpc_atomic_fixed4u_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1456 | kmp_uint32 *lhs, kmp_uint32 rhs, |
| 1457 | int flag); |
| 1458 | kmp_int32 __kmpc_atomic_fixed4_shl_cpt_rev(ident_t *id_ref, int gtid, |
| 1459 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1460 | int flag); |
| 1461 | kmp_int32 __kmpc_atomic_fixed4_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1462 | kmp_int32 *lhs, kmp_int32 rhs, |
| 1463 | int flag); |
| 1464 | kmp_uint32 __kmpc_atomic_fixed4u_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1465 | kmp_uint32 *lhs, kmp_uint32 rhs, |
| 1466 | int flag); |
| 1467 | kmp_int64 __kmpc_atomic_fixed8_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1468 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1469 | int flag); |
| 1470 | kmp_int64 __kmpc_atomic_fixed8_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1471 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1472 | int flag); |
| 1473 | kmp_uint64 __kmpc_atomic_fixed8u_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1474 | kmp_uint64 *lhs, kmp_uint64 rhs, |
| 1475 | int flag); |
| 1476 | kmp_int64 __kmpc_atomic_fixed8_shl_cpt_rev(ident_t *id_ref, int gtid, |
| 1477 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1478 | int flag); |
| 1479 | kmp_int64 __kmpc_atomic_fixed8_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1480 | kmp_int64 *lhs, kmp_int64 rhs, |
| 1481 | int flag); |
| 1482 | kmp_uint64 __kmpc_atomic_fixed8u_shr_cpt_rev(ident_t *id_ref, int gtid, |
| 1483 | kmp_uint64 *lhs, kmp_uint64 rhs, |
| 1484 | int flag); |
| 1485 | float __kmpc_atomic_float4_sub_cpt_rev(ident_t *id_ref, int gtid, float *lhs, |
| 1486 | float rhs, int flag); |
| 1487 | float __kmpc_atomic_float4_div_cpt_rev(ident_t *id_ref, int gtid, float *lhs, |
| 1488 | float rhs, int flag); |
| 1489 | double __kmpc_atomic_float8_sub_cpt_rev(ident_t *id_ref, int gtid, double *lhs, |
| 1490 | double rhs, int flag); |
| 1491 | double __kmpc_atomic_float8_div_cpt_rev(ident_t *id_ref, int gtid, double *lhs, |
| 1492 | double rhs, int flag); |
| 1493 | long double __kmpc_atomic_float10_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1494 | long double *lhs, long double rhs, |
| 1495 | int flag); |
| 1496 | long double __kmpc_atomic_float10_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1497 | long double *lhs, long double rhs, |
| 1498 | int flag); |
| 1499 | #if KMP_HAVE_QUAD |
| 1500 | QUAD_LEGACY __kmpc_atomic_float16_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1501 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1502 | int flag); |
| 1503 | QUAD_LEGACY __kmpc_atomic_float16_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1504 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs, |
| 1505 | int flag); |
| 1506 | #endif |
| 1507 | // Workaround for cmplx4 routines - return void; captured value is returned via |
| 1508 | // the argument |
| 1509 | void __kmpc_atomic_cmplx4_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1510 | kmp_cmplx32 *lhs, kmp_cmplx32 rhs, |
| 1511 | kmp_cmplx32 *out, int flag); |
| 1512 | void __kmpc_atomic_cmplx4_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1513 | kmp_cmplx32 *lhs, kmp_cmplx32 rhs, |
| 1514 | kmp_cmplx32 *out, int flag); |
| 1515 | kmp_cmplx64 __kmpc_atomic_cmplx8_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1516 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1517 | int flag); |
| 1518 | kmp_cmplx64 __kmpc_atomic_cmplx8_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1519 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs, |
| 1520 | int flag); |
| 1521 | kmp_cmplx80 __kmpc_atomic_cmplx10_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1522 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1523 | int flag); |
| 1524 | kmp_cmplx80 __kmpc_atomic_cmplx10_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1525 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs, |
| 1526 | int flag); |
| 1527 | #if KMP_HAVE_QUAD |
| 1528 | CPLX128_LEG __kmpc_atomic_cmplx16_sub_cpt_rev(ident_t *id_ref, int gtid, |
| 1529 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1530 | int flag); |
| 1531 | CPLX128_LEG __kmpc_atomic_cmplx16_div_cpt_rev(ident_t *id_ref, int gtid, |
| 1532 | CPLX128_LEG *lhs, CPLX128_LEG rhs, |
| 1533 | int flag); |
| 1534 | #if (KMP_ARCH_X86) |
| 1535 | Quad_a16_t __kmpc_atomic_float16_sub_a16_cpt_rev(ident_t *id_ref, int gtid, |
| 1536 | Quad_a16_t *lhs, |
| 1537 | Quad_a16_t rhs, int flag); |
| 1538 | Quad_a16_t __kmpc_atomic_float16_div_a16_cpt_rev(ident_t *id_ref, int gtid, |
| 1539 | Quad_a16_t *lhs, |
| 1540 | Quad_a16_t rhs, int flag); |
| 1541 | kmp_cmplx128_a16_t |
| 1542 | __kmpc_atomic_cmplx16_sub_a16_cpt_rev(ident_t *id_ref, int gtid, |
| 1543 | kmp_cmplx128_a16_t *lhs, |
| 1544 | kmp_cmplx128_a16_t rhs, int flag); |
| 1545 | kmp_cmplx128_a16_t |
| 1546 | __kmpc_atomic_cmplx16_div_a16_cpt_rev(ident_t *id_ref, int gtid, |
| 1547 | kmp_cmplx128_a16_t *lhs, |
| 1548 | kmp_cmplx128_a16_t rhs, int flag); |
| 1549 | #endif |
| 1550 | #endif |
| 1551 | |
| 1552 | // OpenMP 4.0 Capture-write (swap): {v = x; x = expr;} |
| 1553 | char __kmpc_atomic_fixed1_swp(ident_t *id_ref, int gtid, char *lhs, char rhs); |
| 1554 | short __kmpc_atomic_fixed2_swp(ident_t *id_ref, int gtid, short *lhs, |
| 1555 | short rhs); |
| 1556 | kmp_int32 __kmpc_atomic_fixed4_swp(ident_t *id_ref, int gtid, kmp_int32 *lhs, |
| 1557 | kmp_int32 rhs); |
| 1558 | kmp_int64 __kmpc_atomic_fixed8_swp(ident_t *id_ref, int gtid, kmp_int64 *lhs, |
| 1559 | kmp_int64 rhs); |
| 1560 | float __kmpc_atomic_float4_swp(ident_t *id_ref, int gtid, float *lhs, |
| 1561 | float rhs); |
| 1562 | double __kmpc_atomic_float8_swp(ident_t *id_ref, int gtid, double *lhs, |
| 1563 | double rhs); |
| 1564 | long double __kmpc_atomic_float10_swp(ident_t *id_ref, int gtid, |
| 1565 | long double *lhs, long double rhs); |
| 1566 | #if KMP_HAVE_QUAD |
| 1567 | QUAD_LEGACY __kmpc_atomic_float16_swp(ident_t *id_ref, int gtid, |
| 1568 | QUAD_LEGACY *lhs, QUAD_LEGACY rhs); |
| 1569 | #endif |
| 1570 | // !!! TODO: check if we need a workaround here |
| 1571 | void __kmpc_atomic_cmplx4_swp(ident_t *id_ref, int gtid, kmp_cmplx32 *lhs, |
| 1572 | kmp_cmplx32 rhs, kmp_cmplx32 *out); |
| 1573 | // kmp_cmplx32 __kmpc_atomic_cmplx4_swp( ident_t *id_ref, int gtid, |
| 1574 | // kmp_cmplx32 * lhs, kmp_cmplx32 rhs ); |
| 1575 | |
| 1576 | kmp_cmplx64 __kmpc_atomic_cmplx8_swp(ident_t *id_ref, int gtid, |
| 1577 | kmp_cmplx64 *lhs, kmp_cmplx64 rhs); |
| 1578 | kmp_cmplx80 __kmpc_atomic_cmplx10_swp(ident_t *id_ref, int gtid, |
| 1579 | kmp_cmplx80 *lhs, kmp_cmplx80 rhs); |
| 1580 | #if KMP_HAVE_QUAD |
| 1581 | CPLX128_LEG __kmpc_atomic_cmplx16_swp(ident_t *id_ref, int gtid, |
| 1582 | CPLX128_LEG *lhs, CPLX128_LEG rhs); |
| 1583 | #if (KMP_ARCH_X86) |
| 1584 | Quad_a16_t __kmpc_atomic_float16_a16_swp(ident_t *id_ref, int gtid, |
| 1585 | Quad_a16_t *lhs, Quad_a16_t rhs); |
| 1586 | kmp_cmplx128_a16_t __kmpc_atomic_cmplx16_a16_swp(ident_t *id_ref, int gtid, |
| 1587 | kmp_cmplx128_a16_t *lhs, |
| 1588 | kmp_cmplx128_a16_t rhs); |
| 1589 | #endif |
| 1590 | #endif |
| 1591 | |
| 1592 | // Capture routines for mixed types (RHS=float16) |
| 1593 | #if KMP_HAVE_QUAD |
| 1594 | |
| 1595 | char __kmpc_atomic_fixed1_add_cpt_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1596 | _Quad rhs, int flag); |
| 1597 | char __kmpc_atomic_fixed1_sub_cpt_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1598 | _Quad rhs, int flag); |
| 1599 | char __kmpc_atomic_fixed1_mul_cpt_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1600 | _Quad rhs, int flag); |
| 1601 | char __kmpc_atomic_fixed1_div_cpt_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1602 | _Quad rhs, int flag); |
| 1603 | unsigned char __kmpc_atomic_fixed1u_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1604 | unsigned char *lhs, _Quad rhs, |
| 1605 | int flag); |
| 1606 | unsigned char __kmpc_atomic_fixed1u_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1607 | unsigned char *lhs, _Quad rhs, |
| 1608 | int flag); |
| 1609 | unsigned char __kmpc_atomic_fixed1u_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1610 | unsigned char *lhs, _Quad rhs, |
| 1611 | int flag); |
| 1612 | unsigned char __kmpc_atomic_fixed1u_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1613 | unsigned char *lhs, _Quad rhs, |
| 1614 | int flag); |
| 1615 | |
| 1616 | short __kmpc_atomic_fixed2_add_cpt_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1617 | _Quad rhs, int flag); |
| 1618 | short __kmpc_atomic_fixed2_sub_cpt_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1619 | _Quad rhs, int flag); |
| 1620 | short __kmpc_atomic_fixed2_mul_cpt_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1621 | _Quad rhs, int flag); |
| 1622 | short __kmpc_atomic_fixed2_div_cpt_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1623 | _Quad rhs, int flag); |
| 1624 | unsigned short __kmpc_atomic_fixed2u_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1625 | unsigned short *lhs, _Quad rhs, |
| 1626 | int flag); |
| 1627 | unsigned short __kmpc_atomic_fixed2u_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1628 | unsigned short *lhs, _Quad rhs, |
| 1629 | int flag); |
| 1630 | unsigned short __kmpc_atomic_fixed2u_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1631 | unsigned short *lhs, _Quad rhs, |
| 1632 | int flag); |
| 1633 | unsigned short __kmpc_atomic_fixed2u_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1634 | unsigned short *lhs, _Quad rhs, |
| 1635 | int flag); |
| 1636 | |
| 1637 | kmp_int32 __kmpc_atomic_fixed4_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1638 | kmp_int32 *lhs, _Quad rhs, int flag); |
| 1639 | kmp_int32 __kmpc_atomic_fixed4_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1640 | kmp_int32 *lhs, _Quad rhs, int flag); |
| 1641 | kmp_int32 __kmpc_atomic_fixed4_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1642 | kmp_int32 *lhs, _Quad rhs, int flag); |
| 1643 | kmp_int32 __kmpc_atomic_fixed4_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1644 | kmp_int32 *lhs, _Quad rhs, int flag); |
| 1645 | kmp_uint32 __kmpc_atomic_fixed4u_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1646 | kmp_uint32 *lhs, _Quad rhs, |
| 1647 | int flag); |
| 1648 | kmp_uint32 __kmpc_atomic_fixed4u_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1649 | kmp_uint32 *lhs, _Quad rhs, |
| 1650 | int flag); |
| 1651 | kmp_uint32 __kmpc_atomic_fixed4u_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1652 | kmp_uint32 *lhs, _Quad rhs, |
| 1653 | int flag); |
| 1654 | kmp_uint32 __kmpc_atomic_fixed4u_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1655 | kmp_uint32 *lhs, _Quad rhs, |
| 1656 | int flag); |
| 1657 | |
| 1658 | kmp_int64 __kmpc_atomic_fixed8_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1659 | kmp_int64 *lhs, _Quad rhs, int flag); |
| 1660 | kmp_int64 __kmpc_atomic_fixed8_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1661 | kmp_int64 *lhs, _Quad rhs, int flag); |
| 1662 | kmp_int64 __kmpc_atomic_fixed8_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1663 | kmp_int64 *lhs, _Quad rhs, int flag); |
| 1664 | kmp_int64 __kmpc_atomic_fixed8_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1665 | kmp_int64 *lhs, _Quad rhs, int flag); |
| 1666 | kmp_uint64 __kmpc_atomic_fixed8u_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1667 | kmp_uint64 *lhs, _Quad rhs, |
| 1668 | int flag); |
| 1669 | kmp_uint64 __kmpc_atomic_fixed8u_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1670 | kmp_uint64 *lhs, _Quad rhs, |
| 1671 | int flag); |
| 1672 | kmp_uint64 __kmpc_atomic_fixed8u_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1673 | kmp_uint64 *lhs, _Quad rhs, |
| 1674 | int flag); |
| 1675 | kmp_uint64 __kmpc_atomic_fixed8u_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1676 | kmp_uint64 *lhs, _Quad rhs, |
| 1677 | int flag); |
| 1678 | |
| 1679 | float __kmpc_atomic_float4_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1680 | kmp_real32 *lhs, _Quad rhs, int flag); |
| 1681 | float __kmpc_atomic_float4_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1682 | kmp_real32 *lhs, _Quad rhs, int flag); |
| 1683 | float __kmpc_atomic_float4_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1684 | kmp_real32 *lhs, _Quad rhs, int flag); |
| 1685 | float __kmpc_atomic_float4_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1686 | kmp_real32 *lhs, _Quad rhs, int flag); |
| 1687 | |
| 1688 | double __kmpc_atomic_float8_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1689 | kmp_real64 *lhs, _Quad rhs, int flag); |
| 1690 | double __kmpc_atomic_float8_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1691 | kmp_real64 *lhs, _Quad rhs, int flag); |
| 1692 | double __kmpc_atomic_float8_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1693 | kmp_real64 *lhs, _Quad rhs, int flag); |
| 1694 | double __kmpc_atomic_float8_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1695 | kmp_real64 *lhs, _Quad rhs, int flag); |
| 1696 | |
| 1697 | long double __kmpc_atomic_float10_add_cpt_fp(ident_t *id_ref, int gtid, |
| 1698 | long double *lhs, _Quad rhs, |
| 1699 | int flag); |
| 1700 | long double __kmpc_atomic_float10_sub_cpt_fp(ident_t *id_ref, int gtid, |
| 1701 | long double *lhs, _Quad rhs, |
| 1702 | int flag); |
| 1703 | long double __kmpc_atomic_float10_mul_cpt_fp(ident_t *id_ref, int gtid, |
| 1704 | long double *lhs, _Quad rhs, |
| 1705 | int flag); |
| 1706 | long double __kmpc_atomic_float10_div_cpt_fp(ident_t *id_ref, int gtid, |
| 1707 | long double *lhs, _Quad rhs, |
| 1708 | int flag); |
| 1709 | |
| 1710 | char __kmpc_atomic_fixed1_sub_cpt_rev_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1711 | _Quad rhs, int flag); |
| 1712 | unsigned char __kmpc_atomic_fixed1u_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1713 | unsigned char *lhs, |
| 1714 | _Quad rhs, int flag); |
| 1715 | char __kmpc_atomic_fixed1_div_cpt_rev_fp(ident_t *id_ref, int gtid, char *lhs, |
| 1716 | _Quad rhs, int flag); |
| 1717 | unsigned char __kmpc_atomic_fixed1u_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1718 | unsigned char *lhs, |
| 1719 | _Quad rhs, int flag); |
| 1720 | short __kmpc_atomic_fixed2_sub_cpt_rev_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1721 | _Quad rhs, int flag); |
| 1722 | unsigned short __kmpc_atomic_fixed2u_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1723 | unsigned short *lhs, |
| 1724 | _Quad rhs, int flag); |
| 1725 | short __kmpc_atomic_fixed2_div_cpt_rev_fp(ident_t *id_ref, int gtid, short *lhs, |
| 1726 | _Quad rhs, int flag); |
| 1727 | unsigned short __kmpc_atomic_fixed2u_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1728 | unsigned short *lhs, |
| 1729 | _Quad rhs, int flag); |
| 1730 | kmp_int32 __kmpc_atomic_fixed4_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1731 | kmp_int32 *lhs, _Quad rhs, |
| 1732 | int flag); |
| 1733 | kmp_uint32 __kmpc_atomic_fixed4u_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1734 | kmp_uint32 *lhs, _Quad rhs, |
| 1735 | int flag); |
| 1736 | kmp_int32 __kmpc_atomic_fixed4_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1737 | kmp_int32 *lhs, _Quad rhs, |
| 1738 | int flag); |
| 1739 | kmp_uint32 __kmpc_atomic_fixed4u_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1740 | kmp_uint32 *lhs, _Quad rhs, |
| 1741 | int flag); |
| 1742 | kmp_int64 __kmpc_atomic_fixed8_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1743 | kmp_int64 *lhs, _Quad rhs, |
| 1744 | int flag); |
| 1745 | kmp_uint64 __kmpc_atomic_fixed8u_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1746 | kmp_uint64 *lhs, _Quad rhs, |
| 1747 | int flag); |
| 1748 | kmp_int64 __kmpc_atomic_fixed8_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1749 | kmp_int64 *lhs, _Quad rhs, |
| 1750 | int flag); |
| 1751 | kmp_uint64 __kmpc_atomic_fixed8u_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1752 | kmp_uint64 *lhs, _Quad rhs, |
| 1753 | int flag); |
| 1754 | float __kmpc_atomic_float4_sub_cpt_rev_fp(ident_t *id_ref, int gtid, float *lhs, |
| 1755 | _Quad rhs, int flag); |
| 1756 | float __kmpc_atomic_float4_div_cpt_rev_fp(ident_t *id_ref, int gtid, float *lhs, |
| 1757 | _Quad rhs, int flag); |
| 1758 | double __kmpc_atomic_float8_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1759 | double *lhs, _Quad rhs, int flag); |
| 1760 | double __kmpc_atomic_float8_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1761 | double *lhs, _Quad rhs, int flag); |
| 1762 | long double __kmpc_atomic_float10_sub_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1763 | long double *lhs, _Quad rhs, |
| 1764 | int flag); |
| 1765 | long double __kmpc_atomic_float10_div_cpt_rev_fp(ident_t *id_ref, int gtid, |
| 1766 | long double *lhs, _Quad rhs, |
| 1767 | int flag); |
| 1768 | |
| 1769 | #endif // KMP_HAVE_QUAD |
| 1770 | |
| 1771 | // End of OpenMP 4.0 capture |
| 1772 | |
| 1773 | // OpenMP 5.1 compare and swap |
| 1774 | /* |
| 1775 | __kmpc_atomic_bool_1_cas |
| 1776 | __kmpc_atomic_bool_2_cas |
| 1777 | __kmpc_atomic_bool_4_cas |
| 1778 | __kmpc_atomic_bool_8_cas |
| 1779 | __kmpc_atomic_val_1_cas |
| 1780 | __kmpc_atomic_val_2_cas |
| 1781 | __kmpc_atomic_val_4_cas |
| 1782 | __kmpc_atomic_val_8_cas |
| 1783 | __kmpc_atomic_bool_1_cas_cpt |
| 1784 | __kmpc_atomic_bool_2_cas_cpt |
| 1785 | __kmpc_atomic_bool_4_cas_cpt |
| 1786 | __kmpc_atomic_bool_8_cas_cpt |
| 1787 | __kmpc_atomic_val_1_cas_cpt |
| 1788 | __kmpc_atomic_val_2_cas_cpt |
| 1789 | __kmpc_atomic_val_4_cas_cpt |
| 1790 | __kmpc_atomic_val_8_cas_cpt |
| 1791 | */ |
| 1792 | // In all interfaces of CAS (Compare And Swap): |
| 1793 | // r is the boolean result of comparison |
| 1794 | // x is memory location to operate on |
| 1795 | // e is expected (old) value |
| 1796 | // d is desired (new) value |
| 1797 | // pv is pointer to captured value v whose location may coincide with e |
| 1798 | |
| 1799 | // { r = x == e; if(r) { x = d; } } |
| 1800 | // functions return result of comparison |
| 1801 | bool __kmpc_atomic_bool_1_cas(ident_t *loc, int gtid, char *x, char e, char d); |
| 1802 | bool __kmpc_atomic_bool_2_cas(ident_t *loc, int gtid, short *x, short e, |
| 1803 | short d); |
| 1804 | bool __kmpc_atomic_bool_4_cas(ident_t *loc, int gtid, kmp_int32 *x, kmp_int32 e, |
| 1805 | kmp_int32 d); |
| 1806 | bool __kmpc_atomic_bool_8_cas(ident_t *loc, int gtid, kmp_int64 *x, kmp_int64 e, |
| 1807 | kmp_int64 d); |
| 1808 | |
| 1809 | // { v = x; if (x == e) { x = d; } } |
| 1810 | // functions return old value |
| 1811 | char __kmpc_atomic_val_1_cas(ident_t *loc, int gtid, char *x, char e, char d); |
| 1812 | short __kmpc_atomic_val_2_cas(ident_t *loc, int gtid, short *x, short e, |
| 1813 | short d); |
| 1814 | kmp_int32 __kmpc_atomic_val_4_cas(ident_t *loc, int gtid, kmp_int32 *x, |
| 1815 | kmp_int32 e, kmp_int32 d); |
| 1816 | kmp_int64 __kmpc_atomic_val_8_cas(ident_t *loc, int gtid, kmp_int64 *x, |
| 1817 | kmp_int64 e, kmp_int64 d); |
| 1818 | |
| 1819 | // { r = x == e; if(r) { x = d; } else { v = x; } } |
| 1820 | // v gets old value if comparison failed, untouched otherwise |
| 1821 | // functions return result of comparison |
| 1822 | bool __kmpc_atomic_bool_1_cas_cpt(ident_t *loc, int gtid, char *x, char e, |
| 1823 | char d, char *pv); |
| 1824 | bool __kmpc_atomic_bool_2_cas_cpt(ident_t *loc, int gtid, short *x, short e, |
| 1825 | short d, short *pv); |
| 1826 | bool __kmpc_atomic_bool_4_cas_cpt(ident_t *loc, int gtid, kmp_int32 *x, |
| 1827 | kmp_int32 e, kmp_int32 d, kmp_int32 *pv); |
| 1828 | bool __kmpc_atomic_bool_8_cas_cpt(ident_t *loc, int gtid, kmp_int64 *x, |
| 1829 | kmp_int64 e, kmp_int64 d, kmp_int64 *pv); |
| 1830 | |
| 1831 | // { if (x == e) { x = d; }; v = x; } |
| 1832 | // v gets old value if comparison failed, new value otherwise |
| 1833 | // functions return old value |
| 1834 | char __kmpc_atomic_val_1_cas_cpt(ident_t *loc, int gtid, char *x, char e, |
| 1835 | char d, char *pv); |
| 1836 | short __kmpc_atomic_val_2_cas_cpt(ident_t *loc, int gtid, short *x, short e, |
| 1837 | short d, short *pv); |
| 1838 | kmp_int32 __kmpc_atomic_val_4_cas_cpt(ident_t *loc, int gtid, kmp_int32 *x, |
| 1839 | kmp_int32 e, kmp_int32 d, kmp_int32 *pv); |
| 1840 | kmp_int64 __kmpc_atomic_val_8_cas_cpt(ident_t *loc, int gtid, kmp_int64 *x, |
| 1841 | kmp_int64 e, kmp_int64 d, kmp_int64 *pv); |
| 1842 | |
| 1843 | // End OpenMP 5.1 compare + capture |
| 1844 | |
| 1845 | #endif // KMP_ARCH_X86 || KMP_ARCH_X86_64 |
| 1846 | |
| 1847 | /* ------------------------------------------------------------------------ */ |
| 1848 | |
| 1849 | #ifdef __cplusplus |
| 1850 | } // extern "C" |
| 1851 | #endif |
| 1852 | |
| 1853 | #endif /* KMP_ATOMIC_H */ |
| 1854 | |
| 1855 | // end of file |
| 1856 | |