libgomp.info revision 1.1.1.10
1 This is libgomp.info, produced by makeinfo version 6.8 from 2 libgomp.texi. 3 4 Copyright (C) 2006-2020 Free Software Foundation, Inc. 5 6 Permission is granted to copy, distribute and/or modify this document 7 under the terms of the GNU Free Documentation License, Version 1.3 or 8 any later version published by the Free Software Foundation; with the 9 Invariant Sections being "Funding Free Software", the Front-Cover texts 10 being (a) (see below), and with the Back-Cover Texts being (b) (see 11 below). A copy of the license is included in the section entitled "GNU 12 Free Documentation License". 13 14 (a) The FSF's Front-Cover Text is: 15 16 A GNU Manual 17 18 (b) The FSF's Back-Cover Text is: 19 20 You have freedom to copy and modify this GNU Manual, like GNU 21 software. Copies published by the Free Software Foundation raise funds 22 for GNU development. 23 INFO-DIR-SECTION GNU Libraries 24 START-INFO-DIR-ENTRY 25 * libgomp: (libgomp). GNU Offloading and Multi Processing Runtime Library. 26 END-INFO-DIR-ENTRY 27 28 This manual documents libgomp, the GNU Offloading and Multi 29 Processing Runtime library. This is the GNU implementation of the 30 OpenMP and OpenACC APIs for parallel and accelerator programming in 31 C/C++ and Fortran. 32 33 Published by the Free Software Foundation 51 Franklin Street, Fifth 34 Floor Boston, MA 02110-1301 USA 35 36 Copyright (C) 2006-2020 Free Software Foundation, Inc. 37 38 Permission is granted to copy, distribute and/or modify this document 39 under the terms of the GNU Free Documentation License, Version 1.3 or 40 any later version published by the Free Software Foundation; with the 41 Invariant Sections being "Funding Free Software", the Front-Cover texts 42 being (a) (see below), and with the Back-Cover Texts being (b) (see 43 below). A copy of the license is included in the section entitled "GNU 44 Free Documentation License". 45 46 (a) The FSF's Front-Cover Text is: 47 48 A GNU Manual 49 50 (b) The FSF's Back-Cover Text is: 51 52 You have freedom to copy and modify this GNU Manual, like GNU 53 software. Copies published by the Free Software Foundation raise funds 54 for GNU development. 55 56 57 File: libgomp.info, Node: Top, Next: Enabling OpenMP, Up: (dir) 58 59 Introduction 60 ************ 61 62 This manual documents the usage of libgomp, the GNU Offloading and Multi 63 Processing Runtime Library. This includes the GNU implementation of the 64 OpenMP (https://www.openmp.org) Application Programming Interface (API) 65 for multi-platform shared-memory parallel programming in C/C++ and 66 Fortran, and the GNU implementation of the OpenACC 67 (https://www.openacc.org) Application Programming Interface (API) for 68 offloading of code to accelerator devices in C/C++ and Fortran. 69 70 Originally, libgomp implemented the GNU OpenMP Runtime Library. 71 Based on this, support for OpenACC and offloading (both OpenACC and 72 OpenMP 4's target construct) has been added later on, and the library's 73 name changed to GNU Offloading and Multi Processing Runtime Library. 74 75 * Menu: 76 77 * Enabling OpenMP:: How to enable OpenMP for your applications. 78 * OpenMP Runtime Library Routines: Runtime Library Routines. 79 The OpenMP runtime application programming 80 interface. 81 * OpenMP Environment Variables: Environment Variables. 82 Influencing OpenMP runtime behavior with 83 environment variables. 84 * Enabling OpenACC:: How to enable OpenACC for your 85 applications. 86 * OpenACC Runtime Library Routines:: The OpenACC runtime application 87 programming interface. 88 * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with 89 environment variables. 90 * CUDA Streams Usage:: Notes on the implementation of 91 asynchronous operations. 92 * OpenACC Library Interoperability:: OpenACC library interoperability with the 93 NVIDIA CUBLAS library. 94 * OpenACC Profiling Interface:: 95 * The libgomp ABI:: Notes on the external ABI presented by libgomp. 96 * Reporting Bugs:: How to report bugs in the GNU Offloading and 97 Multi Processing Runtime Library. 98 * Copying:: GNU general public license says 99 how you can copy and share libgomp. 100 * GNU Free Documentation License:: 101 How you can copy and share this manual. 102 * Funding:: How to help assure continued work for free 103 software. 104 * Library Index:: Index of this documentation. 105 106 107 File: libgomp.info, Node: Enabling OpenMP, Next: Runtime Library Routines, Prev: Top, Up: Top 108 109 1 Enabling OpenMP 110 ***************** 111 112 To activate the OpenMP extensions for C/C++ and Fortran, the 113 compile-time flag '-fopenmp' must be specified. This enables the OpenMP 114 directive '#pragma omp' in C/C++ and '!$omp' directives in free form, 115 'c$omp', '*$omp' and '!$omp' directives in fixed form, '!$' conditional 116 compilation sentinels in free form and 'c$', '*$' and '!$' sentinels in 117 fixed form, for Fortran. The flag also arranges for automatic linking 118 of the OpenMP runtime library (*note Runtime Library Routines::). 119 120 A complete description of all OpenMP directives accepted may be found 121 in the OpenMP Application Program Interface (https://www.openmp.org) 122 manual, version 4.5. 123 124 125 File: libgomp.info, Node: Runtime Library Routines, Next: Environment Variables, Prev: Enabling OpenMP, Up: Top 126 127 2 OpenMP Runtime Library Routines 128 ********************************* 129 130 The runtime routines described here are defined by Section 3 of the 131 OpenMP specification in version 4.5. The routines are structured in 132 following three parts: 133 134 * Menu: 135 136 Control threads, processors and the parallel environment. They have C 137 linkage, and do not throw exceptions. 138 139 * omp_get_active_level:: Number of active parallel regions 140 * omp_get_ancestor_thread_num:: Ancestor thread ID 141 * omp_get_cancellation:: Whether cancellation support is enabled 142 * omp_get_default_device:: Get the default device for target regions 143 * omp_get_dynamic:: Dynamic teams setting 144 * omp_get_level:: Number of parallel regions 145 * omp_get_max_active_levels:: Maximum number of active regions 146 * omp_get_max_task_priority:: Maximum task priority value that can be set 147 * omp_get_max_threads:: Maximum number of threads of parallel region 148 * omp_get_nested:: Nested parallel regions 149 * omp_get_num_devices:: Number of target devices 150 * omp_get_num_procs:: Number of processors online 151 * omp_get_num_teams:: Number of teams 152 * omp_get_num_threads:: Size of the active team 153 * omp_get_proc_bind:: Whether theads may be moved between CPUs 154 * omp_get_schedule:: Obtain the runtime scheduling method 155 * omp_get_team_num:: Get team number 156 * omp_get_team_size:: Number of threads in a team 157 * omp_get_thread_limit:: Maximum number of threads 158 * omp_get_thread_num:: Current thread ID 159 * omp_in_parallel:: Whether a parallel region is active 160 * omp_in_final:: Whether in final or included task region 161 * omp_is_initial_device:: Whether executing on the host device 162 * omp_set_default_device:: Set the default device for target regions 163 * omp_set_dynamic:: Enable/disable dynamic teams 164 * omp_set_max_active_levels:: Limits the number of active parallel regions 165 * omp_set_nested:: Enable/disable nested parallel regions 166 * omp_set_num_threads:: Set upper team size limit 167 * omp_set_schedule:: Set the runtime scheduling method 168 169 Initialize, set, test, unset and destroy simple and nested locks. 170 171 * omp_init_lock:: Initialize simple lock 172 * omp_set_lock:: Wait for and set simple lock 173 * omp_test_lock:: Test and set simple lock if available 174 * omp_unset_lock:: Unset simple lock 175 * omp_destroy_lock:: Destroy simple lock 176 * omp_init_nest_lock:: Initialize nested lock 177 * omp_set_nest_lock:: Wait for and set simple lock 178 * omp_test_nest_lock:: Test and set nested lock if available 179 * omp_unset_nest_lock:: Unset nested lock 180 * omp_destroy_nest_lock:: Destroy nested lock 181 182 Portable, thread-based, wall clock timer. 183 184 * omp_get_wtick:: Get timer precision. 185 * omp_get_wtime:: Elapsed wall clock time. 186 187 188 File: libgomp.info, Node: omp_get_active_level, Next: omp_get_ancestor_thread_num, Up: Runtime Library Routines 189 190 2.1 'omp_get_active_level' - Number of parallel regions 191 ======================================================= 192 193 _Description_: 194 This function returns the nesting level for the active parallel 195 blocks, which enclose the calling call. 196 197 _C/C++_ 198 _Prototype_: 'int omp_get_active_level(void);' 199 200 _Fortran_: 201 _Interface_: 'integer function omp_get_active_level()' 202 203 _See also_: 204 *note omp_get_level::, *note omp_get_max_active_levels::, *note 205 omp_set_max_active_levels:: 206 207 _Reference_: 208 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.20. 209 210 211 File: libgomp.info, Node: omp_get_ancestor_thread_num, Next: omp_get_cancellation, Prev: omp_get_active_level, Up: Runtime Library Routines 212 213 2.2 'omp_get_ancestor_thread_num' - Ancestor thread ID 214 ====================================================== 215 216 _Description_: 217 This function returns the thread identification number for the 218 given nesting level of the current thread. For values of LEVEL 219 outside zero to 'omp_get_level' -1 is returned; if LEVEL is 220 'omp_get_level' the result is identical to 'omp_get_thread_num'. 221 222 _C/C++_ 223 _Prototype_: 'int omp_get_ancestor_thread_num(int level);' 224 225 _Fortran_: 226 _Interface_: 'integer function omp_get_ancestor_thread_num(level)' 227 'integer level' 228 229 _See also_: 230 *note omp_get_level::, *note omp_get_thread_num::, *note 231 omp_get_team_size:: 232 233 _Reference_: 234 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.18. 235 236 237 File: libgomp.info, Node: omp_get_cancellation, Next: omp_get_default_device, Prev: omp_get_ancestor_thread_num, Up: Runtime Library Routines 238 239 2.3 'omp_get_cancellation' - Whether cancellation support is enabled 240 ==================================================================== 241 242 _Description_: 243 This function returns 'true' if cancellation is activated, 'false' 244 otherwise. Here, 'true' and 'false' represent their 245 language-specific counterparts. Unless 'OMP_CANCELLATION' is set 246 true, cancellations are deactivated. 247 248 _C/C++_: 249 _Prototype_: 'int omp_get_cancellation(void);' 250 251 _Fortran_: 252 _Interface_: 'logical function omp_get_cancellation()' 253 254 _See also_: 255 *note OMP_CANCELLATION:: 256 257 _Reference_: 258 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.9. 259 260 261 File: libgomp.info, Node: omp_get_default_device, Next: omp_get_dynamic, Prev: omp_get_cancellation, Up: Runtime Library Routines 262 263 2.4 'omp_get_default_device' - Get the default device for target regions 264 ======================================================================== 265 266 _Description_: 267 Get the default device for target regions without device clause. 268 269 _C/C++_: 270 _Prototype_: 'int omp_get_default_device(void);' 271 272 _Fortran_: 273 _Interface_: 'integer function omp_get_default_device()' 274 275 _See also_: 276 *note OMP_DEFAULT_DEVICE::, *note omp_set_default_device:: 277 278 _Reference_: 279 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.30. 280 281 282 File: libgomp.info, Node: omp_get_dynamic, Next: omp_get_level, Prev: omp_get_default_device, Up: Runtime Library Routines 283 284 2.5 'omp_get_dynamic' - Dynamic teams setting 285 ============================================= 286 287 _Description_: 288 This function returns 'true' if enabled, 'false' otherwise. Here, 289 'true' and 'false' represent their language-specific counterparts. 290 291 The dynamic team setting may be initialized at startup by the 292 'OMP_DYNAMIC' environment variable or at runtime using 293 'omp_set_dynamic'. If undefined, dynamic adjustment is disabled by 294 default. 295 296 _C/C++_: 297 _Prototype_: 'int omp_get_dynamic(void);' 298 299 _Fortran_: 300 _Interface_: 'logical function omp_get_dynamic()' 301 302 _See also_: 303 *note omp_set_dynamic::, *note OMP_DYNAMIC:: 304 305 _Reference_: 306 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.8. 307 308 309 File: libgomp.info, Node: omp_get_level, Next: omp_get_max_active_levels, Prev: omp_get_dynamic, Up: Runtime Library Routines 310 311 2.6 'omp_get_level' - Obtain the current nesting level 312 ====================================================== 313 314 _Description_: 315 This function returns the nesting level for the parallel blocks, 316 which enclose the calling call. 317 318 _C/C++_ 319 _Prototype_: 'int omp_get_level(void);' 320 321 _Fortran_: 322 _Interface_: 'integer function omp_level()' 323 324 _See also_: 325 *note omp_get_active_level:: 326 327 _Reference_: 328 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.17. 329 330 331 File: libgomp.info, Node: omp_get_max_active_levels, Next: omp_get_max_task_priority, Prev: omp_get_level, Up: Runtime Library Routines 332 333 2.7 'omp_get_max_active_levels' - Maximum number of active regions 334 ================================================================== 335 336 _Description_: 337 This function obtains the maximum allowed number of nested, active 338 parallel regions. 339 340 _C/C++_ 341 _Prototype_: 'int omp_get_max_active_levels(void);' 342 343 _Fortran_: 344 _Interface_: 'integer function omp_get_max_active_levels()' 345 346 _See also_: 347 *note omp_set_max_active_levels::, *note omp_get_active_level:: 348 349 _Reference_: 350 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.16. 351 352 353 File: libgomp.info, Node: omp_get_max_task_priority, Next: omp_get_max_threads, Prev: omp_get_max_active_levels, Up: Runtime Library Routines 354 355 2.8 'omp_get_max_task_priority' - Maximum priority value 356 ======================================================== 357 358 that can be set for tasks. 359 _Description_: 360 This function obtains the maximum allowed priority number for 361 tasks. 362 363 _C/C++_ 364 _Prototype_: 'int omp_get_max_task_priority(void);' 365 366 _Fortran_: 367 _Interface_: 'integer function omp_get_max_task_priority()' 368 369 _Reference_: 370 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.29. 371 372 373 File: libgomp.info, Node: omp_get_max_threads, Next: omp_get_nested, Prev: omp_get_max_task_priority, Up: Runtime Library Routines 374 375 2.9 'omp_get_max_threads' - Maximum number of threads of parallel region 376 ======================================================================== 377 378 _Description_: 379 Return the maximum number of threads used for the current parallel 380 region that does not use the clause 'num_threads'. 381 382 _C/C++_: 383 _Prototype_: 'int omp_get_max_threads(void);' 384 385 _Fortran_: 386 _Interface_: 'integer function omp_get_max_threads()' 387 388 _See also_: 389 *note omp_set_num_threads::, *note omp_set_dynamic::, *note 390 omp_get_thread_limit:: 391 392 _Reference_: 393 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.3. 394 395 396 File: libgomp.info, Node: omp_get_nested, Next: omp_get_num_devices, Prev: omp_get_max_threads, Up: Runtime Library Routines 397 398 2.10 'omp_get_nested' - Nested parallel regions 399 =============================================== 400 401 _Description_: 402 This function returns 'true' if nested parallel regions are 403 enabled, 'false' otherwise. Here, 'true' and 'false' represent 404 their language-specific counterparts. 405 406 Nested parallel regions may be initialized at startup by the 407 'OMP_NESTED' environment variable or at runtime using 408 'omp_set_nested'. If undefined, nested parallel regions are 409 disabled by default. 410 411 _C/C++_: 412 _Prototype_: 'int omp_get_nested(void);' 413 414 _Fortran_: 415 _Interface_: 'logical function omp_get_nested()' 416 417 _See also_: 418 *note omp_set_nested::, *note OMP_NESTED:: 419 420 _Reference_: 421 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.11. 422 423 424 File: libgomp.info, Node: omp_get_num_devices, Next: omp_get_num_procs, Prev: omp_get_nested, Up: Runtime Library Routines 425 426 2.11 'omp_get_num_devices' - Number of target devices 427 ===================================================== 428 429 _Description_: 430 Returns the number of target devices. 431 432 _C/C++_: 433 _Prototype_: 'int omp_get_num_devices(void);' 434 435 _Fortran_: 436 _Interface_: 'integer function omp_get_num_devices()' 437 438 _Reference_: 439 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.31. 440 441 442 File: libgomp.info, Node: omp_get_num_procs, Next: omp_get_num_teams, Prev: omp_get_num_devices, Up: Runtime Library Routines 443 444 2.12 'omp_get_num_procs' - Number of processors online 445 ====================================================== 446 447 _Description_: 448 Returns the number of processors online on that device. 449 450 _C/C++_: 451 _Prototype_: 'int omp_get_num_procs(void);' 452 453 _Fortran_: 454 _Interface_: 'integer function omp_get_num_procs()' 455 456 _Reference_: 457 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.5. 458 459 460 File: libgomp.info, Node: omp_get_num_teams, Next: omp_get_num_threads, Prev: omp_get_num_procs, Up: Runtime Library Routines 461 462 2.13 'omp_get_num_teams' - Number of teams 463 ========================================== 464 465 _Description_: 466 Returns the number of teams in the current team region. 467 468 _C/C++_: 469 _Prototype_: 'int omp_get_num_teams(void);' 470 471 _Fortran_: 472 _Interface_: 'integer function omp_get_num_teams()' 473 474 _Reference_: 475 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.32. 476 477 478 File: libgomp.info, Node: omp_get_num_threads, Next: omp_get_proc_bind, Prev: omp_get_num_teams, Up: Runtime Library Routines 479 480 2.14 'omp_get_num_threads' - Size of the active team 481 ==================================================== 482 483 _Description_: 484 Returns the number of threads in the current team. In a sequential 485 section of the program 'omp_get_num_threads' returns 1. 486 487 The default team size may be initialized at startup by the 488 'OMP_NUM_THREADS' environment variable. At runtime, the size of 489 the current team may be set either by the 'NUM_THREADS' clause or 490 by 'omp_set_num_threads'. If none of the above were used to define 491 a specific value and 'OMP_DYNAMIC' is disabled, one thread per CPU 492 online is used. 493 494 _C/C++_: 495 _Prototype_: 'int omp_get_num_threads(void);' 496 497 _Fortran_: 498 _Interface_: 'integer function omp_get_num_threads()' 499 500 _See also_: 501 *note omp_get_max_threads::, *note omp_set_num_threads::, *note 502 OMP_NUM_THREADS:: 503 504 _Reference_: 505 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.2. 506 507 508 File: libgomp.info, Node: omp_get_proc_bind, Next: omp_get_schedule, Prev: omp_get_num_threads, Up: Runtime Library Routines 509 510 2.15 'omp_get_proc_bind' - Whether theads may be moved between CPUs 511 =================================================================== 512 513 _Description_: 514 This functions returns the currently active thread affinity policy, 515 which is set via 'OMP_PROC_BIND'. Possible values are 516 'omp_proc_bind_false', 'omp_proc_bind_true', 517 'omp_proc_bind_master', 'omp_proc_bind_close' and 518 'omp_proc_bind_spread'. 519 520 _C/C++_: 521 _Prototype_: 'omp_proc_bind_t omp_get_proc_bind(void);' 522 523 _Fortran_: 524 _Interface_: 'integer(kind=omp_proc_bind_kind) function 525 omp_get_proc_bind()' 526 527 _See also_: 528 *note OMP_PROC_BIND::, *note OMP_PLACES::, *note 529 GOMP_CPU_AFFINITY::, 530 531 _Reference_: 532 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.22. 533 534 535 File: libgomp.info, Node: omp_get_schedule, Next: omp_get_team_num, Prev: omp_get_proc_bind, Up: Runtime Library Routines 536 537 2.16 'omp_get_schedule' - Obtain the runtime scheduling method 538 ============================================================== 539 540 _Description_: 541 Obtain the runtime scheduling method. The KIND argument will be 542 set to the value 'omp_sched_static', 'omp_sched_dynamic', 543 'omp_sched_guided' or 'omp_sched_auto'. The second argument, 544 CHUNK_SIZE, is set to the chunk size. 545 546 _C/C++_ 547 _Prototype_: 'void omp_get_schedule(omp_sched_t *kind, int 548 *chunk_size);' 549 550 _Fortran_: 551 _Interface_: 'subroutine omp_get_schedule(kind, chunk_size)' 552 'integer(kind=omp_sched_kind) kind' 553 'integer chunk_size' 554 555 _See also_: 556 *note omp_set_schedule::, *note OMP_SCHEDULE:: 557 558 _Reference_: 559 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.13. 560 561 562 File: libgomp.info, Node: omp_get_team_num, Next: omp_get_team_size, Prev: omp_get_schedule, Up: Runtime Library Routines 563 564 2.17 'omp_get_team_num' - Get team number 565 ========================================= 566 567 _Description_: 568 Returns the team number of the calling thread. 569 570 _C/C++_: 571 _Prototype_: 'int omp_get_team_num(void);' 572 573 _Fortran_: 574 _Interface_: 'integer function omp_get_team_num()' 575 576 _Reference_: 577 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.33. 578 579 580 File: libgomp.info, Node: omp_get_team_size, Next: omp_get_thread_limit, Prev: omp_get_team_num, Up: Runtime Library Routines 581 582 2.18 'omp_get_team_size' - Number of threads in a team 583 ====================================================== 584 585 _Description_: 586 This function returns the number of threads in a thread team to 587 which either the current thread or its ancestor belongs. For 588 values of LEVEL outside zero to 'omp_get_level', -1 is returned; if 589 LEVEL is zero, 1 is returned, and for 'omp_get_level', the result 590 is identical to 'omp_get_num_threads'. 591 592 _C/C++_: 593 _Prototype_: 'int omp_get_team_size(int level);' 594 595 _Fortran_: 596 _Interface_: 'integer function omp_get_team_size(level)' 597 'integer level' 598 599 _See also_: 600 *note omp_get_num_threads::, *note omp_get_level::, *note 601 omp_get_ancestor_thread_num:: 602 603 _Reference_: 604 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.19. 605 606 607 File: libgomp.info, Node: omp_get_thread_limit, Next: omp_get_thread_num, Prev: omp_get_team_size, Up: Runtime Library Routines 608 609 2.19 'omp_get_thread_limit' - Maximum number of threads 610 ======================================================= 611 612 _Description_: 613 Return the maximum number of threads of the program. 614 615 _C/C++_: 616 _Prototype_: 'int omp_get_thread_limit(void);' 617 618 _Fortran_: 619 _Interface_: 'integer function omp_get_thread_limit()' 620 621 _See also_: 622 *note omp_get_max_threads::, *note OMP_THREAD_LIMIT:: 623 624 _Reference_: 625 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.14. 626 627 628 File: libgomp.info, Node: omp_get_thread_num, Next: omp_in_parallel, Prev: omp_get_thread_limit, Up: Runtime Library Routines 629 630 2.20 'omp_get_thread_num' - Current thread ID 631 ============================================= 632 633 _Description_: 634 Returns a unique thread identification number within the current 635 team. In a sequential parts of the program, 'omp_get_thread_num' 636 always returns 0. In parallel regions the return value varies from 637 0 to 'omp_get_num_threads'-1 inclusive. The return value of the 638 master thread of a team is always 0. 639 640 _C/C++_: 641 _Prototype_: 'int omp_get_thread_num(void);' 642 643 _Fortran_: 644 _Interface_: 'integer function omp_get_thread_num()' 645 646 _See also_: 647 *note omp_get_num_threads::, *note omp_get_ancestor_thread_num:: 648 649 _Reference_: 650 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.4. 651 652 653 File: libgomp.info, Node: omp_in_parallel, Next: omp_in_final, Prev: omp_get_thread_num, Up: Runtime Library Routines 654 655 2.21 'omp_in_parallel' - Whether a parallel region is active 656 ============================================================ 657 658 _Description_: 659 This function returns 'true' if currently running in parallel, 660 'false' otherwise. Here, 'true' and 'false' represent their 661 language-specific counterparts. 662 663 _C/C++_: 664 _Prototype_: 'int omp_in_parallel(void);' 665 666 _Fortran_: 667 _Interface_: 'logical function omp_in_parallel()' 668 669 _Reference_: 670 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.6. 671 672 673 File: libgomp.info, Node: omp_in_final, Next: omp_is_initial_device, Prev: omp_in_parallel, Up: Runtime Library Routines 674 675 2.22 'omp_in_final' - Whether in final or included task region 676 ============================================================== 677 678 _Description_: 679 This function returns 'true' if currently running in a final or 680 included task region, 'false' otherwise. Here, 'true' and 'false' 681 represent their language-specific counterparts. 682 683 _C/C++_: 684 _Prototype_: 'int omp_in_final(void);' 685 686 _Fortran_: 687 _Interface_: 'logical function omp_in_final()' 688 689 _Reference_: 690 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.21. 691 692 693 File: libgomp.info, Node: omp_is_initial_device, Next: omp_set_default_device, Prev: omp_in_final, Up: Runtime Library Routines 694 695 2.23 'omp_is_initial_device' - Whether executing on the host device 696 =================================================================== 697 698 _Description_: 699 This function returns 'true' if currently running on the host 700 device, 'false' otherwise. Here, 'true' and 'false' represent 701 their language-specific counterparts. 702 703 _C/C++_: 704 _Prototype_: 'int omp_is_initial_device(void);' 705 706 _Fortran_: 707 _Interface_: 'logical function omp_is_initial_device()' 708 709 _Reference_: 710 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.34. 711 712 713 File: libgomp.info, Node: omp_set_default_device, Next: omp_set_dynamic, Prev: omp_is_initial_device, Up: Runtime Library Routines 714 715 2.24 'omp_set_default_device' - Set the default device for target regions 716 ========================================================================= 717 718 _Description_: 719 Set the default device for target regions without device clause. 720 The argument shall be a nonnegative device number. 721 722 _C/C++_: 723 _Prototype_: 'void omp_set_default_device(int device_num);' 724 725 _Fortran_: 726 _Interface_: 'subroutine omp_set_default_device(device_num)' 727 'integer device_num' 728 729 _See also_: 730 *note OMP_DEFAULT_DEVICE::, *note omp_get_default_device:: 731 732 _Reference_: 733 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.29. 734 735 736 File: libgomp.info, Node: omp_set_dynamic, Next: omp_set_max_active_levels, Prev: omp_set_default_device, Up: Runtime Library Routines 737 738 2.25 'omp_set_dynamic' - Enable/disable dynamic teams 739 ===================================================== 740 741 _Description_: 742 Enable or disable the dynamic adjustment of the number of threads 743 within a team. The function takes the language-specific equivalent 744 of 'true' and 'false', where 'true' enables dynamic adjustment of 745 team sizes and 'false' disables it. 746 747 _C/C++_: 748 _Prototype_: 'void omp_set_dynamic(int dynamic_threads);' 749 750 _Fortran_: 751 _Interface_: 'subroutine omp_set_dynamic(dynamic_threads)' 752 'logical, intent(in) :: dynamic_threads' 753 754 _See also_: 755 *note OMP_DYNAMIC::, *note omp_get_dynamic:: 756 757 _Reference_: 758 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.7. 759 760 761 File: libgomp.info, Node: omp_set_max_active_levels, Next: omp_set_nested, Prev: omp_set_dynamic, Up: Runtime Library Routines 762 763 2.26 'omp_set_max_active_levels' - Limits the number of active parallel regions 764 =============================================================================== 765 766 _Description_: 767 This function limits the maximum allowed number of nested, active 768 parallel regions. 769 770 _C/C++_ 771 _Prototype_: 'void omp_set_max_active_levels(int max_levels);' 772 773 _Fortran_: 774 _Interface_: 'subroutine omp_set_max_active_levels(max_levels)' 775 'integer max_levels' 776 777 _See also_: 778 *note omp_get_max_active_levels::, *note omp_get_active_level:: 779 780 _Reference_: 781 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.15. 782 783 784 File: libgomp.info, Node: omp_set_nested, Next: omp_set_num_threads, Prev: omp_set_max_active_levels, Up: Runtime Library Routines 785 786 2.27 'omp_set_nested' - Enable/disable nested parallel regions 787 ============================================================== 788 789 _Description_: 790 Enable or disable nested parallel regions, i.e., whether team 791 members are allowed to create new teams. The function takes the 792 language-specific equivalent of 'true' and 'false', where 'true' 793 enables dynamic adjustment of team sizes and 'false' disables it. 794 795 _C/C++_: 796 _Prototype_: 'void omp_set_nested(int nested);' 797 798 _Fortran_: 799 _Interface_: 'subroutine omp_set_nested(nested)' 800 'logical, intent(in) :: nested' 801 802 _See also_: 803 *note OMP_NESTED::, *note omp_get_nested:: 804 805 _Reference_: 806 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.10. 807 808 809 File: libgomp.info, Node: omp_set_num_threads, Next: omp_set_schedule, Prev: omp_set_nested, Up: Runtime Library Routines 810 811 2.28 'omp_set_num_threads' - Set upper team size limit 812 ====================================================== 813 814 _Description_: 815 Specifies the number of threads used by default in subsequent 816 parallel sections, if those do not specify a 'num_threads' clause. 817 The argument of 'omp_set_num_threads' shall be a positive integer. 818 819 _C/C++_: 820 _Prototype_: 'void omp_set_num_threads(int num_threads);' 821 822 _Fortran_: 823 _Interface_: 'subroutine omp_set_num_threads(num_threads)' 824 'integer, intent(in) :: num_threads' 825 826 _See also_: 827 *note OMP_NUM_THREADS::, *note omp_get_num_threads::, *note 828 omp_get_max_threads:: 829 830 _Reference_: 831 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.1. 832 833 834 File: libgomp.info, Node: omp_set_schedule, Next: omp_init_lock, Prev: omp_set_num_threads, Up: Runtime Library Routines 835 836 2.29 'omp_set_schedule' - Set the runtime scheduling method 837 =========================================================== 838 839 _Description_: 840 Sets the runtime scheduling method. The KIND argument can have the 841 value 'omp_sched_static', 'omp_sched_dynamic', 'omp_sched_guided' 842 or 'omp_sched_auto'. Except for 'omp_sched_auto', the chunk size 843 is set to the value of CHUNK_SIZE if positive, or to the default 844 value if zero or negative. For 'omp_sched_auto' the CHUNK_SIZE 845 argument is ignored. 846 847 _C/C++_ 848 _Prototype_: 'void omp_set_schedule(omp_sched_t kind, int 849 chunk_size);' 850 851 _Fortran_: 852 _Interface_: 'subroutine omp_set_schedule(kind, chunk_size)' 853 'integer(kind=omp_sched_kind) kind' 854 'integer chunk_size' 855 856 _See also_: 857 *note omp_get_schedule:: *note OMP_SCHEDULE:: 858 859 _Reference_: 860 OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.12. 861 862 863 File: libgomp.info, Node: omp_init_lock, Next: omp_set_lock, Prev: omp_set_schedule, Up: Runtime Library Routines 864 865 2.30 'omp_init_lock' - Initialize simple lock 866 ============================================= 867 868 _Description_: 869 Initialize a simple lock. After initialization, the lock is in an 870 unlocked state. 871 872 _C/C++_: 873 _Prototype_: 'void omp_init_lock(omp_lock_t *lock);' 874 875 _Fortran_: 876 _Interface_: 'subroutine omp_init_lock(svar)' 877 'integer(omp_lock_kind), intent(out) :: svar' 878 879 _See also_: 880 *note omp_destroy_lock:: 881 882 _Reference_: 883 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.1. 884 885 886 File: libgomp.info, Node: omp_set_lock, Next: omp_test_lock, Prev: omp_init_lock, Up: Runtime Library Routines 887 888 2.31 'omp_set_lock' - Wait for and set simple lock 889 ================================================== 890 891 _Description_: 892 Before setting a simple lock, the lock variable must be initialized 893 by 'omp_init_lock'. The calling thread is blocked until the lock 894 is available. If the lock is already held by the current thread, a 895 deadlock occurs. 896 897 _C/C++_: 898 _Prototype_: 'void omp_set_lock(omp_lock_t *lock);' 899 900 _Fortran_: 901 _Interface_: 'subroutine omp_set_lock(svar)' 902 'integer(omp_lock_kind), intent(inout) :: svar' 903 904 _See also_: 905 *note omp_init_lock::, *note omp_test_lock::, *note 906 omp_unset_lock:: 907 908 _Reference_: 909 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.4. 910 911 912 File: libgomp.info, Node: omp_test_lock, Next: omp_unset_lock, Prev: omp_set_lock, Up: Runtime Library Routines 913 914 2.32 'omp_test_lock' - Test and set simple lock if available 915 ============================================================ 916 917 _Description_: 918 Before setting a simple lock, the lock variable must be initialized 919 by 'omp_init_lock'. Contrary to 'omp_set_lock', 'omp_test_lock' 920 does not block if the lock is not available. This function returns 921 'true' upon success, 'false' otherwise. Here, 'true' and 'false' 922 represent their language-specific counterparts. 923 924 _C/C++_: 925 _Prototype_: 'int omp_test_lock(omp_lock_t *lock);' 926 927 _Fortran_: 928 _Interface_: 'logical function omp_test_lock(svar)' 929 'integer(omp_lock_kind), intent(inout) :: svar' 930 931 _See also_: 932 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 933 934 _Reference_: 935 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.6. 936 937 938 File: libgomp.info, Node: omp_unset_lock, Next: omp_destroy_lock, Prev: omp_test_lock, Up: Runtime Library Routines 939 940 2.33 'omp_unset_lock' - Unset simple lock 941 ========================================= 942 943 _Description_: 944 A simple lock about to be unset must have been locked by 945 'omp_set_lock' or 'omp_test_lock' before. In addition, the lock 946 must be held by the thread calling 'omp_unset_lock'. Then, the 947 lock becomes unlocked. If one or more threads attempted to set the 948 lock before, one of them is chosen to, again, set the lock to 949 itself. 950 951 _C/C++_: 952 _Prototype_: 'void omp_unset_lock(omp_lock_t *lock);' 953 954 _Fortran_: 955 _Interface_: 'subroutine omp_unset_lock(svar)' 956 'integer(omp_lock_kind), intent(inout) :: svar' 957 958 _See also_: 959 *note omp_set_lock::, *note omp_test_lock:: 960 961 _Reference_: 962 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.5. 963 964 965 File: libgomp.info, Node: omp_destroy_lock, Next: omp_init_nest_lock, Prev: omp_unset_lock, Up: Runtime Library Routines 966 967 2.34 'omp_destroy_lock' - Destroy simple lock 968 ============================================= 969 970 _Description_: 971 Destroy a simple lock. In order to be destroyed, a simple lock 972 must be in the unlocked state. 973 974 _C/C++_: 975 _Prototype_: 'void omp_destroy_lock(omp_lock_t *lock);' 976 977 _Fortran_: 978 _Interface_: 'subroutine omp_destroy_lock(svar)' 979 'integer(omp_lock_kind), intent(inout) :: svar' 980 981 _See also_: 982 *note omp_init_lock:: 983 984 _Reference_: 985 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.3. 986 987 988 File: libgomp.info, Node: omp_init_nest_lock, Next: omp_set_nest_lock, Prev: omp_destroy_lock, Up: Runtime Library Routines 989 990 2.35 'omp_init_nest_lock' - Initialize nested lock 991 ================================================== 992 993 _Description_: 994 Initialize a nested lock. After initialization, the lock is in an 995 unlocked state and the nesting count is set to zero. 996 997 _C/C++_: 998 _Prototype_: 'void omp_init_nest_lock(omp_nest_lock_t *lock);' 999 1000 _Fortran_: 1001 _Interface_: 'subroutine omp_init_nest_lock(nvar)' 1002 'integer(omp_nest_lock_kind), intent(out) :: nvar' 1003 1004 _See also_: 1005 *note omp_destroy_nest_lock:: 1006 1007 _Reference_: 1008 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.1. 1009 1010 1011 File: libgomp.info, Node: omp_set_nest_lock, Next: omp_test_nest_lock, Prev: omp_init_nest_lock, Up: Runtime Library Routines 1012 1013 2.36 'omp_set_nest_lock' - Wait for and set nested lock 1014 ======================================================= 1015 1016 _Description_: 1017 Before setting a nested lock, the lock variable must be initialized 1018 by 'omp_init_nest_lock'. The calling thread is blocked until the 1019 lock is available. If the lock is already held by the current 1020 thread, the nesting count for the lock is incremented. 1021 1022 _C/C++_: 1023 _Prototype_: 'void omp_set_nest_lock(omp_nest_lock_t *lock);' 1024 1025 _Fortran_: 1026 _Interface_: 'subroutine omp_set_nest_lock(nvar)' 1027 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1028 1029 _See also_: 1030 *note omp_init_nest_lock::, *note omp_unset_nest_lock:: 1031 1032 _Reference_: 1033 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.4. 1034 1035 1036 File: libgomp.info, Node: omp_test_nest_lock, Next: omp_unset_nest_lock, Prev: omp_set_nest_lock, Up: Runtime Library Routines 1037 1038 2.37 'omp_test_nest_lock' - Test and set nested lock if available 1039 ================================================================= 1040 1041 _Description_: 1042 Before setting a nested lock, the lock variable must be initialized 1043 by 'omp_init_nest_lock'. Contrary to 'omp_set_nest_lock', 1044 'omp_test_nest_lock' does not block if the lock is not available. 1045 If the lock is already held by the current thread, the new nesting 1046 count is returned. Otherwise, the return value equals zero. 1047 1048 _C/C++_: 1049 _Prototype_: 'int omp_test_nest_lock(omp_nest_lock_t *lock);' 1050 1051 _Fortran_: 1052 _Interface_: 'logical function omp_test_nest_lock(nvar)' 1053 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1054 1055 _See also_: 1056 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 1057 1058 _Reference_: 1059 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.6. 1060 1061 1062 File: libgomp.info, Node: omp_unset_nest_lock, Next: omp_destroy_nest_lock, Prev: omp_test_nest_lock, Up: Runtime Library Routines 1063 1064 2.38 'omp_unset_nest_lock' - Unset nested lock 1065 ============================================== 1066 1067 _Description_: 1068 A nested lock about to be unset must have been locked by 1069 'omp_set_nested_lock' or 'omp_test_nested_lock' before. In 1070 addition, the lock must be held by the thread calling 1071 'omp_unset_nested_lock'. If the nesting count drops to zero, the 1072 lock becomes unlocked. If one ore more threads attempted to set 1073 the lock before, one of them is chosen to, again, set the lock to 1074 itself. 1075 1076 _C/C++_: 1077 _Prototype_: 'void omp_unset_nest_lock(omp_nest_lock_t *lock);' 1078 1079 _Fortran_: 1080 _Interface_: 'subroutine omp_unset_nest_lock(nvar)' 1081 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1082 1083 _See also_: 1084 *note omp_set_nest_lock:: 1085 1086 _Reference_: 1087 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.5. 1088 1089 1090 File: libgomp.info, Node: omp_destroy_nest_lock, Next: omp_get_wtick, Prev: omp_unset_nest_lock, Up: Runtime Library Routines 1091 1092 2.39 'omp_destroy_nest_lock' - Destroy nested lock 1093 ================================================== 1094 1095 _Description_: 1096 Destroy a nested lock. In order to be destroyed, a nested lock 1097 must be in the unlocked state and its nesting count must equal 1098 zero. 1099 1100 _C/C++_: 1101 _Prototype_: 'void omp_destroy_nest_lock(omp_nest_lock_t *);' 1102 1103 _Fortran_: 1104 _Interface_: 'subroutine omp_destroy_nest_lock(nvar)' 1105 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1106 1107 _See also_: 1108 *note omp_init_lock:: 1109 1110 _Reference_: 1111 OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.3. 1112 1113 1114 File: libgomp.info, Node: omp_get_wtick, Next: omp_get_wtime, Prev: omp_destroy_nest_lock, Up: Runtime Library Routines 1115 1116 2.40 'omp_get_wtick' - Get timer precision 1117 ========================================== 1118 1119 _Description_: 1120 Gets the timer precision, i.e., the number of seconds between two 1121 successive clock ticks. 1122 1123 _C/C++_: 1124 _Prototype_: 'double omp_get_wtick(void);' 1125 1126 _Fortran_: 1127 _Interface_: 'double precision function omp_get_wtick()' 1128 1129 _See also_: 1130 *note omp_get_wtime:: 1131 1132 _Reference_: 1133 OpenMP specification v4.5 (https://www.openmp.org), Section 3.4.2. 1134 1135 1136 File: libgomp.info, Node: omp_get_wtime, Prev: omp_get_wtick, Up: Runtime Library Routines 1137 1138 2.41 'omp_get_wtime' - Elapsed wall clock time 1139 ============================================== 1140 1141 _Description_: 1142 Elapsed wall clock time in seconds. The time is measured per 1143 thread, no guarantee can be made that two distinct threads measure 1144 the same time. Time is measured from some "time in the past", 1145 which is an arbitrary time guaranteed not to change during the 1146 execution of the program. 1147 1148 _C/C++_: 1149 _Prototype_: 'double omp_get_wtime(void);' 1150 1151 _Fortran_: 1152 _Interface_: 'double precision function omp_get_wtime()' 1153 1154 _See also_: 1155 *note omp_get_wtick:: 1156 1157 _Reference_: 1158 OpenMP specification v4.5 (https://www.openmp.org), Section 3.4.1. 1159 1160 1161 File: libgomp.info, Node: Environment Variables, Next: Enabling OpenACC, Prev: Runtime Library Routines, Up: Top 1162 1163 3 OpenMP Environment Variables 1164 ****************************** 1165 1166 The environment variables which beginning with 'OMP_' are defined by 1167 section 4 of the OpenMP specification in version 4.5, while those 1168 beginning with 'GOMP_' are GNU extensions. 1169 1170 * Menu: 1171 1172 * OMP_CANCELLATION:: Set whether cancellation is activated 1173 * OMP_DISPLAY_ENV:: Show OpenMP version and environment variables 1174 * OMP_DEFAULT_DEVICE:: Set the device used in target regions 1175 * OMP_DYNAMIC:: Dynamic adjustment of threads 1176 * OMP_MAX_ACTIVE_LEVELS:: Set the maximum number of nested parallel regions 1177 * OMP_MAX_TASK_PRIORITY:: Set the maximum task priority value 1178 * OMP_NESTED:: Nested parallel regions 1179 * OMP_NUM_THREADS:: Specifies the number of threads to use 1180 * OMP_PROC_BIND:: Whether theads may be moved between CPUs 1181 * OMP_PLACES:: Specifies on which CPUs the theads should be placed 1182 * OMP_STACKSIZE:: Set default thread stack size 1183 * OMP_SCHEDULE:: How threads are scheduled 1184 * OMP_THREAD_LIMIT:: Set the maximum number of threads 1185 * OMP_WAIT_POLICY:: How waiting threads are handled 1186 * GOMP_CPU_AFFINITY:: Bind threads to specific CPUs 1187 * GOMP_DEBUG:: Enable debugging output 1188 * GOMP_STACKSIZE:: Set default thread stack size 1189 * GOMP_SPINCOUNT:: Set the busy-wait spin count 1190 * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools 1191 1192 1193 File: libgomp.info, Node: OMP_CANCELLATION, Next: OMP_DISPLAY_ENV, Up: Environment Variables 1194 1195 3.1 'OMP_CANCELLATION' - Set whether cancellation is activated 1196 ============================================================== 1197 1198 _Description_: 1199 If set to 'TRUE', the cancellation is activated. If set to 'FALSE' 1200 or if unset, cancellation is disabled and the 'cancel' construct is 1201 ignored. 1202 1203 _See also_: 1204 *note omp_get_cancellation:: 1205 1206 _Reference_: 1207 OpenMP specification v4.5 (https://www.openmp.org), Section 4.11 1208 1209 1210 File: libgomp.info, Node: OMP_DISPLAY_ENV, Next: OMP_DEFAULT_DEVICE, Prev: OMP_CANCELLATION, Up: Environment Variables 1211 1212 3.2 'OMP_DISPLAY_ENV' - Show OpenMP version and environment variables 1213 ===================================================================== 1214 1215 _Description_: 1216 If set to 'TRUE', the OpenMP version number and the values 1217 associated with the OpenMP environment variables are printed to 1218 'stderr'. If set to 'VERBOSE', it additionally shows the value of 1219 the environment variables which are GNU extensions. If undefined 1220 or set to 'FALSE', this information will not be shown. 1221 1222 _Reference_: 1223 OpenMP specification v4.5 (https://www.openmp.org), Section 4.12 1224 1225 1226 File: libgomp.info, Node: OMP_DEFAULT_DEVICE, Next: OMP_DYNAMIC, Prev: OMP_DISPLAY_ENV, Up: Environment Variables 1227 1228 3.3 'OMP_DEFAULT_DEVICE' - Set the device used in target regions 1229 ================================================================ 1230 1231 _Description_: 1232 Set to choose the device which is used in a 'target' region, unless 1233 the value is overridden by 'omp_set_default_device' or by a 1234 'device' clause. The value shall be the nonnegative device number. 1235 If no device with the given device number exists, the code is 1236 executed on the host. If unset, device number 0 will be used. 1237 1238 _See also_: 1239 *note omp_get_default_device::, *note omp_set_default_device::, 1240 1241 _Reference_: 1242 OpenMP specification v4.5 (https://www.openmp.org), Section 4.13 1243 1244 1245 File: libgomp.info, Node: OMP_DYNAMIC, Next: OMP_MAX_ACTIVE_LEVELS, Prev: OMP_DEFAULT_DEVICE, Up: Environment Variables 1246 1247 3.4 'OMP_DYNAMIC' - Dynamic adjustment of threads 1248 ================================================= 1249 1250 _Description_: 1251 Enable or disable the dynamic adjustment of the number of threads 1252 within a team. The value of this environment variable shall be 1253 'TRUE' or 'FALSE'. If undefined, dynamic adjustment is disabled by 1254 default. 1255 1256 _See also_: 1257 *note omp_set_dynamic:: 1258 1259 _Reference_: 1260 OpenMP specification v4.5 (https://www.openmp.org), Section 4.3 1261 1262 1263 File: libgomp.info, Node: OMP_MAX_ACTIVE_LEVELS, Next: OMP_MAX_TASK_PRIORITY, Prev: OMP_DYNAMIC, Up: Environment Variables 1264 1265 3.5 'OMP_MAX_ACTIVE_LEVELS' - Set the maximum number of nested parallel regions 1266 =============================================================================== 1267 1268 _Description_: 1269 Specifies the initial value for the maximum number of nested 1270 parallel regions. The value of this variable shall be a positive 1271 integer. If undefined, the number of active levels is unlimited. 1272 1273 _See also_: 1274 *note omp_set_max_active_levels:: 1275 1276 _Reference_: 1277 OpenMP specification v4.5 (https://www.openmp.org), Section 4.9 1278 1279 1280 File: libgomp.info, Node: OMP_MAX_TASK_PRIORITY, Next: OMP_NESTED, Prev: OMP_MAX_ACTIVE_LEVELS, Up: Environment Variables 1281 1282 3.6 'OMP_MAX_TASK_PRIORITY' - Set the maximum priority 1283 ====================================================== 1284 1285 number that can be set for a task. 1286 _Description_: 1287 Specifies the initial value for the maximum priority value that can 1288 be set for a task. The value of this variable shall be a 1289 non-negative integer, and zero is allowed. If undefined, the 1290 default priority is 0. 1291 1292 _See also_: 1293 *note omp_get_max_task_priority:: 1294 1295 _Reference_: 1296 OpenMP specification v4.5 (https://www.openmp.org), Section 4.14 1297 1298 1299 File: libgomp.info, Node: OMP_NESTED, Next: OMP_NUM_THREADS, Prev: OMP_MAX_TASK_PRIORITY, Up: Environment Variables 1300 1301 3.7 'OMP_NESTED' - Nested parallel regions 1302 ========================================== 1303 1304 _Description_: 1305 Enable or disable nested parallel regions, i.e., whether team 1306 members are allowed to create new teams. The value of this 1307 environment variable shall be 'TRUE' or 'FALSE'. If undefined, 1308 nested parallel regions are disabled by default. 1309 1310 _See also_: 1311 *note omp_set_nested:: 1312 1313 _Reference_: 1314 OpenMP specification v4.5 (https://www.openmp.org), Section 4.6 1315 1316 1317 File: libgomp.info, Node: OMP_NUM_THREADS, Next: OMP_PROC_BIND, Prev: OMP_NESTED, Up: Environment Variables 1318 1319 3.8 'OMP_NUM_THREADS' - Specifies the number of threads to use 1320 ============================================================== 1321 1322 _Description_: 1323 Specifies the default number of threads to use in parallel regions. 1324 The value of this variable shall be a comma-separated list of 1325 positive integers; the value specified the number of threads to use 1326 for the corresponding nested level. If undefined one thread per 1327 CPU is used. 1328 1329 _See also_: 1330 *note omp_set_num_threads:: 1331 1332 _Reference_: 1333 OpenMP specification v4.5 (https://www.openmp.org), Section 4.2 1334 1335 1336 File: libgomp.info, Node: OMP_PROC_BIND, Next: OMP_PLACES, Prev: OMP_NUM_THREADS, Up: Environment Variables 1337 1338 3.9 'OMP_PROC_BIND' - Whether theads may be moved between CPUs 1339 ============================================================== 1340 1341 _Description_: 1342 Specifies whether threads may be moved between processors. If set 1343 to 'TRUE', OpenMP theads should not be moved; if set to 'FALSE' 1344 they may be moved. Alternatively, a comma separated list with the 1345 values 'MASTER', 'CLOSE' and 'SPREAD' can be used to specify the 1346 thread affinity policy for the corresponding nesting level. With 1347 'MASTER' the worker threads are in the same place partition as the 1348 master thread. With 'CLOSE' those are kept close to the master 1349 thread in contiguous place partitions. And with 'SPREAD' a sparse 1350 distribution across the place partitions is used. 1351 1352 When undefined, 'OMP_PROC_BIND' defaults to 'TRUE' when 1353 'OMP_PLACES' or 'GOMP_CPU_AFFINITY' is set and 'FALSE' otherwise. 1354 1355 _See also_: 1356 *note OMP_PLACES::, *note GOMP_CPU_AFFINITY::, *note 1357 omp_get_proc_bind:: 1358 1359 _Reference_: 1360 OpenMP specification v4.5 (https://www.openmp.org), Section 4.4 1361 1362 1363 File: libgomp.info, Node: OMP_PLACES, Next: OMP_STACKSIZE, Prev: OMP_PROC_BIND, Up: Environment Variables 1364 1365 3.10 'OMP_PLACES' - Specifies on which CPUs the theads should be placed 1366 ======================================================================= 1367 1368 _Description_: 1369 The thread placement can be either specified using an abstract name 1370 or by an explicit list of the places. The abstract names 1371 'threads', 'cores' and 'sockets' can be optionally followed by a 1372 positive number in parentheses, which denotes the how many places 1373 shall be created. With 'threads' each place corresponds to a 1374 single hardware thread; 'cores' to a single core with the 1375 corresponding number of hardware threads; and with 'sockets' the 1376 place corresponds to a single socket. The resulting placement can 1377 be shown by setting the 'OMP_DISPLAY_ENV' environment variable. 1378 1379 Alternatively, the placement can be specified explicitly as 1380 comma-separated list of places. A place is specified by set of 1381 nonnegative numbers in curly braces, denoting the denoting the 1382 hardware threads. The hardware threads belonging to a place can 1383 either be specified as comma-separated list of nonnegative thread 1384 numbers or using an interval. Multiple places can also be either 1385 specified by a comma-separated list of places or by an interval. 1386 To specify an interval, a colon followed by the count is placed 1387 after after the hardware thread number or the place. Optionally, 1388 the length can be followed by a colon and the stride number - 1389 otherwise a unit stride is assumed. For instance, the following 1390 specifies the same places list: '"{0,1,2}, {3,4,6}, {7,8,9}, 1391 {10,11,12}"'; '"{0:3}, {3:3}, {7:3}, {10:3}"'; and '"{0:2}:4:3"'. 1392 1393 If 'OMP_PLACES' and 'GOMP_CPU_AFFINITY' are unset and 1394 'OMP_PROC_BIND' is either unset or 'false', threads may be moved 1395 between CPUs following no placement policy. 1396 1397 _See also_: 1398 *note OMP_PROC_BIND::, *note GOMP_CPU_AFFINITY::, *note 1399 omp_get_proc_bind::, *note OMP_DISPLAY_ENV:: 1400 1401 _Reference_: 1402 OpenMP specification v4.5 (https://www.openmp.org), Section 4.5 1403 1404 1405 File: libgomp.info, Node: OMP_STACKSIZE, Next: OMP_SCHEDULE, Prev: OMP_PLACES, Up: Environment Variables 1406 1407 3.11 'OMP_STACKSIZE' - Set default thread stack size 1408 ==================================================== 1409 1410 _Description_: 1411 Set the default thread stack size in kilobytes, unless the number 1412 is suffixed by 'B', 'K', 'M' or 'G', in which case the size is, 1413 respectively, in bytes, kilobytes, megabytes or gigabytes. This is 1414 different from 'pthread_attr_setstacksize' which gets the number of 1415 bytes as an argument. If the stack size cannot be set due to 1416 system constraints, an error is reported and the initial stack size 1417 is left unchanged. If undefined, the stack size is system 1418 dependent. 1419 1420 _Reference_: 1421 OpenMP specification v4.5 (https://www.openmp.org), Section 4.7 1422 1423 1424 File: libgomp.info, Node: OMP_SCHEDULE, Next: OMP_THREAD_LIMIT, Prev: OMP_STACKSIZE, Up: Environment Variables 1425 1426 3.12 'OMP_SCHEDULE' - How threads are scheduled 1427 =============================================== 1428 1429 _Description_: 1430 Allows to specify 'schedule type' and 'chunk size'. The value of 1431 the variable shall have the form: 'type[,chunk]' where 'type' is 1432 one of 'static', 'dynamic', 'guided' or 'auto' The optional 'chunk' 1433 size shall be a positive integer. If undefined, dynamic scheduling 1434 and a chunk size of 1 is used. 1435 1436 _See also_: 1437 *note omp_set_schedule:: 1438 1439 _Reference_: 1440 OpenMP specification v4.5 (https://www.openmp.org), Sections 1441 2.7.1.1 and 4.1 1442 1443 1444 File: libgomp.info, Node: OMP_THREAD_LIMIT, Next: OMP_WAIT_POLICY, Prev: OMP_SCHEDULE, Up: Environment Variables 1445 1446 3.13 'OMP_THREAD_LIMIT' - Set the maximum number of threads 1447 =========================================================== 1448 1449 _Description_: 1450 Specifies the number of threads to use for the whole program. The 1451 value of this variable shall be a positive integer. If undefined, 1452 the number of threads is not limited. 1453 1454 _See also_: 1455 *note OMP_NUM_THREADS::, *note omp_get_thread_limit:: 1456 1457 _Reference_: 1458 OpenMP specification v4.5 (https://www.openmp.org), Section 4.10 1459 1460 1461 File: libgomp.info, Node: OMP_WAIT_POLICY, Next: GOMP_CPU_AFFINITY, Prev: OMP_THREAD_LIMIT, Up: Environment Variables 1462 1463 3.14 'OMP_WAIT_POLICY' - How waiting threads are handled 1464 ======================================================== 1465 1466 _Description_: 1467 Specifies whether waiting threads should be active or passive. If 1468 the value is 'PASSIVE', waiting threads should not consume CPU 1469 power while waiting; while the value is 'ACTIVE' specifies that 1470 they should. If undefined, threads wait actively for a short time 1471 before waiting passively. 1472 1473 _See also_: 1474 *note GOMP_SPINCOUNT:: 1475 1476 _Reference_: 1477 OpenMP specification v4.5 (https://www.openmp.org), Section 4.8 1478 1479 1480 File: libgomp.info, Node: GOMP_CPU_AFFINITY, Next: GOMP_DEBUG, Prev: OMP_WAIT_POLICY, Up: Environment Variables 1481 1482 3.15 'GOMP_CPU_AFFINITY' - Bind threads to specific CPUs 1483 ======================================================== 1484 1485 _Description_: 1486 Binds threads to specific CPUs. The variable should contain a 1487 space-separated or comma-separated list of CPUs. This list may 1488 contain different kinds of entries: either single CPU numbers in 1489 any order, a range of CPUs (M-N) or a range with some stride 1490 (M-N:S). CPU numbers are zero based. For example, 1491 'GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"' will bind the initial thread 1492 to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to 1493 CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 1494 10, 12, and 14 respectively and then start assigning back from the 1495 beginning of the list. 'GOMP_CPU_AFFINITY=0' binds all threads to 1496 CPU 0. 1497 1498 There is no libgomp library routine to determine whether a CPU 1499 affinity specification is in effect. As a workaround, 1500 language-specific library functions, e.g., 'getenv' in C or 1501 'GET_ENVIRONMENT_VARIABLE' in Fortran, may be used to query the 1502 setting of the 'GOMP_CPU_AFFINITY' environment variable. A defined 1503 CPU affinity on startup cannot be changed or disabled during the 1504 runtime of the application. 1505 1506 If both 'GOMP_CPU_AFFINITY' and 'OMP_PROC_BIND' are set, 1507 'OMP_PROC_BIND' has a higher precedence. If neither has been set 1508 and 'OMP_PROC_BIND' is unset, or when 'OMP_PROC_BIND' is set to 1509 'FALSE', the host system will handle the assignment of threads to 1510 CPUs. 1511 1512 _See also_: 1513 *note OMP_PLACES::, *note OMP_PROC_BIND:: 1514 1515 1516 File: libgomp.info, Node: GOMP_DEBUG, Next: GOMP_STACKSIZE, Prev: GOMP_CPU_AFFINITY, Up: Environment Variables 1517 1518 3.16 'GOMP_DEBUG' - Enable debugging output 1519 =========================================== 1520 1521 _Description_: 1522 Enable debugging output. The variable should be set to '0' 1523 (disabled, also the default if not set), or '1' (enabled). 1524 1525 If enabled, some debugging output will be printed during execution. 1526 This is currently not specified in more detail, and subject to 1527 change. 1528 1529 1530 File: libgomp.info, Node: GOMP_STACKSIZE, Next: GOMP_SPINCOUNT, Prev: GOMP_DEBUG, Up: Environment Variables 1531 1532 3.17 'GOMP_STACKSIZE' - Set default thread stack size 1533 ===================================================== 1534 1535 _Description_: 1536 Set the default thread stack size in kilobytes. This is different 1537 from 'pthread_attr_setstacksize' which gets the number of bytes as 1538 an argument. If the stack size cannot be set due to system 1539 constraints, an error is reported and the initial stack size is 1540 left unchanged. If undefined, the stack size is system dependent. 1541 1542 _See also_: 1543 *note OMP_STACKSIZE:: 1544 1545 _Reference_: 1546 GCC Patches Mailinglist 1547 (https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html), GCC 1548 Patches Mailinglist 1549 (https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html) 1550 1551 1552 File: libgomp.info, Node: GOMP_SPINCOUNT, Next: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_STACKSIZE, Up: Environment Variables 1553 1554 3.18 'GOMP_SPINCOUNT' - Set the busy-wait spin count 1555 ==================================================== 1556 1557 _Description_: 1558 Determines how long a threads waits actively with consuming CPU 1559 power before waiting passively without consuming CPU power. The 1560 value may be either 'INFINITE', 'INFINITY' to always wait actively 1561 or an integer which gives the number of spins of the busy-wait 1562 loop. The integer may optionally be followed by the following 1563 suffixes acting as multiplication factors: 'k' (kilo, thousand), 1564 'M' (mega, million), 'G' (giga, billion), or 'T' (tera, trillion). 1565 If undefined, 0 is used when 'OMP_WAIT_POLICY' is 'PASSIVE', 1566 300,000 is used when 'OMP_WAIT_POLICY' is undefined and 30 billion 1567 is used when 'OMP_WAIT_POLICY' is 'ACTIVE'. If there are more 1568 OpenMP threads than available CPUs, 1000 and 100 spins are used for 1569 'OMP_WAIT_POLICY' being 'ACTIVE' or undefined, respectively; unless 1570 the 'GOMP_SPINCOUNT' is lower or 'OMP_WAIT_POLICY' is 'PASSIVE'. 1571 1572 _See also_: 1573 *note OMP_WAIT_POLICY:: 1574 1575 1576 File: libgomp.info, Node: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_SPINCOUNT, Up: Environment Variables 1577 1578 3.19 'GOMP_RTEMS_THREAD_POOLS' - Set the RTEMS specific thread pools 1579 ==================================================================== 1580 1581 _Description_: 1582 This environment variable is only used on the RTEMS real-time 1583 operating system. It determines the scheduler instance specific 1584 thread pools. The format for 'GOMP_RTEMS_THREAD_POOLS' is a list 1585 of optional '<thread-pool-count>[$<priority>]@<scheduler-name>' 1586 configurations separated by ':' where: 1587 * '<thread-pool-count>' is the thread pool count for this 1588 scheduler instance. 1589 * '$<priority>' is an optional priority for the worker threads 1590 of a thread pool according to 'pthread_setschedparam'. In 1591 case a priority value is omitted, then a worker thread will 1592 inherit the priority of the OpenMP master thread that created 1593 it. The priority of the worker thread is not changed after 1594 creation, even if a new OpenMP master thread using the worker 1595 has a different priority. 1596 * '@<scheduler-name>' is the scheduler instance name according 1597 to the RTEMS application configuration. 1598 In case no thread pool configuration is specified for a scheduler 1599 instance, then each OpenMP master thread of this scheduler instance 1600 will use its own dynamically allocated thread pool. To limit the 1601 worker thread count of the thread pools, each OpenMP master thread 1602 must call 'omp_set_num_threads'. 1603 _Example_: 1604 Lets suppose we have three scheduler instances 'IO', 'WRK0', and 1605 'WRK1' with 'GOMP_RTEMS_THREAD_POOLS' set to '"1@WRK0:3$4@WRK1"'. 1606 Then there are no thread pool restrictions for scheduler instance 1607 'IO'. In the scheduler instance 'WRK0' there is one thread pool 1608 available. Since no priority is specified for this scheduler 1609 instance, the worker thread inherits the priority of the OpenMP 1610 master thread that created it. In the scheduler instance 'WRK1' 1611 there are three thread pools available and their worker threads run 1612 at priority four. 1613 1614 1615 File: libgomp.info, Node: Enabling OpenACC, Next: OpenACC Runtime Library Routines, Prev: Environment Variables, Up: Top 1616 1617 4 Enabling OpenACC 1618 ****************** 1619 1620 To activate the OpenACC extensions for C/C++ and Fortran, the 1621 compile-time flag '-fopenacc' must be specified. This enables the 1622 OpenACC directive '#pragma acc' in C/C++ and '!$acc' directives in free 1623 form, 'c$acc', '*$acc' and '!$acc' directives in fixed form, '!$' 1624 conditional compilation sentinels in free form and 'c$', '*$' and '!$' 1625 sentinels in fixed form, for Fortran. The flag also arranges for 1626 automatic linking of the OpenACC runtime library (*note OpenACC Runtime 1627 Library Routines::). 1628 1629 See <https://gcc.gnu.org/wiki/OpenACC> for more information. 1630 1631 A complete description of all OpenACC directives accepted may be 1632 found in the OpenACC (https://www.openacc.org) Application Programming 1633 Interface manual, version 2.6. 1634 1635 1636 File: libgomp.info, Node: OpenACC Runtime Library Routines, Next: OpenACC Environment Variables, Prev: Enabling OpenACC, Up: Top 1637 1638 5 OpenACC Runtime Library Routines 1639 ********************************** 1640 1641 The runtime routines described here are defined by section 3 of the 1642 OpenACC specifications in version 2.6. They have C linkage, and do not 1643 throw exceptions. Generally, they are available only for the host, with 1644 the exception of 'acc_on_device', which is available for both the host 1645 and the acceleration device. 1646 1647 * Menu: 1648 1649 * acc_get_num_devices:: Get number of devices for the given device 1650 type. 1651 * acc_set_device_type:: Set type of device accelerator to use. 1652 * acc_get_device_type:: Get type of device accelerator to be used. 1653 * acc_set_device_num:: Set device number to use. 1654 * acc_get_device_num:: Get device number to be used. 1655 * acc_get_property:: Get device property. 1656 * acc_async_test:: Tests for completion of a specific asynchronous 1657 operation. 1658 * acc_async_test_all:: Tests for completion of all asynchronous 1659 operations. 1660 * acc_wait:: Wait for completion of a specific asynchronous 1661 operation. 1662 * acc_wait_all:: Waits for completion of all asynchronous 1663 operations. 1664 * acc_wait_all_async:: Wait for completion of all asynchronous 1665 operations. 1666 * acc_wait_async:: Wait for completion of asynchronous operations. 1667 * acc_init:: Initialize runtime for a specific device type. 1668 * acc_shutdown:: Shuts down the runtime for a specific device 1669 type. 1670 * acc_on_device:: Whether executing on a particular device 1671 * acc_malloc:: Allocate device memory. 1672 * acc_free:: Free device memory. 1673 * acc_copyin:: Allocate device memory and copy host memory to 1674 it. 1675 * acc_present_or_copyin:: If the data is not present on the device, 1676 allocate device memory and copy from host 1677 memory. 1678 * acc_create:: Allocate device memory and map it to host 1679 memory. 1680 * acc_present_or_create:: If the data is not present on the device, 1681 allocate device memory and map it to host 1682 memory. 1683 * acc_copyout:: Copy device memory to host memory. 1684 * acc_delete:: Free device memory. 1685 * acc_update_device:: Update device memory from mapped host memory. 1686 * acc_update_self:: Update host memory from mapped device memory. 1687 * acc_map_data:: Map previously allocated device memory to host 1688 memory. 1689 * acc_unmap_data:: Unmap device memory from host memory. 1690 * acc_deviceptr:: Get device pointer associated with specific 1691 host address. 1692 * acc_hostptr:: Get host pointer associated with specific 1693 device address. 1694 * acc_is_present:: Indicate whether host variable / array is 1695 present on device. 1696 * acc_memcpy_to_device:: Copy host memory to device memory. 1697 * acc_memcpy_from_device:: Copy device memory to host memory. 1698 * acc_attach:: Let device pointer point to device-pointer target. 1699 * acc_detach:: Let device pointer point to host-pointer target. 1700 1701 API routines for target platforms. 1702 1703 * acc_get_current_cuda_device:: Get CUDA device handle. 1704 * acc_get_current_cuda_context::Get CUDA context handle. 1705 * acc_get_cuda_stream:: Get CUDA stream handle. 1706 * acc_set_cuda_stream:: Set CUDA stream handle. 1707 1708 API routines for the OpenACC Profiling Interface. 1709 1710 * acc_prof_register:: Register callbacks. 1711 * acc_prof_unregister:: Unregister callbacks. 1712 * acc_prof_lookup:: Obtain inquiry functions. 1713 * acc_register_library:: Library registration. 1714 1715 1716 File: libgomp.info, Node: acc_get_num_devices, Next: acc_set_device_type, Up: OpenACC Runtime Library Routines 1717 1718 5.1 'acc_get_num_devices' - Get number of devices for given device type 1719 ======================================================================= 1720 1721 _Description_ 1722 This function returns a value indicating the number of devices 1723 available for the device type specified in DEVICETYPE. 1724 1725 _C/C++_: 1726 _Prototype_: 'int acc_get_num_devices(acc_device_t devicetype);' 1727 1728 _Fortran_: 1729 _Interface_: 'integer function acc_get_num_devices(devicetype)' 1730 'integer(kind=acc_device_kind) devicetype' 1731 1732 _Reference_: 1733 OpenACC specification v2.6 (https://www.openacc.org), section 1734 3.2.1. 1735 1736 1737 File: libgomp.info, Node: acc_set_device_type, Next: acc_get_device_type, Prev: acc_get_num_devices, Up: OpenACC Runtime Library Routines 1738 1739 5.2 'acc_set_device_type' - Set type of device accelerator to use. 1740 ================================================================== 1741 1742 _Description_ 1743 This function indicates to the runtime library which device type, 1744 specified in DEVICETYPE, to use when executing a parallel or 1745 kernels region. 1746 1747 _C/C++_: 1748 _Prototype_: 'acc_set_device_type(acc_device_t devicetype);' 1749 1750 _Fortran_: 1751 _Interface_: 'subroutine acc_set_device_type(devicetype)' 1752 'integer(kind=acc_device_kind) devicetype' 1753 1754 _Reference_: 1755 OpenACC specification v2.6 (https://www.openacc.org), section 1756 3.2.2. 1757 1758 1759 File: libgomp.info, Node: acc_get_device_type, Next: acc_set_device_num, Prev: acc_set_device_type, Up: OpenACC Runtime Library Routines 1760 1761 5.3 'acc_get_device_type' - Get type of device accelerator to be used. 1762 ====================================================================== 1763 1764 _Description_ 1765 This function returns what device type will be used when executing 1766 a parallel or kernels region. 1767 1768 This function returns 'acc_device_none' if 'acc_get_device_type' is 1769 called from 'acc_ev_device_init_start', 'acc_ev_device_init_end' 1770 callbacks of the OpenACC Profiling Interface (*note OpenACC 1771 Profiling Interface::), that is, if the device is currently being 1772 initialized. 1773 1774 _C/C++_: 1775 _Prototype_: 'acc_device_t acc_get_device_type(void);' 1776 1777 _Fortran_: 1778 _Interface_: 'function acc_get_device_type(void)' 1779 'integer(kind=acc_device_kind) acc_get_device_type' 1780 1781 _Reference_: 1782 OpenACC specification v2.6 (https://www.openacc.org), section 1783 3.2.3. 1784 1785 1786 File: libgomp.info, Node: acc_set_device_num, Next: acc_get_device_num, Prev: acc_get_device_type, Up: OpenACC Runtime Library Routines 1787 1788 5.4 'acc_set_device_num' - Set device number to use. 1789 ==================================================== 1790 1791 _Description_ 1792 This function will indicate to the runtime which device number, 1793 specified by DEVICENUM, associated with the specified device type 1794 DEVICETYPE. 1795 1796 _C/C++_: 1797 _Prototype_: 'acc_set_device_num(int devicenum, acc_device_t 1798 devicetype);' 1799 1800 _Fortran_: 1801 _Interface_: 'subroutine acc_set_device_num(devicenum, devicetype)' 1802 'integer devicenum' 1803 'integer(kind=acc_device_kind) devicetype' 1804 1805 _Reference_: 1806 OpenACC specification v2.6 (https://www.openacc.org), section 1807 3.2.4. 1808 1809 1810 File: libgomp.info, Node: acc_get_device_num, Next: acc_get_property, Prev: acc_set_device_num, Up: OpenACC Runtime Library Routines 1811 1812 5.5 'acc_get_device_num' - Get device number to be used. 1813 ======================================================== 1814 1815 _Description_ 1816 This function returns which device number associated with the 1817 specified device type DEVICETYPE, will be used when executing a 1818 parallel or kernels region. 1819 1820 _C/C++_: 1821 _Prototype_: 'int acc_get_device_num(acc_device_t devicetype);' 1822 1823 _Fortran_: 1824 _Interface_: 'function acc_get_device_num(devicetype)' 1825 'integer(kind=acc_device_kind) devicetype' 1826 'integer acc_get_device_num' 1827 1828 _Reference_: 1829 OpenACC specification v2.6 (https://www.openacc.org), section 1830 3.2.5. 1831 1832 1833 File: libgomp.info, Node: acc_get_property, Next: acc_async_test, Prev: acc_get_device_num, Up: OpenACC Runtime Library Routines 1834 1835 5.6 'acc_get_property' - Get device property. 1836 ============================================= 1837 1838 _Description_ 1839 These routines return the value of the specified PROPERTY for the 1840 device being queried according to DEVICENUM and DEVICETYPE. 1841 Integer-valued and string-valued properties are returned by 1842 'acc_get_property' and 'acc_get_property_string' respectively. The 1843 Fortran 'acc_get_property_string' subroutine returns the string 1844 retrieved in its fourth argument while the remaining entry points 1845 are functions, which pass the return value as their result. 1846 1847 Note for Fortran, only: the OpenACC technical committee corrected 1848 and, hence, modified the interface introduced in OpenACC 2.6. The 1849 kind-value parameter 'acc_device_property' has been renamed to 1850 'acc_device_property_kind' for consistency and the return type of 1851 the 'acc_get_property' function is now a 'c_size_t' integer instead 1852 of a 'acc_device_property' integer. The parameter 1853 'acc_device_property' will continue to be provided, but might be 1854 removed in a future version of GCC. 1855 1856 _C/C++_: 1857 _Prototype_: 'size_t acc_get_property(int devicenum, acc_device_t 1858 devicetype, acc_device_property_t property);' 1859 _Prototype_: 'const char *acc_get_property_string(int devicenum, 1860 acc_device_t devicetype, acc_device_property_t 1861 property);' 1862 1863 _Fortran_: 1864 _Interface_: 'function acc_get_property(devicenum, devicetype, 1865 property)' 1866 _Interface_: 'subroutine acc_get_property_string(devicenum, 1867 devicetype, property, string)' 1868 'use ISO_C_Binding, only: c_size_t' 1869 'integer devicenum' 1870 'integer(kind=acc_device_kind) devicetype' 1871 'integer(kind=acc_device_property_kind) property' 1872 'integer(kind=c_size_t) acc_get_property' 1873 'character(*) string' 1874 1875 _Reference_: 1876 OpenACC specification v2.6 (https://www.openacc.org), section 1877 3.2.6. 1878 1879 1880 File: libgomp.info, Node: acc_async_test, Next: acc_async_test_all, Prev: acc_get_property, Up: OpenACC Runtime Library Routines 1881 1882 5.7 'acc_async_test' - Test for completion of a specific asynchronous operation. 1883 ================================================================================ 1884 1885 _Description_ 1886 This function tests for completion of the asynchronous operation 1887 specified in ARG. In C/C++, a non-zero value will be returned to 1888 indicate the specified asynchronous operation has completed. While 1889 Fortran will return a 'true'. If the asynchronous operation has 1890 not completed, C/C++ returns a zero and Fortran returns a 'false'. 1891 1892 _C/C++_: 1893 _Prototype_: 'int acc_async_test(int arg);' 1894 1895 _Fortran_: 1896 _Interface_: 'function acc_async_test(arg)' 1897 'integer(kind=acc_handle_kind) arg' 1898 'logical acc_async_test' 1899 1900 _Reference_: 1901 OpenACC specification v2.6 (https://www.openacc.org), section 1902 3.2.9. 1903 1904 1905 File: libgomp.info, Node: acc_async_test_all, Next: acc_wait, Prev: acc_async_test, Up: OpenACC Runtime Library Routines 1906 1907 5.8 'acc_async_test_all' - Tests for completion of all asynchronous operations. 1908 =============================================================================== 1909 1910 _Description_ 1911 This function tests for completion of all asynchronous operations. 1912 In C/C++, a non-zero value will be returned to indicate all 1913 asynchronous operations have completed. While Fortran will return 1914 a 'true'. If any asynchronous operation has not completed, C/C++ 1915 returns a zero and Fortran returns a 'false'. 1916 1917 _C/C++_: 1918 _Prototype_: 'int acc_async_test_all(void);' 1919 1920 _Fortran_: 1921 _Interface_: 'function acc_async_test()' 1922 'logical acc_get_device_num' 1923 1924 _Reference_: 1925 OpenACC specification v2.6 (https://www.openacc.org), section 1926 3.2.10. 1927 1928 1929 File: libgomp.info, Node: acc_wait, Next: acc_wait_all, Prev: acc_async_test_all, Up: OpenACC Runtime Library Routines 1930 1931 5.9 'acc_wait' - Wait for completion of a specific asynchronous operation. 1932 ========================================================================== 1933 1934 _Description_ 1935 This function waits for completion of the asynchronous operation 1936 specified in ARG. 1937 1938 _C/C++_: 1939 _Prototype_: 'acc_wait(arg);' 1940 _Prototype 'acc_async_wait(arg);' 1941 (OpenACC 1.0 1942 compatibility)_: 1943 1944 _Fortran_: 1945 _Interface_: 'subroutine acc_wait(arg)' 1946 'integer(acc_handle_kind) arg' 1947 _Interface 'subroutine acc_async_wait(arg)' 1948 (OpenACC 1.0 1949 compatibility)_: 1950 'integer(acc_handle_kind) arg' 1951 1952 _Reference_: 1953 OpenACC specification v2.6 (https://www.openacc.org), section 1954 3.2.11. 1955 1956 1957 File: libgomp.info, Node: acc_wait_all, Next: acc_wait_all_async, Prev: acc_wait, Up: OpenACC Runtime Library Routines 1958 1959 5.10 'acc_wait_all' - Waits for completion of all asynchronous operations. 1960 ========================================================================== 1961 1962 _Description_ 1963 This function waits for the completion of all asynchronous 1964 operations. 1965 1966 _C/C++_: 1967 _Prototype_: 'acc_wait_all(void);' 1968 _Prototype 'acc_async_wait_all(void);' 1969 (OpenACC 1.0 1970 compatibility)_: 1971 1972 _Fortran_: 1973 _Interface_: 'subroutine acc_wait_all()' 1974 _Interface 'subroutine acc_async_wait_all()' 1975 (OpenACC 1.0 1976 compatibility)_: 1977 1978 _Reference_: 1979 OpenACC specification v2.6 (https://www.openacc.org), section 1980 3.2.13. 1981 1982 1983 File: libgomp.info, Node: acc_wait_all_async, Next: acc_wait_async, Prev: acc_wait_all, Up: OpenACC Runtime Library Routines 1984 1985 5.11 'acc_wait_all_async' - Wait for completion of all asynchronous operations. 1986 =============================================================================== 1987 1988 _Description_ 1989 This function enqueues a wait operation on the queue ASYNC for any 1990 and all asynchronous operations that have been previously enqueued 1991 on any queue. 1992 1993 _C/C++_: 1994 _Prototype_: 'acc_wait_all_async(int async);' 1995 1996 _Fortran_: 1997 _Interface_: 'subroutine acc_wait_all_async(async)' 1998 'integer(acc_handle_kind) async' 1999 2000 _Reference_: 2001 OpenACC specification v2.6 (https://www.openacc.org), section 2002 3.2.14. 2003 2004 2005 File: libgomp.info, Node: acc_wait_async, Next: acc_init, Prev: acc_wait_all_async, Up: OpenACC Runtime Library Routines 2006 2007 5.12 'acc_wait_async' - Wait for completion of asynchronous operations. 2008 ======================================================================= 2009 2010 _Description_ 2011 This function enqueues a wait operation on queue ASYNC for any and 2012 all asynchronous operations enqueued on queue ARG. 2013 2014 _C/C++_: 2015 _Prototype_: 'acc_wait_async(int arg, int async);' 2016 2017 _Fortran_: 2018 _Interface_: 'subroutine acc_wait_async(arg, async)' 2019 'integer(acc_handle_kind) arg, async' 2020 2021 _Reference_: 2022 OpenACC specification v2.6 (https://www.openacc.org), section 2023 3.2.12. 2024 2025 2026 File: libgomp.info, Node: acc_init, Next: acc_shutdown, Prev: acc_wait_async, Up: OpenACC Runtime Library Routines 2027 2028 5.13 'acc_init' - Initialize runtime for a specific device type. 2029 ================================================================ 2030 2031 _Description_ 2032 This function initializes the runtime for the device type specified 2033 in DEVICETYPE. 2034 2035 _C/C++_: 2036 _Prototype_: 'acc_init(acc_device_t devicetype);' 2037 2038 _Fortran_: 2039 _Interface_: 'subroutine acc_init(devicetype)' 2040 'integer(acc_device_kind) devicetype' 2041 2042 _Reference_: 2043 OpenACC specification v2.6 (https://www.openacc.org), section 2044 3.2.7. 2045 2046 2047 File: libgomp.info, Node: acc_shutdown, Next: acc_on_device, Prev: acc_init, Up: OpenACC Runtime Library Routines 2048 2049 5.14 'acc_shutdown' - Shuts down the runtime for a specific device type. 2050 ======================================================================== 2051 2052 _Description_ 2053 This function shuts down the runtime for the device type specified 2054 in DEVICETYPE. 2055 2056 _C/C++_: 2057 _Prototype_: 'acc_shutdown(acc_device_t devicetype);' 2058 2059 _Fortran_: 2060 _Interface_: 'subroutine acc_shutdown(devicetype)' 2061 'integer(acc_device_kind) devicetype' 2062 2063 _Reference_: 2064 OpenACC specification v2.6 (https://www.openacc.org), section 2065 3.2.8. 2066 2067 2068 File: libgomp.info, Node: acc_on_device, Next: acc_malloc, Prev: acc_shutdown, Up: OpenACC Runtime Library Routines 2069 2070 5.15 'acc_on_device' - Whether executing on a particular device 2071 =============================================================== 2072 2073 _Description_: 2074 This function returns whether the program is executing on a 2075 particular device specified in DEVICETYPE. In C/C++ a non-zero 2076 value is returned to indicate the device is executing on the 2077 specified device type. In Fortran, 'true' will be returned. If 2078 the program is not executing on the specified device type C/C++ 2079 will return a zero, while Fortran will return 'false'. 2080 2081 _C/C++_: 2082 _Prototype_: 'acc_on_device(acc_device_t devicetype);' 2083 2084 _Fortran_: 2085 _Interface_: 'function acc_on_device(devicetype)' 2086 'integer(acc_device_kind) devicetype' 2087 'logical acc_on_device' 2088 2089 _Reference_: 2090 OpenACC specification v2.6 (https://www.openacc.org), section 2091 3.2.17. 2092 2093 2094 File: libgomp.info, Node: acc_malloc, Next: acc_free, Prev: acc_on_device, Up: OpenACC Runtime Library Routines 2095 2096 5.16 'acc_malloc' - Allocate device memory. 2097 =========================================== 2098 2099 _Description_ 2100 This function allocates LEN bytes of device memory. It returns the 2101 device address of the allocated memory. 2102 2103 _C/C++_: 2104 _Prototype_: 'd_void* acc_malloc(size_t len);' 2105 2106 _Reference_: 2107 OpenACC specification v2.6 (https://www.openacc.org), section 2108 3.2.18. 2109 2110 2111 File: libgomp.info, Node: acc_free, Next: acc_copyin, Prev: acc_malloc, Up: OpenACC Runtime Library Routines 2112 2113 5.17 'acc_free' - Free device memory. 2114 ===================================== 2115 2116 _Description_ 2117 Free previously allocated device memory at the device address 'a'. 2118 2119 _C/C++_: 2120 _Prototype_: 'acc_free(d_void *a);' 2121 2122 _Reference_: 2123 OpenACC specification v2.6 (https://www.openacc.org), section 2124 3.2.19. 2125 2126 2127 File: libgomp.info, Node: acc_copyin, Next: acc_present_or_copyin, Prev: acc_free, Up: OpenACC Runtime Library Routines 2128 2129 5.18 'acc_copyin' - Allocate device memory and copy host memory to it. 2130 ====================================================================== 2131 2132 _Description_ 2133 In C/C++, this function allocates LEN bytes of device memory and 2134 maps it to the specified host address in A. The device address of 2135 the newly allocated device memory is returned. 2136 2137 In Fortran, two (2) forms are supported. In the first form, A 2138 specifies a contiguous array section. The second form A specifies 2139 a variable or array element and LEN specifies the length in bytes. 2140 2141 _C/C++_: 2142 _Prototype_: 'void *acc_copyin(h_void *a, size_t len);' 2143 _Prototype_: 'void *acc_copyin_async(h_void *a, size_t len, int 2144 async);' 2145 2146 _Fortran_: 2147 _Interface_: 'subroutine acc_copyin(a)' 2148 'type, dimension(:[,:]...) :: a' 2149 _Interface_: 'subroutine acc_copyin(a, len)' 2150 'type, dimension(:[,:]...) :: a' 2151 'integer len' 2152 _Interface_: 'subroutine acc_copyin_async(a, async)' 2153 'type, dimension(:[,:]...) :: a' 2154 'integer(acc_handle_kind) :: async' 2155 _Interface_: 'subroutine acc_copyin_async(a, len, async)' 2156 'type, dimension(:[,:]...) :: a' 2157 'integer len' 2158 'integer(acc_handle_kind) :: async' 2159 2160 _Reference_: 2161 OpenACC specification v2.6 (https://www.openacc.org), section 2162 3.2.20. 2163 2164 2165 File: libgomp.info, Node: acc_present_or_copyin, Next: acc_create, Prev: acc_copyin, Up: OpenACC Runtime Library Routines 2166 2167 5.19 'acc_present_or_copyin' - If the data is not present on the device, allocate device memory and copy from host memory. 2168 ========================================================================================================================== 2169 2170 _Description_ 2171 This function tests if the host data specified by A and of length 2172 LEN is present or not. If it is not present, then device memory 2173 will be allocated and the host memory copied. The device address 2174 of the newly allocated device memory is returned. 2175 2176 In Fortran, two (2) forms are supported. In the first form, A 2177 specifies a contiguous array section. The second form A specifies 2178 a variable or array element and LEN specifies the length in bytes. 2179 2180 Note that 'acc_present_or_copyin' and 'acc_pcopyin' exist for 2181 backward compatibility with OpenACC 2.0; use *note acc_copyin:: 2182 instead. 2183 2184 _C/C++_: 2185 _Prototype_: 'void *acc_present_or_copyin(h_void *a, size_t len);' 2186 _Prototype_: 'void *acc_pcopyin(h_void *a, size_t len);' 2187 2188 _Fortran_: 2189 _Interface_: 'subroutine acc_present_or_copyin(a)' 2190 'type, dimension(:[,:]...) :: a' 2191 _Interface_: 'subroutine acc_present_or_copyin(a, len)' 2192 'type, dimension(:[,:]...) :: a' 2193 'integer len' 2194 _Interface_: 'subroutine acc_pcopyin(a)' 2195 'type, dimension(:[,:]...) :: a' 2196 _Interface_: 'subroutine acc_pcopyin(a, len)' 2197 'type, dimension(:[,:]...) :: a' 2198 'integer len' 2199 2200 _Reference_: 2201 OpenACC specification v2.6 (https://www.openacc.org), section 2202 3.2.20. 2203 2204 2205 File: libgomp.info, Node: acc_create, Next: acc_present_or_create, Prev: acc_present_or_copyin, Up: OpenACC Runtime Library Routines 2206 2207 5.20 'acc_create' - Allocate device memory and map it to host memory. 2208 ===================================================================== 2209 2210 _Description_ 2211 This function allocates device memory and maps it to host memory 2212 specified by the host address A with a length of LEN bytes. In 2213 C/C++, the function returns the device address of the allocated 2214 device memory. 2215 2216 In Fortran, two (2) forms are supported. In the first form, A 2217 specifies a contiguous array section. The second form A specifies 2218 a variable or array element and LEN specifies the length in bytes. 2219 2220 _C/C++_: 2221 _Prototype_: 'void *acc_create(h_void *a, size_t len);' 2222 _Prototype_: 'void *acc_create_async(h_void *a, size_t len, int 2223 async);' 2224 2225 _Fortran_: 2226 _Interface_: 'subroutine acc_create(a)' 2227 'type, dimension(:[,:]...) :: a' 2228 _Interface_: 'subroutine acc_create(a, len)' 2229 'type, dimension(:[,:]...) :: a' 2230 'integer len' 2231 _Interface_: 'subroutine acc_create_async(a, async)' 2232 'type, dimension(:[,:]...) :: a' 2233 'integer(acc_handle_kind) :: async' 2234 _Interface_: 'subroutine acc_create_async(a, len, async)' 2235 'type, dimension(:[,:]...) :: a' 2236 'integer len' 2237 'integer(acc_handle_kind) :: async' 2238 2239 _Reference_: 2240 OpenACC specification v2.6 (https://www.openacc.org), section 2241 3.2.21. 2242 2243 2244 File: libgomp.info, Node: acc_present_or_create, Next: acc_copyout, Prev: acc_create, Up: OpenACC Runtime Library Routines 2245 2246 5.21 'acc_present_or_create' - If the data is not present on the device, allocate device memory and map it to host memory. 2247 ========================================================================================================================== 2248 2249 _Description_ 2250 This function tests if the host data specified by A and of length 2251 LEN is present or not. If it is not present, then device memory 2252 will be allocated and mapped to host memory. In C/C++, the device 2253 address of the newly allocated device memory is returned. 2254 2255 In Fortran, two (2) forms are supported. In the first form, A 2256 specifies a contiguous array section. The second form A specifies 2257 a variable or array element and LEN specifies the length in bytes. 2258 2259 Note that 'acc_present_or_create' and 'acc_pcreate' exist for 2260 backward compatibility with OpenACC 2.0; use *note acc_create:: 2261 instead. 2262 2263 _C/C++_: 2264 _Prototype_: 'void *acc_present_or_create(h_void *a, size_t len)' 2265 _Prototype_: 'void *acc_pcreate(h_void *a, size_t len)' 2266 2267 _Fortran_: 2268 _Interface_: 'subroutine acc_present_or_create(a)' 2269 'type, dimension(:[,:]...) :: a' 2270 _Interface_: 'subroutine acc_present_or_create(a, len)' 2271 'type, dimension(:[,:]...) :: a' 2272 'integer len' 2273 _Interface_: 'subroutine acc_pcreate(a)' 2274 'type, dimension(:[,:]...) :: a' 2275 _Interface_: 'subroutine acc_pcreate(a, len)' 2276 'type, dimension(:[,:]...) :: a' 2277 'integer len' 2278 2279 _Reference_: 2280 OpenACC specification v2.6 (https://www.openacc.org), section 2281 3.2.21. 2282 2283 2284 File: libgomp.info, Node: acc_copyout, Next: acc_delete, Prev: acc_present_or_create, Up: OpenACC Runtime Library Routines 2285 2286 5.22 'acc_copyout' - Copy device memory to host memory. 2287 ======================================================= 2288 2289 _Description_ 2290 This function copies mapped device memory to host memory which is 2291 specified by host address A for a length LEN bytes in C/C++. 2292 2293 In Fortran, two (2) forms are supported. In the first form, A 2294 specifies a contiguous array section. The second form A specifies 2295 a variable or array element and LEN specifies the length in bytes. 2296 2297 _C/C++_: 2298 _Prototype_: 'acc_copyout(h_void *a, size_t len);' 2299 _Prototype_: 'acc_copyout_async(h_void *a, size_t len, int async);' 2300 _Prototype_: 'acc_copyout_finalize(h_void *a, size_t len);' 2301 _Prototype_: 'acc_copyout_finalize_async(h_void *a, size_t len, int 2302 async);' 2303 2304 _Fortran_: 2305 _Interface_: 'subroutine acc_copyout(a)' 2306 'type, dimension(:[,:]...) :: a' 2307 _Interface_: 'subroutine acc_copyout(a, len)' 2308 'type, dimension(:[,:]...) :: a' 2309 'integer len' 2310 _Interface_: 'subroutine acc_copyout_async(a, async)' 2311 'type, dimension(:[,:]...) :: a' 2312 'integer(acc_handle_kind) :: async' 2313 _Interface_: 'subroutine acc_copyout_async(a, len, async)' 2314 'type, dimension(:[,:]...) :: a' 2315 'integer len' 2316 'integer(acc_handle_kind) :: async' 2317 _Interface_: 'subroutine acc_copyout_finalize(a)' 2318 'type, dimension(:[,:]...) :: a' 2319 _Interface_: 'subroutine acc_copyout_finalize(a, len)' 2320 'type, dimension(:[,:]...) :: a' 2321 'integer len' 2322 _Interface_: 'subroutine acc_copyout_finalize_async(a, async)' 2323 'type, dimension(:[,:]...) :: a' 2324 'integer(acc_handle_kind) :: async' 2325 _Interface_: 'subroutine acc_copyout_finalize_async(a, len, async)' 2326 'type, dimension(:[,:]...) :: a' 2327 'integer len' 2328 'integer(acc_handle_kind) :: async' 2329 2330 _Reference_: 2331 OpenACC specification v2.6 (https://www.openacc.org), section 2332 3.2.22. 2333 2334 2335 File: libgomp.info, Node: acc_delete, Next: acc_update_device, Prev: acc_copyout, Up: OpenACC Runtime Library Routines 2336 2337 5.23 'acc_delete' - Free device memory. 2338 ======================================= 2339 2340 _Description_ 2341 This function frees previously allocated device memory specified by 2342 the device address A and the length of LEN bytes. 2343 2344 In Fortran, two (2) forms are supported. In the first form, A 2345 specifies a contiguous array section. The second form A specifies 2346 a variable or array element and LEN specifies the length in bytes. 2347 2348 _C/C++_: 2349 _Prototype_: 'acc_delete(h_void *a, size_t len);' 2350 _Prototype_: 'acc_delete_async(h_void *a, size_t len, int async);' 2351 _Prototype_: 'acc_delete_finalize(h_void *a, size_t len);' 2352 _Prototype_: 'acc_delete_finalize_async(h_void *a, size_t len, int 2353 async);' 2354 2355 _Fortran_: 2356 _Interface_: 'subroutine acc_delete(a)' 2357 'type, dimension(:[,:]...) :: a' 2358 _Interface_: 'subroutine acc_delete(a, len)' 2359 'type, dimension(:[,:]...) :: a' 2360 'integer len' 2361 _Interface_: 'subroutine acc_delete_async(a, async)' 2362 'type, dimension(:[,:]...) :: a' 2363 'integer(acc_handle_kind) :: async' 2364 _Interface_: 'subroutine acc_delete_async(a, len, async)' 2365 'type, dimension(:[,:]...) :: a' 2366 'integer len' 2367 'integer(acc_handle_kind) :: async' 2368 _Interface_: 'subroutine acc_delete_finalize(a)' 2369 'type, dimension(:[,:]...) :: a' 2370 _Interface_: 'subroutine acc_delete_finalize(a, len)' 2371 'type, dimension(:[,:]...) :: a' 2372 'integer len' 2373 _Interface_: 'subroutine acc_delete_async_finalize(a, async)' 2374 'type, dimension(:[,:]...) :: a' 2375 'integer(acc_handle_kind) :: async' 2376 _Interface_: 'subroutine acc_delete_async_finalize(a, len, async)' 2377 'type, dimension(:[,:]...) :: a' 2378 'integer len' 2379 'integer(acc_handle_kind) :: async' 2380 2381 _Reference_: 2382 OpenACC specification v2.6 (https://www.openacc.org), section 2383 3.2.23. 2384 2385 2386 File: libgomp.info, Node: acc_update_device, Next: acc_update_self, Prev: acc_delete, Up: OpenACC Runtime Library Routines 2387 2388 5.24 'acc_update_device' - Update device memory from mapped host memory. 2389 ======================================================================== 2390 2391 _Description_ 2392 This function updates the device copy from the previously mapped 2393 host memory. The host memory is specified with the host address A 2394 and a length of LEN bytes. 2395 2396 In Fortran, two (2) forms are supported. In the first form, A 2397 specifies a contiguous array section. The second form A specifies 2398 a variable or array element and LEN specifies the length in bytes. 2399 2400 _C/C++_: 2401 _Prototype_: 'acc_update_device(h_void *a, size_t len);' 2402 _Prototype_: 'acc_update_device(h_void *a, size_t len, async);' 2403 2404 _Fortran_: 2405 _Interface_: 'subroutine acc_update_device(a)' 2406 'type, dimension(:[,:]...) :: a' 2407 _Interface_: 'subroutine acc_update_device(a, len)' 2408 'type, dimension(:[,:]...) :: a' 2409 'integer len' 2410 _Interface_: 'subroutine acc_update_device_async(a, async)' 2411 'type, dimension(:[,:]...) :: a' 2412 'integer(acc_handle_kind) :: async' 2413 _Interface_: 'subroutine acc_update_device_async(a, len, async)' 2414 'type, dimension(:[,:]...) :: a' 2415 'integer len' 2416 'integer(acc_handle_kind) :: async' 2417 2418 _Reference_: 2419 OpenACC specification v2.6 (https://www.openacc.org), section 2420 3.2.24. 2421 2422 2423 File: libgomp.info, Node: acc_update_self, Next: acc_map_data, Prev: acc_update_device, Up: OpenACC Runtime Library Routines 2424 2425 5.25 'acc_update_self' - Update host memory from mapped device memory. 2426 ====================================================================== 2427 2428 _Description_ 2429 This function updates the host copy from the previously mapped 2430 device memory. The host memory is specified with the host address 2431 A and a length of LEN bytes. 2432 2433 In Fortran, two (2) forms are supported. In the first form, A 2434 specifies a contiguous array section. The second form A specifies 2435 a variable or array element and LEN specifies the length in bytes. 2436 2437 _C/C++_: 2438 _Prototype_: 'acc_update_self(h_void *a, size_t len);' 2439 _Prototype_: 'acc_update_self_async(h_void *a, size_t len, int 2440 async);' 2441 2442 _Fortran_: 2443 _Interface_: 'subroutine acc_update_self(a)' 2444 'type, dimension(:[,:]...) :: a' 2445 _Interface_: 'subroutine acc_update_self(a, len)' 2446 'type, dimension(:[,:]...) :: a' 2447 'integer len' 2448 _Interface_: 'subroutine acc_update_self_async(a, async)' 2449 'type, dimension(:[,:]...) :: a' 2450 'integer(acc_handle_kind) :: async' 2451 _Interface_: 'subroutine acc_update_self_async(a, len, async)' 2452 'type, dimension(:[,:]...) :: a' 2453 'integer len' 2454 'integer(acc_handle_kind) :: async' 2455 2456 _Reference_: 2457 OpenACC specification v2.6 (https://www.openacc.org), section 2458 3.2.25. 2459 2460 2461 File: libgomp.info, Node: acc_map_data, Next: acc_unmap_data, Prev: acc_update_self, Up: OpenACC Runtime Library Routines 2462 2463 5.26 'acc_map_data' - Map previously allocated device memory to host memory. 2464 ============================================================================ 2465 2466 _Description_ 2467 This function maps previously allocated device and host memory. 2468 The device memory is specified with the device address D. The host 2469 memory is specified with the host address H and a length of LEN. 2470 2471 _C/C++_: 2472 _Prototype_: 'acc_map_data(h_void *h, d_void *d, size_t len);' 2473 2474 _Reference_: 2475 OpenACC specification v2.6 (https://www.openacc.org), section 2476 3.2.26. 2477 2478 2479 File: libgomp.info, Node: acc_unmap_data, Next: acc_deviceptr, Prev: acc_map_data, Up: OpenACC Runtime Library Routines 2480 2481 5.27 'acc_unmap_data' - Unmap device memory from host memory. 2482 ============================================================= 2483 2484 _Description_ 2485 This function unmaps previously mapped device and host memory. The 2486 latter specified by H. 2487 2488 _C/C++_: 2489 _Prototype_: 'acc_unmap_data(h_void *h);' 2490 2491 _Reference_: 2492 OpenACC specification v2.6 (https://www.openacc.org), section 2493 3.2.27. 2494 2495 2496 File: libgomp.info, Node: acc_deviceptr, Next: acc_hostptr, Prev: acc_unmap_data, Up: OpenACC Runtime Library Routines 2497 2498 5.28 'acc_deviceptr' - Get device pointer associated with specific host address. 2499 ================================================================================ 2500 2501 _Description_ 2502 This function returns the device address that has been mapped to 2503 the host address specified by H. 2504 2505 _C/C++_: 2506 _Prototype_: 'void *acc_deviceptr(h_void *h);' 2507 2508 _Reference_: 2509 OpenACC specification v2.6 (https://www.openacc.org), section 2510 3.2.28. 2511 2512 2513 File: libgomp.info, Node: acc_hostptr, Next: acc_is_present, Prev: acc_deviceptr, Up: OpenACC Runtime Library Routines 2514 2515 5.29 'acc_hostptr' - Get host pointer associated with specific device address. 2516 ============================================================================== 2517 2518 _Description_ 2519 This function returns the host address that has been mapped to the 2520 device address specified by D. 2521 2522 _C/C++_: 2523 _Prototype_: 'void *acc_hostptr(d_void *d);' 2524 2525 _Reference_: 2526 OpenACC specification v2.6 (https://www.openacc.org), section 2527 3.2.29. 2528 2529 2530 File: libgomp.info, Node: acc_is_present, Next: acc_memcpy_to_device, Prev: acc_hostptr, Up: OpenACC Runtime Library Routines 2531 2532 5.30 'acc_is_present' - Indicate whether host variable / array is present on device. 2533 ==================================================================================== 2534 2535 _Description_ 2536 This function indicates whether the specified host address in A and 2537 a length of LEN bytes is present on the device. In C/C++, a 2538 non-zero value is returned to indicate the presence of the mapped 2539 memory on the device. A zero is returned to indicate the memory is 2540 not mapped on the device. 2541 2542 In Fortran, two (2) forms are supported. In the first form, A 2543 specifies a contiguous array section. The second form A specifies 2544 a variable or array element and LEN specifies the length in bytes. 2545 If the host memory is mapped to device memory, then a 'true' is 2546 returned. Otherwise, a 'false' is return to indicate the mapped 2547 memory is not present. 2548 2549 _C/C++_: 2550 _Prototype_: 'int acc_is_present(h_void *a, size_t len);' 2551 2552 _Fortran_: 2553 _Interface_: 'function acc_is_present(a)' 2554 'type, dimension(:[,:]...) :: a' 2555 'logical acc_is_present' 2556 _Interface_: 'function acc_is_present(a, len)' 2557 'type, dimension(:[,:]...) :: a' 2558 'integer len' 2559 'logical acc_is_present' 2560 2561 _Reference_: 2562 OpenACC specification v2.6 (https://www.openacc.org), section 2563 3.2.30. 2564 2565 2566 File: libgomp.info, Node: acc_memcpy_to_device, Next: acc_memcpy_from_device, Prev: acc_is_present, Up: OpenACC Runtime Library Routines 2567 2568 5.31 'acc_memcpy_to_device' - Copy host memory to device memory. 2569 ================================================================ 2570 2571 _Description_ 2572 This function copies host memory specified by host address of SRC 2573 to device memory specified by the device address DEST for a length 2574 of BYTES bytes. 2575 2576 _C/C++_: 2577 _Prototype_: 'acc_memcpy_to_device(d_void *dest, h_void *src, size_t 2578 bytes);' 2579 2580 _Reference_: 2581 OpenACC specification v2.6 (https://www.openacc.org), section 2582 3.2.31. 2583 2584 2585 File: libgomp.info, Node: acc_memcpy_from_device, Next: acc_attach, Prev: acc_memcpy_to_device, Up: OpenACC Runtime Library Routines 2586 2587 5.32 'acc_memcpy_from_device' - Copy device memory to host memory. 2588 ================================================================== 2589 2590 _Description_ 2591 This function copies host memory specified by host address of SRC 2592 from device memory specified by the device address DEST for a 2593 length of BYTES bytes. 2594 2595 _C/C++_: 2596 _Prototype_: 'acc_memcpy_from_device(d_void *dest, h_void *src, 2597 size_t bytes);' 2598 2599 _Reference_: 2600 OpenACC specification v2.6 (https://www.openacc.org), section 2601 3.2.32. 2602 2603 2604 File: libgomp.info, Node: acc_attach, Next: acc_detach, Prev: acc_memcpy_from_device, Up: OpenACC Runtime Library Routines 2605 2606 5.33 'acc_attach' - Let device pointer point to device-pointer target. 2607 ====================================================================== 2608 2609 _Description_ 2610 This function updates a pointer on the device from pointing to a 2611 host-pointer address to pointing to the corresponding device data. 2612 2613 _C/C++_: 2614 _Prototype_: 'acc_attach(h_void **ptr);' 2615 _Prototype_: 'acc_attach_async(h_void **ptr, int async);' 2616 2617 _Reference_: 2618 OpenACC specification v2.6 (https://www.openacc.org), section 2619 3.2.34. 2620 2621 2622 File: libgomp.info, Node: acc_detach, Next: acc_get_current_cuda_device, Prev: acc_attach, Up: OpenACC Runtime Library Routines 2623 2624 5.34 'acc_detach' - Let device pointer point to host-pointer target. 2625 ==================================================================== 2626 2627 _Description_ 2628 This function updates a pointer on the device from pointing to a 2629 device-pointer address to pointing to the corresponding host data. 2630 2631 _C/C++_: 2632 _Prototype_: 'acc_detach(h_void **ptr);' 2633 _Prototype_: 'acc_detach_async(h_void **ptr, int async);' 2634 _Prototype_: 'acc_detach_finalize(h_void **ptr);' 2635 _Prototype_: 'acc_detach_finalize_async(h_void **ptr, int async);' 2636 2637 _Reference_: 2638 OpenACC specification v2.6 (https://www.openacc.org), section 2639 3.2.35. 2640 2641 2642 File: libgomp.info, Node: acc_get_current_cuda_device, Next: acc_get_current_cuda_context, Prev: acc_detach, Up: OpenACC Runtime Library Routines 2643 2644 5.35 'acc_get_current_cuda_device' - Get CUDA device handle. 2645 ============================================================ 2646 2647 _Description_ 2648 This function returns the CUDA device handle. This handle is the 2649 same as used by the CUDA Runtime or Driver API's. 2650 2651 _C/C++_: 2652 _Prototype_: 'void *acc_get_current_cuda_device(void);' 2653 2654 _Reference_: 2655 OpenACC specification v2.6 (https://www.openacc.org), section 2656 A.2.1.1. 2657 2658 2659 File: libgomp.info, Node: acc_get_current_cuda_context, Next: acc_get_cuda_stream, Prev: acc_get_current_cuda_device, Up: OpenACC Runtime Library Routines 2660 2661 5.36 'acc_get_current_cuda_context' - Get CUDA context handle. 2662 ============================================================== 2663 2664 _Description_ 2665 This function returns the CUDA context handle. This handle is the 2666 same as used by the CUDA Runtime or Driver API's. 2667 2668 _C/C++_: 2669 _Prototype_: 'void *acc_get_current_cuda_context(void);' 2670 2671 _Reference_: 2672 OpenACC specification v2.6 (https://www.openacc.org), section 2673 A.2.1.2. 2674 2675 2676 File: libgomp.info, Node: acc_get_cuda_stream, Next: acc_set_cuda_stream, Prev: acc_get_current_cuda_context, Up: OpenACC Runtime Library Routines 2677 2678 5.37 'acc_get_cuda_stream' - Get CUDA stream handle. 2679 ==================================================== 2680 2681 _Description_ 2682 This function returns the CUDA stream handle for the queue ASYNC. 2683 This handle is the same as used by the CUDA Runtime or Driver 2684 API's. 2685 2686 _C/C++_: 2687 _Prototype_: 'void *acc_get_cuda_stream(int async);' 2688 2689 _Reference_: 2690 OpenACC specification v2.6 (https://www.openacc.org), section 2691 A.2.1.3. 2692 2693 2694 File: libgomp.info, Node: acc_set_cuda_stream, Next: acc_prof_register, Prev: acc_get_cuda_stream, Up: OpenACC Runtime Library Routines 2695 2696 5.38 'acc_set_cuda_stream' - Set CUDA stream handle. 2697 ==================================================== 2698 2699 _Description_ 2700 This function associates the stream handle specified by STREAM with 2701 the queue ASYNC. 2702 2703 This cannot be used to change the stream handle associated with 2704 'acc_async_sync'. 2705 2706 The return value is not specified. 2707 2708 _C/C++_: 2709 _Prototype_: 'int acc_set_cuda_stream(int async, void *stream);' 2710 2711 _Reference_: 2712 OpenACC specification v2.6 (https://www.openacc.org), section 2713 A.2.1.4. 2714 2715 2716 File: libgomp.info, Node: acc_prof_register, Next: acc_prof_unregister, Prev: acc_set_cuda_stream, Up: OpenACC Runtime Library Routines 2717 2718 5.39 'acc_prof_register' - Register callbacks. 2719 ============================================== 2720 2721 _Description_: 2722 This function registers callbacks. 2723 2724 _C/C++_: 2725 _Prototype_: 'void acc_prof_register (acc_event_t, acc_prof_callback, 2726 acc_register_t);' 2727 2728 _See also_: 2729 *note OpenACC Profiling Interface:: 2730 2731 _Reference_: 2732 OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 2733 2734 2735 File: libgomp.info, Node: acc_prof_unregister, Next: acc_prof_lookup, Prev: acc_prof_register, Up: OpenACC Runtime Library Routines 2736 2737 5.40 'acc_prof_unregister' - Unregister callbacks. 2738 ================================================== 2739 2740 _Description_: 2741 This function unregisters callbacks. 2742 2743 _C/C++_: 2744 _Prototype_: 'void acc_prof_unregister (acc_event_t, 2745 acc_prof_callback, acc_register_t);' 2746 2747 _See also_: 2748 *note OpenACC Profiling Interface:: 2749 2750 _Reference_: 2751 OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 2752 2753 2754 File: libgomp.info, Node: acc_prof_lookup, Next: acc_register_library, Prev: acc_prof_unregister, Up: OpenACC Runtime Library Routines 2755 2756 5.41 'acc_prof_lookup' - Obtain inquiry functions. 2757 ================================================== 2758 2759 _Description_: 2760 Function to obtain inquiry functions. 2761 2762 _C/C++_: 2763 _Prototype_: 'acc_query_fn acc_prof_lookup (const char *);' 2764 2765 _See also_: 2766 *note OpenACC Profiling Interface:: 2767 2768 _Reference_: 2769 OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 2770 2771 2772 File: libgomp.info, Node: acc_register_library, Prev: acc_prof_lookup, Up: OpenACC Runtime Library Routines 2773 2774 5.42 'acc_register_library' - Library registration. 2775 =================================================== 2776 2777 _Description_: 2778 Function for library registration. 2779 2780 _C/C++_: 2781 _Prototype_: 'void acc_register_library (acc_prof_reg, acc_prof_reg, 2782 acc_prof_lookup_func);' 2783 2784 _See also_: 2785 *note OpenACC Profiling Interface::, *note ACC_PROFLIB:: 2786 2787 _Reference_: 2788 OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 2789 2790 2791 File: libgomp.info, Node: OpenACC Environment Variables, Next: CUDA Streams Usage, Prev: OpenACC Runtime Library Routines, Up: Top 2792 2793 6 OpenACC Environment Variables 2794 ******************************* 2795 2796 The variables 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM' are defined by 2797 section 4 of the OpenACC specification in version 2.0. The variable 2798 'ACC_PROFLIB' is defined by section 4 of the OpenACC specification in 2799 version 2.6. The variable 'GCC_ACC_NOTIFY' is used for diagnostic 2800 purposes. 2801 2802 * Menu: 2803 2804 * ACC_DEVICE_TYPE:: 2805 * ACC_DEVICE_NUM:: 2806 * ACC_PROFLIB:: 2807 * GCC_ACC_NOTIFY:: 2808 2809 2810 File: libgomp.info, Node: ACC_DEVICE_TYPE, Next: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2811 2812 6.1 'ACC_DEVICE_TYPE' 2813 ===================== 2814 2815 _Reference_: 2816 OpenACC specification v2.6 (https://www.openacc.org), section 4.1. 2817 2818 2819 File: libgomp.info, Node: ACC_DEVICE_NUM, Next: ACC_PROFLIB, Prev: ACC_DEVICE_TYPE, Up: OpenACC Environment Variables 2820 2821 6.2 'ACC_DEVICE_NUM' 2822 ==================== 2823 2824 _Reference_: 2825 OpenACC specification v2.6 (https://www.openacc.org), section 4.2. 2826 2827 2828 File: libgomp.info, Node: ACC_PROFLIB, Next: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2829 2830 6.3 'ACC_PROFLIB' 2831 ================= 2832 2833 _See also_: 2834 *note acc_register_library::, *note OpenACC Profiling Interface:: 2835 2836 _Reference_: 2837 OpenACC specification v2.6 (https://www.openacc.org), section 4.3. 2838 2839 2840 File: libgomp.info, Node: GCC_ACC_NOTIFY, Prev: ACC_PROFLIB, Up: OpenACC Environment Variables 2841 2842 6.4 'GCC_ACC_NOTIFY' 2843 ==================== 2844 2845 _Description_: 2846 Print debug information pertaining to the accelerator. 2847 2848 2849 File: libgomp.info, Node: CUDA Streams Usage, Next: OpenACC Library Interoperability, Prev: OpenACC Environment Variables, Up: Top 2850 2851 7 CUDA Streams Usage 2852 ******************** 2853 2854 This applies to the 'nvptx' plugin only. 2855 2856 The library provides elements that perform asynchronous movement of 2857 data and asynchronous operation of computing constructs. This 2858 asynchronous functionality is implemented by making use of CUDA 2859 streams(1). 2860 2861 The primary means by that the asynchronous functionality is accessed 2862 is through the use of those OpenACC directives which make use of the 2863 'async' and 'wait' clauses. When the 'async' clause is first used with 2864 a directive, it creates a CUDA stream. If an 'async-argument' is used 2865 with the 'async' clause, then the stream is associated with the 2866 specified 'async-argument'. 2867 2868 Following the creation of an association between a CUDA stream and 2869 the 'async-argument' of an 'async' clause, both the 'wait' clause and 2870 the 'wait' directive can be used. When either the clause or directive 2871 is used after stream creation, it creates a rendezvous point whereby 2872 execution waits until all operations associated with the 2873 'async-argument', that is, stream, have completed. 2874 2875 Normally, the management of the streams that are created as a result 2876 of using the 'async' clause, is done without any intervention by the 2877 caller. This implies the association between the 'async-argument' and 2878 the CUDA stream will be maintained for the lifetime of the program. 2879 However, this association can be changed through the use of the library 2880 function 'acc_set_cuda_stream'. When the function 'acc_set_cuda_stream' 2881 is called, the CUDA stream that was originally associated with the 2882 'async' clause will be destroyed. Caution should be taken when changing 2883 the association as subsequent references to the 'async-argument' refer 2884 to a different CUDA stream. 2885 2886 ---------- Footnotes ---------- 2887 2888 (1) See "Stream Management" in "CUDA Driver API", TRM-06703-001, 2889 Version 5.5, for additional information 2890 2891 2892 File: libgomp.info, Node: OpenACC Library Interoperability, Next: OpenACC Profiling Interface, Prev: CUDA Streams Usage, Up: Top 2893 2894 8 OpenACC Library Interoperability 2895 ********************************** 2896 2897 8.1 Introduction 2898 ================ 2899 2900 The OpenACC library uses the CUDA Driver API, and may interact with 2901 programs that use the Runtime library directly, or another library based 2902 on the Runtime library, e.g., CUBLAS(1). This chapter describes the use 2903 cases and what changes are required in order to use both the OpenACC 2904 library and the CUBLAS and Runtime libraries within a program. 2905 2906 8.2 First invocation: NVIDIA CUBLAS library API 2907 =============================================== 2908 2909 In this first use case (see below), a function in the CUBLAS library is 2910 called prior to any of the functions in the OpenACC library. More 2911 specifically, the function 'cublasCreate()'. 2912 2913 When invoked, the function initializes the library and allocates the 2914 hardware resources on the host and the device on behalf of the caller. 2915 Once the initialization and allocation has completed, a handle is 2916 returned to the caller. The OpenACC library also requires 2917 initialization and allocation of hardware resources. Since the CUBLAS 2918 library has already allocated the hardware resources for the device, all 2919 that is left to do is to initialize the OpenACC library and acquire the 2920 hardware resources on the host. 2921 2922 Prior to calling the OpenACC function that initializes the library 2923 and allocate the host hardware resources, you need to acquire the device 2924 number that was allocated during the call to 'cublasCreate()'. The 2925 invoking of the runtime library function 'cudaGetDevice()' accomplishes 2926 this. Once acquired, the device number is passed along with the device 2927 type as parameters to the OpenACC library function 2928 'acc_set_device_num()'. 2929 2930 Once the call to 'acc_set_device_num()' has completed, the OpenACC 2931 library uses the context that was created during the call to 2932 'cublasCreate()'. In other words, both libraries will be sharing the 2933 same context. 2934 2935 /* Create the handle */ 2936 s = cublasCreate(&h); 2937 if (s != CUBLAS_STATUS_SUCCESS) 2938 { 2939 fprintf(stderr, "cublasCreate failed %d\n", s); 2940 exit(EXIT_FAILURE); 2941 } 2942 2943 /* Get the device number */ 2944 e = cudaGetDevice(&dev); 2945 if (e != cudaSuccess) 2946 { 2947 fprintf(stderr, "cudaGetDevice failed %d\n", e); 2948 exit(EXIT_FAILURE); 2949 } 2950 2951 /* Initialize OpenACC library and use device 'dev' */ 2952 acc_set_device_num(dev, acc_device_nvidia); 2953 2954 Use Case 1 2955 2956 8.3 First invocation: OpenACC library API 2957 ========================================= 2958 2959 In this second use case (see below), a function in the OpenACC library 2960 is called prior to any of the functions in the CUBLAS library. More 2961 specificially, the function 'acc_set_device_num()'. 2962 2963 In the use case presented here, the function 'acc_set_device_num()' 2964 is used to both initialize the OpenACC library and allocate the hardware 2965 resources on the host and the device. In the call to the function, the 2966 call parameters specify which device to use and what device type to use, 2967 i.e., 'acc_device_nvidia'. It should be noted that this is but one 2968 method to initialize the OpenACC library and allocate the appropriate 2969 hardware resources. Other methods are available through the use of 2970 environment variables and these will be discussed in the next section. 2971 2972 Once the call to 'acc_set_device_num()' has completed, other OpenACC 2973 functions can be called as seen with multiple calls being made to 2974 'acc_copyin()'. In addition, calls can be made to functions in the 2975 CUBLAS library. In the use case a call to 'cublasCreate()' is made 2976 subsequent to the calls to 'acc_copyin()'. As seen in the previous use 2977 case, a call to 'cublasCreate()' initializes the CUBLAS library and 2978 allocates the hardware resources on the host and the device. However, 2979 since the device has already been allocated, 'cublasCreate()' will only 2980 initialize the CUBLAS library and allocate the appropriate hardware 2981 resources on the host. The context that was created as part of the 2982 OpenACC initialization is shared with the CUBLAS library, similarly to 2983 the first use case. 2984 2985 dev = 0; 2986 2987 acc_set_device_num(dev, acc_device_nvidia); 2988 2989 /* Copy the first set to the device */ 2990 d_X = acc_copyin(&h_X[0], N * sizeof (float)); 2991 if (d_X == NULL) 2992 { 2993 fprintf(stderr, "copyin error h_X\n"); 2994 exit(EXIT_FAILURE); 2995 } 2996 2997 /* Copy the second set to the device */ 2998 d_Y = acc_copyin(&h_Y1[0], N * sizeof (float)); 2999 if (d_Y == NULL) 3000 { 3001 fprintf(stderr, "copyin error h_Y1\n"); 3002 exit(EXIT_FAILURE); 3003 } 3004 3005 /* Create the handle */ 3006 s = cublasCreate(&h); 3007 if (s != CUBLAS_STATUS_SUCCESS) 3008 { 3009 fprintf(stderr, "cublasCreate failed %d\n", s); 3010 exit(EXIT_FAILURE); 3011 } 3012 3013 /* Perform saxpy using CUBLAS library function */ 3014 s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1); 3015 if (s != CUBLAS_STATUS_SUCCESS) 3016 { 3017 fprintf(stderr, "cublasSaxpy failed %d\n", s); 3018 exit(EXIT_FAILURE); 3019 } 3020 3021 /* Copy the results from the device */ 3022 acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float)); 3023 3024 Use Case 2 3025 3026 8.4 OpenACC library and environment variables 3027 ============================================= 3028 3029 There are two environment variables associated with the OpenACC library 3030 that may be used to control the device type and device number: 3031 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM', respectively. These two 3032 environment variables can be used as an alternative to calling 3033 'acc_set_device_num()'. As seen in the second use case, the device type 3034 and device number were specified using 'acc_set_device_num()'. If 3035 however, the aforementioned environment variables were set, then the 3036 call to 'acc_set_device_num()' would not be required. 3037 3038 The use of the environment variables is only relevant when an OpenACC 3039 function is called prior to a call to 'cudaCreate()'. If 'cudaCreate()' 3040 is called prior to a call to an OpenACC function, then you must call 3041 'acc_set_device_num()'(2) 3042 3043 ---------- Footnotes ---------- 3044 3045 (1) See section 2.26, "Interactions with the CUDA Driver API" in 3046 "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU 3047 Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5, for 3048 additional information on library interoperability. 3049 3050 (2) More complete information about 'ACC_DEVICE_TYPE' and 3051 'ACC_DEVICE_NUM' can be found in sections 4.1 and 4.2 of the OpenACC 3052 (https://www.openacc.org) Application Programming Interface, Version 3053 2.6. 3054 3055 3056 File: libgomp.info, Node: OpenACC Profiling Interface, Next: The libgomp ABI, Prev: OpenACC Library Interoperability, Up: Top 3057 3058 9 OpenACC Profiling Interface 3059 ***************************** 3060 3061 9.1 Implementation Status and Implementation-Defined Behavior 3062 ============================================================= 3063 3064 We're implementing the OpenACC Profiling Interface as defined by the 3065 OpenACC 2.6 specification. We're clarifying some aspects here as 3066 _implementation-defined behavior_, while they're still under discussion 3067 within the OpenACC Technical Committee. 3068 3069 This implementation is tuned to keep the performance impact as low as 3070 possible for the (very common) case that the Profiling Interface is not 3071 enabled. This is relevant, as the Profiling Interface affects all the 3072 _hot_ code paths (in the target code, not in the offloaded code). Users 3073 of the OpenACC Profiling Interface can be expected to understand that 3074 performance will be impacted to some degree once the Profiling Interface 3075 has gotten enabled: for example, because of the _runtime_ (libgomp) 3076 calling into a third-party _library_ for every event that has been 3077 registered. 3078 3079 We're not yet accounting for the fact that 'OpenACC events may occur 3080 during event processing'. We just handle one case specially, as 3081 required by CUDA 9.0 'nvprof', that 'acc_get_device_type' (*note 3082 acc_get_device_type::)) may be called from 'acc_ev_device_init_start', 3083 'acc_ev_device_init_end' callbacks. 3084 3085 We're not yet implementing initialization via a 3086 'acc_register_library' function that is either statically linked in, or 3087 dynamically via 'LD_PRELOAD'. Initialization via 'acc_register_library' 3088 functions dynamically loaded via the 'ACC_PROFLIB' environment variable 3089 does work, as does directly calling 'acc_prof_register', 3090 'acc_prof_unregister', 'acc_prof_lookup'. 3091 3092 As currently there are no inquiry functions defined, calls to 3093 'acc_prof_lookup' will always return 'NULL'. 3094 3095 There aren't separate _start_, _stop_ events defined for the event 3096 types 'acc_ev_create', 'acc_ev_delete', 'acc_ev_alloc', 'acc_ev_free'. 3097 It's not clear if these should be triggered before or after the actual 3098 device-specific call is made. We trigger them after. 3099 3100 Remarks about data provided to callbacks: 3101 3102 'acc_prof_info.event_type' 3103 It's not clear if for _nested_ event callbacks (for example, 3104 'acc_ev_enqueue_launch_start' as part of a parent compute 3105 construct), this should be set for the nested event 3106 ('acc_ev_enqueue_launch_start'), or if the value of the parent 3107 construct should remain ('acc_ev_compute_construct_start'). In 3108 this implementation, the value will generally correspond to the 3109 innermost nested event type. 3110 3111 'acc_prof_info.device_type' 3112 3113 * For 'acc_ev_compute_construct_start', and in presence of an 3114 'if' clause with _false_ argument, this will still refer to 3115 the offloading device type. It's not clear if that's the 3116 expected behavior. 3117 3118 * Complementary to the item before, for 3119 'acc_ev_compute_construct_end', this is set to 3120 'acc_device_host' in presence of an 'if' clause with _false_ 3121 argument. It's not clear if that's the expected behavior. 3122 3123 'acc_prof_info.thread_id' 3124 Always '-1'; not yet implemented. 3125 3126 'acc_prof_info.async' 3127 3128 * Not yet implemented correctly for 3129 'acc_ev_compute_construct_start'. 3130 3131 * In a compute construct, for host-fallback 3132 execution/'acc_device_host' it will always be 3133 'acc_async_sync'. It's not clear if that's the expected 3134 behavior. 3135 3136 * For 'acc_ev_device_init_start' and 'acc_ev_device_init_end', 3137 it will always be 'acc_async_sync'. It's not clear if that's 3138 the expected behavior. 3139 3140 'acc_prof_info.async_queue' 3141 There is no 'limited number of asynchronous queues' in libgomp. 3142 This will always have the same value as 'acc_prof_info.async'. 3143 3144 'acc_prof_info.src_file' 3145 Always 'NULL'; not yet implemented. 3146 3147 'acc_prof_info.func_name' 3148 Always 'NULL'; not yet implemented. 3149 3150 'acc_prof_info.line_no' 3151 Always '-1'; not yet implemented. 3152 3153 'acc_prof_info.end_line_no' 3154 Always '-1'; not yet implemented. 3155 3156 'acc_prof_info.func_line_no' 3157 Always '-1'; not yet implemented. 3158 3159 'acc_prof_info.func_end_line_no' 3160 Always '-1'; not yet implemented. 3161 3162 'acc_event_info.event_type', 'acc_event_info.*.event_type' 3163 Relating to 'acc_prof_info.event_type' discussed above, in this 3164 implementation, this will always be the same value as 3165 'acc_prof_info.event_type'. 3166 3167 'acc_event_info.*.parent_construct' 3168 3169 * Will be 'acc_construct_parallel' for all OpenACC compute 3170 constructs as well as many OpenACC Runtime API calls; should 3171 be the one matching the actual construct, or 3172 'acc_construct_runtime_api', respectively. 3173 3174 * Will be 'acc_construct_enter_data' or 3175 'acc_construct_exit_data' when processing variable mappings 3176 specified in OpenACC _declare_ directives; should be 3177 'acc_construct_declare'. 3178 3179 * For implicit 'acc_ev_device_init_start', 3180 'acc_ev_device_init_end', and explicit as well as implicit 3181 'acc_ev_alloc', 'acc_ev_free', 'acc_ev_enqueue_upload_start', 3182 'acc_ev_enqueue_upload_end', 'acc_ev_enqueue_download_start', 3183 and 'acc_ev_enqueue_download_end', will be 3184 'acc_construct_parallel'; should reflect the real parent 3185 construct. 3186 3187 'acc_event_info.*.implicit' 3188 For 'acc_ev_alloc', 'acc_ev_free', 'acc_ev_enqueue_upload_start', 3189 'acc_ev_enqueue_upload_end', 'acc_ev_enqueue_download_start', and 3190 'acc_ev_enqueue_download_end', this currently will be '1' also for 3191 explicit usage. 3192 3193 'acc_event_info.data_event.var_name' 3194 Always 'NULL'; not yet implemented. 3195 3196 'acc_event_info.data_event.host_ptr' 3197 For 'acc_ev_alloc', and 'acc_ev_free', this is always 'NULL'. 3198 3199 'typedef union acc_api_info' 3200 ... as printed in '5.2.3. Third Argument: API-Specific 3201 Information'. This should obviously be 'typedef _struct_ 3202 acc_api_info'. 3203 3204 'acc_api_info.device_api' 3205 Possibly not yet implemented correctly for 3206 'acc_ev_compute_construct_start', 'acc_ev_device_init_start', 3207 'acc_ev_device_init_end': will always be 'acc_device_api_none' for 3208 these event types. For 'acc_ev_enter_data_start', it will be 3209 'acc_device_api_none' in some cases. 3210 3211 'acc_api_info.device_type' 3212 Always the same as 'acc_prof_info.device_type'. 3213 3214 'acc_api_info.vendor' 3215 Always '-1'; not yet implemented. 3216 3217 'acc_api_info.device_handle' 3218 Always 'NULL'; not yet implemented. 3219 3220 'acc_api_info.context_handle' 3221 Always 'NULL'; not yet implemented. 3222 3223 'acc_api_info.async_handle' 3224 Always 'NULL'; not yet implemented. 3225 3226 Remarks about certain event types: 3227 3228 'acc_ev_device_init_start', 'acc_ev_device_init_end' 3229 3230 * Whan a compute construct triggers implicit 3231 'acc_ev_device_init_start' and 'acc_ev_device_init_end' 3232 events, they currently aren't _nested within_ the 3233 corresponding 'acc_ev_compute_construct_start' and 3234 'acc_ev_compute_construct_end', but they're currently observed 3235 _before_ 'acc_ev_compute_construct_start'. It's not clear 3236 what to do: the standard asks us provide a lot of details to 3237 the 'acc_ev_compute_construct_start' callback, without 3238 (implicitly) initializing a device before? 3239 3240 * Callbacks for these event types will not be invoked for calls 3241 to the 'acc_set_device_type' and 'acc_set_device_num' 3242 functions. It's not clear if they should be. 3243 3244 'acc_ev_enter_data_start', 'acc_ev_enter_data_end', 'acc_ev_exit_data_start', 'acc_ev_exit_data_end' 3245 3246 * Callbacks for these event types will also be invoked for 3247 OpenACC _host_data_ constructs. It's not clear if they should 3248 be. 3249 3250 * Callbacks for these event types will also be invoked when 3251 processing variable mappings specified in OpenACC _declare_ 3252 directives. It's not clear if they should be. 3253 3254 Callbacks for the following event types will be invoked, but dispatch 3255 and information provided therein has not yet been thoroughly reviewed: 3256 3257 * 'acc_ev_alloc' 3258 * 'acc_ev_free' 3259 * 'acc_ev_update_start', 'acc_ev_update_end' 3260 * 'acc_ev_enqueue_upload_start', 'acc_ev_enqueue_upload_end' 3261 * 'acc_ev_enqueue_download_start', 'acc_ev_enqueue_download_end' 3262 3263 During device initialization, and finalization, respectively, 3264 callbacks for the following event types will not yet be invoked: 3265 3266 * 'acc_ev_alloc' 3267 * 'acc_ev_free' 3268 3269 Callbacks for the following event types have not yet been 3270 implemented, so currently won't be invoked: 3271 3272 * 'acc_ev_device_shutdown_start', 'acc_ev_device_shutdown_end' 3273 * 'acc_ev_runtime_shutdown' 3274 * 'acc_ev_create', 'acc_ev_delete' 3275 * 'acc_ev_wait_start', 'acc_ev_wait_end' 3276 3277 For the following runtime library functions, not all expected 3278 callbacks will be invoked (mostly concerning implicit device 3279 initialization): 3280 3281 * 'acc_get_num_devices' 3282 * 'acc_set_device_type' 3283 * 'acc_get_device_type' 3284 * 'acc_set_device_num' 3285 * 'acc_get_device_num' 3286 * 'acc_init' 3287 * 'acc_shutdown' 3288 3289 Aside from implicit device initialization, for the following runtime 3290 library functions, no callbacks will be invoked for shared-memory 3291 offloading devices (it's not clear if they should be): 3292 3293 * 'acc_malloc' 3294 * 'acc_free' 3295 * 'acc_copyin', 'acc_present_or_copyin', 'acc_copyin_async' 3296 * 'acc_create', 'acc_present_or_create', 'acc_create_async' 3297 * 'acc_copyout', 'acc_copyout_async', 'acc_copyout_finalize', 3298 'acc_copyout_finalize_async' 3299 * 'acc_delete', 'acc_delete_async', 'acc_delete_finalize', 3300 'acc_delete_finalize_async' 3301 * 'acc_update_device', 'acc_update_device_async' 3302 * 'acc_update_self', 'acc_update_self_async' 3303 * 'acc_map_data', 'acc_unmap_data' 3304 * 'acc_memcpy_to_device', 'acc_memcpy_to_device_async' 3305 * 'acc_memcpy_from_device', 'acc_memcpy_from_device_async' 3306 3307 3308 File: libgomp.info, Node: The libgomp ABI, Next: Reporting Bugs, Prev: OpenACC Profiling Interface, Up: Top 3309 3310 10 The libgomp ABI 3311 ****************** 3312 3313 The following sections present notes on the external ABI as presented by 3314 libgomp. Only maintainers should need them. 3315 3316 * Menu: 3317 3318 * Implementing MASTER construct:: 3319 * Implementing CRITICAL construct:: 3320 * Implementing ATOMIC construct:: 3321 * Implementing FLUSH construct:: 3322 * Implementing BARRIER construct:: 3323 * Implementing THREADPRIVATE construct:: 3324 * Implementing PRIVATE clause:: 3325 * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses:: 3326 * Implementing REDUCTION clause:: 3327 * Implementing PARALLEL construct:: 3328 * Implementing FOR construct:: 3329 * Implementing ORDERED construct:: 3330 * Implementing SECTIONS construct:: 3331 * Implementing SINGLE construct:: 3332 * Implementing OpenACC's PARALLEL construct:: 3333 3334 3335 File: libgomp.info, Node: Implementing MASTER construct, Next: Implementing CRITICAL construct, Up: The libgomp ABI 3336 3337 10.1 Implementing MASTER construct 3338 ================================== 3339 3340 if (omp_get_thread_num () == 0) 3341 block 3342 3343 Alternately, we generate two copies of the parallel subfunction and 3344 only include this in the version run by the master thread. Surely this 3345 is not worthwhile though... 3346 3347 3348 File: libgomp.info, Node: Implementing CRITICAL construct, Next: Implementing ATOMIC construct, Prev: Implementing MASTER construct, Up: The libgomp ABI 3349 3350 10.2 Implementing CRITICAL construct 3351 ==================================== 3352 3353 Without a specified name, 3354 3355 void GOMP_critical_start (void); 3356 void GOMP_critical_end (void); 3357 3358 so that we don't get COPY relocations from libgomp to the main 3359 application. 3360 3361 With a specified name, use omp_set_lock and omp_unset_lock with name 3362 being transformed into a variable declared like 3363 3364 omp_lock_t gomp_critical_user_<name> __attribute__((common)) 3365 3366 Ideally the ABI would specify that all zero is a valid unlocked 3367 state, and so we wouldn't need to initialize this at startup. 3368 3369 3370 File: libgomp.info, Node: Implementing ATOMIC construct, Next: Implementing FLUSH construct, Prev: Implementing CRITICAL construct, Up: The libgomp ABI 3371 3372 10.3 Implementing ATOMIC construct 3373 ================================== 3374 3375 The target should implement the '__sync' builtins. 3376 3377 Failing that we could add 3378 3379 void GOMP_atomic_enter (void) 3380 void GOMP_atomic_exit (void) 3381 3382 which reuses the regular lock code, but with yet another lock object 3383 private to the library. 3384 3385 3386 File: libgomp.info, Node: Implementing FLUSH construct, Next: Implementing BARRIER construct, Prev: Implementing ATOMIC construct, Up: The libgomp ABI 3387 3388 10.4 Implementing FLUSH construct 3389 ================================= 3390 3391 Expands to the '__sync_synchronize' builtin. 3392 3393 3394 File: libgomp.info, Node: Implementing BARRIER construct, Next: Implementing THREADPRIVATE construct, Prev: Implementing FLUSH construct, Up: The libgomp ABI 3395 3396 10.5 Implementing BARRIER construct 3397 =================================== 3398 3399 void GOMP_barrier (void) 3400 3401 3402 File: libgomp.info, Node: Implementing THREADPRIVATE construct, Next: Implementing PRIVATE clause, Prev: Implementing BARRIER construct, Up: The libgomp ABI 3403 3404 10.6 Implementing THREADPRIVATE construct 3405 ========================================= 3406 3407 In _most_ cases we can map this directly to '__thread'. Except that OMP 3408 allows constructors for C++ objects. We can either refuse to support 3409 this (how often is it used?) or we can implement something akin to 3410 .ctors. 3411 3412 Even more ideally, this ctor feature is handled by extensions to the 3413 main pthreads library. Failing that, we can have a set of entry points 3414 to register ctor functions to be called. 3415 3416 3417 File: libgomp.info, Node: Implementing PRIVATE clause, Next: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Prev: Implementing THREADPRIVATE construct, Up: The libgomp ABI 3418 3419 10.7 Implementing PRIVATE clause 3420 ================================ 3421 3422 In association with a PARALLEL, or within the lexical extent of a 3423 PARALLEL block, the variable becomes a local variable in the parallel 3424 subfunction. 3425 3426 In association with FOR or SECTIONS blocks, create a new automatic 3427 variable within the current function. This preserves the semantic of 3428 new variable creation. 3429 3430 3431 File: libgomp.info, Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Next: Implementing REDUCTION clause, Prev: Implementing PRIVATE clause, Up: The libgomp ABI 3432 3433 10.8 Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses 3434 ========================================================================= 3435 3436 This seems simple enough for PARALLEL blocks. Create a private struct 3437 for communicating between the parent and subfunction. In the parent, 3438 copy in values for scalar and "small" structs; copy in addresses for 3439 others TREE_ADDRESSABLE types. In the subfunction, copy the value into 3440 the local variable. 3441 3442 It is not clear what to do with bare FOR or SECTION blocks. The only 3443 thing I can figure is that we do something like: 3444 3445 #pragma omp for firstprivate(x) lastprivate(y) 3446 for (int i = 0; i < n; ++i) 3447 body; 3448 3449 which becomes 3450 3451 { 3452 int x = x, y; 3453 3454 // for stuff 3455 3456 if (i == n) 3457 y = y; 3458 } 3459 3460 where the "x=x" and "y=y" assignments actually have different uids 3461 for the two variables, i.e. not something you could write directly in 3462 C. Presumably this only makes sense if the "outer" x and y are global 3463 variables. 3464 3465 COPYPRIVATE would work the same way, except the structure broadcast 3466 would have to happen via SINGLE machinery instead. 3467 3468 3469 File: libgomp.info, Node: Implementing REDUCTION clause, Next: Implementing PARALLEL construct, Prev: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Up: The libgomp ABI 3470 3471 10.9 Implementing REDUCTION clause 3472 ================================== 3473 3474 The private struct mentioned in the previous section should have a 3475 pointer to an array of the type of the variable, indexed by the thread's 3476 TEAM_ID. The thread stores its final value into the array, and after 3477 the barrier, the master thread iterates over the array to collect the 3478 values. 3479 3480 3481 File: libgomp.info, Node: Implementing PARALLEL construct, Next: Implementing FOR construct, Prev: Implementing REDUCTION clause, Up: The libgomp ABI 3482 3483 10.10 Implementing PARALLEL construct 3484 ===================================== 3485 3486 #pragma omp parallel 3487 { 3488 body; 3489 } 3490 3491 becomes 3492 3493 void subfunction (void *data) 3494 { 3495 use data; 3496 body; 3497 } 3498 3499 setup data; 3500 GOMP_parallel_start (subfunction, &data, num_threads); 3501 subfunction (&data); 3502 GOMP_parallel_end (); 3503 3504 void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads) 3505 3506 The FN argument is the subfunction to be run in parallel. 3507 3508 The DATA argument is a pointer to a structure used to communicate 3509 data in and out of the subfunction, as discussed above with respect to 3510 FIRSTPRIVATE et al. 3511 3512 The NUM_THREADS argument is 1 if an IF clause is present and false, 3513 or the value of the NUM_THREADS clause, if present, or 0. 3514 3515 The function needs to create the appropriate number of threads and/or 3516 launch them from the dock. It needs to create the team structure and 3517 assign team ids. 3518 3519 void GOMP_parallel_end (void) 3520 3521 Tears down the team and returns us to the previous 3522 'omp_in_parallel()' state. 3523 3524 3525 File: libgomp.info, Node: Implementing FOR construct, Next: Implementing ORDERED construct, Prev: Implementing PARALLEL construct, Up: The libgomp ABI 3526 3527 10.11 Implementing FOR construct 3528 ================================ 3529 3530 #pragma omp parallel for 3531 for (i = lb; i <= ub; i++) 3532 body; 3533 3534 becomes 3535 3536 void subfunction (void *data) 3537 { 3538 long _s0, _e0; 3539 while (GOMP_loop_static_next (&_s0, &_e0)) 3540 { 3541 long _e1 = _e0, i; 3542 for (i = _s0; i < _e1; i++) 3543 body; 3544 } 3545 GOMP_loop_end_nowait (); 3546 } 3547 3548 GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0); 3549 subfunction (NULL); 3550 GOMP_parallel_end (); 3551 3552 #pragma omp for schedule(runtime) 3553 for (i = 0; i < n; i++) 3554 body; 3555 3556 becomes 3557 3558 { 3559 long i, _s0, _e0; 3560 if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0)) 3561 do { 3562 long _e1 = _e0; 3563 for (i = _s0, i < _e0; i++) 3564 body; 3565 } while (GOMP_loop_runtime_next (&_s0, _&e0)); 3566 GOMP_loop_end (); 3567 } 3568 3569 Note that while it looks like there is trickiness to propagating a 3570 non-constant STEP, there isn't really. We're explicitly allowed to 3571 evaluate it as many times as we want, and any variables involved should 3572 automatically be handled as PRIVATE or SHARED like any other variables. 3573 So the expression should remain evaluable in the subfunction. We can 3574 also pull it into a local variable if we like, but since its supposed to 3575 remain unchanged, we can also not if we like. 3576 3577 If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be able 3578 to get away with no work-sharing context at all, since we can simply 3579 perform the arithmetic directly in each thread to divide up the 3580 iterations. Which would mean that we wouldn't need to call any of these 3581 routines. 3582 3583 There are separate routines for handling loops with an ORDERED 3584 clause. Bookkeeping for that is non-trivial... 3585 3586 3587 File: libgomp.info, Node: Implementing ORDERED construct, Next: Implementing SECTIONS construct, Prev: Implementing FOR construct, Up: The libgomp ABI 3588 3589 10.12 Implementing ORDERED construct 3590 ==================================== 3591 3592 void GOMP_ordered_start (void) 3593 void GOMP_ordered_end (void) 3594 3595 3596 File: libgomp.info, Node: Implementing SECTIONS construct, Next: Implementing SINGLE construct, Prev: Implementing ORDERED construct, Up: The libgomp ABI 3597 3598 10.13 Implementing SECTIONS construct 3599 ===================================== 3600 3601 A block as 3602 3603 #pragma omp sections 3604 { 3605 #pragma omp section 3606 stmt1; 3607 #pragma omp section 3608 stmt2; 3609 #pragma omp section 3610 stmt3; 3611 } 3612 3613 becomes 3614 3615 for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ()) 3616 switch (i) 3617 { 3618 case 1: 3619 stmt1; 3620 break; 3621 case 2: 3622 stmt2; 3623 break; 3624 case 3: 3625 stmt3; 3626 break; 3627 } 3628 GOMP_barrier (); 3629 3630 3631 File: libgomp.info, Node: Implementing SINGLE construct, Next: Implementing OpenACC's PARALLEL construct, Prev: Implementing SECTIONS construct, Up: The libgomp ABI 3632 3633 10.14 Implementing SINGLE construct 3634 =================================== 3635 3636 A block like 3637 3638 #pragma omp single 3639 { 3640 body; 3641 } 3642 3643 becomes 3644 3645 if (GOMP_single_start ()) 3646 body; 3647 GOMP_barrier (); 3648 3649 while 3650 3651 #pragma omp single copyprivate(x) 3652 body; 3653 3654 becomes 3655 3656 datap = GOMP_single_copy_start (); 3657 if (datap == NULL) 3658 { 3659 body; 3660 data.x = x; 3661 GOMP_single_copy_end (&data); 3662 } 3663 else 3664 x = datap->x; 3665 GOMP_barrier (); 3666 3667 3668 File: libgomp.info, Node: Implementing OpenACC's PARALLEL construct, Prev: Implementing SINGLE construct, Up: The libgomp ABI 3669 3670 10.15 Implementing OpenACC's PARALLEL construct 3671 =============================================== 3672 3673 void GOACC_parallel () 3674 3675 3676 File: libgomp.info, Node: Reporting Bugs, Next: Copying, Prev: The libgomp ABI, Up: Top 3677 3678 11 Reporting Bugs 3679 ***************** 3680 3681 Bugs in the GNU Offloading and Multi Processing Runtime Library should 3682 be reported via Bugzilla (https://gcc.gnu.org/bugzilla/). Please add 3683 "openacc", or "openmp", or both to the keywords field in the bug report, 3684 as appropriate. 3685 3686 3687 File: libgomp.info, Node: Copying, Next: GNU Free Documentation License, Prev: Reporting Bugs, Up: Top 3688 3689 GNU General Public License 3690 ************************** 3691 3692 Version 3, 29 June 2007 3693 3694 Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> 3695 3696 Everyone is permitted to copy and distribute verbatim copies of this 3697 license document, but changing it is not allowed. 3698 3699 Preamble 3700 ======== 3701 3702 The GNU General Public License is a free, copyleft license for software 3703 and other kinds of works. 3704 3705 The licenses for most software and other practical works are designed 3706 to take away your freedom to share and change the works. By contrast, 3707 the GNU General Public License is intended to guarantee your freedom to 3708 share and change all versions of a program-to make sure it remains free 3709 software for all its users. We, the Free Software Foundation, use the 3710 GNU General Public License for most of our software; it applies also to 3711 any other work released this way by its authors. You can apply it to 3712 your programs, too. 3713 3714 When we speak of free software, we are referring to freedom, not 3715 price. Our General Public Licenses are designed to make sure that you 3716 have the freedom to distribute copies of free software (and charge for 3717 them if you wish), that you receive source code or can get it if you 3718 want it, that you can change the software or use pieces of it in new 3719 free programs, and that you know you can do these things. 3720 3721 To protect your rights, we need to prevent others from denying you 3722 these rights or asking you to surrender the rights. Therefore, you have 3723 certain responsibilities if you distribute copies of the software, or if 3724 you modify it: responsibilities to respect the freedom of others. 3725 3726 For example, if you distribute copies of such a program, whether 3727 gratis or for a fee, you must pass on to the recipients the same 3728 freedoms that you received. You must make sure that they, too, receive 3729 or can get the source code. And you must show them these terms so they 3730 know their rights. 3731 3732 Developers that use the GNU GPL protect your rights with two steps: 3733 (1) assert copyright on the software, and (2) offer you this License 3734 giving you legal permission to copy, distribute and/or modify it. 3735 3736 For the developers' and authors' protection, the GPL clearly explains 3737 that there is no warranty for this free software. For both users' and 3738 authors' sake, the GPL requires that modified versions be marked as 3739 changed, so that their problems will not be attributed erroneously to 3740 authors of previous versions. 3741 3742 Some devices are designed to deny users access to install or run 3743 modified versions of the software inside them, although the manufacturer 3744 can do so. This is fundamentally incompatible with the aim of 3745 protecting users' freedom to change the software. The systematic 3746 pattern of such abuse occurs in the area of products for individuals to 3747 use, which is precisely where it is most unacceptable. Therefore, we 3748 have designed this version of the GPL to prohibit the practice for those 3749 products. If such problems arise substantially in other domains, we 3750 stand ready to extend this provision to those domains in future versions 3751 of the GPL, as needed to protect the freedom of users. 3752 3753 Finally, every program is threatened constantly by software patents. 3754 States should not allow patents to restrict development and use of 3755 software on general-purpose computers, but in those that do, we wish to 3756 avoid the special danger that patents applied to a free program could 3757 make it effectively proprietary. To prevent this, the GPL assures that 3758 patents cannot be used to render the program non-free. 3759 3760 The precise terms and conditions for copying, distribution and 3761 modification follow. 3762 3763 TERMS AND CONDITIONS 3764 ==================== 3765 3766 0. Definitions. 3767 3768 "This License" refers to version 3 of the GNU General Public 3769 License. 3770 3771 "Copyright" also means copyright-like laws that apply to other 3772 kinds of works, such as semiconductor masks. 3773 3774 "The Program" refers to any copyrightable work licensed under this 3775 License. Each licensee is addressed as "you". "Licensees" and 3776 "recipients" may be individuals or organizations. 3777 3778 To "modify" a work means to copy from or adapt all or part of the 3779 work in a fashion requiring copyright permission, other than the 3780 making of an exact copy. The resulting work is called a "modified 3781 version" of the earlier work or a work "based on" the earlier work. 3782 3783 A "covered work" means either the unmodified Program or a work 3784 based on the Program. 3785 3786 To "propagate" a work means to do anything with it that, without 3787 permission, would make you directly or secondarily liable for 3788 infringement under applicable copyright law, except executing it on 3789 a computer or modifying a private copy. Propagation includes 3790 copying, distribution (with or without modification), making 3791 available to the public, and in some countries other activities as 3792 well. 3793 3794 To "convey" a work means any kind of propagation that enables other 3795 parties to make or receive copies. Mere interaction with a user 3796 through a computer network, with no transfer of a copy, is not 3797 conveying. 3798 3799 An interactive user interface displays "Appropriate Legal Notices" 3800 to the extent that it includes a convenient and prominently visible 3801 feature that (1) displays an appropriate copyright notice, and (2) 3802 tells the user that there is no warranty for the work (except to 3803 the extent that warranties are provided), that licensees may convey 3804 the work under this License, and how to view a copy of this 3805 License. If the interface presents a list of user commands or 3806 options, such as a menu, a prominent item in the list meets this 3807 criterion. 3808 3809 1. Source Code. 3810 3811 The "source code" for a work means the preferred form of the work 3812 for making modifications to it. "Object code" means any non-source 3813 form of a work. 3814 3815 A "Standard Interface" means an interface that either is an 3816 official standard defined by a recognized standards body, or, in 3817 the case of interfaces specified for a particular programming 3818 language, one that is widely used among developers working in that 3819 language. 3820 3821 The "System Libraries" of an executable work include anything, 3822 other than the work as a whole, that (a) is included in the normal 3823 form of packaging a Major Component, but which is not part of that 3824 Major Component, and (b) serves only to enable use of the work with 3825 that Major Component, or to implement a Standard Interface for 3826 which an implementation is available to the public in source code 3827 form. A "Major Component", in this context, means a major 3828 essential component (kernel, window system, and so on) of the 3829 specific operating system (if any) on which the executable work 3830 runs, or a compiler used to produce the work, or an object code 3831 interpreter used to run it. 3832 3833 The "Corresponding Source" for a work in object code form means all 3834 the source code needed to generate, install, and (for an executable 3835 work) run the object code and to modify the work, including scripts 3836 to control those activities. However, it does not include the 3837 work's System Libraries, or general-purpose tools or generally 3838 available free programs which are used unmodified in performing 3839 those activities but which are not part of the work. For example, 3840 Corresponding Source includes interface definition files associated 3841 with source files for the work, and the source code for shared 3842 libraries and dynamically linked subprograms that the work is 3843 specifically designed to require, such as by intimate data 3844 communication or control flow between those subprograms and other 3845 parts of the work. 3846 3847 The Corresponding Source need not include anything that users can 3848 regenerate automatically from other parts of the Corresponding 3849 Source. 3850 3851 The Corresponding Source for a work in source code form is that 3852 same work. 3853 3854 2. Basic Permissions. 3855 3856 All rights granted under this License are granted for the term of 3857 copyright on the Program, and are irrevocable provided the stated 3858 conditions are met. This License explicitly affirms your unlimited 3859 permission to run the unmodified Program. The output from running 3860 a covered work is covered by this License only if the output, given 3861 its content, constitutes a covered work. This License acknowledges 3862 your rights of fair use or other equivalent, as provided by 3863 copyright law. 3864 3865 You may make, run and propagate covered works that you do not 3866 convey, without conditions so long as your license otherwise 3867 remains in force. You may convey covered works to others for the 3868 sole purpose of having them make modifications exclusively for you, 3869 or provide you with facilities for running those works, provided 3870 that you comply with the terms of this License in conveying all 3871 material for which you do not control copyright. Those thus making 3872 or running the covered works for you must do so exclusively on your 3873 behalf, under your direction and control, on terms that prohibit 3874 them from making any copies of your copyrighted material outside 3875 their relationship with you. 3876 3877 Conveying under any other circumstances is permitted solely under 3878 the conditions stated below. Sublicensing is not allowed; section 3879 10 makes it unnecessary. 3880 3881 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 3882 3883 No covered work shall be deemed part of an effective technological 3884 measure under any applicable law fulfilling obligations under 3885 article 11 of the WIPO copyright treaty adopted on 20 December 3886 1996, or similar laws prohibiting or restricting circumvention of 3887 such measures. 3888 3889 When you convey a covered work, you waive any legal power to forbid 3890 circumvention of technological measures to the extent such 3891 circumvention is effected by exercising rights under this License 3892 with respect to the covered work, and you disclaim any intention to 3893 limit operation or modification of the work as a means of 3894 enforcing, against the work's users, your or third parties' legal 3895 rights to forbid circumvention of technological measures. 3896 3897 4. Conveying Verbatim Copies. 3898 3899 You may convey verbatim copies of the Program's source code as you 3900 receive it, in any medium, provided that you conspicuously and 3901 appropriately publish on each copy an appropriate copyright notice; 3902 keep intact all notices stating that this License and any 3903 non-permissive terms added in accord with section 7 apply to the 3904 code; keep intact all notices of the absence of any warranty; and 3905 give all recipients a copy of this License along with the Program. 3906 3907 You may charge any price or no price for each copy that you convey, 3908 and you may offer support or warranty protection for a fee. 3909 3910 5. Conveying Modified Source Versions. 3911 3912 You may convey a work based on the Program, or the modifications to 3913 produce it from the Program, in the form of source code under the 3914 terms of section 4, provided that you also meet all of these 3915 conditions: 3916 3917 a. The work must carry prominent notices stating that you 3918 modified it, and giving a relevant date. 3919 3920 b. The work must carry prominent notices stating that it is 3921 released under this License and any conditions added under 3922 section 7. This requirement modifies the requirement in 3923 section 4 to "keep intact all notices". 3924 3925 c. You must license the entire work, as a whole, under this 3926 License to anyone who comes into possession of a copy. This 3927 License will therefore apply, along with any applicable 3928 section 7 additional terms, to the whole of the work, and all 3929 its parts, regardless of how they are packaged. This License 3930 gives no permission to license the work in any other way, but 3931 it does not invalidate such permission if you have separately 3932 received it. 3933 3934 d. If the work has interactive user interfaces, each must display 3935 Appropriate Legal Notices; however, if the Program has 3936 interactive interfaces that do not display Appropriate Legal 3937 Notices, your work need not make them do so. 3938 3939 A compilation of a covered work with other separate and independent 3940 works, which are not by their nature extensions of the covered 3941 work, and which are not combined with it such as to form a larger 3942 program, in or on a volume of a storage or distribution medium, is 3943 called an "aggregate" if the compilation and its resulting 3944 copyright are not used to limit the access or legal rights of the 3945 compilation's users beyond what the individual works permit. 3946 Inclusion of a covered work in an aggregate does not cause this 3947 License to apply to the other parts of the aggregate. 3948 3949 6. Conveying Non-Source Forms. 3950 3951 You may convey a covered work in object code form under the terms 3952 of sections 4 and 5, provided that you also convey the 3953 machine-readable Corresponding Source under the terms of this 3954 License, in one of these ways: 3955 3956 a. Convey the object code in, or embodied in, a physical product 3957 (including a physical distribution medium), accompanied by the 3958 Corresponding Source fixed on a durable physical medium 3959 customarily used for software interchange. 3960 3961 b. Convey the object code in, or embodied in, a physical product 3962 (including a physical distribution medium), accompanied by a 3963 written offer, valid for at least three years and valid for as 3964 long as you offer spare parts or customer support for that 3965 product model, to give anyone who possesses the object code 3966 either (1) a copy of the Corresponding Source for all the 3967 software in the product that is covered by this License, on a 3968 durable physical medium customarily used for software 3969 interchange, for a price no more than your reasonable cost of 3970 physically performing this conveying of source, or (2) access 3971 to copy the Corresponding Source from a network server at no 3972 charge. 3973 3974 c. Convey individual copies of the object code with a copy of the 3975 written offer to provide the Corresponding Source. This 3976 alternative is allowed only occasionally and noncommercially, 3977 and only if you received the object code with such an offer, 3978 in accord with subsection 6b. 3979 3980 d. Convey the object code by offering access from a designated 3981 place (gratis or for a charge), and offer equivalent access to 3982 the Corresponding Source in the same way through the same 3983 place at no further charge. You need not require recipients 3984 to copy the Corresponding Source along with the object code. 3985 If the place to copy the object code is a network server, the 3986 Corresponding Source may be on a different server (operated by 3987 you or a third party) that supports equivalent copying 3988 facilities, provided you maintain clear directions next to the 3989 object code saying where to find the Corresponding Source. 3990 Regardless of what server hosts the Corresponding Source, you 3991 remain obligated to ensure that it is available for as long as 3992 needed to satisfy these requirements. 3993 3994 e. Convey the object code using peer-to-peer transmission, 3995 provided you inform other peers where the object code and 3996 Corresponding Source of the work are being offered to the 3997 general public at no charge under subsection 6d. 3998 3999 A separable portion of the object code, whose source code is 4000 excluded from the Corresponding Source as a System Library, need 4001 not be included in conveying the object code work. 4002 4003 A "User Product" is either (1) a "consumer product", which means 4004 any tangible personal property which is normally used for personal, 4005 family, or household purposes, or (2) anything designed or sold for 4006 incorporation into a dwelling. In determining whether a product is 4007 a consumer product, doubtful cases shall be resolved in favor of 4008 coverage. For a particular product received by a particular user, 4009 "normally used" refers to a typical or common use of that class of 4010 product, regardless of the status of the particular user or of the 4011 way in which the particular user actually uses, or expects or is 4012 expected to use, the product. A product is a consumer product 4013 regardless of whether the product has substantial commercial, 4014 industrial or non-consumer uses, unless such uses represent the 4015 only significant mode of use of the product. 4016 4017 "Installation Information" for a User Product means any methods, 4018 procedures, authorization keys, or other information required to 4019 install and execute modified versions of a covered work in that 4020 User Product from a modified version of its Corresponding Source. 4021 The information must suffice to ensure that the continued 4022 functioning of the modified object code is in no case prevented or 4023 interfered with solely because modification has been made. 4024 4025 If you convey an object code work under this section in, or with, 4026 or specifically for use in, a User Product, and the conveying 4027 occurs as part of a transaction in which the right of possession 4028 and use of the User Product is transferred to the recipient in 4029 perpetuity or for a fixed term (regardless of how the transaction 4030 is characterized), the Corresponding Source conveyed under this 4031 section must be accompanied by the Installation Information. But 4032 this requirement does not apply if neither you nor any third party 4033 retains the ability to install modified object code on the User 4034 Product (for example, the work has been installed in ROM). 4035 4036 The requirement to provide Installation Information does not 4037 include a requirement to continue to provide support service, 4038 warranty, or updates for a work that has been modified or installed 4039 by the recipient, or for the User Product in which it has been 4040 modified or installed. Access to a network may be denied when the 4041 modification itself materially and adversely affects the operation 4042 of the network or violates the rules and protocols for 4043 communication across the network. 4044 4045 Corresponding Source conveyed, and Installation Information 4046 provided, in accord with this section must be in a format that is 4047 publicly documented (and with an implementation available to the 4048 public in source code form), and must require no special password 4049 or key for unpacking, reading or copying. 4050 4051 7. Additional Terms. 4052 4053 "Additional permissions" are terms that supplement the terms of 4054 this License by making exceptions from one or more of its 4055 conditions. Additional permissions that are applicable to the 4056 entire Program shall be treated as though they were included in 4057 this License, to the extent that they are valid under applicable 4058 law. If additional permissions apply only to part of the Program, 4059 that part may be used separately under those permissions, but the 4060 entire Program remains governed by this License without regard to 4061 the additional permissions. 4062 4063 When you convey a copy of a covered work, you may at your option 4064 remove any additional permissions from that copy, or from any part 4065 of it. (Additional permissions may be written to require their own 4066 removal in certain cases when you modify the work.) You may place 4067 additional permissions on material, added by you to a covered work, 4068 for which you have or can give appropriate copyright permission. 4069 4070 Notwithstanding any other provision of this License, for material 4071 you add to a covered work, you may (if authorized by the copyright 4072 holders of that material) supplement the terms of this License with 4073 terms: 4074 4075 a. Disclaiming warranty or limiting liability differently from 4076 the terms of sections 15 and 16 of this License; or 4077 4078 b. Requiring preservation of specified reasonable legal notices 4079 or author attributions in that material or in the Appropriate 4080 Legal Notices displayed by works containing it; or 4081 4082 c. Prohibiting misrepresentation of the origin of that material, 4083 or requiring that modified versions of such material be marked 4084 in reasonable ways as different from the original version; or 4085 4086 d. Limiting the use for publicity purposes of names of licensors 4087 or authors of the material; or 4088 4089 e. Declining to grant rights under trademark law for use of some 4090 trade names, trademarks, or service marks; or 4091 4092 f. Requiring indemnification of licensors and authors of that 4093 material by anyone who conveys the material (or modified 4094 versions of it) with contractual assumptions of liability to 4095 the recipient, for any liability that these contractual 4096 assumptions directly impose on those licensors and authors. 4097 4098 All other non-permissive additional terms are considered "further 4099 restrictions" within the meaning of section 10. If the Program as 4100 you received it, or any part of it, contains a notice stating that 4101 it is governed by this License along with a term that is a further 4102 restriction, you may remove that term. If a license document 4103 contains a further restriction but permits relicensing or conveying 4104 under this License, you may add to a covered work material governed 4105 by the terms of that license document, provided that the further 4106 restriction does not survive such relicensing or conveying. 4107 4108 If you add terms to a covered work in accord with this section, you 4109 must place, in the relevant source files, a statement of the 4110 additional terms that apply to those files, or a notice indicating 4111 where to find the applicable terms. 4112 4113 Additional terms, permissive or non-permissive, may be stated in 4114 the form of a separately written license, or stated as exceptions; 4115 the above requirements apply either way. 4116 4117 8. Termination. 4118 4119 You may not propagate or modify a covered work except as expressly 4120 provided under this License. Any attempt otherwise to propagate or 4121 modify it is void, and will automatically terminate your rights 4122 under this License (including any patent licenses granted under the 4123 third paragraph of section 11). 4124 4125 However, if you cease all violation of this License, then your 4126 license from a particular copyright holder is reinstated (a) 4127 provisionally, unless and until the copyright holder explicitly and 4128 finally terminates your license, and (b) permanently, if the 4129 copyright holder fails to notify you of the violation by some 4130 reasonable means prior to 60 days after the cessation. 4131 4132 Moreover, your license from a particular copyright holder is 4133 reinstated permanently if the copyright holder notifies you of the 4134 violation by some reasonable means, this is the first time you have 4135 received notice of violation of this License (for any work) from 4136 that copyright holder, and you cure the violation prior to 30 days 4137 after your receipt of the notice. 4138 4139 Termination of your rights under this section does not terminate 4140 the licenses of parties who have received copies or rights from you 4141 under this License. If your rights have been terminated and not 4142 permanently reinstated, you do not qualify to receive new licenses 4143 for the same material under section 10. 4144 4145 9. Acceptance Not Required for Having Copies. 4146 4147 You are not required to accept this License in order to receive or 4148 run a copy of the Program. Ancillary propagation of a covered work 4149 occurring solely as a consequence of using peer-to-peer 4150 transmission to receive a copy likewise does not require 4151 acceptance. However, nothing other than this License grants you 4152 permission to propagate or modify any covered work. These actions 4153 infringe copyright if you do not accept this License. Therefore, 4154 by modifying or propagating a covered work, you indicate your 4155 acceptance of this License to do so. 4156 4157 10. Automatic Licensing of Downstream Recipients. 4158 4159 Each time you convey a covered work, the recipient automatically 4160 receives a license from the original licensors, to run, modify and 4161 propagate that work, subject to this License. You are not 4162 responsible for enforcing compliance by third parties with this 4163 License. 4164 4165 An "entity transaction" is a transaction transferring control of an 4166 organization, or substantially all assets of one, or subdividing an 4167 organization, or merging organizations. If propagation of a 4168 covered work results from an entity transaction, each party to that 4169 transaction who receives a copy of the work also receives whatever 4170 licenses to the work the party's predecessor in interest had or 4171 could give under the previous paragraph, plus a right to possession 4172 of the Corresponding Source of the work from the predecessor in 4173 interest, if the predecessor has it or can get it with reasonable 4174 efforts. 4175 4176 You may not impose any further restrictions on the exercise of the 4177 rights granted or affirmed under this License. For example, you 4178 may not impose a license fee, royalty, or other charge for exercise 4179 of rights granted under this License, and you may not initiate 4180 litigation (including a cross-claim or counterclaim in a lawsuit) 4181 alleging that any patent claim is infringed by making, using, 4182 selling, offering for sale, or importing the Program or any portion 4183 of it. 4184 4185 11. Patents. 4186 4187 A "contributor" is a copyright holder who authorizes use under this 4188 License of the Program or a work on which the Program is based. 4189 The work thus licensed is called the contributor's "contributor 4190 version". 4191 4192 A contributor's "essential patent claims" are all patent claims 4193 owned or controlled by the contributor, whether already acquired or 4194 hereafter acquired, that would be infringed by some manner, 4195 permitted by this License, of making, using, or selling its 4196 contributor version, but do not include claims that would be 4197 infringed only as a consequence of further modification of the 4198 contributor version. For purposes of this definition, "control" 4199 includes the right to grant patent sublicenses in a manner 4200 consistent with the requirements of this License. 4201 4202 Each contributor grants you a non-exclusive, worldwide, 4203 royalty-free patent license under the contributor's essential 4204 patent claims, to make, use, sell, offer for sale, import and 4205 otherwise run, modify and propagate the contents of its contributor 4206 version. 4207 4208 In the following three paragraphs, a "patent license" is any 4209 express agreement or commitment, however denominated, not to 4210 enforce a patent (such as an express permission to practice a 4211 patent or covenant not to sue for patent infringement). To "grant" 4212 such a patent license to a party means to make such an agreement or 4213 commitment not to enforce a patent against the party. 4214 4215 If you convey a covered work, knowingly relying on a patent 4216 license, and the Corresponding Source of the work is not available 4217 for anyone to copy, free of charge and under the terms of this 4218 License, through a publicly available network server or other 4219 readily accessible means, then you must either (1) cause the 4220 Corresponding Source to be so available, or (2) arrange to deprive 4221 yourself of the benefit of the patent license for this particular 4222 work, or (3) arrange, in a manner consistent with the requirements 4223 of this License, to extend the patent license to downstream 4224 recipients. "Knowingly relying" means you have actual knowledge 4225 that, but for the patent license, your conveying the covered work 4226 in a country, or your recipient's use of the covered work in a 4227 country, would infringe one or more identifiable patents in that 4228 country that you have reason to believe are valid. 4229 4230 If, pursuant to or in connection with a single transaction or 4231 arrangement, you convey, or propagate by procuring conveyance of, a 4232 covered work, and grant a patent license to some of the parties 4233 receiving the covered work authorizing them to use, propagate, 4234 modify or convey a specific copy of the covered work, then the 4235 patent license you grant is automatically extended to all 4236 recipients of the covered work and works based on it. 4237 4238 A patent license is "discriminatory" if it does not include within 4239 the scope of its coverage, prohibits the exercise of, or is 4240 conditioned on the non-exercise of one or more of the rights that 4241 are specifically granted under this License. You may not convey a 4242 covered work if you are a party to an arrangement with a third 4243 party that is in the business of distributing software, under which 4244 you make payment to the third party based on the extent of your 4245 activity of conveying the work, and under which the third party 4246 grants, to any of the parties who would receive the covered work 4247 from you, a discriminatory patent license (a) in connection with 4248 copies of the covered work conveyed by you (or copies made from 4249 those copies), or (b) primarily for and in connection with specific 4250 products or compilations that contain the covered work, unless you 4251 entered into that arrangement, or that patent license was granted, 4252 prior to 28 March 2007. 4253 4254 Nothing in this License shall be construed as excluding or limiting 4255 any implied license or other defenses to infringement that may 4256 otherwise be available to you under applicable patent law. 4257 4258 12. No Surrender of Others' Freedom. 4259 4260 If conditions are imposed on you (whether by court order, agreement 4261 or otherwise) that contradict the conditions of this License, they 4262 do not excuse you from the conditions of this License. If you 4263 cannot convey a covered work so as to satisfy simultaneously your 4264 obligations under this License and any other pertinent obligations, 4265 then as a consequence you may not convey it at all. For example, 4266 if you agree to terms that obligate you to collect a royalty for 4267 further conveying from those to whom you convey the Program, the 4268 only way you could satisfy both those terms and this License would 4269 be to refrain entirely from conveying the Program. 4270 4271 13. Use with the GNU Affero General Public License. 4272 4273 Notwithstanding any other provision of this License, you have 4274 permission to link or combine any covered work with a work licensed 4275 under version 3 of the GNU Affero General Public License into a 4276 single combined work, and to convey the resulting work. The terms 4277 of this License will continue to apply to the part which is the 4278 covered work, but the special requirements of the GNU Affero 4279 General Public License, section 13, concerning interaction through 4280 a network will apply to the combination as such. 4281 4282 14. Revised Versions of this License. 4283 4284 The Free Software Foundation may publish revised and/or new 4285 versions of the GNU General Public License from time to time. Such 4286 new versions will be similar in spirit to the present version, but 4287 may differ in detail to address new problems or concerns. 4288 4289 Each version is given a distinguishing version number. If the 4290 Program specifies that a certain numbered version of the GNU 4291 General Public License "or any later version" applies to it, you 4292 have the option of following the terms and conditions either of 4293 that numbered version or of any later version published by the Free 4294 Software Foundation. If the Program does not specify a version 4295 number of the GNU General Public License, you may choose any 4296 version ever published by the Free Software Foundation. 4297 4298 If the Program specifies that a proxy can decide which future 4299 versions of the GNU General Public License can be used, that 4300 proxy's public statement of acceptance of a version permanently 4301 authorizes you to choose that version for the Program. 4302 4303 Later license versions may give you additional or different 4304 permissions. However, no additional obligations are imposed on any 4305 author or copyright holder as a result of your choosing to follow a 4306 later version. 4307 4308 15. Disclaimer of Warranty. 4309 4310 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 4311 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 4312 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 4313 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 4314 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 4315 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 4316 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 4317 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 4318 NECESSARY SERVICING, REPAIR OR CORRECTION. 4319 4320 16. Limitation of Liability. 4321 4322 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 4323 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 4324 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR 4325 DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 4326 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 4327 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 4328 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 4329 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 4330 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 4331 THE POSSIBILITY OF SUCH DAMAGES. 4332 4333 17. Interpretation of Sections 15 and 16. 4334 4335 If the disclaimer of warranty and limitation of liability provided 4336 above cannot be given local legal effect according to their terms, 4337 reviewing courts shall apply local law that most closely 4338 approximates an absolute waiver of all civil liability in 4339 connection with the Program, unless a warranty or assumption of 4340 liability accompanies a copy of the Program in return for a fee. 4341 4342 END OF TERMS AND CONDITIONS 4343 =========================== 4344 4345 How to Apply These Terms to Your New Programs 4346 ============================================= 4347 4348 If you develop a new program, and you want it to be of the greatest 4349 possible use to the public, the best way to achieve this is to make it 4350 free software which everyone can redistribute and change under these 4351 terms. 4352 4353 To do so, attach the following notices to the program. It is safest 4354 to attach them to the start of each source file to most effectively 4355 state the exclusion of warranty; and each file should have at least the 4356 "copyright" line and a pointer to where the full notice is found. 4357 4358 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 4359 Copyright (C) YEAR NAME OF AUTHOR 4360 4361 This program is free software: you can redistribute it and/or modify 4362 it under the terms of the GNU General Public License as published by 4363 the Free Software Foundation, either version 3 of the License, or (at 4364 your option) any later version. 4365 4366 This program is distributed in the hope that it will be useful, but 4367 WITHOUT ANY WARRANTY; without even the implied warranty of 4368 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 4369 General Public License for more details. 4370 4371 You should have received a copy of the GNU General Public License 4372 along with this program. If not, see <http://www.gnu.org/licenses/>. 4373 4374 Also add information on how to contact you by electronic and paper 4375 mail. 4376 4377 If the program does terminal interaction, make it output a short 4378 notice like this when it starts in an interactive mode: 4379 4380 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 4381 This program comes with ABSOLUTELY NO WARRANTY; for details type 'show w'. 4382 This is free software, and you are welcome to redistribute it 4383 under certain conditions; type 'show c' for details. 4384 4385 The hypothetical commands 'show w' and 'show c' should show the 4386 appropriate parts of the General Public License. Of course, your 4387 program's commands might be different; for a GUI interface, you would 4388 use an "about box". 4389 4390 You should also get your employer (if you work as a programmer) or 4391 school, if any, to sign a "copyright disclaimer" for the program, if 4392 necessary. For more information on this, and how to apply and follow 4393 the GNU GPL, see <http://www.gnu.org/licenses/>. 4394 4395 The GNU General Public License does not permit incorporating your 4396 program into proprietary programs. If your program is a subroutine 4397 library, you may consider it more useful to permit linking proprietary 4398 applications with the library. If this is what you want to do, use the 4399 GNU Lesser General Public License instead of this License. But first, 4400 please read <https://www.gnu.org/licenses/why-not-lgpl.html>. 4401 4402 4403 File: libgomp.info, Node: GNU Free Documentation License, Next: Funding, Prev: Copying, Up: Top 4404 4405 GNU Free Documentation License 4406 ****************************** 4407 4408 Version 1.3, 3 November 2008 4409 4410 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. 4411 <http://fsf.org/> 4412 4413 Everyone is permitted to copy and distribute verbatim copies 4414 of this license document, but changing it is not allowed. 4415 4416 0. PREAMBLE 4417 4418 The purpose of this License is to make a manual, textbook, or other 4419 functional and useful document "free" in the sense of freedom: to 4420 assure everyone the effective freedom to copy and redistribute it, 4421 with or without modifying it, either commercially or 4422 noncommercially. Secondarily, this License preserves for the 4423 author and publisher a way to get credit for their work, while not 4424 being considered responsible for modifications made by others. 4425 4426 This License is a kind of "copyleft", which means that derivative 4427 works of the document must themselves be free in the same sense. 4428 It complements the GNU General Public License, which is a copyleft 4429 license designed for free software. 4430 4431 We have designed this License in order to use it for manuals for 4432 free software, because free software needs free documentation: a 4433 free program should come with manuals providing the same freedoms 4434 that the software does. But this License is not limited to 4435 software manuals; it can be used for any textual work, regardless 4436 of subject matter or whether it is published as a printed book. We 4437 recommend this License principally for works whose purpose is 4438 instruction or reference. 4439 4440 1. APPLICABILITY AND DEFINITIONS 4441 4442 This License applies to any manual or other work, in any medium, 4443 that contains a notice placed by the copyright holder saying it can 4444 be distributed under the terms of this License. Such a notice 4445 grants a world-wide, royalty-free license, unlimited in duration, 4446 to use that work under the conditions stated herein. The 4447 "Document", below, refers to any such manual or work. Any member 4448 of the public is a licensee, and is addressed as "you". You accept 4449 the license if you copy, modify or distribute the work in a way 4450 requiring permission under copyright law. 4451 4452 A "Modified Version" of the Document means any work containing the 4453 Document or a portion of it, either copied verbatim, or with 4454 modifications and/or translated into another language. 4455 4456 A "Secondary Section" is a named appendix or a front-matter section 4457 of the Document that deals exclusively with the relationship of the 4458 publishers or authors of the Document to the Document's overall 4459 subject (or to related matters) and contains nothing that could 4460 fall directly within that overall subject. (Thus, if the Document 4461 is in part a textbook of mathematics, a Secondary Section may not 4462 explain any mathematics.) The relationship could be a matter of 4463 historical connection with the subject or with related matters, or 4464 of legal, commercial, philosophical, ethical or political position 4465 regarding them. 4466 4467 The "Invariant Sections" are certain Secondary Sections whose 4468 titles are designated, as being those of Invariant Sections, in the 4469 notice that says that the Document is released under this License. 4470 If a section does not fit the above definition of Secondary then it 4471 is not allowed to be designated as Invariant. The Document may 4472 contain zero Invariant Sections. If the Document does not identify 4473 any Invariant Sections then there are none. 4474 4475 The "Cover Texts" are certain short passages of text that are 4476 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 4477 that says that the Document is released under this License. A 4478 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 4479 be at most 25 words. 4480 4481 A "Transparent" copy of the Document means a machine-readable copy, 4482 represented in a format whose specification is available to the 4483 general public, that is suitable for revising the document 4484 straightforwardly with generic text editors or (for images composed 4485 of pixels) generic paint programs or (for drawings) some widely 4486 available drawing editor, and that is suitable for input to text 4487 formatters or for automatic translation to a variety of formats 4488 suitable for input to text formatters. A copy made in an otherwise 4489 Transparent file format whose markup, or absence of markup, has 4490 been arranged to thwart or discourage subsequent modification by 4491 readers is not Transparent. An image format is not Transparent if 4492 used for any substantial amount of text. A copy that is not 4493 "Transparent" is called "Opaque". 4494 4495 Examples of suitable formats for Transparent copies include plain 4496 ASCII without markup, Texinfo input format, LaTeX input format, 4497 SGML or XML using a publicly available DTD, and standard-conforming 4498 simple HTML, PostScript or PDF designed for human modification. 4499 Examples of transparent image formats include PNG, XCF and JPG. 4500 Opaque formats include proprietary formats that can be read and 4501 edited only by proprietary word processors, SGML or XML for which 4502 the DTD and/or processing tools are not generally available, and 4503 the machine-generated HTML, PostScript or PDF produced by some word 4504 processors for output purposes only. 4505 4506 The "Title Page" means, for a printed book, the title page itself, 4507 plus such following pages as are needed to hold, legibly, the 4508 material this License requires to appear in the title page. For 4509 works in formats which do not have any title page as such, "Title 4510 Page" means the text near the most prominent appearance of the 4511 work's title, preceding the beginning of the body of the text. 4512 4513 The "publisher" means any person or entity that distributes copies 4514 of the Document to the public. 4515 4516 A section "Entitled XYZ" means a named subunit of the Document 4517 whose title either is precisely XYZ or contains XYZ in parentheses 4518 following text that translates XYZ in another language. (Here XYZ 4519 stands for a specific section name mentioned below, such as 4520 "Acknowledgements", "Dedications", "Endorsements", or "History".) 4521 To "Preserve the Title" of such a section when you modify the 4522 Document means that it remains a section "Entitled XYZ" according 4523 to this definition. 4524 4525 The Document may include Warranty Disclaimers next to the notice 4526 which states that this License applies to the Document. These 4527 Warranty Disclaimers are considered to be included by reference in 4528 this License, but only as regards disclaiming warranties: any other 4529 implication that these Warranty Disclaimers may have is void and 4530 has no effect on the meaning of this License. 4531 4532 2. VERBATIM COPYING 4533 4534 You may copy and distribute the Document in any medium, either 4535 commercially or noncommercially, provided that this License, the 4536 copyright notices, and the license notice saying this License 4537 applies to the Document are reproduced in all copies, and that you 4538 add no other conditions whatsoever to those of this License. You 4539 may not use technical measures to obstruct or control the reading 4540 or further copying of the copies you make or distribute. However, 4541 you may accept compensation in exchange for copies. If you 4542 distribute a large enough number of copies you must also follow the 4543 conditions in section 3. 4544 4545 You may also lend copies, under the same conditions stated above, 4546 and you may publicly display copies. 4547 4548 3. COPYING IN QUANTITY 4549 4550 If you publish printed copies (or copies in media that commonly 4551 have printed covers) of the Document, numbering more than 100, and 4552 the Document's license notice requires Cover Texts, you must 4553 enclose the copies in covers that carry, clearly and legibly, all 4554 these Cover Texts: Front-Cover Texts on the front cover, and 4555 Back-Cover Texts on the back cover. Both covers must also clearly 4556 and legibly identify you as the publisher of these copies. The 4557 front cover must present the full title with all words of the title 4558 equally prominent and visible. You may add other material on the 4559 covers in addition. Copying with changes limited to the covers, as 4560 long as they preserve the title of the Document and satisfy these 4561 conditions, can be treated as verbatim copying in other respects. 4562 4563 If the required texts for either cover are too voluminous to fit 4564 legibly, you should put the first ones listed (as many as fit 4565 reasonably) on the actual cover, and continue the rest onto 4566 adjacent pages. 4567 4568 If you publish or distribute Opaque copies of the Document 4569 numbering more than 100, you must either include a machine-readable 4570 Transparent copy along with each Opaque copy, or state in or with 4571 each Opaque copy a computer-network location from which the general 4572 network-using public has access to download using public-standard 4573 network protocols a complete Transparent copy of the Document, free 4574 of added material. If you use the latter option, you must take 4575 reasonably prudent steps, when you begin distribution of Opaque 4576 copies in quantity, to ensure that this Transparent copy will 4577 remain thus accessible at the stated location until at least one 4578 year after the last time you distribute an Opaque copy (directly or 4579 through your agents or retailers) of that edition to the public. 4580 4581 It is requested, but not required, that you contact the authors of 4582 the Document well before redistributing any large number of copies, 4583 to give them a chance to provide you with an updated version of the 4584 Document. 4585 4586 4. MODIFICATIONS 4587 4588 You may copy and distribute a Modified Version of the Document 4589 under the conditions of sections 2 and 3 above, provided that you 4590 release the Modified Version under precisely this License, with the 4591 Modified Version filling the role of the Document, thus licensing 4592 distribution and modification of the Modified Version to whoever 4593 possesses a copy of it. In addition, you must do these things in 4594 the Modified Version: 4595 4596 A. Use in the Title Page (and on the covers, if any) a title 4597 distinct from that of the Document, and from those of previous 4598 versions (which should, if there were any, be listed in the 4599 History section of the Document). You may use the same title 4600 as a previous version if the original publisher of that 4601 version gives permission. 4602 4603 B. List on the Title Page, as authors, one or more persons or 4604 entities responsible for authorship of the modifications in 4605 the Modified Version, together with at least five of the 4606 principal authors of the Document (all of its principal 4607 authors, if it has fewer than five), unless they release you 4608 from this requirement. 4609 4610 C. State on the Title page the name of the publisher of the 4611 Modified Version, as the publisher. 4612 4613 D. Preserve all the copyright notices of the Document. 4614 4615 E. Add an appropriate copyright notice for your modifications 4616 adjacent to the other copyright notices. 4617 4618 F. Include, immediately after the copyright notices, a license 4619 notice giving the public permission to use the Modified 4620 Version under the terms of this License, in the form shown in 4621 the Addendum below. 4622 4623 G. Preserve in that license notice the full lists of Invariant 4624 Sections and required Cover Texts given in the Document's 4625 license notice. 4626 4627 H. Include an unaltered copy of this License. 4628 4629 I. Preserve the section Entitled "History", Preserve its Title, 4630 and add to it an item stating at least the title, year, new 4631 authors, and publisher of the Modified Version as given on the 4632 Title Page. If there is no section Entitled "History" in the 4633 Document, create one stating the title, year, authors, and 4634 publisher of the Document as given on its Title Page, then add 4635 an item describing the Modified Version as stated in the 4636 previous sentence. 4637 4638 J. Preserve the network location, if any, given in the Document 4639 for public access to a Transparent copy of the Document, and 4640 likewise the network locations given in the Document for 4641 previous versions it was based on. These may be placed in the 4642 "History" section. You may omit a network location for a work 4643 that was published at least four years before the Document 4644 itself, or if the original publisher of the version it refers 4645 to gives permission. 4646 4647 K. For any section Entitled "Acknowledgements" or "Dedications", 4648 Preserve the Title of the section, and preserve in the section 4649 all the substance and tone of each of the contributor 4650 acknowledgements and/or dedications given therein. 4651 4652 L. Preserve all the Invariant Sections of the Document, unaltered 4653 in their text and in their titles. Section numbers or the 4654 equivalent are not considered part of the section titles. 4655 4656 M. Delete any section Entitled "Endorsements". Such a section 4657 may not be included in the Modified Version. 4658 4659 N. Do not retitle any existing section to be Entitled 4660 "Endorsements" or to conflict in title with any Invariant 4661 Section. 4662 4663 O. Preserve any Warranty Disclaimers. 4664 4665 If the Modified Version includes new front-matter sections or 4666 appendices that qualify as Secondary Sections and contain no 4667 material copied from the Document, you may at your option designate 4668 some or all of these sections as invariant. To do this, add their 4669 titles to the list of Invariant Sections in the Modified Version's 4670 license notice. These titles must be distinct from any other 4671 section titles. 4672 4673 You may add a section Entitled "Endorsements", provided it contains 4674 nothing but endorsements of your Modified Version by various 4675 parties--for example, statements of peer review or that the text 4676 has been approved by an organization as the authoritative 4677 definition of a standard. 4678 4679 You may add a passage of up to five words as a Front-Cover Text, 4680 and a passage of up to 25 words as a Back-Cover Text, to the end of 4681 the list of Cover Texts in the Modified Version. Only one passage 4682 of Front-Cover Text and one of Back-Cover Text may be added by (or 4683 through arrangements made by) any one entity. If the Document 4684 already includes a cover text for the same cover, previously added 4685 by you or by arrangement made by the same entity you are acting on 4686 behalf of, you may not add another; but you may replace the old 4687 one, on explicit permission from the previous publisher that added 4688 the old one. 4689 4690 The author(s) and publisher(s) of the Document do not by this 4691 License give permission to use their names for publicity for or to 4692 assert or imply endorsement of any Modified Version. 4693 4694 5. COMBINING DOCUMENTS 4695 4696 You may combine the Document with other documents released under 4697 this License, under the terms defined in section 4 above for 4698 modified versions, provided that you include in the combination all 4699 of the Invariant Sections of all of the original documents, 4700 unmodified, and list them all as Invariant Sections of your 4701 combined work in its license notice, and that you preserve all 4702 their Warranty Disclaimers. 4703 4704 The combined work need only contain one copy of this License, and 4705 multiple identical Invariant Sections may be replaced with a single 4706 copy. If there are multiple Invariant Sections with the same name 4707 but different contents, make the title of each such section unique 4708 by adding at the end of it, in parentheses, the name of the 4709 original author or publisher of that section if known, or else a 4710 unique number. Make the same adjustment to the section titles in 4711 the list of Invariant Sections in the license notice of the 4712 combined work. 4713 4714 In the combination, you must combine any sections Entitled 4715 "History" in the various original documents, forming one section 4716 Entitled "History"; likewise combine any sections Entitled 4717 "Acknowledgements", and any sections Entitled "Dedications". You 4718 must delete all sections Entitled "Endorsements." 4719 4720 6. COLLECTIONS OF DOCUMENTS 4721 4722 You may make a collection consisting of the Document and other 4723 documents released under this License, and replace the individual 4724 copies of this License in the various documents with a single copy 4725 that is included in the collection, provided that you follow the 4726 rules of this License for verbatim copying of each of the documents 4727 in all other respects. 4728 4729 You may extract a single document from such a collection, and 4730 distribute it individually under this License, provided you insert 4731 a copy of this License into the extracted document, and follow this 4732 License in all other respects regarding verbatim copying of that 4733 document. 4734 4735 7. AGGREGATION WITH INDEPENDENT WORKS 4736 4737 A compilation of the Document or its derivatives with other 4738 separate and independent documents or works, in or on a volume of a 4739 storage or distribution medium, is called an "aggregate" if the 4740 copyright resulting from the compilation is not used to limit the 4741 legal rights of the compilation's users beyond what the individual 4742 works permit. When the Document is included in an aggregate, this 4743 License does not apply to the other works in the aggregate which 4744 are not themselves derivative works of the Document. 4745 4746 If the Cover Text requirement of section 3 is applicable to these 4747 copies of the Document, then if the Document is less than one half 4748 of the entire aggregate, the Document's Cover Texts may be placed 4749 on covers that bracket the Document within the aggregate, or the 4750 electronic equivalent of covers if the Document is in electronic 4751 form. Otherwise they must appear on printed covers that bracket 4752 the whole aggregate. 4753 4754 8. TRANSLATION 4755 4756 Translation is considered a kind of modification, so you may 4757 distribute translations of the Document under the terms of section 4758 4. Replacing Invariant Sections with translations requires special 4759 permission from their copyright holders, but you may include 4760 translations of some or all Invariant Sections in addition to the 4761 original versions of these Invariant Sections. You may include a 4762 translation of this License, and all the license notices in the 4763 Document, and any Warranty Disclaimers, provided that you also 4764 include the original English version of this License and the 4765 original versions of those notices and disclaimers. In case of a 4766 disagreement between the translation and the original version of 4767 this License or a notice or disclaimer, the original version will 4768 prevail. 4769 4770 If a section in the Document is Entitled "Acknowledgements", 4771 "Dedications", or "History", the requirement (section 4) to 4772 Preserve its Title (section 1) will typically require changing the 4773 actual title. 4774 4775 9. TERMINATION 4776 4777 You may not copy, modify, sublicense, or distribute the Document 4778 except as expressly provided under this License. Any attempt 4779 otherwise to copy, modify, sublicense, or distribute it is void, 4780 and will automatically terminate your rights under this License. 4781 4782 However, if you cease all violation of this License, then your 4783 license from a particular copyright holder is reinstated (a) 4784 provisionally, unless and until the copyright holder explicitly and 4785 finally terminates your license, and (b) permanently, if the 4786 copyright holder fails to notify you of the violation by some 4787 reasonable means prior to 60 days after the cessation. 4788 4789 Moreover, your license from a particular copyright holder is 4790 reinstated permanently if the copyright holder notifies you of the 4791 violation by some reasonable means, this is the first time you have 4792 received notice of violation of this License (for any work) from 4793 that copyright holder, and you cure the violation prior to 30 days 4794 after your receipt of the notice. 4795 4796 Termination of your rights under this section does not terminate 4797 the licenses of parties who have received copies or rights from you 4798 under this License. If your rights have been terminated and not 4799 permanently reinstated, receipt of a copy of some or all of the 4800 same material does not give you any rights to use it. 4801 4802 10. FUTURE REVISIONS OF THIS LICENSE 4803 4804 The Free Software Foundation may publish new, revised versions of 4805 the GNU Free Documentation License from time to time. Such new 4806 versions will be similar in spirit to the present version, but may 4807 differ in detail to address new problems or concerns. See 4808 <http://www.gnu.org/copyleft/>. 4809 4810 Each version of the License is given a distinguishing version 4811 number. If the Document specifies that a particular numbered 4812 version of this License "or any later version" applies to it, you 4813 have the option of following the terms and conditions either of 4814 that specified version or of any later version that has been 4815 published (not as a draft) by the Free Software Foundation. If the 4816 Document does not specify a version number of this License, you may 4817 choose any version ever published (not as a draft) by the Free 4818 Software Foundation. If the Document specifies that a proxy can 4819 decide which future versions of this License can be used, that 4820 proxy's public statement of acceptance of a version permanently 4821 authorizes you to choose that version for the Document. 4822 4823 11. RELICENSING 4824 4825 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any 4826 World Wide Web server that publishes copyrightable works and also 4827 provides prominent facilities for anybody to edit those works. A 4828 public wiki that anybody can edit is an example of such a server. 4829 A "Massive Multiauthor Collaboration" (or "MMC") contained in the 4830 site means any set of copyrightable works thus published on the MMC 4831 site. 4832 4833 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 4834 license published by Creative Commons Corporation, a not-for-profit 4835 corporation with a principal place of business in San Francisco, 4836 California, as well as future copyleft versions of that license 4837 published by that same organization. 4838 4839 "Incorporate" means to publish or republish a Document, in whole or 4840 in part, as part of another Document. 4841 4842 An MMC is "eligible for relicensing" if it is licensed under this 4843 License, and if all works that were first published under this 4844 License somewhere other than this MMC, and subsequently 4845 incorporated in whole or in part into the MMC, (1) had no cover 4846 texts or invariant sections, and (2) were thus incorporated prior 4847 to November 1, 2008. 4848 4849 The operator of an MMC Site may republish an MMC contained in the 4850 site under CC-BY-SA on the same site at any time before August 1, 4851 2009, provided the MMC is eligible for relicensing. 4852 4853 ADDENDUM: How to use this License for your documents 4854 ==================================================== 4855 4856 To use this License in a document you have written, include a copy of 4857 the License in the document and put the following copyright and license 4858 notices just after the title page: 4859 4860 Copyright (C) YEAR YOUR NAME. 4861 Permission is granted to copy, distribute and/or modify this document 4862 under the terms of the GNU Free Documentation License, Version 1.3 4863 or any later version published by the Free Software Foundation; 4864 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 4865 Texts. A copy of the license is included in the section entitled ``GNU 4866 Free Documentation License''. 4867 4868 If you have Invariant Sections, Front-Cover Texts and Back-Cover 4869 Texts, replace the "with...Texts." line with this: 4870 4871 with the Invariant Sections being LIST THEIR TITLES, with 4872 the Front-Cover Texts being LIST, and with the Back-Cover Texts 4873 being LIST. 4874 4875 If you have Invariant Sections without Cover Texts, or some other 4876 combination of the three, merge those two alternatives to suit the 4877 situation. 4878 4879 If your document contains nontrivial examples of program code, we 4880 recommend releasing these examples in parallel under your choice of free 4881 software license, such as the GNU General Public License, to permit 4882 their use in free software. 4883 4884 4885 File: libgomp.info, Node: Funding, Next: Library Index, Prev: GNU Free Documentation License, Up: Top 4886 4887 Funding Free Software 4888 ********************* 4889 4890 If you want to have more free software a few years from now, it makes 4891 sense for you to help encourage people to contribute funds for its 4892 development. The most effective approach known is to encourage 4893 commercial redistributors to donate. 4894 4895 Users of free software systems can boost the pace of development by 4896 encouraging for-a-fee distributors to donate part of their selling price 4897 to free software developers--the Free Software Foundation, and others. 4898 4899 The way to convince distributors to do this is to demand it and 4900 expect it from them. So when you compare distributors, judge them 4901 partly by how much they give to free software development. Show 4902 distributors they must compete to be the one who gives the most. 4903 4904 To make this approach work, you must insist on numbers that you can 4905 compare, such as, "We will donate ten dollars to the Frobnitz project 4906 for each disk sold." Don't be satisfied with a vague promise, such as 4907 "A portion of the profits are donated," since it doesn't give a basis 4908 for comparison. 4909 4910 Even a precise fraction "of the profits from this disk" is not very 4911 meaningful, since creative accounting and unrelated business decisions 4912 can greatly alter what fraction of the sales price counts as profit. If 4913 the price you pay is $50, ten percent of the profit is probably less 4914 than a dollar; it might be a few cents, or nothing at all. 4915 4916 Some redistributors do development work themselves. This is useful 4917 too; but to keep everyone honest, you need to inquire how much they do, 4918 and what kind. Some kinds of development make much more long-term 4919 difference than others. For example, maintaining a separate version of 4920 a program contributes very little; maintaining the standard version of a 4921 program for the whole community contributes much. Easy new ports 4922 contribute little, since someone else would surely do them; difficult 4923 ports such as adding a new CPU to the GNU Compiler Collection contribute 4924 more; major new features or packages contribute the most. 4925 4926 By establishing the idea that supporting further development is "the 4927 proper thing to do" when distributing free software for a fee, we can 4928 assure a steady flow of resources into making more free software. 4929 4930 Copyright (C) 1994 Free Software Foundation, Inc. 4931 Verbatim copying and redistribution of this section is permitted 4932 without royalty; alteration is not permitted. 4933 4934 4935 File: libgomp.info, Node: Library Index, Prev: Funding, Up: Top 4936 4937 Library Index 4938 ************* 4939 4940 [index] 4941 * Menu: 4942 4943 * acc_get_property: acc_get_property. (line 6) 4944 * acc_get_property_string: acc_get_property. (line 6) 4945 * Environment Variable: OMP_CANCELLATION. (line 6) 4946 * Environment Variable <1>: OMP_DISPLAY_ENV. (line 6) 4947 * Environment Variable <2>: OMP_DEFAULT_DEVICE. (line 6) 4948 * Environment Variable <3>: OMP_DYNAMIC. (line 6) 4949 * Environment Variable <4>: OMP_MAX_ACTIVE_LEVELS. (line 6) 4950 * Environment Variable <5>: OMP_MAX_TASK_PRIORITY. (line 6) 4951 * Environment Variable <6>: OMP_NESTED. (line 6) 4952 * Environment Variable <7>: OMP_NUM_THREADS. (line 6) 4953 * Environment Variable <8>: OMP_PROC_BIND. (line 6) 4954 * Environment Variable <9>: OMP_PLACES. (line 6) 4955 * Environment Variable <10>: OMP_STACKSIZE. (line 6) 4956 * Environment Variable <11>: OMP_SCHEDULE. (line 6) 4957 * Environment Variable <12>: OMP_THREAD_LIMIT. (line 6) 4958 * Environment Variable <13>: OMP_WAIT_POLICY. (line 6) 4959 * Environment Variable <14>: GOMP_CPU_AFFINITY. (line 6) 4960 * Environment Variable <15>: GOMP_DEBUG. (line 6) 4961 * Environment Variable <16>: GOMP_STACKSIZE. (line 6) 4962 * Environment Variable <17>: GOMP_SPINCOUNT. (line 6) 4963 * Environment Variable <18>: GOMP_RTEMS_THREAD_POOLS. 4964 (line 6) 4965 * FDL, GNU Free Documentation License: GNU Free Documentation License. 4966 (line 6) 4967 * Implementation specific setting: OMP_NESTED. (line 6) 4968 * Implementation specific setting <1>: OMP_NUM_THREADS. (line 6) 4969 * Implementation specific setting <2>: OMP_SCHEDULE. (line 6) 4970 * Implementation specific setting <3>: GOMP_STACKSIZE. (line 6) 4971 * Implementation specific setting <4>: GOMP_SPINCOUNT. (line 6) 4972 * Implementation specific setting <5>: GOMP_RTEMS_THREAD_POOLS. 4973 (line 6) 4974 * Introduction: Top. (line 6) 4975 4976 4977 4978 Tag Table: 4979 Node: Top2083 4980 Node: Enabling OpenMP4645 4981 Node: Runtime Library Routines5433 4982 Node: omp_get_active_level8510 4983 Node: omp_get_ancestor_thread_num9210 4984 Node: omp_get_cancellation10140 4985 Node: omp_get_default_device10954 4986 Node: omp_get_dynamic11630 4987 Node: omp_get_level12505 4988 Node: omp_get_max_active_levels13125 4989 Node: omp_get_max_task_priority13830 4990 Node: omp_get_max_threads14450 4991 Node: omp_get_nested15207 4992 Node: omp_get_num_devices16121 4993 Node: omp_get_num_procs16642 4994 Node: omp_get_num_teams17181 4995 Node: omp_get_num_threads17697 4996 Node: omp_get_proc_bind18786 4997 Node: omp_get_schedule19707 4998 Node: omp_get_team_num20661 4999 Node: omp_get_team_size21160 5000 Node: omp_get_thread_limit22120 5001 Node: omp_get_thread_num22739 5002 Node: omp_in_parallel23610 5003 Node: omp_in_final24259 5004 Node: omp_is_initial_device24933 5005 Node: omp_set_default_device25626 5006 Node: omp_set_dynamic26417 5007 Node: omp_set_max_active_levels27303 5008 Node: omp_set_nested28080 5009 Node: omp_set_num_threads28972 5010 Node: omp_set_schedule29840 5011 Node: omp_init_lock30921 5012 Node: omp_set_lock31574 5013 Node: omp_test_lock32429 5014 Node: omp_unset_lock33405 5015 Node: omp_destroy_lock34336 5016 Node: omp_init_nest_lock35013 5017 Node: omp_set_nest_lock35748 5018 Node: omp_test_nest_lock36663 5019 Node: omp_unset_nest_lock37690 5020 Node: omp_destroy_nest_lock38705 5021 Node: omp_get_wtick39456 5022 Node: omp_get_wtime40048 5023 Node: Environment Variables40824 5024 Node: OMP_CANCELLATION42393 5025 Node: OMP_DISPLAY_ENV42926 5026 Node: OMP_DEFAULT_DEVICE43629 5027 Node: OMP_DYNAMIC44409 5028 Node: OMP_MAX_ACTIVE_LEVELS45005 5029 Node: OMP_MAX_TASK_PRIORITY45655 5030 Node: OMP_NESTED46313 5031 Node: OMP_NUM_THREADS46918 5032 Node: OMP_PROC_BIND47606 5033 Node: OMP_PLACES48798 5034 Node: OMP_STACKSIZE50975 5035 Node: OMP_SCHEDULE51799 5036 Node: OMP_THREAD_LIMIT52497 5037 Node: OMP_WAIT_POLICY53097 5038 Node: GOMP_CPU_AFFINITY53789 5039 Node: GOMP_DEBUG55519 5040 Node: GOMP_STACKSIZE56026 5041 Node: GOMP_SPINCOUNT56857 5042 Node: GOMP_RTEMS_THREAD_POOLS58061 5043 Node: Enabling OpenACC60239 5044 Node: OpenACC Runtime Library Routines61140 5045 Node: acc_get_num_devices65421 5046 Node: acc_set_device_type66147 5047 Node: acc_get_device_type66911 5048 Node: acc_set_device_num67924 5049 Node: acc_get_device_num68741 5050 Node: acc_get_property69540 5051 Node: acc_async_test71763 5052 Node: acc_async_test_all72751 5053 Node: acc_wait73651 5054 Node: acc_wait_all74514 5055 Node: acc_wait_all_async75275 5056 Node: acc_wait_async76027 5057 Node: acc_init76735 5058 Node: acc_shutdown77380 5059 Node: acc_on_device78047 5060 Node: acc_malloc79051 5061 Node: acc_free79550 5062 Node: acc_copyin79977 5063 Node: acc_present_or_copyin81564 5064 Node: acc_create83342 5065 Node: acc_present_or_create84974 5066 Node: acc_copyout86760 5067 Node: acc_delete89064 5068 Node: acc_update_device91311 5069 Node: acc_update_self92885 5070 Node: acc_map_data94475 5071 Node: acc_unmap_data95160 5072 Node: acc_deviceptr95681 5073 Node: acc_hostptr96251 5074 Node: acc_is_present96815 5075 Node: acc_memcpy_to_device98342 5076 Node: acc_memcpy_from_device99005 5077 Node: acc_attach99672 5078 Node: acc_detach100319 5079 Node: acc_get_current_cuda_device101098 5080 Node: acc_get_current_cuda_context101683 5081 Node: acc_get_cuda_stream102283 5082 Node: acc_set_cuda_stream102874 5083 Node: acc_prof_register103545 5084 Node: acc_prof_unregister104104 5085 Node: acc_prof_lookup104671 5086 Node: acc_register_library105192 5087 Node: OpenACC Environment Variables105758 5088 Node: ACC_DEVICE_TYPE106330 5089 Node: ACC_DEVICE_NUM106566 5090 Node: ACC_PROFLIB106820 5091 Node: GCC_ACC_NOTIFY107151 5092 Node: CUDA Streams Usage107371 5093 Ref: CUDA Streams Usage-Footnote-1109272 5094 Node: OpenACC Library Interoperability109381 5095 Ref: OpenACC Library Interoperability-Footnote-1115749 5096 Ref: OpenACC Library Interoperability-Footnote-2116001 5097 Node: OpenACC Profiling Interface116209 5098 Node: The libgomp ABI126233 5099 Node: Implementing MASTER construct127086 5100 Node: Implementing CRITICAL construct127502 5101 Node: Implementing ATOMIC construct128243 5102 Node: Implementing FLUSH construct128726 5103 Node: Implementing BARRIER construct128999 5104 Node: Implementing THREADPRIVATE construct129270 5105 Node: Implementing PRIVATE clause129925 5106 Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses130508 5107 Node: Implementing REDUCTION clause131834 5108 Node: Implementing PARALLEL construct132393 5109 Node: Implementing FOR construct133652 5110 Node: Implementing ORDERED construct135652 5111 Node: Implementing SECTIONS construct135960 5112 Node: Implementing SINGLE construct136728 5113 Node: Implementing OpenACC's PARALLEL construct137442 5114 Node: Reporting Bugs137702 5115 Node: Copying138065 5116 Node: GNU Free Documentation License175611 5117 Node: Funding200734 5118 Node: Library Index203260 5119 5120 End Tag Table 5121 5122 5123 Local Variables: 5124 coding: utf-8 5125 End: 5126
Indexes created Tue Mar 24 00:24:13 UTC 2026