libgomp.info revision 1.7
1 This is libgomp.info, produced by makeinfo version 4.13 from 2 /space/rguenther/gcc-6.4.0/gcc-6.4.0/libgomp/libgomp.texi. 3 4 Copyright (C) 2006-2016 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 22 funds for GNU development. 23 24 INFO-DIR-SECTION GNU Libraries 25 START-INFO-DIR-ENTRY 26 * libgomp: (libgomp). GNU Offloading and Multi Processing Runtime Library. 27 END-INFO-DIR-ENTRY 28 29 This manual documents libgomp, the GNU Offloading and Multi 30 Processing Runtime library. This is the GNU implementation of the 31 OpenMP and OpenACC APIs for parallel and accelerator programming in 32 C/C++ and Fortran. 33 34 Published by the Free Software Foundation 51 Franklin Street, Fifth 35 Floor Boston, MA 02110-1301 USA 36 37 Copyright (C) 2006-2016 Free Software Foundation, Inc. 38 39 Permission is granted to copy, distribute and/or modify this document 40 under the terms of the GNU Free Documentation License, Version 1.3 or 41 any later version published by the Free Software Foundation; with the 42 Invariant Sections being "Funding Free Software", the Front-Cover texts 43 being (a) (see below), and with the Back-Cover Texts being (b) (see 44 below). A copy of the license is included in the section entitled "GNU 45 Free Documentation License". 46 47 (a) The FSF's Front-Cover Text is: 48 49 A GNU Manual 50 51 (b) The FSF's Back-Cover Text is: 52 53 You have freedom to copy and modify this GNU Manual, like GNU 54 software. Copies published by the Free Software Foundation raise 55 funds for GNU development. 56 57 58 File: libgomp.info, Node: Top, Next: Enabling OpenMP, Up: (dir) 59 60 Introduction 61 ************ 62 63 This manual documents the usage of libgomp, the GNU Offloading and 64 Multi Processing Runtime Library. This includes the GNU implementation 65 of the OpenMP (http://www.openmp.org) Application Programming Interface 66 (API) for multi-platform shared-memory parallel programming in C/C++ 67 and Fortran, and the GNU implementation of the OpenACC 68 (http://www.openacc.org/) Application Programming Interface (API) for 69 offloading of code to accelerator devices in C/C++ and Fortran. 70 71 Originally, libgomp implemented the GNU OpenMP Runtime Library. 72 Based on this, support for OpenACC and offloading (both OpenACC and 73 OpenMP 4's target construct) has been added later on, and the library's 74 name changed to GNU Offloading and Multi Processing Runtime Library. 75 76 * Menu: 77 78 * Enabling OpenMP:: How to enable OpenMP for your applications. 79 * Runtime Library Routines:: The OpenMP runtime application programming 80 interface. 81 * Environment Variables:: Influencing runtime behavior with environment 82 variables. 83 * Enabling OpenACC:: How to enable OpenACC for your 84 applications. 85 * OpenACC Runtime Library Routines:: The OpenACC runtime application 86 programming interface. 87 * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with 88 environment variables. 89 * CUDA Streams Usage:: Notes on the implementation of 90 asynchronous operations. 91 * OpenACC Library Interoperability:: OpenACC library interoperability with the 92 NVIDIA CUBLAS library. 93 * The libgomp ABI:: Notes on the external ABI presented by libgomp. 94 * Reporting Bugs:: How to report bugs in the GNU Offloading and 95 Multi Processing Runtime Library. 96 * Copying:: GNU general public license says 97 how you can copy and share libgomp. 98 * GNU Free Documentation License:: 99 How you can copy and share this manual. 100 * Funding:: How to help assure continued work for free 101 software. 102 * Library Index:: Index of this documentation. 103 104 105 File: libgomp.info, Node: Enabling OpenMP, Next: Runtime Library Routines, Prev: Top, Up: Top 106 107 1 Enabling OpenMP 108 ***************** 109 110 To activate the OpenMP extensions for C/C++ and Fortran, the 111 compile-time flag `-fopenmp' must be specified. This enables the 112 OpenMP directive `#pragma omp' in C/C++ and `!$omp' directives in free 113 form, `c$omp', `*$omp' and `!$omp' directives in fixed form, `!$' 114 conditional compilation sentinels in free form and `c$', `*$' and `!$' 115 sentinels in fixed form, for Fortran. The flag also arranges for 116 automatic linking of the OpenMP runtime library (*note Runtime Library 117 Routines::). 118 119 A complete description of all OpenMP directives accepted may be 120 found in the OpenMP Application Program Interface 121 (http://www.openmp.org) manual, version 4.5. 122 123 124 File: libgomp.info, Node: Runtime Library Routines, Next: Environment Variables, Prev: Enabling OpenMP, Up: Top 125 126 2 Runtime Library Routines 127 ************************** 128 129 The runtime routines described here are defined by Section 3 of the 130 OpenMP specification in version 4.5. The routines are structured in 131 following three parts: 132 133 * Menu: 134 135 Control threads, processors and the parallel environment. They have C 136 linkage, and do not throw exceptions. 137 138 * omp_get_active_level:: Number of active parallel regions 139 * omp_get_ancestor_thread_num:: Ancestor thread ID 140 * omp_get_cancellation:: Whether cancellation support is enabled 141 * omp_get_default_device:: Get the default device for target regions 142 * omp_get_dynamic:: Dynamic teams setting 143 * omp_get_level:: Number of parallel regions 144 * omp_get_max_active_levels:: Maximum number of active regions 145 * omp_get_max_task_priority:: Maximum task priority value that can be set 146 * omp_get_max_threads:: Maximum number of threads of parallel region 147 * omp_get_nested:: Nested parallel regions 148 * omp_get_num_devices:: Number of target devices 149 * omp_get_num_procs:: Number of processors online 150 * omp_get_num_teams:: Number of teams 151 * omp_get_num_threads:: Size of the active team 152 * omp_get_proc_bind:: Whether theads may be moved between CPUs 153 * omp_get_schedule:: Obtain the runtime scheduling method 154 * omp_get_team_num:: Get team number 155 * omp_get_team_size:: Number of threads in a team 156 * omp_get_thread_limit:: Maximum number of threads 157 * omp_get_thread_num:: Current thread ID 158 * omp_in_parallel:: Whether a parallel region is active 159 * omp_in_final:: Whether in final or included task region 160 * omp_is_initial_device:: Whether executing on the host device 161 * omp_set_default_device:: Set the default device for target regions 162 * omp_set_dynamic:: Enable/disable dynamic teams 163 * omp_set_max_active_levels:: Limits the number of active parallel regions 164 * omp_set_nested:: Enable/disable nested parallel regions 165 * omp_set_num_threads:: Set upper team size limit 166 * omp_set_schedule:: Set the runtime scheduling method 167 168 Initialize, set, test, unset and destroy simple and nested locks. 169 170 * omp_init_lock:: Initialize simple lock 171 * omp_set_lock:: Wait for and set simple lock 172 * omp_test_lock:: Test and set simple lock if available 173 * omp_unset_lock:: Unset simple lock 174 * omp_destroy_lock:: Destroy simple lock 175 * omp_init_nest_lock:: Initialize nested lock 176 * omp_set_nest_lock:: Wait for and set simple lock 177 * omp_test_nest_lock:: Test and set nested lock if available 178 * omp_unset_nest_lock:: Unset nested lock 179 * omp_destroy_nest_lock:: Destroy nested lock 180 181 Portable, thread-based, wall clock timer. 182 183 * omp_get_wtick:: Get timer precision. 184 * omp_get_wtime:: Elapsed wall clock time. 185 186 187 File: libgomp.info, Node: omp_get_active_level, Next: omp_get_ancestor_thread_num, Up: Runtime Library Routines 188 189 2.1 `omp_get_active_level' - Number of parallel regions 190 ======================================================= 191 192 _Description_: 193 This function returns the nesting level for the active parallel 194 blocks, which enclose the calling call. 195 196 _C/C++_ 197 _Prototype_: `int omp_get_active_level(void);' 198 199 _Fortran_: 200 _Interface_: `integer function omp_get_active_level()' 201 202 _See also_: 203 *note omp_get_level::, *note omp_get_max_active_levels::, *note 204 omp_set_max_active_levels:: 205 206 _Reference_: 207 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.20. 208 209 210 File: libgomp.info, Node: omp_get_ancestor_thread_num, Next: omp_get_cancellation, Prev: omp_get_active_level, Up: Runtime Library Routines 211 212 2.2 `omp_get_ancestor_thread_num' - Ancestor thread ID 213 ====================================================== 214 215 _Description_: 216 This function returns the thread identification number for the 217 given nesting level of the current thread. For values of LEVEL 218 outside zero to `omp_get_level' -1 is returned; if LEVEL is 219 `omp_get_level' the result is identical to `omp_get_thread_num'. 220 221 _C/C++_ 222 _Prototype_: `int omp_get_ancestor_thread_num(int level);' 223 224 _Fortran_: 225 _Interface_: `integer function omp_get_ancestor_thread_num(level)' 226 `integer level' 227 228 _See also_: 229 *note omp_get_level::, *note omp_get_thread_num::, *note 230 omp_get_team_size:: 231 232 _Reference_: 233 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.18. 234 235 236 File: libgomp.info, Node: omp_get_cancellation, Next: omp_get_default_device, Prev: omp_get_ancestor_thread_num, Up: Runtime Library Routines 237 238 2.3 `omp_get_cancellation' - Whether cancellation support is enabled 239 ==================================================================== 240 241 _Description_: 242 This function returns `true' if cancellation is activated, `false' 243 otherwise. Here, `true' and `false' represent their 244 language-specific counterparts. Unless `OMP_CANCELLATION' is set 245 true, cancellations are deactivated. 246 247 _C/C++_: 248 _Prototype_: `int omp_get_cancellation(void);' 249 250 _Fortran_: 251 _Interface_: `logical function omp_get_cancellation()' 252 253 _See also_: 254 *note OMP_CANCELLATION:: 255 256 _Reference_: 257 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.9. 258 259 260 File: libgomp.info, Node: omp_get_default_device, Next: omp_get_dynamic, Prev: omp_get_cancellation, Up: Runtime Library Routines 261 262 2.4 `omp_get_default_device' - Get the default device for target regions 263 ======================================================================== 264 265 _Description_: 266 Get the default device for target regions without device clause. 267 268 _C/C++_: 269 _Prototype_: `int omp_get_default_device(void);' 270 271 _Fortran_: 272 _Interface_: `integer function omp_get_default_device()' 273 274 _See also_: 275 *note OMP_DEFAULT_DEVICE::, *note omp_set_default_device:: 276 277 _Reference_: 278 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.30. 279 280 281 File: libgomp.info, Node: omp_get_dynamic, Next: omp_get_level, Prev: omp_get_default_device, Up: Runtime Library Routines 282 283 2.5 `omp_get_dynamic' - Dynamic teams setting 284 ============================================= 285 286 _Description_: 287 This function returns `true' if enabled, `false' otherwise. Here, 288 `true' and `false' represent their language-specific counterparts. 289 290 The dynamic team setting may be initialized at startup by the 291 `OMP_DYNAMIC' environment variable or at runtime using 292 `omp_set_dynamic'. If undefined, dynamic adjustment is disabled 293 by default. 294 295 _C/C++_: 296 _Prototype_: `int omp_get_dynamic(void);' 297 298 _Fortran_: 299 _Interface_: `logical function omp_get_dynamic()' 300 301 _See also_: 302 *note omp_set_dynamic::, *note OMP_DYNAMIC:: 303 304 _Reference_: 305 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.8. 306 307 308 File: libgomp.info, Node: omp_get_level, Next: omp_get_max_active_levels, Prev: omp_get_dynamic, Up: Runtime Library Routines 309 310 2.6 `omp_get_level' - Obtain the current nesting level 311 ====================================================== 312 313 _Description_: 314 This function returns the nesting level for the parallel blocks, 315 which enclose the calling call. 316 317 _C/C++_ 318 _Prototype_: `int omp_get_level(void);' 319 320 _Fortran_: 321 _Interface_: `integer function omp_level()' 322 323 _See also_: 324 *note omp_get_active_level:: 325 326 _Reference_: 327 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.17. 328 329 330 File: libgomp.info, Node: omp_get_max_active_levels, Next: omp_get_max_task_priority, Prev: omp_get_level, Up: Runtime Library Routines 331 332 2.7 `omp_get_max_active_levels' - Maximum number of active regions 333 ================================================================== 334 335 _Description_: 336 This function obtains the maximum allowed number of nested, active 337 parallel regions. 338 339 _C/C++_ 340 _Prototype_: `int omp_get_max_active_levels(void);' 341 342 _Fortran_: 343 _Interface_: `integer function omp_get_max_active_levels()' 344 345 _See also_: 346 *note omp_set_max_active_levels::, *note omp_get_active_level:: 347 348 _Reference_: 349 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.16. 350 351 352 File: libgomp.info, Node: omp_get_max_task_priority, Next: omp_get_max_threads, Prev: omp_get_max_active_levels, Up: Runtime Library Routines 353 354 2.8 `omp_get_max_task_priority' - Maximum priority value 355 ======================================================== 356 357 that can be set for tasks. 358 _Description_: 359 This function obtains the maximum allowed priority number for 360 tasks. 361 362 _C/C++_ 363 _Prototype_: `int omp_get_max_task_priority(void);' 364 365 _Fortran_: 366 _Interface_: `integer function omp_get_max_task_priority()' 367 368 _Reference_: 369 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.29. 370 371 372 File: libgomp.info, Node: omp_get_max_threads, Next: omp_get_nested, Prev: omp_get_max_task_priority, Up: Runtime Library Routines 373 374 2.9 `omp_get_max_threads' - Maximum number of threads of parallel region 375 ======================================================================== 376 377 _Description_: 378 Return the maximum number of threads used for the current parallel 379 region that does not use the clause `num_threads'. 380 381 _C/C++_: 382 _Prototype_: `int omp_get_max_threads(void);' 383 384 _Fortran_: 385 _Interface_: `integer function omp_get_max_threads()' 386 387 _See also_: 388 *note omp_set_num_threads::, *note omp_set_dynamic::, *note 389 omp_get_thread_limit:: 390 391 _Reference_: 392 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.3. 393 394 395 File: libgomp.info, Node: omp_get_nested, Next: omp_get_num_devices, Prev: omp_get_max_threads, Up: Runtime Library Routines 396 397 2.10 `omp_get_nested' - Nested parallel regions 398 =============================================== 399 400 _Description_: 401 This function returns `true' if nested parallel regions are 402 enabled, `false' otherwise. Here, `true' and `false' represent 403 their language-specific counterparts. 404 405 Nested parallel regions may be initialized at startup by the 406 `OMP_NESTED' environment variable or at runtime using 407 `omp_set_nested'. If undefined, nested parallel regions are 408 disabled by default. 409 410 _C/C++_: 411 _Prototype_: `int omp_get_nested(void);' 412 413 _Fortran_: 414 _Interface_: `logical function omp_get_nested()' 415 416 _See also_: 417 *note omp_set_nested::, *note OMP_NESTED:: 418 419 _Reference_: 420 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.11. 421 422 423 File: libgomp.info, Node: omp_get_num_devices, Next: omp_get_num_procs, Prev: omp_get_nested, Up: Runtime Library Routines 424 425 2.11 `omp_get_num_devices' - Number of target devices 426 ===================================================== 427 428 _Description_: 429 Returns the number of target devices. 430 431 _C/C++_: 432 _Prototype_: `int omp_get_num_devices(void);' 433 434 _Fortran_: 435 _Interface_: `integer function omp_get_num_devices()' 436 437 _Reference_: 438 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.31. 439 440 441 File: libgomp.info, Node: omp_get_num_procs, Next: omp_get_num_teams, Prev: omp_get_num_devices, Up: Runtime Library Routines 442 443 2.12 `omp_get_num_procs' - Number of processors online 444 ====================================================== 445 446 _Description_: 447 Returns the number of processors online on that device. 448 449 _C/C++_: 450 _Prototype_: `int omp_get_num_procs(void);' 451 452 _Fortran_: 453 _Interface_: `integer function omp_get_num_procs()' 454 455 _Reference_: 456 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.5. 457 458 459 File: libgomp.info, Node: omp_get_num_teams, Next: omp_get_num_threads, Prev: omp_get_num_procs, Up: Runtime Library Routines 460 461 2.13 `omp_get_num_teams' - Number of teams 462 ========================================== 463 464 _Description_: 465 Returns the number of teams in the current team region. 466 467 _C/C++_: 468 _Prototype_: `int omp_get_num_teams(void);' 469 470 _Fortran_: 471 _Interface_: `integer function omp_get_num_teams()' 472 473 _Reference_: 474 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.32. 475 476 477 File: libgomp.info, Node: omp_get_num_threads, Next: omp_get_proc_bind, Prev: omp_get_num_teams, Up: Runtime Library Routines 478 479 2.14 `omp_get_num_threads' - Size of the active team 480 ==================================================== 481 482 _Description_: 483 Returns the number of threads in the current team. In a 484 sequential section of the program `omp_get_num_threads' returns 1. 485 486 The default team size may be initialized at startup by the 487 `OMP_NUM_THREADS' environment variable. At runtime, the size of 488 the current team may be set either by the `NUM_THREADS' clause or 489 by `omp_set_num_threads'. If none of the above were used to 490 define a specific value and `OMP_DYNAMIC' is disabled, one thread 491 per CPU online is used. 492 493 _C/C++_: 494 _Prototype_: `int omp_get_num_threads(void);' 495 496 _Fortran_: 497 _Interface_: `integer function omp_get_num_threads()' 498 499 _See also_: 500 *note omp_get_max_threads::, *note omp_set_num_threads::, *note 501 OMP_NUM_THREADS:: 502 503 _Reference_: 504 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.2. 505 506 507 File: libgomp.info, Node: omp_get_proc_bind, Next: omp_get_schedule, Prev: omp_get_num_threads, Up: Runtime Library Routines 508 509 2.15 `omp_get_proc_bind' - Whether theads may be moved between CPUs 510 =================================================================== 511 512 _Description_: 513 This functions returns the currently active thread affinity 514 policy, which is set via `OMP_PROC_BIND'. Possible values are 515 `omp_proc_bind_false', `omp_proc_bind_true', 516 `omp_proc_bind_master', `omp_proc_bind_close' and 517 `omp_proc_bind_spread'. 518 519 _C/C++_: 520 _Prototype_: `omp_proc_bind_t omp_get_proc_bind(void);' 521 522 _Fortran_: 523 _Interface_: `integer(kind=omp_proc_bind_kind) function 524 omp_get_proc_bind()' 525 526 _See also_: 527 *note OMP_PROC_BIND::, *note OMP_PLACES::, *note 528 GOMP_CPU_AFFINITY::, 529 530 _Reference_: 531 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.22. 532 533 534 File: libgomp.info, Node: omp_get_schedule, Next: omp_get_team_num, Prev: omp_get_proc_bind, Up: Runtime Library Routines 535 536 2.16 `omp_get_schedule' - Obtain the runtime scheduling method 537 ============================================================== 538 539 _Description_: 540 Obtain the runtime scheduling method. The KIND argument will be 541 set to the value `omp_sched_static', `omp_sched_dynamic', 542 `omp_sched_guided' or `omp_sched_auto'. The second argument, 543 CHUNK_SIZE, is set to the chunk size. 544 545 _C/C++_ 546 _Prototype_: `void omp_get_schedule(omp_sched_t *kind, int 547 *chunk_size);' 548 549 _Fortran_: 550 _Interface_: `subroutine omp_get_schedule(kind, chunk_size)' 551 `integer(kind=omp_sched_kind) kind' 552 `integer chunk_size' 553 554 _See also_: 555 *note omp_set_schedule::, *note OMP_SCHEDULE:: 556 557 _Reference_: 558 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.13. 559 560 561 File: libgomp.info, Node: omp_get_team_num, Next: omp_get_team_size, Prev: omp_get_schedule, Up: Runtime Library Routines 562 563 2.17 `omp_get_team_num' - Get team number 564 ========================================= 565 566 _Description_: 567 Returns the team number of the calling thread. 568 569 _C/C++_: 570 _Prototype_: `int omp_get_team_num(void);' 571 572 _Fortran_: 573 _Interface_: `integer function omp_get_team_num()' 574 575 _Reference_: 576 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.33. 577 578 579 File: libgomp.info, Node: omp_get_team_size, Next: omp_get_thread_limit, Prev: omp_get_team_num, Up: Runtime Library Routines 580 581 2.18 `omp_get_team_size' - Number of threads in a team 582 ====================================================== 583 584 _Description_: 585 This function returns the number of threads in a thread team to 586 which either the current thread or its ancestor belongs. For 587 values of LEVEL outside zero to `omp_get_level', -1 is returned; 588 if LEVEL is zero, 1 is returned, and for `omp_get_level', the 589 result is identical to `omp_get_num_threads'. 590 591 _C/C++_: 592 _Prototype_: `int omp_get_team_size(int level);' 593 594 _Fortran_: 595 _Interface_: `integer function omp_get_team_size(level)' 596 `integer level' 597 598 _See also_: 599 *note omp_get_num_threads::, *note omp_get_level::, *note 600 omp_get_ancestor_thread_num:: 601 602 _Reference_: 603 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.19. 604 605 606 File: libgomp.info, Node: omp_get_thread_limit, Next: omp_get_thread_num, Prev: omp_get_team_size, Up: Runtime Library Routines 607 608 2.19 `omp_get_thread_limit' - Maximum number of threads 609 ======================================================= 610 611 _Description_: 612 Return the maximum number of threads of the program. 613 614 _C/C++_: 615 _Prototype_: `int omp_get_thread_limit(void);' 616 617 _Fortran_: 618 _Interface_: `integer function omp_get_thread_limit()' 619 620 _See also_: 621 *note omp_get_max_threads::, *note OMP_THREAD_LIMIT:: 622 623 _Reference_: 624 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.14. 625 626 627 File: libgomp.info, Node: omp_get_thread_num, Next: omp_in_parallel, Prev: omp_get_thread_limit, Up: Runtime Library Routines 628 629 2.20 `omp_get_thread_num' - Current thread ID 630 ============================================= 631 632 _Description_: 633 Returns a unique thread identification number within the current 634 team. In a sequential parts of the program, `omp_get_thread_num' 635 always returns 0. In parallel regions the return value varies 636 from 0 to `omp_get_num_threads'-1 inclusive. The return value of 637 the master thread of a team is always 0. 638 639 _C/C++_: 640 _Prototype_: `int omp_get_thread_num(void);' 641 642 _Fortran_: 643 _Interface_: `integer function omp_get_thread_num()' 644 645 _See also_: 646 *note omp_get_num_threads::, *note omp_get_ancestor_thread_num:: 647 648 _Reference_: 649 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.4. 650 651 652 File: libgomp.info, Node: omp_in_parallel, Next: omp_in_final, Prev: omp_get_thread_num, Up: Runtime Library Routines 653 654 2.21 `omp_in_parallel' - Whether a parallel region is active 655 ============================================================ 656 657 _Description_: 658 This function returns `true' if currently running in parallel, 659 `false' otherwise. Here, `true' and `false' represent their 660 language-specific counterparts. 661 662 _C/C++_: 663 _Prototype_: `int omp_in_parallel(void);' 664 665 _Fortran_: 666 _Interface_: `logical function omp_in_parallel()' 667 668 _Reference_: 669 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.6. 670 671 672 File: libgomp.info, Node: omp_in_final, Next: omp_is_initial_device, Prev: omp_in_parallel, Up: Runtime Library Routines 673 674 2.22 `omp_in_final' - Whether in final or included task region 675 ============================================================== 676 677 _Description_: 678 This function returns `true' if currently running in a final or 679 included task region, `false' otherwise. Here, `true' and `false' 680 represent their language-specific counterparts. 681 682 _C/C++_: 683 _Prototype_: `int omp_in_final(void);' 684 685 _Fortran_: 686 _Interface_: `logical function omp_in_final()' 687 688 _Reference_: 689 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.21. 690 691 692 File: libgomp.info, Node: omp_is_initial_device, Next: omp_set_default_device, Prev: omp_in_final, Up: Runtime Library Routines 693 694 2.23 `omp_is_initial_device' - Whether executing on the host device 695 =================================================================== 696 697 _Description_: 698 This function returns `true' if currently running on the host 699 device, `false' otherwise. Here, `true' and `false' represent 700 their language-specific counterparts. 701 702 _C/C++_: 703 _Prototype_: `int omp_is_initial_device(void);' 704 705 _Fortran_: 706 _Interface_: `logical function omp_is_initial_device()' 707 708 _Reference_: 709 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.34. 710 711 712 File: libgomp.info, Node: omp_set_default_device, Next: omp_set_dynamic, Prev: omp_is_initial_device, Up: Runtime Library Routines 713 714 2.24 `omp_set_default_device' - Set the default device for target regions 715 ========================================================================= 716 717 _Description_: 718 Set the default device for target regions without device clause. 719 The argument shall be a nonnegative device number. 720 721 _C/C++_: 722 _Prototype_: `void omp_set_default_device(int device_num);' 723 724 _Fortran_: 725 _Interface_: `subroutine omp_set_default_device(device_num)' 726 `integer device_num' 727 728 _See also_: 729 *note OMP_DEFAULT_DEVICE::, *note omp_get_default_device:: 730 731 _Reference_: 732 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.29. 733 734 735 File: libgomp.info, Node: omp_set_dynamic, Next: omp_set_max_active_levels, Prev: omp_set_default_device, Up: Runtime Library Routines 736 737 2.25 `omp_set_dynamic' - Enable/disable dynamic teams 738 ===================================================== 739 740 _Description_: 741 Enable or disable the dynamic adjustment of the number of threads 742 within a team. The function takes the language-specific equivalent 743 of `true' and `false', where `true' enables dynamic adjustment of 744 team sizes and `false' disables it. 745 746 _C/C++_: 747 _Prototype_: `void omp_set_dynamic(int dynamic_threads);' 748 749 _Fortran_: 750 _Interface_: `subroutine omp_set_dynamic(dynamic_threads)' 751 `logical, intent(in) :: dynamic_threads' 752 753 _See also_: 754 *note OMP_DYNAMIC::, *note omp_get_dynamic:: 755 756 _Reference_: 757 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.7. 758 759 760 File: libgomp.info, Node: omp_set_max_active_levels, Next: omp_set_nested, Prev: omp_set_dynamic, Up: Runtime Library Routines 761 762 2.26 `omp_set_max_active_levels' - Limits the number of active parallel regions 763 =============================================================================== 764 765 _Description_: 766 This function limits the maximum allowed number of nested, active 767 parallel regions. 768 769 _C/C++_ 770 _Prototype_: `void omp_set_max_active_levels(int max_levels);' 771 772 _Fortran_: 773 _Interface_: `subroutine omp_set_max_active_levels(max_levels)' 774 `integer max_levels' 775 776 _See also_: 777 *note omp_get_max_active_levels::, *note omp_get_active_level:: 778 779 _Reference_: 780 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.15. 781 782 783 File: libgomp.info, Node: omp_set_nested, Next: omp_set_num_threads, Prev: omp_set_max_active_levels, Up: Runtime Library Routines 784 785 2.27 `omp_set_nested' - Enable/disable nested parallel regions 786 ============================================================== 787 788 _Description_: 789 Enable or disable nested parallel regions, i.e., whether team 790 members are allowed to create new teams. The function takes the 791 language-specific equivalent of `true' and `false', where `true' 792 enables dynamic adjustment of team sizes and `false' disables it. 793 794 _C/C++_: 795 _Prototype_: `void omp_set_nested(int nested);' 796 797 _Fortran_: 798 _Interface_: `subroutine omp_set_nested(nested)' 799 `logical, intent(in) :: nested' 800 801 _See also_: 802 *note OMP_NESTED::, *note omp_get_nested:: 803 804 _Reference_: 805 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.10. 806 807 808 File: libgomp.info, Node: omp_set_num_threads, Next: omp_set_schedule, Prev: omp_set_nested, Up: Runtime Library Routines 809 810 2.28 `omp_set_num_threads' - Set upper team size limit 811 ====================================================== 812 813 _Description_: 814 Specifies the number of threads used by default in subsequent 815 parallel sections, if those do not specify a `num_threads' clause. 816 The argument of `omp_set_num_threads' shall be a positive integer. 817 818 _C/C++_: 819 _Prototype_: `void omp_set_num_threads(int num_threads);' 820 821 _Fortran_: 822 _Interface_: `subroutine omp_set_num_threads(num_threads)' 823 `integer, intent(in) :: num_threads' 824 825 _See also_: 826 *note OMP_NUM_THREADS::, *note omp_get_num_threads::, *note 827 omp_get_max_threads:: 828 829 _Reference_: 830 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.1. 831 832 833 File: libgomp.info, Node: omp_set_schedule, Next: omp_init_lock, Prev: omp_set_num_threads, Up: Runtime Library Routines 834 835 2.29 `omp_set_schedule' - Set the runtime scheduling method 836 =========================================================== 837 838 _Description_: 839 Sets the runtime scheduling method. The KIND argument can have the 840 value `omp_sched_static', `omp_sched_dynamic', `omp_sched_guided' 841 or `omp_sched_auto'. Except for `omp_sched_auto', the chunk size 842 is set to the value of CHUNK_SIZE if positive, or to the default 843 value if zero or negative. For `omp_sched_auto' the CHUNK_SIZE 844 argument is ignored. 845 846 _C/C++_ 847 _Prototype_: `void omp_set_schedule(omp_sched_t kind, int 848 chunk_size);' 849 850 _Fortran_: 851 _Interface_: `subroutine omp_set_schedule(kind, chunk_size)' 852 `integer(kind=omp_sched_kind) kind' 853 `integer chunk_size' 854 855 _See also_: 856 *note omp_get_schedule:: *note OMP_SCHEDULE:: 857 858 _Reference_: 859 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.12. 860 861 862 File: libgomp.info, Node: omp_init_lock, Next: omp_set_lock, Prev: omp_set_schedule, Up: Runtime Library Routines 863 864 2.30 `omp_init_lock' - Initialize simple lock 865 ============================================= 866 867 _Description_: 868 Initialize a simple lock. After initialization, the lock is in an 869 unlocked state. 870 871 _C/C++_: 872 _Prototype_: `void omp_init_lock(omp_lock_t *lock);' 873 874 _Fortran_: 875 _Interface_: `subroutine omp_init_lock(svar)' 876 `integer(omp_lock_kind), intent(out) :: svar' 877 878 _See also_: 879 *note omp_destroy_lock:: 880 881 _Reference_: 882 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.1. 883 884 885 File: libgomp.info, Node: omp_set_lock, Next: omp_test_lock, Prev: omp_init_lock, Up: Runtime Library Routines 886 887 2.31 `omp_set_lock' - Wait for and set simple lock 888 ================================================== 889 890 _Description_: 891 Before setting a simple lock, the lock variable must be 892 initialized by `omp_init_lock'. The calling thread is blocked 893 until the lock is available. If the lock is already held by the 894 current thread, a deadlock occurs. 895 896 _C/C++_: 897 _Prototype_: `void omp_set_lock(omp_lock_t *lock);' 898 899 _Fortran_: 900 _Interface_: `subroutine omp_set_lock(svar)' 901 `integer(omp_lock_kind), intent(inout) :: svar' 902 903 _See also_: 904 *note omp_init_lock::, *note omp_test_lock::, *note 905 omp_unset_lock:: 906 907 _Reference_: 908 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.4. 909 910 911 File: libgomp.info, Node: omp_test_lock, Next: omp_unset_lock, Prev: omp_set_lock, Up: Runtime Library Routines 912 913 2.32 `omp_test_lock' - Test and set simple lock if available 914 ============================================================ 915 916 _Description_: 917 Before setting a simple lock, the lock variable must be 918 initialized by `omp_init_lock'. Contrary to `omp_set_lock', 919 `omp_test_lock' does not block if the lock is not available. This 920 function returns `true' upon success, `false' otherwise. Here, 921 `true' and `false' represent their language-specific counterparts. 922 923 _C/C++_: 924 _Prototype_: `int omp_test_lock(omp_lock_t *lock);' 925 926 _Fortran_: 927 _Interface_: `logical function omp_test_lock(svar)' 928 `integer(omp_lock_kind), intent(inout) :: svar' 929 930 _See also_: 931 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 932 933 _Reference_: 934 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.6. 935 936 937 File: libgomp.info, Node: omp_unset_lock, Next: omp_destroy_lock, Prev: omp_test_lock, Up: Runtime Library Routines 938 939 2.33 `omp_unset_lock' - Unset simple lock 940 ========================================= 941 942 _Description_: 943 A simple lock about to be unset must have been locked by 944 `omp_set_lock' or `omp_test_lock' before. In addition, the lock 945 must be held by the thread calling `omp_unset_lock'. Then, the 946 lock becomes unlocked. If one or more threads attempted to set 947 the lock before, one of them is chosen to, again, set the lock to 948 itself. 949 950 _C/C++_: 951 _Prototype_: `void omp_unset_lock(omp_lock_t *lock);' 952 953 _Fortran_: 954 _Interface_: `subroutine omp_unset_lock(svar)' 955 `integer(omp_lock_kind), intent(inout) :: svar' 956 957 _See also_: 958 *note omp_set_lock::, *note omp_test_lock:: 959 960 _Reference_: 961 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.5. 962 963 964 File: libgomp.info, Node: omp_destroy_lock, Next: omp_init_nest_lock, Prev: omp_unset_lock, Up: Runtime Library Routines 965 966 2.34 `omp_destroy_lock' - Destroy simple lock 967 ============================================= 968 969 _Description_: 970 Destroy a simple lock. In order to be destroyed, a simple lock 971 must be in the unlocked state. 972 973 _C/C++_: 974 _Prototype_: `void omp_destroy_lock(omp_lock_t *lock);' 975 976 _Fortran_: 977 _Interface_: `subroutine omp_destroy_lock(svar)' 978 `integer(omp_lock_kind), intent(inout) :: svar' 979 980 _See also_: 981 *note omp_init_lock:: 982 983 _Reference_: 984 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.3. 985 986 987 File: libgomp.info, Node: omp_init_nest_lock, Next: omp_set_nest_lock, Prev: omp_destroy_lock, Up: Runtime Library Routines 988 989 2.35 `omp_init_nest_lock' - Initialize nested lock 990 ================================================== 991 992 _Description_: 993 Initialize a nested lock. After initialization, the lock is in an 994 unlocked state and the nesting count is set to zero. 995 996 _C/C++_: 997 _Prototype_: `void omp_init_nest_lock(omp_nest_lock_t *lock);' 998 999 _Fortran_: 1000 _Interface_: `subroutine omp_init_nest_lock(nvar)' 1001 `integer(omp_nest_lock_kind), intent(out) :: nvar' 1002 1003 _See also_: 1004 *note omp_destroy_nest_lock:: 1005 1006 _Reference_: 1007 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.1. 1008 1009 1010 File: libgomp.info, Node: omp_set_nest_lock, Next: omp_test_nest_lock, Prev: omp_init_nest_lock, Up: Runtime Library Routines 1011 1012 2.36 `omp_set_nest_lock' - Wait for and set nested lock 1013 ======================================================= 1014 1015 _Description_: 1016 Before setting a nested lock, the lock variable must be 1017 initialized by `omp_init_nest_lock'. The calling thread is 1018 blocked until the lock is available. If the lock is already held 1019 by the current thread, the nesting count for the lock is 1020 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 (http://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 1043 initialized by `omp_init_nest_lock'. Contrary to 1044 `omp_set_nest_lock', `omp_test_nest_lock' does not block if the 1045 lock is not available. If the lock is already held by the current 1046 thread, the new nesting count is returned. Otherwise, the return 1047 value equals zero. 1048 1049 _C/C++_: 1050 _Prototype_: `int omp_test_nest_lock(omp_nest_lock_t *lock);' 1051 1052 _Fortran_: 1053 _Interface_: `logical function omp_test_nest_lock(nvar)' 1054 `integer(omp_nest_lock_kind), intent(inout) :: nvar' 1055 1056 _See also_: 1057 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 1058 1059 _Reference_: 1060 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.6. 1061 1062 1063 File: libgomp.info, Node: omp_unset_nest_lock, Next: omp_destroy_nest_lock, Prev: omp_test_nest_lock, Up: Runtime Library Routines 1064 1065 2.38 `omp_unset_nest_lock' - Unset nested lock 1066 ============================================== 1067 1068 _Description_: 1069 A nested lock about to be unset must have been locked by 1070 `omp_set_nested_lock' or `omp_test_nested_lock' before. In 1071 addition, the lock must be held by the thread calling 1072 `omp_unset_nested_lock'. If the nesting count drops to zero, the 1073 lock becomes unlocked. If one ore more threads attempted to set 1074 the lock before, one of them is chosen to, again, set the lock to 1075 itself. 1076 1077 _C/C++_: 1078 _Prototype_: `void omp_unset_nest_lock(omp_nest_lock_t *lock);' 1079 1080 _Fortran_: 1081 _Interface_: `subroutine omp_unset_nest_lock(nvar)' 1082 `integer(omp_nest_lock_kind), intent(inout) :: nvar' 1083 1084 _See also_: 1085 *note omp_set_nest_lock:: 1086 1087 _Reference_: 1088 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.5. 1089 1090 1091 File: libgomp.info, Node: omp_destroy_nest_lock, Next: omp_get_wtick, Prev: omp_unset_nest_lock, Up: Runtime Library Routines 1092 1093 2.39 `omp_destroy_nest_lock' - Destroy nested lock 1094 ================================================== 1095 1096 _Description_: 1097 Destroy a nested lock. In order to be destroyed, a nested lock 1098 must be in the unlocked state and its nesting count must equal 1099 zero. 1100 1101 _C/C++_: 1102 _Prototype_: `void omp_destroy_nest_lock(omp_nest_lock_t *);' 1103 1104 _Fortran_: 1105 _Interface_: `subroutine omp_destroy_nest_lock(nvar)' 1106 `integer(omp_nest_lock_kind), intent(inout) :: nvar' 1107 1108 _See also_: 1109 *note omp_init_lock:: 1110 1111 _Reference_: 1112 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.3. 1113 1114 1115 File: libgomp.info, Node: omp_get_wtick, Next: omp_get_wtime, Prev: omp_destroy_nest_lock, Up: Runtime Library Routines 1116 1117 2.40 `omp_get_wtick' - Get timer precision 1118 ========================================== 1119 1120 _Description_: 1121 Gets the timer precision, i.e., the number of seconds between two 1122 successive clock ticks. 1123 1124 _C/C++_: 1125 _Prototype_: `double omp_get_wtick(void);' 1126 1127 _Fortran_: 1128 _Interface_: `double precision function omp_get_wtick()' 1129 1130 _See also_: 1131 *note omp_get_wtime:: 1132 1133 _Reference_: 1134 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.4.2. 1135 1136 1137 File: libgomp.info, Node: omp_get_wtime, Prev: omp_get_wtick, Up: Runtime Library Routines 1138 1139 2.41 `omp_get_wtime' - Elapsed wall clock time 1140 ============================================== 1141 1142 _Description_: 1143 Elapsed wall clock time in seconds. The time is measured per 1144 thread, no guarantee can be made that two distinct threads measure 1145 the same time. Time is measured from some "time in the past", 1146 which is an arbitrary time guaranteed not to change during the 1147 execution of the program. 1148 1149 _C/C++_: 1150 _Prototype_: `double omp_get_wtime(void);' 1151 1152 _Fortran_: 1153 _Interface_: `double precision function omp_get_wtime()' 1154 1155 _See also_: 1156 *note omp_get_wtick:: 1157 1158 _Reference_: 1159 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.4.1. 1160 1161 1162 File: libgomp.info, Node: Environment Variables, Next: Enabling OpenACC, Prev: Runtime Library Routines, Up: Top 1163 1164 3 Environment Variables 1165 *********************** 1166 1167 The environment variables which beginning with `OMP_' are defined by 1168 section 4 of the OpenMP specification in version 4.5, while those 1169 beginning with `GOMP_' are GNU extensions. 1170 1171 * Menu: 1172 1173 * OMP_CANCELLATION:: Set whether cancellation is activated 1174 * OMP_DISPLAY_ENV:: Show OpenMP version and environment variables 1175 * OMP_DEFAULT_DEVICE:: Set the device used in target regions 1176 * OMP_DYNAMIC:: Dynamic adjustment of threads 1177 * OMP_MAX_ACTIVE_LEVELS:: Set the maximum number of nested parallel regions 1178 * OMP_MAX_TASK_PRIORITY:: Set the maximum task priority value 1179 * OMP_NESTED:: Nested parallel regions 1180 * OMP_NUM_THREADS:: Specifies the number of threads to use 1181 * OMP_PROC_BIND:: Whether theads may be moved between CPUs 1182 * OMP_PLACES:: Specifies on which CPUs the theads should be placed 1183 * OMP_STACKSIZE:: Set default thread stack size 1184 * OMP_SCHEDULE:: How threads are scheduled 1185 * OMP_THREAD_LIMIT:: Set the maximum number of threads 1186 * OMP_WAIT_POLICY:: How waiting threads are handled 1187 * GOMP_CPU_AFFINITY:: Bind threads to specific CPUs 1188 * GOMP_DEBUG:: Enable debugging output 1189 * GOMP_STACKSIZE:: Set default thread stack size 1190 * GOMP_SPINCOUNT:: Set the busy-wait spin count 1191 * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools 1192 1193 1194 File: libgomp.info, Node: OMP_CANCELLATION, Next: OMP_DISPLAY_ENV, Up: Environment Variables 1195 1196 3.1 `OMP_CANCELLATION' - Set whether cancellation is activated 1197 ============================================================== 1198 1199 _Description_: 1200 If set to `TRUE', the cancellation is activated. If set to 1201 `FALSE' or if unset, cancellation is disabled and the `cancel' 1202 construct is ignored. 1203 1204 _See also_: 1205 *note omp_get_cancellation:: 1206 1207 _Reference_: 1208 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.11 1209 1210 1211 File: libgomp.info, Node: OMP_DISPLAY_ENV, Next: OMP_DEFAULT_DEVICE, Prev: OMP_CANCELLATION, Up: Environment Variables 1212 1213 3.2 `OMP_DISPLAY_ENV' - Show OpenMP version and environment variables 1214 ===================================================================== 1215 1216 _Description_: 1217 If set to `TRUE', the OpenMP version number and the values 1218 associated with the OpenMP environment variables are printed to 1219 `stderr'. If set to `VERBOSE', it additionally shows the value of 1220 the environment variables which are GNU extensions. If undefined 1221 or set to `FALSE', this information will not be shown. 1222 1223 _Reference_: 1224 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.12 1225 1226 1227 File: libgomp.info, Node: OMP_DEFAULT_DEVICE, Next: OMP_DYNAMIC, Prev: OMP_DISPLAY_ENV, Up: Environment Variables 1228 1229 3.3 `OMP_DEFAULT_DEVICE' - Set the device used in target regions 1230 ================================================================ 1231 1232 _Description_: 1233 Set to choose the device which is used in a `target' region, 1234 unless the value is overridden by `omp_set_default_device' or by a 1235 `device' clause. The value shall be the nonnegative device 1236 number. If no device with the given device number exists, the code 1237 is executed on the host. If unset, device number 0 will be used. 1238 1239 _See also_: 1240 *note omp_get_default_device::, *note omp_set_default_device::, 1241 1242 _Reference_: 1243 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.13 1244 1245 1246 File: libgomp.info, Node: OMP_DYNAMIC, Next: OMP_MAX_ACTIVE_LEVELS, Prev: OMP_DEFAULT_DEVICE, Up: Environment Variables 1247 1248 3.4 `OMP_DYNAMIC' - Dynamic adjustment of threads 1249 ================================================= 1250 1251 _Description_: 1252 Enable or disable the dynamic adjustment of the number of threads 1253 within a team. The value of this environment variable shall be 1254 `TRUE' or `FALSE'. If undefined, dynamic adjustment is disabled 1255 by default. 1256 1257 _See also_: 1258 *note omp_set_dynamic:: 1259 1260 _Reference_: 1261 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.3 1262 1263 1264 File: libgomp.info, Node: OMP_MAX_ACTIVE_LEVELS, Next: OMP_MAX_TASK_PRIORITY, Prev: OMP_DYNAMIC, Up: Environment Variables 1265 1266 3.5 `OMP_MAX_ACTIVE_LEVELS' - Set the maximum number of nested parallel regions 1267 =============================================================================== 1268 1269 _Description_: 1270 Specifies the initial value for the maximum number of nested 1271 parallel regions. The value of this variable shall be a positive 1272 integer. If undefined, the number of active levels is unlimited. 1273 1274 _See also_: 1275 *note omp_set_max_active_levels:: 1276 1277 _Reference_: 1278 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.9 1279 1280 1281 File: libgomp.info, Node: OMP_MAX_TASK_PRIORITY, Next: OMP_NESTED, Prev: OMP_MAX_ACTIVE_LEVELS, Up: Environment Variables 1282 1283 3.6 `OMP_MAX_TASK_PRIORITY' - Set the maximum priority 1284 ====================================================== 1285 1286 number that can be set for a task. 1287 _Description_: 1288 Specifies the initial value for the maximum priority value that 1289 can be set for a task. The value of this variable shall be a 1290 non-negative integer, and zero is allowed. If undefined, the 1291 default priority is 0. 1292 1293 _See also_: 1294 *note omp_get_max_task_priority:: 1295 1296 _Reference_: 1297 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.14 1298 1299 1300 File: libgomp.info, Node: OMP_NESTED, Next: OMP_NUM_THREADS, Prev: OMP_MAX_TASK_PRIORITY, Up: Environment Variables 1301 1302 3.7 `OMP_NESTED' - Nested parallel regions 1303 ========================================== 1304 1305 _Description_: 1306 Enable or disable nested parallel regions, i.e., whether team 1307 members are allowed to create new teams. The value of this 1308 environment variable shall be `TRUE' or `FALSE'. If undefined, 1309 nested parallel regions are disabled by default. 1310 1311 _See also_: 1312 *note omp_set_nested:: 1313 1314 _Reference_: 1315 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.6 1316 1317 1318 File: libgomp.info, Node: OMP_NUM_THREADS, Next: OMP_PROC_BIND, Prev: OMP_NESTED, Up: Environment Variables 1319 1320 3.8 `OMP_NUM_THREADS' - Specifies the number of threads to use 1321 ============================================================== 1322 1323 _Description_: 1324 Specifies the default number of threads to use in parallel 1325 regions. The value of this variable shall be a comma-separated 1326 list of positive integers; the value specified the number of 1327 threads to use for the corresponding nested level. If undefined 1328 one thread per CPU is used. 1329 1330 _See also_: 1331 *note omp_set_num_threads:: 1332 1333 _Reference_: 1334 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.2 1335 1336 1337 File: libgomp.info, Node: OMP_PROC_BIND, Next: OMP_PLACES, Prev: OMP_NUM_THREADS, Up: Environment Variables 1338 1339 3.9 `OMP_PROC_BIND' - Whether theads may be moved between CPUs 1340 ============================================================== 1341 1342 _Description_: 1343 Specifies whether threads may be moved between processors. If set 1344 to `TRUE', OpenMP theads should not be moved; if set to `FALSE' 1345 they may be moved. Alternatively, a comma separated list with the 1346 values `MASTER', `CLOSE' and `SPREAD' can be used to specify the 1347 thread affinity policy for the corresponding nesting level. With 1348 `MASTER' the worker threads are in the same place partition as the 1349 master thread. With `CLOSE' those are kept close to the master 1350 thread in contiguous place partitions. And with `SPREAD' a sparse 1351 distribution across the place partitions is used. 1352 1353 When undefined, `OMP_PROC_BIND' defaults to `TRUE' when 1354 `OMP_PLACES' or `GOMP_CPU_AFFINITY' is set and `FALSE' otherwise. 1355 1356 _See also_: 1357 *note OMP_PLACES::, *note GOMP_CPU_AFFINITY::, *note 1358 omp_get_proc_bind:: 1359 1360 _Reference_: 1361 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.4 1362 1363 1364 File: libgomp.info, Node: OMP_PLACES, Next: OMP_STACKSIZE, Prev: OMP_PROC_BIND, Up: Environment Variables 1365 1366 3.10 `OMP_PLACES' - Specifies on which CPUs the theads should be placed 1367 ======================================================================= 1368 1369 _Description_: 1370 The thread placement can be either specified using an abstract 1371 name or by an explicit list of the places. The abstract names 1372 `threads', `cores' and `sockets' can be optionally followed by a 1373 positive number in parentheses, which denotes the how many places 1374 shall be created. With `threads' each place corresponds to a 1375 single hardware thread; `cores' to a single core with the 1376 corresponding number of hardware threads; and with `sockets' the 1377 place corresponds to a single socket. The resulting placement can 1378 be shown by setting the `OMP_DISPLAY_ENV' environment variable. 1379 1380 Alternatively, the placement can be specified explicitly as 1381 comma-separated list of places. A place is specified by set of 1382 nonnegative numbers in curly braces, denoting the denoting the 1383 hardware threads. The hardware threads belonging to a place can 1384 either be specified as comma-separated list of nonnegative thread 1385 numbers or using an interval. Multiple places can also be either 1386 specified by a comma-separated list of places or by an interval. 1387 To specify an interval, a colon followed by the count is placed 1388 after after the hardware thread number or the place. Optionally, 1389 the length can be followed by a colon and the stride number - 1390 otherwise a unit stride is assumed. For instance, the following 1391 specifies the same places list: `"{0,1,2}, {3,4,6}, {7,8,9}, 1392 {10,11,12}"'; `"{0:3}, {3:3}, {7:3}, {10:3}"'; and `"{0:2}:4:3"'. 1393 1394 If `OMP_PLACES' and `GOMP_CPU_AFFINITY' are unset and 1395 `OMP_PROC_BIND' is either unset or `false', threads may be moved 1396 between CPUs following no placement policy. 1397 1398 _See also_: 1399 *note OMP_PROC_BIND::, *note GOMP_CPU_AFFINITY::, *note 1400 omp_get_proc_bind::, *note OMP_DISPLAY_ENV:: 1401 1402 _Reference_: 1403 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.5 1404 1405 1406 File: libgomp.info, Node: OMP_STACKSIZE, Next: OMP_SCHEDULE, Prev: OMP_PLACES, Up: Environment Variables 1407 1408 3.11 `OMP_STACKSIZE' - Set default thread stack size 1409 ==================================================== 1410 1411 _Description_: 1412 Set the default thread stack size in kilobytes, unless the number 1413 is suffixed by `B', `K', `M' or `G', in which case the size is, 1414 respectively, in bytes, kilobytes, megabytes or gigabytes. This 1415 is different from `pthread_attr_setstacksize' which gets the 1416 number of bytes as an argument. If the stack size cannot be set 1417 due to system constraints, an error is reported and the initial 1418 stack size is left unchanged. If undefined, the stack size is 1419 system dependent. 1420 1421 _Reference_: 1422 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.7 1423 1424 1425 File: libgomp.info, Node: OMP_SCHEDULE, Next: OMP_THREAD_LIMIT, Prev: OMP_STACKSIZE, Up: Environment Variables 1426 1427 3.12 `OMP_SCHEDULE' - How threads are scheduled 1428 =============================================== 1429 1430 _Description_: 1431 Allows to specify `schedule type' and `chunk size'. The value of 1432 the variable shall have the form: `type[,chunk]' where `type' is 1433 one of `static', `dynamic', `guided' or `auto' The optional 1434 `chunk' size shall be a positive integer. If undefined, dynamic 1435 scheduling and a chunk size of 1 is used. 1436 1437 _See also_: 1438 *note omp_set_schedule:: 1439 1440 _Reference_: 1441 OpenMP specification v4.5 (http://www.openmp.org/), Sections 1442 2.7.1.1 and 4.1 1443 1444 1445 File: libgomp.info, Node: OMP_THREAD_LIMIT, Next: OMP_WAIT_POLICY, Prev: OMP_SCHEDULE, Up: Environment Variables 1446 1447 3.13 `OMP_THREAD_LIMIT' - Set the maximum number of threads 1448 =========================================================== 1449 1450 _Description_: 1451 Specifies the number of threads to use for the whole program. The 1452 value of this variable shall be a positive integer. If undefined, 1453 the number of threads is not limited. 1454 1455 _See also_: 1456 *note OMP_NUM_THREADS::, *note omp_get_thread_limit:: 1457 1458 _Reference_: 1459 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.10 1460 1461 1462 File: libgomp.info, Node: OMP_WAIT_POLICY, Next: GOMP_CPU_AFFINITY, Prev: OMP_THREAD_LIMIT, Up: Environment Variables 1463 1464 3.14 `OMP_WAIT_POLICY' - How waiting threads are handled 1465 ======================================================== 1466 1467 _Description_: 1468 Specifies whether waiting threads should be active or passive. If 1469 the value is `PASSIVE', waiting threads should not consume CPU 1470 power while waiting; while the value is `ACTIVE' specifies that 1471 they should. If undefined, threads wait actively for a short time 1472 before waiting passively. 1473 1474 _See also_: 1475 *note GOMP_SPINCOUNT:: 1476 1477 _Reference_: 1478 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.8 1479 1480 1481 File: libgomp.info, Node: GOMP_CPU_AFFINITY, Next: GOMP_DEBUG, Prev: OMP_WAIT_POLICY, Up: Environment Variables 1482 1483 3.15 `GOMP_CPU_AFFINITY' - Bind threads to specific CPUs 1484 ======================================================== 1485 1486 _Description_: 1487 Binds threads to specific CPUs. The variable should contain a 1488 space-separated or comma-separated list of CPUs. This list may 1489 contain different kinds of entries: either single CPU numbers in 1490 any order, a range of CPUs (M-N) or a range with some stride 1491 (M-N:S). CPU numbers are zero based. For example, 1492 `GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"' will bind the initial thread 1493 to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to 1494 CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 1495 10, 12, and 14 respectively and then start assigning back from the 1496 beginning of the list. `GOMP_CPU_AFFINITY=0' binds all threads to 1497 CPU 0. 1498 1499 There is no libgomp library routine to determine whether a CPU 1500 affinity specification is in effect. As a workaround, 1501 language-specific library functions, e.g., `getenv' in C or 1502 `GET_ENVIRONMENT_VARIABLE' in Fortran, may be used to query the 1503 setting of the `GOMP_CPU_AFFINITY' environment variable. A 1504 defined CPU affinity on startup cannot be changed or disabled 1505 during the runtime of the application. 1506 1507 If both `GOMP_CPU_AFFINITY' and `OMP_PROC_BIND' are set, 1508 `OMP_PROC_BIND' has a higher precedence. If neither has been set 1509 and `OMP_PROC_BIND' is unset, or when `OMP_PROC_BIND' is set to 1510 `FALSE', the host system will handle the assignment of threads to 1511 CPUs. 1512 1513 _See also_: 1514 *note OMP_PLACES::, *note OMP_PROC_BIND:: 1515 1516 1517 File: libgomp.info, Node: GOMP_DEBUG, Next: GOMP_STACKSIZE, Prev: GOMP_CPU_AFFINITY, Up: Environment Variables 1518 1519 3.16 `GOMP_DEBUG' - Enable debugging output 1520 =========================================== 1521 1522 _Description_: 1523 Enable debugging output. The variable should be set to `0' 1524 (disabled, also the default if not set), or `1' (enabled). 1525 1526 If enabled, some debugging output will be printed during execution. 1527 This is currently not specified in more detail, and subject to 1528 change. 1529 1530 1531 File: libgomp.info, Node: GOMP_STACKSIZE, Next: GOMP_SPINCOUNT, Prev: GOMP_DEBUG, Up: Environment Variables 1532 1533 3.17 `GOMP_STACKSIZE' - Set default thread stack size 1534 ===================================================== 1535 1536 _Description_: 1537 Set the default thread stack size in kilobytes. This is different 1538 from `pthread_attr_setstacksize' which gets the number of bytes as 1539 an argument. If the stack size cannot be set due to system 1540 constraints, an error is reported and the initial stack size is 1541 left unchanged. If undefined, the stack size is system dependent. 1542 1543 _See also_: 1544 *note OMP_STACKSIZE:: 1545 1546 _Reference_: 1547 GCC Patches Mailinglist 1548 (http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html), GCC 1549 Patches Mailinglist 1550 (http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html) 1551 1552 1553 File: libgomp.info, Node: GOMP_SPINCOUNT, Next: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_STACKSIZE, Up: Environment Variables 1554 1555 3.18 `GOMP_SPINCOUNT' - Set the busy-wait spin count 1556 ==================================================== 1557 1558 _Description_: 1559 Determines how long a threads waits actively with consuming CPU 1560 power before waiting passively without consuming CPU power. The 1561 value may be either `INFINITE', `INFINITY' to always wait actively 1562 or an integer which gives the number of spins of the busy-wait 1563 loop. The integer may optionally be followed by the following 1564 suffixes acting as multiplication factors: `k' (kilo, thousand), 1565 `M' (mega, million), `G' (giga, billion), or `T' (tera, trillion). 1566 If undefined, 0 is used when `OMP_WAIT_POLICY' is `PASSIVE', 1567 300,000 is used when `OMP_WAIT_POLICY' is undefined and 30 billion 1568 is used when `OMP_WAIT_POLICY' is `ACTIVE'. If there are more 1569 OpenMP threads than available CPUs, 1000 and 100 spins are used 1570 for `OMP_WAIT_POLICY' being `ACTIVE' or undefined, respectively; 1571 unless the `GOMP_SPINCOUNT' is lower or `OMP_WAIT_POLICY' is 1572 `PASSIVE'. 1573 1574 _See also_: 1575 *note OMP_WAIT_POLICY:: 1576 1577 1578 File: libgomp.info, Node: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_SPINCOUNT, Up: Environment Variables 1579 1580 3.19 `GOMP_RTEMS_THREAD_POOLS' - Set the RTEMS specific thread pools 1581 ==================================================================== 1582 1583 _Description_: 1584 This environment variable is only used on the RTEMS real-time 1585 operating system. It determines the scheduler instance specific 1586 thread pools. The format for `GOMP_RTEMS_THREAD_POOLS' is a list 1587 of optional `<thread-pool-count>[$<priority>]@<scheduler-name>' 1588 configurations separated by `:' where: 1589 * `<thread-pool-count>' is the thread pool count for this 1590 scheduler instance. 1591 1592 * `$<priority>' is an optional priority for the worker threads 1593 of a thread pool according to `pthread_setschedparam'. In 1594 case a priority value is omitted, then a worker thread will 1595 inherit the priority of the OpenMP master thread that created 1596 it. The priority of the worker thread is not changed after 1597 creation, even if a new OpenMP master thread using the worker 1598 has a different priority. 1599 1600 * `@<scheduler-name>' is the scheduler instance name according 1601 to the RTEMS application configuration. 1602 In case no thread pool configuration is specified for a scheduler 1603 instance, then each OpenMP master thread of this scheduler 1604 instance will use its own dynamically allocated thread pool. To 1605 limit the worker thread count of the thread pools, each OpenMP 1606 master thread must call `omp_set_num_threads'. 1607 1608 _Example_: 1609 Lets suppose we have three scheduler instances `IO', `WRK0', and 1610 `WRK1' with `GOMP_RTEMS_THREAD_POOLS' set to `"1@WRK0:3$4@WRK1"'. 1611 Then there are no thread pool restrictions for scheduler instance 1612 `IO'. In the scheduler instance `WRK0' there is one thread pool 1613 available. Since no priority is specified for this scheduler 1614 instance, the worker thread inherits the priority of the OpenMP 1615 master thread that created it. In the scheduler instance `WRK1' 1616 there are three thread pools available and their worker threads 1617 run at priority four. 1618 1619 1620 File: libgomp.info, Node: Enabling OpenACC, Next: OpenACC Runtime Library Routines, Prev: Environment Variables, Up: Top 1621 1622 4 Enabling OpenACC 1623 ****************** 1624 1625 To activate the OpenACC extensions for C/C++ and Fortran, the 1626 compile-time flag `-fopenacc' must be specified. This enables the 1627 OpenACC directive `#pragma acc' in C/C++ and `!$accp' directives in 1628 free form, `c$acc', `*$acc' and `!$acc' directives in fixed form, `!$' 1629 conditional compilation sentinels in free form and `c$', `*$' and `!$' 1630 sentinels in fixed form, for Fortran. The flag also arranges for 1631 automatic linking of the OpenACC runtime library (*note OpenACC Runtime 1632 Library Routines::). 1633 1634 A complete description of all OpenACC directives accepted may be 1635 found in the OpenACC (http://www.openacc.org/) Application Programming 1636 Interface manual, version 2.0. 1637 1638 Note that this is an experimental feature and subject to change in 1639 future versions of GCC. See `https://gcc.gnu.org/wiki/OpenACC' for 1640 more information. 1641 1642 1643 File: libgomp.info, Node: OpenACC Runtime Library Routines, Next: OpenACC Environment Variables, Prev: Enabling OpenACC, Up: Top 1644 1645 5 OpenACC Runtime Library Routines 1646 ********************************** 1647 1648 The runtime routines described here are defined by section 3 of the 1649 OpenACC specifications in version 2.0. They have C linkage, and do not 1650 throw exceptions. Generally, they are available only for the host, 1651 with the exception of `acc_on_device', which is available for both the 1652 host and the acceleration device. 1653 1654 * Menu: 1655 1656 * acc_get_num_devices:: Get number of devices for the given device 1657 type. 1658 * acc_set_device_type:: Set type of device accelerator to use. 1659 * acc_get_device_type:: Get type of device accelerator to be used. 1660 * acc_set_device_num:: Set device number to use. 1661 * acc_get_device_num:: Get device number to be used. 1662 * acc_async_test:: Tests for completion of a specific asynchronous 1663 operation. 1664 * acc_async_test_all:: Tests for completion of all asychronous 1665 operations. 1666 * acc_wait:: Wait for completion of a specific asynchronous 1667 operation. 1668 * acc_wait_all:: Waits for completion of all asyncrhonous 1669 operations. 1670 * acc_wait_all_async:: Wait for completion of all asynchronous 1671 operations. 1672 * acc_wait_async:: Wait for completion of asynchronous operations. 1673 * acc_init:: Initialize runtime for a specific device type. 1674 * acc_shutdown:: Shuts down the runtime for a specific device 1675 type. 1676 * acc_on_device:: Whether executing on a particular device 1677 * acc_malloc:: Allocate device memory. 1678 * acc_free:: Free device memory. 1679 * acc_copyin:: Allocate device memory and copy host memory to 1680 it. 1681 * acc_present_or_copyin:: If the data is not present on the device, 1682 allocate device memory and copy from host 1683 memory. 1684 * acc_create:: Allocate device memory and map it to host 1685 memory. 1686 * acc_present_or_create:: If the data is not present on the device, 1687 allocate device memory and map it to host 1688 memory. 1689 * acc_copyout:: Copy device memory to host memory. 1690 * acc_delete:: Free device memory. 1691 * acc_update_device:: Update device memory from mapped host memory. 1692 * acc_update_self:: Update host memory from mapped device memory. 1693 * acc_map_data:: Map previously allocated device memory to host 1694 memory. 1695 * acc_unmap_data:: Unmap device memory from host memory. 1696 * acc_deviceptr:: Get device pointer associated with specific 1697 host address. 1698 * acc_hostptr:: Get host pointer associated with specific 1699 device address. 1700 * acc_is_present:: Indiciate whether host variable / array is 1701 present on device. 1702 * acc_memcpy_to_device:: Copy host memory to device memory. 1703 * acc_memcpy_from_device:: Copy device memory to host memory. 1704 1705 API routines for target platforms. 1706 1707 * acc_get_current_cuda_device:: Get CUDA device handle. 1708 * acc_get_current_cuda_context::Get CUDA context handle. 1709 * acc_get_cuda_stream:: Get CUDA stream handle. 1710 * acc_set_cuda_stream:: Set CUDA stream handle. 1711 1712 1713 File: libgomp.info, Node: acc_get_num_devices, Next: acc_set_device_type, Up: OpenACC Runtime Library Routines 1714 1715 5.1 `acc_get_num_devices' - Get number of devices for given device type 1716 ======================================================================= 1717 1718 _Description_ 1719 This function returns a value indicating the number of devices 1720 available for the device type specified in DEVICETYPE. 1721 1722 _C/C++_: 1723 _Prototype_: `int acc_get_num_devices(acc_device_t devicetype);' 1724 1725 _Fortran_: 1726 _Interface_: `integer function acc_get_num_devices(devicetype)' 1727 `integer(kind=acc_device_kind) devicetype' 1728 1729 _Reference_: 1730 OpenACC specification v2.0 (http://www.openacc.org/), section 1731 3.2.1. 1732 1733 1734 File: libgomp.info, Node: acc_set_device_type, Next: acc_get_device_type, Prev: acc_get_num_devices, Up: OpenACC Runtime Library Routines 1735 1736 5.2 `acc_set_device_type' - Set type of device accelerator to use. 1737 ================================================================== 1738 1739 _Description_ 1740 This function indicates to the runtime library which device typr, 1741 specified in DEVICETYPE, to use when executing a parallel or 1742 kernels region. 1743 1744 _C/C++_: 1745 _Prototype_: `acc_set_device_type(acc_device_t devicetype);' 1746 1747 _Fortran_: 1748 _Interface_: `subroutine acc_set_device_type(devicetype)' 1749 `integer(kind=acc_device_kind) devicetype' 1750 1751 _Reference_: 1752 OpenACC specification v2.0 (http://www.openacc.org/), section 1753 3.2.2. 1754 1755 1756 File: libgomp.info, Node: acc_get_device_type, Next: acc_set_device_num, Prev: acc_set_device_type, Up: OpenACC Runtime Library Routines 1757 1758 5.3 `acc_get_device_type' - Get type of device accelerator to be used. 1759 ====================================================================== 1760 1761 _Description_ 1762 This function returns what device type will be used when executing 1763 a parallel or kernels region. 1764 1765 _C/C++_: 1766 _Prototype_: `acc_device_t acc_get_device_type(void);' 1767 1768 _Fortran_: 1769 _Interface_: `function acc_get_device_type(void)' 1770 `integer(kind=acc_device_kind) acc_get_device_type' 1771 1772 _Reference_: 1773 OpenACC specification v2.0 (http://www.openacc.org/), section 1774 3.2.3. 1775 1776 1777 File: libgomp.info, Node: acc_set_device_num, Next: acc_get_device_num, Prev: acc_get_device_type, Up: OpenACC Runtime Library Routines 1778 1779 5.4 `acc_set_device_num' - Set device number to use. 1780 ==================================================== 1781 1782 _Description_ 1783 This function will indicate to the runtime which device number, 1784 specified by NUM, associated with the specifed device type 1785 DEVICETYPE. 1786 1787 _C/C++_: 1788 _Prototype_: `acc_set_device_num(int num, acc_device_t 1789 devicetype);' 1790 1791 _Fortran_: 1792 _Interface_: `subroutine acc_set_device_num(devicenum, devicetype)' 1793 `integer devicenum' 1794 `integer(kind=acc_device_kind) devicetype' 1795 1796 _Reference_: 1797 OpenACC specification v2.0 (http://www.openacc.org/), section 1798 3.2.4. 1799 1800 1801 File: libgomp.info, Node: acc_get_device_num, Next: acc_async_test, Prev: acc_set_device_num, Up: OpenACC Runtime Library Routines 1802 1803 5.5 `acc_get_device_num' - Get device number to be used. 1804 ======================================================== 1805 1806 _Description_ 1807 This function returns which device number associated with the 1808 specified device type DEVICETYPE, will be used when executing a 1809 parallel or kernels region. 1810 1811 _C/C++_: 1812 _Prototype_: `int acc_get_device_num(acc_device_t devicetype);' 1813 1814 _Fortran_: 1815 _Interface_: `function acc_get_device_num(devicetype)' 1816 `integer(kind=acc_device_kind) devicetype' 1817 `integer acc_get_device_num' 1818 1819 _Reference_: 1820 OpenACC specification v2.0 (http://www.openacc.org/), section 1821 3.2.5. 1822 1823 1824 File: libgomp.info, Node: acc_async_test, Next: acc_async_test_all, Prev: acc_get_device_num, Up: OpenACC Runtime Library Routines 1825 1826 5.6 `acc_async_test' - Test for completion of a specific asynchronous operation. 1827 ================================================================================ 1828 1829 _Description_ 1830 This function tests for completion of the asynchrounous operation 1831 specified in ARG. In C/C++, a non-zero value will be returned to 1832 indicate the specified asynchronous operation has completed. While 1833 Fortran will return a `true'. If the asynchrounous operation has 1834 not completed, C/C++ returns a zero and Fortran returns a `false'. 1835 1836 _C/C++_: 1837 _Prototype_: `int acc_async_test(int arg);' 1838 1839 _Fortran_: 1840 _Interface_: `function acc_async_test(arg)' 1841 `integer(kind=acc_handle_kind) arg' 1842 `logical acc_async_test' 1843 1844 _Reference_: 1845 OpenACC specification v2.0 (http://www.openacc.org/), section 1846 3.2.6. 1847 1848 1849 File: libgomp.info, Node: acc_async_test_all, Next: acc_wait, Prev: acc_async_test, Up: OpenACC Runtime Library Routines 1850 1851 5.7 `acc_async_test_all' - Tests for completion of all asynchronous operations. 1852 =============================================================================== 1853 1854 _Description_ 1855 This function tests for completion of all asynchrounous operations. 1856 In C/C++, a non-zero value will be returned to indicate all 1857 asynchronous operations have completed. While Fortran will return 1858 a `true'. If any asynchronous operation has not completed, C/C++ 1859 returns a zero and Fortran returns a `false'. 1860 1861 _C/C++_: 1862 _Prototype_: `int acc_async_test_all(void);' 1863 1864 _Fortran_: 1865 _Interface_: `function acc_async_test()' 1866 `logical acc_get_device_num' 1867 1868 _Reference_: 1869 OpenACC specification v2.0 (http://www.openacc.org/), section 1870 3.2.7. 1871 1872 1873 File: libgomp.info, Node: acc_wait, Next: acc_wait_all, Prev: acc_async_test_all, Up: OpenACC Runtime Library Routines 1874 1875 5.8 `acc_wait' - Wait for completion of a specific asynchronous operation. 1876 ========================================================================== 1877 1878 _Description_ 1879 This function waits for completion of the asynchronous operation 1880 specified in ARG. 1881 1882 _C/C++_: 1883 _Prototype_: `acc_wait(arg);' 1884 1885 _Fortran_: 1886 _Interface_: `subroutine acc_wait(arg)' 1887 `integer(acc_handle_kind) arg' 1888 1889 _Reference_: 1890 OpenACC specification v2.0 (http://www.openacc.org/), section 1891 3.2.8. 1892 1893 1894 File: libgomp.info, Node: acc_wait_all, Next: acc_wait_all_async, Prev: acc_wait, Up: OpenACC Runtime Library Routines 1895 1896 5.9 `acc_wait_all' - Waits for completion of all asynchronous operations. 1897 ========================================================================= 1898 1899 _Description_ 1900 This function waits for the completion of all asynchronous 1901 operations. 1902 1903 _C/C++_: 1904 _Prototype_: `acc_wait_all(void);' 1905 1906 _Fortran_: 1907 _Interface_: `subroutine acc_wait_async()' 1908 1909 _Reference_: 1910 OpenACC specification v2.0 (http://www.openacc.org/), section 1911 3.2.10. 1912 1913 1914 File: libgomp.info, Node: acc_wait_all_async, Next: acc_wait_async, Prev: acc_wait_all, Up: OpenACC Runtime Library Routines 1915 1916 5.10 `acc_wait_all_async' - Wait for completion of all asynchronous operations. 1917 =============================================================================== 1918 1919 _Description_ 1920 This function enqueues a wait operation on the queue ASYNC for any 1921 and all asynchronous operations that have been previously enqueued 1922 on any queue. 1923 1924 _C/C++_: 1925 _Prototype_: `acc_wait_all_async(int async);' 1926 1927 _Fortran_: 1928 _Interface_: `subroutine acc_wait_all_async(async)' 1929 `integer(acc_handle_kind) async' 1930 1931 _Reference_: 1932 OpenACC specification v2.0 (http://www.openacc.org/), section 1933 3.2.11. 1934 1935 1936 File: libgomp.info, Node: acc_wait_async, Next: acc_init, Prev: acc_wait_all_async, Up: OpenACC Runtime Library Routines 1937 1938 5.11 `acc_wait_async' - Wait for completion of asynchronous operations. 1939 ======================================================================= 1940 1941 _Description_ 1942 This function enqueues a wait operation on queue ASYNC for any and 1943 all asynchronous operations enqueued on queue ARG. 1944 1945 _C/C++_: 1946 _Prototype_: `acc_wait_async(int arg, int async);' 1947 1948 _Fortran_: 1949 _Interface_: `subroutine acc_wait_async(arg, async)' 1950 `integer(acc_handle_kind) arg, async' 1951 1952 _Reference_: 1953 OpenACC specification v2.0 (http://www.openacc.org/), section 1954 3.2.9. 1955 1956 1957 File: libgomp.info, Node: acc_init, Next: acc_shutdown, Prev: acc_wait_async, Up: OpenACC Runtime Library Routines 1958 1959 5.12 `acc_init' - Initialize runtime for a specific device type. 1960 ================================================================ 1961 1962 _Description_ 1963 This function initializes the runtime for the device type 1964 specified in DEVICETYPE. 1965 1966 _C/C++_: 1967 _Prototype_: `acc_init(acc_device_t devicetype);' 1968 1969 _Fortran_: 1970 _Interface_: `subroutine acc_init(devicetype)' 1971 `integer(acc_device_kind) devicetype' 1972 1973 _Reference_: 1974 OpenACC specification v2.0 (http://www.openacc.org/), section 1975 3.2.12. 1976 1977 1978 File: libgomp.info, Node: acc_shutdown, Next: acc_on_device, Prev: acc_init, Up: OpenACC Runtime Library Routines 1979 1980 5.13 `acc_shutdown' - Shuts down the runtime for a specific device type. 1981 ======================================================================== 1982 1983 _Description_ 1984 This function shuts down the runtime for the device type specified 1985 in DEVICETYPE. 1986 1987 _C/C++_: 1988 _Prototype_: `acc_shutdown(acc_device_t devicetype);' 1989 1990 _Fortran_: 1991 _Interface_: `subroutine acc_shutdown(devicetype)' 1992 `integer(acc_device_kind) devicetype' 1993 1994 _Reference_: 1995 OpenACC specification v2.0 (http://www.openacc.org/), section 1996 3.2.13. 1997 1998 1999 File: libgomp.info, Node: acc_on_device, Next: acc_malloc, Prev: acc_shutdown, Up: OpenACC Runtime Library Routines 2000 2001 5.14 `acc_on_device' - Whether executing on a particular device 2002 =============================================================== 2003 2004 _Description_: 2005 This function returns whether the program is executing on a 2006 particular device specified in DEVICETYPE. In C/C++ a non-zero 2007 value is returned to indicate the device is execiting on the 2008 specified device type. In Fortran, `true' will be returned. If 2009 the program is not executing on the specified device type C/C++ 2010 will return a zero, while Fortran will return `false'. 2011 2012 _C/C++_: 2013 _Prototype_: `acc_on_device(acc_device_t devicetype);' 2014 2015 _Fortran_: 2016 _Interface_: `function acc_on_device(devicetype)' 2017 `integer(acc_device_kind) devicetype' 2018 `logical acc_on_device' 2019 2020 _Reference_: 2021 OpenACC specification v2.0 (http://www.openacc.org/), section 2022 3.2.14. 2023 2024 2025 File: libgomp.info, Node: acc_malloc, Next: acc_free, Prev: acc_on_device, Up: OpenACC Runtime Library Routines 2026 2027 5.15 `acc_malloc' - Allocate device memory. 2028 =========================================== 2029 2030 _Description_ 2031 This function allocates LEN bytes of device memory. It returns the 2032 device address of the allocated memory. 2033 2034 _C/C++_: 2035 _Prototype_: `d_void* acc_malloc(size_t len);' 2036 2037 _Reference_: 2038 OpenACC specification v2.0 (http://www.openacc.org/), section 2039 3.2.15. 2040 2041 2042 File: libgomp.info, Node: acc_free, Next: acc_copyin, Prev: acc_malloc, Up: OpenACC Runtime Library Routines 2043 2044 5.16 `acc_free' - Free device memory. 2045 ===================================== 2046 2047 _Description_ 2048 Free previously allocated device memory at the device address `a'. 2049 2050 _C/C++_: 2051 _Prototype_: `acc_free(d_void *a);' 2052 2053 _Reference_: 2054 OpenACC specification v2.0 (http://www.openacc.org/), section 2055 3.2.16. 2056 2057 2058 File: libgomp.info, Node: acc_copyin, Next: acc_present_or_copyin, Prev: acc_free, Up: OpenACC Runtime Library Routines 2059 2060 5.17 `acc_copyin' - Allocate device memory and copy host memory to it. 2061 ====================================================================== 2062 2063 _Description_ 2064 In C/C++, this function allocates LEN bytes of device memory and 2065 maps it to the specified host address in A. The device address of 2066 the newly allocated device memory is returned. 2067 2068 In Fortran, two (2) forms are supported. In the first form, A 2069 specifies a contiguous array section. The second form A specifies a 2070 variable or array element and LEN specifies the length in bytes. 2071 2072 _C/C++_: 2073 _Prototype_: `void *acc_copyin(h_void *a, size_t len);' 2074 2075 _Fortran_: 2076 _Interface_: `subroutine acc_copyin(a)' 2077 `type, dimension(:[,:]...) :: a' 2078 _Interface_: `subroutine acc_copyin(a, len)' 2079 `type, dimension(:[,:]...) :: a' 2080 `integer len' 2081 2082 _Reference_: 2083 OpenACC specification v2.0 (http://www.openacc.org/), section 2084 3.2.17. 2085 2086 2087 File: libgomp.info, Node: acc_present_or_copyin, Next: acc_create, Prev: acc_copyin, Up: OpenACC Runtime Library Routines 2088 2089 5.18 `acc_present_or_copyin' - If the data is not present on the device, allocate device memory and copy from host memory. 2090 ========================================================================================================================== 2091 2092 _Description_ 2093 This function tests if the host data specifed by A and of length 2094 LEN is present or not. If it is not present, then device memory 2095 will be allocated and the host memory copied. The device address of 2096 the newly allocated device memory is returned. 2097 2098 In Fortran, two (2) forms are supported. In the first form, A 2099 specifies a contiguous array section. The second form A specifies 2100 a variable or array element and LEN specifies the length in bytes. 2101 2102 _C/C++_: 2103 _Prototype_: `void *acc_present_or_copyin(h_void *a, size_t len);' 2104 _Prototype_: `void *acc_pcopyin(h_void *a, size_t len);' 2105 2106 _Fortran_: 2107 _Interface_: `subroutine acc_present_or_copyin(a)' 2108 `type, dimension(:[,:]...) :: a' 2109 _Interface_: `subroutine acc_present_or_copyin(a, len)' 2110 `type, dimension(:[,:]...) :: a' 2111 `integer len' 2112 _Interface_: `subroutine acc_pcopyin(a)' 2113 `type, dimension(:[,:]...) :: a' 2114 _Interface_: `subroutine acc_pcopyin(a, len)' 2115 `type, dimension(:[,:]...) :: a' 2116 `integer len' 2117 2118 _Reference_: 2119 OpenACC specification v2.0 (http://www.openacc.org/), section 2120 3.2.18. 2121 2122 2123 File: libgomp.info, Node: acc_create, Next: acc_present_or_create, Prev: acc_present_or_copyin, Up: OpenACC Runtime Library Routines 2124 2125 5.19 `acc_create' - Allocate device memory and map it to host memory. 2126 ===================================================================== 2127 2128 _Description_ 2129 This function allocates device memory and maps it to host memory 2130 specified by the host address A with a length of LEN bytes. In 2131 C/C++, the function returns the device address of the allocated 2132 device memory. 2133 2134 In Fortran, two (2) forms are supported. In the first form, A 2135 specifies a contiguous array section. The second form A specifies 2136 a variable or array element and LEN specifies the length in bytes. 2137 2138 _C/C++_: 2139 _Prototype_: `void *acc_create(h_void *a, size_t len);' 2140 2141 _Fortran_: 2142 _Interface_: `subroutine acc_create(a)' 2143 `type, dimension(:[,:]...) :: a' 2144 _Interface_: `subroutine acc_create(a, len)' 2145 `type, dimension(:[,:]...) :: a' 2146 `integer len' 2147 2148 _Reference_: 2149 OpenACC specification v2.0 (http://www.openacc.org/), section 2150 3.2.19. 2151 2152 2153 File: libgomp.info, Node: acc_present_or_create, Next: acc_copyout, Prev: acc_create, Up: OpenACC Runtime Library Routines 2154 2155 5.20 `acc_present_or_create' - If the data is not present on the device, allocate device memory and map it to host memory. 2156 ========================================================================================================================== 2157 2158 _Description_ 2159 This function tests if the host data specifed by A and of length 2160 LEN is present or not. If it is not present, then device memory 2161 will be allocated and mapped to host memory. In C/C++, the device 2162 address of the newly allocated device memory is returned. 2163 2164 In Fortran, two (2) forms are supported. In the first form, A 2165 specifies a contiguous array section. The second form A specifies 2166 a variable or array element and LEN specifies the length in bytes. 2167 2168 _C/C++_: 2169 _Prototype_: `void *acc_present_or_create(h_void *a, size_t len)' 2170 _Prototype_: `void *acc_pcreate(h_void *a, size_t len)' 2171 2172 _Fortran_: 2173 _Interface_: `subroutine acc_present_or_create(a)' 2174 `type, dimension(:[,:]...) :: a' 2175 _Interface_: `subroutine acc_present_or_create(a, len)' 2176 `type, dimension(:[,:]...) :: a' 2177 `integer len' 2178 _Interface_: `subroutine acc_pcreate(a)' 2179 `type, dimension(:[,:]...) :: a' 2180 _Interface_: `subroutine acc_pcreate(a, len)' 2181 `type, dimension(:[,:]...) :: a' 2182 `integer len' 2183 2184 _Reference_: 2185 OpenACC specification v2.0 (http://www.openacc.org/), section 2186 3.2.20. 2187 2188 2189 File: libgomp.info, Node: acc_copyout, Next: acc_delete, Prev: acc_present_or_create, Up: OpenACC Runtime Library Routines 2190 2191 5.21 `acc_copyout' - Copy device memory to host memory. 2192 ======================================================= 2193 2194 _Description_ 2195 This function copies mapped device memory to host memory which is 2196 specified by host address A for a length LEN bytes in C/C++. 2197 2198 In Fortran, two (2) forms are supported. In the first form, A 2199 specifies a contiguous array section. The second form A specifies 2200 a variable or array element and LEN specifies the length in bytes. 2201 2202 _C/C++_: 2203 _Prototype_: `acc_copyout(h_void *a, size_t len);' 2204 2205 _Fortran_: 2206 _Interface_: `subroutine acc_copyout(a)' 2207 `type, dimension(:[,:]...) :: a' 2208 _Interface_: `subroutine acc_copyout(a, len)' 2209 `type, dimension(:[,:]...) :: a' 2210 `integer len' 2211 2212 _Reference_: 2213 OpenACC specification v2.0 (http://www.openacc.org/), section 2214 3.2.21. 2215 2216 2217 File: libgomp.info, Node: acc_delete, Next: acc_update_device, Prev: acc_copyout, Up: OpenACC Runtime Library Routines 2218 2219 5.22 `acc_delete' - Free device memory. 2220 ======================================= 2221 2222 _Description_ 2223 This function frees previously allocated device memory specified by 2224 the device address A and the length of LEN bytes. 2225 2226 In Fortran, two (2) forms are supported. In the first form, A 2227 specifies a contiguous array section. The second form A specifies 2228 a variable or array element and LEN specifies the length in bytes. 2229 2230 _C/C++_: 2231 _Prototype_: `acc_delete(h_void *a, size_t len);' 2232 2233 _Fortran_: 2234 _Interface_: `subroutine acc_delete(a)' 2235 `type, dimension(:[,:]...) :: a' 2236 _Interface_: `subroutine acc_delete(a, len)' 2237 `type, dimension(:[,:]...) :: a' 2238 `integer len' 2239 2240 _Reference_: 2241 OpenACC specification v2.0 (http://www.openacc.org/), section 2242 3.2.22. 2243 2244 2245 File: libgomp.info, Node: acc_update_device, Next: acc_update_self, Prev: acc_delete, Up: OpenACC Runtime Library Routines 2246 2247 5.23 `acc_update_device' - Update device memory from mapped host memory. 2248 ======================================================================== 2249 2250 _Description_ 2251 This function updates the device copy from the previously mapped 2252 host memory. The host memory is specified with the host address A 2253 and a length of LEN bytes. 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 _C/C++_: 2260 _Prototype_: `acc_update_device(h_void *a, size_t len);' 2261 2262 _Fortran_: 2263 _Interface_: `subroutine acc_update_device(a)' 2264 `type, dimension(:[,:]...) :: a' 2265 _Interface_: `subroutine acc_update_device(a, len)' 2266 `type, dimension(:[,:]...) :: a' 2267 `integer len' 2268 2269 _Reference_: 2270 OpenACC specification v2.0 (http://www.openacc.org/), section 2271 3.2.23. 2272 2273 2274 File: libgomp.info, Node: acc_update_self, Next: acc_map_data, Prev: acc_update_device, Up: OpenACC Runtime Library Routines 2275 2276 5.24 `acc_update_self' - Update host memory from mapped device memory. 2277 ====================================================================== 2278 2279 _Description_ 2280 This function updates the host copy from the previously mapped 2281 device memory. The host memory is specified with the host address 2282 A and a length of LEN bytes. 2283 2284 In Fortran, two (2) forms are supported. In the first form, A 2285 specifies a contiguous array section. The second form A specifies 2286 a variable or array element and LEN specifies the length in bytes. 2287 2288 _C/C++_: 2289 _Prototype_: `acc_update_self(h_void *a, size_t len);' 2290 2291 _Fortran_: 2292 _Interface_: `subroutine acc_update_self(a)' 2293 `type, dimension(:[,:]...) :: a' 2294 _Interface_: `subroutine acc_update_self(a, len)' 2295 `type, dimension(:[,:]...) :: a' 2296 `integer len' 2297 2298 _Reference_: 2299 OpenACC specification v2.0 (http://www.openacc.org/), section 2300 3.2.24. 2301 2302 2303 File: libgomp.info, Node: acc_map_data, Next: acc_unmap_data, Prev: acc_update_self, Up: OpenACC Runtime Library Routines 2304 2305 5.25 `acc_map_data' - Map previously allocated device memory to host memory. 2306 ============================================================================ 2307 2308 _Description_ 2309 This function maps previously allocated device and host memory. 2310 The device memory is specified with the device address D. The host 2311 memory is specified with the host address H and a length of LEN. 2312 2313 _C/C++_: 2314 _Prototype_: `acc_map_data(h_void *h, d_void *d, size_t len);' 2315 2316 _Reference_: 2317 OpenACC specification v2.0 (http://www.openacc.org/), section 2318 3.2.25. 2319 2320 2321 File: libgomp.info, Node: acc_unmap_data, Next: acc_deviceptr, Prev: acc_map_data, Up: OpenACC Runtime Library Routines 2322 2323 5.26 `acc_unmap_data' - Unmap device memory from host memory. 2324 ============================================================= 2325 2326 _Description_ 2327 This function unmaps previously mapped device and host memory. The 2328 latter specified by H. 2329 2330 _C/C++_: 2331 _Prototype_: `acc_unmap_data(h_void *h);' 2332 2333 _Reference_: 2334 OpenACC specification v2.0 (http://www.openacc.org/), section 2335 3.2.26. 2336 2337 2338 File: libgomp.info, Node: acc_deviceptr, Next: acc_hostptr, Prev: acc_unmap_data, Up: OpenACC Runtime Library Routines 2339 2340 5.27 `acc_deviceptr' - Get device pointer associated with specific host address. 2341 ================================================================================ 2342 2343 _Description_ 2344 This function returns the device address that has been mapped to 2345 the host address specified by H. 2346 2347 _C/C++_: 2348 _Prototype_: `void *acc_deviceptr(h_void *h);' 2349 2350 _Reference_: 2351 OpenACC specification v2.0 (http://www.openacc.org/), section 2352 3.2.27. 2353 2354 2355 File: libgomp.info, Node: acc_hostptr, Next: acc_is_present, Prev: acc_deviceptr, Up: OpenACC Runtime Library Routines 2356 2357 5.28 `acc_hostptr' - Get host pointer associated with specific device address. 2358 ============================================================================== 2359 2360 _Description_ 2361 This function returns the host address that has been mapped to the 2362 device address specified by D. 2363 2364 _C/C++_: 2365 _Prototype_: `void *acc_hostptr(d_void *d);' 2366 2367 _Reference_: 2368 OpenACC specification v2.0 (http://www.openacc.org/), section 2369 3.2.28. 2370 2371 2372 File: libgomp.info, Node: acc_is_present, Next: acc_memcpy_to_device, Prev: acc_hostptr, Up: OpenACC Runtime Library Routines 2373 2374 5.29 `acc_is_present' - Indicate whether host variable / array is present on device. 2375 ==================================================================================== 2376 2377 _Description_ 2378 This function indicates whether the specified host address in A 2379 and a length of LEN bytes is present on the device. In C/C++, a 2380 non-zero value is returned to indicate the presence of the mapped 2381 memory on the device. A zero is returned to indicate the memory is 2382 not mapped on the device. 2383 2384 In Fortran, two (2) forms are supported. In the first form, A 2385 specifies a contiguous array section. The second form A specifies 2386 a variable or array element and LEN specifies the length in bytes. 2387 If the host memory is mapped to device memory, then a `true' is 2388 returned. Otherwise, a `false' is return to indicate the mapped 2389 memory is not present. 2390 2391 _C/C++_: 2392 _Prototype_: `int acc_is_present(h_void *a, size_t len);' 2393 2394 _Fortran_: 2395 _Interface_: `function acc_is_present(a)' 2396 `type, dimension(:[,:]...) :: a' 2397 `logical acc_is_present' 2398 _Interface_: `function acc_is_present(a, len)' 2399 `type, dimension(:[,:]...) :: a' 2400 `integer len' 2401 `logical acc_is_present' 2402 2403 _Reference_: 2404 OpenACC specification v2.0 (http://www.openacc.org/), section 2405 3.2.29. 2406 2407 2408 File: libgomp.info, Node: acc_memcpy_to_device, Next: acc_memcpy_from_device, Prev: acc_is_present, Up: OpenACC Runtime Library Routines 2409 2410 5.30 `acc_memcpy_to_device' - Copy host memory to device memory. 2411 ================================================================ 2412 2413 _Description_ 2414 This function copies host memory specified by host address of SRC 2415 to device memory specified by the device address DEST for a length 2416 of BYTES bytes. 2417 2418 _C/C++_: 2419 _Prototype_: `acc_memcpy_to_device(d_void *dest, h_void *src, 2420 size_t bytes);' 2421 2422 _Reference_: 2423 OpenACC specification v2.0 (http://www.openacc.org/), section 2424 3.2.30. 2425 2426 2427 File: libgomp.info, Node: acc_memcpy_from_device, Next: acc_get_current_cuda_device, Prev: acc_memcpy_to_device, Up: OpenACC Runtime Library Routines 2428 2429 5.31 `acc_memcpy_from_device' - Copy device memory to host memory. 2430 ================================================================== 2431 2432 _Description_ 2433 This function copies host memory specified by host address of SRC 2434 from device memory specified by the device address DEST for a 2435 length of BYTES bytes. 2436 2437 _C/C++_: 2438 _Prototype_: `acc_memcpy_from_device(d_void *dest, h_void *src, 2439 size_t bytes);' 2440 2441 _Reference_: 2442 OpenACC specification v2.0 (http://www.openacc.org/), section 2443 3.2.31. 2444 2445 2446 File: libgomp.info, Node: acc_get_current_cuda_device, Next: acc_get_current_cuda_context, Prev: acc_memcpy_from_device, Up: OpenACC Runtime Library Routines 2447 2448 5.32 `acc_get_current_cuda_device' - Get CUDA device handle. 2449 ============================================================ 2450 2451 _Description_ 2452 This function returns the CUDA device handle. This handle is the 2453 same as used by the CUDA Runtime or Driver API's. 2454 2455 _C/C++_: 2456 _Prototype_: `void *acc_get_current_cuda_device(void);' 2457 2458 _Reference_: 2459 OpenACC specification v2.0 (http://www.openacc.org/), section 2460 A.2.1.1. 2461 2462 2463 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 2464 2465 5.33 `acc_get_current_cuda_context' - Get CUDA context handle. 2466 ============================================================== 2467 2468 _Description_ 2469 This function returns the CUDA context handle. This handle is the 2470 same as used by the CUDA Runtime or Driver API's. 2471 2472 _C/C++_: 2473 _Prototype_: `acc_get_current_cuda_context(void);' 2474 2475 _Reference_: 2476 OpenACC specification v2.0 (http://www.openacc.org/), section 2477 A.2.1.2. 2478 2479 2480 File: libgomp.info, Node: acc_get_cuda_stream, Next: acc_set_cuda_stream, Prev: acc_get_current_cuda_context, Up: OpenACC Runtime Library Routines 2481 2482 5.34 `acc_get_cuda_stream' - Get CUDA stream handle. 2483 ==================================================== 2484 2485 _Description_ 2486 This function returns the CUDA stream handle. This handle is the 2487 same as used by the CUDA Runtime or Driver API's. 2488 2489 _C/C++_: 2490 _Prototype_: `acc_get_cuda_stream(void);' 2491 2492 _Reference_: 2493 OpenACC specification v2.0 (http://www.openacc.org/), section 2494 A.2.1.3. 2495 2496 2497 File: libgomp.info, Node: acc_set_cuda_stream, Prev: acc_get_cuda_stream, Up: OpenACC Runtime Library Routines 2498 2499 5.35 `acc_set_cuda_stream' - Set CUDA stream handle. 2500 ==================================================== 2501 2502 _Description_ 2503 This function associates the stream handle specified by STREAM with 2504 the asynchronous value specified by ASYNC. 2505 2506 _C/C++_: 2507 _Prototype_: `acc_set_cuda_stream(int async void *stream);' 2508 2509 _Reference_: 2510 OpenACC specification v2.0 (http://www.openacc.org/), section 2511 A.2.1.4. 2512 2513 2514 File: libgomp.info, Node: OpenACC Environment Variables, Next: CUDA Streams Usage, Prev: OpenACC Runtime Library Routines, Up: Top 2515 2516 6 OpenACC Environment Variables 2517 ******************************* 2518 2519 The variables `ACC_DEVICE_TYPE' and `ACC_DEVICE_NUM' are defined by 2520 section 4 of the OpenACC specification in version 2.0. The variable 2521 `GCC_ACC_NOTIFY' is used for diagnostic purposes. 2522 2523 * Menu: 2524 2525 * ACC_DEVICE_TYPE:: 2526 * ACC_DEVICE_NUM:: 2527 * GCC_ACC_NOTIFY:: 2528 2529 2530 File: libgomp.info, Node: ACC_DEVICE_TYPE, Next: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2531 2532 6.1 `ACC_DEVICE_TYPE' 2533 ===================== 2534 2535 _Reference_: 2536 OpenACC specification v2.0 (http://www.openacc.org/), section 4.1. 2537 2538 2539 File: libgomp.info, Node: ACC_DEVICE_NUM, Next: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_TYPE, Up: OpenACC Environment Variables 2540 2541 6.2 `ACC_DEVICE_NUM' 2542 ==================== 2543 2544 _Reference_: 2545 OpenACC specification v2.0 (http://www.openacc.org/), section 4.2. 2546 2547 2548 File: libgomp.info, Node: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2549 2550 6.3 `GCC_ACC_NOTIFY' 2551 ==================== 2552 2553 _Description_: 2554 Print debug information pertaining to the accelerator. 2555 2556 2557 File: libgomp.info, Node: CUDA Streams Usage, Next: OpenACC Library Interoperability, Prev: OpenACC Environment Variables, Up: Top 2558 2559 7 CUDA Streams Usage 2560 ******************** 2561 2562 This applies to the `nvptx' plugin only. 2563 2564 The library provides elements that perform asynchronous movement of 2565 data and asynchronous operation of computing constructs. This 2566 asynchronous functionality is implemented by making use of CUDA 2567 streams(1). 2568 2569 The primary means by that the asychronous functionality is accessed 2570 is through the use of those OpenACC directives which make use of the 2571 `async' and `wait' clauses. When the `async' clause is first used with 2572 a directive, it creates a CUDA stream. If an `async-argument' is used 2573 with the `async' clause, then the stream is associated with the 2574 specified `async-argument'. 2575 2576 Following the creation of an association between a CUDA stream and 2577 the `async-argument' of an `async' clause, both the `wait' clause and 2578 the `wait' directive can be used. When either the clause or directive 2579 is used after stream creation, it creates a rendezvous point whereby 2580 execution waits until all operations associated with the 2581 `async-argument', that is, stream, have completed. 2582 2583 Normally, the management of the streams that are created as a result 2584 of using the `async' clause, is done without any intervention by the 2585 caller. This implies the association between the `async-argument' and 2586 the CUDA stream will be maintained for the lifetime of the program. 2587 However, this association can be changed through the use of the library 2588 function `acc_set_cuda_stream'. When the function 2589 `acc_set_cuda_stream' is called, the CUDA stream that was originally 2590 associated with the `async' clause will be destroyed. Caution should 2591 be taken when changing the association as subsequent references to the 2592 `async-argument' refer to a different CUDA stream. 2593 2594 ---------- Footnotes ---------- 2595 2596 (1) See "Stream Management" in "CUDA Driver API", TRM-06703-001, 2597 Version 5.5, for additional information 2598 2599 2600 File: libgomp.info, Node: OpenACC Library Interoperability, Next: The libgomp ABI, Prev: CUDA Streams Usage, Up: Top 2601 2602 8 OpenACC Library Interoperability 2603 ********************************** 2604 2605 8.1 Introduction 2606 ================ 2607 2608 The OpenACC library uses the CUDA Driver API, and may interact with 2609 programs that use the Runtime library directly, or another library 2610 based on the Runtime library, e.g., CUBLAS(1). This chapter describes 2611 the use cases and what changes are required in order to use both the 2612 OpenACC library and the CUBLAS and Runtime libraries within a program. 2613 2614 8.2 First invocation: NVIDIA CUBLAS library API 2615 =============================================== 2616 2617 In this first use case (see below), a function in the CUBLAS library is 2618 called prior to any of the functions in the OpenACC library. More 2619 specifically, the function `cublasCreate()'. 2620 2621 When invoked, the function initializes the library and allocates the 2622 hardware resources on the host and the device on behalf of the caller. 2623 Once the initialization and allocation has completed, a handle is 2624 returned to the caller. The OpenACC library also requires 2625 initialization and allocation of hardware resources. Since the CUBLAS 2626 library has already allocated the hardware resources for the device, 2627 all that is left to do is to initialize the OpenACC library and acquire 2628 the hardware resources on the host. 2629 2630 Prior to calling the OpenACC function that initializes the library 2631 and allocate the host hardware resources, you need to acquire the 2632 device number that was allocated during the call to `cublasCreate()'. 2633 The invoking of the runtime library function `cudaGetDevice()' 2634 accomplishes this. Once acquired, the device number is passed along 2635 with the device type as parameters to the OpenACC library function 2636 `acc_set_device_num()'. 2637 2638 Once the call to `acc_set_device_num()' has completed, the OpenACC 2639 library uses the context that was created during the call to 2640 `cublasCreate()'. In other words, both libraries will be sharing the 2641 same context. 2642 2643 /* Create the handle */ 2644 s = cublasCreate(&h); 2645 if (s != CUBLAS_STATUS_SUCCESS) 2646 { 2647 fprintf(stderr, "cublasCreate failed %d\n", s); 2648 exit(EXIT_FAILURE); 2649 } 2650 2651 /* Get the device number */ 2652 e = cudaGetDevice(&dev); 2653 if (e != cudaSuccess) 2654 { 2655 fprintf(stderr, "cudaGetDevice failed %d\n", e); 2656 exit(EXIT_FAILURE); 2657 } 2658 2659 /* Initialize OpenACC library and use device 'dev' */ 2660 acc_set_device_num(dev, acc_device_nvidia); 2661 Use Case 1 2662 2663 8.3 First invocation: OpenACC library API 2664 ========================================= 2665 2666 In this second use case (see below), a function in the OpenACC library 2667 is called prior to any of the functions in the CUBLAS library. More 2668 specificially, the function `acc_set_device_num()'. 2669 2670 In the use case presented here, the function `acc_set_device_num()' 2671 is used to both initialize the OpenACC library and allocate the hardware 2672 resources on the host and the device. In the call to the function, the 2673 call parameters specify which device to use and what device type to 2674 use, i.e., `acc_device_nvidia'. It should be noted that this is but one 2675 method to initialize the OpenACC library and allocate the appropriate 2676 hardware resources. Other methods are available through the use of 2677 environment variables and these will be discussed in the next section. 2678 2679 Once the call to `acc_set_device_num()' has completed, other OpenACC 2680 functions can be called as seen with multiple calls being made to 2681 `acc_copyin()'. In addition, calls can be made to functions in the 2682 CUBLAS library. In the use case a call to `cublasCreate()' is made 2683 subsequent to the calls to `acc_copyin()'. As seen in the previous use 2684 case, a call to `cublasCreate()' initializes the CUBLAS library and 2685 allocates the hardware resources on the host and the device. However, 2686 since the device has already been allocated, `cublasCreate()' will only 2687 initialize the CUBLAS library and allocate the appropriate hardware 2688 resources on the host. The context that was created as part of the 2689 OpenACC initialization is shared with the CUBLAS library, similarly to 2690 the first use case. 2691 2692 dev = 0; 2693 2694 acc_set_device_num(dev, acc_device_nvidia); 2695 2696 /* Copy the first set to the device */ 2697 d_X = acc_copyin(&h_X[0], N * sizeof (float)); 2698 if (d_X == NULL) 2699 { 2700 fprintf(stderr, "copyin error h_X\n"); 2701 exit(EXIT_FAILURE); 2702 } 2703 2704 /* Copy the second set to the device */ 2705 d_Y = acc_copyin(&h_Y1[0], N * sizeof (float)); 2706 if (d_Y == NULL) 2707 { 2708 fprintf(stderr, "copyin error h_Y1\n"); 2709 exit(EXIT_FAILURE); 2710 } 2711 2712 /* Create the handle */ 2713 s = cublasCreate(&h); 2714 if (s != CUBLAS_STATUS_SUCCESS) 2715 { 2716 fprintf(stderr, "cublasCreate failed %d\n", s); 2717 exit(EXIT_FAILURE); 2718 } 2719 2720 /* Perform saxpy using CUBLAS library function */ 2721 s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1); 2722 if (s != CUBLAS_STATUS_SUCCESS) 2723 { 2724 fprintf(stderr, "cublasSaxpy failed %d\n", s); 2725 exit(EXIT_FAILURE); 2726 } 2727 2728 /* Copy the results from the device */ 2729 acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float)); 2730 Use Case 2 2731 2732 8.4 OpenACC library and environment variables 2733 ============================================= 2734 2735 There are two environment variables associated with the OpenACC library 2736 that may be used to control the device type and device number: 2737 `ACC_DEVICE_TYPE' and `ACC_DEVICE_NUM', respecively. These two 2738 environement variables can be used as an alternative to calling 2739 `acc_set_device_num()'. As seen in the second use case, the device type 2740 and device number were specified using `acc_set_device_num()'. If 2741 however, the aforementioned environment variables were set, then the 2742 call to `acc_set_device_num()' would not be required. 2743 2744 The use of the environment variables is only relevant when an 2745 OpenACC function is called prior to a call to `cudaCreate()'. If 2746 `cudaCreate()' is called prior to a call to an OpenACC function, then 2747 you must call `acc_set_device_num()'(2) 2748 2749 ---------- Footnotes ---------- 2750 2751 (1) See section 2.26, "Interactions with the CUDA Driver API" in 2752 "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU 2753 Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5, 2754 for additional information on library interoperability. 2755 2756 (2) More complete information about `ACC_DEVICE_TYPE' and 2757 `ACC_DEVICE_NUM' can be found in sections 4.1 and 4.2 of the OpenACC 2758 (http://www.openacc.org/) Application Programming Interface, Version 2759 2.0. 2760 2761 2762 File: libgomp.info, Node: The libgomp ABI, Next: Reporting Bugs, Prev: OpenACC Library Interoperability, Up: Top 2763 2764 9 The libgomp ABI 2765 ***************** 2766 2767 The following sections present notes on the external ABI as presented 2768 by libgomp. Only maintainers should need them. 2769 2770 * Menu: 2771 2772 * Implementing MASTER construct:: 2773 * Implementing CRITICAL construct:: 2774 * Implementing ATOMIC construct:: 2775 * Implementing FLUSH construct:: 2776 * Implementing BARRIER construct:: 2777 * Implementing THREADPRIVATE construct:: 2778 * Implementing PRIVATE clause:: 2779 * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses:: 2780 * Implementing REDUCTION clause:: 2781 * Implementing PARALLEL construct:: 2782 * Implementing FOR construct:: 2783 * Implementing ORDERED construct:: 2784 * Implementing SECTIONS construct:: 2785 * Implementing SINGLE construct:: 2786 * Implementing OpenACC's PARALLEL construct:: 2787 2788 2789 File: libgomp.info, Node: Implementing MASTER construct, Next: Implementing CRITICAL construct, Up: The libgomp ABI 2790 2791 9.1 Implementing MASTER construct 2792 ================================= 2793 2794 if (omp_get_thread_num () == 0) 2795 block 2796 2797 Alternately, we generate two copies of the parallel subfunction and 2798 only include this in the version run by the master thread. Surely this 2799 is not worthwhile though... 2800 2801 2802 File: libgomp.info, Node: Implementing CRITICAL construct, Next: Implementing ATOMIC construct, Prev: Implementing MASTER construct, Up: The libgomp ABI 2803 2804 9.2 Implementing CRITICAL construct 2805 =================================== 2806 2807 Without a specified name, 2808 2809 void GOMP_critical_start (void); 2810 void GOMP_critical_end (void); 2811 2812 so that we don't get COPY relocations from libgomp to the main 2813 application. 2814 2815 With a specified name, use omp_set_lock and omp_unset_lock with name 2816 being transformed into a variable declared like 2817 2818 omp_lock_t gomp_critical_user_<name> __attribute__((common)) 2819 2820 Ideally the ABI would specify that all zero is a valid unlocked 2821 state, and so we wouldn't need to initialize this at startup. 2822 2823 2824 File: libgomp.info, Node: Implementing ATOMIC construct, Next: Implementing FLUSH construct, Prev: Implementing CRITICAL construct, Up: The libgomp ABI 2825 2826 9.3 Implementing ATOMIC construct 2827 ================================= 2828 2829 The target should implement the `__sync' builtins. 2830 2831 Failing that we could add 2832 2833 void GOMP_atomic_enter (void) 2834 void GOMP_atomic_exit (void) 2835 2836 which reuses the regular lock code, but with yet another lock object 2837 private to the library. 2838 2839 2840 File: libgomp.info, Node: Implementing FLUSH construct, Next: Implementing BARRIER construct, Prev: Implementing ATOMIC construct, Up: The libgomp ABI 2841 2842 9.4 Implementing FLUSH construct 2843 ================================ 2844 2845 Expands to the `__sync_synchronize' builtin. 2846 2847 2848 File: libgomp.info, Node: Implementing BARRIER construct, Next: Implementing THREADPRIVATE construct, Prev: Implementing FLUSH construct, Up: The libgomp ABI 2849 2850 9.5 Implementing BARRIER construct 2851 ================================== 2852 2853 void GOMP_barrier (void) 2854 2855 2856 File: libgomp.info, Node: Implementing THREADPRIVATE construct, Next: Implementing PRIVATE clause, Prev: Implementing BARRIER construct, Up: The libgomp ABI 2857 2858 9.6 Implementing THREADPRIVATE construct 2859 ======================================== 2860 2861 In _most_ cases we can map this directly to `__thread'. Except that 2862 OMP allows constructors for C++ objects. We can either refuse to 2863 support this (how often is it used?) or we can implement something akin 2864 to .ctors. 2865 2866 Even more ideally, this ctor feature is handled by extensions to the 2867 main pthreads library. Failing that, we can have a set of entry points 2868 to register ctor functions to be called. 2869 2870 2871 File: libgomp.info, Node: Implementing PRIVATE clause, Next: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Prev: Implementing THREADPRIVATE construct, Up: The libgomp ABI 2872 2873 9.7 Implementing PRIVATE clause 2874 =============================== 2875 2876 In association with a PARALLEL, or within the lexical extent of a 2877 PARALLEL block, the variable becomes a local variable in the parallel 2878 subfunction. 2879 2880 In association with FOR or SECTIONS blocks, create a new automatic 2881 variable within the current function. This preserves the semantic of 2882 new variable creation. 2883 2884 2885 File: libgomp.info, Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Next: Implementing REDUCTION clause, Prev: Implementing PRIVATE clause, Up: The libgomp ABI 2886 2887 9.8 Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses 2888 ======================================================================== 2889 2890 This seems simple enough for PARALLEL blocks. Create a private struct 2891 for communicating between the parent and subfunction. In the parent, 2892 copy in values for scalar and "small" structs; copy in addresses for 2893 others TREE_ADDRESSABLE types. In the subfunction, copy the value into 2894 the local variable. 2895 2896 It is not clear what to do with bare FOR or SECTION blocks. The 2897 only thing I can figure is that we do something like: 2898 2899 #pragma omp for firstprivate(x) lastprivate(y) 2900 for (int i = 0; i < n; ++i) 2901 body; 2902 2903 which becomes 2904 2905 { 2906 int x = x, y; 2907 2908 // for stuff 2909 2910 if (i == n) 2911 y = y; 2912 } 2913 2914 where the "x=x" and "y=y" assignments actually have different uids 2915 for the two variables, i.e. not something you could write directly in 2916 C. Presumably this only makes sense if the "outer" x and y are global 2917 variables. 2918 2919 COPYPRIVATE would work the same way, except the structure broadcast 2920 would have to happen via SINGLE machinery instead. 2921 2922 2923 File: libgomp.info, Node: Implementing REDUCTION clause, Next: Implementing PARALLEL construct, Prev: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Up: The libgomp ABI 2924 2925 9.9 Implementing REDUCTION clause 2926 ================================= 2927 2928 The private struct mentioned in the previous section should have a 2929 pointer to an array of the type of the variable, indexed by the 2930 thread's TEAM_ID. The thread stores its final value into the array, 2931 and after the barrier, the master thread iterates over the array to 2932 collect the values. 2933 2934 2935 File: libgomp.info, Node: Implementing PARALLEL construct, Next: Implementing FOR construct, Prev: Implementing REDUCTION clause, Up: The libgomp ABI 2936 2937 9.10 Implementing PARALLEL construct 2938 ==================================== 2939 2940 #pragma omp parallel 2941 { 2942 body; 2943 } 2944 2945 becomes 2946 2947 void subfunction (void *data) 2948 { 2949 use data; 2950 body; 2951 } 2952 2953 setup data; 2954 GOMP_parallel_start (subfunction, &data, num_threads); 2955 subfunction (&data); 2956 GOMP_parallel_end (); 2957 2958 void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads) 2959 2960 The FN argument is the subfunction to be run in parallel. 2961 2962 The DATA argument is a pointer to a structure used to communicate 2963 data in and out of the subfunction, as discussed above with respect to 2964 FIRSTPRIVATE et al. 2965 2966 The NUM_THREADS argument is 1 if an IF clause is present and false, 2967 or the value of the NUM_THREADS clause, if present, or 0. 2968 2969 The function needs to create the appropriate number of threads 2970 and/or launch them from the dock. It needs to create the team 2971 structure and assign team ids. 2972 2973 void GOMP_parallel_end (void) 2974 2975 Tears down the team and returns us to the previous 2976 `omp_in_parallel()' state. 2977 2978 2979 File: libgomp.info, Node: Implementing FOR construct, Next: Implementing ORDERED construct, Prev: Implementing PARALLEL construct, Up: The libgomp ABI 2980 2981 9.11 Implementing FOR construct 2982 =============================== 2983 2984 #pragma omp parallel for 2985 for (i = lb; i <= ub; i++) 2986 body; 2987 2988 becomes 2989 2990 void subfunction (void *data) 2991 { 2992 long _s0, _e0; 2993 while (GOMP_loop_static_next (&_s0, &_e0)) 2994 { 2995 long _e1 = _e0, i; 2996 for (i = _s0; i < _e1; i++) 2997 body; 2998 } 2999 GOMP_loop_end_nowait (); 3000 } 3001 3002 GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0); 3003 subfunction (NULL); 3004 GOMP_parallel_end (); 3005 3006 #pragma omp for schedule(runtime) 3007 for (i = 0; i < n; i++) 3008 body; 3009 3010 becomes 3011 3012 { 3013 long i, _s0, _e0; 3014 if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0)) 3015 do { 3016 long _e1 = _e0; 3017 for (i = _s0, i < _e0; i++) 3018 body; 3019 } while (GOMP_loop_runtime_next (&_s0, _&e0)); 3020 GOMP_loop_end (); 3021 } 3022 3023 Note that while it looks like there is trickiness to propagating a 3024 non-constant STEP, there isn't really. We're explicitly allowed to 3025 evaluate it as many times as we want, and any variables involved should 3026 automatically be handled as PRIVATE or SHARED like any other variables. 3027 So the expression should remain evaluable in the subfunction. We can 3028 also pull it into a local variable if we like, but since its supposed 3029 to remain unchanged, we can also not if we like. 3030 3031 If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be 3032 able to get away with no work-sharing context at all, since we can 3033 simply perform the arithmetic directly in each thread to divide up the 3034 iterations. Which would mean that we wouldn't need to call any of 3035 these routines. 3036 3037 There are separate routines for handling loops with an ORDERED 3038 clause. Bookkeeping for that is non-trivial... 3039 3040 3041 File: libgomp.info, Node: Implementing ORDERED construct, Next: Implementing SECTIONS construct, Prev: Implementing FOR construct, Up: The libgomp ABI 3042 3043 9.12 Implementing ORDERED construct 3044 =================================== 3045 3046 void GOMP_ordered_start (void) 3047 void GOMP_ordered_end (void) 3048 3049 3050 File: libgomp.info, Node: Implementing SECTIONS construct, Next: Implementing SINGLE construct, Prev: Implementing ORDERED construct, Up: The libgomp ABI 3051 3052 9.13 Implementing SECTIONS construct 3053 ==================================== 3054 3055 A block as 3056 3057 #pragma omp sections 3058 { 3059 #pragma omp section 3060 stmt1; 3061 #pragma omp section 3062 stmt2; 3063 #pragma omp section 3064 stmt3; 3065 } 3066 3067 becomes 3068 3069 for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ()) 3070 switch (i) 3071 { 3072 case 1: 3073 stmt1; 3074 break; 3075 case 2: 3076 stmt2; 3077 break; 3078 case 3: 3079 stmt3; 3080 break; 3081 } 3082 GOMP_barrier (); 3083 3084 3085 File: libgomp.info, Node: Implementing SINGLE construct, Next: Implementing OpenACC's PARALLEL construct, Prev: Implementing SECTIONS construct, Up: The libgomp ABI 3086 3087 9.14 Implementing SINGLE construct 3088 ================================== 3089 3090 A block like 3091 3092 #pragma omp single 3093 { 3094 body; 3095 } 3096 3097 becomes 3098 3099 if (GOMP_single_start ()) 3100 body; 3101 GOMP_barrier (); 3102 3103 while 3104 3105 #pragma omp single copyprivate(x) 3106 body; 3107 3108 becomes 3109 3110 datap = GOMP_single_copy_start (); 3111 if (datap == NULL) 3112 { 3113 body; 3114 data.x = x; 3115 GOMP_single_copy_end (&data); 3116 } 3117 else 3118 x = datap->x; 3119 GOMP_barrier (); 3120 3121 3122 File: libgomp.info, Node: Implementing OpenACC's PARALLEL construct, Prev: Implementing SINGLE construct, Up: The libgomp ABI 3123 3124 9.15 Implementing OpenACC's PARALLEL construct 3125 ============================================== 3126 3127 void GOACC_parallel () 3128 3129 3130 File: libgomp.info, Node: Reporting Bugs, Next: Copying, Prev: The libgomp ABI, Up: Top 3131 3132 10 Reporting Bugs 3133 ***************** 3134 3135 Bugs in the GNU Offloading and Multi Processing Runtime Library should 3136 be reported via Bugzilla (http://gcc.gnu.org/bugzilla/). Please add 3137 "openacc", or "openmp", or both to the keywords field in the bug 3138 report, as appropriate. 3139 3140 3141 File: libgomp.info, Node: Copying, Next: GNU Free Documentation License, Prev: Reporting Bugs, Up: Top 3142 3143 GNU General Public License 3144 ************************** 3145 3146 Version 3, 29 June 2007 3147 3148 Copyright (C) 2007 Free Software Foundation, Inc. `http://fsf.org/' 3149 3150 Everyone is permitted to copy and distribute verbatim copies of this 3151 license document, but changing it is not allowed. 3152 3153 Preamble 3154 ======== 3155 3156 The GNU General Public License is a free, copyleft license for software 3157 and other kinds of works. 3158 3159 The licenses for most software and other practical works are designed 3160 to take away your freedom to share and change the works. By contrast, 3161 the GNU General Public License is intended to guarantee your freedom to 3162 share and change all versions of a program-to make sure it remains free 3163 software for all its users. We, the Free Software Foundation, use the 3164 GNU General Public License for most of our software; it applies also to 3165 any other work released this way by its authors. You can apply it to 3166 your programs, too. 3167 3168 When we speak of free software, we are referring to freedom, not 3169 price. Our General Public Licenses are designed to make sure that you 3170 have the freedom to distribute copies of free software (and charge for 3171 them if you wish), that you receive source code or can get it if you 3172 want it, that you can change the software or use pieces of it in new 3173 free programs, and that you know you can do these things. 3174 3175 To protect your rights, we need to prevent others from denying you 3176 these rights or asking you to surrender the rights. Therefore, you 3177 have certain responsibilities if you distribute copies of the software, 3178 or if you modify it: responsibilities to respect the freedom of others. 3179 3180 For example, if you distribute copies of such a program, whether 3181 gratis or for a fee, you must pass on to the recipients the same 3182 freedoms that you received. You must make sure that they, too, receive 3183 or can get the source code. And you must show them these terms so they 3184 know their rights. 3185 3186 Developers that use the GNU GPL protect your rights with two steps: 3187 (1) assert copyright on the software, and (2) offer you this License 3188 giving you legal permission to copy, distribute and/or modify it. 3189 3190 For the developers' and authors' protection, the GPL clearly explains 3191 that there is no warranty for this free software. For both users' and 3192 authors' sake, the GPL requires that modified versions be marked as 3193 changed, so that their problems will not be attributed erroneously to 3194 authors of previous versions. 3195 3196 Some devices are designed to deny users access to install or run 3197 modified versions of the software inside them, although the 3198 manufacturer can do so. This is fundamentally incompatible with the 3199 aim of protecting users' freedom to change the software. The 3200 systematic pattern of such abuse occurs in the area of products for 3201 individuals to use, which is precisely where it is most unacceptable. 3202 Therefore, we have designed this version of the GPL to prohibit the 3203 practice for those products. If such problems arise substantially in 3204 other domains, we stand ready to extend this provision to those domains 3205 in future versions of the GPL, as needed to protect the freedom of 3206 users. 3207 3208 Finally, every program is threatened constantly by software patents. 3209 States should not allow patents to restrict development and use of 3210 software on general-purpose computers, but in those that do, we wish to 3211 avoid the special danger that patents applied to a free program could 3212 make it effectively proprietary. To prevent this, the GPL assures that 3213 patents cannot be used to render the program non-free. 3214 3215 The precise terms and conditions for copying, distribution and 3216 modification follow. 3217 3218 TERMS AND CONDITIONS 3219 ==================== 3220 3221 0. Definitions. 3222 3223 "This License" refers to version 3 of the GNU General Public 3224 License. 3225 3226 "Copyright" also means copyright-like laws that apply to other 3227 kinds of works, such as semiconductor masks. 3228 3229 "The Program" refers to any copyrightable work licensed under this 3230 License. Each licensee is addressed as "you". "Licensees" and 3231 "recipients" may be individuals or organizations. 3232 3233 To "modify" a work means to copy from or adapt all or part of the 3234 work in a fashion requiring copyright permission, other than the 3235 making of an exact copy. The resulting work is called a "modified 3236 version" of the earlier work or a work "based on" the earlier work. 3237 3238 A "covered work" means either the unmodified Program or a work 3239 based on the Program. 3240 3241 To "propagate" a work means to do anything with it that, without 3242 permission, would make you directly or secondarily liable for 3243 infringement under applicable copyright law, except executing it 3244 on a computer or modifying a private copy. Propagation includes 3245 copying, distribution (with or without modification), making 3246 available to the public, and in some countries other activities as 3247 well. 3248 3249 To "convey" a work means any kind of propagation that enables other 3250 parties to make or receive copies. Mere interaction with a user 3251 through a computer network, with no transfer of a copy, is not 3252 conveying. 3253 3254 An interactive user interface displays "Appropriate Legal Notices" 3255 to the extent that it includes a convenient and prominently visible 3256 feature that (1) displays an appropriate copyright notice, and (2) 3257 tells the user that there is no warranty for the work (except to 3258 the extent that warranties are provided), that licensees may 3259 convey the work under this License, and how to view a copy of this 3260 License. If the interface presents a list of user commands or 3261 options, such as a menu, a prominent item in the list meets this 3262 criterion. 3263 3264 1. Source Code. 3265 3266 The "source code" for a work means the preferred form of the work 3267 for making modifications to it. "Object code" means any 3268 non-source form of a work. 3269 3270 A "Standard Interface" means an interface that either is an 3271 official standard defined by a recognized standards body, or, in 3272 the case of interfaces specified for a particular programming 3273 language, one that is widely used among developers working in that 3274 language. 3275 3276 The "System Libraries" of an executable work include anything, 3277 other than the work as a whole, that (a) is included in the normal 3278 form of packaging a Major Component, but which is not part of that 3279 Major Component, and (b) serves only to enable use of the work 3280 with that Major Component, or to implement a Standard Interface 3281 for which an implementation is available to the public in source 3282 code form. A "Major Component", in this context, means a major 3283 essential component (kernel, window system, and so on) of the 3284 specific operating system (if any) on which the executable work 3285 runs, or a compiler used to produce the work, or an object code 3286 interpreter used to run it. 3287 3288 The "Corresponding Source" for a work in object code form means all 3289 the source code needed to generate, install, and (for an executable 3290 work) run the object code and to modify the work, including 3291 scripts to control those activities. However, it does not include 3292 the work's System Libraries, or general-purpose tools or generally 3293 available free programs which are used unmodified in performing 3294 those activities but which are not part of the work. For example, 3295 Corresponding Source includes interface definition files 3296 associated with source files for the work, and the source code for 3297 shared libraries and dynamically linked subprograms that the work 3298 is specifically designed to require, such as by intimate data 3299 communication or control flow between those subprograms and other 3300 parts of the work. 3301 3302 The Corresponding Source need not include anything that users can 3303 regenerate automatically from other parts of the Corresponding 3304 Source. 3305 3306 The Corresponding Source for a work in source code form is that 3307 same work. 3308 3309 2. Basic Permissions. 3310 3311 All rights granted under this License are granted for the term of 3312 copyright on the Program, and are irrevocable provided the stated 3313 conditions are met. This License explicitly affirms your unlimited 3314 permission to run the unmodified Program. The output from running 3315 a covered work is covered by this License only if the output, 3316 given its content, constitutes a covered work. This License 3317 acknowledges your rights of fair use or other equivalent, as 3318 provided by copyright law. 3319 3320 You may make, run and propagate covered works that you do not 3321 convey, without conditions so long as your license otherwise 3322 remains in force. You may convey covered works to others for the 3323 sole purpose of having them make modifications exclusively for 3324 you, or provide you with facilities for running those works, 3325 provided that you comply with the terms of this License in 3326 conveying all material for which you do not control copyright. 3327 Those thus making or running the covered works for you must do so 3328 exclusively on your behalf, under your direction and control, on 3329 terms that prohibit them from making any copies of your 3330 copyrighted material outside their relationship with you. 3331 3332 Conveying under any other circumstances is permitted solely under 3333 the conditions stated below. Sublicensing is not allowed; section 3334 10 makes it unnecessary. 3335 3336 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 3337 3338 No covered work shall be deemed part of an effective technological 3339 measure under any applicable law fulfilling obligations under 3340 article 11 of the WIPO copyright treaty adopted on 20 December 3341 1996, or similar laws prohibiting or restricting circumvention of 3342 such measures. 3343 3344 When you convey a covered work, you waive any legal power to forbid 3345 circumvention of technological measures to the extent such 3346 circumvention is effected by exercising rights under this License 3347 with respect to the covered work, and you disclaim any intention 3348 to limit operation or modification of the work as a means of 3349 enforcing, against the work's users, your or third parties' legal 3350 rights to forbid circumvention of technological measures. 3351 3352 4. Conveying Verbatim Copies. 3353 3354 You may convey verbatim copies of the Program's source code as you 3355 receive it, in any medium, provided that you conspicuously and 3356 appropriately publish on each copy an appropriate copyright notice; 3357 keep intact all notices stating that this License and any 3358 non-permissive terms added in accord with section 7 apply to the 3359 code; keep intact all notices of the absence of any warranty; and 3360 give all recipients a copy of this License along with the Program. 3361 3362 You may charge any price or no price for each copy that you convey, 3363 and you may offer support or warranty protection for a fee. 3364 3365 5. Conveying Modified Source Versions. 3366 3367 You may convey a work based on the Program, or the modifications to 3368 produce it from the Program, in the form of source code under the 3369 terms of section 4, provided that you also meet all of these 3370 conditions: 3371 3372 a. The work must carry prominent notices stating that you 3373 modified it, and giving a relevant date. 3374 3375 b. The work must carry prominent notices stating that it is 3376 released under this License and any conditions added under 3377 section 7. This requirement modifies the requirement in 3378 section 4 to "keep intact all notices". 3379 3380 c. You must license the entire work, as a whole, under this 3381 License to anyone who comes into possession of a copy. This 3382 License will therefore apply, along with any applicable 3383 section 7 additional terms, to the whole of the work, and all 3384 its parts, regardless of how they are packaged. This License 3385 gives no permission to license the work in any other way, but 3386 it does not invalidate such permission if you have separately 3387 received it. 3388 3389 d. If the work has interactive user interfaces, each must display 3390 Appropriate Legal Notices; however, if the Program has 3391 interactive interfaces that do not display Appropriate Legal 3392 Notices, your work need not make them do so. 3393 3394 A compilation of a covered work with other separate and independent 3395 works, which are not by their nature extensions of the covered 3396 work, and which are not combined with it such as to form a larger 3397 program, in or on a volume of a storage or distribution medium, is 3398 called an "aggregate" if the compilation and its resulting 3399 copyright are not used to limit the access or legal rights of the 3400 compilation's users beyond what the individual works permit. 3401 Inclusion of a covered work in an aggregate does not cause this 3402 License to apply to the other parts of the aggregate. 3403 3404 6. Conveying Non-Source Forms. 3405 3406 You may convey a covered work in object code form under the terms 3407 of sections 4 and 5, provided that you also convey the 3408 machine-readable Corresponding Source under the terms of this 3409 License, in one of these ways: 3410 3411 a. Convey the object code in, or embodied in, a physical product 3412 (including a physical distribution medium), accompanied by the 3413 Corresponding Source fixed on a durable physical medium 3414 customarily used for software interchange. 3415 3416 b. Convey the object code in, or embodied in, a physical product 3417 (including a physical distribution medium), accompanied by a 3418 written offer, valid for at least three years and valid for 3419 as long as you offer spare parts or customer support for that 3420 product model, to give anyone who possesses the object code 3421 either (1) a copy of the Corresponding Source for all the 3422 software in the product that is covered by this License, on a 3423 durable physical medium customarily used for software 3424 interchange, for a price no more than your reasonable cost of 3425 physically performing this conveying of source, or (2) access 3426 to copy the Corresponding Source from a network server at no 3427 charge. 3428 3429 c. Convey individual copies of the object code with a copy of 3430 the written offer to provide the Corresponding Source. This 3431 alternative is allowed only occasionally and noncommercially, 3432 and only if you received the object code with such an offer, 3433 in accord with subsection 6b. 3434 3435 d. Convey the object code by offering access from a designated 3436 place (gratis or for a charge), and offer equivalent access 3437 to the Corresponding Source in the same way through the same 3438 place at no further charge. You need not require recipients 3439 to copy the Corresponding Source along with the object code. 3440 If the place to copy the object code is a network server, the 3441 Corresponding Source may be on a different server (operated 3442 by you or a third party) that supports equivalent copying 3443 facilities, provided you maintain clear directions next to 3444 the object code saying where to find the Corresponding Source. 3445 Regardless of what server hosts the Corresponding Source, you 3446 remain obligated to ensure that it is available for as long 3447 as needed to satisfy these requirements. 3448 3449 e. Convey the object code using peer-to-peer transmission, 3450 provided you inform other peers where the object code and 3451 Corresponding Source of the work are being offered to the 3452 general public at no charge under subsection 6d. 3453 3454 3455 A separable portion of the object code, whose source code is 3456 excluded from the Corresponding Source as a System Library, need 3457 not be included in conveying the object code work. 3458 3459 A "User Product" is either (1) a "consumer product", which means 3460 any tangible personal property which is normally used for personal, 3461 family, or household purposes, or (2) anything designed or sold for 3462 incorporation into a dwelling. In determining whether a product 3463 is a consumer product, doubtful cases shall be resolved in favor of 3464 coverage. For a particular product received by a particular user, 3465 "normally used" refers to a typical or common use of that class of 3466 product, regardless of the status of the particular user or of the 3467 way in which the particular user actually uses, or expects or is 3468 expected to use, the product. A product is a consumer product 3469 regardless of whether the product has substantial commercial, 3470 industrial or non-consumer uses, unless such uses represent the 3471 only significant mode of use of the product. 3472 3473 "Installation Information" for a User Product means any methods, 3474 procedures, authorization keys, or other information required to 3475 install and execute modified versions of a covered work in that 3476 User Product from a modified version of its Corresponding Source. 3477 The information must suffice to ensure that the continued 3478 functioning of the modified object code is in no case prevented or 3479 interfered with solely because modification has been made. 3480 3481 If you convey an object code work under this section in, or with, 3482 or specifically for use in, a User Product, and the conveying 3483 occurs as part of a transaction in which the right of possession 3484 and use of the User Product is transferred to the recipient in 3485 perpetuity or for a fixed term (regardless of how the transaction 3486 is characterized), the Corresponding Source conveyed under this 3487 section must be accompanied by the Installation Information. But 3488 this requirement does not apply if neither you nor any third party 3489 retains the ability to install modified object code on the User 3490 Product (for example, the work has been installed in ROM). 3491 3492 The requirement to provide Installation Information does not 3493 include a requirement to continue to provide support service, 3494 warranty, or updates for a work that has been modified or 3495 installed by the recipient, or for the User Product in which it 3496 has been modified or installed. Access to a network may be denied 3497 when the modification itself materially and adversely affects the 3498 operation of the network or violates the rules and protocols for 3499 communication across the network. 3500 3501 Corresponding Source conveyed, and Installation Information 3502 provided, in accord with this section must be in a format that is 3503 publicly documented (and with an implementation available to the 3504 public in source code form), and must require no special password 3505 or key for unpacking, reading or copying. 3506 3507 7. Additional Terms. 3508 3509 "Additional permissions" are terms that supplement the terms of 3510 this License by making exceptions from one or more of its 3511 conditions. Additional permissions that are applicable to the 3512 entire Program shall be treated as though they were included in 3513 this License, to the extent that they are valid under applicable 3514 law. If additional permissions apply only to part of the Program, 3515 that part may be used separately under those permissions, but the 3516 entire Program remains governed by this License without regard to 3517 the additional permissions. 3518 3519 When you convey a copy of a covered work, you may at your option 3520 remove any additional permissions from that copy, or from any part 3521 of it. (Additional permissions may be written to require their own 3522 removal in certain cases when you modify the work.) You may place 3523 additional permissions on material, added by you to a covered work, 3524 for which you have or can give appropriate copyright permission. 3525 3526 Notwithstanding any other provision of this License, for material 3527 you add to a covered work, you may (if authorized by the copyright 3528 holders of that material) supplement the terms of this License 3529 with terms: 3530 3531 a. Disclaiming warranty or limiting liability differently from 3532 the terms of sections 15 and 16 of this License; or 3533 3534 b. Requiring preservation of specified reasonable legal notices 3535 or author attributions in that material or in the Appropriate 3536 Legal Notices displayed by works containing it; or 3537 3538 c. Prohibiting misrepresentation of the origin of that material, 3539 or requiring that modified versions of such material be 3540 marked in reasonable ways as different from the original 3541 version; or 3542 3543 d. Limiting the use for publicity purposes of names of licensors 3544 or authors of the material; or 3545 3546 e. Declining to grant rights under trademark law for use of some 3547 trade names, trademarks, or service marks; or 3548 3549 f. Requiring indemnification of licensors and authors of that 3550 material by anyone who conveys the material (or modified 3551 versions of it) with contractual assumptions of liability to 3552 the recipient, for any liability that these contractual 3553 assumptions directly impose on those licensors and authors. 3554 3555 All other non-permissive additional terms are considered "further 3556 restrictions" within the meaning of section 10. If the Program as 3557 you received it, or any part of it, contains a notice stating that 3558 it is governed by this License along with a term that is a further 3559 restriction, you may remove that term. If a license document 3560 contains a further restriction but permits relicensing or 3561 conveying under this License, you may add to a covered work 3562 material governed by the terms of that license document, provided 3563 that the further restriction does not survive such relicensing or 3564 conveying. 3565 3566 If you add terms to a covered work in accord with this section, you 3567 must place, in the relevant source files, a statement of the 3568 additional terms that apply to those files, or a notice indicating 3569 where to find the applicable terms. 3570 3571 Additional terms, permissive or non-permissive, may be stated in 3572 the form of a separately written license, or stated as exceptions; 3573 the above requirements apply either way. 3574 3575 8. Termination. 3576 3577 You may not propagate or modify a covered work except as expressly 3578 provided under this License. Any attempt otherwise to propagate or 3579 modify it is void, and will automatically terminate your rights 3580 under this License (including any patent licenses granted under 3581 the third paragraph of section 11). 3582 3583 However, if you cease all violation of this License, then your 3584 license from a particular copyright holder is reinstated (a) 3585 provisionally, unless and until the copyright holder explicitly 3586 and finally terminates your license, and (b) permanently, if the 3587 copyright holder fails to notify you of the violation by some 3588 reasonable means prior to 60 days after the cessation. 3589 3590 Moreover, your license from a particular copyright holder is 3591 reinstated permanently if the copyright holder notifies you of the 3592 violation by some reasonable means, this is the first time you have 3593 received notice of violation of this License (for any work) from 3594 that copyright holder, and you cure the violation prior to 30 days 3595 after your receipt of the notice. 3596 3597 Termination of your rights under this section does not terminate 3598 the licenses of parties who have received copies or rights from 3599 you under this License. If your rights have been terminated and 3600 not permanently reinstated, you do not qualify to receive new 3601 licenses for the same material under section 10. 3602 3603 9. Acceptance Not Required for Having Copies. 3604 3605 You are not required to accept this License in order to receive or 3606 run a copy of the Program. Ancillary propagation of a covered work 3607 occurring solely as a consequence of using peer-to-peer 3608 transmission to receive a copy likewise does not require 3609 acceptance. However, nothing other than this License grants you 3610 permission to propagate or modify any covered work. These actions 3611 infringe copyright if you do not accept this License. Therefore, 3612 by modifying or propagating a covered work, you indicate your 3613 acceptance of this License to do so. 3614 3615 10. Automatic Licensing of Downstream Recipients. 3616 3617 Each time you convey a covered work, the recipient automatically 3618 receives a license from the original licensors, to run, modify and 3619 propagate that work, subject to this License. You are not 3620 responsible for enforcing compliance by third parties with this 3621 License. 3622 3623 An "entity transaction" is a transaction transferring control of an 3624 organization, or substantially all assets of one, or subdividing an 3625 organization, or merging organizations. If propagation of a 3626 covered work results from an entity transaction, each party to that 3627 transaction who receives a copy of the work also receives whatever 3628 licenses to the work the party's predecessor in interest had or 3629 could give under the previous paragraph, plus a right to 3630 possession of the Corresponding Source of the work from the 3631 predecessor in interest, if the predecessor has it or can get it 3632 with reasonable efforts. 3633 3634 You may not impose any further restrictions on the exercise of the 3635 rights granted or affirmed under this License. For example, you 3636 may not impose a license fee, royalty, or other charge for 3637 exercise of rights granted under this License, and you may not 3638 initiate litigation (including a cross-claim or counterclaim in a 3639 lawsuit) alleging that any patent claim is infringed by making, 3640 using, selling, offering for sale, or importing the Program or any 3641 portion of it. 3642 3643 11. Patents. 3644 3645 A "contributor" is a copyright holder who authorizes use under this 3646 License of the Program or a work on which the Program is based. 3647 The work thus licensed is called the contributor's "contributor 3648 version". 3649 3650 A contributor's "essential patent claims" are all patent claims 3651 owned or controlled by the contributor, whether already acquired or 3652 hereafter acquired, that would be infringed by some manner, 3653 permitted by this License, of making, using, or selling its 3654 contributor version, but do not include claims that would be 3655 infringed only as a consequence of further modification of the 3656 contributor version. For purposes of this definition, "control" 3657 includes the right to grant patent sublicenses in a manner 3658 consistent with the requirements of this License. 3659 3660 Each contributor grants you a non-exclusive, worldwide, 3661 royalty-free patent license under the contributor's essential 3662 patent claims, to make, use, sell, offer for sale, import and 3663 otherwise run, modify and propagate the contents of its 3664 contributor version. 3665 3666 In the following three paragraphs, a "patent license" is any 3667 express agreement or commitment, however denominated, not to 3668 enforce a patent (such as an express permission to practice a 3669 patent or covenant not to sue for patent infringement). To 3670 "grant" such a patent license to a party means to make such an 3671 agreement or commitment not to enforce a patent against the party. 3672 3673 If you convey a covered work, knowingly relying on a patent 3674 license, and the Corresponding Source of the work is not available 3675 for anyone to copy, free of charge and under the terms of this 3676 License, through a publicly available network server or other 3677 readily accessible means, then you must either (1) cause the 3678 Corresponding Source to be so available, or (2) arrange to deprive 3679 yourself of the benefit of the patent license for this particular 3680 work, or (3) arrange, in a manner consistent with the requirements 3681 of this License, to extend the patent license to downstream 3682 recipients. "Knowingly relying" means you have actual knowledge 3683 that, but for the patent license, your conveying the covered work 3684 in a country, or your recipient's use of the covered work in a 3685 country, would infringe one or more identifiable patents in that 3686 country that you have reason to believe are valid. 3687 3688 If, pursuant to or in connection with a single transaction or 3689 arrangement, you convey, or propagate by procuring conveyance of, a 3690 covered work, and grant a patent license to some of the parties 3691 receiving the covered work authorizing them to use, propagate, 3692 modify or convey a specific copy of the covered work, then the 3693 patent license you grant is automatically extended to all 3694 recipients of the covered work and works based on it. 3695 3696 A patent license is "discriminatory" if it does not include within 3697 the scope of its coverage, prohibits the exercise of, or is 3698 conditioned on the non-exercise of one or more of the rights that 3699 are specifically granted under this License. You may not convey a 3700 covered work if you are a party to an arrangement with a third 3701 party that is in the business of distributing software, under 3702 which you make payment to the third party based on the extent of 3703 your activity of conveying the work, and under which the third 3704 party grants, to any of the parties who would receive the covered 3705 work from you, a discriminatory patent license (a) in connection 3706 with copies of the covered work conveyed by you (or copies made 3707 from those copies), or (b) primarily for and in connection with 3708 specific products or compilations that contain the covered work, 3709 unless you entered into that arrangement, or that patent license 3710 was granted, prior to 28 March 2007. 3711 3712 Nothing in this License shall be construed as excluding or limiting 3713 any implied license or other defenses to infringement that may 3714 otherwise be available to you under applicable patent law. 3715 3716 12. No Surrender of Others' Freedom. 3717 3718 If conditions are imposed on you (whether by court order, 3719 agreement or otherwise) that contradict the conditions of this 3720 License, they do not excuse you from the conditions of this 3721 License. If you cannot convey a covered work so as to satisfy 3722 simultaneously your obligations under this License and any other 3723 pertinent obligations, then as a consequence you may not convey it 3724 at all. For example, if you agree to terms that obligate you to 3725 collect a royalty for further conveying from those to whom you 3726 convey the Program, the only way you could satisfy both those 3727 terms and this License would be to refrain entirely from conveying 3728 the Program. 3729 3730 13. Use with the GNU Affero General Public License. 3731 3732 Notwithstanding any other provision of this License, you have 3733 permission to link or combine any covered work with a work licensed 3734 under version 3 of the GNU Affero General Public License into a 3735 single combined work, and to convey the resulting work. The terms 3736 of this License will continue to apply to the part which is the 3737 covered work, but the special requirements of the GNU Affero 3738 General Public License, section 13, concerning interaction through 3739 a network will apply to the combination as such. 3740 3741 14. Revised Versions of this License. 3742 3743 The Free Software Foundation may publish revised and/or new 3744 versions of the GNU General Public License from time to time. 3745 Such new versions will be similar in spirit to the present 3746 version, but may differ in detail to address new problems or 3747 concerns. 3748 3749 Each version is given a distinguishing version number. If the 3750 Program specifies that a certain numbered version of the GNU 3751 General Public License "or any later version" applies to it, you 3752 have the option of following the terms and conditions either of 3753 that numbered version or of any later version published by the 3754 Free Software Foundation. If the Program does not specify a 3755 version number of the GNU General Public License, you may choose 3756 any version ever published by the Free Software Foundation. 3757 3758 If the Program specifies that a proxy can decide which future 3759 versions of the GNU General Public License can be used, that 3760 proxy's public statement of acceptance of a version permanently 3761 authorizes you to choose that version for the Program. 3762 3763 Later license versions may give you additional or different 3764 permissions. However, no additional obligations are imposed on any 3765 author or copyright holder as a result of your choosing to follow a 3766 later version. 3767 3768 15. Disclaimer of Warranty. 3769 3770 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 3771 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 3772 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 3773 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 3774 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 3775 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 3776 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 3777 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 3778 NECESSARY SERVICING, REPAIR OR CORRECTION. 3779 3780 16. Limitation of Liability. 3781 3782 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 3783 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 3784 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU 3785 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 3786 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 3787 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 3788 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 3789 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 3790 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 3791 THE POSSIBILITY OF SUCH DAMAGES. 3792 3793 17. Interpretation of Sections 15 and 16. 3794 3795 If the disclaimer of warranty and limitation of liability provided 3796 above cannot be given local legal effect according to their terms, 3797 reviewing courts shall apply local law that most closely 3798 approximates an absolute waiver of all civil liability in 3799 connection with the Program, unless a warranty or assumption of 3800 liability accompanies a copy of the Program in return for a fee. 3801 3802 3803 END OF TERMS AND CONDITIONS 3804 =========================== 3805 3806 How to Apply These Terms to Your New Programs 3807 ============================================= 3808 3809 If you develop a new program, and you want it to be of the greatest 3810 possible use to the public, the best way to achieve this is to make it 3811 free software which everyone can redistribute and change under these 3812 terms. 3813 3814 To do so, attach the following notices to the program. It is safest 3815 to attach them to the start of each source file to most effectively 3816 state the exclusion of warranty; and each file should have at least the 3817 "copyright" line and a pointer to where the full notice is found. 3818 3819 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 3820 Copyright (C) YEAR NAME OF AUTHOR 3821 3822 This program is free software: you can redistribute it and/or modify 3823 it under the terms of the GNU General Public License as published by 3824 the Free Software Foundation, either version 3 of the License, or (at 3825 your option) any later version. 3826 3827 This program is distributed in the hope that it will be useful, but 3828 WITHOUT ANY WARRANTY; without even the implied warranty of 3829 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 3830 General Public License for more details. 3831 3832 You should have received a copy of the GNU General Public License 3833 along with this program. If not, see `http://www.gnu.org/licenses/'. 3834 3835 Also add information on how to contact you by electronic and paper 3836 mail. 3837 3838 If the program does terminal interaction, make it output a short 3839 notice like this when it starts in an interactive mode: 3840 3841 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 3842 This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 3843 This is free software, and you are welcome to redistribute it 3844 under certain conditions; type `show c' for details. 3845 3846 The hypothetical commands `show w' and `show c' should show the 3847 appropriate parts of the General Public License. Of course, your 3848 program's commands might be different; for a GUI interface, you would 3849 use an "about box". 3850 3851 You should also get your employer (if you work as a programmer) or 3852 school, if any, to sign a "copyright disclaimer" for the program, if 3853 necessary. For more information on this, and how to apply and follow 3854 the GNU GPL, see `http://www.gnu.org/licenses/'. 3855 3856 The GNU General Public License does not permit incorporating your 3857 program into proprietary programs. If your program is a subroutine 3858 library, you may consider it more useful to permit linking proprietary 3859 applications with the library. If this is what you want to do, use the 3860 GNU Lesser General Public License instead of this License. But first, 3861 please read `http://www.gnu.org/philosophy/why-not-lgpl.html'. 3862 3863 3864 File: libgomp.info, Node: GNU Free Documentation License, Next: Funding, Prev: Copying, Up: Top 3865 3866 GNU Free Documentation License 3867 ****************************** 3868 3869 Version 1.3, 3 November 2008 3870 3871 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. 3872 `http://fsf.org/' 3873 3874 Everyone is permitted to copy and distribute verbatim copies 3875 of this license document, but changing it is not allowed. 3876 3877 0. PREAMBLE 3878 3879 The purpose of this License is to make a manual, textbook, or other 3880 functional and useful document "free" in the sense of freedom: to 3881 assure everyone the effective freedom to copy and redistribute it, 3882 with or without modifying it, either commercially or 3883 noncommercially. Secondarily, this License preserves for the 3884 author and publisher a way to get credit for their work, while not 3885 being considered responsible for modifications made by others. 3886 3887 This License is a kind of "copyleft", which means that derivative 3888 works of the document must themselves be free in the same sense. 3889 It complements the GNU General Public License, which is a copyleft 3890 license designed for free software. 3891 3892 We have designed this License in order to use it for manuals for 3893 free software, because free software needs free documentation: a 3894 free program should come with manuals providing the same freedoms 3895 that the software does. But this License is not limited to 3896 software manuals; it can be used for any textual work, regardless 3897 of subject matter or whether it is published as a printed book. 3898 We recommend this License principally for works whose purpose is 3899 instruction or reference. 3900 3901 1. APPLICABILITY AND DEFINITIONS 3902 3903 This License applies to any manual or other work, in any medium, 3904 that contains a notice placed by the copyright holder saying it 3905 can be distributed under the terms of this License. Such a notice 3906 grants a world-wide, royalty-free license, unlimited in duration, 3907 to use that work under the conditions stated herein. The 3908 "Document", below, refers to any such manual or work. Any member 3909 of the public is a licensee, and is addressed as "you". You 3910 accept the license if you copy, modify or distribute the work in a 3911 way requiring permission under copyright law. 3912 3913 A "Modified Version" of the Document means any work containing the 3914 Document or a portion of it, either copied verbatim, or with 3915 modifications and/or translated into another language. 3916 3917 A "Secondary Section" is a named appendix or a front-matter section 3918 of the Document that deals exclusively with the relationship of the 3919 publishers or authors of the Document to the Document's overall 3920 subject (or to related matters) and contains nothing that could 3921 fall directly within that overall subject. (Thus, if the Document 3922 is in part a textbook of mathematics, a Secondary Section may not 3923 explain any mathematics.) The relationship could be a matter of 3924 historical connection with the subject or with related matters, or 3925 of legal, commercial, philosophical, ethical or political position 3926 regarding them. 3927 3928 The "Invariant Sections" are certain Secondary Sections whose 3929 titles are designated, as being those of Invariant Sections, in 3930 the notice that says that the Document is released under this 3931 License. If a section does not fit the above definition of 3932 Secondary then it is not allowed to be designated as Invariant. 3933 The Document may contain zero Invariant Sections. If the Document 3934 does not identify any Invariant Sections then there are none. 3935 3936 The "Cover Texts" are certain short passages of text that are 3937 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 3938 that says that the Document is released under this License. A 3939 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 3940 be at most 25 words. 3941 3942 A "Transparent" copy of the Document means a machine-readable copy, 3943 represented in a format whose specification is available to the 3944 general public, that is suitable for revising the document 3945 straightforwardly with generic text editors or (for images 3946 composed of pixels) generic paint programs or (for drawings) some 3947 widely available drawing editor, and that is suitable for input to 3948 text formatters or for automatic translation to a variety of 3949 formats suitable for input to text formatters. A copy made in an 3950 otherwise Transparent file format whose markup, or absence of 3951 markup, has been arranged to thwart or discourage subsequent 3952 modification by readers is not Transparent. An image format is 3953 not Transparent if used for any substantial amount of text. A 3954 copy that is not "Transparent" is called "Opaque". 3955 3956 Examples of suitable formats for Transparent copies include plain 3957 ASCII without markup, Texinfo input format, LaTeX input format, 3958 SGML or XML using a publicly available DTD, and 3959 standard-conforming simple HTML, PostScript or PDF designed for 3960 human modification. Examples of transparent image formats include 3961 PNG, XCF and JPG. Opaque formats include proprietary formats that 3962 can be read and edited only by proprietary word processors, SGML or 3963 XML for which the DTD and/or processing tools are not generally 3964 available, and the machine-generated HTML, PostScript or PDF 3965 produced by some word processors for output purposes only. 3966 3967 The "Title Page" means, for a printed book, the title page itself, 3968 plus such following pages as are needed to hold, legibly, the 3969 material this License requires to appear in the title page. For 3970 works in formats which do not have any title page as such, "Title 3971 Page" means the text near the most prominent appearance of the 3972 work's title, preceding the beginning of the body of the text. 3973 3974 The "publisher" means any person or entity that distributes copies 3975 of the Document to the public. 3976 3977 A section "Entitled XYZ" means a named subunit of the Document 3978 whose title either is precisely XYZ or contains XYZ in parentheses 3979 following text that translates XYZ in another language. (Here XYZ 3980 stands for a specific section name mentioned below, such as 3981 "Acknowledgements", "Dedications", "Endorsements", or "History".) 3982 To "Preserve the Title" of such a section when you modify the 3983 Document means that it remains a section "Entitled XYZ" according 3984 to this definition. 3985 3986 The Document may include Warranty Disclaimers next to the notice 3987 which states that this License applies to the Document. These 3988 Warranty Disclaimers are considered to be included by reference in 3989 this License, but only as regards disclaiming warranties: any other 3990 implication that these Warranty Disclaimers may have is void and 3991 has no effect on the meaning of this License. 3992 3993 2. VERBATIM COPYING 3994 3995 You may copy and distribute the Document in any medium, either 3996 commercially or noncommercially, provided that this License, the 3997 copyright notices, and the license notice saying this License 3998 applies to the Document are reproduced in all copies, and that you 3999 add no other conditions whatsoever to those of this License. You 4000 may not use technical measures to obstruct or control the reading 4001 or further copying of the copies you make or distribute. However, 4002 you may accept compensation in exchange for copies. If you 4003 distribute a large enough number of copies you must also follow 4004 the conditions in section 3. 4005 4006 You may also lend copies, under the same conditions stated above, 4007 and you may publicly display copies. 4008 4009 3. COPYING IN QUANTITY 4010 4011 If you publish printed copies (or copies in media that commonly 4012 have printed covers) of the Document, numbering more than 100, and 4013 the Document's license notice requires Cover Texts, you must 4014 enclose the copies in covers that carry, clearly and legibly, all 4015 these Cover Texts: Front-Cover Texts on the front cover, and 4016 Back-Cover Texts on the back cover. Both covers must also clearly 4017 and legibly identify you as the publisher of these copies. The 4018 front cover must present the full title with all words of the 4019 title equally prominent and visible. You may add other material 4020 on the covers in addition. Copying with changes limited to the 4021 covers, as long as they preserve the title of the Document and 4022 satisfy these conditions, can be treated as verbatim copying in 4023 other respects. 4024 4025 If the required texts for either cover are too voluminous to fit 4026 legibly, you should put the first ones listed (as many as fit 4027 reasonably) on the actual cover, and continue the rest onto 4028 adjacent pages. 4029 4030 If you publish or distribute Opaque copies of the Document 4031 numbering more than 100, you must either include a 4032 machine-readable Transparent copy along with each Opaque copy, or 4033 state in or with each Opaque copy a computer-network location from 4034 which the general network-using public has access to download 4035 using public-standard network protocols a complete Transparent 4036 copy of the Document, free of added material. If you use the 4037 latter option, you must take reasonably prudent steps, when you 4038 begin distribution of Opaque copies in quantity, to ensure that 4039 this Transparent copy will remain thus accessible at the stated 4040 location until at least one year after the last time you 4041 distribute an Opaque copy (directly or through your agents or 4042 retailers) of that edition to the public. 4043 4044 It is requested, but not required, that you contact the authors of 4045 the Document well before redistributing any large number of 4046 copies, to give them a chance to provide you with an updated 4047 version of the Document. 4048 4049 4. MODIFICATIONS 4050 4051 You may copy and distribute a Modified Version of the Document 4052 under the conditions of sections 2 and 3 above, provided that you 4053 release the Modified Version under precisely this License, with 4054 the Modified Version filling the role of the Document, thus 4055 licensing distribution and modification of the Modified Version to 4056 whoever possesses a copy of it. In addition, you must do these 4057 things in the Modified Version: 4058 4059 A. Use in the Title Page (and on the covers, if any) a title 4060 distinct from that of the Document, and from those of 4061 previous versions (which should, if there were any, be listed 4062 in the History section of the Document). You may use the 4063 same title as a previous version if the original publisher of 4064 that version gives permission. 4065 4066 B. List on the Title Page, as authors, one or more persons or 4067 entities responsible for authorship of the modifications in 4068 the Modified Version, together with at least five of the 4069 principal authors of the Document (all of its principal 4070 authors, if it has fewer than five), unless they release you 4071 from this requirement. 4072 4073 C. State on the Title page the name of the publisher of the 4074 Modified Version, as the publisher. 4075 4076 D. Preserve all the copyright notices of the Document. 4077 4078 E. Add an appropriate copyright notice for your modifications 4079 adjacent to the other copyright notices. 4080 4081 F. Include, immediately after the copyright notices, a license 4082 notice giving the public permission to use the Modified 4083 Version under the terms of this License, in the form shown in 4084 the Addendum below. 4085 4086 G. Preserve in that license notice the full lists of Invariant 4087 Sections and required Cover Texts given in the Document's 4088 license notice. 4089 4090 H. Include an unaltered copy of this License. 4091 4092 I. Preserve the section Entitled "History", Preserve its Title, 4093 and add to it an item stating at least the title, year, new 4094 authors, and publisher of the Modified Version as given on 4095 the Title Page. If there is no section Entitled "History" in 4096 the Document, create one stating the title, year, authors, 4097 and publisher of the Document as given on its Title Page, 4098 then add an item describing the Modified Version as stated in 4099 the previous sentence. 4100 4101 J. Preserve the network location, if any, given in the Document 4102 for public access to a Transparent copy of the Document, and 4103 likewise the network locations given in the Document for 4104 previous versions it was based on. These may be placed in 4105 the "History" section. You may omit a network location for a 4106 work that was published at least four years before the 4107 Document itself, or if the original publisher of the version 4108 it refers to gives permission. 4109 4110 K. For any section Entitled "Acknowledgements" or "Dedications", 4111 Preserve the Title of the section, and preserve in the 4112 section all the substance and tone of each of the contributor 4113 acknowledgements and/or dedications given therein. 4114 4115 L. Preserve all the Invariant Sections of the Document, 4116 unaltered in their text and in their titles. Section numbers 4117 or the equivalent are not considered part of the section 4118 titles. 4119 4120 M. Delete any section Entitled "Endorsements". Such a section 4121 may not be included in the Modified Version. 4122 4123 N. Do not retitle any existing section to be Entitled 4124 "Endorsements" or to conflict in title with any Invariant 4125 Section. 4126 4127 O. Preserve any Warranty Disclaimers. 4128 4129 If the Modified Version includes new front-matter sections or 4130 appendices that qualify as Secondary Sections and contain no 4131 material copied from the Document, you may at your option 4132 designate some or all of these sections as invariant. To do this, 4133 add their titles to the list of Invariant Sections in the Modified 4134 Version's license notice. These titles must be distinct from any 4135 other section titles. 4136 4137 You may add a section Entitled "Endorsements", provided it contains 4138 nothing but endorsements of your Modified Version by various 4139 parties--for example, statements of peer review or that the text 4140 has been approved by an organization as the authoritative 4141 definition of a standard. 4142 4143 You may add a passage of up to five words as a Front-Cover Text, 4144 and a passage of up to 25 words as a Back-Cover Text, to the end 4145 of the list of Cover Texts in the Modified Version. Only one 4146 passage of Front-Cover Text and one of Back-Cover Text may be 4147 added by (or through arrangements made by) any one entity. If the 4148 Document already includes a cover text for the same cover, 4149 previously added by you or by arrangement made by the same entity 4150 you are acting on behalf of, you may not add another; but you may 4151 replace the old one, on explicit permission from the previous 4152 publisher that added the old one. 4153 4154 The author(s) and publisher(s) of the Document do not by this 4155 License give permission to use their names for publicity for or to 4156 assert or imply endorsement of any Modified Version. 4157 4158 5. COMBINING DOCUMENTS 4159 4160 You may combine the Document with other documents released under 4161 this License, under the terms defined in section 4 above for 4162 modified versions, provided that you include in the combination 4163 all of the Invariant Sections of all of the original documents, 4164 unmodified, and list them all as Invariant Sections of your 4165 combined work in its license notice, and that you preserve all 4166 their Warranty Disclaimers. 4167 4168 The combined work need only contain one copy of this License, and 4169 multiple identical Invariant Sections may be replaced with a single 4170 copy. If there are multiple Invariant Sections with the same name 4171 but different contents, make the title of each such section unique 4172 by adding at the end of it, in parentheses, the name of the 4173 original author or publisher of that section if known, or else a 4174 unique number. Make the same adjustment to the section titles in 4175 the list of Invariant Sections in the license notice of the 4176 combined work. 4177 4178 In the combination, you must combine any sections Entitled 4179 "History" in the various original documents, forming one section 4180 Entitled "History"; likewise combine any sections Entitled 4181 "Acknowledgements", and any sections Entitled "Dedications". You 4182 must delete all sections Entitled "Endorsements." 4183 4184 6. COLLECTIONS OF DOCUMENTS 4185 4186 You may make a collection consisting of the Document and other 4187 documents released under this License, and replace the individual 4188 copies of this License in the various documents with a single copy 4189 that is included in the collection, provided that you follow the 4190 rules of this License for verbatim copying of each of the 4191 documents in all other respects. 4192 4193 You may extract a single document from such a collection, and 4194 distribute it individually under this License, provided you insert 4195 a copy of this License into the extracted document, and follow 4196 this License in all other respects regarding verbatim copying of 4197 that document. 4198 4199 7. AGGREGATION WITH INDEPENDENT WORKS 4200 4201 A compilation of the Document or its derivatives with other 4202 separate and independent documents or works, in or on a volume of 4203 a storage or distribution medium, is called an "aggregate" if the 4204 copyright resulting from the compilation is not used to limit the 4205 legal rights of the compilation's users beyond what the individual 4206 works permit. When the Document is included in an aggregate, this 4207 License does not apply to the other works in the aggregate which 4208 are not themselves derivative works of the Document. 4209 4210 If the Cover Text requirement of section 3 is applicable to these 4211 copies of the Document, then if the Document is less than one half 4212 of the entire aggregate, the Document's Cover Texts may be placed 4213 on covers that bracket the Document within the aggregate, or the 4214 electronic equivalent of covers if the Document is in electronic 4215 form. Otherwise they must appear on printed covers that bracket 4216 the whole aggregate. 4217 4218 8. TRANSLATION 4219 4220 Translation is considered a kind of modification, so you may 4221 distribute translations of the Document under the terms of section 4222 4. Replacing Invariant Sections with translations requires special 4223 permission from their copyright holders, but you may include 4224 translations of some or all Invariant Sections in addition to the 4225 original versions of these Invariant Sections. You may include a 4226 translation of this License, and all the license notices in the 4227 Document, and any Warranty Disclaimers, provided that you also 4228 include the original English version of this License and the 4229 original versions of those notices and disclaimers. In case of a 4230 disagreement between the translation and the original version of 4231 this License or a notice or disclaimer, the original version will 4232 prevail. 4233 4234 If a section in the Document is Entitled "Acknowledgements", 4235 "Dedications", or "History", the requirement (section 4) to 4236 Preserve its Title (section 1) will typically require changing the 4237 actual title. 4238 4239 9. TERMINATION 4240 4241 You may not copy, modify, sublicense, or distribute the Document 4242 except as expressly provided under this License. Any attempt 4243 otherwise to copy, modify, sublicense, or distribute it is void, 4244 and will automatically terminate your rights under this License. 4245 4246 However, if you cease all violation of this License, then your 4247 license from a particular copyright holder is reinstated (a) 4248 provisionally, unless and until the copyright holder explicitly 4249 and finally terminates your license, and (b) permanently, if the 4250 copyright holder fails to notify you of the violation by some 4251 reasonable means prior to 60 days after the cessation. 4252 4253 Moreover, your license from a particular copyright holder is 4254 reinstated permanently if the copyright holder notifies you of the 4255 violation by some reasonable means, this is the first time you have 4256 received notice of violation of this License (for any work) from 4257 that copyright holder, and you cure the violation prior to 30 days 4258 after your receipt of the notice. 4259 4260 Termination of your rights under this section does not terminate 4261 the licenses of parties who have received copies or rights from 4262 you under this License. If your rights have been terminated and 4263 not permanently reinstated, receipt of a copy of some or all of 4264 the same material does not give you any rights to use it. 4265 4266 10. FUTURE REVISIONS OF THIS LICENSE 4267 4268 The Free Software Foundation may publish new, revised versions of 4269 the GNU Free Documentation License from time to time. Such new 4270 versions will be similar in spirit to the present version, but may 4271 differ in detail to address new problems or concerns. See 4272 `http://www.gnu.org/copyleft/'. 4273 4274 Each version of the License is given a distinguishing version 4275 number. If the Document specifies that a particular numbered 4276 version of this License "or any later version" applies to it, you 4277 have the option of following the terms and conditions either of 4278 that specified version or of any later version that has been 4279 published (not as a draft) by the Free Software Foundation. If 4280 the Document does not specify a version number of this License, 4281 you may choose any version ever published (not as a draft) by the 4282 Free Software Foundation. If the Document specifies that a proxy 4283 can decide which future versions of this License can be used, that 4284 proxy's public statement of acceptance of a version permanently 4285 authorizes you to choose that version for the Document. 4286 4287 11. RELICENSING 4288 4289 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any 4290 World Wide Web server that publishes copyrightable works and also 4291 provides prominent facilities for anybody to edit those works. A 4292 public wiki that anybody can edit is an example of such a server. 4293 A "Massive Multiauthor Collaboration" (or "MMC") contained in the 4294 site means any set of copyrightable works thus published on the MMC 4295 site. 4296 4297 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 4298 license published by Creative Commons Corporation, a not-for-profit 4299 corporation with a principal place of business in San Francisco, 4300 California, as well as future copyleft versions of that license 4301 published by that same organization. 4302 4303 "Incorporate" means to publish or republish a Document, in whole or 4304 in part, as part of another Document. 4305 4306 An MMC is "eligible for relicensing" if it is licensed under this 4307 License, and if all works that were first published under this 4308 License somewhere other than this MMC, and subsequently 4309 incorporated in whole or in part into the MMC, (1) had no cover 4310 texts or invariant sections, and (2) were thus incorporated prior 4311 to November 1, 2008. 4312 4313 The operator of an MMC Site may republish an MMC contained in the 4314 site under CC-BY-SA on the same site at any time before August 1, 4315 2009, provided the MMC is eligible for relicensing. 4316 4317 4318 ADDENDUM: How to use this License for your documents 4319 ==================================================== 4320 4321 To use this License in a document you have written, include a copy of 4322 the License in the document and put the following copyright and license 4323 notices just after the title page: 4324 4325 Copyright (C) YEAR YOUR NAME. 4326 Permission is granted to copy, distribute and/or modify this document 4327 under the terms of the GNU Free Documentation License, Version 1.3 4328 or any later version published by the Free Software Foundation; 4329 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 4330 Texts. A copy of the license is included in the section entitled ``GNU 4331 Free Documentation License''. 4332 4333 If you have Invariant Sections, Front-Cover Texts and Back-Cover 4334 Texts, replace the "with...Texts." line with this: 4335 4336 with the Invariant Sections being LIST THEIR TITLES, with 4337 the Front-Cover Texts being LIST, and with the Back-Cover Texts 4338 being LIST. 4339 4340 If you have Invariant Sections without Cover Texts, or some other 4341 combination of the three, merge those two alternatives to suit the 4342 situation. 4343 4344 If your document contains nontrivial examples of program code, we 4345 recommend releasing these examples in parallel under your choice of 4346 free software license, such as the GNU General Public License, to 4347 permit their use in free software. 4348 4349 4350 File: libgomp.info, Node: Funding, Next: Library Index, Prev: GNU Free Documentation License, Up: Top 4351 4352 Funding Free Software 4353 ********************* 4354 4355 If you want to have more free software a few years from now, it makes 4356 sense for you to help encourage people to contribute funds for its 4357 development. The most effective approach known is to encourage 4358 commercial redistributors to donate. 4359 4360 Users of free software systems can boost the pace of development by 4361 encouraging for-a-fee distributors to donate part of their selling price 4362 to free software developers--the Free Software Foundation, and others. 4363 4364 The way to convince distributors to do this is to demand it and 4365 expect it from them. So when you compare distributors, judge them 4366 partly by how much they give to free software development. Show 4367 distributors they must compete to be the one who gives the most. 4368 4369 To make this approach work, you must insist on numbers that you can 4370 compare, such as, "We will donate ten dollars to the Frobnitz project 4371 for each disk sold." Don't be satisfied with a vague promise, such as 4372 "A portion of the profits are donated," since it doesn't give a basis 4373 for comparison. 4374 4375 Even a precise fraction "of the profits from this disk" is not very 4376 meaningful, since creative accounting and unrelated business decisions 4377 can greatly alter what fraction of the sales price counts as profit. 4378 If the price you pay is $50, ten percent of the profit is probably less 4379 than a dollar; it might be a few cents, or nothing at all. 4380 4381 Some redistributors do development work themselves. This is useful 4382 too; but to keep everyone honest, you need to inquire how much they do, 4383 and what kind. Some kinds of development make much more long-term 4384 difference than others. For example, maintaining a separate version of 4385 a program contributes very little; maintaining the standard version of a 4386 program for the whole community contributes much. Easy new ports 4387 contribute little, since someone else would surely do them; difficult 4388 ports such as adding a new CPU to the GNU Compiler Collection 4389 contribute more; major new features or packages contribute the most. 4390 4391 By establishing the idea that supporting further development is "the 4392 proper thing to do" when distributing free software for a fee, we can 4393 assure a steady flow of resources into making more free software. 4394 4395 Copyright (C) 1994 Free Software Foundation, Inc. 4396 Verbatim copying and redistribution of this section is permitted 4397 without royalty; alteration is not permitted. 4398 4399 4400 File: libgomp.info, Node: Library Index, Prev: Funding, Up: Top 4401 4402 Library Index 4403 ************* 4404 4405 [index] 4406 * Menu: 4407 4408 * Environment Variable <1>: GOMP_RTEMS_THREAD_POOLS. 4409 (line 6) 4410 * Environment Variable <2>: GOMP_SPINCOUNT. (line 6) 4411 * Environment Variable <3>: GOMP_STACKSIZE. (line 6) 4412 * Environment Variable <4>: GOMP_DEBUG. (line 6) 4413 * Environment Variable <5>: GOMP_CPU_AFFINITY. (line 6) 4414 * Environment Variable <6>: OMP_WAIT_POLICY. (line 6) 4415 * Environment Variable <7>: OMP_THREAD_LIMIT. (line 6) 4416 * Environment Variable <8>: OMP_SCHEDULE. (line 6) 4417 * Environment Variable <9>: OMP_STACKSIZE. (line 6) 4418 * Environment Variable <10>: OMP_PLACES. (line 6) 4419 * Environment Variable <11>: OMP_PROC_BIND. (line 6) 4420 * Environment Variable <12>: OMP_NUM_THREADS. (line 6) 4421 * Environment Variable <13>: OMP_NESTED. (line 6) 4422 * Environment Variable <14>: OMP_MAX_TASK_PRIORITY. (line 6) 4423 * Environment Variable <15>: OMP_MAX_ACTIVE_LEVELS. (line 6) 4424 * Environment Variable <16>: OMP_DYNAMIC. (line 6) 4425 * Environment Variable <17>: OMP_DEFAULT_DEVICE. (line 6) 4426 * Environment Variable <18>: OMP_DISPLAY_ENV. (line 6) 4427 * Environment Variable: OMP_CANCELLATION. (line 6) 4428 * FDL, GNU Free Documentation License: GNU Free Documentation License. 4429 (line 6) 4430 * Implementation specific setting <1>: GOMP_RTEMS_THREAD_POOLS. 4431 (line 6) 4432 * Implementation specific setting <2>: GOMP_SPINCOUNT. (line 6) 4433 * Implementation specific setting <3>: GOMP_STACKSIZE. (line 6) 4434 * Implementation specific setting <4>: OMP_SCHEDULE. (line 6) 4435 * Implementation specific setting <5>: OMP_NUM_THREADS. (line 6) 4436 * Implementation specific setting: OMP_NESTED. (line 6) 4437 * Introduction: Top. (line 6) 4438 4439 4440 4441 Tag Table: 4442 Node: Top2130 4443 Node: Enabling OpenMP4536 4444 Node: Runtime Library Routines5323 4445 Node: omp_get_active_level8386 4446 Node: omp_get_ancestor_thread_num9084 4447 Node: omp_get_cancellation10011 4448 Node: omp_get_default_device10823 4449 Node: omp_get_dynamic11497 4450 Node: omp_get_level12370 4451 Node: omp_get_max_active_levels12988 4452 Node: omp_get_max_task_priority13691 4453 Node: omp_get_max_threads14309 4454 Node: omp_get_nested15064 4455 Node: omp_get_num_devices15976 4456 Node: omp_get_num_procs16495 4457 Node: omp_get_num_teams17032 4458 Node: omp_get_num_threads17546 4459 Node: omp_get_proc_bind18633 4460 Node: omp_get_schedule19551 4461 Node: omp_get_team_num20500 4462 Node: omp_get_team_size20997 4463 Node: omp_get_thread_limit21954 4464 Node: omp_get_thread_num22571 4465 Node: omp_in_parallel23440 4466 Node: omp_in_final24087 4467 Node: omp_is_initial_device24759 4468 Node: omp_set_default_device25450 4469 Node: omp_set_dynamic26238 4470 Node: omp_set_max_active_levels27121 4471 Node: omp_set_nested27895 4472 Node: omp_set_num_threads28784 4473 Node: omp_set_schedule29649 4474 Node: omp_init_lock30725 4475 Node: omp_set_lock31375 4476 Node: omp_test_lock32227 4477 Node: omp_unset_lock33200 4478 Node: omp_destroy_lock34128 4479 Node: omp_init_nest_lock34802 4480 Node: omp_set_nest_lock35534 4481 Node: omp_test_nest_lock36451 4482 Node: omp_unset_nest_lock37481 4483 Node: omp_destroy_nest_lock38493 4484 Node: omp_get_wtick39241 4485 Node: omp_get_wtime39831 4486 Node: Environment Variables40605 4487 Node: OMP_CANCELLATION42160 4488 Node: OMP_DISPLAY_ENV42693 4489 Node: OMP_DEFAULT_DEVICE43396 4490 Node: OMP_DYNAMIC44176 4491 Node: OMP_MAX_ACTIVE_LEVELS44772 4492 Node: OMP_MAX_TASK_PRIORITY45422 4493 Node: OMP_NESTED46082 4494 Node: OMP_NUM_THREADS46687 4495 Node: OMP_PROC_BIND47376 4496 Node: OMP_PLACES48568 4497 Node: OMP_STACKSIZE50745 4498 Node: OMP_SCHEDULE51569 4499 Node: OMP_THREAD_LIMIT52267 4500 Node: OMP_WAIT_POLICY52867 4501 Node: GOMP_CPU_AFFINITY53559 4502 Node: GOMP_DEBUG55290 4503 Node: GOMP_STACKSIZE55797 4504 Node: GOMP_SPINCOUNT56626 4505 Node: GOMP_RTEMS_THREAD_POOLS57835 4506 Node: Enabling OpenACC60016 4507 Node: OpenACC Runtime Library Routines61010 4508 Node: acc_get_num_devices64804 4509 Node: acc_set_device_type65527 4510 Node: acc_get_device_type66288 4511 Node: acc_set_device_num67000 4512 Node: acc_get_device_num67799 4513 Node: acc_async_test68592 4514 Node: acc_async_test_all69577 4515 Node: acc_wait70472 4516 Node: acc_wait_all71104 4517 Node: acc_wait_all_async71681 4518 Node: acc_wait_async72430 4519 Node: acc_init73134 4520 Node: acc_shutdown73777 4521 Node: acc_on_device74442 4522 Node: acc_malloc75440 4523 Node: acc_free75937 4524 Node: acc_copyin76363 4525 Node: acc_present_or_copyin77464 4526 Node: acc_create79074 4527 Node: acc_present_or_create80220 4528 Node: acc_copyout81838 4529 Node: acc_delete82853 4530 Node: acc_update_device83820 4531 Node: acc_update_self84924 4532 Node: acc_map_data86020 4533 Node: acc_unmap_data86703 4534 Node: acc_deviceptr87222 4535 Node: acc_hostptr87791 4536 Node: acc_is_present88354 4537 Node: acc_memcpy_to_device89868 4538 Node: acc_memcpy_from_device90529 4539 Node: acc_get_current_cuda_device91211 4540 Node: acc_get_current_cuda_context91806 4541 Node: acc_get_cuda_stream92398 4542 Node: acc_set_cuda_stream92952 4543 Node: OpenACC Environment Variables93483 4544 Node: ACC_DEVICE_TYPE93942 4545 Node: ACC_DEVICE_NUM94178 4546 Node: GCC_ACC_NOTIFY94435 4547 Node: CUDA Streams Usage94658 4548 Ref: CUDA Streams Usage-Footnote-196558 4549 Node: OpenACC Library Interoperability96667 4550 Ref: OpenACC Library Interoperability-Footnote-1103013 4551 Ref: OpenACC Library Interoperability-Footnote-2103265 4552 Node: The libgomp ABI103473 4553 Node: Implementing MASTER construct104329 4554 Node: Implementing CRITICAL construct104743 4555 Node: Implementing ATOMIC construct105482 4556 Node: Implementing FLUSH construct105963 4557 Node: Implementing BARRIER construct106234 4558 Node: Implementing THREADPRIVATE construct106503 4559 Node: Implementing PRIVATE clause107155 4560 Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses107736 4561 Node: Implementing REDUCTION clause109060 4562 Node: Implementing PARALLEL construct109617 4563 Node: Implementing FOR construct110874 4564 Node: Implementing ORDERED construct112872 4565 Node: Implementing SECTIONS construct113178 4566 Node: Implementing SINGLE construct113944 4567 Node: Implementing OpenACC's PARALLEL construct114656 4568 Node: Reporting Bugs114914 4569 Node: Copying115276 4570 Node: GNU Free Documentation License152841 4571 Node: Funding177983 4572 Node: Library Index180508 4573 4574 End Tag Table 4575
Indexes created Fri Mar 20 00:23:26 UTC 2026