libgomp.info revision 1.1.1.8
1 This is libgomp.info, produced by makeinfo version 6.5 from 2 libgomp.texi. 3 4 Copyright (C) 2006-2018 Free Software Foundation, Inc. 5 6 Permission is granted to copy, distribute and/or modify this document 7 under the terms of the GNU Free Documentation License, Version 1.3 or 8 any later version published by the Free Software Foundation; with the 9 Invariant Sections being "Funding Free Software", the Front-Cover texts 10 being (a) (see below), and with the Back-Cover Texts being (b) (see 11 below). A copy of the license is included in the section entitled "GNU 12 Free Documentation License". 13 14 (a) The FSF's Front-Cover Text is: 15 16 A GNU Manual 17 18 (b) The FSF's Back-Cover Text is: 19 20 You have freedom to copy and modify this GNU Manual, like GNU 21 software. Copies published by the Free Software Foundation raise funds 22 for GNU development. 23 INFO-DIR-SECTION GNU Libraries 24 START-INFO-DIR-ENTRY 25 * libgomp: (libgomp). GNU Offloading and Multi Processing Runtime Library. 26 END-INFO-DIR-ENTRY 27 28 This manual documents libgomp, the GNU Offloading and Multi 29 Processing Runtime library. This is the GNU implementation of the 30 OpenMP and OpenACC APIs for parallel and accelerator programming in 31 C/C++ and Fortran. 32 33 Published by the Free Software Foundation 51 Franklin Street, Fifth 34 Floor Boston, MA 02110-1301 USA 35 36 Copyright (C) 2006-2018 Free Software Foundation, Inc. 37 38 Permission is granted to copy, distribute and/or modify this document 39 under the terms of the GNU Free Documentation License, Version 1.3 or 40 any later version published by the Free Software Foundation; with the 41 Invariant Sections being "Funding Free Software", the Front-Cover texts 42 being (a) (see below), and with the Back-Cover Texts being (b) (see 43 below). A copy of the license is included in the section entitled "GNU 44 Free Documentation License". 45 46 (a) The FSF's Front-Cover Text is: 47 48 A GNU Manual 49 50 (b) The FSF's Back-Cover Text is: 51 52 You have freedom to copy and modify this GNU Manual, like GNU 53 software. Copies published by the Free Software Foundation raise funds 54 for GNU development. 55 56 57 File: libgomp.info, Node: Top, Next: Enabling OpenMP, Up: (dir) 58 59 Introduction 60 ************ 61 62 This manual documents the usage of libgomp, the GNU Offloading and Multi 63 Processing Runtime Library. This includes the GNU implementation of the 64 OpenMP (http://www.openmp.org) Application Programming Interface (API) 65 for multi-platform shared-memory parallel programming in C/C++ and 66 Fortran, and the GNU implementation of the OpenACC 67 (https://www.openacc.org) Application Programming Interface (API) for 68 offloading of code to accelerator devices in C/C++ and Fortran. 69 70 Originally, libgomp implemented the GNU OpenMP Runtime Library. 71 Based on this, support for OpenACC and offloading (both OpenACC and 72 OpenMP 4's target construct) has been added later on, and the library's 73 name changed to GNU Offloading and Multi Processing Runtime Library. 74 75 * Menu: 76 77 * Enabling OpenMP:: How to enable OpenMP for your applications. 78 * Runtime Library Routines:: The OpenMP runtime application programming 79 interface. 80 * Environment Variables:: Influencing runtime behavior with environment 81 variables. 82 * Enabling OpenACC:: How to enable OpenACC for your 83 applications. 84 * OpenACC Runtime Library Routines:: The OpenACC runtime application 85 programming interface. 86 * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with 87 environment variables. 88 * CUDA Streams Usage:: Notes on the implementation of 89 asynchronous operations. 90 * OpenACC Library Interoperability:: OpenACC library interoperability with the 91 NVIDIA CUBLAS library. 92 * The libgomp ABI:: Notes on the external ABI presented by libgomp. 93 * Reporting Bugs:: How to report bugs in the GNU Offloading and 94 Multi Processing Runtime Library. 95 * Copying:: GNU general public license says 96 how you can copy and share libgomp. 97 * GNU Free Documentation License:: 98 How you can copy and share this manual. 99 * Funding:: How to help assure continued work for free 100 software. 101 * Library Index:: Index of this documentation. 102 103 104 File: libgomp.info, Node: Enabling OpenMP, Next: Runtime Library Routines, Prev: Top, Up: Top 105 106 1 Enabling OpenMP 107 ***************** 108 109 To activate the OpenMP extensions for C/C++ and Fortran, the 110 compile-time flag '-fopenmp' must be specified. This enables the OpenMP 111 directive '#pragma omp' in C/C++ and '!$omp' directives in free form, 112 'c$omp', '*$omp' and '!$omp' directives in fixed form, '!$' conditional 113 compilation sentinels in free form and 'c$', '*$' and '!$' sentinels in 114 fixed form, for Fortran. The flag also arranges for automatic linking 115 of the OpenMP runtime library (*note Runtime Library Routines::). 116 117 A complete description of all OpenMP directives accepted may be found 118 in the OpenMP Application Program Interface (http://www.openmp.org) 119 manual, version 4.5. 120 121 122 File: libgomp.info, Node: Runtime Library Routines, Next: Environment Variables, Prev: Enabling OpenMP, Up: Top 123 124 2 Runtime Library Routines 125 ************************** 126 127 The runtime routines described here are defined by Section 3 of the 128 OpenMP specification in version 4.5. The routines are structured in 129 following three parts: 130 131 * Menu: 132 133 Control threads, processors and the parallel environment. They have C 134 linkage, and do not throw exceptions. 135 136 * omp_get_active_level:: Number of active parallel regions 137 * omp_get_ancestor_thread_num:: Ancestor thread ID 138 * omp_get_cancellation:: Whether cancellation support is enabled 139 * omp_get_default_device:: Get the default device for target regions 140 * omp_get_dynamic:: Dynamic teams setting 141 * omp_get_level:: Number of parallel regions 142 * omp_get_max_active_levels:: Maximum number of active regions 143 * omp_get_max_task_priority:: Maximum task priority value that can be set 144 * omp_get_max_threads:: Maximum number of threads of parallel region 145 * omp_get_nested:: Nested parallel regions 146 * omp_get_num_devices:: Number of target devices 147 * omp_get_num_procs:: Number of processors online 148 * omp_get_num_teams:: Number of teams 149 * omp_get_num_threads:: Size of the active team 150 * omp_get_proc_bind:: Whether theads may be moved between CPUs 151 * omp_get_schedule:: Obtain the runtime scheduling method 152 * omp_get_team_num:: Get team number 153 * omp_get_team_size:: Number of threads in a team 154 * omp_get_thread_limit:: Maximum number of threads 155 * omp_get_thread_num:: Current thread ID 156 * omp_in_parallel:: Whether a parallel region is active 157 * omp_in_final:: Whether in final or included task region 158 * omp_is_initial_device:: Whether executing on the host device 159 * omp_set_default_device:: Set the default device for target regions 160 * omp_set_dynamic:: Enable/disable dynamic teams 161 * omp_set_max_active_levels:: Limits the number of active parallel regions 162 * omp_set_nested:: Enable/disable nested parallel regions 163 * omp_set_num_threads:: Set upper team size limit 164 * omp_set_schedule:: Set the runtime scheduling method 165 166 Initialize, set, test, unset and destroy simple and nested locks. 167 168 * omp_init_lock:: Initialize simple lock 169 * omp_set_lock:: Wait for and set simple lock 170 * omp_test_lock:: Test and set simple lock if available 171 * omp_unset_lock:: Unset simple lock 172 * omp_destroy_lock:: Destroy simple lock 173 * omp_init_nest_lock:: Initialize nested lock 174 * omp_set_nest_lock:: Wait for and set simple lock 175 * omp_test_nest_lock:: Test and set nested lock if available 176 * omp_unset_nest_lock:: Unset nested lock 177 * omp_destroy_nest_lock:: Destroy nested lock 178 179 Portable, thread-based, wall clock timer. 180 181 * omp_get_wtick:: Get timer precision. 182 * omp_get_wtime:: Elapsed wall clock time. 183 184 185 File: libgomp.info, Node: omp_get_active_level, Next: omp_get_ancestor_thread_num, Up: Runtime Library Routines 186 187 2.1 'omp_get_active_level' - Number of parallel regions 188 ======================================================= 189 190 _Description_: 191 This function returns the nesting level for the active parallel 192 blocks, which enclose the calling call. 193 194 _C/C++_ 195 _Prototype_: 'int omp_get_active_level(void);' 196 197 _Fortran_: 198 _Interface_: 'integer function omp_get_active_level()' 199 200 _See also_: 201 *note omp_get_level::, *note omp_get_max_active_levels::, *note 202 omp_set_max_active_levels:: 203 204 _Reference_: 205 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.20. 206 207 208 File: libgomp.info, Node: omp_get_ancestor_thread_num, Next: omp_get_cancellation, Prev: omp_get_active_level, Up: Runtime Library Routines 209 210 2.2 'omp_get_ancestor_thread_num' - Ancestor thread ID 211 ====================================================== 212 213 _Description_: 214 This function returns the thread identification number for the 215 given nesting level of the current thread. For values of LEVEL 216 outside zero to 'omp_get_level' -1 is returned; if LEVEL is 217 'omp_get_level' the result is identical to 'omp_get_thread_num'. 218 219 _C/C++_ 220 _Prototype_: 'int omp_get_ancestor_thread_num(int level);' 221 222 _Fortran_: 223 _Interface_: 'integer function omp_get_ancestor_thread_num(level)' 224 'integer level' 225 226 _See also_: 227 *note omp_get_level::, *note omp_get_thread_num::, *note 228 omp_get_team_size:: 229 230 _Reference_: 231 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.18. 232 233 234 File: libgomp.info, Node: omp_get_cancellation, Next: omp_get_default_device, Prev: omp_get_ancestor_thread_num, Up: Runtime Library Routines 235 236 2.3 'omp_get_cancellation' - Whether cancellation support is enabled 237 ==================================================================== 238 239 _Description_: 240 This function returns 'true' if cancellation is activated, 'false' 241 otherwise. Here, 'true' and 'false' represent their 242 language-specific counterparts. Unless 'OMP_CANCELLATION' is set 243 true, cancellations are deactivated. 244 245 _C/C++_: 246 _Prototype_: 'int omp_get_cancellation(void);' 247 248 _Fortran_: 249 _Interface_: 'logical function omp_get_cancellation()' 250 251 _See also_: 252 *note OMP_CANCELLATION:: 253 254 _Reference_: 255 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.9. 256 257 258 File: libgomp.info, Node: omp_get_default_device, Next: omp_get_dynamic, Prev: omp_get_cancellation, Up: Runtime Library Routines 259 260 2.4 'omp_get_default_device' - Get the default device for target regions 261 ======================================================================== 262 263 _Description_: 264 Get the default device for target regions without device clause. 265 266 _C/C++_: 267 _Prototype_: 'int omp_get_default_device(void);' 268 269 _Fortran_: 270 _Interface_: 'integer function omp_get_default_device()' 271 272 _See also_: 273 *note OMP_DEFAULT_DEVICE::, *note omp_set_default_device:: 274 275 _Reference_: 276 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.30. 277 278 279 File: libgomp.info, Node: omp_get_dynamic, Next: omp_get_level, Prev: omp_get_default_device, Up: Runtime Library Routines 280 281 2.5 'omp_get_dynamic' - Dynamic teams setting 282 ============================================= 283 284 _Description_: 285 This function returns 'true' if enabled, 'false' otherwise. Here, 286 'true' and 'false' represent their language-specific counterparts. 287 288 The dynamic team setting may be initialized at startup by the 289 'OMP_DYNAMIC' environment variable or at runtime using 290 'omp_set_dynamic'. If undefined, dynamic adjustment is disabled by 291 default. 292 293 _C/C++_: 294 _Prototype_: 'int omp_get_dynamic(void);' 295 296 _Fortran_: 297 _Interface_: 'logical function omp_get_dynamic()' 298 299 _See also_: 300 *note omp_set_dynamic::, *note OMP_DYNAMIC:: 301 302 _Reference_: 303 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.8. 304 305 306 File: libgomp.info, Node: omp_get_level, Next: omp_get_max_active_levels, Prev: omp_get_dynamic, Up: Runtime Library Routines 307 308 2.6 'omp_get_level' - Obtain the current nesting level 309 ====================================================== 310 311 _Description_: 312 This function returns the nesting level for the parallel blocks, 313 which enclose the calling call. 314 315 _C/C++_ 316 _Prototype_: 'int omp_get_level(void);' 317 318 _Fortran_: 319 _Interface_: 'integer function omp_level()' 320 321 _See also_: 322 *note omp_get_active_level:: 323 324 _Reference_: 325 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.17. 326 327 328 File: libgomp.info, Node: omp_get_max_active_levels, Next: omp_get_max_task_priority, Prev: omp_get_level, Up: Runtime Library Routines 329 330 2.7 'omp_get_max_active_levels' - Maximum number of active regions 331 ================================================================== 332 333 _Description_: 334 This function obtains the maximum allowed number of nested, active 335 parallel regions. 336 337 _C/C++_ 338 _Prototype_: 'int omp_get_max_active_levels(void);' 339 340 _Fortran_: 341 _Interface_: 'integer function omp_get_max_active_levels()' 342 343 _See also_: 344 *note omp_set_max_active_levels::, *note omp_get_active_level:: 345 346 _Reference_: 347 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.16. 348 349 350 File: libgomp.info, Node: omp_get_max_task_priority, Next: omp_get_max_threads, Prev: omp_get_max_active_levels, Up: Runtime Library Routines 351 352 2.8 'omp_get_max_task_priority' - Maximum priority value 353 ======================================================== 354 355 that can be set for tasks. 356 _Description_: 357 This function obtains the maximum allowed priority number for 358 tasks. 359 360 _C/C++_ 361 _Prototype_: 'int omp_get_max_task_priority(void);' 362 363 _Fortran_: 364 _Interface_: 'integer function omp_get_max_task_priority()' 365 366 _Reference_: 367 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.29. 368 369 370 File: libgomp.info, Node: omp_get_max_threads, Next: omp_get_nested, Prev: omp_get_max_task_priority, Up: Runtime Library Routines 371 372 2.9 'omp_get_max_threads' - Maximum number of threads of parallel region 373 ======================================================================== 374 375 _Description_: 376 Return the maximum number of threads used for the current parallel 377 region that does not use the clause 'num_threads'. 378 379 _C/C++_: 380 _Prototype_: 'int omp_get_max_threads(void);' 381 382 _Fortran_: 383 _Interface_: 'integer function omp_get_max_threads()' 384 385 _See also_: 386 *note omp_set_num_threads::, *note omp_set_dynamic::, *note 387 omp_get_thread_limit:: 388 389 _Reference_: 390 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.3. 391 392 393 File: libgomp.info, Node: omp_get_nested, Next: omp_get_num_devices, Prev: omp_get_max_threads, Up: Runtime Library Routines 394 395 2.10 'omp_get_nested' - Nested parallel regions 396 =============================================== 397 398 _Description_: 399 This function returns 'true' if nested parallel regions are 400 enabled, 'false' otherwise. Here, 'true' and 'false' represent 401 their language-specific counterparts. 402 403 Nested parallel regions may be initialized at startup by the 404 'OMP_NESTED' environment variable or at runtime using 405 'omp_set_nested'. If undefined, nested parallel regions are 406 disabled by default. 407 408 _C/C++_: 409 _Prototype_: 'int omp_get_nested(void);' 410 411 _Fortran_: 412 _Interface_: 'logical function omp_get_nested()' 413 414 _See also_: 415 *note omp_set_nested::, *note OMP_NESTED:: 416 417 _Reference_: 418 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.11. 419 420 421 File: libgomp.info, Node: omp_get_num_devices, Next: omp_get_num_procs, Prev: omp_get_nested, Up: Runtime Library Routines 422 423 2.11 'omp_get_num_devices' - Number of target devices 424 ===================================================== 425 426 _Description_: 427 Returns the number of target devices. 428 429 _C/C++_: 430 _Prototype_: 'int omp_get_num_devices(void);' 431 432 _Fortran_: 433 _Interface_: 'integer function omp_get_num_devices()' 434 435 _Reference_: 436 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.31. 437 438 439 File: libgomp.info, Node: omp_get_num_procs, Next: omp_get_num_teams, Prev: omp_get_num_devices, Up: Runtime Library Routines 440 441 2.12 'omp_get_num_procs' - Number of processors online 442 ====================================================== 443 444 _Description_: 445 Returns the number of processors online on that device. 446 447 _C/C++_: 448 _Prototype_: 'int omp_get_num_procs(void);' 449 450 _Fortran_: 451 _Interface_: 'integer function omp_get_num_procs()' 452 453 _Reference_: 454 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.5. 455 456 457 File: libgomp.info, Node: omp_get_num_teams, Next: omp_get_num_threads, Prev: omp_get_num_procs, Up: Runtime Library Routines 458 459 2.13 'omp_get_num_teams' - Number of teams 460 ========================================== 461 462 _Description_: 463 Returns the number of teams in the current team region. 464 465 _C/C++_: 466 _Prototype_: 'int omp_get_num_teams(void);' 467 468 _Fortran_: 469 _Interface_: 'integer function omp_get_num_teams()' 470 471 _Reference_: 472 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.32. 473 474 475 File: libgomp.info, Node: omp_get_num_threads, Next: omp_get_proc_bind, Prev: omp_get_num_teams, Up: Runtime Library Routines 476 477 2.14 'omp_get_num_threads' - Size of the active team 478 ==================================================== 479 480 _Description_: 481 Returns the number of threads in the current team. In a sequential 482 section of the program 'omp_get_num_threads' returns 1. 483 484 The default team size may be initialized at startup by the 485 'OMP_NUM_THREADS' environment variable. At runtime, the size of 486 the current team may be set either by the 'NUM_THREADS' clause or 487 by 'omp_set_num_threads'. If none of the above were used to define 488 a specific value and 'OMP_DYNAMIC' is disabled, one thread per CPU 489 online is used. 490 491 _C/C++_: 492 _Prototype_: 'int omp_get_num_threads(void);' 493 494 _Fortran_: 495 _Interface_: 'integer function omp_get_num_threads()' 496 497 _See also_: 498 *note omp_get_max_threads::, *note omp_set_num_threads::, *note 499 OMP_NUM_THREADS:: 500 501 _Reference_: 502 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.2. 503 504 505 File: libgomp.info, Node: omp_get_proc_bind, Next: omp_get_schedule, Prev: omp_get_num_threads, Up: Runtime Library Routines 506 507 2.15 'omp_get_proc_bind' - Whether theads may be moved between CPUs 508 =================================================================== 509 510 _Description_: 511 This functions returns the currently active thread affinity policy, 512 which is set via 'OMP_PROC_BIND'. Possible values are 513 'omp_proc_bind_false', 'omp_proc_bind_true', 514 'omp_proc_bind_master', 'omp_proc_bind_close' and 515 'omp_proc_bind_spread'. 516 517 _C/C++_: 518 _Prototype_: 'omp_proc_bind_t omp_get_proc_bind(void);' 519 520 _Fortran_: 521 _Interface_: 'integer(kind=omp_proc_bind_kind) function 522 omp_get_proc_bind()' 523 524 _See also_: 525 *note OMP_PROC_BIND::, *note OMP_PLACES::, *note 526 GOMP_CPU_AFFINITY::, 527 528 _Reference_: 529 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.22. 530 531 532 File: libgomp.info, Node: omp_get_schedule, Next: omp_get_team_num, Prev: omp_get_proc_bind, Up: Runtime Library Routines 533 534 2.16 'omp_get_schedule' - Obtain the runtime scheduling method 535 ============================================================== 536 537 _Description_: 538 Obtain the runtime scheduling method. The KIND argument will be 539 set to the value 'omp_sched_static', 'omp_sched_dynamic', 540 'omp_sched_guided' or 'omp_sched_auto'. The second argument, 541 CHUNK_SIZE, is set to the chunk size. 542 543 _C/C++_ 544 _Prototype_: 'void omp_get_schedule(omp_sched_t *kind, int 545 *chunk_size);' 546 547 _Fortran_: 548 _Interface_: 'subroutine omp_get_schedule(kind, chunk_size)' 549 'integer(kind=omp_sched_kind) kind' 550 'integer chunk_size' 551 552 _See also_: 553 *note omp_set_schedule::, *note OMP_SCHEDULE:: 554 555 _Reference_: 556 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.13. 557 558 559 File: libgomp.info, Node: omp_get_team_num, Next: omp_get_team_size, Prev: omp_get_schedule, Up: Runtime Library Routines 560 561 2.17 'omp_get_team_num' - Get team number 562 ========================================= 563 564 _Description_: 565 Returns the team number of the calling thread. 566 567 _C/C++_: 568 _Prototype_: 'int omp_get_team_num(void);' 569 570 _Fortran_: 571 _Interface_: 'integer function omp_get_team_num()' 572 573 _Reference_: 574 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.33. 575 576 577 File: libgomp.info, Node: omp_get_team_size, Next: omp_get_thread_limit, Prev: omp_get_team_num, Up: Runtime Library Routines 578 579 2.18 'omp_get_team_size' - Number of threads in a team 580 ====================================================== 581 582 _Description_: 583 This function returns the number of threads in a thread team to 584 which either the current thread or its ancestor belongs. For 585 values of LEVEL outside zero to 'omp_get_level', -1 is returned; if 586 LEVEL is zero, 1 is returned, and for 'omp_get_level', the result 587 is identical to 'omp_get_num_threads'. 588 589 _C/C++_: 590 _Prototype_: 'int omp_get_team_size(int level);' 591 592 _Fortran_: 593 _Interface_: 'integer function omp_get_team_size(level)' 594 'integer level' 595 596 _See also_: 597 *note omp_get_num_threads::, *note omp_get_level::, *note 598 omp_get_ancestor_thread_num:: 599 600 _Reference_: 601 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.19. 602 603 604 File: libgomp.info, Node: omp_get_thread_limit, Next: omp_get_thread_num, Prev: omp_get_team_size, Up: Runtime Library Routines 605 606 2.19 'omp_get_thread_limit' - Maximum number of threads 607 ======================================================= 608 609 _Description_: 610 Return the maximum number of threads of the program. 611 612 _C/C++_: 613 _Prototype_: 'int omp_get_thread_limit(void);' 614 615 _Fortran_: 616 _Interface_: 'integer function omp_get_thread_limit()' 617 618 _See also_: 619 *note omp_get_max_threads::, *note OMP_THREAD_LIMIT:: 620 621 _Reference_: 622 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.14. 623 624 625 File: libgomp.info, Node: omp_get_thread_num, Next: omp_in_parallel, Prev: omp_get_thread_limit, Up: Runtime Library Routines 626 627 2.20 'omp_get_thread_num' - Current thread ID 628 ============================================= 629 630 _Description_: 631 Returns a unique thread identification number within the current 632 team. In a sequential parts of the program, 'omp_get_thread_num' 633 always returns 0. In parallel regions the return value varies from 634 0 to 'omp_get_num_threads'-1 inclusive. The return value of the 635 master thread of a team is always 0. 636 637 _C/C++_: 638 _Prototype_: 'int omp_get_thread_num(void);' 639 640 _Fortran_: 641 _Interface_: 'integer function omp_get_thread_num()' 642 643 _See also_: 644 *note omp_get_num_threads::, *note omp_get_ancestor_thread_num:: 645 646 _Reference_: 647 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.4. 648 649 650 File: libgomp.info, Node: omp_in_parallel, Next: omp_in_final, Prev: omp_get_thread_num, Up: Runtime Library Routines 651 652 2.21 'omp_in_parallel' - Whether a parallel region is active 653 ============================================================ 654 655 _Description_: 656 This function returns 'true' if currently running in parallel, 657 'false' otherwise. Here, 'true' and 'false' represent their 658 language-specific counterparts. 659 660 _C/C++_: 661 _Prototype_: 'int omp_in_parallel(void);' 662 663 _Fortran_: 664 _Interface_: 'logical function omp_in_parallel()' 665 666 _Reference_: 667 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.6. 668 669 670 File: libgomp.info, Node: omp_in_final, Next: omp_is_initial_device, Prev: omp_in_parallel, Up: Runtime Library Routines 671 672 2.22 'omp_in_final' - Whether in final or included task region 673 ============================================================== 674 675 _Description_: 676 This function returns 'true' if currently running in a final or 677 included task region, 'false' otherwise. Here, 'true' and 'false' 678 represent their language-specific counterparts. 679 680 _C/C++_: 681 _Prototype_: 'int omp_in_final(void);' 682 683 _Fortran_: 684 _Interface_: 'logical function omp_in_final()' 685 686 _Reference_: 687 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.21. 688 689 690 File: libgomp.info, Node: omp_is_initial_device, Next: omp_set_default_device, Prev: omp_in_final, Up: Runtime Library Routines 691 692 2.23 'omp_is_initial_device' - Whether executing on the host device 693 =================================================================== 694 695 _Description_: 696 This function returns 'true' if currently running on the host 697 device, 'false' otherwise. Here, 'true' and 'false' represent 698 their language-specific counterparts. 699 700 _C/C++_: 701 _Prototype_: 'int omp_is_initial_device(void);' 702 703 _Fortran_: 704 _Interface_: 'logical function omp_is_initial_device()' 705 706 _Reference_: 707 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.34. 708 709 710 File: libgomp.info, Node: omp_set_default_device, Next: omp_set_dynamic, Prev: omp_is_initial_device, Up: Runtime Library Routines 711 712 2.24 'omp_set_default_device' - Set the default device for target regions 713 ========================================================================= 714 715 _Description_: 716 Set the default device for target regions without device clause. 717 The argument shall be a nonnegative device number. 718 719 _C/C++_: 720 _Prototype_: 'void omp_set_default_device(int device_num);' 721 722 _Fortran_: 723 _Interface_: 'subroutine omp_set_default_device(device_num)' 724 'integer device_num' 725 726 _See also_: 727 *note OMP_DEFAULT_DEVICE::, *note omp_get_default_device:: 728 729 _Reference_: 730 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.29. 731 732 733 File: libgomp.info, Node: omp_set_dynamic, Next: omp_set_max_active_levels, Prev: omp_set_default_device, Up: Runtime Library Routines 734 735 2.25 'omp_set_dynamic' - Enable/disable dynamic teams 736 ===================================================== 737 738 _Description_: 739 Enable or disable the dynamic adjustment of the number of threads 740 within a team. The function takes the language-specific equivalent 741 of 'true' and 'false', where 'true' enables dynamic adjustment of 742 team sizes and 'false' disables it. 743 744 _C/C++_: 745 _Prototype_: 'void omp_set_dynamic(int dynamic_threads);' 746 747 _Fortran_: 748 _Interface_: 'subroutine omp_set_dynamic(dynamic_threads)' 749 'logical, intent(in) :: dynamic_threads' 750 751 _See also_: 752 *note OMP_DYNAMIC::, *note omp_get_dynamic:: 753 754 _Reference_: 755 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.7. 756 757 758 File: libgomp.info, Node: omp_set_max_active_levels, Next: omp_set_nested, Prev: omp_set_dynamic, Up: Runtime Library Routines 759 760 2.26 'omp_set_max_active_levels' - Limits the number of active parallel regions 761 =============================================================================== 762 763 _Description_: 764 This function limits the maximum allowed number of nested, active 765 parallel regions. 766 767 _C/C++_ 768 _Prototype_: 'void omp_set_max_active_levels(int max_levels);' 769 770 _Fortran_: 771 _Interface_: 'subroutine omp_set_max_active_levels(max_levels)' 772 'integer max_levels' 773 774 _See also_: 775 *note omp_get_max_active_levels::, *note omp_get_active_level:: 776 777 _Reference_: 778 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.15. 779 780 781 File: libgomp.info, Node: omp_set_nested, Next: omp_set_num_threads, Prev: omp_set_max_active_levels, Up: Runtime Library Routines 782 783 2.27 'omp_set_nested' - Enable/disable nested parallel regions 784 ============================================================== 785 786 _Description_: 787 Enable or disable nested parallel regions, i.e., whether team 788 members are allowed to create new teams. The function takes the 789 language-specific equivalent of 'true' and 'false', where 'true' 790 enables dynamic adjustment of team sizes and 'false' disables it. 791 792 _C/C++_: 793 _Prototype_: 'void omp_set_nested(int nested);' 794 795 _Fortran_: 796 _Interface_: 'subroutine omp_set_nested(nested)' 797 'logical, intent(in) :: nested' 798 799 _See also_: 800 *note OMP_NESTED::, *note omp_get_nested:: 801 802 _Reference_: 803 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.10. 804 805 806 File: libgomp.info, Node: omp_set_num_threads, Next: omp_set_schedule, Prev: omp_set_nested, Up: Runtime Library Routines 807 808 2.28 'omp_set_num_threads' - Set upper team size limit 809 ====================================================== 810 811 _Description_: 812 Specifies the number of threads used by default in subsequent 813 parallel sections, if those do not specify a 'num_threads' clause. 814 The argument of 'omp_set_num_threads' shall be a positive integer. 815 816 _C/C++_: 817 _Prototype_: 'void omp_set_num_threads(int num_threads);' 818 819 _Fortran_: 820 _Interface_: 'subroutine omp_set_num_threads(num_threads)' 821 'integer, intent(in) :: num_threads' 822 823 _See also_: 824 *note OMP_NUM_THREADS::, *note omp_get_num_threads::, *note 825 omp_get_max_threads:: 826 827 _Reference_: 828 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.1. 829 830 831 File: libgomp.info, Node: omp_set_schedule, Next: omp_init_lock, Prev: omp_set_num_threads, Up: Runtime Library Routines 832 833 2.29 'omp_set_schedule' - Set the runtime scheduling method 834 =========================================================== 835 836 _Description_: 837 Sets the runtime scheduling method. The KIND argument can have the 838 value 'omp_sched_static', 'omp_sched_dynamic', 'omp_sched_guided' 839 or 'omp_sched_auto'. Except for 'omp_sched_auto', the chunk size 840 is set to the value of CHUNK_SIZE if positive, or to the default 841 value if zero or negative. For 'omp_sched_auto' the CHUNK_SIZE 842 argument is ignored. 843 844 _C/C++_ 845 _Prototype_: 'void omp_set_schedule(omp_sched_t kind, int 846 chunk_size);' 847 848 _Fortran_: 849 _Interface_: 'subroutine omp_set_schedule(kind, chunk_size)' 850 'integer(kind=omp_sched_kind) kind' 851 'integer chunk_size' 852 853 _See also_: 854 *note omp_get_schedule:: *note OMP_SCHEDULE:: 855 856 _Reference_: 857 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.2.12. 858 859 860 File: libgomp.info, Node: omp_init_lock, Next: omp_set_lock, Prev: omp_set_schedule, Up: Runtime Library Routines 861 862 2.30 'omp_init_lock' - Initialize simple lock 863 ============================================= 864 865 _Description_: 866 Initialize a simple lock. After initialization, the lock is in an 867 unlocked state. 868 869 _C/C++_: 870 _Prototype_: 'void omp_init_lock(omp_lock_t *lock);' 871 872 _Fortran_: 873 _Interface_: 'subroutine omp_init_lock(svar)' 874 'integer(omp_lock_kind), intent(out) :: svar' 875 876 _See also_: 877 *note omp_destroy_lock:: 878 879 _Reference_: 880 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.1. 881 882 883 File: libgomp.info, Node: omp_set_lock, Next: omp_test_lock, Prev: omp_init_lock, Up: Runtime Library Routines 884 885 2.31 'omp_set_lock' - Wait for and set simple lock 886 ================================================== 887 888 _Description_: 889 Before setting a simple lock, the lock variable must be initialized 890 by 'omp_init_lock'. The calling thread is blocked until the lock 891 is available. If the lock is already held by the current thread, a 892 deadlock occurs. 893 894 _C/C++_: 895 _Prototype_: 'void omp_set_lock(omp_lock_t *lock);' 896 897 _Fortran_: 898 _Interface_: 'subroutine omp_set_lock(svar)' 899 'integer(omp_lock_kind), intent(inout) :: svar' 900 901 _See also_: 902 *note omp_init_lock::, *note omp_test_lock::, *note 903 omp_unset_lock:: 904 905 _Reference_: 906 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.4. 907 908 909 File: libgomp.info, Node: omp_test_lock, Next: omp_unset_lock, Prev: omp_set_lock, Up: Runtime Library Routines 910 911 2.32 'omp_test_lock' - Test and set simple lock if available 912 ============================================================ 913 914 _Description_: 915 Before setting a simple lock, the lock variable must be initialized 916 by 'omp_init_lock'. Contrary to 'omp_set_lock', 'omp_test_lock' 917 does not block if the lock is not available. This function returns 918 'true' upon success, 'false' otherwise. Here, 'true' and 'false' 919 represent their language-specific counterparts. 920 921 _C/C++_: 922 _Prototype_: 'int omp_test_lock(omp_lock_t *lock);' 923 924 _Fortran_: 925 _Interface_: 'logical function omp_test_lock(svar)' 926 'integer(omp_lock_kind), intent(inout) :: svar' 927 928 _See also_: 929 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 930 931 _Reference_: 932 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.6. 933 934 935 File: libgomp.info, Node: omp_unset_lock, Next: omp_destroy_lock, Prev: omp_test_lock, Up: Runtime Library Routines 936 937 2.33 'omp_unset_lock' - Unset simple lock 938 ========================================= 939 940 _Description_: 941 A simple lock about to be unset must have been locked by 942 'omp_set_lock' or 'omp_test_lock' before. In addition, the lock 943 must be held by the thread calling 'omp_unset_lock'. Then, the 944 lock becomes unlocked. If one or more threads attempted to set the 945 lock before, one of them is chosen to, again, set the lock to 946 itself. 947 948 _C/C++_: 949 _Prototype_: 'void omp_unset_lock(omp_lock_t *lock);' 950 951 _Fortran_: 952 _Interface_: 'subroutine omp_unset_lock(svar)' 953 'integer(omp_lock_kind), intent(inout) :: svar' 954 955 _See also_: 956 *note omp_set_lock::, *note omp_test_lock:: 957 958 _Reference_: 959 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.5. 960 961 962 File: libgomp.info, Node: omp_destroy_lock, Next: omp_init_nest_lock, Prev: omp_unset_lock, Up: Runtime Library Routines 963 964 2.34 'omp_destroy_lock' - Destroy simple lock 965 ============================================= 966 967 _Description_: 968 Destroy a simple lock. In order to be destroyed, a simple lock 969 must be in the unlocked state. 970 971 _C/C++_: 972 _Prototype_: 'void omp_destroy_lock(omp_lock_t *lock);' 973 974 _Fortran_: 975 _Interface_: 'subroutine omp_destroy_lock(svar)' 976 'integer(omp_lock_kind), intent(inout) :: svar' 977 978 _See also_: 979 *note omp_init_lock:: 980 981 _Reference_: 982 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.3. 983 984 985 File: libgomp.info, Node: omp_init_nest_lock, Next: omp_set_nest_lock, Prev: omp_destroy_lock, Up: Runtime Library Routines 986 987 2.35 'omp_init_nest_lock' - Initialize nested lock 988 ================================================== 989 990 _Description_: 991 Initialize a nested lock. After initialization, the lock is in an 992 unlocked state and the nesting count is set to zero. 993 994 _C/C++_: 995 _Prototype_: 'void omp_init_nest_lock(omp_nest_lock_t *lock);' 996 997 _Fortran_: 998 _Interface_: 'subroutine omp_init_nest_lock(nvar)' 999 'integer(omp_nest_lock_kind), intent(out) :: nvar' 1000 1001 _See also_: 1002 *note omp_destroy_nest_lock:: 1003 1004 _Reference_: 1005 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.1. 1006 1007 1008 File: libgomp.info, Node: omp_set_nest_lock, Next: omp_test_nest_lock, Prev: omp_init_nest_lock, Up: Runtime Library Routines 1009 1010 2.36 'omp_set_nest_lock' - Wait for and set nested lock 1011 ======================================================= 1012 1013 _Description_: 1014 Before setting a nested lock, the lock variable must be initialized 1015 by 'omp_init_nest_lock'. The calling thread is blocked until the 1016 lock is available. If the lock is already held by the current 1017 thread, the nesting count for the lock is incremented. 1018 1019 _C/C++_: 1020 _Prototype_: 'void omp_set_nest_lock(omp_nest_lock_t *lock);' 1021 1022 _Fortran_: 1023 _Interface_: 'subroutine omp_set_nest_lock(nvar)' 1024 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1025 1026 _See also_: 1027 *note omp_init_nest_lock::, *note omp_unset_nest_lock:: 1028 1029 _Reference_: 1030 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.4. 1031 1032 1033 File: libgomp.info, Node: omp_test_nest_lock, Next: omp_unset_nest_lock, Prev: omp_set_nest_lock, Up: Runtime Library Routines 1034 1035 2.37 'omp_test_nest_lock' - Test and set nested lock if available 1036 ================================================================= 1037 1038 _Description_: 1039 Before setting a nested lock, the lock variable must be initialized 1040 by 'omp_init_nest_lock'. Contrary to 'omp_set_nest_lock', 1041 'omp_test_nest_lock' does not block if the lock is not available. 1042 If the lock is already held by the current thread, the new nesting 1043 count is returned. Otherwise, the return value equals zero. 1044 1045 _C/C++_: 1046 _Prototype_: 'int omp_test_nest_lock(omp_nest_lock_t *lock);' 1047 1048 _Fortran_: 1049 _Interface_: 'logical function omp_test_nest_lock(nvar)' 1050 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1051 1052 _See also_: 1053 *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 1054 1055 _Reference_: 1056 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.6. 1057 1058 1059 File: libgomp.info, Node: omp_unset_nest_lock, Next: omp_destroy_nest_lock, Prev: omp_test_nest_lock, Up: Runtime Library Routines 1060 1061 2.38 'omp_unset_nest_lock' - Unset nested lock 1062 ============================================== 1063 1064 _Description_: 1065 A nested lock about to be unset must have been locked by 1066 'omp_set_nested_lock' or 'omp_test_nested_lock' before. In 1067 addition, the lock must be held by the thread calling 1068 'omp_unset_nested_lock'. If the nesting count drops to zero, the 1069 lock becomes unlocked. If one ore more threads attempted to set 1070 the lock before, one of them is chosen to, again, set the lock to 1071 itself. 1072 1073 _C/C++_: 1074 _Prototype_: 'void omp_unset_nest_lock(omp_nest_lock_t *lock);' 1075 1076 _Fortran_: 1077 _Interface_: 'subroutine omp_unset_nest_lock(nvar)' 1078 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1079 1080 _See also_: 1081 *note omp_set_nest_lock:: 1082 1083 _Reference_: 1084 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.5. 1085 1086 1087 File: libgomp.info, Node: omp_destroy_nest_lock, Next: omp_get_wtick, Prev: omp_unset_nest_lock, Up: Runtime Library Routines 1088 1089 2.39 'omp_destroy_nest_lock' - Destroy nested lock 1090 ================================================== 1091 1092 _Description_: 1093 Destroy a nested lock. In order to be destroyed, a nested lock 1094 must be in the unlocked state and its nesting count must equal 1095 zero. 1096 1097 _C/C++_: 1098 _Prototype_: 'void omp_destroy_nest_lock(omp_nest_lock_t *);' 1099 1100 _Fortran_: 1101 _Interface_: 'subroutine omp_destroy_nest_lock(nvar)' 1102 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1103 1104 _See also_: 1105 *note omp_init_lock:: 1106 1107 _Reference_: 1108 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.3.3. 1109 1110 1111 File: libgomp.info, Node: omp_get_wtick, Next: omp_get_wtime, Prev: omp_destroy_nest_lock, Up: Runtime Library Routines 1112 1113 2.40 'omp_get_wtick' - Get timer precision 1114 ========================================== 1115 1116 _Description_: 1117 Gets the timer precision, i.e., the number of seconds between two 1118 successive clock ticks. 1119 1120 _C/C++_: 1121 _Prototype_: 'double omp_get_wtick(void);' 1122 1123 _Fortran_: 1124 _Interface_: 'double precision function omp_get_wtick()' 1125 1126 _See also_: 1127 *note omp_get_wtime:: 1128 1129 _Reference_: 1130 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.4.2. 1131 1132 1133 File: libgomp.info, Node: omp_get_wtime, Prev: omp_get_wtick, Up: Runtime Library Routines 1134 1135 2.41 'omp_get_wtime' - Elapsed wall clock time 1136 ============================================== 1137 1138 _Description_: 1139 Elapsed wall clock time in seconds. The time is measured per 1140 thread, no guarantee can be made that two distinct threads measure 1141 the same time. Time is measured from some "time in the past", 1142 which is an arbitrary time guaranteed not to change during the 1143 execution of the program. 1144 1145 _C/C++_: 1146 _Prototype_: 'double omp_get_wtime(void);' 1147 1148 _Fortran_: 1149 _Interface_: 'double precision function omp_get_wtime()' 1150 1151 _See also_: 1152 *note omp_get_wtick:: 1153 1154 _Reference_: 1155 OpenMP specification v4.5 (http://www.openmp.org/), Section 3.4.1. 1156 1157 1158 File: libgomp.info, Node: Environment Variables, Next: Enabling OpenACC, Prev: Runtime Library Routines, Up: Top 1159 1160 3 Environment Variables 1161 *********************** 1162 1163 The environment variables which beginning with 'OMP_' are defined by 1164 section 4 of the OpenMP specification in version 4.5, while those 1165 beginning with 'GOMP_' are GNU extensions. 1166 1167 * Menu: 1168 1169 * OMP_CANCELLATION:: Set whether cancellation is activated 1170 * OMP_DISPLAY_ENV:: Show OpenMP version and environment variables 1171 * OMP_DEFAULT_DEVICE:: Set the device used in target regions 1172 * OMP_DYNAMIC:: Dynamic adjustment of threads 1173 * OMP_MAX_ACTIVE_LEVELS:: Set the maximum number of nested parallel regions 1174 * OMP_MAX_TASK_PRIORITY:: Set the maximum task priority value 1175 * OMP_NESTED:: Nested parallel regions 1176 * OMP_NUM_THREADS:: Specifies the number of threads to use 1177 * OMP_PROC_BIND:: Whether theads may be moved between CPUs 1178 * OMP_PLACES:: Specifies on which CPUs the theads should be placed 1179 * OMP_STACKSIZE:: Set default thread stack size 1180 * OMP_SCHEDULE:: How threads are scheduled 1181 * OMP_THREAD_LIMIT:: Set the maximum number of threads 1182 * OMP_WAIT_POLICY:: How waiting threads are handled 1183 * GOMP_CPU_AFFINITY:: Bind threads to specific CPUs 1184 * GOMP_DEBUG:: Enable debugging output 1185 * GOMP_STACKSIZE:: Set default thread stack size 1186 * GOMP_SPINCOUNT:: Set the busy-wait spin count 1187 * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools 1188 1189 1190 File: libgomp.info, Node: OMP_CANCELLATION, Next: OMP_DISPLAY_ENV, Up: Environment Variables 1191 1192 3.1 'OMP_CANCELLATION' - Set whether cancellation is activated 1193 ============================================================== 1194 1195 _Description_: 1196 If set to 'TRUE', the cancellation is activated. If set to 'FALSE' 1197 or if unset, cancellation is disabled and the 'cancel' construct is 1198 ignored. 1199 1200 _See also_: 1201 *note omp_get_cancellation:: 1202 1203 _Reference_: 1204 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.11 1205 1206 1207 File: libgomp.info, Node: OMP_DISPLAY_ENV, Next: OMP_DEFAULT_DEVICE, Prev: OMP_CANCELLATION, Up: Environment Variables 1208 1209 3.2 'OMP_DISPLAY_ENV' - Show OpenMP version and environment variables 1210 ===================================================================== 1211 1212 _Description_: 1213 If set to 'TRUE', the OpenMP version number and the values 1214 associated with the OpenMP environment variables are printed to 1215 'stderr'. If set to 'VERBOSE', it additionally shows the value of 1216 the environment variables which are GNU extensions. If undefined 1217 or set to 'FALSE', this information will not be shown. 1218 1219 _Reference_: 1220 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.12 1221 1222 1223 File: libgomp.info, Node: OMP_DEFAULT_DEVICE, Next: OMP_DYNAMIC, Prev: OMP_DISPLAY_ENV, Up: Environment Variables 1224 1225 3.3 'OMP_DEFAULT_DEVICE' - Set the device used in target regions 1226 ================================================================ 1227 1228 _Description_: 1229 Set to choose the device which is used in a 'target' region, unless 1230 the value is overridden by 'omp_set_default_device' or by a 1231 'device' clause. The value shall be the nonnegative device number. 1232 If no device with the given device number exists, the code is 1233 executed on the host. If unset, device number 0 will be used. 1234 1235 _See also_: 1236 *note omp_get_default_device::, *note omp_set_default_device::, 1237 1238 _Reference_: 1239 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.13 1240 1241 1242 File: libgomp.info, Node: OMP_DYNAMIC, Next: OMP_MAX_ACTIVE_LEVELS, Prev: OMP_DEFAULT_DEVICE, Up: Environment Variables 1243 1244 3.4 'OMP_DYNAMIC' - Dynamic adjustment of threads 1245 ================================================= 1246 1247 _Description_: 1248 Enable or disable the dynamic adjustment of the number of threads 1249 within a team. The value of this environment variable shall be 1250 'TRUE' or 'FALSE'. If undefined, dynamic adjustment is disabled by 1251 default. 1252 1253 _See also_: 1254 *note omp_set_dynamic:: 1255 1256 _Reference_: 1257 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.3 1258 1259 1260 File: libgomp.info, Node: OMP_MAX_ACTIVE_LEVELS, Next: OMP_MAX_TASK_PRIORITY, Prev: OMP_DYNAMIC, Up: Environment Variables 1261 1262 3.5 'OMP_MAX_ACTIVE_LEVELS' - Set the maximum number of nested parallel regions 1263 =============================================================================== 1264 1265 _Description_: 1266 Specifies the initial value for the maximum number of nested 1267 parallel regions. The value of this variable shall be a positive 1268 integer. If undefined, the number of active levels is unlimited. 1269 1270 _See also_: 1271 *note omp_set_max_active_levels:: 1272 1273 _Reference_: 1274 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.9 1275 1276 1277 File: libgomp.info, Node: OMP_MAX_TASK_PRIORITY, Next: OMP_NESTED, Prev: OMP_MAX_ACTIVE_LEVELS, Up: Environment Variables 1278 1279 3.6 'OMP_MAX_TASK_PRIORITY' - Set the maximum priority 1280 ====================================================== 1281 1282 number that can be set for a task. 1283 _Description_: 1284 Specifies the initial value for the maximum priority value that can 1285 be set for a task. The value of this variable shall be a 1286 non-negative integer, and zero is allowed. If undefined, the 1287 default priority is 0. 1288 1289 _See also_: 1290 *note omp_get_max_task_priority:: 1291 1292 _Reference_: 1293 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.14 1294 1295 1296 File: libgomp.info, Node: OMP_NESTED, Next: OMP_NUM_THREADS, Prev: OMP_MAX_TASK_PRIORITY, Up: Environment Variables 1297 1298 3.7 'OMP_NESTED' - Nested parallel regions 1299 ========================================== 1300 1301 _Description_: 1302 Enable or disable nested parallel regions, i.e., whether team 1303 members are allowed to create new teams. The value of this 1304 environment variable shall be 'TRUE' or 'FALSE'. If undefined, 1305 nested parallel regions are disabled by default. 1306 1307 _See also_: 1308 *note omp_set_nested:: 1309 1310 _Reference_: 1311 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.6 1312 1313 1314 File: libgomp.info, Node: OMP_NUM_THREADS, Next: OMP_PROC_BIND, Prev: OMP_NESTED, Up: Environment Variables 1315 1316 3.8 'OMP_NUM_THREADS' - Specifies the number of threads to use 1317 ============================================================== 1318 1319 _Description_: 1320 Specifies the default number of threads to use in parallel regions. 1321 The value of this variable shall be a comma-separated list of 1322 positive integers; the value specified the number of threads to use 1323 for the corresponding nested level. If undefined one thread per 1324 CPU is used. 1325 1326 _See also_: 1327 *note omp_set_num_threads:: 1328 1329 _Reference_: 1330 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.2 1331 1332 1333 File: libgomp.info, Node: OMP_PROC_BIND, Next: OMP_PLACES, Prev: OMP_NUM_THREADS, Up: Environment Variables 1334 1335 3.9 'OMP_PROC_BIND' - Whether theads may be moved between CPUs 1336 ============================================================== 1337 1338 _Description_: 1339 Specifies whether threads may be moved between processors. If set 1340 to 'TRUE', OpenMP theads should not be moved; if set to 'FALSE' 1341 they may be moved. Alternatively, a comma separated list with the 1342 values 'MASTER', 'CLOSE' and 'SPREAD' can be used to specify the 1343 thread affinity policy for the corresponding nesting level. With 1344 'MASTER' the worker threads are in the same place partition as the 1345 master thread. With 'CLOSE' those are kept close to the master 1346 thread in contiguous place partitions. And with 'SPREAD' a sparse 1347 distribution across the place partitions is used. 1348 1349 When undefined, 'OMP_PROC_BIND' defaults to 'TRUE' when 1350 'OMP_PLACES' or 'GOMP_CPU_AFFINITY' is set and 'FALSE' otherwise. 1351 1352 _See also_: 1353 *note OMP_PLACES::, *note GOMP_CPU_AFFINITY::, *note 1354 omp_get_proc_bind:: 1355 1356 _Reference_: 1357 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.4 1358 1359 1360 File: libgomp.info, Node: OMP_PLACES, Next: OMP_STACKSIZE, Prev: OMP_PROC_BIND, Up: Environment Variables 1361 1362 3.10 'OMP_PLACES' - Specifies on which CPUs the theads should be placed 1363 ======================================================================= 1364 1365 _Description_: 1366 The thread placement can be either specified using an abstract name 1367 or by an explicit list of the places. The abstract names 1368 'threads', 'cores' and 'sockets' can be optionally followed by a 1369 positive number in parentheses, which denotes the how many places 1370 shall be created. With 'threads' each place corresponds to a 1371 single hardware thread; 'cores' to a single core with the 1372 corresponding number of hardware threads; and with 'sockets' the 1373 place corresponds to a single socket. The resulting placement can 1374 be shown by setting the 'OMP_DISPLAY_ENV' environment variable. 1375 1376 Alternatively, the placement can be specified explicitly as 1377 comma-separated list of places. A place is specified by set of 1378 nonnegative numbers in curly braces, denoting the denoting the 1379 hardware threads. The hardware threads belonging to a place can 1380 either be specified as comma-separated list of nonnegative thread 1381 numbers or using an interval. Multiple places can also be either 1382 specified by a comma-separated list of places or by an interval. 1383 To specify an interval, a colon followed by the count is placed 1384 after after the hardware thread number or the place. Optionally, 1385 the length can be followed by a colon and the stride number - 1386 otherwise a unit stride is assumed. For instance, the following 1387 specifies the same places list: '"{0,1,2}, {3,4,6}, {7,8,9}, 1388 {10,11,12}"'; '"{0:3}, {3:3}, {7:3}, {10:3}"'; and '"{0:2}:4:3"'. 1389 1390 If 'OMP_PLACES' and 'GOMP_CPU_AFFINITY' are unset and 1391 'OMP_PROC_BIND' is either unset or 'false', threads may be moved 1392 between CPUs following no placement policy. 1393 1394 _See also_: 1395 *note OMP_PROC_BIND::, *note GOMP_CPU_AFFINITY::, *note 1396 omp_get_proc_bind::, *note OMP_DISPLAY_ENV:: 1397 1398 _Reference_: 1399 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.5 1400 1401 1402 File: libgomp.info, Node: OMP_STACKSIZE, Next: OMP_SCHEDULE, Prev: OMP_PLACES, Up: Environment Variables 1403 1404 3.11 'OMP_STACKSIZE' - Set default thread stack size 1405 ==================================================== 1406 1407 _Description_: 1408 Set the default thread stack size in kilobytes, unless the number 1409 is suffixed by 'B', 'K', 'M' or 'G', in which case the size is, 1410 respectively, in bytes, kilobytes, megabytes or gigabytes. This is 1411 different from 'pthread_attr_setstacksize' which gets the number of 1412 bytes as an argument. If the stack size cannot be set due to 1413 system constraints, an error is reported and the initial stack size 1414 is left unchanged. If undefined, the stack size is system 1415 dependent. 1416 1417 _Reference_: 1418 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.7 1419 1420 1421 File: libgomp.info, Node: OMP_SCHEDULE, Next: OMP_THREAD_LIMIT, Prev: OMP_STACKSIZE, Up: Environment Variables 1422 1423 3.12 'OMP_SCHEDULE' - How threads are scheduled 1424 =============================================== 1425 1426 _Description_: 1427 Allows to specify 'schedule type' and 'chunk size'. The value of 1428 the variable shall have the form: 'type[,chunk]' where 'type' is 1429 one of 'static', 'dynamic', 'guided' or 'auto' The optional 'chunk' 1430 size shall be a positive integer. If undefined, dynamic scheduling 1431 and a chunk size of 1 is used. 1432 1433 _See also_: 1434 *note omp_set_schedule:: 1435 1436 _Reference_: 1437 OpenMP specification v4.5 (http://www.openmp.org/), Sections 1438 2.7.1.1 and 4.1 1439 1440 1441 File: libgomp.info, Node: OMP_THREAD_LIMIT, Next: OMP_WAIT_POLICY, Prev: OMP_SCHEDULE, Up: Environment Variables 1442 1443 3.13 'OMP_THREAD_LIMIT' - Set the maximum number of threads 1444 =========================================================== 1445 1446 _Description_: 1447 Specifies the number of threads to use for the whole program. The 1448 value of this variable shall be a positive integer. If undefined, 1449 the number of threads is not limited. 1450 1451 _See also_: 1452 *note OMP_NUM_THREADS::, *note omp_get_thread_limit:: 1453 1454 _Reference_: 1455 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.10 1456 1457 1458 File: libgomp.info, Node: OMP_WAIT_POLICY, Next: GOMP_CPU_AFFINITY, Prev: OMP_THREAD_LIMIT, Up: Environment Variables 1459 1460 3.14 'OMP_WAIT_POLICY' - How waiting threads are handled 1461 ======================================================== 1462 1463 _Description_: 1464 Specifies whether waiting threads should be active or passive. If 1465 the value is 'PASSIVE', waiting threads should not consume CPU 1466 power while waiting; while the value is 'ACTIVE' specifies that 1467 they should. If undefined, threads wait actively for a short time 1468 before waiting passively. 1469 1470 _See also_: 1471 *note GOMP_SPINCOUNT:: 1472 1473 _Reference_: 1474 OpenMP specification v4.5 (http://www.openmp.org/), Section 4.8 1475 1476 1477 File: libgomp.info, Node: GOMP_CPU_AFFINITY, Next: GOMP_DEBUG, Prev: OMP_WAIT_POLICY, Up: Environment Variables 1478 1479 3.15 'GOMP_CPU_AFFINITY' - Bind threads to specific CPUs 1480 ======================================================== 1481 1482 _Description_: 1483 Binds threads to specific CPUs. The variable should contain a 1484 space-separated or comma-separated list of CPUs. This list may 1485 contain different kinds of entries: either single CPU numbers in 1486 any order, a range of CPUs (M-N) or a range with some stride 1487 (M-N:S). CPU numbers are zero based. For example, 1488 'GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"' will bind the initial thread 1489 to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to 1490 CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 1491 10, 12, and 14 respectively and then start assigning back from the 1492 beginning of the list. 'GOMP_CPU_AFFINITY=0' binds all threads to 1493 CPU 0. 1494 1495 There is no libgomp library routine to determine whether a CPU 1496 affinity specification is in effect. As a workaround, 1497 language-specific library functions, e.g., 'getenv' in C or 1498 'GET_ENVIRONMENT_VARIABLE' in Fortran, may be used to query the 1499 setting of the 'GOMP_CPU_AFFINITY' environment variable. A defined 1500 CPU affinity on startup cannot be changed or disabled during the 1501 runtime of the application. 1502 1503 If both 'GOMP_CPU_AFFINITY' and 'OMP_PROC_BIND' are set, 1504 'OMP_PROC_BIND' has a higher precedence. If neither has been set 1505 and 'OMP_PROC_BIND' is unset, or when 'OMP_PROC_BIND' is set to 1506 'FALSE', the host system will handle the assignment of threads to 1507 CPUs. 1508 1509 _See also_: 1510 *note OMP_PLACES::, *note OMP_PROC_BIND:: 1511 1512 1513 File: libgomp.info, Node: GOMP_DEBUG, Next: GOMP_STACKSIZE, Prev: GOMP_CPU_AFFINITY, Up: Environment Variables 1514 1515 3.16 'GOMP_DEBUG' - Enable debugging output 1516 =========================================== 1517 1518 _Description_: 1519 Enable debugging output. The variable should be set to '0' 1520 (disabled, also the default if not set), or '1' (enabled). 1521 1522 If enabled, some debugging output will be printed during execution. 1523 This is currently not specified in more detail, and subject to 1524 change. 1525 1526 1527 File: libgomp.info, Node: GOMP_STACKSIZE, Next: GOMP_SPINCOUNT, Prev: GOMP_DEBUG, Up: Environment Variables 1528 1529 3.17 'GOMP_STACKSIZE' - Set default thread stack size 1530 ===================================================== 1531 1532 _Description_: 1533 Set the default thread stack size in kilobytes. This is different 1534 from 'pthread_attr_setstacksize' which gets the number of bytes as 1535 an argument. If the stack size cannot be set due to system 1536 constraints, an error is reported and the initial stack size is 1537 left unchanged. If undefined, the stack size is system dependent. 1538 1539 _See also_: 1540 *note OMP_STACKSIZE:: 1541 1542 _Reference_: 1543 GCC Patches Mailinglist 1544 (http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html), GCC 1545 Patches Mailinglist 1546 (http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html) 1547 1548 1549 File: libgomp.info, Node: GOMP_SPINCOUNT, Next: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_STACKSIZE, Up: Environment Variables 1550 1551 3.18 'GOMP_SPINCOUNT' - Set the busy-wait spin count 1552 ==================================================== 1553 1554 _Description_: 1555 Determines how long a threads waits actively with consuming CPU 1556 power before waiting passively without consuming CPU power. The 1557 value may be either 'INFINITE', 'INFINITY' to always wait actively 1558 or an integer which gives the number of spins of the busy-wait 1559 loop. The integer may optionally be followed by the following 1560 suffixes acting as multiplication factors: 'k' (kilo, thousand), 1561 'M' (mega, million), 'G' (giga, billion), or 'T' (tera, trillion). 1562 If undefined, 0 is used when 'OMP_WAIT_POLICY' is 'PASSIVE', 1563 300,000 is used when 'OMP_WAIT_POLICY' is undefined and 30 billion 1564 is used when 'OMP_WAIT_POLICY' is 'ACTIVE'. If there are more 1565 OpenMP threads than available CPUs, 1000 and 100 spins are used for 1566 'OMP_WAIT_POLICY' being 'ACTIVE' or undefined, respectively; unless 1567 the 'GOMP_SPINCOUNT' is lower or 'OMP_WAIT_POLICY' is 'PASSIVE'. 1568 1569 _See also_: 1570 *note OMP_WAIT_POLICY:: 1571 1572 1573 File: libgomp.info, Node: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_SPINCOUNT, Up: Environment Variables 1574 1575 3.19 'GOMP_RTEMS_THREAD_POOLS' - Set the RTEMS specific thread pools 1576 ==================================================================== 1577 1578 _Description_: 1579 This environment variable is only used on the RTEMS real-time 1580 operating system. It determines the scheduler instance specific 1581 thread pools. The format for 'GOMP_RTEMS_THREAD_POOLS' is a list 1582 of optional '<thread-pool-count>[$<priority>]@<scheduler-name>' 1583 configurations separated by ':' where: 1584 * '<thread-pool-count>' is the thread pool count for this 1585 scheduler instance. 1586 * '$<priority>' is an optional priority for the worker threads 1587 of a thread pool according to 'pthread_setschedparam'. In 1588 case a priority value is omitted, then a worker thread will 1589 inherit the priority of the OpenMP master thread that created 1590 it. The priority of the worker thread is not changed after 1591 creation, even if a new OpenMP master thread using the worker 1592 has a different priority. 1593 * '@<scheduler-name>' is the scheduler instance name according 1594 to the RTEMS application configuration. 1595 In case no thread pool configuration is specified for a scheduler 1596 instance, then each OpenMP master thread of this scheduler instance 1597 will use its own dynamically allocated thread pool. To limit the 1598 worker thread count of the thread pools, each OpenMP master thread 1599 must call 'omp_set_num_threads'. 1600 _Example_: 1601 Lets suppose we have three scheduler instances 'IO', 'WRK0', and 1602 'WRK1' with 'GOMP_RTEMS_THREAD_POOLS' set to '"1@WRK0:3$4@WRK1"'. 1603 Then there are no thread pool restrictions for scheduler instance 1604 'IO'. In the scheduler instance 'WRK0' there is one thread pool 1605 available. Since no priority is specified for this scheduler 1606 instance, the worker thread inherits the priority of the OpenMP 1607 master thread that created it. In the scheduler instance 'WRK1' 1608 there are three thread pools available and their worker threads run 1609 at priority four. 1610 1611 1612 File: libgomp.info, Node: Enabling OpenACC, Next: OpenACC Runtime Library Routines, Prev: Environment Variables, Up: Top 1613 1614 4 Enabling OpenACC 1615 ****************** 1616 1617 To activate the OpenACC extensions for C/C++ and Fortran, the 1618 compile-time flag '-fopenacc' must be specified. This enables the 1619 OpenACC directive '#pragma acc' in C/C++ and '!$accp' directives in free 1620 form, 'c$acc', '*$acc' and '!$acc' directives in fixed form, '!$' 1621 conditional compilation sentinels in free form and 'c$', '*$' and '!$' 1622 sentinels in fixed form, for Fortran. The flag also arranges for 1623 automatic linking of the OpenACC runtime library (*note OpenACC Runtime 1624 Library Routines::). 1625 1626 A complete description of all OpenACC directives accepted may be 1627 found in the OpenACC (https://www.openacc.org) Application Programming 1628 Interface manual, version 2.0. 1629 1630 Note that this is an experimental feature and subject to change in 1631 future versions of GCC. See <https://gcc.gnu.org/wiki/OpenACC> for more 1632 information. 1633 1634 1635 File: libgomp.info, Node: OpenACC Runtime Library Routines, Next: OpenACC Environment Variables, Prev: Enabling OpenACC, Up: Top 1636 1637 5 OpenACC Runtime Library Routines 1638 ********************************** 1639 1640 The runtime routines described here are defined by section 3 of the 1641 OpenACC specifications in version 2.0. They have C linkage, and do not 1642 throw exceptions. Generally, they are available only for the host, with 1643 the exception of 'acc_on_device', which is available for both the host 1644 and the acceleration device. 1645 1646 * Menu: 1647 1648 * acc_get_num_devices:: Get number of devices for the given device 1649 type. 1650 * acc_set_device_type:: Set type of device accelerator to use. 1651 * acc_get_device_type:: Get type of device accelerator to be used. 1652 * acc_set_device_num:: Set device number to use. 1653 * acc_get_device_num:: Get device number to be used. 1654 * acc_async_test:: Tests for completion of a specific asynchronous 1655 operation. 1656 * acc_async_test_all:: Tests for completion of all asychronous 1657 operations. 1658 * acc_wait:: Wait for completion of a specific asynchronous 1659 operation. 1660 * acc_wait_all:: Waits for completion of all asyncrhonous 1661 operations. 1662 * acc_wait_all_async:: Wait for completion of all asynchronous 1663 operations. 1664 * acc_wait_async:: Wait for completion of asynchronous operations. 1665 * acc_init:: Initialize runtime for a specific device type. 1666 * acc_shutdown:: Shuts down the runtime for a specific device 1667 type. 1668 * acc_on_device:: Whether executing on a particular device 1669 * acc_malloc:: Allocate device memory. 1670 * acc_free:: Free device memory. 1671 * acc_copyin:: Allocate device memory and copy host memory to 1672 it. 1673 * acc_present_or_copyin:: If the data is not present on the device, 1674 allocate device memory and copy from host 1675 memory. 1676 * acc_create:: Allocate device memory and map it to host 1677 memory. 1678 * acc_present_or_create:: If the data is not present on the device, 1679 allocate device memory and map it to host 1680 memory. 1681 * acc_copyout:: Copy device memory to host memory. 1682 * acc_delete:: Free device memory. 1683 * acc_update_device:: Update device memory from mapped host memory. 1684 * acc_update_self:: Update host memory from mapped device memory. 1685 * acc_map_data:: Map previously allocated device memory to host 1686 memory. 1687 * acc_unmap_data:: Unmap device memory from host memory. 1688 * acc_deviceptr:: Get device pointer associated with specific 1689 host address. 1690 * acc_hostptr:: Get host pointer associated with specific 1691 device address. 1692 * acc_is_present:: Indiciate whether host variable / array is 1693 present on device. 1694 * acc_memcpy_to_device:: Copy host memory to device memory. 1695 * acc_memcpy_from_device:: Copy device memory to host memory. 1696 1697 API routines for target platforms. 1698 1699 * acc_get_current_cuda_device:: Get CUDA device handle. 1700 * acc_get_current_cuda_context::Get CUDA context handle. 1701 * acc_get_cuda_stream:: Get CUDA stream handle. 1702 * acc_set_cuda_stream:: Set CUDA stream handle. 1703 1704 1705 File: libgomp.info, Node: acc_get_num_devices, Next: acc_set_device_type, Up: OpenACC Runtime Library Routines 1706 1707 5.1 'acc_get_num_devices' - Get number of devices for given device type 1708 ======================================================================= 1709 1710 _Description_ 1711 This function returns a value indicating the number of devices 1712 available for the device type specified in DEVICETYPE. 1713 1714 _C/C++_: 1715 _Prototype_: 'int acc_get_num_devices(acc_device_t devicetype);' 1716 1717 _Fortran_: 1718 _Interface_: 'integer function acc_get_num_devices(devicetype)' 1719 'integer(kind=acc_device_kind) devicetype' 1720 1721 _Reference_: 1722 OpenACC specification v2.0 (https://www.openacc.org), section 1723 3.2.1. 1724 1725 1726 File: libgomp.info, Node: acc_set_device_type, Next: acc_get_device_type, Prev: acc_get_num_devices, Up: OpenACC Runtime Library Routines 1727 1728 5.2 'acc_set_device_type' - Set type of device accelerator to use. 1729 ================================================================== 1730 1731 _Description_ 1732 This function indicates to the runtime library which device typr, 1733 specified in DEVICETYPE, to use when executing a parallel or 1734 kernels region. 1735 1736 _C/C++_: 1737 _Prototype_: 'acc_set_device_type(acc_device_t devicetype);' 1738 1739 _Fortran_: 1740 _Interface_: 'subroutine acc_set_device_type(devicetype)' 1741 'integer(kind=acc_device_kind) devicetype' 1742 1743 _Reference_: 1744 OpenACC specification v2.0 (https://www.openacc.org), section 1745 3.2.2. 1746 1747 1748 File: libgomp.info, Node: acc_get_device_type, Next: acc_set_device_num, Prev: acc_set_device_type, Up: OpenACC Runtime Library Routines 1749 1750 5.3 'acc_get_device_type' - Get type of device accelerator to be used. 1751 ====================================================================== 1752 1753 _Description_ 1754 This function returns what device type will be used when executing 1755 a parallel or kernels region. 1756 1757 _C/C++_: 1758 _Prototype_: 'acc_device_t acc_get_device_type(void);' 1759 1760 _Fortran_: 1761 _Interface_: 'function acc_get_device_type(void)' 1762 'integer(kind=acc_device_kind) acc_get_device_type' 1763 1764 _Reference_: 1765 OpenACC specification v2.0 (https://www.openacc.org), section 1766 3.2.3. 1767 1768 1769 File: libgomp.info, Node: acc_set_device_num, Next: acc_get_device_num, Prev: acc_get_device_type, Up: OpenACC Runtime Library Routines 1770 1771 5.4 'acc_set_device_num' - Set device number to use. 1772 ==================================================== 1773 1774 _Description_ 1775 This function will indicate to the runtime which device number, 1776 specified by NUM, associated with the specifed device type 1777 DEVICETYPE. 1778 1779 _C/C++_: 1780 _Prototype_: 'acc_set_device_num(int num, acc_device_t devicetype);' 1781 1782 _Fortran_: 1783 _Interface_: 'subroutine acc_set_device_num(devicenum, devicetype)' 1784 'integer devicenum' 1785 'integer(kind=acc_device_kind) devicetype' 1786 1787 _Reference_: 1788 OpenACC specification v2.0 (https://www.openacc.org), section 1789 3.2.4. 1790 1791 1792 File: libgomp.info, Node: acc_get_device_num, Next: acc_async_test, Prev: acc_set_device_num, Up: OpenACC Runtime Library Routines 1793 1794 5.5 'acc_get_device_num' - Get device number to be used. 1795 ======================================================== 1796 1797 _Description_ 1798 This function returns which device number associated with the 1799 specified device type DEVICETYPE, will be used when executing a 1800 parallel or kernels region. 1801 1802 _C/C++_: 1803 _Prototype_: 'int acc_get_device_num(acc_device_t devicetype);' 1804 1805 _Fortran_: 1806 _Interface_: 'function acc_get_device_num(devicetype)' 1807 'integer(kind=acc_device_kind) devicetype' 1808 'integer acc_get_device_num' 1809 1810 _Reference_: 1811 OpenACC specification v2.0 (https://www.openacc.org), section 1812 3.2.5. 1813 1814 1815 File: libgomp.info, Node: acc_async_test, Next: acc_async_test_all, Prev: acc_get_device_num, Up: OpenACC Runtime Library Routines 1816 1817 5.6 'acc_async_test' - Test for completion of a specific asynchronous operation. 1818 ================================================================================ 1819 1820 _Description_ 1821 This function tests for completion of the asynchrounous operation 1822 specified in ARG. In C/C++, a non-zero value will be returned to 1823 indicate the specified asynchronous operation has completed. While 1824 Fortran will return a 'true'. If the asynchrounous operation has 1825 not completed, C/C++ returns a zero and Fortran returns a 'false'. 1826 1827 _C/C++_: 1828 _Prototype_: 'int acc_async_test(int arg);' 1829 1830 _Fortran_: 1831 _Interface_: 'function acc_async_test(arg)' 1832 'integer(kind=acc_handle_kind) arg' 1833 'logical acc_async_test' 1834 1835 _Reference_: 1836 OpenACC specification v2.0 (https://www.openacc.org), section 1837 3.2.6. 1838 1839 1840 File: libgomp.info, Node: acc_async_test_all, Next: acc_wait, Prev: acc_async_test, Up: OpenACC Runtime Library Routines 1841 1842 5.7 'acc_async_test_all' - Tests for completion of all asynchronous operations. 1843 =============================================================================== 1844 1845 _Description_ 1846 This function tests for completion of all asynchrounous operations. 1847 In C/C++, a non-zero value will be returned to indicate all 1848 asynchronous operations have completed. While Fortran will return 1849 a 'true'. If any asynchronous operation has not completed, C/C++ 1850 returns a zero and Fortran returns a 'false'. 1851 1852 _C/C++_: 1853 _Prototype_: 'int acc_async_test_all(void);' 1854 1855 _Fortran_: 1856 _Interface_: 'function acc_async_test()' 1857 'logical acc_get_device_num' 1858 1859 _Reference_: 1860 OpenACC specification v2.0 (https://www.openacc.org), section 1861 3.2.7. 1862 1863 1864 File: libgomp.info, Node: acc_wait, Next: acc_wait_all, Prev: acc_async_test_all, Up: OpenACC Runtime Library Routines 1865 1866 5.8 'acc_wait' - Wait for completion of a specific asynchronous operation. 1867 ========================================================================== 1868 1869 _Description_ 1870 This function waits for completion of the asynchronous operation 1871 specified in ARG. 1872 1873 _C/C++_: 1874 _Prototype_: 'acc_wait(arg);' 1875 _Prototype 'acc_async_wait(arg);' 1876 (OpenACC 1.0 1877 compatibility)_: 1878 1879 _Fortran_: 1880 _Interface_: 'subroutine acc_wait(arg)' 1881 'integer(acc_handle_kind) arg' 1882 _Interface 'subroutine acc_async_wait(arg)' 1883 (OpenACC 1.0 1884 compatibility)_: 1885 'integer(acc_handle_kind) arg' 1886 1887 _Reference_: 1888 OpenACC specification v2.0 (https://www.openacc.org), section 1889 3.2.8. 1890 1891 1892 File: libgomp.info, Node: acc_wait_all, Next: acc_wait_all_async, Prev: acc_wait, Up: OpenACC Runtime Library Routines 1893 1894 5.9 'acc_wait_all' - Waits for completion of all asynchronous operations. 1895 ========================================================================= 1896 1897 _Description_ 1898 This function waits for the completion of all asynchronous 1899 operations. 1900 1901 _C/C++_: 1902 _Prototype_: 'acc_wait_all(void);' 1903 _Prototype 'acc_async_wait_all(void);' 1904 (OpenACC 1.0 1905 compatibility)_: 1906 1907 _Fortran_: 1908 _Interface_: 'subroutine acc_wait_all()' 1909 _Interface 'subroutine acc_async_wait_all()' 1910 (OpenACC 1.0 1911 compatibility)_: 1912 1913 _Reference_: 1914 OpenACC specification v2.0 (https://www.openacc.org), section 1915 3.2.10. 1916 1917 1918 File: libgomp.info, Node: acc_wait_all_async, Next: acc_wait_async, Prev: acc_wait_all, Up: OpenACC Runtime Library Routines 1919 1920 5.10 'acc_wait_all_async' - Wait for completion of all asynchronous operations. 1921 =============================================================================== 1922 1923 _Description_ 1924 This function enqueues a wait operation on the queue ASYNC for any 1925 and all asynchronous operations that have been previously enqueued 1926 on any queue. 1927 1928 _C/C++_: 1929 _Prototype_: 'acc_wait_all_async(int async);' 1930 1931 _Fortran_: 1932 _Interface_: 'subroutine acc_wait_all_async(async)' 1933 'integer(acc_handle_kind) async' 1934 1935 _Reference_: 1936 OpenACC specification v2.0 (https://www.openacc.org), section 1937 3.2.11. 1938 1939 1940 File: libgomp.info, Node: acc_wait_async, Next: acc_init, Prev: acc_wait_all_async, Up: OpenACC Runtime Library Routines 1941 1942 5.11 'acc_wait_async' - Wait for completion of asynchronous operations. 1943 ======================================================================= 1944 1945 _Description_ 1946 This function enqueues a wait operation on queue ASYNC for any and 1947 all asynchronous operations enqueued on queue ARG. 1948 1949 _C/C++_: 1950 _Prototype_: 'acc_wait_async(int arg, int async);' 1951 1952 _Fortran_: 1953 _Interface_: 'subroutine acc_wait_async(arg, async)' 1954 'integer(acc_handle_kind) arg, async' 1955 1956 _Reference_: 1957 OpenACC specification v2.0 (https://www.openacc.org), section 1958 3.2.9. 1959 1960 1961 File: libgomp.info, Node: acc_init, Next: acc_shutdown, Prev: acc_wait_async, Up: OpenACC Runtime Library Routines 1962 1963 5.12 'acc_init' - Initialize runtime for a specific device type. 1964 ================================================================ 1965 1966 _Description_ 1967 This function initializes the runtime for the device type specified 1968 in DEVICETYPE. 1969 1970 _C/C++_: 1971 _Prototype_: 'acc_init(acc_device_t devicetype);' 1972 1973 _Fortran_: 1974 _Interface_: 'subroutine acc_init(devicetype)' 1975 'integer(acc_device_kind) devicetype' 1976 1977 _Reference_: 1978 OpenACC specification v2.0 (https://www.openacc.org), section 1979 3.2.12. 1980 1981 1982 File: libgomp.info, Node: acc_shutdown, Next: acc_on_device, Prev: acc_init, Up: OpenACC Runtime Library Routines 1983 1984 5.13 'acc_shutdown' - Shuts down the runtime for a specific device type. 1985 ======================================================================== 1986 1987 _Description_ 1988 This function shuts down the runtime for the device type specified 1989 in DEVICETYPE. 1990 1991 _C/C++_: 1992 _Prototype_: 'acc_shutdown(acc_device_t devicetype);' 1993 1994 _Fortran_: 1995 _Interface_: 'subroutine acc_shutdown(devicetype)' 1996 'integer(acc_device_kind) devicetype' 1997 1998 _Reference_: 1999 OpenACC specification v2.0 (https://www.openacc.org), section 2000 3.2.13. 2001 2002 2003 File: libgomp.info, Node: acc_on_device, Next: acc_malloc, Prev: acc_shutdown, Up: OpenACC Runtime Library Routines 2004 2005 5.14 'acc_on_device' - Whether executing on a particular device 2006 =============================================================== 2007 2008 _Description_: 2009 This function returns whether the program is executing on a 2010 particular device specified in DEVICETYPE. In C/C++ a non-zero 2011 value is returned to indicate the device is execiting on the 2012 specified device type. In Fortran, 'true' will be returned. If 2013 the program is not executing on the specified device type C/C++ 2014 will return a zero, while Fortran will return 'false'. 2015 2016 _C/C++_: 2017 _Prototype_: 'acc_on_device(acc_device_t devicetype);' 2018 2019 _Fortran_: 2020 _Interface_: 'function acc_on_device(devicetype)' 2021 'integer(acc_device_kind) devicetype' 2022 'logical acc_on_device' 2023 2024 _Reference_: 2025 OpenACC specification v2.0 (https://www.openacc.org), section 2026 3.2.14. 2027 2028 2029 File: libgomp.info, Node: acc_malloc, Next: acc_free, Prev: acc_on_device, Up: OpenACC Runtime Library Routines 2030 2031 5.15 'acc_malloc' - Allocate device memory. 2032 =========================================== 2033 2034 _Description_ 2035 This function allocates LEN bytes of device memory. It returns the 2036 device address of the allocated memory. 2037 2038 _C/C++_: 2039 _Prototype_: 'd_void* acc_malloc(size_t len);' 2040 2041 _Reference_: 2042 OpenACC specification v2.0 (https://www.openacc.org), section 2043 3.2.15. 2044 2045 2046 File: libgomp.info, Node: acc_free, Next: acc_copyin, Prev: acc_malloc, Up: OpenACC Runtime Library Routines 2047 2048 5.16 'acc_free' - Free device memory. 2049 ===================================== 2050 2051 _Description_ 2052 Free previously allocated device memory at the device address 'a'. 2053 2054 _C/C++_: 2055 _Prototype_: 'acc_free(d_void *a);' 2056 2057 _Reference_: 2058 OpenACC specification v2.0 (https://www.openacc.org), section 2059 3.2.16. 2060 2061 2062 File: libgomp.info, Node: acc_copyin, Next: acc_present_or_copyin, Prev: acc_free, Up: OpenACC Runtime Library Routines 2063 2064 5.17 'acc_copyin' - Allocate device memory and copy host memory to it. 2065 ====================================================================== 2066 2067 _Description_ 2068 In C/C++, this function allocates LEN bytes of device memory and 2069 maps it to the specified host address in A. The device address of 2070 the newly allocated device memory is returned. 2071 2072 In Fortran, two (2) forms are supported. In the first form, A 2073 specifies a contiguous array section. The second form A specifies 2074 a variable or array element and LEN specifies the length in bytes. 2075 2076 _C/C++_: 2077 _Prototype_: 'void *acc_copyin(h_void *a, size_t len);' 2078 2079 _Fortran_: 2080 _Interface_: 'subroutine acc_copyin(a)' 2081 'type, dimension(:[,:]...) :: a' 2082 _Interface_: 'subroutine acc_copyin(a, len)' 2083 'type, dimension(:[,:]...) :: a' 2084 'integer len' 2085 2086 _Reference_: 2087 OpenACC specification v2.0 (https://www.openacc.org), section 2088 3.2.17. 2089 2090 2091 File: libgomp.info, Node: acc_present_or_copyin, Next: acc_create, Prev: acc_copyin, Up: OpenACC Runtime Library Routines 2092 2093 5.18 'acc_present_or_copyin' - If the data is not present on the device, allocate device memory and copy from host memory. 2094 ========================================================================================================================== 2095 2096 _Description_ 2097 This function tests if the host data specifed by A and of length 2098 LEN is present or not. If it is not present, then device memory 2099 will be allocated and the host memory copied. The device address 2100 of the newly allocated device memory is returned. 2101 2102 In Fortran, two (2) forms are supported. In the first form, A 2103 specifies a contiguous array section. The second form A specifies 2104 a variable or array element and LEN specifies the length in bytes. 2105 2106 _C/C++_: 2107 _Prototype_: 'void *acc_present_or_copyin(h_void *a, size_t len);' 2108 _Prototype_: 'void *acc_pcopyin(h_void *a, size_t len);' 2109 2110 _Fortran_: 2111 _Interface_: 'subroutine acc_present_or_copyin(a)' 2112 'type, dimension(:[,:]...) :: a' 2113 _Interface_: 'subroutine acc_present_or_copyin(a, len)' 2114 'type, dimension(:[,:]...) :: a' 2115 'integer len' 2116 _Interface_: 'subroutine acc_pcopyin(a)' 2117 'type, dimension(:[,:]...) :: a' 2118 _Interface_: 'subroutine acc_pcopyin(a, len)' 2119 'type, dimension(:[,:]...) :: a' 2120 'integer len' 2121 2122 _Reference_: 2123 OpenACC specification v2.0 (https://www.openacc.org), section 2124 3.2.18. 2125 2126 2127 File: libgomp.info, Node: acc_create, Next: acc_present_or_create, Prev: acc_present_or_copyin, Up: OpenACC Runtime Library Routines 2128 2129 5.19 'acc_create' - Allocate device memory and map it to host memory. 2130 ===================================================================== 2131 2132 _Description_ 2133 This function allocates device memory and maps it to host memory 2134 specified by the host address A with a length of LEN bytes. In 2135 C/C++, the function returns the device address of the allocated 2136 device memory. 2137 2138 In Fortran, two (2) forms are supported. In the first form, A 2139 specifies a contiguous array section. The second form A specifies 2140 a variable or array element and LEN specifies the length in bytes. 2141 2142 _C/C++_: 2143 _Prototype_: 'void *acc_create(h_void *a, size_t len);' 2144 2145 _Fortran_: 2146 _Interface_: 'subroutine acc_create(a)' 2147 'type, dimension(:[,:]...) :: a' 2148 _Interface_: 'subroutine acc_create(a, len)' 2149 'type, dimension(:[,:]...) :: a' 2150 'integer len' 2151 2152 _Reference_: 2153 OpenACC specification v2.0 (https://www.openacc.org), section 2154 3.2.19. 2155 2156 2157 File: libgomp.info, Node: acc_present_or_create, Next: acc_copyout, Prev: acc_create, Up: OpenACC Runtime Library Routines 2158 2159 5.20 'acc_present_or_create' - If the data is not present on the device, allocate device memory and map it to host memory. 2160 ========================================================================================================================== 2161 2162 _Description_ 2163 This function tests if the host data specifed by A and of length 2164 LEN is present or not. If it is not present, then device memory 2165 will be allocated and mapped to host memory. In C/C++, the device 2166 address of the newly allocated device memory is returned. 2167 2168 In Fortran, two (2) forms are supported. In the first form, A 2169 specifies a contiguous array section. The second form A specifies 2170 a variable or array element and LEN specifies the length in bytes. 2171 2172 _C/C++_: 2173 _Prototype_: 'void *acc_present_or_create(h_void *a, size_t len)' 2174 _Prototype_: 'void *acc_pcreate(h_void *a, size_t len)' 2175 2176 _Fortran_: 2177 _Interface_: 'subroutine acc_present_or_create(a)' 2178 'type, dimension(:[,:]...) :: a' 2179 _Interface_: 'subroutine acc_present_or_create(a, len)' 2180 'type, dimension(:[,:]...) :: a' 2181 'integer len' 2182 _Interface_: 'subroutine acc_pcreate(a)' 2183 'type, dimension(:[,:]...) :: a' 2184 _Interface_: 'subroutine acc_pcreate(a, len)' 2185 'type, dimension(:[,:]...) :: a' 2186 'integer len' 2187 2188 _Reference_: 2189 OpenACC specification v2.0 (https://www.openacc.org), section 2190 3.2.20. 2191 2192 2193 File: libgomp.info, Node: acc_copyout, Next: acc_delete, Prev: acc_present_or_create, Up: OpenACC Runtime Library Routines 2194 2195 5.21 'acc_copyout' - Copy device memory to host memory. 2196 ======================================================= 2197 2198 _Description_ 2199 This function copies mapped device memory to host memory which is 2200 specified by host address A for a length LEN bytes in C/C++. 2201 2202 In Fortran, two (2) forms are supported. In the first form, A 2203 specifies a contiguous array section. The second form A specifies 2204 a variable or array element and LEN specifies the length in bytes. 2205 2206 _C/C++_: 2207 _Prototype_: 'acc_copyout(h_void *a, size_t len);' 2208 2209 _Fortran_: 2210 _Interface_: 'subroutine acc_copyout(a)' 2211 'type, dimension(:[,:]...) :: a' 2212 _Interface_: 'subroutine acc_copyout(a, len)' 2213 'type, dimension(:[,:]...) :: a' 2214 'integer len' 2215 2216 _Reference_: 2217 OpenACC specification v2.0 (https://www.openacc.org), section 2218 3.2.21. 2219 2220 2221 File: libgomp.info, Node: acc_delete, Next: acc_update_device, Prev: acc_copyout, Up: OpenACC Runtime Library Routines 2222 2223 5.22 'acc_delete' - Free device memory. 2224 ======================================= 2225 2226 _Description_ 2227 This function frees previously allocated device memory specified by 2228 the device address A and the length of LEN bytes. 2229 2230 In Fortran, two (2) forms are supported. In the first form, A 2231 specifies a contiguous array section. The second form A specifies 2232 a variable or array element and LEN specifies the length in bytes. 2233 2234 _C/C++_: 2235 _Prototype_: 'acc_delete(h_void *a, size_t len);' 2236 2237 _Fortran_: 2238 _Interface_: 'subroutine acc_delete(a)' 2239 'type, dimension(:[,:]...) :: a' 2240 _Interface_: 'subroutine acc_delete(a, len)' 2241 'type, dimension(:[,:]...) :: a' 2242 'integer len' 2243 2244 _Reference_: 2245 OpenACC specification v2.0 (https://www.openacc.org), section 2246 3.2.22. 2247 2248 2249 File: libgomp.info, Node: acc_update_device, Next: acc_update_self, Prev: acc_delete, Up: OpenACC Runtime Library Routines 2250 2251 5.23 'acc_update_device' - Update device memory from mapped host memory. 2252 ======================================================================== 2253 2254 _Description_ 2255 This function updates the device copy from the previously mapped 2256 host memory. The host memory is specified with the host address A 2257 and a length of LEN bytes. 2258 2259 In Fortran, two (2) forms are supported. In the first form, A 2260 specifies a contiguous array section. The second form A specifies 2261 a variable or array element and LEN specifies the length in bytes. 2262 2263 _C/C++_: 2264 _Prototype_: 'acc_update_device(h_void *a, size_t len);' 2265 2266 _Fortran_: 2267 _Interface_: 'subroutine acc_update_device(a)' 2268 'type, dimension(:[,:]...) :: a' 2269 _Interface_: 'subroutine acc_update_device(a, len)' 2270 'type, dimension(:[,:]...) :: a' 2271 'integer len' 2272 2273 _Reference_: 2274 OpenACC specification v2.0 (https://www.openacc.org), section 2275 3.2.23. 2276 2277 2278 File: libgomp.info, Node: acc_update_self, Next: acc_map_data, Prev: acc_update_device, Up: OpenACC Runtime Library Routines 2279 2280 5.24 'acc_update_self' - Update host memory from mapped device memory. 2281 ====================================================================== 2282 2283 _Description_ 2284 This function updates the host copy from the previously mapped 2285 device memory. The host memory is specified with the host address 2286 A and a length of LEN bytes. 2287 2288 In Fortran, two (2) forms are supported. In the first form, A 2289 specifies a contiguous array section. The second form A specifies 2290 a variable or array element and LEN specifies the length in bytes. 2291 2292 _C/C++_: 2293 _Prototype_: 'acc_update_self(h_void *a, size_t len);' 2294 2295 _Fortran_: 2296 _Interface_: 'subroutine acc_update_self(a)' 2297 'type, dimension(:[,:]...) :: a' 2298 _Interface_: 'subroutine acc_update_self(a, len)' 2299 'type, dimension(:[,:]...) :: a' 2300 'integer len' 2301 2302 _Reference_: 2303 OpenACC specification v2.0 (https://www.openacc.org), section 2304 3.2.24. 2305 2306 2307 File: libgomp.info, Node: acc_map_data, Next: acc_unmap_data, Prev: acc_update_self, Up: OpenACC Runtime Library Routines 2308 2309 5.25 'acc_map_data' - Map previously allocated device memory to host memory. 2310 ============================================================================ 2311 2312 _Description_ 2313 This function maps previously allocated device and host memory. 2314 The device memory is specified with the device address D. The host 2315 memory is specified with the host address H and a length of LEN. 2316 2317 _C/C++_: 2318 _Prototype_: 'acc_map_data(h_void *h, d_void *d, size_t len);' 2319 2320 _Reference_: 2321 OpenACC specification v2.0 (https://www.openacc.org), section 2322 3.2.25. 2323 2324 2325 File: libgomp.info, Node: acc_unmap_data, Next: acc_deviceptr, Prev: acc_map_data, Up: OpenACC Runtime Library Routines 2326 2327 5.26 'acc_unmap_data' - Unmap device memory from host memory. 2328 ============================================================= 2329 2330 _Description_ 2331 This function unmaps previously mapped device and host memory. The 2332 latter specified by H. 2333 2334 _C/C++_: 2335 _Prototype_: 'acc_unmap_data(h_void *h);' 2336 2337 _Reference_: 2338 OpenACC specification v2.0 (https://www.openacc.org), section 2339 3.2.26. 2340 2341 2342 File: libgomp.info, Node: acc_deviceptr, Next: acc_hostptr, Prev: acc_unmap_data, Up: OpenACC Runtime Library Routines 2343 2344 5.27 'acc_deviceptr' - Get device pointer associated with specific host address. 2345 ================================================================================ 2346 2347 _Description_ 2348 This function returns the device address that has been mapped to 2349 the host address specified by H. 2350 2351 _C/C++_: 2352 _Prototype_: 'void *acc_deviceptr(h_void *h);' 2353 2354 _Reference_: 2355 OpenACC specification v2.0 (https://www.openacc.org), section 2356 3.2.27. 2357 2358 2359 File: libgomp.info, Node: acc_hostptr, Next: acc_is_present, Prev: acc_deviceptr, Up: OpenACC Runtime Library Routines 2360 2361 5.28 'acc_hostptr' - Get host pointer associated with specific device address. 2362 ============================================================================== 2363 2364 _Description_ 2365 This function returns the host address that has been mapped to the 2366 device address specified by D. 2367 2368 _C/C++_: 2369 _Prototype_: 'void *acc_hostptr(d_void *d);' 2370 2371 _Reference_: 2372 OpenACC specification v2.0 (https://www.openacc.org), section 2373 3.2.28. 2374 2375 2376 File: libgomp.info, Node: acc_is_present, Next: acc_memcpy_to_device, Prev: acc_hostptr, Up: OpenACC Runtime Library Routines 2377 2378 5.29 'acc_is_present' - Indicate whether host variable / array is present on device. 2379 ==================================================================================== 2380 2381 _Description_ 2382 This function indicates whether the specified host address in A and 2383 a length of LEN bytes is present on the device. In C/C++, a 2384 non-zero value is returned to indicate the presence of the mapped 2385 memory on the device. A zero is returned to indicate the memory is 2386 not mapped on the device. 2387 2388 In Fortran, two (2) forms are supported. In the first form, A 2389 specifies a contiguous array section. The second form A specifies 2390 a variable or array element and LEN specifies the length in bytes. 2391 If the host memory is mapped to device memory, then a 'true' is 2392 returned. Otherwise, a 'false' is return to indicate the mapped 2393 memory is not present. 2394 2395 _C/C++_: 2396 _Prototype_: 'int acc_is_present(h_void *a, size_t len);' 2397 2398 _Fortran_: 2399 _Interface_: 'function acc_is_present(a)' 2400 'type, dimension(:[,:]...) :: a' 2401 'logical acc_is_present' 2402 _Interface_: 'function acc_is_present(a, len)' 2403 'type, dimension(:[,:]...) :: a' 2404 'integer len' 2405 'logical acc_is_present' 2406 2407 _Reference_: 2408 OpenACC specification v2.0 (https://www.openacc.org), section 2409 3.2.29. 2410 2411 2412 File: libgomp.info, Node: acc_memcpy_to_device, Next: acc_memcpy_from_device, Prev: acc_is_present, Up: OpenACC Runtime Library Routines 2413 2414 5.30 'acc_memcpy_to_device' - Copy host memory to device memory. 2415 ================================================================ 2416 2417 _Description_ 2418 This function copies host memory specified by host address of SRC 2419 to device memory specified by the device address DEST for a length 2420 of BYTES bytes. 2421 2422 _C/C++_: 2423 _Prototype_: 'acc_memcpy_to_device(d_void *dest, h_void *src, size_t 2424 bytes);' 2425 2426 _Reference_: 2427 OpenACC specification v2.0 (https://www.openacc.org), section 2428 3.2.30. 2429 2430 2431 File: libgomp.info, Node: acc_memcpy_from_device, Next: acc_get_current_cuda_device, Prev: acc_memcpy_to_device, Up: OpenACC Runtime Library Routines 2432 2433 5.31 'acc_memcpy_from_device' - Copy device memory to host memory. 2434 ================================================================== 2435 2436 _Description_ 2437 This function copies host memory specified by host address of SRC 2438 from device memory specified by the device address DEST for a 2439 length of BYTES bytes. 2440 2441 _C/C++_: 2442 _Prototype_: 'acc_memcpy_from_device(d_void *dest, h_void *src, 2443 size_t bytes);' 2444 2445 _Reference_: 2446 OpenACC specification v2.0 (https://www.openacc.org), section 2447 3.2.31. 2448 2449 2450 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 2451 2452 5.32 'acc_get_current_cuda_device' - Get CUDA device handle. 2453 ============================================================ 2454 2455 _Description_ 2456 This function returns the CUDA device handle. This handle is the 2457 same as used by the CUDA Runtime or Driver API's. 2458 2459 _C/C++_: 2460 _Prototype_: 'void *acc_get_current_cuda_device(void);' 2461 2462 _Reference_: 2463 OpenACC specification v2.0 (https://www.openacc.org), section 2464 A.2.1.1. 2465 2466 2467 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 2468 2469 5.33 'acc_get_current_cuda_context' - Get CUDA context handle. 2470 ============================================================== 2471 2472 _Description_ 2473 This function returns the CUDA context handle. This handle is the 2474 same as used by the CUDA Runtime or Driver API's. 2475 2476 _C/C++_: 2477 _Prototype_: 'acc_get_current_cuda_context(void);' 2478 2479 _Reference_: 2480 OpenACC specification v2.0 (https://www.openacc.org), section 2481 A.2.1.2. 2482 2483 2484 File: libgomp.info, Node: acc_get_cuda_stream, Next: acc_set_cuda_stream, Prev: acc_get_current_cuda_context, Up: OpenACC Runtime Library Routines 2485 2486 5.34 'acc_get_cuda_stream' - Get CUDA stream handle. 2487 ==================================================== 2488 2489 _Description_ 2490 This function returns the CUDA stream handle. This handle is the 2491 same as used by the CUDA Runtime or Driver API's. 2492 2493 _C/C++_: 2494 _Prototype_: 'acc_get_cuda_stream(void);' 2495 2496 _Reference_: 2497 OpenACC specification v2.0 (https://www.openacc.org), section 2498 A.2.1.3. 2499 2500 2501 File: libgomp.info, Node: acc_set_cuda_stream, Prev: acc_get_cuda_stream, Up: OpenACC Runtime Library Routines 2502 2503 5.35 'acc_set_cuda_stream' - Set CUDA stream handle. 2504 ==================================================== 2505 2506 _Description_ 2507 This function associates the stream handle specified by STREAM with 2508 the asynchronous value specified by ASYNC. 2509 2510 _C/C++_: 2511 _Prototype_: 'acc_set_cuda_stream(int async void *stream);' 2512 2513 _Reference_: 2514 OpenACC specification v2.0 (https://www.openacc.org), section 2515 A.2.1.4. 2516 2517 2518 File: libgomp.info, Node: OpenACC Environment Variables, Next: CUDA Streams Usage, Prev: OpenACC Runtime Library Routines, Up: Top 2519 2520 6 OpenACC Environment Variables 2521 ******************************* 2522 2523 The variables 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM' are defined by 2524 section 4 of the OpenACC specification in version 2.0. The variable 2525 'GCC_ACC_NOTIFY' is used for diagnostic purposes. 2526 2527 * Menu: 2528 2529 * ACC_DEVICE_TYPE:: 2530 * ACC_DEVICE_NUM:: 2531 * GCC_ACC_NOTIFY:: 2532 2533 2534 File: libgomp.info, Node: ACC_DEVICE_TYPE, Next: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2535 2536 6.1 'ACC_DEVICE_TYPE' 2537 ===================== 2538 2539 _Reference_: 2540 OpenACC specification v2.0 (https://www.openacc.org), section 4.1. 2541 2542 2543 File: libgomp.info, Node: ACC_DEVICE_NUM, Next: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_TYPE, Up: OpenACC Environment Variables 2544 2545 6.2 'ACC_DEVICE_NUM' 2546 ==================== 2547 2548 _Reference_: 2549 OpenACC specification v2.0 (https://www.openacc.org), section 4.2. 2550 2551 2552 File: libgomp.info, Node: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 2553 2554 6.3 'GCC_ACC_NOTIFY' 2555 ==================== 2556 2557 _Description_: 2558 Print debug information pertaining to the accelerator. 2559 2560 2561 File: libgomp.info, Node: CUDA Streams Usage, Next: OpenACC Library Interoperability, Prev: OpenACC Environment Variables, Up: Top 2562 2563 7 CUDA Streams Usage 2564 ******************** 2565 2566 This applies to the 'nvptx' plugin only. 2567 2568 The library provides elements that perform asynchronous movement of 2569 data and asynchronous operation of computing constructs. This 2570 asynchronous functionality is implemented by making use of CUDA 2571 streams(1). 2572 2573 The primary means by that the asychronous functionality is accessed 2574 is through the use of those OpenACC directives which make use of the 2575 'async' and 'wait' clauses. When the 'async' clause is first used with 2576 a directive, it creates a CUDA stream. If an 'async-argument' is used 2577 with the 'async' clause, then the stream is associated with the 2578 specified 'async-argument'. 2579 2580 Following the creation of an association between a CUDA stream and 2581 the 'async-argument' of an 'async' clause, both the 'wait' clause and 2582 the 'wait' directive can be used. When either the clause or directive 2583 is used after stream creation, it creates a rendezvous point whereby 2584 execution waits until all operations associated with the 2585 'async-argument', that is, stream, have completed. 2586 2587 Normally, the management of the streams that are created as a result 2588 of using the 'async' clause, is done without any intervention by the 2589 caller. This implies the association between the 'async-argument' and 2590 the CUDA stream will be maintained for the lifetime of the program. 2591 However, this association can be changed through the use of the library 2592 function 'acc_set_cuda_stream'. When the function 'acc_set_cuda_stream' 2593 is called, the CUDA stream that was originally associated with the 2594 'async' clause will be destroyed. Caution should be taken when changing 2595 the association as subsequent references to the 'async-argument' refer 2596 to a different CUDA stream. 2597 2598 ---------- Footnotes ---------- 2599 2600 (1) See "Stream Management" in "CUDA Driver API", TRM-06703-001, 2601 Version 5.5, for additional information 2602 2603 2604 File: libgomp.info, Node: OpenACC Library Interoperability, Next: The libgomp ABI, Prev: CUDA Streams Usage, Up: Top 2605 2606 8 OpenACC Library Interoperability 2607 ********************************** 2608 2609 8.1 Introduction 2610 ================ 2611 2612 The OpenACC library uses the CUDA Driver API, and may interact with 2613 programs that use the Runtime library directly, or another library based 2614 on the Runtime library, e.g., CUBLAS(1). This chapter describes the use 2615 cases and what changes are required in order to use both the OpenACC 2616 library and the CUBLAS and Runtime libraries within a program. 2617 2618 8.2 First invocation: NVIDIA CUBLAS library API 2619 =============================================== 2620 2621 In this first use case (see below), a function in the CUBLAS library is 2622 called prior to any of the functions in the OpenACC library. More 2623 specifically, the function 'cublasCreate()'. 2624 2625 When invoked, the function initializes the library and allocates the 2626 hardware resources on the host and the device on behalf of the caller. 2627 Once the initialization and allocation has completed, a handle is 2628 returned to the caller. The OpenACC library also requires 2629 initialization and allocation of hardware resources. Since the CUBLAS 2630 library has already allocated the hardware resources for the device, all 2631 that is left to do is to initialize the OpenACC library and acquire the 2632 hardware resources on the host. 2633 2634 Prior to calling the OpenACC function that initializes the library 2635 and allocate the host hardware resources, you need to acquire the device 2636 number that was allocated during the call to 'cublasCreate()'. The 2637 invoking of the runtime library function 'cudaGetDevice()' accomplishes 2638 this. Once acquired, the device number is passed along with the device 2639 type as parameters to the OpenACC library function 2640 'acc_set_device_num()'. 2641 2642 Once the call to 'acc_set_device_num()' has completed, the OpenACC 2643 library uses the context that was created during the call to 2644 'cublasCreate()'. In other words, both libraries will be sharing the 2645 same context. 2646 2647 /* Create the handle */ 2648 s = cublasCreate(&h); 2649 if (s != CUBLAS_STATUS_SUCCESS) 2650 { 2651 fprintf(stderr, "cublasCreate failed %d\n", s); 2652 exit(EXIT_FAILURE); 2653 } 2654 2655 /* Get the device number */ 2656 e = cudaGetDevice(&dev); 2657 if (e != cudaSuccess) 2658 { 2659 fprintf(stderr, "cudaGetDevice failed %d\n", e); 2660 exit(EXIT_FAILURE); 2661 } 2662 2663 /* Initialize OpenACC library and use device 'dev' */ 2664 acc_set_device_num(dev, acc_device_nvidia); 2665 2666 Use Case 1 2667 2668 8.3 First invocation: OpenACC library API 2669 ========================================= 2670 2671 In this second use case (see below), a function in the OpenACC library 2672 is called prior to any of the functions in the CUBLAS library. More 2673 specificially, the function 'acc_set_device_num()'. 2674 2675 In the use case presented here, the function 'acc_set_device_num()' 2676 is used to both initialize the OpenACC library and allocate the hardware 2677 resources on the host and the device. In the call to the function, the 2678 call parameters specify which device to use and what device type to use, 2679 i.e., 'acc_device_nvidia'. It should be noted that this is but one 2680 method to initialize the OpenACC library and allocate the appropriate 2681 hardware resources. Other methods are available through the use of 2682 environment variables and these will be discussed in the next section. 2683 2684 Once the call to 'acc_set_device_num()' has completed, other OpenACC 2685 functions can be called as seen with multiple calls being made to 2686 'acc_copyin()'. In addition, calls can be made to functions in the 2687 CUBLAS library. In the use case a call to 'cublasCreate()' is made 2688 subsequent to the calls to 'acc_copyin()'. As seen in the previous use 2689 case, a call to 'cublasCreate()' initializes the CUBLAS library and 2690 allocates the hardware resources on the host and the device. However, 2691 since the device has already been allocated, 'cublasCreate()' will only 2692 initialize the CUBLAS library and allocate the appropriate hardware 2693 resources on the host. The context that was created as part of the 2694 OpenACC initialization is shared with the CUBLAS library, similarly to 2695 the first use case. 2696 2697 dev = 0; 2698 2699 acc_set_device_num(dev, acc_device_nvidia); 2700 2701 /* Copy the first set to the device */ 2702 d_X = acc_copyin(&h_X[0], N * sizeof (float)); 2703 if (d_X == NULL) 2704 { 2705 fprintf(stderr, "copyin error h_X\n"); 2706 exit(EXIT_FAILURE); 2707 } 2708 2709 /* Copy the second set to the device */ 2710 d_Y = acc_copyin(&h_Y1[0], N * sizeof (float)); 2711 if (d_Y == NULL) 2712 { 2713 fprintf(stderr, "copyin error h_Y1\n"); 2714 exit(EXIT_FAILURE); 2715 } 2716 2717 /* Create the handle */ 2718 s = cublasCreate(&h); 2719 if (s != CUBLAS_STATUS_SUCCESS) 2720 { 2721 fprintf(stderr, "cublasCreate failed %d\n", s); 2722 exit(EXIT_FAILURE); 2723 } 2724 2725 /* Perform saxpy using CUBLAS library function */ 2726 s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1); 2727 if (s != CUBLAS_STATUS_SUCCESS) 2728 { 2729 fprintf(stderr, "cublasSaxpy failed %d\n", s); 2730 exit(EXIT_FAILURE); 2731 } 2732 2733 /* Copy the results from the device */ 2734 acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float)); 2735 2736 Use Case 2 2737 2738 8.4 OpenACC library and environment variables 2739 ============================================= 2740 2741 There are two environment variables associated with the OpenACC library 2742 that may be used to control the device type and device number: 2743 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM', respecively. These two 2744 environement variables can be used as an alternative to calling 2745 'acc_set_device_num()'. As seen in the second use case, the device type 2746 and device number were specified using 'acc_set_device_num()'. If 2747 however, the aforementioned environment variables were set, then the 2748 call to 'acc_set_device_num()' would not be required. 2749 2750 The use of the environment variables is only relevant when an OpenACC 2751 function is called prior to a call to 'cudaCreate()'. If 'cudaCreate()' 2752 is called prior to a call to an OpenACC function, then you must call 2753 'acc_set_device_num()'(2) 2754 2755 ---------- Footnotes ---------- 2756 2757 (1) See section 2.26, "Interactions with the CUDA Driver API" in 2758 "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU 2759 Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5, for 2760 additional information on library interoperability. 2761 2762 (2) More complete information about 'ACC_DEVICE_TYPE' and 2763 'ACC_DEVICE_NUM' can be found in sections 4.1 and 4.2 of the OpenACC 2764 (https://www.openacc.org) Application Programming Interface, Version 2765 2.0. 2766 2767 2768 File: libgomp.info, Node: The libgomp ABI, Next: Reporting Bugs, Prev: OpenACC Library Interoperability, Up: Top 2769 2770 9 The libgomp ABI 2771 ***************** 2772 2773 The following sections present notes on the external ABI as presented by 2774 libgomp. Only maintainers should need them. 2775 2776 * Menu: 2777 2778 * Implementing MASTER construct:: 2779 * Implementing CRITICAL construct:: 2780 * Implementing ATOMIC construct:: 2781 * Implementing FLUSH construct:: 2782 * Implementing BARRIER construct:: 2783 * Implementing THREADPRIVATE construct:: 2784 * Implementing PRIVATE clause:: 2785 * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses:: 2786 * Implementing REDUCTION clause:: 2787 * Implementing PARALLEL construct:: 2788 * Implementing FOR construct:: 2789 * Implementing ORDERED construct:: 2790 * Implementing SECTIONS construct:: 2791 * Implementing SINGLE construct:: 2792 * Implementing OpenACC's PARALLEL construct:: 2793 2794 2795 File: libgomp.info, Node: Implementing MASTER construct, Next: Implementing CRITICAL construct, Up: The libgomp ABI 2796 2797 9.1 Implementing MASTER construct 2798 ================================= 2799 2800 if (omp_get_thread_num () == 0) 2801 block 2802 2803 Alternately, we generate two copies of the parallel subfunction and 2804 only include this in the version run by the master thread. Surely this 2805 is not worthwhile though... 2806 2807 2808 File: libgomp.info, Node: Implementing CRITICAL construct, Next: Implementing ATOMIC construct, Prev: Implementing MASTER construct, Up: The libgomp ABI 2809 2810 9.2 Implementing CRITICAL construct 2811 =================================== 2812 2813 Without a specified name, 2814 2815 void GOMP_critical_start (void); 2816 void GOMP_critical_end (void); 2817 2818 so that we don't get COPY relocations from libgomp to the main 2819 application. 2820 2821 With a specified name, use omp_set_lock and omp_unset_lock with name 2822 being transformed into a variable declared like 2823 2824 omp_lock_t gomp_critical_user_<name> __attribute__((common)) 2825 2826 Ideally the ABI would specify that all zero is a valid unlocked 2827 state, and so we wouldn't need to initialize this at startup. 2828 2829 2830 File: libgomp.info, Node: Implementing ATOMIC construct, Next: Implementing FLUSH construct, Prev: Implementing CRITICAL construct, Up: The libgomp ABI 2831 2832 9.3 Implementing ATOMIC construct 2833 ================================= 2834 2835 The target should implement the '__sync' builtins. 2836 2837 Failing that we could add 2838 2839 void GOMP_atomic_enter (void) 2840 void GOMP_atomic_exit (void) 2841 2842 which reuses the regular lock code, but with yet another lock object 2843 private to the library. 2844 2845 2846 File: libgomp.info, Node: Implementing FLUSH construct, Next: Implementing BARRIER construct, Prev: Implementing ATOMIC construct, Up: The libgomp ABI 2847 2848 9.4 Implementing FLUSH construct 2849 ================================ 2850 2851 Expands to the '__sync_synchronize' builtin. 2852 2853 2854 File: libgomp.info, Node: Implementing BARRIER construct, Next: Implementing THREADPRIVATE construct, Prev: Implementing FLUSH construct, Up: The libgomp ABI 2855 2856 9.5 Implementing BARRIER construct 2857 ================================== 2858 2859 void GOMP_barrier (void) 2860 2861 2862 File: libgomp.info, Node: Implementing THREADPRIVATE construct, Next: Implementing PRIVATE clause, Prev: Implementing BARRIER construct, Up: The libgomp ABI 2863 2864 9.6 Implementing THREADPRIVATE construct 2865 ======================================== 2866 2867 In _most_ cases we can map this directly to '__thread'. Except that OMP 2868 allows constructors for C++ objects. We can either refuse to support 2869 this (how often is it used?) or we can implement something akin to 2870 .ctors. 2871 2872 Even more ideally, this ctor feature is handled by extensions to the 2873 main pthreads library. Failing that, we can have a set of entry points 2874 to register ctor functions to be called. 2875 2876 2877 File: libgomp.info, Node: Implementing PRIVATE clause, Next: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Prev: Implementing THREADPRIVATE construct, Up: The libgomp ABI 2878 2879 9.7 Implementing PRIVATE clause 2880 =============================== 2881 2882 In association with a PARALLEL, or within the lexical extent of a 2883 PARALLEL block, the variable becomes a local variable in the parallel 2884 subfunction. 2885 2886 In association with FOR or SECTIONS blocks, create a new automatic 2887 variable within the current function. This preserves the semantic of 2888 new variable creation. 2889 2890 2891 File: libgomp.info, Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Next: Implementing REDUCTION clause, Prev: Implementing PRIVATE clause, Up: The libgomp ABI 2892 2893 9.8 Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses 2894 ======================================================================== 2895 2896 This seems simple enough for PARALLEL blocks. Create a private struct 2897 for communicating between the parent and subfunction. In the parent, 2898 copy in values for scalar and "small" structs; copy in addresses for 2899 others TREE_ADDRESSABLE types. In the subfunction, copy the value into 2900 the local variable. 2901 2902 It is not clear what to do with bare FOR or SECTION blocks. The only 2903 thing I can figure is that we do something like: 2904 2905 #pragma omp for firstprivate(x) lastprivate(y) 2906 for (int i = 0; i < n; ++i) 2907 body; 2908 2909 which becomes 2910 2911 { 2912 int x = x, y; 2913 2914 // for stuff 2915 2916 if (i == n) 2917 y = y; 2918 } 2919 2920 where the "x=x" and "y=y" assignments actually have different uids 2921 for the two variables, i.e. not something you could write directly in 2922 C. Presumably this only makes sense if the "outer" x and y are global 2923 variables. 2924 2925 COPYPRIVATE would work the same way, except the structure broadcast 2926 would have to happen via SINGLE machinery instead. 2927 2928 2929 File: libgomp.info, Node: Implementing REDUCTION clause, Next: Implementing PARALLEL construct, Prev: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Up: The libgomp ABI 2930 2931 9.9 Implementing REDUCTION clause 2932 ================================= 2933 2934 The private struct mentioned in the previous section should have a 2935 pointer to an array of the type of the variable, indexed by the thread's 2936 TEAM_ID. The thread stores its final value into the array, and after 2937 the barrier, the master thread iterates over the array to collect the 2938 values. 2939 2940 2941 File: libgomp.info, Node: Implementing PARALLEL construct, Next: Implementing FOR construct, Prev: Implementing REDUCTION clause, Up: The libgomp ABI 2942 2943 9.10 Implementing PARALLEL construct 2944 ==================================== 2945 2946 #pragma omp parallel 2947 { 2948 body; 2949 } 2950 2951 becomes 2952 2953 void subfunction (void *data) 2954 { 2955 use data; 2956 body; 2957 } 2958 2959 setup data; 2960 GOMP_parallel_start (subfunction, &data, num_threads); 2961 subfunction (&data); 2962 GOMP_parallel_end (); 2963 2964 void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads) 2965 2966 The FN argument is the subfunction to be run in parallel. 2967 2968 The DATA argument is a pointer to a structure used to communicate 2969 data in and out of the subfunction, as discussed above with respect to 2970 FIRSTPRIVATE et al. 2971 2972 The NUM_THREADS argument is 1 if an IF clause is present and false, 2973 or the value of the NUM_THREADS clause, if present, or 0. 2974 2975 The function needs to create the appropriate number of threads and/or 2976 launch them from the dock. It needs to create the team structure and 2977 assign team ids. 2978 2979 void GOMP_parallel_end (void) 2980 2981 Tears down the team and returns us to the previous 2982 'omp_in_parallel()' state. 2983 2984 2985 File: libgomp.info, Node: Implementing FOR construct, Next: Implementing ORDERED construct, Prev: Implementing PARALLEL construct, Up: The libgomp ABI 2986 2987 9.11 Implementing FOR construct 2988 =============================== 2989 2990 #pragma omp parallel for 2991 for (i = lb; i <= ub; i++) 2992 body; 2993 2994 becomes 2995 2996 void subfunction (void *data) 2997 { 2998 long _s0, _e0; 2999 while (GOMP_loop_static_next (&_s0, &_e0)) 3000 { 3001 long _e1 = _e0, i; 3002 for (i = _s0; i < _e1; i++) 3003 body; 3004 } 3005 GOMP_loop_end_nowait (); 3006 } 3007 3008 GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0); 3009 subfunction (NULL); 3010 GOMP_parallel_end (); 3011 3012 #pragma omp for schedule(runtime) 3013 for (i = 0; i < n; i++) 3014 body; 3015 3016 becomes 3017 3018 { 3019 long i, _s0, _e0; 3020 if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0)) 3021 do { 3022 long _e1 = _e0; 3023 for (i = _s0, i < _e0; i++) 3024 body; 3025 } while (GOMP_loop_runtime_next (&_s0, _&e0)); 3026 GOMP_loop_end (); 3027 } 3028 3029 Note that while it looks like there is trickiness to propagating a 3030 non-constant STEP, there isn't really. We're explicitly allowed to 3031 evaluate it as many times as we want, and any variables involved should 3032 automatically be handled as PRIVATE or SHARED like any other variables. 3033 So the expression should remain evaluable in the subfunction. We can 3034 also pull it into a local variable if we like, but since its supposed to 3035 remain unchanged, we can also not if we like. 3036 3037 If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be able 3038 to get away with no work-sharing context at all, since we can simply 3039 perform the arithmetic directly in each thread to divide up the 3040 iterations. Which would mean that we wouldn't need to call any of these 3041 routines. 3042 3043 There are separate routines for handling loops with an ORDERED 3044 clause. Bookkeeping for that is non-trivial... 3045 3046 3047 File: libgomp.info, Node: Implementing ORDERED construct, Next: Implementing SECTIONS construct, Prev: Implementing FOR construct, Up: The libgomp ABI 3048 3049 9.12 Implementing ORDERED construct 3050 =================================== 3051 3052 void GOMP_ordered_start (void) 3053 void GOMP_ordered_end (void) 3054 3055 3056 File: libgomp.info, Node: Implementing SECTIONS construct, Next: Implementing SINGLE construct, Prev: Implementing ORDERED construct, Up: The libgomp ABI 3057 3058 9.13 Implementing SECTIONS construct 3059 ==================================== 3060 3061 A block as 3062 3063 #pragma omp sections 3064 { 3065 #pragma omp section 3066 stmt1; 3067 #pragma omp section 3068 stmt2; 3069 #pragma omp section 3070 stmt3; 3071 } 3072 3073 becomes 3074 3075 for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ()) 3076 switch (i) 3077 { 3078 case 1: 3079 stmt1; 3080 break; 3081 case 2: 3082 stmt2; 3083 break; 3084 case 3: 3085 stmt3; 3086 break; 3087 } 3088 GOMP_barrier (); 3089 3090 3091 File: libgomp.info, Node: Implementing SINGLE construct, Next: Implementing OpenACC's PARALLEL construct, Prev: Implementing SECTIONS construct, Up: The libgomp ABI 3092 3093 9.14 Implementing SINGLE construct 3094 ================================== 3095 3096 A block like 3097 3098 #pragma omp single 3099 { 3100 body; 3101 } 3102 3103 becomes 3104 3105 if (GOMP_single_start ()) 3106 body; 3107 GOMP_barrier (); 3108 3109 while 3110 3111 #pragma omp single copyprivate(x) 3112 body; 3113 3114 becomes 3115 3116 datap = GOMP_single_copy_start (); 3117 if (datap == NULL) 3118 { 3119 body; 3120 data.x = x; 3121 GOMP_single_copy_end (&data); 3122 } 3123 else 3124 x = datap->x; 3125 GOMP_barrier (); 3126 3127 3128 File: libgomp.info, Node: Implementing OpenACC's PARALLEL construct, Prev: Implementing SINGLE construct, Up: The libgomp ABI 3129 3130 9.15 Implementing OpenACC's PARALLEL construct 3131 ============================================== 3132 3133 void GOACC_parallel () 3134 3135 3136 File: libgomp.info, Node: Reporting Bugs, Next: Copying, Prev: The libgomp ABI, Up: Top 3137 3138 10 Reporting Bugs 3139 ***************** 3140 3141 Bugs in the GNU Offloading and Multi Processing Runtime Library should 3142 be reported via Bugzilla (http://gcc.gnu.org/bugzilla/). Please add 3143 "openacc", or "openmp", or both to the keywords field in the bug report, 3144 as appropriate. 3145 3146 3147 File: libgomp.info, Node: Copying, Next: GNU Free Documentation License, Prev: Reporting Bugs, Up: Top 3148 3149 GNU General Public License 3150 ************************** 3151 3152 Version 3, 29 June 2007 3153 3154 Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> 3155 3156 Everyone is permitted to copy and distribute verbatim copies of this 3157 license document, but changing it is not allowed. 3158 3159 Preamble 3160 ======== 3161 3162 The GNU General Public License is a free, copyleft license for software 3163 and other kinds of works. 3164 3165 The licenses for most software and other practical works are designed 3166 to take away your freedom to share and change the works. By contrast, 3167 the GNU General Public License is intended to guarantee your freedom to 3168 share and change all versions of a program-to make sure it remains free 3169 software for all its users. We, the Free Software Foundation, use the 3170 GNU General Public License for most of our software; it applies also to 3171 any other work released this way by its authors. You can apply it to 3172 your programs, too. 3173 3174 When we speak of free software, we are referring to freedom, not 3175 price. Our General Public Licenses are designed to make sure that you 3176 have the freedom to distribute copies of free software (and charge for 3177 them if you wish), that you receive source code or can get it if you 3178 want it, that you can change the software or use pieces of it in new 3179 free programs, and that you know you can do these things. 3180 3181 To protect your rights, we need to prevent others from denying you 3182 these rights or asking you to surrender the rights. Therefore, you have 3183 certain responsibilities if you distribute copies of the software, or if 3184 you modify it: responsibilities to respect the freedom of others. 3185 3186 For example, if you distribute copies of such a program, whether 3187 gratis or for a fee, you must pass on to the recipients the same 3188 freedoms that you received. You must make sure that they, too, receive 3189 or can get the source code. And you must show them these terms so they 3190 know their rights. 3191 3192 Developers that use the GNU GPL protect your rights with two steps: 3193 (1) assert copyright on the software, and (2) offer you this License 3194 giving you legal permission to copy, distribute and/or modify it. 3195 3196 For the developers' and authors' protection, the GPL clearly explains 3197 that there is no warranty for this free software. For both users' and 3198 authors' sake, the GPL requires that modified versions be marked as 3199 changed, so that their problems will not be attributed erroneously to 3200 authors of previous versions. 3201 3202 Some devices are designed to deny users access to install or run 3203 modified versions of the software inside them, although the manufacturer 3204 can do so. This is fundamentally incompatible with the aim of 3205 protecting users' freedom to change the software. The systematic 3206 pattern of such abuse occurs in the area of products for individuals to 3207 use, which is precisely where it is most unacceptable. Therefore, we 3208 have designed this version of the GPL to prohibit the practice for those 3209 products. If such problems arise substantially in other domains, we 3210 stand ready to extend this provision to those domains in future versions 3211 of the GPL, as needed to protect the freedom of users. 3212 3213 Finally, every program is threatened constantly by software patents. 3214 States should not allow patents to restrict development and use of 3215 software on general-purpose computers, but in those that do, we wish to 3216 avoid the special danger that patents applied to a free program could 3217 make it effectively proprietary. To prevent this, the GPL assures that 3218 patents cannot be used to render the program non-free. 3219 3220 The precise terms and conditions for copying, distribution and 3221 modification follow. 3222 3223 TERMS AND CONDITIONS 3224 ==================== 3225 3226 0. Definitions. 3227 3228 "This License" refers to version 3 of the GNU General Public 3229 License. 3230 3231 "Copyright" also means copyright-like laws that apply to other 3232 kinds of works, such as semiconductor masks. 3233 3234 "The Program" refers to any copyrightable work licensed under this 3235 License. Each licensee is addressed as "you". "Licensees" and 3236 "recipients" may be individuals or organizations. 3237 3238 To "modify" a work means to copy from or adapt all or part of the 3239 work in a fashion requiring copyright permission, other than the 3240 making of an exact copy. The resulting work is called a "modified 3241 version" of the earlier work or a work "based on" the earlier work. 3242 3243 A "covered work" means either the unmodified Program or a work 3244 based on the Program. 3245 3246 To "propagate" a work means to do anything with it that, without 3247 permission, would make you directly or secondarily liable for 3248 infringement under applicable copyright law, except executing it on 3249 a computer or modifying a private copy. Propagation includes 3250 copying, distribution (with or without modification), making 3251 available to the public, and in some countries other activities as 3252 well. 3253 3254 To "convey" a work means any kind of propagation that enables other 3255 parties to make or receive copies. Mere interaction with a user 3256 through a computer network, with no transfer of a copy, is not 3257 conveying. 3258 3259 An interactive user interface displays "Appropriate Legal Notices" 3260 to the extent that it includes a convenient and prominently visible 3261 feature that (1) displays an appropriate copyright notice, and (2) 3262 tells the user that there is no warranty for the work (except to 3263 the extent that warranties are provided), that licensees may convey 3264 the work under this License, and how to view a copy of this 3265 License. If the interface presents a list of user commands or 3266 options, such as a menu, a prominent item in the list meets this 3267 criterion. 3268 3269 1. Source Code. 3270 3271 The "source code" for a work means the preferred form of the work 3272 for making modifications to it. "Object code" means any non-source 3273 form of a work. 3274 3275 A "Standard Interface" means an interface that either is an 3276 official standard defined by a recognized standards body, or, in 3277 the case of interfaces specified for a particular programming 3278 language, one that is widely used among developers working in that 3279 language. 3280 3281 The "System Libraries" of an executable work include anything, 3282 other than the work as a whole, that (a) is included in the normal 3283 form of packaging a Major Component, but which is not part of that 3284 Major Component, and (b) serves only to enable use of the work with 3285 that Major Component, or to implement a Standard Interface for 3286 which an implementation is available to the public in source code 3287 form. A "Major Component", in this context, means a major 3288 essential component (kernel, window system, and so on) of the 3289 specific operating system (if any) on which the executable work 3290 runs, or a compiler used to produce the work, or an object code 3291 interpreter used to run it. 3292 3293 The "Corresponding Source" for a work in object code form means all 3294 the source code needed to generate, install, and (for an executable 3295 work) run the object code and to modify the work, including scripts 3296 to control those activities. However, it does not include the 3297 work's System Libraries, or general-purpose tools or generally 3298 available free programs which are used unmodified in performing 3299 those activities but which are not part of the work. For example, 3300 Corresponding Source includes interface definition files associated 3301 with source files for the work, and the source code for shared 3302 libraries and dynamically linked subprograms that the work is 3303 specifically designed to require, such as by intimate data 3304 communication or control flow between those subprograms and other 3305 parts of the work. 3306 3307 The Corresponding Source need not include anything that users can 3308 regenerate automatically from other parts of the Corresponding 3309 Source. 3310 3311 The Corresponding Source for a work in source code form is that 3312 same work. 3313 3314 2. Basic Permissions. 3315 3316 All rights granted under this License are granted for the term of 3317 copyright on the Program, and are irrevocable provided the stated 3318 conditions are met. This License explicitly affirms your unlimited 3319 permission to run the unmodified Program. The output from running 3320 a covered work is covered by this License only if the output, given 3321 its content, constitutes a covered work. This License acknowledges 3322 your rights of fair use or other equivalent, as provided by 3323 copyright law. 3324 3325 You may make, run and propagate covered works that you do not 3326 convey, without conditions so long as your license otherwise 3327 remains in force. You may convey covered works to others for the 3328 sole purpose of having them make modifications exclusively for you, 3329 or provide you with facilities for running those works, provided 3330 that you comply with the terms of this License in conveying all 3331 material for which you do not control copyright. Those thus making 3332 or running the covered works for you must do so exclusively on your 3333 behalf, under your direction and control, on terms that prohibit 3334 them from making any copies of your copyrighted material outside 3335 their relationship with you. 3336 3337 Conveying under any other circumstances is permitted solely under 3338 the conditions stated below. Sublicensing is not allowed; section 3339 10 makes it unnecessary. 3340 3341 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 3342 3343 No covered work shall be deemed part of an effective technological 3344 measure under any applicable law fulfilling obligations under 3345 article 11 of the WIPO copyright treaty adopted on 20 December 3346 1996, or similar laws prohibiting or restricting circumvention of 3347 such measures. 3348 3349 When you convey a covered work, you waive any legal power to forbid 3350 circumvention of technological measures to the extent such 3351 circumvention is effected by exercising rights under this License 3352 with respect to the covered work, and you disclaim any intention to 3353 limit operation or modification of the work as a means of 3354 enforcing, against the work's users, your or third parties' legal 3355 rights to forbid circumvention of technological measures. 3356 3357 4. Conveying Verbatim Copies. 3358 3359 You may convey verbatim copies of the Program's source code as you 3360 receive it, in any medium, provided that you conspicuously and 3361 appropriately publish on each copy an appropriate copyright notice; 3362 keep intact all notices stating that this License and any 3363 non-permissive terms added in accord with section 7 apply to the 3364 code; keep intact all notices of the absence of any warranty; and 3365 give all recipients a copy of this License along with the Program. 3366 3367 You may charge any price or no price for each copy that you convey, 3368 and you may offer support or warranty protection for a fee. 3369 3370 5. Conveying Modified Source Versions. 3371 3372 You may convey a work based on the Program, or the modifications to 3373 produce it from the Program, in the form of source code under the 3374 terms of section 4, provided that you also meet all of these 3375 conditions: 3376 3377 a. The work must carry prominent notices stating that you 3378 modified it, and giving a relevant date. 3379 3380 b. The work must carry prominent notices stating that it is 3381 released under this License and any conditions added under 3382 section 7. This requirement modifies the requirement in 3383 section 4 to "keep intact all notices". 3384 3385 c. You must license the entire work, as a whole, under this 3386 License to anyone who comes into possession of a copy. This 3387 License will therefore apply, along with any applicable 3388 section 7 additional terms, to the whole of the work, and all 3389 its parts, regardless of how they are packaged. This License 3390 gives no permission to license the work in any other way, but 3391 it does not invalidate such permission if you have separately 3392 received it. 3393 3394 d. If the work has interactive user interfaces, each must display 3395 Appropriate Legal Notices; however, if the Program has 3396 interactive interfaces that do not display Appropriate Legal 3397 Notices, your work need not make them do so. 3398 3399 A compilation of a covered work with other separate and independent 3400 works, which are not by their nature extensions of the covered 3401 work, and which are not combined with it such as to form a larger 3402 program, in or on a volume of a storage or distribution medium, is 3403 called an "aggregate" if the compilation and its resulting 3404 copyright are not used to limit the access or legal rights of the 3405 compilation's users beyond what the individual works permit. 3406 Inclusion of a covered work in an aggregate does not cause this 3407 License to apply to the other parts of the aggregate. 3408 3409 6. Conveying Non-Source Forms. 3410 3411 You may convey a covered work in object code form under the terms 3412 of sections 4 and 5, provided that you also convey the 3413 machine-readable Corresponding Source under the terms of this 3414 License, in one of these ways: 3415 3416 a. Convey the object code in, or embodied in, a physical product 3417 (including a physical distribution medium), accompanied by the 3418 Corresponding Source fixed on a durable physical medium 3419 customarily used for software interchange. 3420 3421 b. Convey the object code in, or embodied in, a physical product 3422 (including a physical distribution medium), accompanied by a 3423 written offer, valid for at least three years and valid for as 3424 long as you offer spare parts or customer support for that 3425 product model, to give anyone who possesses the object code 3426 either (1) a copy of the Corresponding Source for all the 3427 software in the product that is covered by this License, on a 3428 durable physical medium customarily used for software 3429 interchange, for a price no more than your reasonable cost of 3430 physically performing this conveying of source, or (2) access 3431 to copy the Corresponding Source from a network server at no 3432 charge. 3433 3434 c. Convey individual copies of the object code with a copy of the 3435 written offer to provide the Corresponding Source. This 3436 alternative is allowed only occasionally and noncommercially, 3437 and only if you received the object code with such an offer, 3438 in accord with subsection 6b. 3439 3440 d. Convey the object code by offering access from a designated 3441 place (gratis or for a charge), and offer equivalent access to 3442 the Corresponding Source in the same way through the same 3443 place at no further charge. You need not require recipients 3444 to copy the Corresponding Source along with the object code. 3445 If the place to copy the object code is a network server, the 3446 Corresponding Source may be on a different server (operated by 3447 you or a third party) that supports equivalent copying 3448 facilities, provided you maintain clear directions next to the 3449 object code saying where to find the Corresponding Source. 3450 Regardless of what server hosts the Corresponding Source, you 3451 remain obligated to ensure that it is available for as long as 3452 needed to satisfy these requirements. 3453 3454 e. Convey the object code using peer-to-peer transmission, 3455 provided you inform other peers where the object code and 3456 Corresponding Source of the work are being offered to the 3457 general public at no charge under subsection 6d. 3458 3459 A separable portion of the object code, whose source code is 3460 excluded from the Corresponding Source as a System Library, need 3461 not be included in conveying the object code work. 3462 3463 A "User Product" is either (1) a "consumer product", which means 3464 any tangible personal property which is normally used for personal, 3465 family, or household purposes, or (2) anything designed or sold for 3466 incorporation into a dwelling. In determining whether a product is 3467 a consumer product, doubtful cases shall be resolved in favor of 3468 coverage. For a particular product received by a particular user, 3469 "normally used" refers to a typical or common use of that class of 3470 product, regardless of the status of the particular user or of the 3471 way in which the particular user actually uses, or expects or is 3472 expected to use, the product. A product is a consumer product 3473 regardless of whether the product has substantial commercial, 3474 industrial or non-consumer uses, unless such uses represent the 3475 only significant mode of use of the product. 3476 3477 "Installation Information" for a User Product means any methods, 3478 procedures, authorization keys, or other information required to 3479 install and execute modified versions of a covered work in that 3480 User Product from a modified version of its Corresponding Source. 3481 The information must suffice to ensure that the continued 3482 functioning of the modified object code is in no case prevented or 3483 interfered with solely because modification has been made. 3484 3485 If you convey an object code work under this section in, or with, 3486 or specifically for use in, a User Product, and the conveying 3487 occurs as part of a transaction in which the right of possession 3488 and use of the User Product is transferred to the recipient in 3489 perpetuity or for a fixed term (regardless of how the transaction 3490 is characterized), the Corresponding Source conveyed under this 3491 section must be accompanied by the Installation Information. But 3492 this requirement does not apply if neither you nor any third party 3493 retains the ability to install modified object code on the User 3494 Product (for example, the work has been installed in ROM). 3495 3496 The requirement to provide Installation Information does not 3497 include a requirement to continue to provide support service, 3498 warranty, or updates for a work that has been modified or installed 3499 by the recipient, or for the User Product in which it has been 3500 modified or installed. Access to a network may be denied when the 3501 modification itself materially and adversely affects the operation 3502 of the network or violates the rules and protocols for 3503 communication across the network. 3504 3505 Corresponding Source conveyed, and Installation Information 3506 provided, in accord with this section must be in a format that is 3507 publicly documented (and with an implementation available to the 3508 public in source code form), and must require no special password 3509 or key for unpacking, reading or copying. 3510 3511 7. Additional Terms. 3512 3513 "Additional permissions" are terms that supplement the terms of 3514 this License by making exceptions from one or more of its 3515 conditions. Additional permissions that are applicable to the 3516 entire Program shall be treated as though they were included in 3517 this License, to the extent that they are valid under applicable 3518 law. If additional permissions apply only to part of the Program, 3519 that part may be used separately under those permissions, but the 3520 entire Program remains governed by this License without regard to 3521 the additional permissions. 3522 3523 When you convey a copy of a covered work, you may at your option 3524 remove any additional permissions from that copy, or from any part 3525 of it. (Additional permissions may be written to require their own 3526 removal in certain cases when you modify the work.) You may place 3527 additional permissions on material, added by you to a covered work, 3528 for which you have or can give appropriate copyright permission. 3529 3530 Notwithstanding any other provision of this License, for material 3531 you add to a covered work, you may (if authorized by the copyright 3532 holders of that material) supplement the terms of this License with 3533 terms: 3534 3535 a. Disclaiming warranty or limiting liability differently from 3536 the terms of sections 15 and 16 of this License; or 3537 3538 b. Requiring preservation of specified reasonable legal notices 3539 or author attributions in that material or in the Appropriate 3540 Legal Notices displayed by works containing it; or 3541 3542 c. Prohibiting misrepresentation of the origin of that material, 3543 or requiring that modified versions of such material be marked 3544 in reasonable ways as different from the original version; or 3545 3546 d. Limiting the use for publicity purposes of names of licensors 3547 or authors of the material; or 3548 3549 e. Declining to grant rights under trademark law for use of some 3550 trade names, trademarks, or service marks; or 3551 3552 f. Requiring indemnification of licensors and authors of that 3553 material by anyone who conveys the material (or modified 3554 versions of it) with contractual assumptions of liability to 3555 the recipient, for any liability that these contractual 3556 assumptions directly impose on those licensors and authors. 3557 3558 All other non-permissive additional terms are considered "further 3559 restrictions" within the meaning of section 10. If the Program as 3560 you received it, or any part of it, contains a notice stating that 3561 it is governed by this License along with a term that is a further 3562 restriction, you may remove that term. If a license document 3563 contains a further restriction but permits relicensing or conveying 3564 under this License, you may add to a covered work material governed 3565 by the terms of that license document, provided that the further 3566 restriction does not survive such relicensing or conveying. 3567 3568 If you add terms to a covered work in accord with this section, you 3569 must place, in the relevant source files, a statement of the 3570 additional terms that apply to those files, or a notice indicating 3571 where to find the applicable terms. 3572 3573 Additional terms, permissive or non-permissive, may be stated in 3574 the form of a separately written license, or stated as exceptions; 3575 the above requirements apply either way. 3576 3577 8. Termination. 3578 3579 You may not propagate or modify a covered work except as expressly 3580 provided under this License. Any attempt otherwise to propagate or 3581 modify it is void, and will automatically terminate your rights 3582 under this License (including any patent licenses granted under the 3583 third paragraph of section 11). 3584 3585 However, if you cease all violation of this License, then your 3586 license from a particular copyright holder is reinstated (a) 3587 provisionally, unless and until the copyright holder explicitly and 3588 finally terminates your license, and (b) permanently, if the 3589 copyright holder fails to notify you of the violation by some 3590 reasonable means prior to 60 days after the cessation. 3591 3592 Moreover, your license from a particular copyright holder is 3593 reinstated permanently if the copyright holder notifies you of the 3594 violation by some reasonable means, this is the first time you have 3595 received notice of violation of this License (for any work) from 3596 that copyright holder, and you cure the violation prior to 30 days 3597 after your receipt of the notice. 3598 3599 Termination of your rights under this section does not terminate 3600 the licenses of parties who have received copies or rights from you 3601 under this License. If your rights have been terminated and not 3602 permanently reinstated, you do not qualify to receive new licenses 3603 for the same material under section 10. 3604 3605 9. Acceptance Not Required for Having Copies. 3606 3607 You are not required to accept this License in order to receive or 3608 run a copy of the Program. Ancillary propagation of a covered work 3609 occurring solely as a consequence of using peer-to-peer 3610 transmission to receive a copy likewise does not require 3611 acceptance. However, nothing other than this License grants you 3612 permission to propagate or modify any covered work. These actions 3613 infringe copyright if you do not accept this License. Therefore, 3614 by modifying or propagating a covered work, you indicate your 3615 acceptance of this License to do so. 3616 3617 10. Automatic Licensing of Downstream Recipients. 3618 3619 Each time you convey a covered work, the recipient automatically 3620 receives a license from the original licensors, to run, modify and 3621 propagate that work, subject to this License. You are not 3622 responsible for enforcing compliance by third parties with this 3623 License. 3624 3625 An "entity transaction" is a transaction transferring control of an 3626 organization, or substantially all assets of one, or subdividing an 3627 organization, or merging organizations. If propagation of a 3628 covered work results from an entity transaction, each party to that 3629 transaction who receives a copy of the work also receives whatever 3630 licenses to the work the party's predecessor in interest had or 3631 could give under the previous paragraph, plus a right to possession 3632 of the Corresponding Source of the work from the predecessor in 3633 interest, if the predecessor has it or can get it with reasonable 3634 efforts. 3635 3636 You may not impose any further restrictions on the exercise of the 3637 rights granted or affirmed under this License. For example, you 3638 may not impose a license fee, royalty, or other charge for exercise 3639 of rights granted under this License, and you may not initiate 3640 litigation (including a cross-claim or counterclaim in a lawsuit) 3641 alleging that any patent claim is infringed by making, using, 3642 selling, offering for sale, or importing the Program or any portion 3643 of it. 3644 3645 11. Patents. 3646 3647 A "contributor" is a copyright holder who authorizes use under this 3648 License of the Program or a work on which the Program is based. 3649 The work thus licensed is called the contributor's "contributor 3650 version". 3651 3652 A contributor's "essential patent claims" are all patent claims 3653 owned or controlled by the contributor, whether already acquired or 3654 hereafter acquired, that would be infringed by some manner, 3655 permitted by this License, of making, using, or selling its 3656 contributor version, but do not include claims that would be 3657 infringed only as a consequence of further modification of the 3658 contributor version. For purposes of this definition, "control" 3659 includes the right to grant patent sublicenses in a manner 3660 consistent with the requirements of this License. 3661 3662 Each contributor grants you a non-exclusive, worldwide, 3663 royalty-free patent license under the contributor's essential 3664 patent claims, to make, use, sell, offer for sale, import and 3665 otherwise run, modify and propagate the contents of its contributor 3666 version. 3667 3668 In the following three paragraphs, a "patent license" is any 3669 express agreement or commitment, however denominated, not to 3670 enforce a patent (such as an express permission to practice a 3671 patent or covenant not to sue for patent infringement). To "grant" 3672 such a patent license to a party means to make such an agreement or 3673 commitment not to enforce a patent against the party. 3674 3675 If you convey a covered work, knowingly relying on a patent 3676 license, and the Corresponding Source of the work is not available 3677 for anyone to copy, free of charge and under the terms of this 3678 License, through a publicly available network server or other 3679 readily accessible means, then you must either (1) cause the 3680 Corresponding Source to be so available, or (2) arrange to deprive 3681 yourself of the benefit of the patent license for this particular 3682 work, or (3) arrange, in a manner consistent with the requirements 3683 of this License, to extend the patent license to downstream 3684 recipients. "Knowingly relying" means you have actual knowledge 3685 that, but for the patent license, your conveying the covered work 3686 in a country, or your recipient's use of the covered work in a 3687 country, would infringe one or more identifiable patents in that 3688 country that you have reason to believe are valid. 3689 3690 If, pursuant to or in connection with a single transaction or 3691 arrangement, you convey, or propagate by procuring conveyance of, a 3692 covered work, and grant a patent license to some of the parties 3693 receiving the covered work authorizing them to use, propagate, 3694 modify or convey a specific copy of the covered work, then the 3695 patent license you grant is automatically extended to all 3696 recipients of the covered work and works based on it. 3697 3698 A patent license is "discriminatory" if it does not include within 3699 the scope of its coverage, prohibits the exercise of, or is 3700 conditioned on the non-exercise of one or more of the rights that 3701 are specifically granted under this License. You may not convey a 3702 covered work if you are a party to an arrangement with a third 3703 party that is in the business of distributing software, under which 3704 you make payment to the third party based on the extent of your 3705 activity of conveying the work, and under which the third party 3706 grants, to any of the parties who would receive the covered work 3707 from you, a discriminatory patent license (a) in connection with 3708 copies of the covered work conveyed by you (or copies made from 3709 those copies), or (b) primarily for and in connection with specific 3710 products or compilations that contain the covered work, unless you 3711 entered into that arrangement, or that patent license was granted, 3712 prior to 28 March 2007. 3713 3714 Nothing in this License shall be construed as excluding or limiting 3715 any implied license or other defenses to infringement that may 3716 otherwise be available to you under applicable patent law. 3717 3718 12. No Surrender of Others' Freedom. 3719 3720 If conditions are imposed on you (whether by court order, agreement 3721 or otherwise) that contradict the conditions of this License, they 3722 do not excuse you from the conditions of this License. If you 3723 cannot convey a covered work so as to satisfy simultaneously your 3724 obligations under this License and any other pertinent obligations, 3725 then as a consequence you may not convey it at all. For example, 3726 if you agree to terms that obligate you to collect a royalty for 3727 further conveying from those to whom you convey the Program, the 3728 only way you could satisfy both those terms and this License would 3729 be to refrain entirely from conveying the Program. 3730 3731 13. Use with the GNU Affero General Public License. 3732 3733 Notwithstanding any other provision of this License, you have 3734 permission to link or combine any covered work with a work licensed 3735 under version 3 of the GNU Affero General Public License into a 3736 single combined work, and to convey the resulting work. The terms 3737 of this License will continue to apply to the part which is the 3738 covered work, but the special requirements of the GNU Affero 3739 General Public License, section 13, concerning interaction through 3740 a network will apply to the combination as such. 3741 3742 14. Revised Versions of this License. 3743 3744 The Free Software Foundation may publish revised and/or new 3745 versions of the GNU General Public License from time to time. Such 3746 new versions will be similar in spirit to the present version, but 3747 may differ in detail to address new problems or 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 Free 3754 Software Foundation. If the Program does not specify a version 3755 number of the GNU General Public License, you may choose any 3756 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 FOR 3785 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 END OF TERMS AND CONDITIONS 3803 =========================== 3804 3805 How to Apply These Terms to Your New Programs 3806 ============================================= 3807 3808 If you develop a new program, and you want it to be of the greatest 3809 possible use to the public, the best way to achieve this is to make it 3810 free software which everyone can redistribute and change under these 3811 terms. 3812 3813 To do so, attach the following notices to the program. It is safest 3814 to attach them to the start of each source file to most effectively 3815 state the exclusion of warranty; and each file should have at least the 3816 "copyright" line and a pointer to where the full notice is found. 3817 3818 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 3819 Copyright (C) YEAR NAME OF AUTHOR 3820 3821 This program is free software: you can redistribute it and/or modify 3822 it under the terms of the GNU General Public License as published by 3823 the Free Software Foundation, either version 3 of the License, or (at 3824 your option) any later version. 3825 3826 This program is distributed in the hope that it will be useful, but 3827 WITHOUT ANY WARRANTY; without even the implied warranty of 3828 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 3829 General Public License for more details. 3830 3831 You should have received a copy of the GNU General Public License 3832 along with this program. If not, see <http://www.gnu.org/licenses/>. 3833 3834 Also add information on how to contact you by electronic and paper 3835 mail. 3836 3837 If the program does terminal interaction, make it output a short 3838 notice like this when it starts in an interactive mode: 3839 3840 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 3841 This program comes with ABSOLUTELY NO WARRANTY; for details type 'show w'. 3842 This is free software, and you are welcome to redistribute it 3843 under certain conditions; type 'show c' for details. 3844 3845 The hypothetical commands 'show w' and 'show c' should show the 3846 appropriate parts of the General Public License. Of course, your 3847 program's commands might be different; for a GUI interface, you would 3848 use an "about box". 3849 3850 You should also get your employer (if you work as a programmer) or 3851 school, if any, to sign a "copyright disclaimer" for the program, if 3852 necessary. For more information on this, and how to apply and follow 3853 the GNU GPL, see <http://www.gnu.org/licenses/>. 3854 3855 The GNU General Public License does not permit incorporating your 3856 program into proprietary programs. If your program is a subroutine 3857 library, you may consider it more useful to permit linking proprietary 3858 applications with the library. If this is what you want to do, use the 3859 GNU Lesser General Public License instead of this License. But first, 3860 please read <http://www.gnu.org/philosophy/why-not-lgpl.html>. 3861 3862 3863 File: libgomp.info, Node: GNU Free Documentation License, Next: Funding, Prev: Copying, Up: Top 3864 3865 GNU Free Documentation License 3866 ****************************** 3867 3868 Version 1.3, 3 November 2008 3869 3870 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. 3871 <http://fsf.org/> 3872 3873 Everyone is permitted to copy and distribute verbatim copies 3874 of this license document, but changing it is not allowed. 3875 3876 0. PREAMBLE 3877 3878 The purpose of this License is to make a manual, textbook, or other 3879 functional and useful document "free" in the sense of freedom: to 3880 assure everyone the effective freedom to copy and redistribute it, 3881 with or without modifying it, either commercially or 3882 noncommercially. Secondarily, this License preserves for the 3883 author and publisher a way to get credit for their work, while not 3884 being considered responsible for modifications made by others. 3885 3886 This License is a kind of "copyleft", which means that derivative 3887 works of the document must themselves be free in the same sense. 3888 It complements the GNU General Public License, which is a copyleft 3889 license designed for free software. 3890 3891 We have designed this License in order to use it for manuals for 3892 free software, because free software needs free documentation: a 3893 free program should come with manuals providing the same freedoms 3894 that the software does. But this License is not limited to 3895 software manuals; it can be used for any textual work, regardless 3896 of subject matter or whether it is published as a printed book. We 3897 recommend this License principally for works whose purpose is 3898 instruction or reference. 3899 3900 1. APPLICABILITY AND DEFINITIONS 3901 3902 This License applies to any manual or other work, in any medium, 3903 that contains a notice placed by the copyright holder saying it can 3904 be distributed under the terms of this License. Such a notice 3905 grants a world-wide, royalty-free license, unlimited in duration, 3906 to use that work under the conditions stated herein. The 3907 "Document", below, refers to any such manual or work. Any member 3908 of the public is a licensee, and is addressed as "you". You accept 3909 the license if you copy, modify or distribute the work in a way 3910 requiring permission under copyright law. 3911 3912 A "Modified Version" of the Document means any work containing the 3913 Document or a portion of it, either copied verbatim, or with 3914 modifications and/or translated into another language. 3915 3916 A "Secondary Section" is a named appendix or a front-matter section 3917 of the Document that deals exclusively with the relationship of the 3918 publishers or authors of the Document to the Document's overall 3919 subject (or to related matters) and contains nothing that could 3920 fall directly within that overall subject. (Thus, if the Document 3921 is in part a textbook of mathematics, a Secondary Section may not 3922 explain any mathematics.) The relationship could be a matter of 3923 historical connection with the subject or with related matters, or 3924 of legal, commercial, philosophical, ethical or political position 3925 regarding them. 3926 3927 The "Invariant Sections" are certain Secondary Sections whose 3928 titles are designated, as being those of Invariant Sections, in the 3929 notice that says that the Document is released under this License. 3930 If a section does not fit the above definition of Secondary then it 3931 is not allowed to be designated as Invariant. The Document may 3932 contain zero Invariant Sections. If the Document does not identify 3933 any Invariant Sections then there are none. 3934 3935 The "Cover Texts" are certain short passages of text that are 3936 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 3937 that says that the Document is released under this License. A 3938 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 3939 be at most 25 words. 3940 3941 A "Transparent" copy of the Document means a machine-readable copy, 3942 represented in a format whose specification is available to the 3943 general public, that is suitable for revising the document 3944 straightforwardly with generic text editors or (for images composed 3945 of pixels) generic paint programs or (for drawings) some widely 3946 available drawing editor, and that is suitable for input to text 3947 formatters or for automatic translation to a variety of formats 3948 suitable for input to text formatters. A copy made in an otherwise 3949 Transparent file format whose markup, or absence of markup, has 3950 been arranged to thwart or discourage subsequent modification by 3951 readers is not Transparent. An image format is not Transparent if 3952 used for any substantial amount of text. A copy that is not 3953 "Transparent" is called "Opaque". 3954 3955 Examples of suitable formats for Transparent copies include plain 3956 ASCII without markup, Texinfo input format, LaTeX input format, 3957 SGML or XML using a publicly available DTD, and standard-conforming 3958 simple HTML, PostScript or PDF designed for human modification. 3959 Examples of transparent image formats include PNG, XCF and JPG. 3960 Opaque formats include proprietary formats that can be read and 3961 edited only by proprietary word processors, SGML or XML for which 3962 the DTD and/or processing tools are not generally available, and 3963 the machine-generated HTML, PostScript or PDF produced by some word 3964 processors for output purposes only. 3965 3966 The "Title Page" means, for a printed book, the title page itself, 3967 plus such following pages as are needed to hold, legibly, the 3968 material this License requires to appear in the title page. For 3969 works in formats which do not have any title page as such, "Title 3970 Page" means the text near the most prominent appearance of the 3971 work's title, preceding the beginning of the body of the text. 3972 3973 The "publisher" means any person or entity that distributes copies 3974 of the Document to the public. 3975 3976 A section "Entitled XYZ" means a named subunit of the Document 3977 whose title either is precisely XYZ or contains XYZ in parentheses 3978 following text that translates XYZ in another language. (Here XYZ 3979 stands for a specific section name mentioned below, such as 3980 "Acknowledgements", "Dedications", "Endorsements", or "History".) 3981 To "Preserve the Title" of such a section when you modify the 3982 Document means that it remains a section "Entitled XYZ" according 3983 to this definition. 3984 3985 The Document may include Warranty Disclaimers next to the notice 3986 which states that this License applies to the Document. These 3987 Warranty Disclaimers are considered to be included by reference in 3988 this License, but only as regards disclaiming warranties: any other 3989 implication that these Warranty Disclaimers may have is void and 3990 has no effect on the meaning of this License. 3991 3992 2. VERBATIM COPYING 3993 3994 You may copy and distribute the Document in any medium, either 3995 commercially or noncommercially, provided that this License, the 3996 copyright notices, and the license notice saying this License 3997 applies to the Document are reproduced in all copies, and that you 3998 add no other conditions whatsoever to those of this License. You 3999 may not use technical measures to obstruct or control the reading 4000 or further copying of the copies you make or distribute. However, 4001 you may accept compensation in exchange for copies. If you 4002 distribute a large enough number of copies you must also follow the 4003 conditions in section 3. 4004 4005 You may also lend copies, under the same conditions stated above, 4006 and you may publicly display copies. 4007 4008 3. COPYING IN QUANTITY 4009 4010 If you publish printed copies (or copies in media that commonly 4011 have printed covers) of the Document, numbering more than 100, and 4012 the Document's license notice requires Cover Texts, you must 4013 enclose the copies in covers that carry, clearly and legibly, all 4014 these Cover Texts: Front-Cover Texts on the front cover, and 4015 Back-Cover Texts on the back cover. Both covers must also clearly 4016 and legibly identify you as the publisher of these copies. The 4017 front cover must present the full title with all words of the title 4018 equally prominent and visible. You may add other material on the 4019 covers in addition. Copying with changes limited to the covers, as 4020 long as they preserve the title of the Document and satisfy these 4021 conditions, can be treated as verbatim copying in other respects. 4022 4023 If the required texts for either cover are too voluminous to fit 4024 legibly, you should put the first ones listed (as many as fit 4025 reasonably) on the actual cover, and continue the rest onto 4026 adjacent pages. 4027 4028 If you publish or distribute Opaque copies of the Document 4029 numbering more than 100, you must either include a machine-readable 4030 Transparent copy along with each Opaque copy, or state in or with 4031 each Opaque copy a computer-network location from which the general 4032 network-using public has access to download using public-standard 4033 network protocols a complete Transparent copy of the Document, free 4034 of added material. If you use the latter option, you must take 4035 reasonably prudent steps, when you begin distribution of Opaque 4036 copies in quantity, to ensure that this Transparent copy will 4037 remain thus accessible at the stated location until at least one 4038 year after the last time you distribute an Opaque copy (directly or 4039 through your agents or retailers) of that edition to the public. 4040 4041 It is requested, but not required, that you contact the authors of 4042 the Document well before redistributing any large number of copies, 4043 to give them a chance to provide you with an updated version of the 4044 Document. 4045 4046 4. MODIFICATIONS 4047 4048 You may copy and distribute a Modified Version of the Document 4049 under the conditions of sections 2 and 3 above, provided that you 4050 release the Modified Version under precisely this License, with the 4051 Modified Version filling the role of the Document, thus licensing 4052 distribution and modification of the Modified Version to whoever 4053 possesses a copy of it. In addition, you must do these things in 4054 the Modified Version: 4055 4056 A. Use in the Title Page (and on the covers, if any) a title 4057 distinct from that of the Document, and from those of previous 4058 versions (which should, if there were any, be listed in the 4059 History section of the Document). You may use the same title 4060 as a previous version if the original publisher of that 4061 version gives permission. 4062 4063 B. List on the Title Page, as authors, one or more persons or 4064 entities responsible for authorship of the modifications in 4065 the Modified Version, together with at least five of the 4066 principal authors of the Document (all of its principal 4067 authors, if it has fewer than five), unless they release you 4068 from this requirement. 4069 4070 C. State on the Title page the name of the publisher of the 4071 Modified Version, as the publisher. 4072 4073 D. Preserve all the copyright notices of the Document. 4074 4075 E. Add an appropriate copyright notice for your modifications 4076 adjacent to the other copyright notices. 4077 4078 F. Include, immediately after the copyright notices, a license 4079 notice giving the public permission to use the Modified 4080 Version under the terms of this License, in the form shown in 4081 the Addendum below. 4082 4083 G. Preserve in that license notice the full lists of Invariant 4084 Sections and required Cover Texts given in the Document's 4085 license notice. 4086 4087 H. Include an unaltered copy of this License. 4088 4089 I. Preserve the section Entitled "History", Preserve its Title, 4090 and add to it an item stating at least the title, year, new 4091 authors, and publisher of the Modified Version as given on the 4092 Title Page. If there is no section Entitled "History" in the 4093 Document, create one stating the title, year, authors, and 4094 publisher of the Document as given on its Title Page, then add 4095 an item describing the Modified Version as stated in the 4096 previous sentence. 4097 4098 J. Preserve the network location, if any, given in the Document 4099 for public access to a Transparent copy of the Document, and 4100 likewise the network locations given in the Document for 4101 previous versions it was based on. These may be placed in the 4102 "History" section. You may omit a network location for a work 4103 that was published at least four years before the Document 4104 itself, or if the original publisher of the version it refers 4105 to gives permission. 4106 4107 K. For any section Entitled "Acknowledgements" or "Dedications", 4108 Preserve the Title of the section, and preserve in the section 4109 all the substance and tone of each of the contributor 4110 acknowledgements and/or dedications given therein. 4111 4112 L. Preserve all the Invariant Sections of the Document, unaltered 4113 in their text and in their titles. Section numbers or the 4114 equivalent are not considered part of the section titles. 4115 4116 M. Delete any section Entitled "Endorsements". Such a section 4117 may not be included in the Modified Version. 4118 4119 N. Do not retitle any existing section to be Entitled 4120 "Endorsements" or to conflict in title with any Invariant 4121 Section. 4122 4123 O. Preserve any Warranty Disclaimers. 4124 4125 If the Modified Version includes new front-matter sections or 4126 appendices that qualify as Secondary Sections and contain no 4127 material copied from the Document, you may at your option designate 4128 some or all of these sections as invariant. To do this, add their 4129 titles to the list of Invariant Sections in the Modified Version's 4130 license notice. These titles must be distinct from any other 4131 section titles. 4132 4133 You may add a section Entitled "Endorsements", provided it contains 4134 nothing but endorsements of your Modified Version by various 4135 parties--for example, statements of peer review or that the text 4136 has been approved by an organization as the authoritative 4137 definition of a standard. 4138 4139 You may add a passage of up to five words as a Front-Cover Text, 4140 and a passage of up to 25 words as a Back-Cover Text, to the end of 4141 the list of Cover Texts in the Modified Version. Only one passage 4142 of Front-Cover Text and one of Back-Cover Text may be added by (or 4143 through arrangements made by) any one entity. If the Document 4144 already includes a cover text for the same cover, previously added 4145 by you or by arrangement made by the same entity you are acting on 4146 behalf of, you may not add another; but you may replace the old 4147 one, on explicit permission from the previous publisher that added 4148 the old one. 4149 4150 The author(s) and publisher(s) of the Document do not by this 4151 License give permission to use their names for publicity for or to 4152 assert or imply endorsement of any Modified Version. 4153 4154 5. COMBINING DOCUMENTS 4155 4156 You may combine the Document with other documents released under 4157 this License, under the terms defined in section 4 above for 4158 modified versions, provided that you include in the combination all 4159 of the Invariant Sections of all of the original documents, 4160 unmodified, and list them all as Invariant Sections of your 4161 combined work in its license notice, and that you preserve all 4162 their Warranty Disclaimers. 4163 4164 The combined work need only contain one copy of this License, and 4165 multiple identical Invariant Sections may be replaced with a single 4166 copy. If there are multiple Invariant Sections with the same name 4167 but different contents, make the title of each such section unique 4168 by adding at the end of it, in parentheses, the name of the 4169 original author or publisher of that section if known, or else a 4170 unique number. Make the same adjustment to the section titles in 4171 the list of Invariant Sections in the license notice of the 4172 combined work. 4173 4174 In the combination, you must combine any sections Entitled 4175 "History" in the various original documents, forming one section 4176 Entitled "History"; likewise combine any sections Entitled 4177 "Acknowledgements", and any sections Entitled "Dedications". You 4178 must delete all sections Entitled "Endorsements." 4179 4180 6. COLLECTIONS OF DOCUMENTS 4181 4182 You may make a collection consisting of the Document and other 4183 documents released under this License, and replace the individual 4184 copies of this License in the various documents with a single copy 4185 that is included in the collection, provided that you follow the 4186 rules of this License for verbatim copying of each of the documents 4187 in all other respects. 4188 4189 You may extract a single document from such a collection, and 4190 distribute it individually under this License, provided you insert 4191 a copy of this License into the extracted document, and follow this 4192 License in all other respects regarding verbatim copying of that 4193 document. 4194 4195 7. AGGREGATION WITH INDEPENDENT WORKS 4196 4197 A compilation of the Document or its derivatives with other 4198 separate and independent documents or works, in or on a volume of a 4199 storage or distribution medium, is called an "aggregate" if the 4200 copyright resulting from the compilation is not used to limit the 4201 legal rights of the compilation's users beyond what the individual 4202 works permit. When the Document is included in an aggregate, this 4203 License does not apply to the other works in the aggregate which 4204 are not themselves derivative works of the Document. 4205 4206 If the Cover Text requirement of section 3 is applicable to these 4207 copies of the Document, then if the Document is less than one half 4208 of the entire aggregate, the Document's Cover Texts may be placed 4209 on covers that bracket the Document within the aggregate, or the 4210 electronic equivalent of covers if the Document is in electronic 4211 form. Otherwise they must appear on printed covers that bracket 4212 the whole aggregate. 4213 4214 8. TRANSLATION 4215 4216 Translation is considered a kind of modification, so you may 4217 distribute translations of the Document under the terms of section 4218 4. Replacing Invariant Sections with translations requires special 4219 permission from their copyright holders, but you may include 4220 translations of some or all Invariant Sections in addition to the 4221 original versions of these Invariant Sections. You may include a 4222 translation of this License, and all the license notices in the 4223 Document, and any Warranty Disclaimers, provided that you also 4224 include the original English version of this License and the 4225 original versions of those notices and disclaimers. In case of a 4226 disagreement between the translation and the original version of 4227 this License or a notice or disclaimer, the original version will 4228 prevail. 4229 4230 If a section in the Document is Entitled "Acknowledgements", 4231 "Dedications", or "History", the requirement (section 4) to 4232 Preserve its Title (section 1) will typically require changing the 4233 actual title. 4234 4235 9. TERMINATION 4236 4237 You may not copy, modify, sublicense, or distribute the Document 4238 except as expressly provided under this License. Any attempt 4239 otherwise to copy, modify, sublicense, or distribute it is void, 4240 and will automatically terminate your rights under this License. 4241 4242 However, if you cease all violation of this License, then your 4243 license from a particular copyright holder is reinstated (a) 4244 provisionally, unless and until the copyright holder explicitly and 4245 finally terminates your license, and (b) permanently, if the 4246 copyright holder fails to notify you of the violation by some 4247 reasonable means prior to 60 days after the cessation. 4248 4249 Moreover, your license from a particular copyright holder is 4250 reinstated permanently if the copyright holder notifies you of the 4251 violation by some reasonable means, this is the first time you have 4252 received notice of violation of this License (for any work) from 4253 that copyright holder, and you cure the violation prior to 30 days 4254 after your receipt of the notice. 4255 4256 Termination of your rights under this section does not terminate 4257 the licenses of parties who have received copies or rights from you 4258 under this License. If your rights have been terminated and not 4259 permanently reinstated, receipt of a copy of some or all of the 4260 same material does not give you any rights to use it. 4261 4262 10. FUTURE REVISIONS OF THIS LICENSE 4263 4264 The Free Software Foundation may publish new, revised versions of 4265 the GNU Free Documentation License from time to time. Such new 4266 versions will be similar in spirit to the present version, but may 4267 differ in detail to address new problems or concerns. See 4268 <http://www.gnu.org/copyleft/>. 4269 4270 Each version of the License is given a distinguishing version 4271 number. If the Document specifies that a particular numbered 4272 version of this License "or any later version" applies to it, you 4273 have the option of following the terms and conditions either of 4274 that specified version or of any later version that has been 4275 published (not as a draft) by the Free Software Foundation. If the 4276 Document does not specify a version number of this License, you may 4277 choose any version ever published (not as a draft) by the Free 4278 Software Foundation. If the Document specifies that a proxy can 4279 decide which future versions of this License can be used, that 4280 proxy's public statement of acceptance of a version permanently 4281 authorizes you to choose that version for the Document. 4282 4283 11. RELICENSING 4284 4285 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any 4286 World Wide Web server that publishes copyrightable works and also 4287 provides prominent facilities for anybody to edit those works. A 4288 public wiki that anybody can edit is an example of such a server. 4289 A "Massive Multiauthor Collaboration" (or "MMC") contained in the 4290 site means any set of copyrightable works thus published on the MMC 4291 site. 4292 4293 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 4294 license published by Creative Commons Corporation, a not-for-profit 4295 corporation with a principal place of business in San Francisco, 4296 California, as well as future copyleft versions of that license 4297 published by that same organization. 4298 4299 "Incorporate" means to publish or republish a Document, in whole or 4300 in part, as part of another Document. 4301 4302 An MMC is "eligible for relicensing" if it is licensed under this 4303 License, and if all works that were first published under this 4304 License somewhere other than this MMC, and subsequently 4305 incorporated in whole or in part into the MMC, (1) had no cover 4306 texts or invariant sections, and (2) were thus incorporated prior 4307 to November 1, 2008. 4308 4309 The operator of an MMC Site may republish an MMC contained in the 4310 site under CC-BY-SA on the same site at any time before August 1, 4311 2009, provided the MMC is eligible for relicensing. 4312 4313 ADDENDUM: How to use this License for your documents 4314 ==================================================== 4315 4316 To use this License in a document you have written, include a copy of 4317 the License in the document and put the following copyright and license 4318 notices just after the title page: 4319 4320 Copyright (C) YEAR YOUR NAME. 4321 Permission is granted to copy, distribute and/or modify this document 4322 under the terms of the GNU Free Documentation License, Version 1.3 4323 or any later version published by the Free Software Foundation; 4324 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 4325 Texts. A copy of the license is included in the section entitled ``GNU 4326 Free Documentation License''. 4327 4328 If you have Invariant Sections, Front-Cover Texts and Back-Cover 4329 Texts, replace the "with...Texts." line with this: 4330 4331 with the Invariant Sections being LIST THEIR TITLES, with 4332 the Front-Cover Texts being LIST, and with the Back-Cover Texts 4333 being LIST. 4334 4335 If you have Invariant Sections without Cover Texts, or some other 4336 combination of the three, merge those two alternatives to suit the 4337 situation. 4338 4339 If your document contains nontrivial examples of program code, we 4340 recommend releasing these examples in parallel under your choice of free 4341 software license, such as the GNU General Public License, to permit 4342 their use in free software. 4343 4344 4345 File: libgomp.info, Node: Funding, Next: Library Index, Prev: GNU Free Documentation License, Up: Top 4346 4347 Funding Free Software 4348 ********************* 4349 4350 If you want to have more free software a few years from now, it makes 4351 sense for you to help encourage people to contribute funds for its 4352 development. The most effective approach known is to encourage 4353 commercial redistributors to donate. 4354 4355 Users of free software systems can boost the pace of development by 4356 encouraging for-a-fee distributors to donate part of their selling price 4357 to free software developers--the Free Software Foundation, and others. 4358 4359 The way to convince distributors to do this is to demand it and 4360 expect it from them. So when you compare distributors, judge them 4361 partly by how much they give to free software development. Show 4362 distributors they must compete to be the one who gives the most. 4363 4364 To make this approach work, you must insist on numbers that you can 4365 compare, such as, "We will donate ten dollars to the Frobnitz project 4366 for each disk sold." Don't be satisfied with a vague promise, such as 4367 "A portion of the profits are donated," since it doesn't give a basis 4368 for comparison. 4369 4370 Even a precise fraction "of the profits from this disk" is not very 4371 meaningful, since creative accounting and unrelated business decisions 4372 can greatly alter what fraction of the sales price counts as profit. If 4373 the price you pay is $50, ten percent of the profit is probably less 4374 than a dollar; it might be a few cents, or nothing at all. 4375 4376 Some redistributors do development work themselves. This is useful 4377 too; but to keep everyone honest, you need to inquire how much they do, 4378 and what kind. Some kinds of development make much more long-term 4379 difference than others. For example, maintaining a separate version of 4380 a program contributes very little; maintaining the standard version of a 4381 program for the whole community contributes much. Easy new ports 4382 contribute little, since someone else would surely do them; difficult 4383 ports such as adding a new CPU to the GNU Compiler Collection contribute 4384 more; major new features or packages contribute the most. 4385 4386 By establishing the idea that supporting further development is "the 4387 proper thing to do" when distributing free software for a fee, we can 4388 assure a steady flow of resources into making more free software. 4389 4390 Copyright (C) 1994 Free Software Foundation, Inc. 4391 Verbatim copying and redistribution of this section is permitted 4392 without royalty; alteration is not permitted. 4393 4394 4395 File: libgomp.info, Node: Library Index, Prev: Funding, Up: Top 4396 4397 Library Index 4398 ************* 4399 4400 [index] 4401 * Menu: 4402 4403 * Environment Variable: OMP_CANCELLATION. (line 6) 4404 * Environment Variable <1>: OMP_DISPLAY_ENV. (line 6) 4405 * Environment Variable <2>: OMP_DEFAULT_DEVICE. (line 6) 4406 * Environment Variable <3>: OMP_DYNAMIC. (line 6) 4407 * Environment Variable <4>: OMP_MAX_ACTIVE_LEVELS. (line 6) 4408 * Environment Variable <5>: OMP_MAX_TASK_PRIORITY. (line 6) 4409 * Environment Variable <6>: OMP_NESTED. (line 6) 4410 * Environment Variable <7>: OMP_NUM_THREADS. (line 6) 4411 * Environment Variable <8>: OMP_PROC_BIND. (line 6) 4412 * Environment Variable <9>: OMP_PLACES. (line 6) 4413 * Environment Variable <10>: OMP_STACKSIZE. (line 6) 4414 * Environment Variable <11>: OMP_SCHEDULE. (line 6) 4415 * Environment Variable <12>: OMP_THREAD_LIMIT. (line 6) 4416 * Environment Variable <13>: OMP_WAIT_POLICY. (line 6) 4417 * Environment Variable <14>: GOMP_CPU_AFFINITY. (line 6) 4418 * Environment Variable <15>: GOMP_DEBUG. (line 6) 4419 * Environment Variable <16>: GOMP_STACKSIZE. (line 6) 4420 * Environment Variable <17>: GOMP_SPINCOUNT. (line 6) 4421 * Environment Variable <18>: GOMP_RTEMS_THREAD_POOLS. 4422 (line 6) 4423 * FDL, GNU Free Documentation License: GNU Free Documentation License. 4424 (line 6) 4425 * Implementation specific setting: OMP_NESTED. (line 6) 4426 * Implementation specific setting <1>: OMP_NUM_THREADS. (line 6) 4427 * Implementation specific setting <2>: OMP_SCHEDULE. (line 6) 4428 * Implementation specific setting <3>: GOMP_STACKSIZE. (line 6) 4429 * Implementation specific setting <4>: GOMP_SPINCOUNT. (line 6) 4430 * Implementation specific setting <5>: GOMP_RTEMS_THREAD_POOLS. 4431 (line 6) 4432 * Introduction: Top. (line 6) 4433 4434 4435 4436 Tag Table: 4437 Node: Top2083 4438 Node: Enabling OpenMP4489 4439 Node: Runtime Library Routines5276 4440 Node: omp_get_active_level8339 4441 Node: omp_get_ancestor_thread_num9039 4442 Node: omp_get_cancellation9969 4443 Node: omp_get_default_device10783 4444 Node: omp_get_dynamic11459 4445 Node: omp_get_level12334 4446 Node: omp_get_max_active_levels12954 4447 Node: omp_get_max_task_priority13659 4448 Node: omp_get_max_threads14279 4449 Node: omp_get_nested15036 4450 Node: omp_get_num_devices15950 4451 Node: omp_get_num_procs16471 4452 Node: omp_get_num_teams17010 4453 Node: omp_get_num_threads17526 4454 Node: omp_get_proc_bind18615 4455 Node: omp_get_schedule19536 4456 Node: omp_get_team_num20490 4457 Node: omp_get_team_size20989 4458 Node: omp_get_thread_limit21949 4459 Node: omp_get_thread_num22568 4460 Node: omp_in_parallel23439 4461 Node: omp_in_final24088 4462 Node: omp_is_initial_device24762 4463 Node: omp_set_default_device25455 4464 Node: omp_set_dynamic26246 4465 Node: omp_set_max_active_levels27132 4466 Node: omp_set_nested27909 4467 Node: omp_set_num_threads28801 4468 Node: omp_set_schedule29669 4469 Node: omp_init_lock30750 4470 Node: omp_set_lock31403 4471 Node: omp_test_lock32258 4472 Node: omp_unset_lock33234 4473 Node: omp_destroy_lock34165 4474 Node: omp_init_nest_lock34842 4475 Node: omp_set_nest_lock35577 4476 Node: omp_test_nest_lock36492 4477 Node: omp_unset_nest_lock37519 4478 Node: omp_destroy_nest_lock38534 4479 Node: omp_get_wtick39285 4480 Node: omp_get_wtime39877 4481 Node: Environment Variables40653 4482 Node: OMP_CANCELLATION42208 4483 Node: OMP_DISPLAY_ENV42741 4484 Node: OMP_DEFAULT_DEVICE43444 4485 Node: OMP_DYNAMIC44224 4486 Node: OMP_MAX_ACTIVE_LEVELS44820 4487 Node: OMP_MAX_TASK_PRIORITY45470 4488 Node: OMP_NESTED46128 4489 Node: OMP_NUM_THREADS46733 4490 Node: OMP_PROC_BIND47421 4491 Node: OMP_PLACES48613 4492 Node: OMP_STACKSIZE50790 4493 Node: OMP_SCHEDULE51614 4494 Node: OMP_THREAD_LIMIT52312 4495 Node: OMP_WAIT_POLICY52912 4496 Node: GOMP_CPU_AFFINITY53604 4497 Node: GOMP_DEBUG55334 4498 Node: GOMP_STACKSIZE55841 4499 Node: GOMP_SPINCOUNT56670 4500 Node: GOMP_RTEMS_THREAD_POOLS57874 4501 Node: Enabling OpenACC60052 4502 Node: OpenACC Runtime Library Routines61045 4503 Node: acc_get_num_devices64839 4504 Node: acc_set_device_type65565 4505 Node: acc_get_device_type66329 4506 Node: acc_set_device_num67044 4507 Node: acc_get_device_num67828 4508 Node: acc_async_test68625 4509 Node: acc_async_test_all69617 4510 Node: acc_wait70517 4511 Node: acc_wait_all71379 4512 Node: acc_wait_all_async72138 4513 Node: acc_wait_async72890 4514 Node: acc_init73597 4515 Node: acc_shutdown74243 4516 Node: acc_on_device74911 4517 Node: acc_malloc75915 4518 Node: acc_free76414 4519 Node: acc_copyin76841 4520 Node: acc_present_or_copyin77951 4521 Node: acc_create79577 4522 Node: acc_present_or_create80732 4523 Node: acc_copyout82366 4524 Node: acc_delete83389 4525 Node: acc_update_device84364 4526 Node: acc_update_self85476 4527 Node: acc_map_data86580 4528 Node: acc_unmap_data87265 4529 Node: acc_deviceptr87786 4530 Node: acc_hostptr88356 4531 Node: acc_is_present88920 4532 Node: acc_memcpy_to_device90447 4533 Node: acc_memcpy_from_device91110 4534 Node: acc_get_current_cuda_device91794 4535 Node: acc_get_current_cuda_context92391 4536 Node: acc_get_cuda_stream92985 4537 Node: acc_set_cuda_stream93541 4538 Node: OpenACC Environment Variables94073 4539 Node: ACC_DEVICE_TYPE94532 4540 Node: ACC_DEVICE_NUM94768 4541 Node: GCC_ACC_NOTIFY95025 4542 Node: CUDA Streams Usage95248 4543 Ref: CUDA Streams Usage-Footnote-197148 4544 Node: OpenACC Library Interoperability97257 4545 Ref: OpenACC Library Interoperability-Footnote-1103613 4546 Ref: OpenACC Library Interoperability-Footnote-2103865 4547 Node: The libgomp ABI104073 4548 Node: Implementing MASTER construct104929 4549 Node: Implementing CRITICAL construct105343 4550 Node: Implementing ATOMIC construct106082 4551 Node: Implementing FLUSH construct106563 4552 Node: Implementing BARRIER construct106834 4553 Node: Implementing THREADPRIVATE construct107103 4554 Node: Implementing PRIVATE clause107756 4555 Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses108337 4556 Node: Implementing REDUCTION clause109661 4557 Node: Implementing PARALLEL construct110218 4558 Node: Implementing FOR construct111475 4559 Node: Implementing ORDERED construct113473 4560 Node: Implementing SECTIONS construct113779 4561 Node: Implementing SINGLE construct114545 4562 Node: Implementing OpenACC's PARALLEL construct115257 4563 Node: Reporting Bugs115515 4564 Node: Copying115877 4565 Node: GNU Free Documentation License153424 4566 Node: Funding178547 4567 Node: Library Index181073 4568 4569 End Tag Table 4570
Indexes created Fri Mar 20 00:23:26 UTC 2026