libgomp.info revision 1.15
1 1.14 mrg This is libgomp.info, produced by makeinfo version 6.5 from 2 1.11 mrg libgomp.texi. 3 1.1 mrg 4 1.15 mrg Copyright (C) 2006-2022 Free Software Foundation, Inc. 5 1.1 mrg 6 1.1 mrg Permission is granted to copy, distribute and/or modify this document 7 1.3 mrg under the terms of the GNU Free Documentation License, Version 1.3 or 8 1.1 mrg any later version published by the Free Software Foundation; with the 9 1.1 mrg Invariant Sections being "Funding Free Software", the Front-Cover texts 10 1.1 mrg being (a) (see below), and with the Back-Cover Texts being (b) (see 11 1.1 mrg below). A copy of the license is included in the section entitled "GNU 12 1.1 mrg Free Documentation License". 13 1.1 mrg 14 1.1 mrg (a) The FSF's Front-Cover Text is: 15 1.1 mrg 16 1.1 mrg A GNU Manual 17 1.1 mrg 18 1.1 mrg (b) The FSF's Back-Cover Text is: 19 1.1 mrg 20 1.1 mrg You have freedom to copy and modify this GNU Manual, like GNU 21 1.11 mrg software. Copies published by the Free Software Foundation raise funds 22 1.11 mrg for GNU development. 23 1.1 mrg INFO-DIR-SECTION GNU Libraries 24 1.1 mrg START-INFO-DIR-ENTRY 25 1.5 mrg * libgomp: (libgomp). GNU Offloading and Multi Processing Runtime Library. 26 1.1 mrg END-INFO-DIR-ENTRY 27 1.1 mrg 28 1.5 mrg This manual documents libgomp, the GNU Offloading and Multi 29 1.5 mrg Processing Runtime library. This is the GNU implementation of the 30 1.5 mrg OpenMP and OpenACC APIs for parallel and accelerator programming in 31 1.5 mrg C/C++ and Fortran. 32 1.1 mrg 33 1.1 mrg Published by the Free Software Foundation 51 Franklin Street, Fifth 34 1.1 mrg Floor Boston, MA 02110-1301 USA 35 1.1 mrg 36 1.15 mrg Copyright (C) 2006-2022 Free Software Foundation, Inc. 37 1.1 mrg 38 1.1 mrg Permission is granted to copy, distribute and/or modify this document 39 1.3 mrg under the terms of the GNU Free Documentation License, Version 1.3 or 40 1.1 mrg any later version published by the Free Software Foundation; with the 41 1.1 mrg Invariant Sections being "Funding Free Software", the Front-Cover texts 42 1.1 mrg being (a) (see below), and with the Back-Cover Texts being (b) (see 43 1.1 mrg below). A copy of the license is included in the section entitled "GNU 44 1.1 mrg Free Documentation License". 45 1.1 mrg 46 1.1 mrg (a) The FSF's Front-Cover Text is: 47 1.1 mrg 48 1.1 mrg A GNU Manual 49 1.1 mrg 50 1.1 mrg (b) The FSF's Back-Cover Text is: 51 1.1 mrg 52 1.1 mrg You have freedom to copy and modify this GNU Manual, like GNU 53 1.11 mrg software. Copies published by the Free Software Foundation raise funds 54 1.11 mrg for GNU development. 55 1.1 mrg 56 1.1 mrg 57 1.15 mrg File: libgomp.info, Node: Top, Next: Enabling OpenMP 58 1.1 mrg 59 1.1 mrg Introduction 60 1.1 mrg ************ 61 1.1 mrg 62 1.11 mrg This manual documents the usage of libgomp, the GNU Offloading and Multi 63 1.11 mrg Processing Runtime Library. This includes the GNU implementation of the 64 1.12 mrg OpenMP (https://www.openmp.org) Application Programming Interface (API) 65 1.11 mrg for multi-platform shared-memory parallel programming in C/C++ and 66 1.11 mrg Fortran, and the GNU implementation of the OpenACC 67 1.11 mrg (https://www.openacc.org) Application Programming Interface (API) for 68 1.5 mrg offloading of code to accelerator devices in C/C++ and Fortran. 69 1.5 mrg 70 1.5 mrg Originally, libgomp implemented the GNU OpenMP Runtime Library. 71 1.5 mrg Based on this, support for OpenACC and offloading (both OpenACC and 72 1.5 mrg OpenMP 4's target construct) has been added later on, and the library's 73 1.5 mrg name changed to GNU Offloading and Multi Processing Runtime Library. 74 1.1 mrg 75 1.1 mrg * Menu: 76 1.1 mrg 77 1.1 mrg * Enabling OpenMP:: How to enable OpenMP for your applications. 78 1.15 mrg * OpenMP Implementation Status:: List of implemented features by OpenMP version 79 1.12 mrg * OpenMP Runtime Library Routines: Runtime Library Routines. 80 1.12 mrg The OpenMP runtime application programming 81 1.1 mrg interface. 82 1.12 mrg * OpenMP Environment Variables: Environment Variables. 83 1.12 mrg Influencing OpenMP runtime behavior with 84 1.12 mrg environment variables. 85 1.7 mrg * Enabling OpenACC:: How to enable OpenACC for your 86 1.7 mrg applications. 87 1.7 mrg * OpenACC Runtime Library Routines:: The OpenACC runtime application 88 1.7 mrg programming interface. 89 1.7 mrg * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with 90 1.7 mrg environment variables. 91 1.7 mrg * CUDA Streams Usage:: Notes on the implementation of 92 1.7 mrg asynchronous operations. 93 1.7 mrg * OpenACC Library Interoperability:: OpenACC library interoperability with the 94 1.7 mrg NVIDIA CUBLAS library. 95 1.13 mrg * OpenACC Profiling Interface:: 96 1.1 mrg * The libgomp ABI:: Notes on the external ABI presented by libgomp. 97 1.5 mrg * Reporting Bugs:: How to report bugs in the GNU Offloading and 98 1.5 mrg Multi Processing Runtime Library. 99 1.1 mrg * Copying:: GNU general public license says 100 1.1 mrg how you can copy and share libgomp. 101 1.1 mrg * GNU Free Documentation License:: 102 1.1 mrg How you can copy and share this manual. 103 1.1 mrg * Funding:: How to help assure continued work for free 104 1.1 mrg software. 105 1.3 mrg * Library Index:: Index of this documentation. 106 1.1 mrg 107 1.1 mrg 108 1.15 mrg File: libgomp.info, Node: Enabling OpenMP, Next: OpenMP Implementation Status, Up: Top 109 1.1 mrg 110 1.1 mrg 1 Enabling OpenMP 111 1.1 mrg ***************** 112 1.1 mrg 113 1.1 mrg To activate the OpenMP extensions for C/C++ and Fortran, the 114 1.11 mrg compile-time flag '-fopenmp' must be specified. This enables the OpenMP 115 1.11 mrg directive '#pragma omp' in C/C++ and '!$omp' directives in free form, 116 1.11 mrg 'c$omp', '*$omp' and '!$omp' directives in fixed form, '!$' conditional 117 1.11 mrg compilation sentinels in free form and 'c$', '*$' and '!$' sentinels in 118 1.11 mrg fixed form, for Fortran. The flag also arranges for automatic linking 119 1.11 mrg of the OpenMP runtime library (*note Runtime Library Routines::). 120 1.11 mrg 121 1.15 mrg A complete description of all OpenMP directives may be found in the 122 1.15 mrg OpenMP Application Program Interface (https://www.openmp.org) manuals. 123 1.15 mrg See also *note OpenMP Implementation Status::. 124 1.1 mrg 125 1.1 mrg 126 1.15 mrg File: libgomp.info, Node: OpenMP Implementation Status, Next: Runtime Library Routines, Prev: Enabling OpenMP, Up: Top 127 1.1 mrg 128 1.15 mrg 2 OpenMP Implementation Status 129 1.15 mrg ****************************** 130 1.15 mrg 131 1.15 mrg * Menu: 132 1.15 mrg 133 1.15 mrg * OpenMP 4.5:: Feature completion status to 4.5 specification 134 1.15 mrg * OpenMP 5.0:: Feature completion status to 5.0 specification 135 1.15 mrg * OpenMP 5.1:: Feature completion status to 5.1 specification 136 1.15 mrg 137 1.15 mrg The '_OPENMP' preprocessor macro and Fortran's 'openmp_version' 138 1.15 mrg parameter, provided by 'omp_lib.h' and the 'omp_lib' module, have the 139 1.15 mrg value '201511' (i.e. OpenMP 4.5). 140 1.15 mrg 141 1.15 mrg 142 1.15 mrg File: libgomp.info, Node: OpenMP 4.5, Next: OpenMP 5.0, Up: OpenMP Implementation Status 143 1.15 mrg 144 1.15 mrg 2.1 OpenMP 4.5 145 1.15 mrg ============== 146 1.15 mrg 147 1.15 mrg The OpenMP 4.5 specification is fully supported. 148 1.15 mrg 149 1.15 mrg 150 1.15 mrg File: libgomp.info, Node: OpenMP 5.0, Next: OpenMP 5.1, Prev: OpenMP 4.5, Up: OpenMP Implementation Status 151 1.15 mrg 152 1.15 mrg 2.2 OpenMP 5.0 153 1.15 mrg ============== 154 1.15 mrg 155 1.15 mrg New features listed in Appendix B of the OpenMP specification 156 1.15 mrg ------------------------------------------------------------- 157 1.15 mrg 158 1.15 mrg Description Status Comments 159 1.15 mrg ----------------------------------------------------------------------- 160 1.15 mrg Array shaping N 161 1.15 mrg Array sections with non-unit strides in C N 162 1.15 mrg and C++ 163 1.15 mrg Iterators Y 164 1.15 mrg 'metadirective' directive N 165 1.15 mrg 'declare variant' directive P simd traits not 166 1.15 mrg handled 167 1.15 mrg correctly 168 1.15 mrg _target-offload-var_ ICV and Y 169 1.15 mrg 'OMP_TARGET_OFFLOAD' env variable 170 1.15 mrg Nested-parallel changes to Y 171 1.15 mrg _max-active-levels-var_ ICV 172 1.15 mrg 'requires' directive P Only fulfillable 173 1.15 mrg requirement are 174 1.15 mrg 'atomic_default_mem_order' 175 1.15 mrg and 176 1.15 mrg 'dynamic_allocators' 177 1.15 mrg 'teams' construct outside an enclosing Y 178 1.15 mrg target region 179 1.15 mrg Non-rectangular loop nests P Only C/C++ 180 1.15 mrg '!=' as relational-op in canonical loop Y 181 1.15 mrg form for C/C++ 182 1.15 mrg 'nonmonotonic' as default loop schedule Y 183 1.15 mrg modifier for worksharing-loop constructs 184 1.15 mrg Collapse of associated loops that are N 185 1.15 mrg imperfectly nested loops 186 1.15 mrg Clauses 'if', 'nontemporal' and Y 187 1.15 mrg 'order(concurrent)' in 'simd' construct 188 1.15 mrg 'atomic' constructs in 'simd' Y 189 1.15 mrg 'loop' construct Y 190 1.15 mrg 'order(concurrent)' clause Y 191 1.15 mrg 'scan' directive and 'in_scan' modifier Y 192 1.15 mrg for the 'reduction' clause 193 1.15 mrg 'in_reduction' clause on 'task' Y 194 1.15 mrg constructs 195 1.15 mrg 'in_reduction' clause on 'target' P 'nowait' only 196 1.15 mrg constructs stub 197 1.15 mrg 'task_reduction' clause with 'taskgroup' Y 198 1.15 mrg 'task' modifier to 'reduction' clause Y 199 1.15 mrg 'affinity' clause to 'task' construct Y Stub only 200 1.15 mrg 'detach' clause to 'task' construct Y 201 1.15 mrg 'omp_fulfill_event' runtime routine Y 202 1.15 mrg 'reduction' and 'in_reduction' clauses on Y 203 1.15 mrg 'taskloop' and 'taskloop simd' constructs 204 1.15 mrg 'taskloop' construct cancelable by Y 205 1.15 mrg 'cancel' construct 206 1.15 mrg 'mutexinoutset' _dependence-type_ for Y 207 1.15 mrg 'depend' clause 208 1.15 mrg Predefined memory spaces, memory Y Some are only 209 1.15 mrg allocators, allocator traits stubs 210 1.15 mrg Memory management routines Y 211 1.15 mrg 'allocate' directive N 212 1.15 mrg 'allocate' clause P initial support 213 1.15 mrg 'use_device_addr' clause on 'target data' Y 214 1.15 mrg 'ancestor' modifier on 'device' clause P Reverse offload 215 1.15 mrg unsupported 216 1.15 mrg Implicit declare target directive Y 217 1.15 mrg Discontiguous array section with 'target N 218 1.15 mrg update' construct 219 1.15 mrg C/C++'s lvalue expressions in 'to', N 220 1.15 mrg 'from' and 'map' clauses 221 1.15 mrg C/C++'s lvalue expressions in 'depend' Y 222 1.15 mrg clauses 223 1.15 mrg Nested 'declare target' directive Y 224 1.15 mrg Combined 'master' constructs Y 225 1.15 mrg 'depend' clause on 'taskwait' Y 226 1.15 mrg Weak memory ordering clauses on 'atomic' Y 227 1.15 mrg and 'flush' construct 228 1.15 mrg 'hint' clause on the 'atomic' construct Y Stub only 229 1.15 mrg 'depobj' construct and depend objects Y 230 1.15 mrg Lock hints were renamed to Y 231 1.15 mrg synchronization hints 232 1.15 mrg 'conditional' modifier to 'lastprivate' Y 233 1.15 mrg clause 234 1.15 mrg Map-order clarifications P 235 1.15 mrg 'close' _map-type-modifier_ Y 236 1.15 mrg Mapping C/C++ pointer variables and to P 237 1.15 mrg assign the address of device memory 238 1.15 mrg mapped by an array section 239 1.15 mrg Mapping of Fortran pointer and P Mapping of vars 240 1.15 mrg allocatable variables, including pointer with allocatable 241 1.15 mrg and allocatable components of variables components 242 1.15 mrg unsupported 243 1.15 mrg 'defaultmap' extensions Y 244 1.15 mrg 'declare mapper' directive N 245 1.15 mrg 'omp_get_supported_active_levels' routine Y 246 1.15 mrg Runtime routines and environment Y 247 1.15 mrg variables to display runtime thread 248 1.15 mrg affinity information 249 1.15 mrg 'omp_pause_resource' and Y 250 1.15 mrg 'omp_pause_resource_all' runtime routines 251 1.15 mrg 'omp_get_device_num' runtime routine Y 252 1.15 mrg OMPT interface N 253 1.15 mrg OMPD interface N 254 1.15 mrg 255 1.15 mrg Other new OpenMP 5.0 features 256 1.15 mrg ----------------------------- 257 1.15 mrg 258 1.15 mrg Description Status Comments 259 1.15 mrg ----------------------------------------------------------------------- 260 1.15 mrg Supporting C++'s range-based for loop Y 261 1.15 mrg 262 1.15 mrg 263 1.15 mrg File: libgomp.info, Node: OpenMP 5.1, Prev: OpenMP 5.0, Up: OpenMP Implementation Status 264 1.15 mrg 265 1.15 mrg 2.3 OpenMP 5.1 266 1.15 mrg ============== 267 1.15 mrg 268 1.15 mrg New features listed in Appendix B of the OpenMP specification 269 1.15 mrg ------------------------------------------------------------- 270 1.15 mrg 271 1.15 mrg Description Status Comments 272 1.15 mrg ----------------------------------------------------------------------- 273 1.15 mrg OpenMP directive as C++ attribute Y 274 1.15 mrg specifiers 275 1.15 mrg 'omp_all_memory' reserved locator N 276 1.15 mrg _target_device trait_ in OpenMP Context N 277 1.15 mrg 'target_device' selector set in context N 278 1.15 mrg selectors 279 1.15 mrg C/C++'s 'declare variant' directive: N 280 1.15 mrg elision support of preprocessed code 281 1.15 mrg 'declare variant': new clauses N 282 1.15 mrg 'adjust_args' and 'append_args' 283 1.15 mrg 'dispatch' construct N 284 1.15 mrg device-specific ICV settings the N 285 1.15 mrg environment variables 286 1.15 mrg assume directive N 287 1.15 mrg 'nothing' directive Y 288 1.15 mrg 'error' directive Y 289 1.15 mrg 'masked' construct Y 290 1.15 mrg 'scope' directive Y 291 1.15 mrg Loop transformation constructs N 292 1.15 mrg 'strict' modifier in the 'grainsize' and Y 293 1.15 mrg 'num_tasks' clauses of the taskloop 294 1.15 mrg construct 295 1.15 mrg 'align' clause/modifier in 'allocate' P C/C++ on clause 296 1.15 mrg directive/clause and 'allocator' only 297 1.15 mrg directive 298 1.15 mrg 'thread_limit' clause to 'target' Y 299 1.15 mrg construct 300 1.15 mrg 'has_device_addr' clause to 'target' Y 301 1.15 mrg construct 302 1.15 mrg iterators in 'target update' motion N 303 1.15 mrg clauses and 'map' clauses 304 1.15 mrg indirect calls to the device version of a N 305 1.15 mrg procedure or function in 'target' regions 306 1.15 mrg 'interop' directive N 307 1.15 mrg 'omp_interop_t' object support in runtime N 308 1.15 mrg routines 309 1.15 mrg 'nowait' clause in 'taskwait' directive N 310 1.15 mrg Extensions to the 'atomic' directive Y 311 1.15 mrg 'seq_cst' clause on a 'flush' construct Y 312 1.15 mrg 'inoutset' argument to the 'depend' N 313 1.15 mrg clause 314 1.15 mrg 'private' and 'firstprivate' argument to Y 315 1.15 mrg 'default' clause in C and C++ 316 1.15 mrg 'present' argument to 'defaultmap' clause N 317 1.15 mrg 'omp_set_num_teams', Y 318 1.15 mrg 'omp_set_teams_thread_limit', 319 1.15 mrg 'omp_get_max_teams', 320 1.15 mrg 'omp_get_teams_thread_limit' runtime 321 1.15 mrg routines 322 1.15 mrg 'omp_target_is_accessible' runtime N 323 1.15 mrg routine 324 1.15 mrg 'omp_target_memcpy_async' and N 325 1.15 mrg 'omp_target_memcpy_rect_async' runtime 326 1.15 mrg routines 327 1.15 mrg 'omp_get_mapped_ptr' runtime routine N 328 1.15 mrg 'omp_calloc', 'omp_realloc', Y 329 1.15 mrg 'omp_aligned_alloc' and 330 1.15 mrg 'omp_aligned_calloc' runtime routines 331 1.15 mrg 'omp_alloctrait_key_t' enum: Y 332 1.15 mrg 'omp_atv_serialized' added, 333 1.15 mrg 'omp_atv_default' changed 334 1.15 mrg 'omp_display_env' runtime routine Y 335 1.15 mrg 'ompt_scope_endpoint_t' enum: N 336 1.15 mrg 'ompt_scope_beginend' 337 1.15 mrg 'ompt_sync_region_t' enum additions N 338 1.15 mrg 'ompt_state_t' enum: N 339 1.15 mrg 'ompt_state_wait_barrier_implementation' 340 1.15 mrg and 'ompt_state_wait_barrier_teams' 341 1.15 mrg 'ompt_callback_target_data_op_emi_t', N 342 1.15 mrg 'ompt_callback_target_emi_t', 343 1.15 mrg 'ompt_callback_target_map_emi_t' and 344 1.15 mrg 'ompt_callback_target_submit_emi_t' 345 1.15 mrg 'ompt_callback_error_t' type N 346 1.15 mrg 'OMP_PLACES' syntax extensions Y 347 1.15 mrg 'OMP_NUM_TEAMS' and Y 348 1.15 mrg 'OMP_TEAMS_THREAD_LIMIT' environment 349 1.15 mrg variables 350 1.15 mrg 351 1.15 mrg Other new OpenMP 5.1 features 352 1.15 mrg ----------------------------- 353 1.15 mrg 354 1.15 mrg Description Status Comments 355 1.15 mrg ----------------------------------------------------------------------- 356 1.15 mrg Support of strictly structured blocks in Y 357 1.15 mrg Fortran 358 1.15 mrg Support of structured block sequences in Y 359 1.15 mrg C/C++ 360 1.15 mrg 'unconstrained' and 'reproducible' Y 361 1.15 mrg modifiers on 'order' clause 362 1.15 mrg 363 1.15 mrg 364 1.15 mrg File: libgomp.info, Node: Runtime Library Routines, Next: Environment Variables, Prev: OpenMP Implementation Status, Up: Top 365 1.15 mrg 366 1.15 mrg 3 OpenMP Runtime Library Routines 367 1.12 mrg ********************************* 368 1.1 mrg 369 1.5 mrg The runtime routines described here are defined by Section 3 of the 370 1.7 mrg OpenMP specification in version 4.5. The routines are structured in 371 1.1 mrg following three parts: 372 1.1 mrg 373 1.5 mrg * Menu: 374 1.1 mrg 375 1.5 mrg Control threads, processors and the parallel environment. They have C 376 1.5 mrg linkage, and do not throw exceptions. 377 1.1 mrg 378 1.1 mrg * omp_get_active_level:: Number of active parallel regions 379 1.1 mrg * omp_get_ancestor_thread_num:: Ancestor thread ID 380 1.5 mrg * omp_get_cancellation:: Whether cancellation support is enabled 381 1.5 mrg * omp_get_default_device:: Get the default device for target regions 382 1.15 mrg * omp_get_device_num:: Get device that current thread is running on 383 1.1 mrg * omp_get_dynamic:: Dynamic teams setting 384 1.15 mrg * omp_get_initial_device:: Device number of host device 385 1.1 mrg * omp_get_level:: Number of parallel regions 386 1.15 mrg * omp_get_max_active_levels:: Current maximum number of active regions 387 1.7 mrg * omp_get_max_task_priority:: Maximum task priority value that can be set 388 1.15 mrg * omp_get_max_teams:: Maximum number of teams for teams region 389 1.3 mrg * omp_get_max_threads:: Maximum number of threads of parallel region 390 1.1 mrg * omp_get_nested:: Nested parallel regions 391 1.5 mrg * omp_get_num_devices:: Number of target devices 392 1.1 mrg * omp_get_num_procs:: Number of processors online 393 1.5 mrg * omp_get_num_teams:: Number of teams 394 1.1 mrg * omp_get_num_threads:: Size of the active team 395 1.14 mrg * omp_get_proc_bind:: Whether threads may be moved between CPUs 396 1.1 mrg * omp_get_schedule:: Obtain the runtime scheduling method 397 1.15 mrg * omp_get_supported_active_levels:: Maximum number of active regions supported 398 1.5 mrg * omp_get_team_num:: Get team number 399 1.1 mrg * omp_get_team_size:: Number of threads in a team 400 1.15 mrg * omp_get_teams_thread_limit:: Maximum number of threads imposed by teams 401 1.3 mrg * omp_get_thread_limit:: Maximum number of threads 402 1.1 mrg * omp_get_thread_num:: Current thread ID 403 1.1 mrg * omp_in_parallel:: Whether a parallel region is active 404 1.3 mrg * omp_in_final:: Whether in final or included task region 405 1.5 mrg * omp_is_initial_device:: Whether executing on the host device 406 1.5 mrg * omp_set_default_device:: Set the default device for target regions 407 1.1 mrg * omp_set_dynamic:: Enable/disable dynamic teams 408 1.1 mrg * omp_set_max_active_levels:: Limits the number of active parallel regions 409 1.1 mrg * omp_set_nested:: Enable/disable nested parallel regions 410 1.15 mrg * omp_set_num_teams:: Set upper teams limit for teams region 411 1.1 mrg * omp_set_num_threads:: Set upper team size limit 412 1.1 mrg * omp_set_schedule:: Set the runtime scheduling method 413 1.15 mrg * omp_set_teams_thread_limit:: Set upper thread limit for teams construct 414 1.1 mrg 415 1.5 mrg Initialize, set, test, unset and destroy simple and nested locks. 416 1.1 mrg 417 1.1 mrg * omp_init_lock:: Initialize simple lock 418 1.1 mrg * omp_set_lock:: Wait for and set simple lock 419 1.1 mrg * omp_test_lock:: Test and set simple lock if available 420 1.1 mrg * omp_unset_lock:: Unset simple lock 421 1.1 mrg * omp_destroy_lock:: Destroy simple lock 422 1.1 mrg * omp_init_nest_lock:: Initialize nested lock 423 1.1 mrg * omp_set_nest_lock:: Wait for and set simple lock 424 1.1 mrg * omp_test_nest_lock:: Test and set nested lock if available 425 1.1 mrg * omp_unset_nest_lock:: Unset nested lock 426 1.1 mrg * omp_destroy_nest_lock:: Destroy nested lock 427 1.1 mrg 428 1.5 mrg Portable, thread-based, wall clock timer. 429 1.1 mrg 430 1.1 mrg * omp_get_wtick:: Get timer precision. 431 1.1 mrg * omp_get_wtime:: Elapsed wall clock time. 432 1.1 mrg 433 1.15 mrg Support for event objects. 434 1.15 mrg 435 1.15 mrg * omp_fulfill_event:: Fulfill and destroy an OpenMP event. 436 1.15 mrg 437 1.1 mrg 438 1.1 mrg File: libgomp.info, Node: omp_get_active_level, Next: omp_get_ancestor_thread_num, Up: Runtime Library Routines 439 1.1 mrg 440 1.15 mrg 3.1 'omp_get_active_level' - Number of parallel regions 441 1.1 mrg ======================================================= 442 1.1 mrg 443 1.1 mrg _Description_: 444 1.1 mrg This function returns the nesting level for the active parallel 445 1.1 mrg blocks, which enclose the calling call. 446 1.1 mrg 447 1.1 mrg _C/C++_ 448 1.11 mrg _Prototype_: 'int omp_get_active_level(void);' 449 1.1 mrg 450 1.1 mrg _Fortran_: 451 1.11 mrg _Interface_: 'integer function omp_get_active_level()' 452 1.1 mrg 453 1.1 mrg _See also_: 454 1.1 mrg *note omp_get_level::, *note omp_get_max_active_levels::, *note 455 1.1 mrg omp_set_max_active_levels:: 456 1.1 mrg 457 1.1 mrg _Reference_: 458 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.20. 459 1.1 mrg 460 1.1 mrg 461 1.5 mrg File: libgomp.info, Node: omp_get_ancestor_thread_num, Next: omp_get_cancellation, Prev: omp_get_active_level, Up: Runtime Library Routines 462 1.1 mrg 463 1.15 mrg 3.2 'omp_get_ancestor_thread_num' - Ancestor thread ID 464 1.1 mrg ====================================================== 465 1.1 mrg 466 1.1 mrg _Description_: 467 1.1 mrg This function returns the thread identification number for the 468 1.5 mrg given nesting level of the current thread. For values of LEVEL 469 1.11 mrg outside zero to 'omp_get_level' -1 is returned; if LEVEL is 470 1.11 mrg 'omp_get_level' the result is identical to 'omp_get_thread_num'. 471 1.1 mrg 472 1.1 mrg _C/C++_ 473 1.11 mrg _Prototype_: 'int omp_get_ancestor_thread_num(int level);' 474 1.1 mrg 475 1.1 mrg _Fortran_: 476 1.11 mrg _Interface_: 'integer function omp_get_ancestor_thread_num(level)' 477 1.11 mrg 'integer level' 478 1.1 mrg 479 1.1 mrg _See also_: 480 1.1 mrg *note omp_get_level::, *note omp_get_thread_num::, *note 481 1.1 mrg omp_get_team_size:: 482 1.1 mrg 483 1.1 mrg _Reference_: 484 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.18. 485 1.1 mrg 486 1.1 mrg 487 1.5 mrg File: libgomp.info, Node: omp_get_cancellation, Next: omp_get_default_device, Prev: omp_get_ancestor_thread_num, Up: Runtime Library Routines 488 1.5 mrg 489 1.15 mrg 3.3 'omp_get_cancellation' - Whether cancellation support is enabled 490 1.5 mrg ==================================================================== 491 1.5 mrg 492 1.5 mrg _Description_: 493 1.11 mrg This function returns 'true' if cancellation is activated, 'false' 494 1.11 mrg otherwise. Here, 'true' and 'false' represent their 495 1.11 mrg language-specific counterparts. Unless 'OMP_CANCELLATION' is set 496 1.5 mrg true, cancellations are deactivated. 497 1.5 mrg 498 1.5 mrg _C/C++_: 499 1.11 mrg _Prototype_: 'int omp_get_cancellation(void);' 500 1.5 mrg 501 1.5 mrg _Fortran_: 502 1.11 mrg _Interface_: 'logical function omp_get_cancellation()' 503 1.5 mrg 504 1.5 mrg _See also_: 505 1.5 mrg *note OMP_CANCELLATION:: 506 1.5 mrg 507 1.5 mrg _Reference_: 508 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.9. 509 1.5 mrg 510 1.5 mrg 511 1.15 mrg File: libgomp.info, Node: omp_get_default_device, Next: omp_get_device_num, Prev: omp_get_cancellation, Up: Runtime Library Routines 512 1.5 mrg 513 1.15 mrg 3.4 'omp_get_default_device' - Get the default device for target regions 514 1.5 mrg ======================================================================== 515 1.5 mrg 516 1.5 mrg _Description_: 517 1.5 mrg Get the default device for target regions without device clause. 518 1.5 mrg 519 1.5 mrg _C/C++_: 520 1.11 mrg _Prototype_: 'int omp_get_default_device(void);' 521 1.1 mrg 522 1.5 mrg _Fortran_: 523 1.11 mrg _Interface_: 'integer function omp_get_default_device()' 524 1.5 mrg 525 1.5 mrg _See also_: 526 1.5 mrg *note OMP_DEFAULT_DEVICE::, *note omp_set_default_device:: 527 1.5 mrg 528 1.5 mrg _Reference_: 529 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.30. 530 1.5 mrg 531 1.5 mrg 532 1.15 mrg File: libgomp.info, Node: omp_get_device_num, Next: omp_get_dynamic, Prev: omp_get_default_device, Up: Runtime Library Routines 533 1.15 mrg 534 1.15 mrg 3.5 'omp_get_device_num' - Return device number of current device 535 1.15 mrg ================================================================= 536 1.15 mrg 537 1.15 mrg _Description_: 538 1.15 mrg This function returns a device number that represents the device 539 1.15 mrg that the current thread is executing on. For OpenMP 5.0, this must 540 1.15 mrg be equal to the value returned by the 'omp_get_initial_device' 541 1.15 mrg function when called from the host. 542 1.15 mrg 543 1.15 mrg _C/C++_ 544 1.15 mrg _Prototype_: 'int omp_get_device_num(void);' 545 1.15 mrg 546 1.15 mrg _Fortran_: 547 1.15 mrg _Interface_: 'integer function omp_get_device_num()' 548 1.15 mrg 549 1.15 mrg _See also_: 550 1.15 mrg *note omp_get_initial_device:: 551 1.15 mrg 552 1.15 mrg _Reference_: 553 1.15 mrg OpenMP specification v5.0 (https://www.openmp.org), Section 3.2.37. 554 1.5 mrg 555 1.15 mrg 556 1.15 mrg File: libgomp.info, Node: omp_get_dynamic, Next: omp_get_initial_device, Prev: omp_get_device_num, Up: Runtime Library Routines 557 1.15 mrg 558 1.15 mrg 3.6 'omp_get_dynamic' - Dynamic teams setting 559 1.1 mrg ============================================= 560 1.1 mrg 561 1.1 mrg _Description_: 562 1.11 mrg This function returns 'true' if enabled, 'false' otherwise. Here, 563 1.11 mrg 'true' and 'false' represent their language-specific counterparts. 564 1.1 mrg 565 1.1 mrg The dynamic team setting may be initialized at startup by the 566 1.11 mrg 'OMP_DYNAMIC' environment variable or at runtime using 567 1.11 mrg 'omp_set_dynamic'. If undefined, dynamic adjustment is disabled by 568 1.11 mrg default. 569 1.1 mrg 570 1.1 mrg _C/C++_: 571 1.11 mrg _Prototype_: 'int omp_get_dynamic(void);' 572 1.1 mrg 573 1.1 mrg _Fortran_: 574 1.11 mrg _Interface_: 'logical function omp_get_dynamic()' 575 1.1 mrg 576 1.1 mrg _See also_: 577 1.1 mrg *note omp_set_dynamic::, *note OMP_DYNAMIC:: 578 1.1 mrg 579 1.1 mrg _Reference_: 580 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.8. 581 1.1 mrg 582 1.1 mrg 583 1.15 mrg File: libgomp.info, Node: omp_get_initial_device, Next: omp_get_level, Prev: omp_get_dynamic, Up: Runtime Library Routines 584 1.15 mrg 585 1.15 mrg 3.7 'omp_get_initial_device' - Return device number of initial device 586 1.15 mrg ===================================================================== 587 1.15 mrg 588 1.15 mrg _Description_: 589 1.15 mrg This function returns a device number that represents the host 590 1.15 mrg device. For OpenMP 5.1, this must be equal to the value returned 591 1.15 mrg by the 'omp_get_num_devices' function. 592 1.15 mrg 593 1.15 mrg _C/C++_ 594 1.15 mrg _Prototype_: 'int omp_get_initial_device(void);' 595 1.15 mrg 596 1.15 mrg _Fortran_: 597 1.15 mrg _Interface_: 'integer function omp_get_initial_device()' 598 1.15 mrg 599 1.15 mrg _See also_: 600 1.15 mrg *note omp_get_num_devices:: 601 1.15 mrg 602 1.15 mrg _Reference_: 603 1.15 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.35. 604 1.15 mrg 605 1.15 mrg 606 1.15 mrg File: libgomp.info, Node: omp_get_level, Next: omp_get_max_active_levels, Prev: omp_get_initial_device, Up: Runtime Library Routines 607 1.1 mrg 608 1.15 mrg 3.8 'omp_get_level' - Obtain the current nesting level 609 1.1 mrg ====================================================== 610 1.1 mrg 611 1.1 mrg _Description_: 612 1.1 mrg This function returns the nesting level for the parallel blocks, 613 1.1 mrg which enclose the calling call. 614 1.1 mrg 615 1.1 mrg _C/C++_ 616 1.11 mrg _Prototype_: 'int omp_get_level(void);' 617 1.1 mrg 618 1.1 mrg _Fortran_: 619 1.11 mrg _Interface_: 'integer function omp_level()' 620 1.1 mrg 621 1.1 mrg _See also_: 622 1.1 mrg *note omp_get_active_level:: 623 1.1 mrg 624 1.1 mrg _Reference_: 625 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.17. 626 1.1 mrg 627 1.1 mrg 628 1.7 mrg File: libgomp.info, Node: omp_get_max_active_levels, Next: omp_get_max_task_priority, Prev: omp_get_level, Up: Runtime Library Routines 629 1.1 mrg 630 1.15 mrg 3.9 'omp_get_max_active_levels' - Current maximum number of active regions 631 1.15 mrg ========================================================================== 632 1.1 mrg 633 1.1 mrg _Description_: 634 1.3 mrg This function obtains the maximum allowed number of nested, active 635 1.3 mrg parallel regions. 636 1.1 mrg 637 1.1 mrg _C/C++_ 638 1.11 mrg _Prototype_: 'int omp_get_max_active_levels(void);' 639 1.1 mrg 640 1.1 mrg _Fortran_: 641 1.11 mrg _Interface_: 'integer function omp_get_max_active_levels()' 642 1.1 mrg 643 1.1 mrg _See also_: 644 1.1 mrg *note omp_set_max_active_levels::, *note omp_get_active_level:: 645 1.1 mrg 646 1.1 mrg _Reference_: 647 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.16. 648 1.1 mrg 649 1.1 mrg 650 1.15 mrg File: libgomp.info, Node: omp_get_max_task_priority, Next: omp_get_max_teams, Prev: omp_get_max_active_levels, Up: Runtime Library Routines 651 1.7 mrg 652 1.15 mrg 3.10 'omp_get_max_task_priority' - Maximum priority value 653 1.15 mrg ========================================================= 654 1.7 mrg 655 1.7 mrg that can be set for tasks. 656 1.7 mrg _Description_: 657 1.7 mrg This function obtains the maximum allowed priority number for 658 1.7 mrg tasks. 659 1.7 mrg 660 1.7 mrg _C/C++_ 661 1.11 mrg _Prototype_: 'int omp_get_max_task_priority(void);' 662 1.1 mrg 663 1.7 mrg _Fortran_: 664 1.11 mrg _Interface_: 'integer function omp_get_max_task_priority()' 665 1.7 mrg 666 1.7 mrg _Reference_: 667 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.29. 668 1.7 mrg 669 1.7 mrg 670 1.15 mrg File: libgomp.info, Node: omp_get_max_teams, Next: omp_get_max_threads, Prev: omp_get_max_task_priority, Up: Runtime Library Routines 671 1.15 mrg 672 1.15 mrg 3.11 'omp_get_max_teams' - Maximum number of teams of teams region 673 1.15 mrg ================================================================== 674 1.15 mrg 675 1.15 mrg _Description_: 676 1.15 mrg Return the maximum number of teams used for the teams region that 677 1.15 mrg does not use the clause 'num_teams'. 678 1.15 mrg 679 1.15 mrg _C/C++_: 680 1.15 mrg _Prototype_: 'int omp_get_max_teams(void);' 681 1.15 mrg 682 1.15 mrg _Fortran_: 683 1.15 mrg _Interface_: 'integer function omp_get_max_teams()' 684 1.15 mrg 685 1.15 mrg _See also_: 686 1.15 mrg *note omp_set_num_teams::, *note omp_get_num_teams:: 687 1.15 mrg 688 1.15 mrg _Reference_: 689 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 3.4.4. 690 1.15 mrg 691 1.15 mrg 692 1.15 mrg File: libgomp.info, Node: omp_get_max_threads, Next: omp_get_nested, Prev: omp_get_max_teams, Up: Runtime Library Routines 693 1.7 mrg 694 1.15 mrg 3.12 'omp_get_max_threads' - Maximum number of threads of parallel region 695 1.15 mrg ========================================================================= 696 1.1 mrg 697 1.1 mrg _Description_: 698 1.3 mrg Return the maximum number of threads used for the current parallel 699 1.11 mrg region that does not use the clause 'num_threads'. 700 1.1 mrg 701 1.1 mrg _C/C++_: 702 1.11 mrg _Prototype_: 'int omp_get_max_threads(void);' 703 1.1 mrg 704 1.1 mrg _Fortran_: 705 1.11 mrg _Interface_: 'integer function omp_get_max_threads()' 706 1.1 mrg 707 1.1 mrg _See also_: 708 1.1 mrg *note omp_set_num_threads::, *note omp_set_dynamic::, *note 709 1.1 mrg omp_get_thread_limit:: 710 1.1 mrg 711 1.1 mrg _Reference_: 712 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.3. 713 1.1 mrg 714 1.1 mrg 715 1.5 mrg File: libgomp.info, Node: omp_get_nested, Next: omp_get_num_devices, Prev: omp_get_max_threads, Up: Runtime Library Routines 716 1.1 mrg 717 1.15 mrg 3.13 'omp_get_nested' - Nested parallel regions 718 1.7 mrg =============================================== 719 1.1 mrg 720 1.1 mrg _Description_: 721 1.11 mrg This function returns 'true' if nested parallel regions are 722 1.11 mrg enabled, 'false' otherwise. Here, 'true' and 'false' represent 723 1.1 mrg their language-specific counterparts. 724 1.1 mrg 725 1.15 mrg The state of nested parallel regions at startup depends on several 726 1.15 mrg environment variables. If 'OMP_MAX_ACTIVE_LEVELS' is defined and 727 1.15 mrg is set to greater than one, then nested parallel regions will be 728 1.15 mrg enabled. If not defined, then the value of the 'OMP_NESTED' 729 1.15 mrg environment variable will be followed if defined. If neither are 730 1.15 mrg defined, then if either 'OMP_NUM_THREADS' or 'OMP_PROC_BIND' are 731 1.15 mrg defined with a list of more than one value, then nested parallel 732 1.15 mrg regions are enabled. If none of these are defined, then nested 733 1.15 mrg parallel regions are disabled by default. 734 1.15 mrg 735 1.15 mrg Nested parallel regions can be enabled or disabled at runtime using 736 1.15 mrg 'omp_set_nested', or by setting the maximum number of nested 737 1.15 mrg regions with 'omp_set_max_active_levels' to one to disable, or 738 1.15 mrg above one to enable. 739 1.1 mrg 740 1.1 mrg _C/C++_: 741 1.11 mrg _Prototype_: 'int omp_get_nested(void);' 742 1.1 mrg 743 1.1 mrg _Fortran_: 744 1.11 mrg _Interface_: 'logical function omp_get_nested()' 745 1.1 mrg 746 1.1 mrg _See also_: 747 1.15 mrg *note omp_set_max_active_levels::, *note omp_set_nested::, *note 748 1.15 mrg OMP_MAX_ACTIVE_LEVELS::, *note OMP_NESTED:: 749 1.1 mrg 750 1.1 mrg _Reference_: 751 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.11. 752 1.1 mrg 753 1.1 mrg 754 1.5 mrg File: libgomp.info, Node: omp_get_num_devices, Next: omp_get_num_procs, Prev: omp_get_nested, Up: Runtime Library Routines 755 1.1 mrg 756 1.15 mrg 3.14 'omp_get_num_devices' - Number of target devices 757 1.1 mrg ===================================================== 758 1.1 mrg 759 1.1 mrg _Description_: 760 1.5 mrg Returns the number of target devices. 761 1.5 mrg 762 1.5 mrg _C/C++_: 763 1.11 mrg _Prototype_: 'int omp_get_num_devices(void);' 764 1.5 mrg 765 1.5 mrg _Fortran_: 766 1.11 mrg _Interface_: 'integer function omp_get_num_devices()' 767 1.5 mrg 768 1.5 mrg _Reference_: 769 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.31. 770 1.5 mrg 771 1.5 mrg 772 1.5 mrg File: libgomp.info, Node: omp_get_num_procs, Next: omp_get_num_teams, Prev: omp_get_num_devices, Up: Runtime Library Routines 773 1.5 mrg 774 1.15 mrg 3.15 'omp_get_num_procs' - Number of processors online 775 1.5 mrg ====================================================== 776 1.5 mrg 777 1.5 mrg _Description_: 778 1.5 mrg Returns the number of processors online on that device. 779 1.1 mrg 780 1.1 mrg _C/C++_: 781 1.11 mrg _Prototype_: 'int omp_get_num_procs(void);' 782 1.1 mrg 783 1.1 mrg _Fortran_: 784 1.11 mrg _Interface_: 'integer function omp_get_num_procs()' 785 1.1 mrg 786 1.1 mrg _Reference_: 787 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.5. 788 1.1 mrg 789 1.1 mrg 790 1.5 mrg File: libgomp.info, Node: omp_get_num_teams, Next: omp_get_num_threads, Prev: omp_get_num_procs, Up: Runtime Library Routines 791 1.1 mrg 792 1.15 mrg 3.16 'omp_get_num_teams' - Number of teams 793 1.5 mrg ========================================== 794 1.1 mrg 795 1.1 mrg _Description_: 796 1.5 mrg Returns the number of teams in the current team region. 797 1.5 mrg 798 1.5 mrg _C/C++_: 799 1.11 mrg _Prototype_: 'int omp_get_num_teams(void);' 800 1.5 mrg 801 1.5 mrg _Fortran_: 802 1.11 mrg _Interface_: 'integer function omp_get_num_teams()' 803 1.5 mrg 804 1.5 mrg _Reference_: 805 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.32. 806 1.5 mrg 807 1.5 mrg 808 1.5 mrg File: libgomp.info, Node: omp_get_num_threads, Next: omp_get_proc_bind, Prev: omp_get_num_teams, Up: Runtime Library Routines 809 1.5 mrg 810 1.15 mrg 3.17 'omp_get_num_threads' - Size of the active team 811 1.5 mrg ==================================================== 812 1.5 mrg 813 1.5 mrg _Description_: 814 1.11 mrg Returns the number of threads in the current team. In a sequential 815 1.11 mrg section of the program 'omp_get_num_threads' returns 1. 816 1.1 mrg 817 1.1 mrg The default team size may be initialized at startup by the 818 1.11 mrg 'OMP_NUM_THREADS' environment variable. At runtime, the size of 819 1.11 mrg the current team may be set either by the 'NUM_THREADS' clause or 820 1.11 mrg by 'omp_set_num_threads'. If none of the above were used to define 821 1.11 mrg a specific value and 'OMP_DYNAMIC' is disabled, one thread per CPU 822 1.11 mrg online is used. 823 1.1 mrg 824 1.1 mrg _C/C++_: 825 1.11 mrg _Prototype_: 'int omp_get_num_threads(void);' 826 1.1 mrg 827 1.1 mrg _Fortran_: 828 1.11 mrg _Interface_: 'integer function omp_get_num_threads()' 829 1.1 mrg 830 1.1 mrg _See also_: 831 1.1 mrg *note omp_get_max_threads::, *note omp_set_num_threads::, *note 832 1.1 mrg OMP_NUM_THREADS:: 833 1.1 mrg 834 1.1 mrg _Reference_: 835 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.2. 836 1.5 mrg 837 1.5 mrg 838 1.5 mrg File: libgomp.info, Node: omp_get_proc_bind, Next: omp_get_schedule, Prev: omp_get_num_threads, Up: Runtime Library Routines 839 1.5 mrg 840 1.15 mrg 3.18 'omp_get_proc_bind' - Whether threads may be moved between CPUs 841 1.14 mrg ==================================================================== 842 1.5 mrg 843 1.5 mrg _Description_: 844 1.11 mrg This functions returns the currently active thread affinity policy, 845 1.11 mrg which is set via 'OMP_PROC_BIND'. Possible values are 846 1.11 mrg 'omp_proc_bind_false', 'omp_proc_bind_true', 847 1.15 mrg 'omp_proc_bind_primary', 'omp_proc_bind_master', 848 1.15 mrg 'omp_proc_bind_close' and 'omp_proc_bind_spread', where 849 1.15 mrg 'omp_proc_bind_master' is an alias for 'omp_proc_bind_primary'. 850 1.5 mrg 851 1.5 mrg _C/C++_: 852 1.11 mrg _Prototype_: 'omp_proc_bind_t omp_get_proc_bind(void);' 853 1.5 mrg 854 1.5 mrg _Fortran_: 855 1.11 mrg _Interface_: 'integer(kind=omp_proc_bind_kind) function 856 1.11 mrg omp_get_proc_bind()' 857 1.5 mrg 858 1.5 mrg _See also_: 859 1.5 mrg *note OMP_PROC_BIND::, *note OMP_PLACES::, *note 860 1.5 mrg GOMP_CPU_AFFINITY::, 861 1.5 mrg 862 1.5 mrg _Reference_: 863 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.22. 864 1.1 mrg 865 1.1 mrg 866 1.15 mrg File: libgomp.info, Node: omp_get_schedule, Next: omp_get_supported_active_levels, Prev: omp_get_proc_bind, Up: Runtime Library Routines 867 1.1 mrg 868 1.15 mrg 3.19 'omp_get_schedule' - Obtain the runtime scheduling method 869 1.1 mrg ============================================================== 870 1.1 mrg 871 1.1 mrg _Description_: 872 1.5 mrg Obtain the runtime scheduling method. The KIND argument will be 873 1.11 mrg set to the value 'omp_sched_static', 'omp_sched_dynamic', 874 1.11 mrg 'omp_sched_guided' or 'omp_sched_auto'. The second argument, 875 1.7 mrg CHUNK_SIZE, is set to the chunk size. 876 1.1 mrg 877 1.1 mrg _C/C++_ 878 1.11 mrg _Prototype_: 'void omp_get_schedule(omp_sched_t *kind, int 879 1.11 mrg *chunk_size);' 880 1.1 mrg 881 1.1 mrg _Fortran_: 882 1.11 mrg _Interface_: 'subroutine omp_get_schedule(kind, chunk_size)' 883 1.11 mrg 'integer(kind=omp_sched_kind) kind' 884 1.11 mrg 'integer chunk_size' 885 1.1 mrg 886 1.1 mrg _See also_: 887 1.1 mrg *note omp_set_schedule::, *note OMP_SCHEDULE:: 888 1.1 mrg 889 1.1 mrg _Reference_: 890 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.13. 891 1.5 mrg 892 1.5 mrg 893 1.15 mrg File: libgomp.info, Node: omp_get_supported_active_levels, Next: omp_get_team_num, Prev: omp_get_schedule, Up: Runtime Library Routines 894 1.15 mrg 895 1.15 mrg 3.20 'omp_get_supported_active_levels' - Maximum number of active regions supported 896 1.15 mrg =================================================================================== 897 1.15 mrg 898 1.15 mrg _Description_: 899 1.15 mrg This function returns the maximum number of nested, active parallel 900 1.15 mrg regions supported by this implementation. 901 1.15 mrg 902 1.15 mrg _C/C++_ 903 1.15 mrg _Prototype_: 'int omp_get_supported_active_levels(void);' 904 1.5 mrg 905 1.15 mrg _Fortran_: 906 1.15 mrg _Interface_: 'integer function omp_get_supported_active_levels()' 907 1.15 mrg 908 1.15 mrg _See also_: 909 1.15 mrg *note omp_get_max_active_levels::, *note 910 1.15 mrg omp_set_max_active_levels:: 911 1.15 mrg 912 1.15 mrg _Reference_: 913 1.15 mrg OpenMP specification v5.0 (https://www.openmp.org), Section 3.2.15. 914 1.15 mrg 915 1.15 mrg 916 1.15 mrg File: libgomp.info, Node: omp_get_team_num, Next: omp_get_team_size, Prev: omp_get_supported_active_levels, Up: Runtime Library Routines 917 1.15 mrg 918 1.15 mrg 3.21 'omp_get_team_num' - Get team number 919 1.5 mrg ========================================= 920 1.5 mrg 921 1.5 mrg _Description_: 922 1.5 mrg Returns the team number of the calling thread. 923 1.5 mrg 924 1.5 mrg _C/C++_: 925 1.11 mrg _Prototype_: 'int omp_get_team_num(void);' 926 1.5 mrg 927 1.5 mrg _Fortran_: 928 1.11 mrg _Interface_: 'integer function omp_get_team_num()' 929 1.5 mrg 930 1.5 mrg _Reference_: 931 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.33. 932 1.1 mrg 933 1.1 mrg 934 1.15 mrg File: libgomp.info, Node: omp_get_team_size, Next: omp_get_teams_thread_limit, Prev: omp_get_team_num, Up: Runtime Library Routines 935 1.1 mrg 936 1.15 mrg 3.22 'omp_get_team_size' - Number of threads in a team 937 1.1 mrg ====================================================== 938 1.1 mrg 939 1.1 mrg _Description_: 940 1.1 mrg This function returns the number of threads in a thread team to 941 1.5 mrg which either the current thread or its ancestor belongs. For 942 1.11 mrg values of LEVEL outside zero to 'omp_get_level', -1 is returned; if 943 1.11 mrg LEVEL is zero, 1 is returned, and for 'omp_get_level', the result 944 1.11 mrg is identical to 'omp_get_num_threads'. 945 1.1 mrg 946 1.1 mrg _C/C++_: 947 1.11 mrg _Prototype_: 'int omp_get_team_size(int level);' 948 1.1 mrg 949 1.1 mrg _Fortran_: 950 1.11 mrg _Interface_: 'integer function omp_get_team_size(level)' 951 1.11 mrg 'integer level' 952 1.1 mrg 953 1.1 mrg _See also_: 954 1.1 mrg *note omp_get_num_threads::, *note omp_get_level::, *note 955 1.1 mrg omp_get_ancestor_thread_num:: 956 1.1 mrg 957 1.1 mrg _Reference_: 958 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.19. 959 1.1 mrg 960 1.1 mrg 961 1.15 mrg File: libgomp.info, Node: omp_get_teams_thread_limit, Next: omp_get_thread_limit, Prev: omp_get_team_size, Up: Runtime Library Routines 962 1.15 mrg 963 1.15 mrg 3.23 'omp_get_teams_thread_limit' - Maximum number of threads imposed by teams 964 1.15 mrg ============================================================================== 965 1.15 mrg 966 1.15 mrg _Description_: 967 1.15 mrg Return the maximum number of threads that will be able to 968 1.15 mrg participate in each team created by a teams construct. 969 1.15 mrg 970 1.15 mrg _C/C++_: 971 1.15 mrg _Prototype_: 'int omp_get_teams_thread_limit(void);' 972 1.1 mrg 973 1.15 mrg _Fortran_: 974 1.15 mrg _Interface_: 'integer function omp_get_teams_thread_limit()' 975 1.15 mrg 976 1.15 mrg _See also_: 977 1.15 mrg *note omp_set_teams_thread_limit::, *note OMP_TEAMS_THREAD_LIMIT:: 978 1.15 mrg 979 1.15 mrg _Reference_: 980 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 3.4.6. 981 1.15 mrg 982 1.15 mrg 983 1.15 mrg File: libgomp.info, Node: omp_get_thread_limit, Next: omp_get_thread_num, Prev: omp_get_teams_thread_limit, Up: Runtime Library Routines 984 1.15 mrg 985 1.15 mrg 3.24 'omp_get_thread_limit' - Maximum number of threads 986 1.1 mrg ======================================================= 987 1.1 mrg 988 1.1 mrg _Description_: 989 1.3 mrg Return the maximum number of threads of the program. 990 1.1 mrg 991 1.1 mrg _C/C++_: 992 1.11 mrg _Prototype_: 'int omp_get_thread_limit(void);' 993 1.1 mrg 994 1.1 mrg _Fortran_: 995 1.11 mrg _Interface_: 'integer function omp_get_thread_limit()' 996 1.1 mrg 997 1.1 mrg _See also_: 998 1.1 mrg *note omp_get_max_threads::, *note OMP_THREAD_LIMIT:: 999 1.1 mrg 1000 1.1 mrg _Reference_: 1001 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.14. 1002 1.1 mrg 1003 1.1 mrg 1004 1.1 mrg File: libgomp.info, Node: omp_get_thread_num, Next: omp_in_parallel, Prev: omp_get_thread_limit, Up: Runtime Library Routines 1005 1.1 mrg 1006 1.15 mrg 3.25 'omp_get_thread_num' - Current thread ID 1007 1.1 mrg ============================================= 1008 1.1 mrg 1009 1.1 mrg _Description_: 1010 1.3 mrg Returns a unique thread identification number within the current 1011 1.11 mrg team. In a sequential parts of the program, 'omp_get_thread_num' 1012 1.11 mrg always returns 0. In parallel regions the return value varies from 1013 1.11 mrg 0 to 'omp_get_num_threads'-1 inclusive. The return value of the 1014 1.15 mrg primary thread of a team is always 0. 1015 1.1 mrg 1016 1.1 mrg _C/C++_: 1017 1.11 mrg _Prototype_: 'int omp_get_thread_num(void);' 1018 1.1 mrg 1019 1.1 mrg _Fortran_: 1020 1.11 mrg _Interface_: 'integer function omp_get_thread_num()' 1021 1.1 mrg 1022 1.1 mrg _See also_: 1023 1.1 mrg *note omp_get_num_threads::, *note omp_get_ancestor_thread_num:: 1024 1.1 mrg 1025 1.1 mrg _Reference_: 1026 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.4. 1027 1.1 mrg 1028 1.1 mrg 1029 1.3 mrg File: libgomp.info, Node: omp_in_parallel, Next: omp_in_final, Prev: omp_get_thread_num, Up: Runtime Library Routines 1030 1.1 mrg 1031 1.15 mrg 3.26 'omp_in_parallel' - Whether a parallel region is active 1032 1.1 mrg ============================================================ 1033 1.1 mrg 1034 1.1 mrg _Description_: 1035 1.11 mrg This function returns 'true' if currently running in parallel, 1036 1.11 mrg 'false' otherwise. Here, 'true' and 'false' represent their 1037 1.1 mrg language-specific counterparts. 1038 1.1 mrg 1039 1.1 mrg _C/C++_: 1040 1.11 mrg _Prototype_: 'int omp_in_parallel(void);' 1041 1.1 mrg 1042 1.1 mrg _Fortran_: 1043 1.11 mrg _Interface_: 'logical function omp_in_parallel()' 1044 1.1 mrg 1045 1.1 mrg _Reference_: 1046 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.6. 1047 1.1 mrg 1048 1.1 mrg 1049 1.5 mrg File: libgomp.info, Node: omp_in_final, Next: omp_is_initial_device, Prev: omp_in_parallel, Up: Runtime Library Routines 1050 1.3 mrg 1051 1.15 mrg 3.27 'omp_in_final' - Whether in final or included task region 1052 1.3 mrg ============================================================== 1053 1.3 mrg 1054 1.3 mrg _Description_: 1055 1.11 mrg This function returns 'true' if currently running in a final or 1056 1.11 mrg included task region, 'false' otherwise. Here, 'true' and 'false' 1057 1.3 mrg represent their language-specific counterparts. 1058 1.3 mrg 1059 1.3 mrg _C/C++_: 1060 1.11 mrg _Prototype_: 'int omp_in_final(void);' 1061 1.1 mrg 1062 1.3 mrg _Fortran_: 1063 1.11 mrg _Interface_: 'logical function omp_in_final()' 1064 1.3 mrg 1065 1.3 mrg _Reference_: 1066 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.21. 1067 1.5 mrg 1068 1.5 mrg 1069 1.5 mrg File: libgomp.info, Node: omp_is_initial_device, Next: omp_set_default_device, Prev: omp_in_final, Up: Runtime Library Routines 1070 1.5 mrg 1071 1.15 mrg 3.28 'omp_is_initial_device' - Whether executing on the host device 1072 1.5 mrg =================================================================== 1073 1.5 mrg 1074 1.5 mrg _Description_: 1075 1.11 mrg This function returns 'true' if currently running on the host 1076 1.11 mrg device, 'false' otherwise. Here, 'true' and 'false' represent 1077 1.5 mrg their language-specific counterparts. 1078 1.5 mrg 1079 1.5 mrg _C/C++_: 1080 1.11 mrg _Prototype_: 'int omp_is_initial_device(void);' 1081 1.5 mrg 1082 1.5 mrg _Fortran_: 1083 1.11 mrg _Interface_: 'logical function omp_is_initial_device()' 1084 1.5 mrg 1085 1.5 mrg _Reference_: 1086 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.34. 1087 1.5 mrg 1088 1.5 mrg 1089 1.5 mrg File: libgomp.info, Node: omp_set_default_device, Next: omp_set_dynamic, Prev: omp_is_initial_device, Up: Runtime Library Routines 1090 1.5 mrg 1091 1.15 mrg 3.29 'omp_set_default_device' - Set the default device for target regions 1092 1.5 mrg ========================================================================= 1093 1.5 mrg 1094 1.5 mrg _Description_: 1095 1.5 mrg Set the default device for target regions without device clause. 1096 1.5 mrg The argument shall be a nonnegative device number. 1097 1.5 mrg 1098 1.5 mrg _C/C++_: 1099 1.11 mrg _Prototype_: 'void omp_set_default_device(int device_num);' 1100 1.5 mrg 1101 1.5 mrg _Fortran_: 1102 1.11 mrg _Interface_: 'subroutine omp_set_default_device(device_num)' 1103 1.11 mrg 'integer device_num' 1104 1.5 mrg 1105 1.5 mrg _See also_: 1106 1.5 mrg *note OMP_DEFAULT_DEVICE::, *note omp_get_default_device:: 1107 1.5 mrg 1108 1.5 mrg _Reference_: 1109 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.29. 1110 1.3 mrg 1111 1.3 mrg 1112 1.5 mrg File: libgomp.info, Node: omp_set_dynamic, Next: omp_set_max_active_levels, Prev: omp_set_default_device, Up: Runtime Library Routines 1113 1.3 mrg 1114 1.15 mrg 3.30 'omp_set_dynamic' - Enable/disable dynamic teams 1115 1.1 mrg ===================================================== 1116 1.1 mrg 1117 1.1 mrg _Description_: 1118 1.1 mrg Enable or disable the dynamic adjustment of the number of threads 1119 1.5 mrg within a team. The function takes the language-specific equivalent 1120 1.11 mrg of 'true' and 'false', where 'true' enables dynamic adjustment of 1121 1.11 mrg team sizes and 'false' disables it. 1122 1.1 mrg 1123 1.1 mrg _C/C++_: 1124 1.11 mrg _Prototype_: 'void omp_set_dynamic(int dynamic_threads);' 1125 1.1 mrg 1126 1.1 mrg _Fortran_: 1127 1.11 mrg _Interface_: 'subroutine omp_set_dynamic(dynamic_threads)' 1128 1.11 mrg 'logical, intent(in) :: dynamic_threads' 1129 1.1 mrg 1130 1.1 mrg _See also_: 1131 1.1 mrg *note OMP_DYNAMIC::, *note omp_get_dynamic:: 1132 1.1 mrg 1133 1.1 mrg _Reference_: 1134 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.7. 1135 1.1 mrg 1136 1.1 mrg 1137 1.1 mrg File: libgomp.info, Node: omp_set_max_active_levels, Next: omp_set_nested, Prev: omp_set_dynamic, Up: Runtime Library Routines 1138 1.1 mrg 1139 1.15 mrg 3.31 'omp_set_max_active_levels' - Limits the number of active parallel regions 1140 1.1 mrg =============================================================================== 1141 1.1 mrg 1142 1.1 mrg _Description_: 1143 1.3 mrg This function limits the maximum allowed number of nested, active 1144 1.15 mrg parallel regions. MAX_LEVELS must be less or equal to the value 1145 1.15 mrg returned by 'omp_get_supported_active_levels'. 1146 1.1 mrg 1147 1.1 mrg _C/C++_ 1148 1.11 mrg _Prototype_: 'void omp_set_max_active_levels(int max_levels);' 1149 1.1 mrg 1150 1.1 mrg _Fortran_: 1151 1.11 mrg _Interface_: 'subroutine omp_set_max_active_levels(max_levels)' 1152 1.11 mrg 'integer max_levels' 1153 1.1 mrg 1154 1.1 mrg _See also_: 1155 1.15 mrg *note omp_get_max_active_levels::, *note omp_get_active_level::, 1156 1.15 mrg *note omp_get_supported_active_levels:: 1157 1.1 mrg 1158 1.1 mrg _Reference_: 1159 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.15. 1160 1.1 mrg 1161 1.1 mrg 1162 1.15 mrg File: libgomp.info, Node: omp_set_nested, Next: omp_set_num_teams, Prev: omp_set_max_active_levels, Up: Runtime Library Routines 1163 1.1 mrg 1164 1.15 mrg 3.32 'omp_set_nested' - Enable/disable nested parallel regions 1165 1.1 mrg ============================================================== 1166 1.1 mrg 1167 1.1 mrg _Description_: 1168 1.1 mrg Enable or disable nested parallel regions, i.e., whether team 1169 1.5 mrg members are allowed to create new teams. The function takes the 1170 1.11 mrg language-specific equivalent of 'true' and 'false', where 'true' 1171 1.11 mrg enables dynamic adjustment of team sizes and 'false' disables it. 1172 1.1 mrg 1173 1.15 mrg Enabling nested parallel regions will also set the maximum number 1174 1.15 mrg of active nested regions to the maximum supported. Disabling 1175 1.15 mrg nested parallel regions will set the maximum number of active 1176 1.15 mrg nested regions to one. 1177 1.15 mrg 1178 1.1 mrg _C/C++_: 1179 1.11 mrg _Prototype_: 'void omp_set_nested(int nested);' 1180 1.1 mrg 1181 1.1 mrg _Fortran_: 1182 1.11 mrg _Interface_: 'subroutine omp_set_nested(nested)' 1183 1.11 mrg 'logical, intent(in) :: nested' 1184 1.1 mrg 1185 1.1 mrg _See also_: 1186 1.15 mrg *note omp_get_nested::, *note omp_set_max_active_levels::, *note 1187 1.15 mrg OMP_MAX_ACTIVE_LEVELS::, *note OMP_NESTED:: 1188 1.1 mrg 1189 1.1 mrg _Reference_: 1190 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.10. 1191 1.1 mrg 1192 1.1 mrg 1193 1.15 mrg File: libgomp.info, Node: omp_set_num_teams, Next: omp_set_num_threads, Prev: omp_set_nested, Up: Runtime Library Routines 1194 1.15 mrg 1195 1.15 mrg 3.33 'omp_set_num_teams' - Set upper teams limit for teams construct 1196 1.15 mrg ==================================================================== 1197 1.15 mrg 1198 1.15 mrg _Description_: 1199 1.15 mrg Specifies the upper bound for number of teams created by the teams 1200 1.15 mrg construct which does not specify a 'num_teams' clause. The 1201 1.15 mrg argument of 'omp_set_num_teams' shall be a positive integer. 1202 1.15 mrg 1203 1.15 mrg _C/C++_: 1204 1.15 mrg _Prototype_: 'void omp_set_num_teams(int num_teams);' 1205 1.1 mrg 1206 1.15 mrg _Fortran_: 1207 1.15 mrg _Interface_: 'subroutine omp_set_num_teams(num_teams)' 1208 1.15 mrg 'integer, intent(in) :: num_teams' 1209 1.15 mrg 1210 1.15 mrg _See also_: 1211 1.15 mrg *note OMP_NUM_TEAMS::, *note omp_get_num_teams::, *note 1212 1.15 mrg omp_get_max_teams:: 1213 1.15 mrg 1214 1.15 mrg _Reference_: 1215 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 3.4.3. 1216 1.15 mrg 1217 1.15 mrg 1218 1.15 mrg File: libgomp.info, Node: omp_set_num_threads, Next: omp_set_schedule, Prev: omp_set_num_teams, Up: Runtime Library Routines 1219 1.15 mrg 1220 1.15 mrg 3.34 'omp_set_num_threads' - Set upper team size limit 1221 1.1 mrg ====================================================== 1222 1.1 mrg 1223 1.1 mrg _Description_: 1224 1.1 mrg Specifies the number of threads used by default in subsequent 1225 1.11 mrg parallel sections, if those do not specify a 'num_threads' clause. 1226 1.11 mrg The argument of 'omp_set_num_threads' shall be a positive integer. 1227 1.1 mrg 1228 1.1 mrg _C/C++_: 1229 1.11 mrg _Prototype_: 'void omp_set_num_threads(int num_threads);' 1230 1.1 mrg 1231 1.1 mrg _Fortran_: 1232 1.11 mrg _Interface_: 'subroutine omp_set_num_threads(num_threads)' 1233 1.11 mrg 'integer, intent(in) :: num_threads' 1234 1.1 mrg 1235 1.1 mrg _See also_: 1236 1.1 mrg *note OMP_NUM_THREADS::, *note omp_get_num_threads::, *note 1237 1.1 mrg omp_get_max_threads:: 1238 1.1 mrg 1239 1.1 mrg _Reference_: 1240 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.1. 1241 1.1 mrg 1242 1.1 mrg 1243 1.15 mrg File: libgomp.info, Node: omp_set_schedule, Next: omp_set_teams_thread_limit, Prev: omp_set_num_threads, Up: Runtime Library Routines 1244 1.1 mrg 1245 1.15 mrg 3.35 'omp_set_schedule' - Set the runtime scheduling method 1246 1.1 mrg =========================================================== 1247 1.1 mrg 1248 1.1 mrg _Description_: 1249 1.5 mrg Sets the runtime scheduling method. The KIND argument can have the 1250 1.11 mrg value 'omp_sched_static', 'omp_sched_dynamic', 'omp_sched_guided' 1251 1.11 mrg or 'omp_sched_auto'. Except for 'omp_sched_auto', the chunk size 1252 1.7 mrg is set to the value of CHUNK_SIZE if positive, or to the default 1253 1.11 mrg value if zero or negative. For 'omp_sched_auto' the CHUNK_SIZE 1254 1.1 mrg argument is ignored. 1255 1.1 mrg 1256 1.1 mrg _C/C++_ 1257 1.11 mrg _Prototype_: 'void omp_set_schedule(omp_sched_t kind, int 1258 1.11 mrg chunk_size);' 1259 1.1 mrg 1260 1.1 mrg _Fortran_: 1261 1.11 mrg _Interface_: 'subroutine omp_set_schedule(kind, chunk_size)' 1262 1.11 mrg 'integer(kind=omp_sched_kind) kind' 1263 1.11 mrg 'integer chunk_size' 1264 1.1 mrg 1265 1.1 mrg _See also_: 1266 1.1 mrg *note omp_get_schedule:: *note OMP_SCHEDULE:: 1267 1.1 mrg 1268 1.1 mrg _Reference_: 1269 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.2.12. 1270 1.1 mrg 1271 1.1 mrg 1272 1.15 mrg File: libgomp.info, Node: omp_set_teams_thread_limit, Next: omp_init_lock, Prev: omp_set_schedule, Up: Runtime Library Routines 1273 1.15 mrg 1274 1.15 mrg 3.36 'omp_set_teams_thread_limit' - Set upper thread limit for teams construct 1275 1.15 mrg ============================================================================== 1276 1.1 mrg 1277 1.15 mrg _Description_: 1278 1.15 mrg Specifies the upper bound for number of threads that will be 1279 1.15 mrg available for each team created by the teams construct which does 1280 1.15 mrg not specify a 'thread_limit' clause. The argument of 1281 1.15 mrg 'omp_set_teams_thread_limit' shall be a positive integer. 1282 1.15 mrg 1283 1.15 mrg _C/C++_: 1284 1.15 mrg _Prototype_: 'void omp_set_teams_thread_limit(int thread_limit);' 1285 1.15 mrg 1286 1.15 mrg _Fortran_: 1287 1.15 mrg _Interface_: 'subroutine omp_set_teams_thread_limit(thread_limit)' 1288 1.15 mrg 'integer, intent(in) :: thread_limit' 1289 1.15 mrg 1290 1.15 mrg _See also_: 1291 1.15 mrg *note OMP_TEAMS_THREAD_LIMIT::, *note omp_get_teams_thread_limit::, 1292 1.15 mrg *note omp_get_thread_limit:: 1293 1.15 mrg 1294 1.15 mrg _Reference_: 1295 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 3.4.5. 1296 1.15 mrg 1297 1.15 mrg 1298 1.15 mrg File: libgomp.info, Node: omp_init_lock, Next: omp_set_lock, Prev: omp_set_teams_thread_limit, Up: Runtime Library Routines 1299 1.15 mrg 1300 1.15 mrg 3.37 'omp_init_lock' - Initialize simple lock 1301 1.1 mrg ============================================= 1302 1.1 mrg 1303 1.1 mrg _Description_: 1304 1.3 mrg Initialize a simple lock. After initialization, the lock is in an 1305 1.1 mrg unlocked state. 1306 1.1 mrg 1307 1.1 mrg _C/C++_: 1308 1.11 mrg _Prototype_: 'void omp_init_lock(omp_lock_t *lock);' 1309 1.1 mrg 1310 1.1 mrg _Fortran_: 1311 1.11 mrg _Interface_: 'subroutine omp_init_lock(svar)' 1312 1.11 mrg 'integer(omp_lock_kind), intent(out) :: svar' 1313 1.1 mrg 1314 1.1 mrg _See also_: 1315 1.1 mrg *note omp_destroy_lock:: 1316 1.1 mrg 1317 1.1 mrg _Reference_: 1318 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.1. 1319 1.1 mrg 1320 1.1 mrg 1321 1.1 mrg File: libgomp.info, Node: omp_set_lock, Next: omp_test_lock, Prev: omp_init_lock, Up: Runtime Library Routines 1322 1.1 mrg 1323 1.15 mrg 3.38 'omp_set_lock' - Wait for and set simple lock 1324 1.1 mrg ================================================== 1325 1.1 mrg 1326 1.1 mrg _Description_: 1327 1.11 mrg Before setting a simple lock, the lock variable must be initialized 1328 1.11 mrg by 'omp_init_lock'. The calling thread is blocked until the lock 1329 1.11 mrg is available. If the lock is already held by the current thread, a 1330 1.11 mrg deadlock occurs. 1331 1.1 mrg 1332 1.1 mrg _C/C++_: 1333 1.11 mrg _Prototype_: 'void omp_set_lock(omp_lock_t *lock);' 1334 1.1 mrg 1335 1.1 mrg _Fortran_: 1336 1.11 mrg _Interface_: 'subroutine omp_set_lock(svar)' 1337 1.11 mrg 'integer(omp_lock_kind), intent(inout) :: svar' 1338 1.1 mrg 1339 1.1 mrg _See also_: 1340 1.1 mrg *note omp_init_lock::, *note omp_test_lock::, *note 1341 1.1 mrg omp_unset_lock:: 1342 1.1 mrg 1343 1.1 mrg _Reference_: 1344 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.4. 1345 1.1 mrg 1346 1.1 mrg 1347 1.1 mrg File: libgomp.info, Node: omp_test_lock, Next: omp_unset_lock, Prev: omp_set_lock, Up: Runtime Library Routines 1348 1.1 mrg 1349 1.15 mrg 3.39 'omp_test_lock' - Test and set simple lock if available 1350 1.1 mrg ============================================================ 1351 1.1 mrg 1352 1.1 mrg _Description_: 1353 1.11 mrg Before setting a simple lock, the lock variable must be initialized 1354 1.11 mrg by 'omp_init_lock'. Contrary to 'omp_set_lock', 'omp_test_lock' 1355 1.11 mrg does not block if the lock is not available. This function returns 1356 1.11 mrg 'true' upon success, 'false' otherwise. Here, 'true' and 'false' 1357 1.11 mrg represent their language-specific counterparts. 1358 1.1 mrg 1359 1.1 mrg _C/C++_: 1360 1.11 mrg _Prototype_: 'int omp_test_lock(omp_lock_t *lock);' 1361 1.1 mrg 1362 1.1 mrg _Fortran_: 1363 1.11 mrg _Interface_: 'logical function omp_test_lock(svar)' 1364 1.11 mrg 'integer(omp_lock_kind), intent(inout) :: svar' 1365 1.1 mrg 1366 1.1 mrg _See also_: 1367 1.1 mrg *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 1368 1.1 mrg 1369 1.1 mrg _Reference_: 1370 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.6. 1371 1.1 mrg 1372 1.1 mrg 1373 1.1 mrg File: libgomp.info, Node: omp_unset_lock, Next: omp_destroy_lock, Prev: omp_test_lock, Up: Runtime Library Routines 1374 1.1 mrg 1375 1.15 mrg 3.40 'omp_unset_lock' - Unset simple lock 1376 1.1 mrg ========================================= 1377 1.1 mrg 1378 1.1 mrg _Description_: 1379 1.1 mrg A simple lock about to be unset must have been locked by 1380 1.11 mrg 'omp_set_lock' or 'omp_test_lock' before. In addition, the lock 1381 1.11 mrg must be held by the thread calling 'omp_unset_lock'. Then, the 1382 1.11 mrg lock becomes unlocked. If one or more threads attempted to set the 1383 1.11 mrg lock before, one of them is chosen to, again, set the lock to 1384 1.1 mrg itself. 1385 1.1 mrg 1386 1.1 mrg _C/C++_: 1387 1.11 mrg _Prototype_: 'void omp_unset_lock(omp_lock_t *lock);' 1388 1.1 mrg 1389 1.1 mrg _Fortran_: 1390 1.11 mrg _Interface_: 'subroutine omp_unset_lock(svar)' 1391 1.11 mrg 'integer(omp_lock_kind), intent(inout) :: svar' 1392 1.1 mrg 1393 1.1 mrg _See also_: 1394 1.1 mrg *note omp_set_lock::, *note omp_test_lock:: 1395 1.1 mrg 1396 1.1 mrg _Reference_: 1397 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.5. 1398 1.1 mrg 1399 1.1 mrg 1400 1.1 mrg File: libgomp.info, Node: omp_destroy_lock, Next: omp_init_nest_lock, Prev: omp_unset_lock, Up: Runtime Library Routines 1401 1.1 mrg 1402 1.15 mrg 3.41 'omp_destroy_lock' - Destroy simple lock 1403 1.1 mrg ============================================= 1404 1.1 mrg 1405 1.1 mrg _Description_: 1406 1.5 mrg Destroy a simple lock. In order to be destroyed, a simple lock 1407 1.1 mrg must be in the unlocked state. 1408 1.1 mrg 1409 1.1 mrg _C/C++_: 1410 1.11 mrg _Prototype_: 'void omp_destroy_lock(omp_lock_t *lock);' 1411 1.1 mrg 1412 1.1 mrg _Fortran_: 1413 1.11 mrg _Interface_: 'subroutine omp_destroy_lock(svar)' 1414 1.11 mrg 'integer(omp_lock_kind), intent(inout) :: svar' 1415 1.1 mrg 1416 1.1 mrg _See also_: 1417 1.1 mrg *note omp_init_lock:: 1418 1.1 mrg 1419 1.1 mrg _Reference_: 1420 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.3. 1421 1.1 mrg 1422 1.1 mrg 1423 1.1 mrg File: libgomp.info, Node: omp_init_nest_lock, Next: omp_set_nest_lock, Prev: omp_destroy_lock, Up: Runtime Library Routines 1424 1.1 mrg 1425 1.15 mrg 3.42 'omp_init_nest_lock' - Initialize nested lock 1426 1.1 mrg ================================================== 1427 1.1 mrg 1428 1.1 mrg _Description_: 1429 1.3 mrg Initialize a nested lock. After initialization, the lock is in an 1430 1.1 mrg unlocked state and the nesting count is set to zero. 1431 1.1 mrg 1432 1.1 mrg _C/C++_: 1433 1.11 mrg _Prototype_: 'void omp_init_nest_lock(omp_nest_lock_t *lock);' 1434 1.1 mrg 1435 1.1 mrg _Fortran_: 1436 1.11 mrg _Interface_: 'subroutine omp_init_nest_lock(nvar)' 1437 1.11 mrg 'integer(omp_nest_lock_kind), intent(out) :: nvar' 1438 1.1 mrg 1439 1.1 mrg _See also_: 1440 1.1 mrg *note omp_destroy_nest_lock:: 1441 1.1 mrg 1442 1.1 mrg _Reference_: 1443 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.1. 1444 1.1 mrg 1445 1.1 mrg 1446 1.1 mrg File: libgomp.info, Node: omp_set_nest_lock, Next: omp_test_nest_lock, Prev: omp_init_nest_lock, Up: Runtime Library Routines 1447 1.1 mrg 1448 1.15 mrg 3.43 'omp_set_nest_lock' - Wait for and set nested lock 1449 1.1 mrg ======================================================= 1450 1.1 mrg 1451 1.1 mrg _Description_: 1452 1.11 mrg Before setting a nested lock, the lock variable must be initialized 1453 1.11 mrg by 'omp_init_nest_lock'. The calling thread is blocked until the 1454 1.11 mrg lock is available. If the lock is already held by the current 1455 1.11 mrg thread, the nesting count for the lock is incremented. 1456 1.1 mrg 1457 1.1 mrg _C/C++_: 1458 1.11 mrg _Prototype_: 'void omp_set_nest_lock(omp_nest_lock_t *lock);' 1459 1.1 mrg 1460 1.1 mrg _Fortran_: 1461 1.11 mrg _Interface_: 'subroutine omp_set_nest_lock(nvar)' 1462 1.11 mrg 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1463 1.1 mrg 1464 1.1 mrg _See also_: 1465 1.1 mrg *note omp_init_nest_lock::, *note omp_unset_nest_lock:: 1466 1.1 mrg 1467 1.1 mrg _Reference_: 1468 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.4. 1469 1.1 mrg 1470 1.1 mrg 1471 1.1 mrg File: libgomp.info, Node: omp_test_nest_lock, Next: omp_unset_nest_lock, Prev: omp_set_nest_lock, Up: Runtime Library Routines 1472 1.1 mrg 1473 1.15 mrg 3.44 'omp_test_nest_lock' - Test and set nested lock if available 1474 1.1 mrg ================================================================= 1475 1.1 mrg 1476 1.1 mrg _Description_: 1477 1.11 mrg Before setting a nested lock, the lock variable must be initialized 1478 1.11 mrg by 'omp_init_nest_lock'. Contrary to 'omp_set_nest_lock', 1479 1.11 mrg 'omp_test_nest_lock' does not block if the lock is not available. 1480 1.11 mrg If the lock is already held by the current thread, the new nesting 1481 1.11 mrg count is returned. Otherwise, the return value equals zero. 1482 1.1 mrg 1483 1.1 mrg _C/C++_: 1484 1.11 mrg _Prototype_: 'int omp_test_nest_lock(omp_nest_lock_t *lock);' 1485 1.1 mrg 1486 1.1 mrg _Fortran_: 1487 1.11 mrg _Interface_: 'logical function omp_test_nest_lock(nvar)' 1488 1.11 mrg 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1489 1.1 mrg 1490 1.1 mrg _See also_: 1491 1.1 mrg *note omp_init_lock::, *note omp_set_lock::, *note omp_set_lock:: 1492 1.1 mrg 1493 1.1 mrg _Reference_: 1494 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.6. 1495 1.1 mrg 1496 1.1 mrg 1497 1.1 mrg File: libgomp.info, Node: omp_unset_nest_lock, Next: omp_destroy_nest_lock, Prev: omp_test_nest_lock, Up: Runtime Library Routines 1498 1.1 mrg 1499 1.15 mrg 3.45 'omp_unset_nest_lock' - Unset nested lock 1500 1.1 mrg ============================================== 1501 1.1 mrg 1502 1.1 mrg _Description_: 1503 1.1 mrg A nested lock about to be unset must have been locked by 1504 1.11 mrg 'omp_set_nested_lock' or 'omp_test_nested_lock' before. In 1505 1.1 mrg addition, the lock must be held by the thread calling 1506 1.11 mrg 'omp_unset_nested_lock'. If the nesting count drops to zero, the 1507 1.5 mrg lock becomes unlocked. If one ore more threads attempted to set 1508 1.3 mrg the lock before, one of them is chosen to, again, set the lock to 1509 1.1 mrg itself. 1510 1.1 mrg 1511 1.1 mrg _C/C++_: 1512 1.11 mrg _Prototype_: 'void omp_unset_nest_lock(omp_nest_lock_t *lock);' 1513 1.1 mrg 1514 1.1 mrg _Fortran_: 1515 1.11 mrg _Interface_: 'subroutine omp_unset_nest_lock(nvar)' 1516 1.11 mrg 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1517 1.1 mrg 1518 1.1 mrg _See also_: 1519 1.1 mrg *note omp_set_nest_lock:: 1520 1.1 mrg 1521 1.1 mrg _Reference_: 1522 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.5. 1523 1.1 mrg 1524 1.1 mrg 1525 1.1 mrg File: libgomp.info, Node: omp_destroy_nest_lock, Next: omp_get_wtick, Prev: omp_unset_nest_lock, Up: Runtime Library Routines 1526 1.1 mrg 1527 1.15 mrg 3.46 'omp_destroy_nest_lock' - Destroy nested lock 1528 1.1 mrg ================================================== 1529 1.1 mrg 1530 1.1 mrg _Description_: 1531 1.5 mrg Destroy a nested lock. In order to be destroyed, a nested lock 1532 1.1 mrg must be in the unlocked state and its nesting count must equal 1533 1.1 mrg zero. 1534 1.1 mrg 1535 1.1 mrg _C/C++_: 1536 1.11 mrg _Prototype_: 'void omp_destroy_nest_lock(omp_nest_lock_t *);' 1537 1.1 mrg 1538 1.1 mrg _Fortran_: 1539 1.11 mrg _Interface_: 'subroutine omp_destroy_nest_lock(nvar)' 1540 1.11 mrg 'integer(omp_nest_lock_kind), intent(inout) :: nvar' 1541 1.1 mrg 1542 1.1 mrg _See also_: 1543 1.1 mrg *note omp_init_lock:: 1544 1.1 mrg 1545 1.1 mrg _Reference_: 1546 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.3.3. 1547 1.1 mrg 1548 1.1 mrg 1549 1.1 mrg File: libgomp.info, Node: omp_get_wtick, Next: omp_get_wtime, Prev: omp_destroy_nest_lock, Up: Runtime Library Routines 1550 1.1 mrg 1551 1.15 mrg 3.47 'omp_get_wtick' - Get timer precision 1552 1.1 mrg ========================================== 1553 1.1 mrg 1554 1.1 mrg _Description_: 1555 1.1 mrg Gets the timer precision, i.e., the number of seconds between two 1556 1.1 mrg successive clock ticks. 1557 1.1 mrg 1558 1.1 mrg _C/C++_: 1559 1.11 mrg _Prototype_: 'double omp_get_wtick(void);' 1560 1.1 mrg 1561 1.1 mrg _Fortran_: 1562 1.11 mrg _Interface_: 'double precision function omp_get_wtick()' 1563 1.1 mrg 1564 1.1 mrg _See also_: 1565 1.1 mrg *note omp_get_wtime:: 1566 1.1 mrg 1567 1.1 mrg _Reference_: 1568 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.4.2. 1569 1.1 mrg 1570 1.1 mrg 1571 1.15 mrg File: libgomp.info, Node: omp_get_wtime, Next: omp_fulfill_event, Prev: omp_get_wtick, Up: Runtime Library Routines 1572 1.1 mrg 1573 1.15 mrg 3.48 'omp_get_wtime' - Elapsed wall clock time 1574 1.1 mrg ============================================== 1575 1.1 mrg 1576 1.1 mrg _Description_: 1577 1.5 mrg Elapsed wall clock time in seconds. The time is measured per 1578 1.3 mrg thread, no guarantee can be made that two distinct threads measure 1579 1.3 mrg the same time. Time is measured from some "time in the past", 1580 1.3 mrg which is an arbitrary time guaranteed not to change during the 1581 1.3 mrg execution of the program. 1582 1.1 mrg 1583 1.1 mrg _C/C++_: 1584 1.11 mrg _Prototype_: 'double omp_get_wtime(void);' 1585 1.1 mrg 1586 1.1 mrg _Fortran_: 1587 1.11 mrg _Interface_: 'double precision function omp_get_wtime()' 1588 1.1 mrg 1589 1.1 mrg _See also_: 1590 1.1 mrg *note omp_get_wtick:: 1591 1.1 mrg 1592 1.1 mrg _Reference_: 1593 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 3.4.1. 1594 1.1 mrg 1595 1.1 mrg 1596 1.15 mrg File: libgomp.info, Node: omp_fulfill_event, Prev: omp_get_wtime, Up: Runtime Library Routines 1597 1.15 mrg 1598 1.15 mrg 3.49 'omp_fulfill_event' - Fulfill and destroy an OpenMP event 1599 1.15 mrg ============================================================== 1600 1.15 mrg 1601 1.15 mrg _Description_: 1602 1.15 mrg Fulfill the event associated with the event handle argument. 1603 1.15 mrg Currently, it is only used to fulfill events generated by detach 1604 1.15 mrg clauses on task constructs - the effect of fulfilling the event is 1605 1.15 mrg to allow the task to complete. 1606 1.15 mrg 1607 1.15 mrg The result of calling 'omp_fulfill_event' with an event handle 1608 1.15 mrg other than that generated by a detach clause is undefined. Calling 1609 1.15 mrg it with an event handle that has already been fulfilled is also 1610 1.15 mrg undefined. 1611 1.15 mrg 1612 1.15 mrg _C/C++_: 1613 1.15 mrg _Prototype_: 'void omp_fulfill_event(omp_event_handle_t event);' 1614 1.15 mrg 1615 1.15 mrg _Fortran_: 1616 1.15 mrg _Interface_: 'subroutine omp_fulfill_event(event)' 1617 1.15 mrg 'integer (kind=omp_event_handle_kind) :: event' 1618 1.15 mrg 1619 1.15 mrg _Reference_: 1620 1.15 mrg OpenMP specification v5.0 (https://www.openmp.org), Section 3.5.1. 1621 1.15 mrg 1622 1.15 mrg 1623 1.7 mrg File: libgomp.info, Node: Environment Variables, Next: Enabling OpenACC, Prev: Runtime Library Routines, Up: Top 1624 1.1 mrg 1625 1.15 mrg 4 OpenMP Environment Variables 1626 1.12 mrg ****************************** 1627 1.1 mrg 1628 1.11 mrg The environment variables which beginning with 'OMP_' are defined by 1629 1.7 mrg section 4 of the OpenMP specification in version 4.5, while those 1630 1.11 mrg beginning with 'GOMP_' are GNU extensions. 1631 1.1 mrg 1632 1.1 mrg * Menu: 1633 1.1 mrg 1634 1.7 mrg * OMP_CANCELLATION:: Set whether cancellation is activated 1635 1.7 mrg * OMP_DISPLAY_ENV:: Show OpenMP version and environment variables 1636 1.7 mrg * OMP_DEFAULT_DEVICE:: Set the device used in target regions 1637 1.7 mrg * OMP_DYNAMIC:: Dynamic adjustment of threads 1638 1.7 mrg * OMP_MAX_ACTIVE_LEVELS:: Set the maximum number of nested parallel regions 1639 1.7 mrg * OMP_MAX_TASK_PRIORITY:: Set the maximum task priority value 1640 1.7 mrg * OMP_NESTED:: Nested parallel regions 1641 1.15 mrg * OMP_NUM_TEAMS:: Specifies the number of teams to use by teams region 1642 1.7 mrg * OMP_NUM_THREADS:: Specifies the number of threads to use 1643 1.14 mrg * OMP_PROC_BIND:: Whether threads may be moved between CPUs 1644 1.14 mrg * OMP_PLACES:: Specifies on which CPUs the threads should be placed 1645 1.7 mrg * OMP_STACKSIZE:: Set default thread stack size 1646 1.7 mrg * OMP_SCHEDULE:: How threads are scheduled 1647 1.15 mrg * OMP_TARGET_OFFLOAD:: Controls offloading behaviour 1648 1.15 mrg * OMP_TEAMS_THREAD_LIMIT:: Set the maximum number of threads imposed by teams 1649 1.7 mrg * OMP_THREAD_LIMIT:: Set the maximum number of threads 1650 1.7 mrg * OMP_WAIT_POLICY:: How waiting threads are handled 1651 1.7 mrg * GOMP_CPU_AFFINITY:: Bind threads to specific CPUs 1652 1.7 mrg * GOMP_DEBUG:: Enable debugging output 1653 1.7 mrg * GOMP_STACKSIZE:: Set default thread stack size 1654 1.7 mrg * GOMP_SPINCOUNT:: Set the busy-wait spin count 1655 1.7 mrg * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools 1656 1.1 mrg 1657 1.1 mrg 1658 1.5 mrg File: libgomp.info, Node: OMP_CANCELLATION, Next: OMP_DISPLAY_ENV, Up: Environment Variables 1659 1.5 mrg 1660 1.15 mrg 4.1 'OMP_CANCELLATION' - Set whether cancellation is activated 1661 1.5 mrg ============================================================== 1662 1.5 mrg 1663 1.5 mrg _Description_: 1664 1.11 mrg If set to 'TRUE', the cancellation is activated. If set to 'FALSE' 1665 1.11 mrg or if unset, cancellation is disabled and the 'cancel' construct is 1666 1.11 mrg ignored. 1667 1.5 mrg 1668 1.5 mrg _See also_: 1669 1.5 mrg *note omp_get_cancellation:: 1670 1.1 mrg 1671 1.5 mrg _Reference_: 1672 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.11 1673 1.5 mrg 1674 1.5 mrg 1675 1.5 mrg File: libgomp.info, Node: OMP_DISPLAY_ENV, Next: OMP_DEFAULT_DEVICE, Prev: OMP_CANCELLATION, Up: Environment Variables 1676 1.5 mrg 1677 1.15 mrg 4.2 'OMP_DISPLAY_ENV' - Show OpenMP version and environment variables 1678 1.5 mrg ===================================================================== 1679 1.5 mrg 1680 1.5 mrg _Description_: 1681 1.11 mrg If set to 'TRUE', the OpenMP version number and the values 1682 1.5 mrg associated with the OpenMP environment variables are printed to 1683 1.11 mrg 'stderr'. If set to 'VERBOSE', it additionally shows the value of 1684 1.5 mrg the environment variables which are GNU extensions. If undefined 1685 1.11 mrg or set to 'FALSE', this information will not be shown. 1686 1.5 mrg 1687 1.5 mrg _Reference_: 1688 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.12 1689 1.5 mrg 1690 1.5 mrg 1691 1.5 mrg File: libgomp.info, Node: OMP_DEFAULT_DEVICE, Next: OMP_DYNAMIC, Prev: OMP_DISPLAY_ENV, Up: Environment Variables 1692 1.5 mrg 1693 1.15 mrg 4.3 'OMP_DEFAULT_DEVICE' - Set the device used in target regions 1694 1.5 mrg ================================================================ 1695 1.5 mrg 1696 1.5 mrg _Description_: 1697 1.11 mrg Set to choose the device which is used in a 'target' region, unless 1698 1.11 mrg the value is overridden by 'omp_set_default_device' or by a 1699 1.11 mrg 'device' clause. The value shall be the nonnegative device number. 1700 1.11 mrg If no device with the given device number exists, the code is 1701 1.11 mrg executed on the host. If unset, device number 0 will be used. 1702 1.5 mrg 1703 1.5 mrg _See also_: 1704 1.5 mrg *note omp_get_default_device::, *note omp_set_default_device::, 1705 1.5 mrg 1706 1.5 mrg _Reference_: 1707 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.13 1708 1.5 mrg 1709 1.5 mrg 1710 1.5 mrg File: libgomp.info, Node: OMP_DYNAMIC, Next: OMP_MAX_ACTIVE_LEVELS, Prev: OMP_DEFAULT_DEVICE, Up: Environment Variables 1711 1.5 mrg 1712 1.15 mrg 4.4 'OMP_DYNAMIC' - Dynamic adjustment of threads 1713 1.1 mrg ================================================= 1714 1.1 mrg 1715 1.1 mrg _Description_: 1716 1.1 mrg Enable or disable the dynamic adjustment of the number of threads 1717 1.5 mrg within a team. The value of this environment variable shall be 1718 1.11 mrg 'TRUE' or 'FALSE'. If undefined, dynamic adjustment is disabled by 1719 1.11 mrg default. 1720 1.1 mrg 1721 1.1 mrg _See also_: 1722 1.1 mrg *note omp_set_dynamic:: 1723 1.1 mrg 1724 1.1 mrg _Reference_: 1725 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.3 1726 1.1 mrg 1727 1.1 mrg 1728 1.7 mrg File: libgomp.info, Node: OMP_MAX_ACTIVE_LEVELS, Next: OMP_MAX_TASK_PRIORITY, Prev: OMP_DYNAMIC, Up: Environment Variables 1729 1.1 mrg 1730 1.15 mrg 4.5 'OMP_MAX_ACTIVE_LEVELS' - Set the maximum number of nested parallel regions 1731 1.1 mrg =============================================================================== 1732 1.1 mrg 1733 1.1 mrg _Description_: 1734 1.3 mrg Specifies the initial value for the maximum number of nested 1735 1.5 mrg parallel regions. The value of this variable shall be a positive 1736 1.15 mrg integer. If undefined, then if 'OMP_NESTED' is defined and set to 1737 1.15 mrg true, or if 'OMP_NUM_THREADS' or 'OMP_PROC_BIND' are defined and 1738 1.15 mrg set to a list with more than one item, the maximum number of nested 1739 1.15 mrg parallel regions will be initialized to the largest number 1740 1.15 mrg supported, otherwise it will be set to one. 1741 1.1 mrg 1742 1.1 mrg _See also_: 1743 1.15 mrg *note omp_set_max_active_levels::, *note OMP_NESTED:: 1744 1.1 mrg 1745 1.1 mrg _Reference_: 1746 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.9 1747 1.1 mrg 1748 1.1 mrg 1749 1.7 mrg File: libgomp.info, Node: OMP_MAX_TASK_PRIORITY, Next: OMP_NESTED, Prev: OMP_MAX_ACTIVE_LEVELS, Up: Environment Variables 1750 1.7 mrg 1751 1.15 mrg 4.6 'OMP_MAX_TASK_PRIORITY' - Set the maximum priority 1752 1.7 mrg ====================================================== 1753 1.7 mrg 1754 1.11 mrg number that can be set for a task. 1755 1.7 mrg _Description_: 1756 1.11 mrg Specifies the initial value for the maximum priority value that can 1757 1.11 mrg be set for a task. The value of this variable shall be a 1758 1.7 mrg non-negative integer, and zero is allowed. If undefined, the 1759 1.7 mrg default priority is 0. 1760 1.7 mrg 1761 1.7 mrg _See also_: 1762 1.7 mrg *note omp_get_max_task_priority:: 1763 1.1 mrg 1764 1.7 mrg _Reference_: 1765 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.14 1766 1.7 mrg 1767 1.7 mrg 1768 1.15 mrg File: libgomp.info, Node: OMP_NESTED, Next: OMP_NUM_TEAMS, Prev: OMP_MAX_TASK_PRIORITY, Up: Environment Variables 1769 1.7 mrg 1770 1.15 mrg 4.7 'OMP_NESTED' - Nested parallel regions 1771 1.1 mrg ========================================== 1772 1.1 mrg 1773 1.1 mrg _Description_: 1774 1.1 mrg Enable or disable nested parallel regions, i.e., whether team 1775 1.5 mrg members are allowed to create new teams. The value of this 1776 1.15 mrg environment variable shall be 'TRUE' or 'FALSE'. If set to 'TRUE', 1777 1.15 mrg the number of maximum active nested regions supported will by 1778 1.15 mrg default be set to the maximum supported, otherwise it will be set 1779 1.15 mrg to one. If 'OMP_MAX_ACTIVE_LEVELS' is defined, its setting will 1780 1.15 mrg override this setting. If both are undefined, nested parallel 1781 1.15 mrg regions are enabled if 'OMP_NUM_THREADS' or 'OMP_PROC_BINDS' are 1782 1.15 mrg defined to a list with more than one item, otherwise they are 1783 1.15 mrg disabled by default. 1784 1.1 mrg 1785 1.1 mrg _See also_: 1786 1.15 mrg *note omp_set_max_active_levels::, *note omp_set_nested:: 1787 1.1 mrg 1788 1.1 mrg _Reference_: 1789 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.6 1790 1.1 mrg 1791 1.1 mrg 1792 1.15 mrg File: libgomp.info, Node: OMP_NUM_TEAMS, Next: OMP_NUM_THREADS, Prev: OMP_NESTED, Up: Environment Variables 1793 1.15 mrg 1794 1.15 mrg 4.8 'OMP_NUM_TEAMS' - Specifies the number of teams to use by teams region 1795 1.15 mrg ========================================================================== 1796 1.15 mrg 1797 1.15 mrg _Description_: 1798 1.15 mrg Specifies the upper bound for number of teams to use in teams 1799 1.15 mrg regions without explicit 'num_teams' clause. The value of this 1800 1.15 mrg variable shall be a positive integer. If undefined it defaults to 1801 1.15 mrg 0 which means implementation defined upper bound. 1802 1.15 mrg 1803 1.15 mrg _See also_: 1804 1.15 mrg *note omp_set_num_teams:: 1805 1.15 mrg 1806 1.15 mrg _Reference_: 1807 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 6.23 1808 1.15 mrg 1809 1.15 mrg 1810 1.15 mrg File: libgomp.info, Node: OMP_NUM_THREADS, Next: OMP_PROC_BIND, Prev: OMP_NUM_TEAMS, Up: Environment Variables 1811 1.1 mrg 1812 1.15 mrg 4.9 'OMP_NUM_THREADS' - Specifies the number of threads to use 1813 1.1 mrg ============================================================== 1814 1.1 mrg 1815 1.1 mrg _Description_: 1816 1.11 mrg Specifies the default number of threads to use in parallel regions. 1817 1.11 mrg The value of this variable shall be a comma-separated list of 1818 1.15 mrg positive integers; the value specifies the number of threads to use 1819 1.15 mrg for the corresponding nested level. Specifying more than one item 1820 1.15 mrg in the list will automatically enable nesting by default. If 1821 1.15 mrg undefined one thread per CPU is used. 1822 1.1 mrg 1823 1.1 mrg _See also_: 1824 1.15 mrg *note omp_set_num_threads::, *note OMP_NESTED:: 1825 1.1 mrg 1826 1.1 mrg _Reference_: 1827 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.2 1828 1.1 mrg 1829 1.1 mrg 1830 1.5 mrg File: libgomp.info, Node: OMP_PROC_BIND, Next: OMP_PLACES, Prev: OMP_NUM_THREADS, Up: Environment Variables 1831 1.5 mrg 1832 1.15 mrg 4.10 'OMP_PROC_BIND' - Whether threads may be moved between CPUs 1833 1.15 mrg ================================================================ 1834 1.5 mrg 1835 1.5 mrg _Description_: 1836 1.5 mrg Specifies whether threads may be moved between processors. If set 1837 1.14 mrg to 'TRUE', OpenMP threads should not be moved; if set to 'FALSE' 1838 1.5 mrg they may be moved. Alternatively, a comma separated list with the 1839 1.15 mrg values 'PRIMARY', 'MASTER', 'CLOSE' and 'SPREAD' can be used to 1840 1.15 mrg specify the thread affinity policy for the corresponding nesting 1841 1.15 mrg level. With 'PRIMARY' and 'MASTER' the worker threads are in the 1842 1.15 mrg same place partition as the primary thread. With 'CLOSE' those are 1843 1.15 mrg kept close to the primary thread in contiguous place partitions. 1844 1.15 mrg And with 'SPREAD' a sparse distribution across the place partitions 1845 1.15 mrg is used. Specifying more than one item in the list will 1846 1.15 mrg automatically enable nesting by default. 1847 1.5 mrg 1848 1.11 mrg When undefined, 'OMP_PROC_BIND' defaults to 'TRUE' when 1849 1.11 mrg 'OMP_PLACES' or 'GOMP_CPU_AFFINITY' is set and 'FALSE' otherwise. 1850 1.5 mrg 1851 1.5 mrg _See also_: 1852 1.15 mrg *note omp_get_proc_bind::, *note GOMP_CPU_AFFINITY::, *note 1853 1.15 mrg OMP_NESTED::, *note OMP_PLACES:: 1854 1.5 mrg 1855 1.5 mrg _Reference_: 1856 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.4 1857 1.5 mrg 1858 1.5 mrg 1859 1.5 mrg File: libgomp.info, Node: OMP_PLACES, Next: OMP_STACKSIZE, Prev: OMP_PROC_BIND, Up: Environment Variables 1860 1.1 mrg 1861 1.15 mrg 4.11 'OMP_PLACES' - Specifies on which CPUs the threads should be placed 1862 1.14 mrg ======================================================================== 1863 1.1 mrg 1864 1.1 mrg _Description_: 1865 1.11 mrg The thread placement can be either specified using an abstract name 1866 1.11 mrg or by an explicit list of the places. The abstract names 1867 1.15 mrg 'threads', 'cores', 'sockets', 'll_caches' and 'numa_domains' can 1868 1.15 mrg be optionally followed by a positive number in parentheses, which 1869 1.15 mrg denotes the how many places shall be created. With 'threads' each 1870 1.15 mrg place corresponds to a single hardware thread; 'cores' to a single 1871 1.15 mrg core with the corresponding number of hardware threads; with 1872 1.15 mrg 'sockets' the place corresponds to a single socket; with 1873 1.15 mrg 'll_caches' to a set of cores that shares the last level cache on 1874 1.15 mrg the device; and 'numa_domains' to a set of cores for which their 1875 1.15 mrg closest memory on the device is the same memory and at a similar 1876 1.15 mrg distance from the cores. The resulting placement can be shown by 1877 1.15 mrg setting the 'OMP_DISPLAY_ENV' environment variable. 1878 1.5 mrg 1879 1.5 mrg Alternatively, the placement can be specified explicitly as 1880 1.5 mrg comma-separated list of places. A place is specified by set of 1881 1.15 mrg nonnegative numbers in curly braces, denoting the hardware threads. 1882 1.15 mrg The curly braces can be omitted when only a single number has been 1883 1.15 mrg specified. The hardware threads belonging to a place can either be 1884 1.15 mrg specified as comma-separated list of nonnegative thread numbers or 1885 1.15 mrg using an interval. Multiple places can also be either specified by 1886 1.15 mrg a comma-separated list of places or by an interval. To specify an 1887 1.15 mrg interval, a colon followed by the count is placed after the 1888 1.15 mrg hardware thread number or the place. Optionally, the length can be 1889 1.15 mrg followed by a colon and the stride number - otherwise a unit stride 1890 1.15 mrg is assumed. Placing an exclamation mark ('!') directly before a 1891 1.15 mrg curly brace or numbers inside the curly braces (excluding 1892 1.15 mrg intervals) will exclude those hardware threads. 1893 1.15 mrg 1894 1.15 mrg For instance, the following specifies the same places list: 1895 1.15 mrg '"{0,1,2}, {3,4,6}, {7,8,9}, {10,11,12}"'; '"{0:3}, {3:3}, {7:3}, 1896 1.15 mrg {10:3}"'; and '"{0:2}:4:3"'. 1897 1.5 mrg 1898 1.11 mrg If 'OMP_PLACES' and 'GOMP_CPU_AFFINITY' are unset and 1899 1.11 mrg 'OMP_PROC_BIND' is either unset or 'false', threads may be moved 1900 1.5 mrg between CPUs following no placement policy. 1901 1.1 mrg 1902 1.1 mrg _See also_: 1903 1.5 mrg *note OMP_PROC_BIND::, *note GOMP_CPU_AFFINITY::, *note 1904 1.5 mrg omp_get_proc_bind::, *note OMP_DISPLAY_ENV:: 1905 1.1 mrg 1906 1.1 mrg _Reference_: 1907 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.5 1908 1.1 mrg 1909 1.1 mrg 1910 1.5 mrg File: libgomp.info, Node: OMP_STACKSIZE, Next: OMP_SCHEDULE, Prev: OMP_PLACES, Up: Environment Variables 1911 1.1 mrg 1912 1.15 mrg 4.12 'OMP_STACKSIZE' - Set default thread stack size 1913 1.5 mrg ==================================================== 1914 1.1 mrg 1915 1.1 mrg _Description_: 1916 1.1 mrg Set the default thread stack size in kilobytes, unless the number 1917 1.11 mrg is suffixed by 'B', 'K', 'M' or 'G', in which case the size is, 1918 1.11 mrg respectively, in bytes, kilobytes, megabytes or gigabytes. This is 1919 1.11 mrg different from 'pthread_attr_setstacksize' which gets the number of 1920 1.11 mrg bytes as an argument. If the stack size cannot be set due to 1921 1.11 mrg system constraints, an error is reported and the initial stack size 1922 1.11 mrg is left unchanged. If undefined, the stack size is system 1923 1.11 mrg dependent. 1924 1.1 mrg 1925 1.1 mrg _Reference_: 1926 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.7 1927 1.1 mrg 1928 1.1 mrg 1929 1.15 mrg File: libgomp.info, Node: OMP_SCHEDULE, Next: OMP_TARGET_OFFLOAD, Prev: OMP_STACKSIZE, Up: Environment Variables 1930 1.1 mrg 1931 1.15 mrg 4.13 'OMP_SCHEDULE' - How threads are scheduled 1932 1.5 mrg =============================================== 1933 1.1 mrg 1934 1.1 mrg _Description_: 1935 1.11 mrg Allows to specify 'schedule type' and 'chunk size'. The value of 1936 1.11 mrg the variable shall have the form: 'type[,chunk]' where 'type' is 1937 1.11 mrg one of 'static', 'dynamic', 'guided' or 'auto' The optional 'chunk' 1938 1.11 mrg size shall be a positive integer. If undefined, dynamic scheduling 1939 1.11 mrg and a chunk size of 1 is used. 1940 1.1 mrg 1941 1.1 mrg _See also_: 1942 1.5 mrg *note omp_set_schedule:: 1943 1.1 mrg 1944 1.1 mrg _Reference_: 1945 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Sections 1946 1.7 mrg 2.7.1.1 and 4.1 1947 1.1 mrg 1948 1.1 mrg 1949 1.15 mrg File: libgomp.info, Node: OMP_TARGET_OFFLOAD, Next: OMP_TEAMS_THREAD_LIMIT, Prev: OMP_SCHEDULE, Up: Environment Variables 1950 1.15 mrg 1951 1.15 mrg 4.14 'OMP_TARGET_OFFLOAD' - Controls offloading behaviour 1952 1.15 mrg ========================================================= 1953 1.15 mrg 1954 1.15 mrg _Description_: 1955 1.15 mrg Specifies the behaviour with regard to offloading code to a device. 1956 1.15 mrg This variable can be set to one of three values - 'MANDATORY', 1957 1.15 mrg 'DISABLED' or 'DEFAULT'. 1958 1.15 mrg 1959 1.15 mrg If set to 'MANDATORY', the program will terminate with an error if 1960 1.15 mrg the offload device is not present or is not supported. If set to 1961 1.15 mrg 'DISABLED', then offloading is disabled and all code will run on 1962 1.15 mrg the host. If set to 'DEFAULT', the program will try offloading to 1963 1.15 mrg the device first, then fall back to running code on the host if it 1964 1.15 mrg cannot. 1965 1.15 mrg 1966 1.15 mrg If undefined, then the program will behave as if 'DEFAULT' was set. 1967 1.15 mrg 1968 1.15 mrg _Reference_: 1969 1.15 mrg OpenMP specification v5.0 (https://www.openmp.org), Section 6.17 1970 1.15 mrg 1971 1.15 mrg 1972 1.15 mrg File: libgomp.info, Node: OMP_TEAMS_THREAD_LIMIT, Next: OMP_THREAD_LIMIT, Prev: OMP_TARGET_OFFLOAD, Up: Environment Variables 1973 1.15 mrg 1974 1.15 mrg 4.15 'OMP_TEAMS_THREAD_LIMIT' - Set the maximum number of threads imposed by teams 1975 1.15 mrg ================================================================================== 1976 1.15 mrg 1977 1.15 mrg _Description_: 1978 1.15 mrg Specifies an upper bound for the number of threads to use by each 1979 1.15 mrg contention group created by a teams construct without explicit 1980 1.15 mrg 'thread_limit' clause. The value of this variable shall be a 1981 1.15 mrg positive integer. If undefined, the value of 0 is used which 1982 1.15 mrg stands for an implementation defined upper limit. 1983 1.15 mrg 1984 1.15 mrg _See also_: 1985 1.15 mrg *note OMP_THREAD_LIMIT::, *note omp_set_teams_thread_limit:: 1986 1.15 mrg 1987 1.15 mrg _Reference_: 1988 1.15 mrg OpenMP specification v5.1 (https://www.openmp.org), Section 6.24 1989 1.15 mrg 1990 1.15 mrg 1991 1.15 mrg File: libgomp.info, Node: OMP_THREAD_LIMIT, Next: OMP_WAIT_POLICY, Prev: OMP_TEAMS_THREAD_LIMIT, Up: Environment Variables 1992 1.1 mrg 1993 1.15 mrg 4.16 'OMP_THREAD_LIMIT' - Set the maximum number of threads 1994 1.5 mrg =========================================================== 1995 1.1 mrg 1996 1.1 mrg _Description_: 1997 1.5 mrg Specifies the number of threads to use for the whole program. The 1998 1.5 mrg value of this variable shall be a positive integer. If undefined, 1999 1.5 mrg the number of threads is not limited. 2000 1.5 mrg 2001 1.5 mrg _See also_: 2002 1.5 mrg *note OMP_NUM_THREADS::, *note omp_get_thread_limit:: 2003 1.1 mrg 2004 1.1 mrg _Reference_: 2005 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.10 2006 1.3 mrg 2007 1.3 mrg 2008 1.5 mrg File: libgomp.info, Node: OMP_WAIT_POLICY, Next: GOMP_CPU_AFFINITY, Prev: OMP_THREAD_LIMIT, Up: Environment Variables 2009 1.3 mrg 2010 1.15 mrg 4.17 'OMP_WAIT_POLICY' - How waiting threads are handled 2011 1.5 mrg ======================================================== 2012 1.3 mrg 2013 1.3 mrg _Description_: 2014 1.5 mrg Specifies whether waiting threads should be active or passive. If 2015 1.11 mrg the value is 'PASSIVE', waiting threads should not consume CPU 2016 1.11 mrg power while waiting; while the value is 'ACTIVE' specifies that 2017 1.5 mrg they should. If undefined, threads wait actively for a short time 2018 1.5 mrg before waiting passively. 2019 1.3 mrg 2020 1.3 mrg _See also_: 2021 1.5 mrg *note GOMP_SPINCOUNT:: 2022 1.3 mrg 2023 1.3 mrg _Reference_: 2024 1.12 mrg OpenMP specification v4.5 (https://www.openmp.org), Section 4.8 2025 1.1 mrg 2026 1.1 mrg 2027 1.5 mrg File: libgomp.info, Node: GOMP_CPU_AFFINITY, Next: GOMP_DEBUG, Prev: OMP_WAIT_POLICY, Up: Environment Variables 2028 1.1 mrg 2029 1.15 mrg 4.18 'GOMP_CPU_AFFINITY' - Bind threads to specific CPUs 2030 1.3 mrg ======================================================== 2031 1.1 mrg 2032 1.1 mrg _Description_: 2033 1.5 mrg Binds threads to specific CPUs. The variable should contain a 2034 1.5 mrg space-separated or comma-separated list of CPUs. This list may 2035 1.3 mrg contain different kinds of entries: either single CPU numbers in 2036 1.3 mrg any order, a range of CPUs (M-N) or a range with some stride 2037 1.11 mrg (M-N:S). CPU numbers are zero based. For example, 2038 1.11 mrg 'GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"' will bind the initial thread 2039 1.3 mrg to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to 2040 1.3 mrg CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 2041 1.3 mrg 10, 12, and 14 respectively and then start assigning back from the 2042 1.11 mrg beginning of the list. 'GOMP_CPU_AFFINITY=0' binds all threads to 2043 1.3 mrg CPU 0. 2044 1.1 mrg 2045 1.5 mrg There is no libgomp library routine to determine whether a CPU 2046 1.5 mrg affinity specification is in effect. As a workaround, 2047 1.11 mrg language-specific library functions, e.g., 'getenv' in C or 2048 1.11 mrg 'GET_ENVIRONMENT_VARIABLE' in Fortran, may be used to query the 2049 1.11 mrg setting of the 'GOMP_CPU_AFFINITY' environment variable. A defined 2050 1.11 mrg CPU affinity on startup cannot be changed or disabled during the 2051 1.11 mrg runtime of the application. 2052 1.11 mrg 2053 1.11 mrg If both 'GOMP_CPU_AFFINITY' and 'OMP_PROC_BIND' are set, 2054 1.11 mrg 'OMP_PROC_BIND' has a higher precedence. If neither has been set 2055 1.11 mrg and 'OMP_PROC_BIND' is unset, or when 'OMP_PROC_BIND' is set to 2056 1.11 mrg 'FALSE', the host system will handle the assignment of threads to 2057 1.5 mrg CPUs. 2058 1.1 mrg 2059 1.3 mrg _See also_: 2060 1.5 mrg *note OMP_PLACES::, *note OMP_PROC_BIND:: 2061 1.5 mrg 2062 1.5 mrg 2063 1.5 mrg File: libgomp.info, Node: GOMP_DEBUG, Next: GOMP_STACKSIZE, Prev: GOMP_CPU_AFFINITY, Up: Environment Variables 2064 1.5 mrg 2065 1.15 mrg 4.19 'GOMP_DEBUG' - Enable debugging output 2066 1.5 mrg =========================================== 2067 1.5 mrg 2068 1.5 mrg _Description_: 2069 1.11 mrg Enable debugging output. The variable should be set to '0' 2070 1.11 mrg (disabled, also the default if not set), or '1' (enabled). 2071 1.5 mrg 2072 1.5 mrg If enabled, some debugging output will be printed during execution. 2073 1.5 mrg This is currently not specified in more detail, and subject to 2074 1.5 mrg change. 2075 1.3 mrg 2076 1.1 mrg 2077 1.5 mrg File: libgomp.info, Node: GOMP_STACKSIZE, Next: GOMP_SPINCOUNT, Prev: GOMP_DEBUG, Up: Environment Variables 2078 1.1 mrg 2079 1.15 mrg 4.20 'GOMP_STACKSIZE' - Set default thread stack size 2080 1.1 mrg ===================================================== 2081 1.1 mrg 2082 1.1 mrg _Description_: 2083 1.5 mrg Set the default thread stack size in kilobytes. This is different 2084 1.11 mrg from 'pthread_attr_setstacksize' which gets the number of bytes as 2085 1.5 mrg an argument. If the stack size cannot be set due to system 2086 1.3 mrg constraints, an error is reported and the initial stack size is 2087 1.5 mrg left unchanged. If undefined, the stack size is system dependent. 2088 1.1 mrg 2089 1.1 mrg _See also_: 2090 1.3 mrg *note OMP_STACKSIZE:: 2091 1.1 mrg 2092 1.1 mrg _Reference_: 2093 1.1 mrg GCC Patches Mailinglist 2094 1.13 mrg (https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html), GCC 2095 1.1 mrg Patches Mailinglist 2096 1.13 mrg (https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html) 2097 1.1 mrg 2098 1.1 mrg 2099 1.7 mrg File: libgomp.info, Node: GOMP_SPINCOUNT, Next: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_STACKSIZE, Up: Environment Variables 2100 1.5 mrg 2101 1.15 mrg 4.21 'GOMP_SPINCOUNT' - Set the busy-wait spin count 2102 1.5 mrg ==================================================== 2103 1.5 mrg 2104 1.5 mrg _Description_: 2105 1.5 mrg Determines how long a threads waits actively with consuming CPU 2106 1.5 mrg power before waiting passively without consuming CPU power. The 2107 1.11 mrg value may be either 'INFINITE', 'INFINITY' to always wait actively 2108 1.5 mrg or an integer which gives the number of spins of the busy-wait 2109 1.5 mrg loop. The integer may optionally be followed by the following 2110 1.11 mrg suffixes acting as multiplication factors: 'k' (kilo, thousand), 2111 1.11 mrg 'M' (mega, million), 'G' (giga, billion), or 'T' (tera, trillion). 2112 1.11 mrg If undefined, 0 is used when 'OMP_WAIT_POLICY' is 'PASSIVE', 2113 1.11 mrg 300,000 is used when 'OMP_WAIT_POLICY' is undefined and 30 billion 2114 1.11 mrg is used when 'OMP_WAIT_POLICY' is 'ACTIVE'. If there are more 2115 1.11 mrg OpenMP threads than available CPUs, 1000 and 100 spins are used for 2116 1.11 mrg 'OMP_WAIT_POLICY' being 'ACTIVE' or undefined, respectively; unless 2117 1.11 mrg the 'GOMP_SPINCOUNT' is lower or 'OMP_WAIT_POLICY' is 'PASSIVE'. 2118 1.5 mrg 2119 1.5 mrg _See also_: 2120 1.5 mrg *note OMP_WAIT_POLICY:: 2121 1.5 mrg 2122 1.5 mrg 2123 1.7 mrg File: libgomp.info, Node: GOMP_RTEMS_THREAD_POOLS, Prev: GOMP_SPINCOUNT, Up: Environment Variables 2124 1.7 mrg 2125 1.15 mrg 4.22 'GOMP_RTEMS_THREAD_POOLS' - Set the RTEMS specific thread pools 2126 1.7 mrg ==================================================================== 2127 1.7 mrg 2128 1.7 mrg _Description_: 2129 1.7 mrg This environment variable is only used on the RTEMS real-time 2130 1.7 mrg operating system. It determines the scheduler instance specific 2131 1.11 mrg thread pools. The format for 'GOMP_RTEMS_THREAD_POOLS' is a list 2132 1.11 mrg of optional '<thread-pool-count>[$<priority>]@<scheduler-name>' 2133 1.11 mrg configurations separated by ':' where: 2134 1.11 mrg * '<thread-pool-count>' is the thread pool count for this 2135 1.7 mrg scheduler instance. 2136 1.11 mrg * '$<priority>' is an optional priority for the worker threads 2137 1.11 mrg of a thread pool according to 'pthread_setschedparam'. In 2138 1.7 mrg case a priority value is omitted, then a worker thread will 2139 1.15 mrg inherit the priority of the OpenMP primary thread that created 2140 1.7 mrg it. The priority of the worker thread is not changed after 2141 1.15 mrg creation, even if a new OpenMP primary thread using the worker 2142 1.7 mrg has a different priority. 2143 1.11 mrg * '@<scheduler-name>' is the scheduler instance name according 2144 1.7 mrg to the RTEMS application configuration. 2145 1.7 mrg In case no thread pool configuration is specified for a scheduler 2146 1.15 mrg instance, then each OpenMP primary thread of this scheduler 2147 1.15 mrg instance will use its own dynamically allocated thread pool. To 2148 1.15 mrg limit the worker thread count of the thread pools, each OpenMP 2149 1.15 mrg primary thread must call 'omp_set_num_threads'. 2150 1.7 mrg _Example_: 2151 1.11 mrg Lets suppose we have three scheduler instances 'IO', 'WRK0', and 2152 1.11 mrg 'WRK1' with 'GOMP_RTEMS_THREAD_POOLS' set to '"1@WRK0:3$4@WRK1"'. 2153 1.7 mrg Then there are no thread pool restrictions for scheduler instance 2154 1.11 mrg 'IO'. In the scheduler instance 'WRK0' there is one thread pool 2155 1.7 mrg available. Since no priority is specified for this scheduler 2156 1.7 mrg instance, the worker thread inherits the priority of the OpenMP 2157 1.15 mrg primary thread that created it. In the scheduler instance 'WRK1' 2158 1.11 mrg there are three thread pools available and their worker threads run 2159 1.11 mrg at priority four. 2160 1.7 mrg 2161 1.7 mrg 2162 1.7 mrg File: libgomp.info, Node: Enabling OpenACC, Next: OpenACC Runtime Library Routines, Prev: Environment Variables, Up: Top 2163 1.7 mrg 2164 1.15 mrg 5 Enabling OpenACC 2165 1.7 mrg ****************** 2166 1.7 mrg 2167 1.7 mrg To activate the OpenACC extensions for C/C++ and Fortran, the 2168 1.11 mrg compile-time flag '-fopenacc' must be specified. This enables the 2169 1.13 mrg OpenACC directive '#pragma acc' in C/C++ and '!$acc' directives in free 2170 1.11 mrg form, 'c$acc', '*$acc' and '!$acc' directives in fixed form, '!$' 2171 1.11 mrg conditional compilation sentinels in free form and 'c$', '*$' and '!$' 2172 1.7 mrg sentinels in fixed form, for Fortran. The flag also arranges for 2173 1.7 mrg automatic linking of the OpenACC runtime library (*note OpenACC Runtime 2174 1.7 mrg Library Routines::). 2175 1.7 mrg 2176 1.13 mrg See <https://gcc.gnu.org/wiki/OpenACC> for more information. 2177 1.13 mrg 2178 1.7 mrg A complete description of all OpenACC directives accepted may be 2179 1.11 mrg found in the OpenACC (https://www.openacc.org) Application Programming 2180 1.13 mrg Interface manual, version 2.6. 2181 1.7 mrg 2182 1.7 mrg 2183 1.7 mrg File: libgomp.info, Node: OpenACC Runtime Library Routines, Next: OpenACC Environment Variables, Prev: Enabling OpenACC, Up: Top 2184 1.7 mrg 2185 1.15 mrg 6 OpenACC Runtime Library Routines 2186 1.7 mrg ********************************** 2187 1.7 mrg 2188 1.7 mrg The runtime routines described here are defined by section 3 of the 2189 1.13 mrg OpenACC specifications in version 2.6. They have C linkage, and do not 2190 1.11 mrg throw exceptions. Generally, they are available only for the host, with 2191 1.11 mrg the exception of 'acc_on_device', which is available for both the host 2192 1.11 mrg and the acceleration device. 2193 1.7 mrg 2194 1.7 mrg * Menu: 2195 1.7 mrg 2196 1.7 mrg * acc_get_num_devices:: Get number of devices for the given device 2197 1.7 mrg type. 2198 1.7 mrg * acc_set_device_type:: Set type of device accelerator to use. 2199 1.7 mrg * acc_get_device_type:: Get type of device accelerator to be used. 2200 1.7 mrg * acc_set_device_num:: Set device number to use. 2201 1.7 mrg * acc_get_device_num:: Get device number to be used. 2202 1.13 mrg * acc_get_property:: Get device property. 2203 1.7 mrg * acc_async_test:: Tests for completion of a specific asynchronous 2204 1.7 mrg operation. 2205 1.13 mrg * acc_async_test_all:: Tests for completion of all asynchronous 2206 1.7 mrg operations. 2207 1.7 mrg * acc_wait:: Wait for completion of a specific asynchronous 2208 1.7 mrg operation. 2209 1.13 mrg * acc_wait_all:: Waits for completion of all asynchronous 2210 1.7 mrg operations. 2211 1.7 mrg * acc_wait_all_async:: Wait for completion of all asynchronous 2212 1.7 mrg operations. 2213 1.7 mrg * acc_wait_async:: Wait for completion of asynchronous operations. 2214 1.7 mrg * acc_init:: Initialize runtime for a specific device type. 2215 1.7 mrg * acc_shutdown:: Shuts down the runtime for a specific device 2216 1.7 mrg type. 2217 1.7 mrg * acc_on_device:: Whether executing on a particular device 2218 1.7 mrg * acc_malloc:: Allocate device memory. 2219 1.7 mrg * acc_free:: Free device memory. 2220 1.7 mrg * acc_copyin:: Allocate device memory and copy host memory to 2221 1.7 mrg it. 2222 1.7 mrg * acc_present_or_copyin:: If the data is not present on the device, 2223 1.7 mrg allocate device memory and copy from host 2224 1.7 mrg memory. 2225 1.7 mrg * acc_create:: Allocate device memory and map it to host 2226 1.7 mrg memory. 2227 1.7 mrg * acc_present_or_create:: If the data is not present on the device, 2228 1.7 mrg allocate device memory and map it to host 2229 1.7 mrg memory. 2230 1.7 mrg * acc_copyout:: Copy device memory to host memory. 2231 1.7 mrg * acc_delete:: Free device memory. 2232 1.7 mrg * acc_update_device:: Update device memory from mapped host memory. 2233 1.7 mrg * acc_update_self:: Update host memory from mapped device memory. 2234 1.7 mrg * acc_map_data:: Map previously allocated device memory to host 2235 1.7 mrg memory. 2236 1.7 mrg * acc_unmap_data:: Unmap device memory from host memory. 2237 1.7 mrg * acc_deviceptr:: Get device pointer associated with specific 2238 1.7 mrg host address. 2239 1.7 mrg * acc_hostptr:: Get host pointer associated with specific 2240 1.7 mrg device address. 2241 1.13 mrg * acc_is_present:: Indicate whether host variable / array is 2242 1.7 mrg present on device. 2243 1.7 mrg * acc_memcpy_to_device:: Copy host memory to device memory. 2244 1.7 mrg * acc_memcpy_from_device:: Copy device memory to host memory. 2245 1.13 mrg * acc_attach:: Let device pointer point to device-pointer target. 2246 1.13 mrg * acc_detach:: Let device pointer point to host-pointer target. 2247 1.7 mrg 2248 1.7 mrg API routines for target platforms. 2249 1.7 mrg 2250 1.7 mrg * acc_get_current_cuda_device:: Get CUDA device handle. 2251 1.7 mrg * acc_get_current_cuda_context::Get CUDA context handle. 2252 1.7 mrg * acc_get_cuda_stream:: Get CUDA stream handle. 2253 1.7 mrg * acc_set_cuda_stream:: Set CUDA stream handle. 2254 1.7 mrg 2255 1.13 mrg API routines for the OpenACC Profiling Interface. 2256 1.13 mrg 2257 1.13 mrg * acc_prof_register:: Register callbacks. 2258 1.13 mrg * acc_prof_unregister:: Unregister callbacks. 2259 1.13 mrg * acc_prof_lookup:: Obtain inquiry functions. 2260 1.13 mrg * acc_register_library:: Library registration. 2261 1.13 mrg 2262 1.7 mrg 2263 1.7 mrg File: libgomp.info, Node: acc_get_num_devices, Next: acc_set_device_type, Up: OpenACC Runtime Library Routines 2264 1.7 mrg 2265 1.15 mrg 6.1 'acc_get_num_devices' - Get number of devices for given device type 2266 1.7 mrg ======================================================================= 2267 1.7 mrg 2268 1.7 mrg _Description_ 2269 1.7 mrg This function returns a value indicating the number of devices 2270 1.7 mrg available for the device type specified in DEVICETYPE. 2271 1.7 mrg 2272 1.7 mrg _C/C++_: 2273 1.11 mrg _Prototype_: 'int acc_get_num_devices(acc_device_t devicetype);' 2274 1.7 mrg 2275 1.7 mrg _Fortran_: 2276 1.11 mrg _Interface_: 'integer function acc_get_num_devices(devicetype)' 2277 1.11 mrg 'integer(kind=acc_device_kind) devicetype' 2278 1.7 mrg 2279 1.7 mrg _Reference_: 2280 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2281 1.7 mrg 3.2.1. 2282 1.7 mrg 2283 1.7 mrg 2284 1.7 mrg File: libgomp.info, Node: acc_set_device_type, Next: acc_get_device_type, Prev: acc_get_num_devices, Up: OpenACC Runtime Library Routines 2285 1.7 mrg 2286 1.15 mrg 6.2 'acc_set_device_type' - Set type of device accelerator to use. 2287 1.7 mrg ================================================================== 2288 1.7 mrg 2289 1.7 mrg _Description_ 2290 1.13 mrg This function indicates to the runtime library which device type, 2291 1.7 mrg specified in DEVICETYPE, to use when executing a parallel or 2292 1.7 mrg kernels region. 2293 1.7 mrg 2294 1.7 mrg _C/C++_: 2295 1.11 mrg _Prototype_: 'acc_set_device_type(acc_device_t devicetype);' 2296 1.7 mrg 2297 1.7 mrg _Fortran_: 2298 1.11 mrg _Interface_: 'subroutine acc_set_device_type(devicetype)' 2299 1.11 mrg 'integer(kind=acc_device_kind) devicetype' 2300 1.7 mrg 2301 1.7 mrg _Reference_: 2302 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2303 1.7 mrg 3.2.2. 2304 1.7 mrg 2305 1.7 mrg 2306 1.7 mrg File: libgomp.info, Node: acc_get_device_type, Next: acc_set_device_num, Prev: acc_set_device_type, Up: OpenACC Runtime Library Routines 2307 1.7 mrg 2308 1.15 mrg 6.3 'acc_get_device_type' - Get type of device accelerator to be used. 2309 1.7 mrg ====================================================================== 2310 1.7 mrg 2311 1.7 mrg _Description_ 2312 1.7 mrg This function returns what device type will be used when executing 2313 1.7 mrg a parallel or kernels region. 2314 1.7 mrg 2315 1.13 mrg This function returns 'acc_device_none' if 'acc_get_device_type' is 2316 1.13 mrg called from 'acc_ev_device_init_start', 'acc_ev_device_init_end' 2317 1.13 mrg callbacks of the OpenACC Profiling Interface (*note OpenACC 2318 1.13 mrg Profiling Interface::), that is, if the device is currently being 2319 1.13 mrg initialized. 2320 1.13 mrg 2321 1.7 mrg _C/C++_: 2322 1.11 mrg _Prototype_: 'acc_device_t acc_get_device_type(void);' 2323 1.7 mrg 2324 1.7 mrg _Fortran_: 2325 1.11 mrg _Interface_: 'function acc_get_device_type(void)' 2326 1.11 mrg 'integer(kind=acc_device_kind) acc_get_device_type' 2327 1.7 mrg 2328 1.7 mrg _Reference_: 2329 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2330 1.7 mrg 3.2.3. 2331 1.7 mrg 2332 1.7 mrg 2333 1.7 mrg File: libgomp.info, Node: acc_set_device_num, Next: acc_get_device_num, Prev: acc_get_device_type, Up: OpenACC Runtime Library Routines 2334 1.7 mrg 2335 1.15 mrg 6.4 'acc_set_device_num' - Set device number to use. 2336 1.7 mrg ==================================================== 2337 1.7 mrg 2338 1.7 mrg _Description_ 2339 1.7 mrg This function will indicate to the runtime which device number, 2340 1.13 mrg specified by DEVICENUM, associated with the specified device type 2341 1.7 mrg DEVICETYPE. 2342 1.7 mrg 2343 1.7 mrg _C/C++_: 2344 1.13 mrg _Prototype_: 'acc_set_device_num(int devicenum, acc_device_t 2345 1.13 mrg devicetype);' 2346 1.7 mrg 2347 1.7 mrg _Fortran_: 2348 1.11 mrg _Interface_: 'subroutine acc_set_device_num(devicenum, devicetype)' 2349 1.11 mrg 'integer devicenum' 2350 1.11 mrg 'integer(kind=acc_device_kind) devicetype' 2351 1.7 mrg 2352 1.7 mrg _Reference_: 2353 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2354 1.7 mrg 3.2.4. 2355 1.7 mrg 2356 1.7 mrg 2357 1.13 mrg File: libgomp.info, Node: acc_get_device_num, Next: acc_get_property, Prev: acc_set_device_num, Up: OpenACC Runtime Library Routines 2358 1.7 mrg 2359 1.15 mrg 6.5 'acc_get_device_num' - Get device number to be used. 2360 1.7 mrg ======================================================== 2361 1.7 mrg 2362 1.7 mrg _Description_ 2363 1.7 mrg This function returns which device number associated with the 2364 1.7 mrg specified device type DEVICETYPE, will be used when executing a 2365 1.7 mrg parallel or kernels region. 2366 1.7 mrg 2367 1.7 mrg _C/C++_: 2368 1.11 mrg _Prototype_: 'int acc_get_device_num(acc_device_t devicetype);' 2369 1.7 mrg 2370 1.7 mrg _Fortran_: 2371 1.11 mrg _Interface_: 'function acc_get_device_num(devicetype)' 2372 1.11 mrg 'integer(kind=acc_device_kind) devicetype' 2373 1.11 mrg 'integer acc_get_device_num' 2374 1.7 mrg 2375 1.7 mrg _Reference_: 2376 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2377 1.7 mrg 3.2.5. 2378 1.7 mrg 2379 1.7 mrg 2380 1.13 mrg File: libgomp.info, Node: acc_get_property, Next: acc_async_test, Prev: acc_get_device_num, Up: OpenACC Runtime Library Routines 2381 1.13 mrg 2382 1.15 mrg 6.6 'acc_get_property' - Get device property. 2383 1.13 mrg ============================================= 2384 1.13 mrg 2385 1.13 mrg _Description_ 2386 1.13 mrg These routines return the value of the specified PROPERTY for the 2387 1.13 mrg device being queried according to DEVICENUM and DEVICETYPE. 2388 1.13 mrg Integer-valued and string-valued properties are returned by 2389 1.13 mrg 'acc_get_property' and 'acc_get_property_string' respectively. The 2390 1.13 mrg Fortran 'acc_get_property_string' subroutine returns the string 2391 1.13 mrg retrieved in its fourth argument while the remaining entry points 2392 1.13 mrg are functions, which pass the return value as their result. 2393 1.13 mrg 2394 1.13 mrg Note for Fortran, only: the OpenACC technical committee corrected 2395 1.13 mrg and, hence, modified the interface introduced in OpenACC 2.6. The 2396 1.13 mrg kind-value parameter 'acc_device_property' has been renamed to 2397 1.13 mrg 'acc_device_property_kind' for consistency and the return type of 2398 1.13 mrg the 'acc_get_property' function is now a 'c_size_t' integer instead 2399 1.13 mrg of a 'acc_device_property' integer. The parameter 2400 1.13 mrg 'acc_device_property' will continue to be provided, but might be 2401 1.13 mrg removed in a future version of GCC. 2402 1.13 mrg 2403 1.13 mrg _C/C++_: 2404 1.13 mrg _Prototype_: 'size_t acc_get_property(int devicenum, acc_device_t 2405 1.13 mrg devicetype, acc_device_property_t property);' 2406 1.13 mrg _Prototype_: 'const char *acc_get_property_string(int devicenum, 2407 1.13 mrg acc_device_t devicetype, acc_device_property_t 2408 1.13 mrg property);' 2409 1.13 mrg 2410 1.13 mrg _Fortran_: 2411 1.13 mrg _Interface_: 'function acc_get_property(devicenum, devicetype, 2412 1.13 mrg property)' 2413 1.13 mrg _Interface_: 'subroutine acc_get_property_string(devicenum, 2414 1.13 mrg devicetype, property, string)' 2415 1.13 mrg 'use ISO_C_Binding, only: c_size_t' 2416 1.13 mrg 'integer devicenum' 2417 1.13 mrg 'integer(kind=acc_device_kind) devicetype' 2418 1.13 mrg 'integer(kind=acc_device_property_kind) property' 2419 1.13 mrg 'integer(kind=c_size_t) acc_get_property' 2420 1.13 mrg 'character(*) string' 2421 1.13 mrg 2422 1.13 mrg _Reference_: 2423 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2424 1.13 mrg 3.2.6. 2425 1.7 mrg 2426 1.13 mrg 2427 1.13 mrg File: libgomp.info, Node: acc_async_test, Next: acc_async_test_all, Prev: acc_get_property, Up: OpenACC Runtime Library Routines 2428 1.13 mrg 2429 1.15 mrg 6.7 'acc_async_test' - Test for completion of a specific asynchronous operation. 2430 1.7 mrg ================================================================================ 2431 1.7 mrg 2432 1.7 mrg _Description_ 2433 1.13 mrg This function tests for completion of the asynchronous operation 2434 1.11 mrg specified in ARG. In C/C++, a non-zero value will be returned to 2435 1.11 mrg indicate the specified asynchronous operation has completed. While 2436 1.13 mrg Fortran will return a 'true'. If the asynchronous operation has 2437 1.11 mrg not completed, C/C++ returns a zero and Fortran returns a 'false'. 2438 1.7 mrg 2439 1.7 mrg _C/C++_: 2440 1.11 mrg _Prototype_: 'int acc_async_test(int arg);' 2441 1.7 mrg 2442 1.7 mrg _Fortran_: 2443 1.11 mrg _Interface_: 'function acc_async_test(arg)' 2444 1.11 mrg 'integer(kind=acc_handle_kind) arg' 2445 1.11 mrg 'logical acc_async_test' 2446 1.7 mrg 2447 1.7 mrg _Reference_: 2448 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2449 1.13 mrg 3.2.9. 2450 1.7 mrg 2451 1.7 mrg 2452 1.7 mrg File: libgomp.info, Node: acc_async_test_all, Next: acc_wait, Prev: acc_async_test, Up: OpenACC Runtime Library Routines 2453 1.7 mrg 2454 1.15 mrg 6.8 'acc_async_test_all' - Tests for completion of all asynchronous operations. 2455 1.7 mrg =============================================================================== 2456 1.7 mrg 2457 1.7 mrg _Description_ 2458 1.13 mrg This function tests for completion of all asynchronous operations. 2459 1.7 mrg In C/C++, a non-zero value will be returned to indicate all 2460 1.11 mrg asynchronous operations have completed. While Fortran will return 2461 1.11 mrg a 'true'. If any asynchronous operation has not completed, C/C++ 2462 1.11 mrg returns a zero and Fortran returns a 'false'. 2463 1.7 mrg 2464 1.7 mrg _C/C++_: 2465 1.11 mrg _Prototype_: 'int acc_async_test_all(void);' 2466 1.7 mrg 2467 1.7 mrg _Fortran_: 2468 1.11 mrg _Interface_: 'function acc_async_test()' 2469 1.11 mrg 'logical acc_get_device_num' 2470 1.7 mrg 2471 1.7 mrg _Reference_: 2472 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2473 1.13 mrg 3.2.10. 2474 1.7 mrg 2475 1.7 mrg 2476 1.7 mrg File: libgomp.info, Node: acc_wait, Next: acc_wait_all, Prev: acc_async_test_all, Up: OpenACC Runtime Library Routines 2477 1.7 mrg 2478 1.15 mrg 6.9 'acc_wait' - Wait for completion of a specific asynchronous operation. 2479 1.7 mrg ========================================================================== 2480 1.7 mrg 2481 1.7 mrg _Description_ 2482 1.7 mrg This function waits for completion of the asynchronous operation 2483 1.7 mrg specified in ARG. 2484 1.7 mrg 2485 1.7 mrg _C/C++_: 2486 1.11 mrg _Prototype_: 'acc_wait(arg);' 2487 1.11 mrg _Prototype 'acc_async_wait(arg);' 2488 1.11 mrg (OpenACC 1.0 2489 1.11 mrg compatibility)_: 2490 1.7 mrg 2491 1.7 mrg _Fortran_: 2492 1.11 mrg _Interface_: 'subroutine acc_wait(arg)' 2493 1.11 mrg 'integer(acc_handle_kind) arg' 2494 1.11 mrg _Interface 'subroutine acc_async_wait(arg)' 2495 1.11 mrg (OpenACC 1.0 2496 1.11 mrg compatibility)_: 2497 1.11 mrg 'integer(acc_handle_kind) arg' 2498 1.7 mrg 2499 1.7 mrg _Reference_: 2500 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2501 1.13 mrg 3.2.11. 2502 1.7 mrg 2503 1.7 mrg 2504 1.7 mrg File: libgomp.info, Node: acc_wait_all, Next: acc_wait_all_async, Prev: acc_wait, Up: OpenACC Runtime Library Routines 2505 1.7 mrg 2506 1.15 mrg 6.10 'acc_wait_all' - Waits for completion of all asynchronous operations. 2507 1.13 mrg ========================================================================== 2508 1.7 mrg 2509 1.7 mrg _Description_ 2510 1.7 mrg This function waits for the completion of all asynchronous 2511 1.7 mrg operations. 2512 1.7 mrg 2513 1.7 mrg _C/C++_: 2514 1.11 mrg _Prototype_: 'acc_wait_all(void);' 2515 1.11 mrg _Prototype 'acc_async_wait_all(void);' 2516 1.11 mrg (OpenACC 1.0 2517 1.11 mrg compatibility)_: 2518 1.7 mrg 2519 1.7 mrg _Fortran_: 2520 1.11 mrg _Interface_: 'subroutine acc_wait_all()' 2521 1.11 mrg _Interface 'subroutine acc_async_wait_all()' 2522 1.11 mrg (OpenACC 1.0 2523 1.11 mrg compatibility)_: 2524 1.7 mrg 2525 1.7 mrg _Reference_: 2526 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2527 1.13 mrg 3.2.13. 2528 1.7 mrg 2529 1.7 mrg 2530 1.7 mrg File: libgomp.info, Node: acc_wait_all_async, Next: acc_wait_async, Prev: acc_wait_all, Up: OpenACC Runtime Library Routines 2531 1.7 mrg 2532 1.15 mrg 6.11 'acc_wait_all_async' - Wait for completion of all asynchronous operations. 2533 1.7 mrg =============================================================================== 2534 1.7 mrg 2535 1.7 mrg _Description_ 2536 1.7 mrg This function enqueues a wait operation on the queue ASYNC for any 2537 1.7 mrg and all asynchronous operations that have been previously enqueued 2538 1.7 mrg on any queue. 2539 1.7 mrg 2540 1.7 mrg _C/C++_: 2541 1.11 mrg _Prototype_: 'acc_wait_all_async(int async);' 2542 1.7 mrg 2543 1.7 mrg _Fortran_: 2544 1.11 mrg _Interface_: 'subroutine acc_wait_all_async(async)' 2545 1.11 mrg 'integer(acc_handle_kind) async' 2546 1.7 mrg 2547 1.7 mrg _Reference_: 2548 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2549 1.13 mrg 3.2.14. 2550 1.7 mrg 2551 1.7 mrg 2552 1.7 mrg File: libgomp.info, Node: acc_wait_async, Next: acc_init, Prev: acc_wait_all_async, Up: OpenACC Runtime Library Routines 2553 1.7 mrg 2554 1.15 mrg 6.12 'acc_wait_async' - Wait for completion of asynchronous operations. 2555 1.7 mrg ======================================================================= 2556 1.7 mrg 2557 1.7 mrg _Description_ 2558 1.7 mrg This function enqueues a wait operation on queue ASYNC for any and 2559 1.7 mrg all asynchronous operations enqueued on queue ARG. 2560 1.7 mrg 2561 1.7 mrg _C/C++_: 2562 1.11 mrg _Prototype_: 'acc_wait_async(int arg, int async);' 2563 1.7 mrg 2564 1.7 mrg _Fortran_: 2565 1.11 mrg _Interface_: 'subroutine acc_wait_async(arg, async)' 2566 1.11 mrg 'integer(acc_handle_kind) arg, async' 2567 1.7 mrg 2568 1.7 mrg _Reference_: 2569 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2570 1.13 mrg 3.2.12. 2571 1.7 mrg 2572 1.7 mrg 2573 1.7 mrg File: libgomp.info, Node: acc_init, Next: acc_shutdown, Prev: acc_wait_async, Up: OpenACC Runtime Library Routines 2574 1.7 mrg 2575 1.15 mrg 6.13 'acc_init' - Initialize runtime for a specific device type. 2576 1.7 mrg ================================================================ 2577 1.7 mrg 2578 1.7 mrg _Description_ 2579 1.11 mrg This function initializes the runtime for the device type specified 2580 1.11 mrg in DEVICETYPE. 2581 1.7 mrg 2582 1.7 mrg _C/C++_: 2583 1.11 mrg _Prototype_: 'acc_init(acc_device_t devicetype);' 2584 1.7 mrg 2585 1.7 mrg _Fortran_: 2586 1.11 mrg _Interface_: 'subroutine acc_init(devicetype)' 2587 1.11 mrg 'integer(acc_device_kind) devicetype' 2588 1.7 mrg 2589 1.7 mrg _Reference_: 2590 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2591 1.13 mrg 3.2.7. 2592 1.7 mrg 2593 1.7 mrg 2594 1.7 mrg File: libgomp.info, Node: acc_shutdown, Next: acc_on_device, Prev: acc_init, Up: OpenACC Runtime Library Routines 2595 1.7 mrg 2596 1.15 mrg 6.14 'acc_shutdown' - Shuts down the runtime for a specific device type. 2597 1.7 mrg ======================================================================== 2598 1.7 mrg 2599 1.7 mrg _Description_ 2600 1.7 mrg This function shuts down the runtime for the device type specified 2601 1.7 mrg in DEVICETYPE. 2602 1.7 mrg 2603 1.7 mrg _C/C++_: 2604 1.11 mrg _Prototype_: 'acc_shutdown(acc_device_t devicetype);' 2605 1.7 mrg 2606 1.7 mrg _Fortran_: 2607 1.11 mrg _Interface_: 'subroutine acc_shutdown(devicetype)' 2608 1.11 mrg 'integer(acc_device_kind) devicetype' 2609 1.7 mrg 2610 1.7 mrg _Reference_: 2611 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2612 1.13 mrg 3.2.8. 2613 1.7 mrg 2614 1.7 mrg 2615 1.7 mrg File: libgomp.info, Node: acc_on_device, Next: acc_malloc, Prev: acc_shutdown, Up: OpenACC Runtime Library Routines 2616 1.7 mrg 2617 1.15 mrg 6.15 'acc_on_device' - Whether executing on a particular device 2618 1.7 mrg =============================================================== 2619 1.7 mrg 2620 1.7 mrg _Description_: 2621 1.7 mrg This function returns whether the program is executing on a 2622 1.11 mrg particular device specified in DEVICETYPE. In C/C++ a non-zero 2623 1.13 mrg value is returned to indicate the device is executing on the 2624 1.11 mrg specified device type. In Fortran, 'true' will be returned. If 2625 1.7 mrg the program is not executing on the specified device type C/C++ 2626 1.11 mrg will return a zero, while Fortran will return 'false'. 2627 1.7 mrg 2628 1.7 mrg _C/C++_: 2629 1.11 mrg _Prototype_: 'acc_on_device(acc_device_t devicetype);' 2630 1.7 mrg 2631 1.7 mrg _Fortran_: 2632 1.11 mrg _Interface_: 'function acc_on_device(devicetype)' 2633 1.11 mrg 'integer(acc_device_kind) devicetype' 2634 1.11 mrg 'logical acc_on_device' 2635 1.7 mrg 2636 1.7 mrg _Reference_: 2637 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2638 1.13 mrg 3.2.17. 2639 1.7 mrg 2640 1.7 mrg 2641 1.7 mrg File: libgomp.info, Node: acc_malloc, Next: acc_free, Prev: acc_on_device, Up: OpenACC Runtime Library Routines 2642 1.7 mrg 2643 1.15 mrg 6.16 'acc_malloc' - Allocate device memory. 2644 1.7 mrg =========================================== 2645 1.7 mrg 2646 1.7 mrg _Description_ 2647 1.11 mrg This function allocates LEN bytes of device memory. It returns the 2648 1.7 mrg device address of the allocated memory. 2649 1.7 mrg 2650 1.7 mrg _C/C++_: 2651 1.11 mrg _Prototype_: 'd_void* acc_malloc(size_t len);' 2652 1.7 mrg 2653 1.7 mrg _Reference_: 2654 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2655 1.13 mrg 3.2.18. 2656 1.7 mrg 2657 1.7 mrg 2658 1.7 mrg File: libgomp.info, Node: acc_free, Next: acc_copyin, Prev: acc_malloc, Up: OpenACC Runtime Library Routines 2659 1.7 mrg 2660 1.15 mrg 6.17 'acc_free' - Free device memory. 2661 1.7 mrg ===================================== 2662 1.7 mrg 2663 1.7 mrg _Description_ 2664 1.11 mrg Free previously allocated device memory at the device address 'a'. 2665 1.7 mrg 2666 1.7 mrg _C/C++_: 2667 1.11 mrg _Prototype_: 'acc_free(d_void *a);' 2668 1.7 mrg 2669 1.7 mrg _Reference_: 2670 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2671 1.13 mrg 3.2.19. 2672 1.7 mrg 2673 1.7 mrg 2674 1.7 mrg File: libgomp.info, Node: acc_copyin, Next: acc_present_or_copyin, Prev: acc_free, Up: OpenACC Runtime Library Routines 2675 1.7 mrg 2676 1.15 mrg 6.18 'acc_copyin' - Allocate device memory and copy host memory to it. 2677 1.7 mrg ====================================================================== 2678 1.7 mrg 2679 1.7 mrg _Description_ 2680 1.7 mrg In C/C++, this function allocates LEN bytes of device memory and 2681 1.11 mrg maps it to the specified host address in A. The device address of 2682 1.7 mrg the newly allocated device memory is returned. 2683 1.7 mrg 2684 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2685 1.11 mrg specifies a contiguous array section. The second form A specifies 2686 1.11 mrg a variable or array element and LEN specifies the length in bytes. 2687 1.7 mrg 2688 1.7 mrg _C/C++_: 2689 1.11 mrg _Prototype_: 'void *acc_copyin(h_void *a, size_t len);' 2690 1.13 mrg _Prototype_: 'void *acc_copyin_async(h_void *a, size_t len, int 2691 1.13 mrg async);' 2692 1.7 mrg 2693 1.7 mrg _Fortran_: 2694 1.11 mrg _Interface_: 'subroutine acc_copyin(a)' 2695 1.11 mrg 'type, dimension(:[,:]...) :: a' 2696 1.11 mrg _Interface_: 'subroutine acc_copyin(a, len)' 2697 1.11 mrg 'type, dimension(:[,:]...) :: a' 2698 1.11 mrg 'integer len' 2699 1.13 mrg _Interface_: 'subroutine acc_copyin_async(a, async)' 2700 1.13 mrg 'type, dimension(:[,:]...) :: a' 2701 1.13 mrg 'integer(acc_handle_kind) :: async' 2702 1.13 mrg _Interface_: 'subroutine acc_copyin_async(a, len, async)' 2703 1.13 mrg 'type, dimension(:[,:]...) :: a' 2704 1.13 mrg 'integer len' 2705 1.13 mrg 'integer(acc_handle_kind) :: async' 2706 1.7 mrg 2707 1.7 mrg _Reference_: 2708 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2709 1.13 mrg 3.2.20. 2710 1.7 mrg 2711 1.7 mrg 2712 1.7 mrg File: libgomp.info, Node: acc_present_or_copyin, Next: acc_create, Prev: acc_copyin, Up: OpenACC Runtime Library Routines 2713 1.7 mrg 2714 1.15 mrg 6.19 'acc_present_or_copyin' - If the data is not present on the device, allocate device memory and copy from host memory. 2715 1.7 mrg ========================================================================================================================== 2716 1.7 mrg 2717 1.7 mrg _Description_ 2718 1.13 mrg This function tests if the host data specified by A and of length 2719 1.11 mrg LEN is present or not. If it is not present, then device memory 2720 1.11 mrg will be allocated and the host memory copied. The device address 2721 1.11 mrg of the newly allocated device memory is returned. 2722 1.7 mrg 2723 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2724 1.11 mrg specifies a contiguous array section. The second form A specifies 2725 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2726 1.7 mrg 2727 1.13 mrg Note that 'acc_present_or_copyin' and 'acc_pcopyin' exist for 2728 1.13 mrg backward compatibility with OpenACC 2.0; use *note acc_copyin:: 2729 1.13 mrg instead. 2730 1.13 mrg 2731 1.7 mrg _C/C++_: 2732 1.11 mrg _Prototype_: 'void *acc_present_or_copyin(h_void *a, size_t len);' 2733 1.11 mrg _Prototype_: 'void *acc_pcopyin(h_void *a, size_t len);' 2734 1.7 mrg 2735 1.7 mrg _Fortran_: 2736 1.11 mrg _Interface_: 'subroutine acc_present_or_copyin(a)' 2737 1.11 mrg 'type, dimension(:[,:]...) :: a' 2738 1.11 mrg _Interface_: 'subroutine acc_present_or_copyin(a, len)' 2739 1.11 mrg 'type, dimension(:[,:]...) :: a' 2740 1.11 mrg 'integer len' 2741 1.11 mrg _Interface_: 'subroutine acc_pcopyin(a)' 2742 1.11 mrg 'type, dimension(:[,:]...) :: a' 2743 1.11 mrg _Interface_: 'subroutine acc_pcopyin(a, len)' 2744 1.11 mrg 'type, dimension(:[,:]...) :: a' 2745 1.11 mrg 'integer len' 2746 1.7 mrg 2747 1.7 mrg _Reference_: 2748 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2749 1.13 mrg 3.2.20. 2750 1.7 mrg 2751 1.7 mrg 2752 1.7 mrg File: libgomp.info, Node: acc_create, Next: acc_present_or_create, Prev: acc_present_or_copyin, Up: OpenACC Runtime Library Routines 2753 1.7 mrg 2754 1.15 mrg 6.20 'acc_create' - Allocate device memory and map it to host memory. 2755 1.7 mrg ===================================================================== 2756 1.7 mrg 2757 1.7 mrg _Description_ 2758 1.7 mrg This function allocates device memory and maps it to host memory 2759 1.11 mrg specified by the host address A with a length of LEN bytes. In 2760 1.7 mrg C/C++, the function returns the device address of the allocated 2761 1.7 mrg device memory. 2762 1.7 mrg 2763 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2764 1.11 mrg specifies a contiguous array section. The second form A specifies 2765 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2766 1.7 mrg 2767 1.7 mrg _C/C++_: 2768 1.11 mrg _Prototype_: 'void *acc_create(h_void *a, size_t len);' 2769 1.13 mrg _Prototype_: 'void *acc_create_async(h_void *a, size_t len, int 2770 1.13 mrg async);' 2771 1.7 mrg 2772 1.7 mrg _Fortran_: 2773 1.11 mrg _Interface_: 'subroutine acc_create(a)' 2774 1.11 mrg 'type, dimension(:[,:]...) :: a' 2775 1.11 mrg _Interface_: 'subroutine acc_create(a, len)' 2776 1.11 mrg 'type, dimension(:[,:]...) :: a' 2777 1.11 mrg 'integer len' 2778 1.13 mrg _Interface_: 'subroutine acc_create_async(a, async)' 2779 1.13 mrg 'type, dimension(:[,:]...) :: a' 2780 1.13 mrg 'integer(acc_handle_kind) :: async' 2781 1.13 mrg _Interface_: 'subroutine acc_create_async(a, len, async)' 2782 1.13 mrg 'type, dimension(:[,:]...) :: a' 2783 1.13 mrg 'integer len' 2784 1.13 mrg 'integer(acc_handle_kind) :: async' 2785 1.7 mrg 2786 1.7 mrg _Reference_: 2787 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2788 1.13 mrg 3.2.21. 2789 1.7 mrg 2790 1.7 mrg 2791 1.7 mrg File: libgomp.info, Node: acc_present_or_create, Next: acc_copyout, Prev: acc_create, Up: OpenACC Runtime Library Routines 2792 1.7 mrg 2793 1.15 mrg 6.21 'acc_present_or_create' - If the data is not present on the device, allocate device memory and map it to host memory. 2794 1.7 mrg ========================================================================================================================== 2795 1.1 mrg 2796 1.7 mrg _Description_ 2797 1.13 mrg This function tests if the host data specified by A and of length 2798 1.11 mrg LEN is present or not. If it is not present, then device memory 2799 1.11 mrg will be allocated and mapped to host memory. In C/C++, the device 2800 1.7 mrg address of the newly allocated device memory is returned. 2801 1.7 mrg 2802 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2803 1.11 mrg specifies a contiguous array section. The second form A specifies 2804 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2805 1.7 mrg 2806 1.13 mrg Note that 'acc_present_or_create' and 'acc_pcreate' exist for 2807 1.13 mrg backward compatibility with OpenACC 2.0; use *note acc_create:: 2808 1.13 mrg instead. 2809 1.13 mrg 2810 1.7 mrg _C/C++_: 2811 1.11 mrg _Prototype_: 'void *acc_present_or_create(h_void *a, size_t len)' 2812 1.11 mrg _Prototype_: 'void *acc_pcreate(h_void *a, size_t len)' 2813 1.7 mrg 2814 1.7 mrg _Fortran_: 2815 1.11 mrg _Interface_: 'subroutine acc_present_or_create(a)' 2816 1.11 mrg 'type, dimension(:[,:]...) :: a' 2817 1.11 mrg _Interface_: 'subroutine acc_present_or_create(a, len)' 2818 1.11 mrg 'type, dimension(:[,:]...) :: a' 2819 1.11 mrg 'integer len' 2820 1.11 mrg _Interface_: 'subroutine acc_pcreate(a)' 2821 1.11 mrg 'type, dimension(:[,:]...) :: a' 2822 1.11 mrg _Interface_: 'subroutine acc_pcreate(a, len)' 2823 1.11 mrg 'type, dimension(:[,:]...) :: a' 2824 1.11 mrg 'integer len' 2825 1.7 mrg 2826 1.7 mrg _Reference_: 2827 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2828 1.13 mrg 3.2.21. 2829 1.7 mrg 2830 1.7 mrg 2831 1.7 mrg File: libgomp.info, Node: acc_copyout, Next: acc_delete, Prev: acc_present_or_create, Up: OpenACC Runtime Library Routines 2832 1.7 mrg 2833 1.15 mrg 6.22 'acc_copyout' - Copy device memory to host memory. 2834 1.7 mrg ======================================================= 2835 1.7 mrg 2836 1.7 mrg _Description_ 2837 1.7 mrg This function copies mapped device memory to host memory which is 2838 1.7 mrg specified by host address A for a length LEN bytes in C/C++. 2839 1.7 mrg 2840 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2841 1.11 mrg specifies a contiguous array section. The second form A specifies 2842 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2843 1.7 mrg 2844 1.7 mrg _C/C++_: 2845 1.11 mrg _Prototype_: 'acc_copyout(h_void *a, size_t len);' 2846 1.13 mrg _Prototype_: 'acc_copyout_async(h_void *a, size_t len, int async);' 2847 1.13 mrg _Prototype_: 'acc_copyout_finalize(h_void *a, size_t len);' 2848 1.13 mrg _Prototype_: 'acc_copyout_finalize_async(h_void *a, size_t len, int 2849 1.13 mrg async);' 2850 1.7 mrg 2851 1.7 mrg _Fortran_: 2852 1.11 mrg _Interface_: 'subroutine acc_copyout(a)' 2853 1.11 mrg 'type, dimension(:[,:]...) :: a' 2854 1.11 mrg _Interface_: 'subroutine acc_copyout(a, len)' 2855 1.11 mrg 'type, dimension(:[,:]...) :: a' 2856 1.11 mrg 'integer len' 2857 1.13 mrg _Interface_: 'subroutine acc_copyout_async(a, async)' 2858 1.13 mrg 'type, dimension(:[,:]...) :: a' 2859 1.13 mrg 'integer(acc_handle_kind) :: async' 2860 1.13 mrg _Interface_: 'subroutine acc_copyout_async(a, len, async)' 2861 1.13 mrg 'type, dimension(:[,:]...) :: a' 2862 1.13 mrg 'integer len' 2863 1.13 mrg 'integer(acc_handle_kind) :: async' 2864 1.13 mrg _Interface_: 'subroutine acc_copyout_finalize(a)' 2865 1.13 mrg 'type, dimension(:[,:]...) :: a' 2866 1.13 mrg _Interface_: 'subroutine acc_copyout_finalize(a, len)' 2867 1.13 mrg 'type, dimension(:[,:]...) :: a' 2868 1.13 mrg 'integer len' 2869 1.13 mrg _Interface_: 'subroutine acc_copyout_finalize_async(a, async)' 2870 1.13 mrg 'type, dimension(:[,:]...) :: a' 2871 1.13 mrg 'integer(acc_handle_kind) :: async' 2872 1.13 mrg _Interface_: 'subroutine acc_copyout_finalize_async(a, len, async)' 2873 1.13 mrg 'type, dimension(:[,:]...) :: a' 2874 1.13 mrg 'integer len' 2875 1.13 mrg 'integer(acc_handle_kind) :: async' 2876 1.7 mrg 2877 1.7 mrg _Reference_: 2878 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2879 1.13 mrg 3.2.22. 2880 1.7 mrg 2881 1.7 mrg 2882 1.7 mrg File: libgomp.info, Node: acc_delete, Next: acc_update_device, Prev: acc_copyout, Up: OpenACC Runtime Library Routines 2883 1.7 mrg 2884 1.15 mrg 6.23 'acc_delete' - Free device memory. 2885 1.7 mrg ======================================= 2886 1.7 mrg 2887 1.7 mrg _Description_ 2888 1.7 mrg This function frees previously allocated device memory specified by 2889 1.7 mrg the device address A and the length of LEN bytes. 2890 1.7 mrg 2891 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2892 1.11 mrg specifies a contiguous array section. The second form A specifies 2893 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2894 1.7 mrg 2895 1.7 mrg _C/C++_: 2896 1.11 mrg _Prototype_: 'acc_delete(h_void *a, size_t len);' 2897 1.13 mrg _Prototype_: 'acc_delete_async(h_void *a, size_t len, int async);' 2898 1.13 mrg _Prototype_: 'acc_delete_finalize(h_void *a, size_t len);' 2899 1.13 mrg _Prototype_: 'acc_delete_finalize_async(h_void *a, size_t len, int 2900 1.13 mrg async);' 2901 1.7 mrg 2902 1.7 mrg _Fortran_: 2903 1.11 mrg _Interface_: 'subroutine acc_delete(a)' 2904 1.11 mrg 'type, dimension(:[,:]...) :: a' 2905 1.11 mrg _Interface_: 'subroutine acc_delete(a, len)' 2906 1.11 mrg 'type, dimension(:[,:]...) :: a' 2907 1.11 mrg 'integer len' 2908 1.13 mrg _Interface_: 'subroutine acc_delete_async(a, async)' 2909 1.13 mrg 'type, dimension(:[,:]...) :: a' 2910 1.13 mrg 'integer(acc_handle_kind) :: async' 2911 1.13 mrg _Interface_: 'subroutine acc_delete_async(a, len, async)' 2912 1.13 mrg 'type, dimension(:[,:]...) :: a' 2913 1.13 mrg 'integer len' 2914 1.13 mrg 'integer(acc_handle_kind) :: async' 2915 1.13 mrg _Interface_: 'subroutine acc_delete_finalize(a)' 2916 1.13 mrg 'type, dimension(:[,:]...) :: a' 2917 1.13 mrg _Interface_: 'subroutine acc_delete_finalize(a, len)' 2918 1.13 mrg 'type, dimension(:[,:]...) :: a' 2919 1.13 mrg 'integer len' 2920 1.13 mrg _Interface_: 'subroutine acc_delete_async_finalize(a, async)' 2921 1.13 mrg 'type, dimension(:[,:]...) :: a' 2922 1.13 mrg 'integer(acc_handle_kind) :: async' 2923 1.13 mrg _Interface_: 'subroutine acc_delete_async_finalize(a, len, async)' 2924 1.13 mrg 'type, dimension(:[,:]...) :: a' 2925 1.13 mrg 'integer len' 2926 1.13 mrg 'integer(acc_handle_kind) :: async' 2927 1.7 mrg 2928 1.7 mrg _Reference_: 2929 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2930 1.13 mrg 3.2.23. 2931 1.7 mrg 2932 1.7 mrg 2933 1.7 mrg File: libgomp.info, Node: acc_update_device, Next: acc_update_self, Prev: acc_delete, Up: OpenACC Runtime Library Routines 2934 1.7 mrg 2935 1.15 mrg 6.24 'acc_update_device' - Update device memory from mapped host memory. 2936 1.7 mrg ======================================================================== 2937 1.7 mrg 2938 1.7 mrg _Description_ 2939 1.7 mrg This function updates the device copy from the previously mapped 2940 1.7 mrg host memory. The host memory is specified with the host address A 2941 1.7 mrg and a length of LEN bytes. 2942 1.7 mrg 2943 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2944 1.11 mrg specifies a contiguous array section. The second form A specifies 2945 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2946 1.7 mrg 2947 1.7 mrg _C/C++_: 2948 1.11 mrg _Prototype_: 'acc_update_device(h_void *a, size_t len);' 2949 1.13 mrg _Prototype_: 'acc_update_device(h_void *a, size_t len, async);' 2950 1.7 mrg 2951 1.7 mrg _Fortran_: 2952 1.11 mrg _Interface_: 'subroutine acc_update_device(a)' 2953 1.11 mrg 'type, dimension(:[,:]...) :: a' 2954 1.11 mrg _Interface_: 'subroutine acc_update_device(a, len)' 2955 1.11 mrg 'type, dimension(:[,:]...) :: a' 2956 1.11 mrg 'integer len' 2957 1.13 mrg _Interface_: 'subroutine acc_update_device_async(a, async)' 2958 1.13 mrg 'type, dimension(:[,:]...) :: a' 2959 1.13 mrg 'integer(acc_handle_kind) :: async' 2960 1.13 mrg _Interface_: 'subroutine acc_update_device_async(a, len, async)' 2961 1.13 mrg 'type, dimension(:[,:]...) :: a' 2962 1.13 mrg 'integer len' 2963 1.13 mrg 'integer(acc_handle_kind) :: async' 2964 1.7 mrg 2965 1.7 mrg _Reference_: 2966 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 2967 1.13 mrg 3.2.24. 2968 1.7 mrg 2969 1.7 mrg 2970 1.7 mrg File: libgomp.info, Node: acc_update_self, Next: acc_map_data, Prev: acc_update_device, Up: OpenACC Runtime Library Routines 2971 1.7 mrg 2972 1.15 mrg 6.25 'acc_update_self' - Update host memory from mapped device memory. 2973 1.7 mrg ====================================================================== 2974 1.7 mrg 2975 1.7 mrg _Description_ 2976 1.7 mrg This function updates the host copy from the previously mapped 2977 1.7 mrg device memory. The host memory is specified with the host address 2978 1.7 mrg A and a length of LEN bytes. 2979 1.7 mrg 2980 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 2981 1.11 mrg specifies a contiguous array section. The second form A specifies 2982 1.7 mrg a variable or array element and LEN specifies the length in bytes. 2983 1.7 mrg 2984 1.7 mrg _C/C++_: 2985 1.11 mrg _Prototype_: 'acc_update_self(h_void *a, size_t len);' 2986 1.13 mrg _Prototype_: 'acc_update_self_async(h_void *a, size_t len, int 2987 1.13 mrg async);' 2988 1.7 mrg 2989 1.7 mrg _Fortran_: 2990 1.11 mrg _Interface_: 'subroutine acc_update_self(a)' 2991 1.11 mrg 'type, dimension(:[,:]...) :: a' 2992 1.11 mrg _Interface_: 'subroutine acc_update_self(a, len)' 2993 1.11 mrg 'type, dimension(:[,:]...) :: a' 2994 1.11 mrg 'integer len' 2995 1.13 mrg _Interface_: 'subroutine acc_update_self_async(a, async)' 2996 1.13 mrg 'type, dimension(:[,:]...) :: a' 2997 1.13 mrg 'integer(acc_handle_kind) :: async' 2998 1.13 mrg _Interface_: 'subroutine acc_update_self_async(a, len, async)' 2999 1.13 mrg 'type, dimension(:[,:]...) :: a' 3000 1.13 mrg 'integer len' 3001 1.13 mrg 'integer(acc_handle_kind) :: async' 3002 1.7 mrg 3003 1.7 mrg _Reference_: 3004 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3005 1.13 mrg 3.2.25. 3006 1.7 mrg 3007 1.7 mrg 3008 1.7 mrg File: libgomp.info, Node: acc_map_data, Next: acc_unmap_data, Prev: acc_update_self, Up: OpenACC Runtime Library Routines 3009 1.7 mrg 3010 1.15 mrg 6.26 'acc_map_data' - Map previously allocated device memory to host memory. 3011 1.7 mrg ============================================================================ 3012 1.7 mrg 3013 1.7 mrg _Description_ 3014 1.7 mrg This function maps previously allocated device and host memory. 3015 1.11 mrg The device memory is specified with the device address D. The host 3016 1.7 mrg memory is specified with the host address H and a length of LEN. 3017 1.7 mrg 3018 1.7 mrg _C/C++_: 3019 1.11 mrg _Prototype_: 'acc_map_data(h_void *h, d_void *d, size_t len);' 3020 1.7 mrg 3021 1.7 mrg _Reference_: 3022 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3023 1.13 mrg 3.2.26. 3024 1.7 mrg 3025 1.7 mrg 3026 1.7 mrg File: libgomp.info, Node: acc_unmap_data, Next: acc_deviceptr, Prev: acc_map_data, Up: OpenACC Runtime Library Routines 3027 1.7 mrg 3028 1.15 mrg 6.27 'acc_unmap_data' - Unmap device memory from host memory. 3029 1.7 mrg ============================================================= 3030 1.7 mrg 3031 1.7 mrg _Description_ 3032 1.11 mrg This function unmaps previously mapped device and host memory. The 3033 1.7 mrg latter specified by H. 3034 1.7 mrg 3035 1.7 mrg _C/C++_: 3036 1.11 mrg _Prototype_: 'acc_unmap_data(h_void *h);' 3037 1.7 mrg 3038 1.7 mrg _Reference_: 3039 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3040 1.13 mrg 3.2.27. 3041 1.7 mrg 3042 1.7 mrg 3043 1.7 mrg File: libgomp.info, Node: acc_deviceptr, Next: acc_hostptr, Prev: acc_unmap_data, Up: OpenACC Runtime Library Routines 3044 1.7 mrg 3045 1.15 mrg 6.28 'acc_deviceptr' - Get device pointer associated with specific host address. 3046 1.7 mrg ================================================================================ 3047 1.7 mrg 3048 1.7 mrg _Description_ 3049 1.7 mrg This function returns the device address that has been mapped to 3050 1.7 mrg the host address specified by H. 3051 1.7 mrg 3052 1.7 mrg _C/C++_: 3053 1.11 mrg _Prototype_: 'void *acc_deviceptr(h_void *h);' 3054 1.7 mrg 3055 1.7 mrg _Reference_: 3056 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3057 1.13 mrg 3.2.28. 3058 1.7 mrg 3059 1.7 mrg 3060 1.7 mrg File: libgomp.info, Node: acc_hostptr, Next: acc_is_present, Prev: acc_deviceptr, Up: OpenACC Runtime Library Routines 3061 1.7 mrg 3062 1.15 mrg 6.29 'acc_hostptr' - Get host pointer associated with specific device address. 3063 1.7 mrg ============================================================================== 3064 1.7 mrg 3065 1.7 mrg _Description_ 3066 1.7 mrg This function returns the host address that has been mapped to the 3067 1.7 mrg device address specified by D. 3068 1.7 mrg 3069 1.7 mrg _C/C++_: 3070 1.11 mrg _Prototype_: 'void *acc_hostptr(d_void *d);' 3071 1.7 mrg 3072 1.7 mrg _Reference_: 3073 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3074 1.13 mrg 3.2.29. 3075 1.7 mrg 3076 1.7 mrg 3077 1.7 mrg File: libgomp.info, Node: acc_is_present, Next: acc_memcpy_to_device, Prev: acc_hostptr, Up: OpenACC Runtime Library Routines 3078 1.7 mrg 3079 1.15 mrg 6.30 'acc_is_present' - Indicate whether host variable / array is present on device. 3080 1.7 mrg ==================================================================================== 3081 1.7 mrg 3082 1.7 mrg _Description_ 3083 1.11 mrg This function indicates whether the specified host address in A and 3084 1.11 mrg a length of LEN bytes is present on the device. In C/C++, a 3085 1.7 mrg non-zero value is returned to indicate the presence of the mapped 3086 1.11 mrg memory on the device. A zero is returned to indicate the memory is 3087 1.7 mrg not mapped on the device. 3088 1.7 mrg 3089 1.11 mrg In Fortran, two (2) forms are supported. In the first form, A 3090 1.11 mrg specifies a contiguous array section. The second form A specifies 3091 1.7 mrg a variable or array element and LEN specifies the length in bytes. 3092 1.11 mrg If the host memory is mapped to device memory, then a 'true' is 3093 1.11 mrg returned. Otherwise, a 'false' is return to indicate the mapped 3094 1.7 mrg memory is not present. 3095 1.7 mrg 3096 1.7 mrg _C/C++_: 3097 1.11 mrg _Prototype_: 'int acc_is_present(h_void *a, size_t len);' 3098 1.7 mrg 3099 1.7 mrg _Fortran_: 3100 1.11 mrg _Interface_: 'function acc_is_present(a)' 3101 1.11 mrg 'type, dimension(:[,:]...) :: a' 3102 1.11 mrg 'logical acc_is_present' 3103 1.11 mrg _Interface_: 'function acc_is_present(a, len)' 3104 1.11 mrg 'type, dimension(:[,:]...) :: a' 3105 1.11 mrg 'integer len' 3106 1.11 mrg 'logical acc_is_present' 3107 1.7 mrg 3108 1.7 mrg _Reference_: 3109 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3110 1.13 mrg 3.2.30. 3111 1.7 mrg 3112 1.7 mrg 3113 1.7 mrg File: libgomp.info, Node: acc_memcpy_to_device, Next: acc_memcpy_from_device, Prev: acc_is_present, Up: OpenACC Runtime Library Routines 3114 1.7 mrg 3115 1.15 mrg 6.31 'acc_memcpy_to_device' - Copy host memory to device memory. 3116 1.7 mrg ================================================================ 3117 1.7 mrg 3118 1.7 mrg _Description_ 3119 1.7 mrg This function copies host memory specified by host address of SRC 3120 1.7 mrg to device memory specified by the device address DEST for a length 3121 1.7 mrg of BYTES bytes. 3122 1.7 mrg 3123 1.7 mrg _C/C++_: 3124 1.11 mrg _Prototype_: 'acc_memcpy_to_device(d_void *dest, h_void *src, size_t 3125 1.11 mrg bytes);' 3126 1.7 mrg 3127 1.7 mrg _Reference_: 3128 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3129 1.13 mrg 3.2.31. 3130 1.7 mrg 3131 1.7 mrg 3132 1.13 mrg File: libgomp.info, Node: acc_memcpy_from_device, Next: acc_attach, Prev: acc_memcpy_to_device, Up: OpenACC Runtime Library Routines 3133 1.7 mrg 3134 1.15 mrg 6.32 'acc_memcpy_from_device' - Copy device memory to host memory. 3135 1.7 mrg ================================================================== 3136 1.7 mrg 3137 1.7 mrg _Description_ 3138 1.7 mrg This function copies host memory specified by host address of SRC 3139 1.7 mrg from device memory specified by the device address DEST for a 3140 1.7 mrg length of BYTES bytes. 3141 1.7 mrg 3142 1.7 mrg _C/C++_: 3143 1.11 mrg _Prototype_: 'acc_memcpy_from_device(d_void *dest, h_void *src, 3144 1.11 mrg size_t bytes);' 3145 1.7 mrg 3146 1.7 mrg _Reference_: 3147 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3148 1.13 mrg 3.2.32. 3149 1.13 mrg 3150 1.13 mrg 3151 1.13 mrg File: libgomp.info, Node: acc_attach, Next: acc_detach, Prev: acc_memcpy_from_device, Up: OpenACC Runtime Library Routines 3152 1.13 mrg 3153 1.15 mrg 6.33 'acc_attach' - Let device pointer point to device-pointer target. 3154 1.13 mrg ====================================================================== 3155 1.13 mrg 3156 1.13 mrg _Description_ 3157 1.13 mrg This function updates a pointer on the device from pointing to a 3158 1.13 mrg host-pointer address to pointing to the corresponding device data. 3159 1.13 mrg 3160 1.13 mrg _C/C++_: 3161 1.13 mrg _Prototype_: 'acc_attach(h_void **ptr);' 3162 1.13 mrg _Prototype_: 'acc_attach_async(h_void **ptr, int async);' 3163 1.13 mrg 3164 1.13 mrg _Reference_: 3165 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3166 1.13 mrg 3.2.34. 3167 1.13 mrg 3168 1.13 mrg 3169 1.13 mrg File: libgomp.info, Node: acc_detach, Next: acc_get_current_cuda_device, Prev: acc_attach, Up: OpenACC Runtime Library Routines 3170 1.13 mrg 3171 1.15 mrg 6.34 'acc_detach' - Let device pointer point to host-pointer target. 3172 1.13 mrg ==================================================================== 3173 1.13 mrg 3174 1.13 mrg _Description_ 3175 1.13 mrg This function updates a pointer on the device from pointing to a 3176 1.13 mrg device-pointer address to pointing to the corresponding host data. 3177 1.13 mrg 3178 1.13 mrg _C/C++_: 3179 1.13 mrg _Prototype_: 'acc_detach(h_void **ptr);' 3180 1.13 mrg _Prototype_: 'acc_detach_async(h_void **ptr, int async);' 3181 1.13 mrg _Prototype_: 'acc_detach_finalize(h_void **ptr);' 3182 1.13 mrg _Prototype_: 'acc_detach_finalize_async(h_void **ptr, int async);' 3183 1.13 mrg 3184 1.13 mrg _Reference_: 3185 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3186 1.13 mrg 3.2.35. 3187 1.7 mrg 3188 1.7 mrg 3189 1.13 mrg File: libgomp.info, Node: acc_get_current_cuda_device, Next: acc_get_current_cuda_context, Prev: acc_detach, Up: OpenACC Runtime Library Routines 3190 1.7 mrg 3191 1.15 mrg 6.35 'acc_get_current_cuda_device' - Get CUDA device handle. 3192 1.7 mrg ============================================================ 3193 1.7 mrg 3194 1.7 mrg _Description_ 3195 1.11 mrg This function returns the CUDA device handle. This handle is the 3196 1.7 mrg same as used by the CUDA Runtime or Driver API's. 3197 1.7 mrg 3198 1.7 mrg _C/C++_: 3199 1.11 mrg _Prototype_: 'void *acc_get_current_cuda_device(void);' 3200 1.7 mrg 3201 1.7 mrg _Reference_: 3202 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3203 1.7 mrg A.2.1.1. 3204 1.7 mrg 3205 1.7 mrg 3206 1.7 mrg 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 3207 1.7 mrg 3208 1.15 mrg 6.36 'acc_get_current_cuda_context' - Get CUDA context handle. 3209 1.7 mrg ============================================================== 3210 1.7 mrg 3211 1.7 mrg _Description_ 3212 1.11 mrg This function returns the CUDA context handle. This handle is the 3213 1.7 mrg same as used by the CUDA Runtime or Driver API's. 3214 1.7 mrg 3215 1.7 mrg _C/C++_: 3216 1.12 mrg _Prototype_: 'void *acc_get_current_cuda_context(void);' 3217 1.7 mrg 3218 1.7 mrg _Reference_: 3219 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3220 1.7 mrg A.2.1.2. 3221 1.7 mrg 3222 1.7 mrg 3223 1.7 mrg File: libgomp.info, Node: acc_get_cuda_stream, Next: acc_set_cuda_stream, Prev: acc_get_current_cuda_context, Up: OpenACC Runtime Library Routines 3224 1.7 mrg 3225 1.15 mrg 6.37 'acc_get_cuda_stream' - Get CUDA stream handle. 3226 1.7 mrg ==================================================== 3227 1.7 mrg 3228 1.7 mrg _Description_ 3229 1.12 mrg This function returns the CUDA stream handle for the queue ASYNC. 3230 1.12 mrg This handle is the same as used by the CUDA Runtime or Driver 3231 1.12 mrg API's. 3232 1.7 mrg 3233 1.7 mrg _C/C++_: 3234 1.12 mrg _Prototype_: 'void *acc_get_cuda_stream(int async);' 3235 1.7 mrg 3236 1.7 mrg _Reference_: 3237 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3238 1.7 mrg A.2.1.3. 3239 1.7 mrg 3240 1.7 mrg 3241 1.13 mrg File: libgomp.info, Node: acc_set_cuda_stream, Next: acc_prof_register, Prev: acc_get_cuda_stream, Up: OpenACC Runtime Library Routines 3242 1.7 mrg 3243 1.15 mrg 6.38 'acc_set_cuda_stream' - Set CUDA stream handle. 3244 1.7 mrg ==================================================== 3245 1.7 mrg 3246 1.7 mrg _Description_ 3247 1.7 mrg This function associates the stream handle specified by STREAM with 3248 1.12 mrg the queue ASYNC. 3249 1.12 mrg 3250 1.12 mrg This cannot be used to change the stream handle associated with 3251 1.12 mrg 'acc_async_sync'. 3252 1.12 mrg 3253 1.12 mrg The return value is not specified. 3254 1.7 mrg 3255 1.7 mrg _C/C++_: 3256 1.12 mrg _Prototype_: 'int acc_set_cuda_stream(int async, void *stream);' 3257 1.7 mrg 3258 1.7 mrg _Reference_: 3259 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 3260 1.7 mrg A.2.1.4. 3261 1.7 mrg 3262 1.7 mrg 3263 1.13 mrg File: libgomp.info, Node: acc_prof_register, Next: acc_prof_unregister, Prev: acc_set_cuda_stream, Up: OpenACC Runtime Library Routines 3264 1.13 mrg 3265 1.15 mrg 6.39 'acc_prof_register' - Register callbacks. 3266 1.13 mrg ============================================== 3267 1.13 mrg 3268 1.13 mrg _Description_: 3269 1.13 mrg This function registers callbacks. 3270 1.13 mrg 3271 1.13 mrg _C/C++_: 3272 1.13 mrg _Prototype_: 'void acc_prof_register (acc_event_t, acc_prof_callback, 3273 1.13 mrg acc_register_t);' 3274 1.13 mrg 3275 1.13 mrg _See also_: 3276 1.13 mrg *note OpenACC Profiling Interface:: 3277 1.13 mrg 3278 1.13 mrg _Reference_: 3279 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 3280 1.13 mrg 3281 1.13 mrg 3282 1.13 mrg File: libgomp.info, Node: acc_prof_unregister, Next: acc_prof_lookup, Prev: acc_prof_register, Up: OpenACC Runtime Library Routines 3283 1.13 mrg 3284 1.15 mrg 6.40 'acc_prof_unregister' - Unregister callbacks. 3285 1.13 mrg ================================================== 3286 1.13 mrg 3287 1.13 mrg _Description_: 3288 1.13 mrg This function unregisters callbacks. 3289 1.13 mrg 3290 1.13 mrg _C/C++_: 3291 1.13 mrg _Prototype_: 'void acc_prof_unregister (acc_event_t, 3292 1.13 mrg acc_prof_callback, acc_register_t);' 3293 1.13 mrg 3294 1.13 mrg _See also_: 3295 1.13 mrg *note OpenACC Profiling Interface:: 3296 1.13 mrg 3297 1.13 mrg _Reference_: 3298 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 3299 1.13 mrg 3300 1.13 mrg 3301 1.13 mrg File: libgomp.info, Node: acc_prof_lookup, Next: acc_register_library, Prev: acc_prof_unregister, Up: OpenACC Runtime Library Routines 3302 1.13 mrg 3303 1.15 mrg 6.41 'acc_prof_lookup' - Obtain inquiry functions. 3304 1.13 mrg ================================================== 3305 1.13 mrg 3306 1.13 mrg _Description_: 3307 1.13 mrg Function to obtain inquiry functions. 3308 1.13 mrg 3309 1.13 mrg _C/C++_: 3310 1.13 mrg _Prototype_: 'acc_query_fn acc_prof_lookup (const char *);' 3311 1.13 mrg 3312 1.13 mrg _See also_: 3313 1.13 mrg *note OpenACC Profiling Interface:: 3314 1.13 mrg 3315 1.13 mrg _Reference_: 3316 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 3317 1.13 mrg 3318 1.13 mrg 3319 1.13 mrg File: libgomp.info, Node: acc_register_library, Prev: acc_prof_lookup, Up: OpenACC Runtime Library Routines 3320 1.13 mrg 3321 1.15 mrg 6.42 'acc_register_library' - Library registration. 3322 1.13 mrg =================================================== 3323 1.13 mrg 3324 1.13 mrg _Description_: 3325 1.13 mrg Function for library registration. 3326 1.13 mrg 3327 1.13 mrg _C/C++_: 3328 1.13 mrg _Prototype_: 'void acc_register_library (acc_prof_reg, acc_prof_reg, 3329 1.13 mrg acc_prof_lookup_func);' 3330 1.13 mrg 3331 1.13 mrg _See also_: 3332 1.13 mrg *note OpenACC Profiling Interface::, *note ACC_PROFLIB:: 3333 1.13 mrg 3334 1.13 mrg _Reference_: 3335 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 5.3. 3336 1.13 mrg 3337 1.13 mrg 3338 1.7 mrg File: libgomp.info, Node: OpenACC Environment Variables, Next: CUDA Streams Usage, Prev: OpenACC Runtime Library Routines, Up: Top 3339 1.7 mrg 3340 1.15 mrg 7 OpenACC Environment Variables 3341 1.7 mrg ******************************* 3342 1.7 mrg 3343 1.11 mrg The variables 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM' are defined by 3344 1.7 mrg section 4 of the OpenACC specification in version 2.0. The variable 3345 1.13 mrg 'ACC_PROFLIB' is defined by section 4 of the OpenACC specification in 3346 1.13 mrg version 2.6. The variable 'GCC_ACC_NOTIFY' is used for diagnostic 3347 1.13 mrg purposes. 3348 1.7 mrg 3349 1.7 mrg * Menu: 3350 1.7 mrg 3351 1.7 mrg * ACC_DEVICE_TYPE:: 3352 1.7 mrg * ACC_DEVICE_NUM:: 3353 1.13 mrg * ACC_PROFLIB:: 3354 1.7 mrg * GCC_ACC_NOTIFY:: 3355 1.7 mrg 3356 1.7 mrg 3357 1.7 mrg File: libgomp.info, Node: ACC_DEVICE_TYPE, Next: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 3358 1.7 mrg 3359 1.15 mrg 7.1 'ACC_DEVICE_TYPE' 3360 1.7 mrg ===================== 3361 1.7 mrg 3362 1.7 mrg _Reference_: 3363 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 4.1. 3364 1.7 mrg 3365 1.7 mrg 3366 1.13 mrg File: libgomp.info, Node: ACC_DEVICE_NUM, Next: ACC_PROFLIB, Prev: ACC_DEVICE_TYPE, Up: OpenACC Environment Variables 3367 1.7 mrg 3368 1.15 mrg 7.2 'ACC_DEVICE_NUM' 3369 1.7 mrg ==================== 3370 1.7 mrg 3371 1.7 mrg _Reference_: 3372 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 4.2. 3373 1.13 mrg 3374 1.13 mrg 3375 1.13 mrg File: libgomp.info, Node: ACC_PROFLIB, Next: GCC_ACC_NOTIFY, Prev: ACC_DEVICE_NUM, Up: OpenACC Environment Variables 3376 1.13 mrg 3377 1.15 mrg 7.3 'ACC_PROFLIB' 3378 1.13 mrg ================= 3379 1.13 mrg 3380 1.13 mrg _See also_: 3381 1.13 mrg *note acc_register_library::, *note OpenACC Profiling Interface:: 3382 1.13 mrg 3383 1.13 mrg _Reference_: 3384 1.13 mrg OpenACC specification v2.6 (https://www.openacc.org), section 4.3. 3385 1.7 mrg 3386 1.7 mrg 3387 1.13 mrg File: libgomp.info, Node: GCC_ACC_NOTIFY, Prev: ACC_PROFLIB, Up: OpenACC Environment Variables 3388 1.7 mrg 3389 1.15 mrg 7.4 'GCC_ACC_NOTIFY' 3390 1.7 mrg ==================== 3391 1.7 mrg 3392 1.7 mrg _Description_: 3393 1.7 mrg Print debug information pertaining to the accelerator. 3394 1.7 mrg 3395 1.7 mrg 3396 1.7 mrg File: libgomp.info, Node: CUDA Streams Usage, Next: OpenACC Library Interoperability, Prev: OpenACC Environment Variables, Up: Top 3397 1.7 mrg 3398 1.15 mrg 8 CUDA Streams Usage 3399 1.7 mrg ******************** 3400 1.7 mrg 3401 1.11 mrg This applies to the 'nvptx' plugin only. 3402 1.7 mrg 3403 1.7 mrg The library provides elements that perform asynchronous movement of 3404 1.7 mrg data and asynchronous operation of computing constructs. This 3405 1.7 mrg asynchronous functionality is implemented by making use of CUDA 3406 1.7 mrg streams(1). 3407 1.7 mrg 3408 1.13 mrg The primary means by that the asynchronous functionality is accessed 3409 1.7 mrg is through the use of those OpenACC directives which make use of the 3410 1.11 mrg 'async' and 'wait' clauses. When the 'async' clause is first used with 3411 1.11 mrg a directive, it creates a CUDA stream. If an 'async-argument' is used 3412 1.11 mrg with the 'async' clause, then the stream is associated with the 3413 1.11 mrg specified 'async-argument'. 3414 1.7 mrg 3415 1.7 mrg Following the creation of an association between a CUDA stream and 3416 1.11 mrg the 'async-argument' of an 'async' clause, both the 'wait' clause and 3417 1.11 mrg the 'wait' directive can be used. When either the clause or directive 3418 1.7 mrg is used after stream creation, it creates a rendezvous point whereby 3419 1.7 mrg execution waits until all operations associated with the 3420 1.11 mrg 'async-argument', that is, stream, have completed. 3421 1.7 mrg 3422 1.7 mrg Normally, the management of the streams that are created as a result 3423 1.11 mrg of using the 'async' clause, is done without any intervention by the 3424 1.11 mrg caller. This implies the association between the 'async-argument' and 3425 1.7 mrg the CUDA stream will be maintained for the lifetime of the program. 3426 1.7 mrg However, this association can be changed through the use of the library 3427 1.11 mrg function 'acc_set_cuda_stream'. When the function 'acc_set_cuda_stream' 3428 1.11 mrg is called, the CUDA stream that was originally associated with the 3429 1.11 mrg 'async' clause will be destroyed. Caution should be taken when changing 3430 1.11 mrg the association as subsequent references to the 'async-argument' refer 3431 1.11 mrg to a different CUDA stream. 3432 1.7 mrg 3433 1.7 mrg ---------- Footnotes ---------- 3434 1.7 mrg 3435 1.7 mrg (1) See "Stream Management" in "CUDA Driver API", TRM-06703-001, 3436 1.7 mrg Version 5.5, for additional information 3437 1.7 mrg 3438 1.7 mrg 3439 1.13 mrg File: libgomp.info, Node: OpenACC Library Interoperability, Next: OpenACC Profiling Interface, Prev: CUDA Streams Usage, Up: Top 3440 1.7 mrg 3441 1.15 mrg 9 OpenACC Library Interoperability 3442 1.7 mrg ********************************** 3443 1.7 mrg 3444 1.15 mrg 9.1 Introduction 3445 1.7 mrg ================ 3446 1.7 mrg 3447 1.7 mrg The OpenACC library uses the CUDA Driver API, and may interact with 3448 1.11 mrg programs that use the Runtime library directly, or another library based 3449 1.11 mrg on the Runtime library, e.g., CUBLAS(1). This chapter describes the use 3450 1.11 mrg cases and what changes are required in order to use both the OpenACC 3451 1.11 mrg library and the CUBLAS and Runtime libraries within a program. 3452 1.7 mrg 3453 1.15 mrg 9.2 First invocation: NVIDIA CUBLAS library API 3454 1.7 mrg =============================================== 3455 1.7 mrg 3456 1.7 mrg In this first use case (see below), a function in the CUBLAS library is 3457 1.11 mrg called prior to any of the functions in the OpenACC library. More 3458 1.11 mrg specifically, the function 'cublasCreate()'. 3459 1.7 mrg 3460 1.7 mrg When invoked, the function initializes the library and allocates the 3461 1.7 mrg hardware resources on the host and the device on behalf of the caller. 3462 1.7 mrg Once the initialization and allocation has completed, a handle is 3463 1.11 mrg returned to the caller. The OpenACC library also requires 3464 1.11 mrg initialization and allocation of hardware resources. Since the CUBLAS 3465 1.11 mrg library has already allocated the hardware resources for the device, all 3466 1.11 mrg that is left to do is to initialize the OpenACC library and acquire the 3467 1.11 mrg hardware resources on the host. 3468 1.7 mrg 3469 1.7 mrg Prior to calling the OpenACC function that initializes the library 3470 1.11 mrg and allocate the host hardware resources, you need to acquire the device 3471 1.11 mrg number that was allocated during the call to 'cublasCreate()'. The 3472 1.11 mrg invoking of the runtime library function 'cudaGetDevice()' accomplishes 3473 1.11 mrg this. Once acquired, the device number is passed along with the device 3474 1.11 mrg type as parameters to the OpenACC library function 3475 1.11 mrg 'acc_set_device_num()'. 3476 1.11 mrg 3477 1.11 mrg Once the call to 'acc_set_device_num()' has completed, the OpenACC 3478 1.11 mrg library uses the context that was created during the call to 3479 1.11 mrg 'cublasCreate()'. In other words, both libraries will be sharing the 3480 1.7 mrg same context. 3481 1.7 mrg 3482 1.7 mrg /* Create the handle */ 3483 1.7 mrg s = cublasCreate(&h); 3484 1.7 mrg if (s != CUBLAS_STATUS_SUCCESS) 3485 1.7 mrg { 3486 1.7 mrg fprintf(stderr, "cublasCreate failed %d\n", s); 3487 1.7 mrg exit(EXIT_FAILURE); 3488 1.7 mrg } 3489 1.7 mrg 3490 1.7 mrg /* Get the device number */ 3491 1.7 mrg e = cudaGetDevice(&dev); 3492 1.7 mrg if (e != cudaSuccess) 3493 1.7 mrg { 3494 1.7 mrg fprintf(stderr, "cudaGetDevice failed %d\n", e); 3495 1.7 mrg exit(EXIT_FAILURE); 3496 1.7 mrg } 3497 1.7 mrg 3498 1.7 mrg /* Initialize OpenACC library and use device 'dev' */ 3499 1.7 mrg acc_set_device_num(dev, acc_device_nvidia); 3500 1.11 mrg 3501 1.11 mrg Use Case 1 3502 1.7 mrg 3503 1.15 mrg 9.3 First invocation: OpenACC library API 3504 1.7 mrg ========================================= 3505 1.7 mrg 3506 1.7 mrg In this second use case (see below), a function in the OpenACC library 3507 1.11 mrg is called prior to any of the functions in the CUBLAS library. More 3508 1.11 mrg specificially, the function 'acc_set_device_num()'. 3509 1.7 mrg 3510 1.11 mrg In the use case presented here, the function 'acc_set_device_num()' 3511 1.7 mrg is used to both initialize the OpenACC library and allocate the hardware 3512 1.11 mrg resources on the host and the device. In the call to the function, the 3513 1.11 mrg call parameters specify which device to use and what device type to use, 3514 1.11 mrg i.e., 'acc_device_nvidia'. It should be noted that this is but one 3515 1.7 mrg method to initialize the OpenACC library and allocate the appropriate 3516 1.11 mrg hardware resources. Other methods are available through the use of 3517 1.7 mrg environment variables and these will be discussed in the next section. 3518 1.7 mrg 3519 1.11 mrg Once the call to 'acc_set_device_num()' has completed, other OpenACC 3520 1.7 mrg functions can be called as seen with multiple calls being made to 3521 1.11 mrg 'acc_copyin()'. In addition, calls can be made to functions in the 3522 1.11 mrg CUBLAS library. In the use case a call to 'cublasCreate()' is made 3523 1.11 mrg subsequent to the calls to 'acc_copyin()'. As seen in the previous use 3524 1.11 mrg case, a call to 'cublasCreate()' initializes the CUBLAS library and 3525 1.7 mrg allocates the hardware resources on the host and the device. However, 3526 1.11 mrg since the device has already been allocated, 'cublasCreate()' will only 3527 1.7 mrg initialize the CUBLAS library and allocate the appropriate hardware 3528 1.11 mrg resources on the host. The context that was created as part of the 3529 1.7 mrg OpenACC initialization is shared with the CUBLAS library, similarly to 3530 1.7 mrg the first use case. 3531 1.7 mrg 3532 1.7 mrg dev = 0; 3533 1.7 mrg 3534 1.7 mrg acc_set_device_num(dev, acc_device_nvidia); 3535 1.7 mrg 3536 1.7 mrg /* Copy the first set to the device */ 3537 1.7 mrg d_X = acc_copyin(&h_X[0], N * sizeof (float)); 3538 1.7 mrg if (d_X == NULL) 3539 1.7 mrg { 3540 1.7 mrg fprintf(stderr, "copyin error h_X\n"); 3541 1.7 mrg exit(EXIT_FAILURE); 3542 1.7 mrg } 3543 1.7 mrg 3544 1.7 mrg /* Copy the second set to the device */ 3545 1.7 mrg d_Y = acc_copyin(&h_Y1[0], N * sizeof (float)); 3546 1.7 mrg if (d_Y == NULL) 3547 1.7 mrg { 3548 1.7 mrg fprintf(stderr, "copyin error h_Y1\n"); 3549 1.7 mrg exit(EXIT_FAILURE); 3550 1.7 mrg } 3551 1.7 mrg 3552 1.7 mrg /* Create the handle */ 3553 1.7 mrg s = cublasCreate(&h); 3554 1.7 mrg if (s != CUBLAS_STATUS_SUCCESS) 3555 1.7 mrg { 3556 1.7 mrg fprintf(stderr, "cublasCreate failed %d\n", s); 3557 1.7 mrg exit(EXIT_FAILURE); 3558 1.7 mrg } 3559 1.7 mrg 3560 1.7 mrg /* Perform saxpy using CUBLAS library function */ 3561 1.7 mrg s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1); 3562 1.7 mrg if (s != CUBLAS_STATUS_SUCCESS) 3563 1.7 mrg { 3564 1.7 mrg fprintf(stderr, "cublasSaxpy failed %d\n", s); 3565 1.7 mrg exit(EXIT_FAILURE); 3566 1.7 mrg } 3567 1.7 mrg 3568 1.7 mrg /* Copy the results from the device */ 3569 1.7 mrg acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float)); 3570 1.11 mrg 3571 1.11 mrg Use Case 2 3572 1.7 mrg 3573 1.15 mrg 9.4 OpenACC library and environment variables 3574 1.7 mrg ============================================= 3575 1.7 mrg 3576 1.7 mrg There are two environment variables associated with the OpenACC library 3577 1.7 mrg that may be used to control the device type and device number: 3578 1.13 mrg 'ACC_DEVICE_TYPE' and 'ACC_DEVICE_NUM', respectively. These two 3579 1.13 mrg environment variables can be used as an alternative to calling 3580 1.11 mrg 'acc_set_device_num()'. As seen in the second use case, the device type 3581 1.11 mrg and device number were specified using 'acc_set_device_num()'. If 3582 1.7 mrg however, the aforementioned environment variables were set, then the 3583 1.11 mrg call to 'acc_set_device_num()' would not be required. 3584 1.7 mrg 3585 1.11 mrg The use of the environment variables is only relevant when an OpenACC 3586 1.11 mrg function is called prior to a call to 'cudaCreate()'. If 'cudaCreate()' 3587 1.11 mrg is called prior to a call to an OpenACC function, then you must call 3588 1.11 mrg 'acc_set_device_num()'(2) 3589 1.7 mrg 3590 1.7 mrg ---------- Footnotes ---------- 3591 1.7 mrg 3592 1.7 mrg (1) See section 2.26, "Interactions with the CUDA Driver API" in 3593 1.7 mrg "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU 3594 1.11 mrg Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5, for 3595 1.11 mrg additional information on library interoperability. 3596 1.7 mrg 3597 1.11 mrg (2) More complete information about 'ACC_DEVICE_TYPE' and 3598 1.11 mrg 'ACC_DEVICE_NUM' can be found in sections 4.1 and 4.2 of the OpenACC 3599 1.11 mrg (https://www.openacc.org) Application Programming Interface, Version 3600 1.13 mrg 2.6. 3601 1.7 mrg 3602 1.7 mrg 3603 1.13 mrg File: libgomp.info, Node: OpenACC Profiling Interface, Next: The libgomp ABI, Prev: OpenACC Library Interoperability, Up: Top 3604 1.7 mrg 3605 1.15 mrg 10 OpenACC Profiling Interface 3606 1.15 mrg ****************************** 3607 1.13 mrg 3608 1.15 mrg 10.1 Implementation Status and Implementation-Defined Behavior 3609 1.15 mrg ============================================================== 3610 1.13 mrg 3611 1.13 mrg We're implementing the OpenACC Profiling Interface as defined by the 3612 1.13 mrg OpenACC 2.6 specification. We're clarifying some aspects here as 3613 1.13 mrg _implementation-defined behavior_, while they're still under discussion 3614 1.13 mrg within the OpenACC Technical Committee. 3615 1.13 mrg 3616 1.13 mrg This implementation is tuned to keep the performance impact as low as 3617 1.13 mrg possible for the (very common) case that the Profiling Interface is not 3618 1.13 mrg enabled. This is relevant, as the Profiling Interface affects all the 3619 1.13 mrg _hot_ code paths (in the target code, not in the offloaded code). Users 3620 1.13 mrg of the OpenACC Profiling Interface can be expected to understand that 3621 1.13 mrg performance will be impacted to some degree once the Profiling Interface 3622 1.13 mrg has gotten enabled: for example, because of the _runtime_ (libgomp) 3623 1.13 mrg calling into a third-party _library_ for every event that has been 3624 1.13 mrg registered. 3625 1.13 mrg 3626 1.13 mrg We're not yet accounting for the fact that 'OpenACC events may occur 3627 1.13 mrg during event processing'. We just handle one case specially, as 3628 1.13 mrg required by CUDA 9.0 'nvprof', that 'acc_get_device_type' (*note 3629 1.13 mrg acc_get_device_type::)) may be called from 'acc_ev_device_init_start', 3630 1.13 mrg 'acc_ev_device_init_end' callbacks. 3631 1.13 mrg 3632 1.13 mrg We're not yet implementing initialization via a 3633 1.13 mrg 'acc_register_library' function that is either statically linked in, or 3634 1.13 mrg dynamically via 'LD_PRELOAD'. Initialization via 'acc_register_library' 3635 1.13 mrg functions dynamically loaded via the 'ACC_PROFLIB' environment variable 3636 1.13 mrg does work, as does directly calling 'acc_prof_register', 3637 1.13 mrg 'acc_prof_unregister', 'acc_prof_lookup'. 3638 1.13 mrg 3639 1.13 mrg As currently there are no inquiry functions defined, calls to 3640 1.13 mrg 'acc_prof_lookup' will always return 'NULL'. 3641 1.13 mrg 3642 1.13 mrg There aren't separate _start_, _stop_ events defined for the event 3643 1.13 mrg types 'acc_ev_create', 'acc_ev_delete', 'acc_ev_alloc', 'acc_ev_free'. 3644 1.13 mrg It's not clear if these should be triggered before or after the actual 3645 1.13 mrg device-specific call is made. We trigger them after. 3646 1.13 mrg 3647 1.13 mrg Remarks about data provided to callbacks: 3648 1.13 mrg 3649 1.13 mrg 'acc_prof_info.event_type' 3650 1.13 mrg It's not clear if for _nested_ event callbacks (for example, 3651 1.13 mrg 'acc_ev_enqueue_launch_start' as part of a parent compute 3652 1.13 mrg construct), this should be set for the nested event 3653 1.13 mrg ('acc_ev_enqueue_launch_start'), or if the value of the parent 3654 1.13 mrg construct should remain ('acc_ev_compute_construct_start'). In 3655 1.13 mrg this implementation, the value will generally correspond to the 3656 1.13 mrg innermost nested event type. 3657 1.13 mrg 3658 1.13 mrg 'acc_prof_info.device_type' 3659 1.13 mrg 3660 1.13 mrg * For 'acc_ev_compute_construct_start', and in presence of an 3661 1.13 mrg 'if' clause with _false_ argument, this will still refer to 3662 1.13 mrg the offloading device type. It's not clear if that's the 3663 1.13 mrg expected behavior. 3664 1.13 mrg 3665 1.13 mrg * Complementary to the item before, for 3666 1.13 mrg 'acc_ev_compute_construct_end', this is set to 3667 1.13 mrg 'acc_device_host' in presence of an 'if' clause with _false_ 3668 1.13 mrg argument. It's not clear if that's the expected behavior. 3669 1.13 mrg 3670 1.13 mrg 'acc_prof_info.thread_id' 3671 1.13 mrg Always '-1'; not yet implemented. 3672 1.13 mrg 3673 1.13 mrg 'acc_prof_info.async' 3674 1.13 mrg 3675 1.13 mrg * Not yet implemented correctly for 3676 1.13 mrg 'acc_ev_compute_construct_start'. 3677 1.13 mrg 3678 1.13 mrg * In a compute construct, for host-fallback 3679 1.13 mrg execution/'acc_device_host' it will always be 3680 1.13 mrg 'acc_async_sync'. It's not clear if that's the expected 3681 1.13 mrg behavior. 3682 1.13 mrg 3683 1.13 mrg * For 'acc_ev_device_init_start' and 'acc_ev_device_init_end', 3684 1.13 mrg it will always be 'acc_async_sync'. It's not clear if that's 3685 1.13 mrg the expected behavior. 3686 1.13 mrg 3687 1.13 mrg 'acc_prof_info.async_queue' 3688 1.13 mrg There is no 'limited number of asynchronous queues' in libgomp. 3689 1.13 mrg This will always have the same value as 'acc_prof_info.async'. 3690 1.13 mrg 3691 1.13 mrg 'acc_prof_info.src_file' 3692 1.13 mrg Always 'NULL'; not yet implemented. 3693 1.13 mrg 3694 1.13 mrg 'acc_prof_info.func_name' 3695 1.13 mrg Always 'NULL'; not yet implemented. 3696 1.13 mrg 3697 1.13 mrg 'acc_prof_info.line_no' 3698 1.13 mrg Always '-1'; not yet implemented. 3699 1.13 mrg 3700 1.13 mrg 'acc_prof_info.end_line_no' 3701 1.13 mrg Always '-1'; not yet implemented. 3702 1.13 mrg 3703 1.13 mrg 'acc_prof_info.func_line_no' 3704 1.13 mrg Always '-1'; not yet implemented. 3705 1.13 mrg 3706 1.13 mrg 'acc_prof_info.func_end_line_no' 3707 1.13 mrg Always '-1'; not yet implemented. 3708 1.13 mrg 3709 1.13 mrg 'acc_event_info.event_type', 'acc_event_info.*.event_type' 3710 1.13 mrg Relating to 'acc_prof_info.event_type' discussed above, in this 3711 1.13 mrg implementation, this will always be the same value as 3712 1.13 mrg 'acc_prof_info.event_type'. 3713 1.13 mrg 3714 1.13 mrg 'acc_event_info.*.parent_construct' 3715 1.13 mrg 3716 1.13 mrg * Will be 'acc_construct_parallel' for all OpenACC compute 3717 1.13 mrg constructs as well as many OpenACC Runtime API calls; should 3718 1.13 mrg be the one matching the actual construct, or 3719 1.13 mrg 'acc_construct_runtime_api', respectively. 3720 1.13 mrg 3721 1.13 mrg * Will be 'acc_construct_enter_data' or 3722 1.13 mrg 'acc_construct_exit_data' when processing variable mappings 3723 1.13 mrg specified in OpenACC _declare_ directives; should be 3724 1.13 mrg 'acc_construct_declare'. 3725 1.13 mrg 3726 1.13 mrg * For implicit 'acc_ev_device_init_start', 3727 1.13 mrg 'acc_ev_device_init_end', and explicit as well as implicit 3728 1.13 mrg 'acc_ev_alloc', 'acc_ev_free', 'acc_ev_enqueue_upload_start', 3729 1.13 mrg 'acc_ev_enqueue_upload_end', 'acc_ev_enqueue_download_start', 3730 1.13 mrg and 'acc_ev_enqueue_download_end', will be 3731 1.13 mrg 'acc_construct_parallel'; should reflect the real parent 3732 1.13 mrg construct. 3733 1.13 mrg 3734 1.13 mrg 'acc_event_info.*.implicit' 3735 1.13 mrg For 'acc_ev_alloc', 'acc_ev_free', 'acc_ev_enqueue_upload_start', 3736 1.13 mrg 'acc_ev_enqueue_upload_end', 'acc_ev_enqueue_download_start', and 3737 1.13 mrg 'acc_ev_enqueue_download_end', this currently will be '1' also for 3738 1.13 mrg explicit usage. 3739 1.13 mrg 3740 1.13 mrg 'acc_event_info.data_event.var_name' 3741 1.13 mrg Always 'NULL'; not yet implemented. 3742 1.13 mrg 3743 1.13 mrg 'acc_event_info.data_event.host_ptr' 3744 1.13 mrg For 'acc_ev_alloc', and 'acc_ev_free', this is always 'NULL'. 3745 1.13 mrg 3746 1.13 mrg 'typedef union acc_api_info' 3747 1.13 mrg ... as printed in '5.2.3. Third Argument: API-Specific 3748 1.13 mrg Information'. This should obviously be 'typedef _struct_ 3749 1.13 mrg acc_api_info'. 3750 1.13 mrg 3751 1.13 mrg 'acc_api_info.device_api' 3752 1.13 mrg Possibly not yet implemented correctly for 3753 1.13 mrg 'acc_ev_compute_construct_start', 'acc_ev_device_init_start', 3754 1.13 mrg 'acc_ev_device_init_end': will always be 'acc_device_api_none' for 3755 1.13 mrg these event types. For 'acc_ev_enter_data_start', it will be 3756 1.13 mrg 'acc_device_api_none' in some cases. 3757 1.13 mrg 3758 1.13 mrg 'acc_api_info.device_type' 3759 1.13 mrg Always the same as 'acc_prof_info.device_type'. 3760 1.13 mrg 3761 1.13 mrg 'acc_api_info.vendor' 3762 1.13 mrg Always '-1'; not yet implemented. 3763 1.13 mrg 3764 1.13 mrg 'acc_api_info.device_handle' 3765 1.13 mrg Always 'NULL'; not yet implemented. 3766 1.13 mrg 3767 1.13 mrg 'acc_api_info.context_handle' 3768 1.13 mrg Always 'NULL'; not yet implemented. 3769 1.13 mrg 3770 1.13 mrg 'acc_api_info.async_handle' 3771 1.13 mrg Always 'NULL'; not yet implemented. 3772 1.13 mrg 3773 1.13 mrg Remarks about certain event types: 3774 1.13 mrg 3775 1.13 mrg 'acc_ev_device_init_start', 'acc_ev_device_init_end' 3776 1.13 mrg 3777 1.15 mrg * When a compute construct triggers implicit 3778 1.13 mrg 'acc_ev_device_init_start' and 'acc_ev_device_init_end' 3779 1.13 mrg events, they currently aren't _nested within_ the 3780 1.13 mrg corresponding 'acc_ev_compute_construct_start' and 3781 1.13 mrg 'acc_ev_compute_construct_end', but they're currently observed 3782 1.13 mrg _before_ 'acc_ev_compute_construct_start'. It's not clear 3783 1.13 mrg what to do: the standard asks us provide a lot of details to 3784 1.13 mrg the 'acc_ev_compute_construct_start' callback, without 3785 1.13 mrg (implicitly) initializing a device before? 3786 1.13 mrg 3787 1.13 mrg * Callbacks for these event types will not be invoked for calls 3788 1.13 mrg to the 'acc_set_device_type' and 'acc_set_device_num' 3789 1.13 mrg functions. It's not clear if they should be. 3790 1.13 mrg 3791 1.13 mrg 'acc_ev_enter_data_start', 'acc_ev_enter_data_end', 'acc_ev_exit_data_start', 'acc_ev_exit_data_end' 3792 1.13 mrg 3793 1.13 mrg * Callbacks for these event types will also be invoked for 3794 1.13 mrg OpenACC _host_data_ constructs. It's not clear if they should 3795 1.13 mrg be. 3796 1.13 mrg 3797 1.13 mrg * Callbacks for these event types will also be invoked when 3798 1.13 mrg processing variable mappings specified in OpenACC _declare_ 3799 1.13 mrg directives. It's not clear if they should be. 3800 1.13 mrg 3801 1.13 mrg Callbacks for the following event types will be invoked, but dispatch 3802 1.13 mrg and information provided therein has not yet been thoroughly reviewed: 3803 1.13 mrg 3804 1.13 mrg * 'acc_ev_alloc' 3805 1.13 mrg * 'acc_ev_free' 3806 1.13 mrg * 'acc_ev_update_start', 'acc_ev_update_end' 3807 1.13 mrg * 'acc_ev_enqueue_upload_start', 'acc_ev_enqueue_upload_end' 3808 1.13 mrg * 'acc_ev_enqueue_download_start', 'acc_ev_enqueue_download_end' 3809 1.13 mrg 3810 1.13 mrg During device initialization, and finalization, respectively, 3811 1.13 mrg callbacks for the following event types will not yet be invoked: 3812 1.13 mrg 3813 1.13 mrg * 'acc_ev_alloc' 3814 1.13 mrg * 'acc_ev_free' 3815 1.13 mrg 3816 1.13 mrg Callbacks for the following event types have not yet been 3817 1.13 mrg implemented, so currently won't be invoked: 3818 1.13 mrg 3819 1.13 mrg * 'acc_ev_device_shutdown_start', 'acc_ev_device_shutdown_end' 3820 1.13 mrg * 'acc_ev_runtime_shutdown' 3821 1.13 mrg * 'acc_ev_create', 'acc_ev_delete' 3822 1.13 mrg * 'acc_ev_wait_start', 'acc_ev_wait_end' 3823 1.13 mrg 3824 1.13 mrg For the following runtime library functions, not all expected 3825 1.13 mrg callbacks will be invoked (mostly concerning implicit device 3826 1.13 mrg initialization): 3827 1.13 mrg 3828 1.13 mrg * 'acc_get_num_devices' 3829 1.13 mrg * 'acc_set_device_type' 3830 1.13 mrg * 'acc_get_device_type' 3831 1.13 mrg * 'acc_set_device_num' 3832 1.13 mrg * 'acc_get_device_num' 3833 1.13 mrg * 'acc_init' 3834 1.13 mrg * 'acc_shutdown' 3835 1.13 mrg 3836 1.13 mrg Aside from implicit device initialization, for the following runtime 3837 1.13 mrg library functions, no callbacks will be invoked for shared-memory 3838 1.13 mrg offloading devices (it's not clear if they should be): 3839 1.13 mrg 3840 1.13 mrg * 'acc_malloc' 3841 1.13 mrg * 'acc_free' 3842 1.13 mrg * 'acc_copyin', 'acc_present_or_copyin', 'acc_copyin_async' 3843 1.13 mrg * 'acc_create', 'acc_present_or_create', 'acc_create_async' 3844 1.13 mrg * 'acc_copyout', 'acc_copyout_async', 'acc_copyout_finalize', 3845 1.13 mrg 'acc_copyout_finalize_async' 3846 1.13 mrg * 'acc_delete', 'acc_delete_async', 'acc_delete_finalize', 3847 1.13 mrg 'acc_delete_finalize_async' 3848 1.13 mrg * 'acc_update_device', 'acc_update_device_async' 3849 1.13 mrg * 'acc_update_self', 'acc_update_self_async' 3850 1.13 mrg * 'acc_map_data', 'acc_unmap_data' 3851 1.13 mrg * 'acc_memcpy_to_device', 'acc_memcpy_to_device_async' 3852 1.13 mrg * 'acc_memcpy_from_device', 'acc_memcpy_from_device_async' 3853 1.13 mrg 3854 1.13 mrg 3855 1.13 mrg File: libgomp.info, Node: The libgomp ABI, Next: Reporting Bugs, Prev: OpenACC Profiling Interface, Up: Top 3856 1.13 mrg 3857 1.15 mrg 11 The libgomp ABI 3858 1.13 mrg ****************** 3859 1.1 mrg 3860 1.11 mrg The following sections present notes on the external ABI as presented by 3861 1.11 mrg libgomp. Only maintainers should need them. 3862 1.1 mrg 3863 1.1 mrg * Menu: 3864 1.1 mrg 3865 1.1 mrg * Implementing MASTER construct:: 3866 1.1 mrg * Implementing CRITICAL construct:: 3867 1.1 mrg * Implementing ATOMIC construct:: 3868 1.1 mrg * Implementing FLUSH construct:: 3869 1.1 mrg * Implementing BARRIER construct:: 3870 1.1 mrg * Implementing THREADPRIVATE construct:: 3871 1.1 mrg * Implementing PRIVATE clause:: 3872 1.1 mrg * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses:: 3873 1.1 mrg * Implementing REDUCTION clause:: 3874 1.1 mrg * Implementing PARALLEL construct:: 3875 1.1 mrg * Implementing FOR construct:: 3876 1.1 mrg * Implementing ORDERED construct:: 3877 1.1 mrg * Implementing SECTIONS construct:: 3878 1.1 mrg * Implementing SINGLE construct:: 3879 1.7 mrg * Implementing OpenACC's PARALLEL construct:: 3880 1.1 mrg 3881 1.1 mrg 3882 1.1 mrg File: libgomp.info, Node: Implementing MASTER construct, Next: Implementing CRITICAL construct, Up: The libgomp ABI 3883 1.1 mrg 3884 1.15 mrg 11.1 Implementing MASTER construct 3885 1.13 mrg ================================== 3886 1.1 mrg 3887 1.1 mrg if (omp_get_thread_num () == 0) 3888 1.1 mrg block 3889 1.1 mrg 3890 1.1 mrg Alternately, we generate two copies of the parallel subfunction and 3891 1.15 mrg only include this in the version run by the primary thread. Surely this 3892 1.3 mrg is not worthwhile though... 3893 1.1 mrg 3894 1.1 mrg 3895 1.1 mrg File: libgomp.info, Node: Implementing CRITICAL construct, Next: Implementing ATOMIC construct, Prev: Implementing MASTER construct, Up: The libgomp ABI 3896 1.1 mrg 3897 1.15 mrg 11.2 Implementing CRITICAL construct 3898 1.13 mrg ==================================== 3899 1.1 mrg 3900 1.1 mrg Without a specified name, 3901 1.1 mrg 3902 1.1 mrg void GOMP_critical_start (void); 3903 1.1 mrg void GOMP_critical_end (void); 3904 1.1 mrg 3905 1.1 mrg so that we don't get COPY relocations from libgomp to the main 3906 1.1 mrg application. 3907 1.1 mrg 3908 1.1 mrg With a specified name, use omp_set_lock and omp_unset_lock with name 3909 1.1 mrg being transformed into a variable declared like 3910 1.1 mrg 3911 1.1 mrg omp_lock_t gomp_critical_user_<name> __attribute__((common)) 3912 1.1 mrg 3913 1.1 mrg Ideally the ABI would specify that all zero is a valid unlocked 3914 1.3 mrg state, and so we wouldn't need to initialize this at startup. 3915 1.1 mrg 3916 1.1 mrg 3917 1.1 mrg File: libgomp.info, Node: Implementing ATOMIC construct, Next: Implementing FLUSH construct, Prev: Implementing CRITICAL construct, Up: The libgomp ABI 3918 1.1 mrg 3919 1.15 mrg 11.3 Implementing ATOMIC construct 3920 1.13 mrg ================================== 3921 1.1 mrg 3922 1.11 mrg The target should implement the '__sync' builtins. 3923 1.1 mrg 3924 1.1 mrg Failing that we could add 3925 1.1 mrg 3926 1.1 mrg void GOMP_atomic_enter (void) 3927 1.1 mrg void GOMP_atomic_exit (void) 3928 1.1 mrg 3929 1.1 mrg which reuses the regular lock code, but with yet another lock object 3930 1.1 mrg private to the library. 3931 1.1 mrg 3932 1.1 mrg 3933 1.1 mrg File: libgomp.info, Node: Implementing FLUSH construct, Next: Implementing BARRIER construct, Prev: Implementing ATOMIC construct, Up: The libgomp ABI 3934 1.1 mrg 3935 1.15 mrg 11.4 Implementing FLUSH construct 3936 1.13 mrg ================================= 3937 1.1 mrg 3938 1.11 mrg Expands to the '__sync_synchronize' builtin. 3939 1.1 mrg 3940 1.1 mrg 3941 1.1 mrg File: libgomp.info, Node: Implementing BARRIER construct, Next: Implementing THREADPRIVATE construct, Prev: Implementing FLUSH construct, Up: The libgomp ABI 3942 1.1 mrg 3943 1.15 mrg 11.5 Implementing BARRIER construct 3944 1.13 mrg =================================== 3945 1.1 mrg 3946 1.1 mrg void GOMP_barrier (void) 3947 1.1 mrg 3948 1.1 mrg 3949 1.1 mrg File: libgomp.info, Node: Implementing THREADPRIVATE construct, Next: Implementing PRIVATE clause, Prev: Implementing BARRIER construct, Up: The libgomp ABI 3950 1.1 mrg 3951 1.15 mrg 11.6 Implementing THREADPRIVATE construct 3952 1.13 mrg ========================================= 3953 1.1 mrg 3954 1.11 mrg In _most_ cases we can map this directly to '__thread'. Except that OMP 3955 1.11 mrg allows constructors for C++ objects. We can either refuse to support 3956 1.11 mrg this (how often is it used?) or we can implement something akin to 3957 1.11 mrg .ctors. 3958 1.1 mrg 3959 1.1 mrg Even more ideally, this ctor feature is handled by extensions to the 3960 1.1 mrg main pthreads library. Failing that, we can have a set of entry points 3961 1.1 mrg to register ctor functions to be called. 3962 1.1 mrg 3963 1.1 mrg 3964 1.1 mrg File: libgomp.info, Node: Implementing PRIVATE clause, Next: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Prev: Implementing THREADPRIVATE construct, Up: The libgomp ABI 3965 1.1 mrg 3966 1.15 mrg 11.7 Implementing PRIVATE clause 3967 1.13 mrg ================================ 3968 1.1 mrg 3969 1.1 mrg In association with a PARALLEL, or within the lexical extent of a 3970 1.1 mrg PARALLEL block, the variable becomes a local variable in the parallel 3971 1.1 mrg subfunction. 3972 1.1 mrg 3973 1.1 mrg In association with FOR or SECTIONS blocks, create a new automatic 3974 1.1 mrg variable within the current function. This preserves the semantic of 3975 1.1 mrg new variable creation. 3976 1.1 mrg 3977 1.1 mrg 3978 1.1 mrg File: libgomp.info, Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Next: Implementing REDUCTION clause, Prev: Implementing PRIVATE clause, Up: The libgomp ABI 3979 1.1 mrg 3980 1.15 mrg 11.8 Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses 3981 1.13 mrg ========================================================================= 3982 1.1 mrg 3983 1.3 mrg This seems simple enough for PARALLEL blocks. Create a private struct 3984 1.3 mrg for communicating between the parent and subfunction. In the parent, 3985 1.3 mrg copy in values for scalar and "small" structs; copy in addresses for 3986 1.3 mrg others TREE_ADDRESSABLE types. In the subfunction, copy the value into 3987 1.3 mrg the local variable. 3988 1.1 mrg 3989 1.11 mrg It is not clear what to do with bare FOR or SECTION blocks. The only 3990 1.11 mrg thing I can figure is that we do something like: 3991 1.1 mrg 3992 1.1 mrg #pragma omp for firstprivate(x) lastprivate(y) 3993 1.1 mrg for (int i = 0; i < n; ++i) 3994 1.1 mrg body; 3995 1.1 mrg 3996 1.1 mrg which becomes 3997 1.1 mrg 3998 1.1 mrg { 3999 1.1 mrg int x = x, y; 4000 1.1 mrg 4001 1.1 mrg // for stuff 4002 1.1 mrg 4003 1.1 mrg if (i == n) 4004 1.1 mrg y = y; 4005 1.1 mrg } 4006 1.1 mrg 4007 1.1 mrg where the "x=x" and "y=y" assignments actually have different uids 4008 1.11 mrg for the two variables, i.e. not something you could write directly in 4009 1.11 mrg C. Presumably this only makes sense if the "outer" x and y are global 4010 1.1 mrg variables. 4011 1.1 mrg 4012 1.1 mrg COPYPRIVATE would work the same way, except the structure broadcast 4013 1.1 mrg would have to happen via SINGLE machinery instead. 4014 1.1 mrg 4015 1.1 mrg 4016 1.1 mrg File: libgomp.info, Node: Implementing REDUCTION clause, Next: Implementing PARALLEL construct, Prev: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses, Up: The libgomp ABI 4017 1.1 mrg 4018 1.15 mrg 11.9 Implementing REDUCTION clause 4019 1.13 mrg ================================== 4020 1.1 mrg 4021 1.1 mrg The private struct mentioned in the previous section should have a 4022 1.11 mrg pointer to an array of the type of the variable, indexed by the thread's 4023 1.11 mrg TEAM_ID. The thread stores its final value into the array, and after 4024 1.15 mrg the barrier, the primary thread iterates over the array to collect the 4025 1.11 mrg values. 4026 1.1 mrg 4027 1.1 mrg 4028 1.1 mrg File: libgomp.info, Node: Implementing PARALLEL construct, Next: Implementing FOR construct, Prev: Implementing REDUCTION clause, Up: The libgomp ABI 4029 1.1 mrg 4030 1.15 mrg 11.10 Implementing PARALLEL construct 4031 1.13 mrg ===================================== 4032 1.1 mrg 4033 1.1 mrg #pragma omp parallel 4034 1.1 mrg { 4035 1.1 mrg body; 4036 1.1 mrg } 4037 1.1 mrg 4038 1.1 mrg becomes 4039 1.1 mrg 4040 1.1 mrg void subfunction (void *data) 4041 1.1 mrg { 4042 1.1 mrg use data; 4043 1.1 mrg body; 4044 1.1 mrg } 4045 1.1 mrg 4046 1.1 mrg setup data; 4047 1.1 mrg GOMP_parallel_start (subfunction, &data, num_threads); 4048 1.1 mrg subfunction (&data); 4049 1.1 mrg GOMP_parallel_end (); 4050 1.1 mrg 4051 1.1 mrg void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads) 4052 1.1 mrg 4053 1.1 mrg The FN argument is the subfunction to be run in parallel. 4054 1.1 mrg 4055 1.1 mrg The DATA argument is a pointer to a structure used to communicate 4056 1.1 mrg data in and out of the subfunction, as discussed above with respect to 4057 1.1 mrg FIRSTPRIVATE et al. 4058 1.1 mrg 4059 1.1 mrg The NUM_THREADS argument is 1 if an IF clause is present and false, 4060 1.1 mrg or the value of the NUM_THREADS clause, if present, or 0. 4061 1.1 mrg 4062 1.11 mrg The function needs to create the appropriate number of threads and/or 4063 1.11 mrg launch them from the dock. It needs to create the team structure and 4064 1.11 mrg assign team ids. 4065 1.1 mrg 4066 1.1 mrg void GOMP_parallel_end (void) 4067 1.1 mrg 4068 1.1 mrg Tears down the team and returns us to the previous 4069 1.11 mrg 'omp_in_parallel()' state. 4070 1.1 mrg 4071 1.1 mrg 4072 1.1 mrg File: libgomp.info, Node: Implementing FOR construct, Next: Implementing ORDERED construct, Prev: Implementing PARALLEL construct, Up: The libgomp ABI 4073 1.1 mrg 4074 1.15 mrg 11.11 Implementing FOR construct 4075 1.13 mrg ================================ 4076 1.1 mrg 4077 1.1 mrg #pragma omp parallel for 4078 1.1 mrg for (i = lb; i <= ub; i++) 4079 1.1 mrg body; 4080 1.1 mrg 4081 1.1 mrg becomes 4082 1.1 mrg 4083 1.1 mrg void subfunction (void *data) 4084 1.1 mrg { 4085 1.1 mrg long _s0, _e0; 4086 1.1 mrg while (GOMP_loop_static_next (&_s0, &_e0)) 4087 1.1 mrg { 4088 1.1 mrg long _e1 = _e0, i; 4089 1.1 mrg for (i = _s0; i < _e1; i++) 4090 1.1 mrg body; 4091 1.1 mrg } 4092 1.1 mrg GOMP_loop_end_nowait (); 4093 1.1 mrg } 4094 1.1 mrg 4095 1.1 mrg GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0); 4096 1.1 mrg subfunction (NULL); 4097 1.1 mrg GOMP_parallel_end (); 4098 1.1 mrg 4099 1.1 mrg #pragma omp for schedule(runtime) 4100 1.1 mrg for (i = 0; i < n; i++) 4101 1.1 mrg body; 4102 1.1 mrg 4103 1.1 mrg becomes 4104 1.1 mrg 4105 1.1 mrg { 4106 1.1 mrg long i, _s0, _e0; 4107 1.1 mrg if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0)) 4108 1.1 mrg do { 4109 1.1 mrg long _e1 = _e0; 4110 1.1 mrg for (i = _s0, i < _e0; i++) 4111 1.1 mrg body; 4112 1.1 mrg } while (GOMP_loop_runtime_next (&_s0, _&e0)); 4113 1.1 mrg GOMP_loop_end (); 4114 1.1 mrg } 4115 1.1 mrg 4116 1.3 mrg Note that while it looks like there is trickiness to propagating a 4117 1.1 mrg non-constant STEP, there isn't really. We're explicitly allowed to 4118 1.1 mrg evaluate it as many times as we want, and any variables involved should 4119 1.1 mrg automatically be handled as PRIVATE or SHARED like any other variables. 4120 1.1 mrg So the expression should remain evaluable in the subfunction. We can 4121 1.11 mrg also pull it into a local variable if we like, but since its supposed to 4122 1.11 mrg remain unchanged, we can also not if we like. 4123 1.1 mrg 4124 1.11 mrg If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be able 4125 1.11 mrg to get away with no work-sharing context at all, since we can simply 4126 1.11 mrg perform the arithmetic directly in each thread to divide up the 4127 1.11 mrg iterations. Which would mean that we wouldn't need to call any of these 4128 1.11 mrg routines. 4129 1.1 mrg 4130 1.1 mrg There are separate routines for handling loops with an ORDERED 4131 1.1 mrg clause. Bookkeeping for that is non-trivial... 4132 1.1 mrg 4133 1.1 mrg 4134 1.1 mrg File: libgomp.info, Node: Implementing ORDERED construct, Next: Implementing SECTIONS construct, Prev: Implementing FOR construct, Up: The libgomp ABI 4135 1.1 mrg 4136 1.15 mrg 11.12 Implementing ORDERED construct 4137 1.13 mrg ==================================== 4138 1.1 mrg 4139 1.1 mrg void GOMP_ordered_start (void) 4140 1.1 mrg void GOMP_ordered_end (void) 4141 1.1 mrg 4142 1.1 mrg 4143 1.1 mrg File: libgomp.info, Node: Implementing SECTIONS construct, Next: Implementing SINGLE construct, Prev: Implementing ORDERED construct, Up: The libgomp ABI 4144 1.1 mrg 4145 1.15 mrg 11.13 Implementing SECTIONS construct 4146 1.13 mrg ===================================== 4147 1.1 mrg 4148 1.1 mrg A block as 4149 1.1 mrg 4150 1.1 mrg #pragma omp sections 4151 1.1 mrg { 4152 1.1 mrg #pragma omp section 4153 1.1 mrg stmt1; 4154 1.1 mrg #pragma omp section 4155 1.1 mrg stmt2; 4156 1.1 mrg #pragma omp section 4157 1.1 mrg stmt3; 4158 1.1 mrg } 4159 1.1 mrg 4160 1.1 mrg becomes 4161 1.1 mrg 4162 1.1 mrg for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ()) 4163 1.1 mrg switch (i) 4164 1.1 mrg { 4165 1.1 mrg case 1: 4166 1.1 mrg stmt1; 4167 1.1 mrg break; 4168 1.1 mrg case 2: 4169 1.1 mrg stmt2; 4170 1.1 mrg break; 4171 1.1 mrg case 3: 4172 1.1 mrg stmt3; 4173 1.1 mrg break; 4174 1.1 mrg } 4175 1.1 mrg GOMP_barrier (); 4176 1.1 mrg 4177 1.1 mrg 4178 1.7 mrg File: libgomp.info, Node: Implementing SINGLE construct, Next: Implementing OpenACC's PARALLEL construct, Prev: Implementing SECTIONS construct, Up: The libgomp ABI 4179 1.1 mrg 4180 1.15 mrg 11.14 Implementing SINGLE construct 4181 1.13 mrg =================================== 4182 1.1 mrg 4183 1.1 mrg A block like 4184 1.1 mrg 4185 1.1 mrg #pragma omp single 4186 1.1 mrg { 4187 1.1 mrg body; 4188 1.1 mrg } 4189 1.1 mrg 4190 1.1 mrg becomes 4191 1.1 mrg 4192 1.1 mrg if (GOMP_single_start ()) 4193 1.1 mrg body; 4194 1.1 mrg GOMP_barrier (); 4195 1.1 mrg 4196 1.1 mrg while 4197 1.1 mrg 4198 1.1 mrg #pragma omp single copyprivate(x) 4199 1.1 mrg body; 4200 1.1 mrg 4201 1.1 mrg becomes 4202 1.1 mrg 4203 1.1 mrg datap = GOMP_single_copy_start (); 4204 1.1 mrg if (datap == NULL) 4205 1.1 mrg { 4206 1.1 mrg body; 4207 1.1 mrg data.x = x; 4208 1.1 mrg GOMP_single_copy_end (&data); 4209 1.1 mrg } 4210 1.1 mrg else 4211 1.1 mrg x = datap->x; 4212 1.1 mrg GOMP_barrier (); 4213 1.1 mrg 4214 1.1 mrg 4215 1.7 mrg File: libgomp.info, Node: Implementing OpenACC's PARALLEL construct, Prev: Implementing SINGLE construct, Up: The libgomp ABI 4216 1.7 mrg 4217 1.15 mrg 11.15 Implementing OpenACC's PARALLEL construct 4218 1.13 mrg =============================================== 4219 1.7 mrg 4220 1.7 mrg void GOACC_parallel () 4221 1.7 mrg 4222 1.7 mrg 4223 1.1 mrg File: libgomp.info, Node: Reporting Bugs, Next: Copying, Prev: The libgomp ABI, Up: Top 4224 1.1 mrg 4225 1.15 mrg 12 Reporting Bugs 4226 1.7 mrg ***************** 4227 1.1 mrg 4228 1.5 mrg Bugs in the GNU Offloading and Multi Processing Runtime Library should 4229 1.13 mrg be reported via Bugzilla (https://gcc.gnu.org/bugzilla/). Please add 4230 1.11 mrg "openacc", or "openmp", or both to the keywords field in the bug report, 4231 1.11 mrg as appropriate. 4232 1.1 mrg 4233 1.1 mrg 4234 1.1 mrg File: libgomp.info, Node: Copying, Next: GNU Free Documentation License, Prev: Reporting Bugs, Up: Top 4235 1.1 mrg 4236 1.3 mrg GNU General Public License 4237 1.1 mrg ************************** 4238 1.1 mrg 4239 1.3 mrg Version 3, 29 June 2007 4240 1.1 mrg 4241 1.15 mrg Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/> 4242 1.1 mrg 4243 1.3 mrg Everyone is permitted to copy and distribute verbatim copies of this 4244 1.3 mrg license document, but changing it is not allowed. 4245 1.1 mrg 4246 1.1 mrg Preamble 4247 1.1 mrg ======== 4248 1.1 mrg 4249 1.3 mrg The GNU General Public License is a free, copyleft license for software 4250 1.3 mrg and other kinds of works. 4251 1.3 mrg 4252 1.3 mrg The licenses for most software and other practical works are designed 4253 1.3 mrg to take away your freedom to share and change the works. By contrast, 4254 1.3 mrg the GNU General Public License is intended to guarantee your freedom to 4255 1.3 mrg share and change all versions of a program-to make sure it remains free 4256 1.3 mrg software for all its users. We, the Free Software Foundation, use the 4257 1.3 mrg GNU General Public License for most of our software; it applies also to 4258 1.3 mrg any other work released this way by its authors. You can apply it to 4259 1.1 mrg your programs, too. 4260 1.1 mrg 4261 1.1 mrg When we speak of free software, we are referring to freedom, not 4262 1.1 mrg price. Our General Public Licenses are designed to make sure that you 4263 1.1 mrg have the freedom to distribute copies of free software (and charge for 4264 1.3 mrg them if you wish), that you receive source code or can get it if you 4265 1.3 mrg want it, that you can change the software or use pieces of it in new 4266 1.3 mrg free programs, and that you know you can do these things. 4267 1.3 mrg 4268 1.3 mrg To protect your rights, we need to prevent others from denying you 4269 1.11 mrg these rights or asking you to surrender the rights. Therefore, you have 4270 1.11 mrg certain responsibilities if you distribute copies of the software, or if 4271 1.11 mrg you modify it: responsibilities to respect the freedom of others. 4272 1.1 mrg 4273 1.1 mrg For example, if you distribute copies of such a program, whether 4274 1.3 mrg gratis or for a fee, you must pass on to the recipients the same 4275 1.3 mrg freedoms that you received. You must make sure that they, too, receive 4276 1.3 mrg or can get the source code. And you must show them these terms so they 4277 1.3 mrg know their rights. 4278 1.3 mrg 4279 1.3 mrg Developers that use the GNU GPL protect your rights with two steps: 4280 1.3 mrg (1) assert copyright on the software, and (2) offer you this License 4281 1.3 mrg giving you legal permission to copy, distribute and/or modify it. 4282 1.3 mrg 4283 1.3 mrg For the developers' and authors' protection, the GPL clearly explains 4284 1.3 mrg that there is no warranty for this free software. For both users' and 4285 1.3 mrg authors' sake, the GPL requires that modified versions be marked as 4286 1.3 mrg changed, so that their problems will not be attributed erroneously to 4287 1.3 mrg authors of previous versions. 4288 1.3 mrg 4289 1.3 mrg Some devices are designed to deny users access to install or run 4290 1.11 mrg modified versions of the software inside them, although the manufacturer 4291 1.11 mrg can do so. This is fundamentally incompatible with the aim of 4292 1.11 mrg protecting users' freedom to change the software. The systematic 4293 1.11 mrg pattern of such abuse occurs in the area of products for individuals to 4294 1.11 mrg use, which is precisely where it is most unacceptable. Therefore, we 4295 1.11 mrg have designed this version of the GPL to prohibit the practice for those 4296 1.11 mrg products. If such problems arise substantially in other domains, we 4297 1.11 mrg stand ready to extend this provision to those domains in future versions 4298 1.11 mrg of the GPL, as needed to protect the freedom of users. 4299 1.3 mrg 4300 1.3 mrg Finally, every program is threatened constantly by software patents. 4301 1.3 mrg States should not allow patents to restrict development and use of 4302 1.3 mrg software on general-purpose computers, but in those that do, we wish to 4303 1.3 mrg avoid the special danger that patents applied to a free program could 4304 1.3 mrg make it effectively proprietary. To prevent this, the GPL assures that 4305 1.3 mrg patents cannot be used to render the program non-free. 4306 1.1 mrg 4307 1.1 mrg The precise terms and conditions for copying, distribution and 4308 1.1 mrg modification follow. 4309 1.1 mrg 4310 1.3 mrg TERMS AND CONDITIONS 4311 1.3 mrg ==================== 4312 1.3 mrg 4313 1.3 mrg 0. Definitions. 4314 1.3 mrg 4315 1.3 mrg "This License" refers to version 3 of the GNU General Public 4316 1.3 mrg License. 4317 1.1 mrg 4318 1.3 mrg "Copyright" also means copyright-like laws that apply to other 4319 1.3 mrg kinds of works, such as semiconductor masks. 4320 1.3 mrg 4321 1.3 mrg "The Program" refers to any copyrightable work licensed under this 4322 1.3 mrg License. Each licensee is addressed as "you". "Licensees" and 4323 1.3 mrg "recipients" may be individuals or organizations. 4324 1.3 mrg 4325 1.3 mrg To "modify" a work means to copy from or adapt all or part of the 4326 1.3 mrg work in a fashion requiring copyright permission, other than the 4327 1.3 mrg making of an exact copy. The resulting work is called a "modified 4328 1.3 mrg version" of the earlier work or a work "based on" the earlier work. 4329 1.3 mrg 4330 1.3 mrg A "covered work" means either the unmodified Program or a work 4331 1.3 mrg based on the Program. 4332 1.3 mrg 4333 1.3 mrg To "propagate" a work means to do anything with it that, without 4334 1.3 mrg permission, would make you directly or secondarily liable for 4335 1.11 mrg infringement under applicable copyright law, except executing it on 4336 1.11 mrg a computer or modifying a private copy. Propagation includes 4337 1.3 mrg copying, distribution (with or without modification), making 4338 1.3 mrg available to the public, and in some countries other activities as 4339 1.3 mrg well. 4340 1.3 mrg 4341 1.3 mrg To "convey" a work means any kind of propagation that enables other 4342 1.3 mrg parties to make or receive copies. Mere interaction with a user 4343 1.3 mrg through a computer network, with no transfer of a copy, is not 4344 1.3 mrg conveying. 4345 1.3 mrg 4346 1.3 mrg An interactive user interface displays "Appropriate Legal Notices" 4347 1.3 mrg to the extent that it includes a convenient and prominently visible 4348 1.3 mrg feature that (1) displays an appropriate copyright notice, and (2) 4349 1.3 mrg tells the user that there is no warranty for the work (except to 4350 1.11 mrg the extent that warranties are provided), that licensees may convey 4351 1.11 mrg the work under this License, and how to view a copy of this 4352 1.3 mrg License. If the interface presents a list of user commands or 4353 1.3 mrg options, such as a menu, a prominent item in the list meets this 4354 1.3 mrg criterion. 4355 1.3 mrg 4356 1.3 mrg 1. Source Code. 4357 1.3 mrg 4358 1.3 mrg The "source code" for a work means the preferred form of the work 4359 1.11 mrg for making modifications to it. "Object code" means any non-source 4360 1.11 mrg form of a work. 4361 1.3 mrg 4362 1.3 mrg A "Standard Interface" means an interface that either is an 4363 1.3 mrg official standard defined by a recognized standards body, or, in 4364 1.3 mrg the case of interfaces specified for a particular programming 4365 1.3 mrg language, one that is widely used among developers working in that 4366 1.3 mrg language. 4367 1.3 mrg 4368 1.3 mrg The "System Libraries" of an executable work include anything, 4369 1.3 mrg other than the work as a whole, that (a) is included in the normal 4370 1.3 mrg form of packaging a Major Component, but which is not part of that 4371 1.11 mrg Major Component, and (b) serves only to enable use of the work with 4372 1.11 mrg that Major Component, or to implement a Standard Interface for 4373 1.11 mrg which an implementation is available to the public in source code 4374 1.11 mrg form. A "Major Component", in this context, means a major 4375 1.3 mrg essential component (kernel, window system, and so on) of the 4376 1.3 mrg specific operating system (if any) on which the executable work 4377 1.3 mrg runs, or a compiler used to produce the work, or an object code 4378 1.3 mrg interpreter used to run it. 4379 1.3 mrg 4380 1.3 mrg The "Corresponding Source" for a work in object code form means all 4381 1.3 mrg the source code needed to generate, install, and (for an executable 4382 1.11 mrg work) run the object code and to modify the work, including scripts 4383 1.11 mrg to control those activities. However, it does not include the 4384 1.11 mrg work's System Libraries, or general-purpose tools or generally 4385 1.3 mrg available free programs which are used unmodified in performing 4386 1.3 mrg those activities but which are not part of the work. For example, 4387 1.11 mrg Corresponding Source includes interface definition files associated 4388 1.11 mrg with source files for the work, and the source code for shared 4389 1.11 mrg libraries and dynamically linked subprograms that the work is 4390 1.11 mrg specifically designed to require, such as by intimate data 4391 1.3 mrg communication or control flow between those subprograms and other 4392 1.3 mrg parts of the work. 4393 1.3 mrg 4394 1.3 mrg The Corresponding Source need not include anything that users can 4395 1.3 mrg regenerate automatically from other parts of the Corresponding 4396 1.3 mrg Source. 4397 1.3 mrg 4398 1.3 mrg The Corresponding Source for a work in source code form is that 4399 1.3 mrg same work. 4400 1.3 mrg 4401 1.3 mrg 2. Basic Permissions. 4402 1.3 mrg 4403 1.3 mrg All rights granted under this License are granted for the term of 4404 1.3 mrg copyright on the Program, and are irrevocable provided the stated 4405 1.3 mrg conditions are met. This License explicitly affirms your unlimited 4406 1.3 mrg permission to run the unmodified Program. The output from running 4407 1.11 mrg a covered work is covered by this License only if the output, given 4408 1.11 mrg its content, constitutes a covered work. This License acknowledges 4409 1.11 mrg your rights of fair use or other equivalent, as provided by 4410 1.11 mrg copyright law. 4411 1.3 mrg 4412 1.3 mrg You may make, run and propagate covered works that you do not 4413 1.3 mrg convey, without conditions so long as your license otherwise 4414 1.3 mrg remains in force. You may convey covered works to others for the 4415 1.11 mrg sole purpose of having them make modifications exclusively for you, 4416 1.11 mrg or provide you with facilities for running those works, provided 4417 1.11 mrg that you comply with the terms of this License in conveying all 4418 1.11 mrg material for which you do not control copyright. Those thus making 4419 1.11 mrg or running the covered works for you must do so exclusively on your 4420 1.11 mrg behalf, under your direction and control, on terms that prohibit 4421 1.11 mrg them from making any copies of your copyrighted material outside 4422 1.11 mrg their relationship with you. 4423 1.3 mrg 4424 1.3 mrg Conveying under any other circumstances is permitted solely under 4425 1.3 mrg the conditions stated below. Sublicensing is not allowed; section 4426 1.3 mrg 10 makes it unnecessary. 4427 1.3 mrg 4428 1.3 mrg 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 4429 1.3 mrg 4430 1.3 mrg No covered work shall be deemed part of an effective technological 4431 1.3 mrg measure under any applicable law fulfilling obligations under 4432 1.3 mrg article 11 of the WIPO copyright treaty adopted on 20 December 4433 1.3 mrg 1996, or similar laws prohibiting or restricting circumvention of 4434 1.3 mrg such measures. 4435 1.3 mrg 4436 1.3 mrg When you convey a covered work, you waive any legal power to forbid 4437 1.3 mrg circumvention of technological measures to the extent such 4438 1.3 mrg circumvention is effected by exercising rights under this License 4439 1.11 mrg with respect to the covered work, and you disclaim any intention to 4440 1.11 mrg limit operation or modification of the work as a means of 4441 1.3 mrg enforcing, against the work's users, your or third parties' legal 4442 1.3 mrg rights to forbid circumvention of technological measures. 4443 1.3 mrg 4444 1.3 mrg 4. Conveying Verbatim Copies. 4445 1.3 mrg 4446 1.3 mrg You may convey verbatim copies of the Program's source code as you 4447 1.3 mrg receive it, in any medium, provided that you conspicuously and 4448 1.3 mrg appropriately publish on each copy an appropriate copyright notice; 4449 1.3 mrg keep intact all notices stating that this License and any 4450 1.3 mrg non-permissive terms added in accord with section 7 apply to the 4451 1.3 mrg code; keep intact all notices of the absence of any warranty; and 4452 1.3 mrg give all recipients a copy of this License along with the Program. 4453 1.3 mrg 4454 1.3 mrg You may charge any price or no price for each copy that you convey, 4455 1.3 mrg and you may offer support or warranty protection for a fee. 4456 1.3 mrg 4457 1.3 mrg 5. Conveying Modified Source Versions. 4458 1.3 mrg 4459 1.3 mrg You may convey a work based on the Program, or the modifications to 4460 1.3 mrg produce it from the Program, in the form of source code under the 4461 1.3 mrg terms of section 4, provided that you also meet all of these 4462 1.3 mrg conditions: 4463 1.3 mrg 4464 1.3 mrg a. The work must carry prominent notices stating that you 4465 1.3 mrg modified it, and giving a relevant date. 4466 1.3 mrg 4467 1.3 mrg b. The work must carry prominent notices stating that it is 4468 1.3 mrg released under this License and any conditions added under 4469 1.3 mrg section 7. This requirement modifies the requirement in 4470 1.3 mrg section 4 to "keep intact all notices". 4471 1.3 mrg 4472 1.3 mrg c. You must license the entire work, as a whole, under this 4473 1.3 mrg License to anyone who comes into possession of a copy. This 4474 1.3 mrg License will therefore apply, along with any applicable 4475 1.3 mrg section 7 additional terms, to the whole of the work, and all 4476 1.3 mrg its parts, regardless of how they are packaged. This License 4477 1.3 mrg gives no permission to license the work in any other way, but 4478 1.3 mrg it does not invalidate such permission if you have separately 4479 1.3 mrg received it. 4480 1.3 mrg 4481 1.3 mrg d. If the work has interactive user interfaces, each must display 4482 1.3 mrg Appropriate Legal Notices; however, if the Program has 4483 1.3 mrg interactive interfaces that do not display Appropriate Legal 4484 1.3 mrg Notices, your work need not make them do so. 4485 1.3 mrg 4486 1.3 mrg A compilation of a covered work with other separate and independent 4487 1.3 mrg works, which are not by their nature extensions of the covered 4488 1.3 mrg work, and which are not combined with it such as to form a larger 4489 1.3 mrg program, in or on a volume of a storage or distribution medium, is 4490 1.3 mrg called an "aggregate" if the compilation and its resulting 4491 1.3 mrg copyright are not used to limit the access or legal rights of the 4492 1.3 mrg compilation's users beyond what the individual works permit. 4493 1.3 mrg Inclusion of a covered work in an aggregate does not cause this 4494 1.3 mrg License to apply to the other parts of the aggregate. 4495 1.3 mrg 4496 1.3 mrg 6. Conveying Non-Source Forms. 4497 1.3 mrg 4498 1.3 mrg You may convey a covered work in object code form under the terms 4499 1.3 mrg of sections 4 and 5, provided that you also convey the 4500 1.3 mrg machine-readable Corresponding Source under the terms of this 4501 1.3 mrg License, in one of these ways: 4502 1.3 mrg 4503 1.3 mrg a. Convey the object code in, or embodied in, a physical product 4504 1.3 mrg (including a physical distribution medium), accompanied by the 4505 1.3 mrg Corresponding Source fixed on a durable physical medium 4506 1.3 mrg customarily used for software interchange. 4507 1.3 mrg 4508 1.3 mrg b. Convey the object code in, or embodied in, a physical product 4509 1.3 mrg (including a physical distribution medium), accompanied by a 4510 1.11 mrg written offer, valid for at least three years and valid for as 4511 1.11 mrg long as you offer spare parts or customer support for that 4512 1.3 mrg product model, to give anyone who possesses the object code 4513 1.3 mrg either (1) a copy of the Corresponding Source for all the 4514 1.3 mrg software in the product that is covered by this License, on a 4515 1.3 mrg durable physical medium customarily used for software 4516 1.3 mrg interchange, for a price no more than your reasonable cost of 4517 1.3 mrg physically performing this conveying of source, or (2) access 4518 1.3 mrg to copy the Corresponding Source from a network server at no 4519 1.3 mrg charge. 4520 1.3 mrg 4521 1.11 mrg c. Convey individual copies of the object code with a copy of the 4522 1.11 mrg written offer to provide the Corresponding Source. This 4523 1.3 mrg alternative is allowed only occasionally and noncommercially, 4524 1.3 mrg and only if you received the object code with such an offer, 4525 1.3 mrg in accord with subsection 6b. 4526 1.3 mrg 4527 1.3 mrg d. Convey the object code by offering access from a designated 4528 1.11 mrg place (gratis or for a charge), and offer equivalent access to 4529 1.11 mrg the Corresponding Source in the same way through the same 4530 1.3 mrg place at no further charge. You need not require recipients 4531 1.3 mrg to copy the Corresponding Source along with the object code. 4532 1.3 mrg If the place to copy the object code is a network server, the 4533 1.11 mrg Corresponding Source may be on a different server (operated by 4534 1.11 mrg you or a third party) that supports equivalent copying 4535 1.11 mrg facilities, provided you maintain clear directions next to the 4536 1.11 mrg object code saying where to find the Corresponding Source. 4537 1.3 mrg Regardless of what server hosts the Corresponding Source, you 4538 1.11 mrg remain obligated to ensure that it is available for as long as 4539 1.11 mrg needed to satisfy these requirements. 4540 1.3 mrg 4541 1.3 mrg e. Convey the object code using peer-to-peer transmission, 4542 1.3 mrg provided you inform other peers where the object code and 4543 1.3 mrg Corresponding Source of the work are being offered to the 4544 1.3 mrg general public at no charge under subsection 6d. 4545 1.3 mrg 4546 1.3 mrg A separable portion of the object code, whose source code is 4547 1.3 mrg excluded from the Corresponding Source as a System Library, need 4548 1.3 mrg not be included in conveying the object code work. 4549 1.3 mrg 4550 1.3 mrg A "User Product" is either (1) a "consumer product", which means 4551 1.3 mrg any tangible personal property which is normally used for personal, 4552 1.3 mrg family, or household purposes, or (2) anything designed or sold for 4553 1.11 mrg incorporation into a dwelling. In determining whether a product is 4554 1.11 mrg a consumer product, doubtful cases shall be resolved in favor of 4555 1.3 mrg coverage. For a particular product received by a particular user, 4556 1.3 mrg "normally used" refers to a typical or common use of that class of 4557 1.3 mrg product, regardless of the status of the particular user or of the 4558 1.3 mrg way in which the particular user actually uses, or expects or is 4559 1.3 mrg expected to use, the product. A product is a consumer product 4560 1.3 mrg regardless of whether the product has substantial commercial, 4561 1.3 mrg industrial or non-consumer uses, unless such uses represent the 4562 1.3 mrg only significant mode of use of the product. 4563 1.3 mrg 4564 1.3 mrg "Installation Information" for a User Product means any methods, 4565 1.3 mrg procedures, authorization keys, or other information required to 4566 1.3 mrg install and execute modified versions of a covered work in that 4567 1.3 mrg User Product from a modified version of its Corresponding Source. 4568 1.3 mrg The information must suffice to ensure that the continued 4569 1.3 mrg functioning of the modified object code is in no case prevented or 4570 1.3 mrg interfered with solely because modification has been made. 4571 1.3 mrg 4572 1.3 mrg If you convey an object code work under this section in, or with, 4573 1.3 mrg or specifically for use in, a User Product, and the conveying 4574 1.3 mrg occurs as part of a transaction in which the right of possession 4575 1.3 mrg and use of the User Product is transferred to the recipient in 4576 1.3 mrg perpetuity or for a fixed term (regardless of how the transaction 4577 1.3 mrg is characterized), the Corresponding Source conveyed under this 4578 1.3 mrg section must be accompanied by the Installation Information. But 4579 1.3 mrg this requirement does not apply if neither you nor any third party 4580 1.3 mrg retains the ability to install modified object code on the User 4581 1.3 mrg Product (for example, the work has been installed in ROM). 4582 1.3 mrg 4583 1.3 mrg The requirement to provide Installation Information does not 4584 1.3 mrg include a requirement to continue to provide support service, 4585 1.11 mrg warranty, or updates for a work that has been modified or installed 4586 1.11 mrg by the recipient, or for the User Product in which it has been 4587 1.11 mrg modified or installed. Access to a network may be denied when the 4588 1.11 mrg modification itself materially and adversely affects the operation 4589 1.11 mrg of the network or violates the rules and protocols for 4590 1.3 mrg communication across the network. 4591 1.3 mrg 4592 1.3 mrg Corresponding Source conveyed, and Installation Information 4593 1.3 mrg provided, in accord with this section must be in a format that is 4594 1.3 mrg publicly documented (and with an implementation available to the 4595 1.3 mrg public in source code form), and must require no special password 4596 1.3 mrg or key for unpacking, reading or copying. 4597 1.3 mrg 4598 1.3 mrg 7. Additional Terms. 4599 1.3 mrg 4600 1.3 mrg "Additional permissions" are terms that supplement the terms of 4601 1.3 mrg this License by making exceptions from one or more of its 4602 1.3 mrg conditions. Additional permissions that are applicable to the 4603 1.3 mrg entire Program shall be treated as though they were included in 4604 1.3 mrg this License, to the extent that they are valid under applicable 4605 1.3 mrg law. If additional permissions apply only to part of the Program, 4606 1.3 mrg that part may be used separately under those permissions, but the 4607 1.3 mrg entire Program remains governed by this License without regard to 4608 1.3 mrg the additional permissions. 4609 1.3 mrg 4610 1.3 mrg When you convey a copy of a covered work, you may at your option 4611 1.3 mrg remove any additional permissions from that copy, or from any part 4612 1.3 mrg of it. (Additional permissions may be written to require their own 4613 1.3 mrg removal in certain cases when you modify the work.) You may place 4614 1.3 mrg additional permissions on material, added by you to a covered work, 4615 1.3 mrg for which you have or can give appropriate copyright permission. 4616 1.3 mrg 4617 1.3 mrg Notwithstanding any other provision of this License, for material 4618 1.3 mrg you add to a covered work, you may (if authorized by the copyright 4619 1.11 mrg holders of that material) supplement the terms of this License with 4620 1.11 mrg terms: 4621 1.3 mrg 4622 1.3 mrg a. Disclaiming warranty or limiting liability differently from 4623 1.3 mrg the terms of sections 15 and 16 of this License; or 4624 1.3 mrg 4625 1.3 mrg b. Requiring preservation of specified reasonable legal notices 4626 1.3 mrg or author attributions in that material or in the Appropriate 4627 1.3 mrg Legal Notices displayed by works containing it; or 4628 1.3 mrg 4629 1.3 mrg c. Prohibiting misrepresentation of the origin of that material, 4630 1.11 mrg or requiring that modified versions of such material be marked 4631 1.11 mrg in reasonable ways as different from the original version; or 4632 1.3 mrg 4633 1.3 mrg d. Limiting the use for publicity purposes of names of licensors 4634 1.3 mrg or authors of the material; or 4635 1.3 mrg 4636 1.3 mrg e. Declining to grant rights under trademark law for use of some 4637 1.3 mrg trade names, trademarks, or service marks; or 4638 1.3 mrg 4639 1.3 mrg f. Requiring indemnification of licensors and authors of that 4640 1.3 mrg material by anyone who conveys the material (or modified 4641 1.3 mrg versions of it) with contractual assumptions of liability to 4642 1.3 mrg the recipient, for any liability that these contractual 4643 1.3 mrg assumptions directly impose on those licensors and authors. 4644 1.3 mrg 4645 1.3 mrg All other non-permissive additional terms are considered "further 4646 1.3 mrg restrictions" within the meaning of section 10. If the Program as 4647 1.3 mrg you received it, or any part of it, contains a notice stating that 4648 1.3 mrg it is governed by this License along with a term that is a further 4649 1.3 mrg restriction, you may remove that term. If a license document 4650 1.11 mrg contains a further restriction but permits relicensing or conveying 4651 1.11 mrg under this License, you may add to a covered work material governed 4652 1.11 mrg by the terms of that license document, provided that the further 4653 1.11 mrg restriction does not survive such relicensing or conveying. 4654 1.3 mrg 4655 1.3 mrg If you add terms to a covered work in accord with this section, you 4656 1.3 mrg must place, in the relevant source files, a statement of the 4657 1.3 mrg additional terms that apply to those files, or a notice indicating 4658 1.3 mrg where to find the applicable terms. 4659 1.3 mrg 4660 1.3 mrg Additional terms, permissive or non-permissive, may be stated in 4661 1.3 mrg the form of a separately written license, or stated as exceptions; 4662 1.3 mrg the above requirements apply either way. 4663 1.3 mrg 4664 1.3 mrg 8. Termination. 4665 1.3 mrg 4666 1.3 mrg You may not propagate or modify a covered work except as expressly 4667 1.3 mrg provided under this License. Any attempt otherwise to propagate or 4668 1.3 mrg modify it is void, and will automatically terminate your rights 4669 1.11 mrg under this License (including any patent licenses granted under the 4670 1.11 mrg third paragraph of section 11). 4671 1.3 mrg 4672 1.3 mrg However, if you cease all violation of this License, then your 4673 1.3 mrg license from a particular copyright holder is reinstated (a) 4674 1.11 mrg provisionally, unless and until the copyright holder explicitly and 4675 1.11 mrg finally terminates your license, and (b) permanently, if the 4676 1.3 mrg copyright holder fails to notify you of the violation by some 4677 1.3 mrg reasonable means prior to 60 days after the cessation. 4678 1.3 mrg 4679 1.3 mrg Moreover, your license from a particular copyright holder is 4680 1.3 mrg reinstated permanently if the copyright holder notifies you of the 4681 1.3 mrg violation by some reasonable means, this is the first time you have 4682 1.3 mrg received notice of violation of this License (for any work) from 4683 1.3 mrg that copyright holder, and you cure the violation prior to 30 days 4684 1.3 mrg after your receipt of the notice. 4685 1.3 mrg 4686 1.3 mrg Termination of your rights under this section does not terminate 4687 1.11 mrg the licenses of parties who have received copies or rights from you 4688 1.11 mrg under this License. If your rights have been terminated and not 4689 1.11 mrg permanently reinstated, you do not qualify to receive new licenses 4690 1.11 mrg for the same material under section 10. 4691 1.3 mrg 4692 1.3 mrg 9. Acceptance Not Required for Having Copies. 4693 1.3 mrg 4694 1.3 mrg You are not required to accept this License in order to receive or 4695 1.3 mrg run a copy of the Program. Ancillary propagation of a covered work 4696 1.3 mrg occurring solely as a consequence of using peer-to-peer 4697 1.3 mrg transmission to receive a copy likewise does not require 4698 1.3 mrg acceptance. However, nothing other than this License grants you 4699 1.3 mrg permission to propagate or modify any covered work. These actions 4700 1.3 mrg infringe copyright if you do not accept this License. Therefore, 4701 1.3 mrg by modifying or propagating a covered work, you indicate your 4702 1.3 mrg acceptance of this License to do so. 4703 1.3 mrg 4704 1.11 mrg 10. Automatic Licensing of Downstream Recipients. 4705 1.3 mrg 4706 1.3 mrg Each time you convey a covered work, the recipient automatically 4707 1.3 mrg receives a license from the original licensors, to run, modify and 4708 1.3 mrg propagate that work, subject to this License. You are not 4709 1.3 mrg responsible for enforcing compliance by third parties with this 4710 1.3 mrg License. 4711 1.3 mrg 4712 1.3 mrg An "entity transaction" is a transaction transferring control of an 4713 1.3 mrg organization, or substantially all assets of one, or subdividing an 4714 1.3 mrg organization, or merging organizations. If propagation of a 4715 1.3 mrg covered work results from an entity transaction, each party to that 4716 1.3 mrg transaction who receives a copy of the work also receives whatever 4717 1.3 mrg licenses to the work the party's predecessor in interest had or 4718 1.11 mrg could give under the previous paragraph, plus a right to possession 4719 1.11 mrg of the Corresponding Source of the work from the predecessor in 4720 1.11 mrg interest, if the predecessor has it or can get it with reasonable 4721 1.11 mrg efforts. 4722 1.3 mrg 4723 1.3 mrg You may not impose any further restrictions on the exercise of the 4724 1.3 mrg rights granted or affirmed under this License. For example, you 4725 1.11 mrg may not impose a license fee, royalty, or other charge for exercise 4726 1.11 mrg of rights granted under this License, and you may not initiate 4727 1.11 mrg litigation (including a cross-claim or counterclaim in a lawsuit) 4728 1.11 mrg alleging that any patent claim is infringed by making, using, 4729 1.11 mrg selling, offering for sale, or importing the Program or any portion 4730 1.11 mrg of it. 4731 1.3 mrg 4732 1.11 mrg 11. Patents. 4733 1.3 mrg 4734 1.3 mrg A "contributor" is a copyright holder who authorizes use under this 4735 1.3 mrg License of the Program or a work on which the Program is based. 4736 1.3 mrg The work thus licensed is called the contributor's "contributor 4737 1.3 mrg version". 4738 1.3 mrg 4739 1.3 mrg A contributor's "essential patent claims" are all patent claims 4740 1.3 mrg owned or controlled by the contributor, whether already acquired or 4741 1.3 mrg hereafter acquired, that would be infringed by some manner, 4742 1.3 mrg permitted by this License, of making, using, or selling its 4743 1.3 mrg contributor version, but do not include claims that would be 4744 1.3 mrg infringed only as a consequence of further modification of the 4745 1.3 mrg contributor version. For purposes of this definition, "control" 4746 1.3 mrg includes the right to grant patent sublicenses in a manner 4747 1.3 mrg consistent with the requirements of this License. 4748 1.3 mrg 4749 1.3 mrg Each contributor grants you a non-exclusive, worldwide, 4750 1.3 mrg royalty-free patent license under the contributor's essential 4751 1.3 mrg patent claims, to make, use, sell, offer for sale, import and 4752 1.11 mrg otherwise run, modify and propagate the contents of its contributor 4753 1.11 mrg version. 4754 1.3 mrg 4755 1.3 mrg In the following three paragraphs, a "patent license" is any 4756 1.3 mrg express agreement or commitment, however denominated, not to 4757 1.3 mrg enforce a patent (such as an express permission to practice a 4758 1.11 mrg patent or covenant not to sue for patent infringement). To "grant" 4759 1.11 mrg such a patent license to a party means to make such an agreement or 4760 1.11 mrg commitment not to enforce a patent against the party. 4761 1.3 mrg 4762 1.3 mrg If you convey a covered work, knowingly relying on a patent 4763 1.3 mrg license, and the Corresponding Source of the work is not available 4764 1.3 mrg for anyone to copy, free of charge and under the terms of this 4765 1.3 mrg License, through a publicly available network server or other 4766 1.3 mrg readily accessible means, then you must either (1) cause the 4767 1.3 mrg Corresponding Source to be so available, or (2) arrange to deprive 4768 1.3 mrg yourself of the benefit of the patent license for this particular 4769 1.3 mrg work, or (3) arrange, in a manner consistent with the requirements 4770 1.3 mrg of this License, to extend the patent license to downstream 4771 1.3 mrg recipients. "Knowingly relying" means you have actual knowledge 4772 1.3 mrg that, but for the patent license, your conveying the covered work 4773 1.3 mrg in a country, or your recipient's use of the covered work in a 4774 1.3 mrg country, would infringe one or more identifiable patents in that 4775 1.3 mrg country that you have reason to believe are valid. 4776 1.3 mrg 4777 1.3 mrg If, pursuant to or in connection with a single transaction or 4778 1.3 mrg arrangement, you convey, or propagate by procuring conveyance of, a 4779 1.3 mrg covered work, and grant a patent license to some of the parties 4780 1.3 mrg receiving the covered work authorizing them to use, propagate, 4781 1.3 mrg modify or convey a specific copy of the covered work, then the 4782 1.3 mrg patent license you grant is automatically extended to all 4783 1.3 mrg recipients of the covered work and works based on it. 4784 1.3 mrg 4785 1.3 mrg A patent license is "discriminatory" if it does not include within 4786 1.3 mrg the scope of its coverage, prohibits the exercise of, or is 4787 1.3 mrg conditioned on the non-exercise of one or more of the rights that 4788 1.3 mrg are specifically granted under this License. You may not convey a 4789 1.3 mrg covered work if you are a party to an arrangement with a third 4790 1.11 mrg party that is in the business of distributing software, under which 4791 1.11 mrg you make payment to the third party based on the extent of your 4792 1.11 mrg activity of conveying the work, and under which the third party 4793 1.11 mrg grants, to any of the parties who would receive the covered work 4794 1.11 mrg from you, a discriminatory patent license (a) in connection with 4795 1.11 mrg copies of the covered work conveyed by you (or copies made from 4796 1.11 mrg those copies), or (b) primarily for and in connection with specific 4797 1.11 mrg products or compilations that contain the covered work, unless you 4798 1.11 mrg entered into that arrangement, or that patent license was granted, 4799 1.11 mrg prior to 28 March 2007. 4800 1.3 mrg 4801 1.3 mrg Nothing in this License shall be construed as excluding or limiting 4802 1.3 mrg any implied license or other defenses to infringement that may 4803 1.3 mrg otherwise be available to you under applicable patent law. 4804 1.3 mrg 4805 1.11 mrg 12. No Surrender of Others' Freedom. 4806 1.3 mrg 4807 1.11 mrg If conditions are imposed on you (whether by court order, agreement 4808 1.11 mrg or otherwise) that contradict the conditions of this License, they 4809 1.11 mrg do not excuse you from the conditions of this License. If you 4810 1.11 mrg cannot convey a covered work so as to satisfy simultaneously your 4811 1.11 mrg obligations under this License and any other pertinent obligations, 4812 1.11 mrg then as a consequence you may not convey it at all. For example, 4813 1.11 mrg if you agree to terms that obligate you to collect a royalty for 4814 1.11 mrg further conveying from those to whom you convey the Program, the 4815 1.11 mrg only way you could satisfy both those terms and this License would 4816 1.11 mrg be to refrain entirely from conveying the Program. 4817 1.3 mrg 4818 1.11 mrg 13. Use with the GNU Affero General Public License. 4819 1.3 mrg 4820 1.3 mrg Notwithstanding any other provision of this License, you have 4821 1.3 mrg permission to link or combine any covered work with a work licensed 4822 1.3 mrg under version 3 of the GNU Affero General Public License into a 4823 1.3 mrg single combined work, and to convey the resulting work. The terms 4824 1.3 mrg of this License will continue to apply to the part which is the 4825 1.3 mrg covered work, but the special requirements of the GNU Affero 4826 1.3 mrg General Public License, section 13, concerning interaction through 4827 1.3 mrg a network will apply to the combination as such. 4828 1.3 mrg 4829 1.11 mrg 14. Revised Versions of this License. 4830 1.3 mrg 4831 1.3 mrg The Free Software Foundation may publish revised and/or new 4832 1.11 mrg versions of the GNU General Public License from time to time. Such 4833 1.11 mrg new versions will be similar in spirit to the present version, but 4834 1.11 mrg may differ in detail to address new problems or concerns. 4835 1.1 mrg 4836 1.1 mrg Each version is given a distinguishing version number. If the 4837 1.3 mrg Program specifies that a certain numbered version of the GNU 4838 1.3 mrg General Public License "or any later version" applies to it, you 4839 1.3 mrg have the option of following the terms and conditions either of 4840 1.11 mrg that numbered version or of any later version published by the Free 4841 1.11 mrg Software Foundation. If the Program does not specify a version 4842 1.11 mrg number of the GNU General Public License, you may choose any 4843 1.11 mrg version ever published by the Free Software Foundation. 4844 1.1 mrg 4845 1.3 mrg If the Program specifies that a proxy can decide which future 4846 1.3 mrg versions of the GNU General Public License can be used, that 4847 1.3 mrg proxy's public statement of acceptance of a version permanently 4848 1.3 mrg authorizes you to choose that version for the Program. 4849 1.3 mrg 4850 1.3 mrg Later license versions may give you additional or different 4851 1.3 mrg permissions. However, no additional obligations are imposed on any 4852 1.3 mrg author or copyright holder as a result of your choosing to follow a 4853 1.3 mrg later version. 4854 1.3 mrg 4855 1.11 mrg 15. Disclaimer of Warranty. 4856 1.3 mrg 4857 1.3 mrg THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 4858 1.11 mrg APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 4859 1.3 mrg COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 4860 1.3 mrg WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 4861 1.3 mrg INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 4862 1.11 mrg MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 4863 1.3 mrg RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 4864 1.3 mrg SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 4865 1.3 mrg NECESSARY SERVICING, REPAIR OR CORRECTION. 4866 1.3 mrg 4867 1.11 mrg 16. Limitation of Liability. 4868 1.3 mrg 4869 1.3 mrg IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 4870 1.3 mrg WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 4871 1.11 mrg AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR 4872 1.11 mrg DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 4873 1.3 mrg CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 4874 1.3 mrg THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 4875 1.3 mrg BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 4876 1.3 mrg PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 4877 1.3 mrg PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 4878 1.3 mrg THE POSSIBILITY OF SUCH DAMAGES. 4879 1.3 mrg 4880 1.11 mrg 17. Interpretation of Sections 15 and 16. 4881 1.3 mrg 4882 1.3 mrg If the disclaimer of warranty and limitation of liability provided 4883 1.3 mrg above cannot be given local legal effect according to their terms, 4884 1.3 mrg reviewing courts shall apply local law that most closely 4885 1.3 mrg approximates an absolute waiver of all civil liability in 4886 1.3 mrg connection with the Program, unless a warranty or assumption of 4887 1.3 mrg liability accompanies a copy of the Program in return for a fee. 4888 1.1 mrg 4889 1.3 mrg END OF TERMS AND CONDITIONS 4890 1.3 mrg =========================== 4891 1.3 mrg 4892 1.3 mrg How to Apply These Terms to Your New Programs 4893 1.3 mrg ============================================= 4894 1.1 mrg 4895 1.1 mrg If you develop a new program, and you want it to be of the greatest 4896 1.1 mrg possible use to the public, the best way to achieve this is to make it 4897 1.1 mrg free software which everyone can redistribute and change under these 4898 1.1 mrg terms. 4899 1.1 mrg 4900 1.1 mrg To do so, attach the following notices to the program. It is safest 4901 1.1 mrg to attach them to the start of each source file to most effectively 4902 1.3 mrg state the exclusion of warranty; and each file should have at least the 4903 1.3 mrg "copyright" line and a pointer to where the full notice is found. 4904 1.1 mrg 4905 1.1 mrg ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 4906 1.3 mrg Copyright (C) YEAR NAME OF AUTHOR 4907 1.1 mrg 4908 1.3 mrg This program is free software: you can redistribute it and/or modify 4909 1.1 mrg it under the terms of the GNU General Public License as published by 4910 1.3 mrg the Free Software Foundation, either version 3 of the License, or (at 4911 1.3 mrg your option) any later version. 4912 1.1 mrg 4913 1.3 mrg This program is distributed in the hope that it will be useful, but 4914 1.3 mrg WITHOUT ANY WARRANTY; without even the implied warranty of 4915 1.3 mrg MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 4916 1.3 mrg General Public License for more details. 4917 1.1 mrg 4918 1.1 mrg You should have received a copy of the GNU General Public License 4919 1.15 mrg along with this program. If not, see <https://www.gnu.org/licenses/>. 4920 1.1 mrg 4921 1.1 mrg Also add information on how to contact you by electronic and paper 4922 1.1 mrg mail. 4923 1.1 mrg 4924 1.3 mrg If the program does terminal interaction, make it output a short 4925 1.3 mrg notice like this when it starts in an interactive mode: 4926 1.1 mrg 4927 1.3 mrg PROGRAM Copyright (C) YEAR NAME OF AUTHOR 4928 1.11 mrg This program comes with ABSOLUTELY NO WARRANTY; for details type 'show w'. 4929 1.1 mrg This is free software, and you are welcome to redistribute it 4930 1.11 mrg under certain conditions; type 'show c' for details. 4931 1.1 mrg 4932 1.11 mrg The hypothetical commands 'show w' and 'show c' should show the 4933 1.3 mrg appropriate parts of the General Public License. Of course, your 4934 1.3 mrg program's commands might be different; for a GUI interface, you would 4935 1.3 mrg use an "about box". 4936 1.1 mrg 4937 1.1 mrg You should also get your employer (if you work as a programmer) or 4938 1.3 mrg school, if any, to sign a "copyright disclaimer" for the program, if 4939 1.3 mrg necessary. For more information on this, and how to apply and follow 4940 1.15 mrg the GNU GPL, see <https://www.gnu.org/licenses/>. 4941 1.1 mrg 4942 1.3 mrg The GNU General Public License does not permit incorporating your 4943 1.1 mrg program into proprietary programs. If your program is a subroutine 4944 1.1 mrg library, you may consider it more useful to permit linking proprietary 4945 1.1 mrg applications with the library. If this is what you want to do, use the 4946 1.3 mrg GNU Lesser General Public License instead of this License. But first, 4947 1.13 mrg please read <https://www.gnu.org/licenses/why-not-lgpl.html>. 4948 1.1 mrg 4949 1.1 mrg 4950 1.1 mrg File: libgomp.info, Node: GNU Free Documentation License, Next: Funding, Prev: Copying, Up: Top 4951 1.1 mrg 4952 1.1 mrg GNU Free Documentation License 4953 1.1 mrg ****************************** 4954 1.1 mrg 4955 1.3 mrg Version 1.3, 3 November 2008 4956 1.1 mrg 4957 1.3 mrg Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. 4958 1.15 mrg <https://fsf.org/> 4959 1.1 mrg 4960 1.1 mrg Everyone is permitted to copy and distribute verbatim copies 4961 1.1 mrg of this license document, but changing it is not allowed. 4962 1.1 mrg 4963 1.1 mrg 0. PREAMBLE 4964 1.1 mrg 4965 1.1 mrg The purpose of this License is to make a manual, textbook, or other 4966 1.1 mrg functional and useful document "free" in the sense of freedom: to 4967 1.1 mrg assure everyone the effective freedom to copy and redistribute it, 4968 1.1 mrg with or without modifying it, either commercially or 4969 1.1 mrg noncommercially. Secondarily, this License preserves for the 4970 1.1 mrg author and publisher a way to get credit for their work, while not 4971 1.1 mrg being considered responsible for modifications made by others. 4972 1.1 mrg 4973 1.1 mrg This License is a kind of "copyleft", which means that derivative 4974 1.1 mrg works of the document must themselves be free in the same sense. 4975 1.1 mrg It complements the GNU General Public License, which is a copyleft 4976 1.1 mrg license designed for free software. 4977 1.1 mrg 4978 1.1 mrg We have designed this License in order to use it for manuals for 4979 1.1 mrg free software, because free software needs free documentation: a 4980 1.1 mrg free program should come with manuals providing the same freedoms 4981 1.1 mrg that the software does. But this License is not limited to 4982 1.1 mrg software manuals; it can be used for any textual work, regardless 4983 1.11 mrg of subject matter or whether it is published as a printed book. We 4984 1.11 mrg recommend this License principally for works whose purpose is 4985 1.1 mrg instruction or reference. 4986 1.1 mrg 4987 1.1 mrg 1. APPLICABILITY AND DEFINITIONS 4988 1.1 mrg 4989 1.1 mrg This License applies to any manual or other work, in any medium, 4990 1.11 mrg that contains a notice placed by the copyright holder saying it can 4991 1.11 mrg be distributed under the terms of this License. Such a notice 4992 1.1 mrg grants a world-wide, royalty-free license, unlimited in duration, 4993 1.1 mrg to use that work under the conditions stated herein. The 4994 1.1 mrg "Document", below, refers to any such manual or work. Any member 4995 1.11 mrg of the public is a licensee, and is addressed as "you". You accept 4996 1.11 mrg the license if you copy, modify or distribute the work in a way 4997 1.11 mrg requiring permission under copyright law. 4998 1.1 mrg 4999 1.1 mrg A "Modified Version" of the Document means any work containing the 5000 1.1 mrg Document or a portion of it, either copied verbatim, or with 5001 1.1 mrg modifications and/or translated into another language. 5002 1.1 mrg 5003 1.1 mrg A "Secondary Section" is a named appendix or a front-matter section 5004 1.1 mrg of the Document that deals exclusively with the relationship of the 5005 1.1 mrg publishers or authors of the Document to the Document's overall 5006 1.1 mrg subject (or to related matters) and contains nothing that could 5007 1.1 mrg fall directly within that overall subject. (Thus, if the Document 5008 1.1 mrg is in part a textbook of mathematics, a Secondary Section may not 5009 1.1 mrg explain any mathematics.) The relationship could be a matter of 5010 1.1 mrg historical connection with the subject or with related matters, or 5011 1.1 mrg of legal, commercial, philosophical, ethical or political position 5012 1.1 mrg regarding them. 5013 1.1 mrg 5014 1.1 mrg The "Invariant Sections" are certain Secondary Sections whose 5015 1.11 mrg titles are designated, as being those of Invariant Sections, in the 5016 1.11 mrg notice that says that the Document is released under this License. 5017 1.11 mrg If a section does not fit the above definition of Secondary then it 5018 1.11 mrg is not allowed to be designated as Invariant. The Document may 5019 1.11 mrg contain zero Invariant Sections. If the Document does not identify 5020 1.11 mrg any Invariant Sections then there are none. 5021 1.1 mrg 5022 1.1 mrg The "Cover Texts" are certain short passages of text that are 5023 1.1 mrg listed, as Front-Cover Texts or Back-Cover Texts, in the notice 5024 1.1 mrg that says that the Document is released under this License. A 5025 1.1 mrg Front-Cover Text may be at most 5 words, and a Back-Cover Text may 5026 1.1 mrg be at most 25 words. 5027 1.1 mrg 5028 1.1 mrg A "Transparent" copy of the Document means a machine-readable copy, 5029 1.1 mrg represented in a format whose specification is available to the 5030 1.1 mrg general public, that is suitable for revising the document 5031 1.11 mrg straightforwardly with generic text editors or (for images composed 5032 1.11 mrg of pixels) generic paint programs or (for drawings) some widely 5033 1.11 mrg available drawing editor, and that is suitable for input to text 5034 1.11 mrg formatters or for automatic translation to a variety of formats 5035 1.11 mrg suitable for input to text formatters. A copy made in an otherwise 5036 1.11 mrg Transparent file format whose markup, or absence of markup, has 5037 1.11 mrg been arranged to thwart or discourage subsequent modification by 5038 1.11 mrg readers is not Transparent. An image format is not Transparent if 5039 1.11 mrg used for any substantial amount of text. A copy that is not 5040 1.11 mrg "Transparent" is called "Opaque". 5041 1.1 mrg 5042 1.1 mrg Examples of suitable formats for Transparent copies include plain 5043 1.1 mrg ASCII without markup, Texinfo input format, LaTeX input format, 5044 1.11 mrg SGML or XML using a publicly available DTD, and standard-conforming 5045 1.11 mrg simple HTML, PostScript or PDF designed for human modification. 5046 1.11 mrg Examples of transparent image formats include PNG, XCF and JPG. 5047 1.11 mrg Opaque formats include proprietary formats that can be read and 5048 1.11 mrg edited only by proprietary word processors, SGML or XML for which 5049 1.11 mrg the DTD and/or processing tools are not generally available, and 5050 1.11 mrg the machine-generated HTML, PostScript or PDF produced by some word 5051 1.11 mrg processors for output purposes only. 5052 1.1 mrg 5053 1.1 mrg The "Title Page" means, for a printed book, the title page itself, 5054 1.1 mrg plus such following pages as are needed to hold, legibly, the 5055 1.1 mrg material this License requires to appear in the title page. For 5056 1.1 mrg works in formats which do not have any title page as such, "Title 5057 1.1 mrg Page" means the text near the most prominent appearance of the 5058 1.1 mrg work's title, preceding the beginning of the body of the text. 5059 1.1 mrg 5060 1.3 mrg The "publisher" means any person or entity that distributes copies 5061 1.3 mrg of the Document to the public. 5062 1.3 mrg 5063 1.1 mrg A section "Entitled XYZ" means a named subunit of the Document 5064 1.1 mrg whose title either is precisely XYZ or contains XYZ in parentheses 5065 1.1 mrg following text that translates XYZ in another language. (Here XYZ 5066 1.1 mrg stands for a specific section name mentioned below, such as 5067 1.1 mrg "Acknowledgements", "Dedications", "Endorsements", or "History".) 5068 1.1 mrg To "Preserve the Title" of such a section when you modify the 5069 1.1 mrg Document means that it remains a section "Entitled XYZ" according 5070 1.1 mrg to this definition. 5071 1.1 mrg 5072 1.1 mrg The Document may include Warranty Disclaimers next to the notice 5073 1.1 mrg which states that this License applies to the Document. These 5074 1.1 mrg Warranty Disclaimers are considered to be included by reference in 5075 1.1 mrg this License, but only as regards disclaiming warranties: any other 5076 1.1 mrg implication that these Warranty Disclaimers may have is void and 5077 1.1 mrg has no effect on the meaning of this License. 5078 1.1 mrg 5079 1.1 mrg 2. VERBATIM COPYING 5080 1.1 mrg 5081 1.1 mrg You may copy and distribute the Document in any medium, either 5082 1.1 mrg commercially or noncommercially, provided that this License, the 5083 1.1 mrg copyright notices, and the license notice saying this License 5084 1.1 mrg applies to the Document are reproduced in all copies, and that you 5085 1.1 mrg add no other conditions whatsoever to those of this License. You 5086 1.1 mrg may not use technical measures to obstruct or control the reading 5087 1.1 mrg or further copying of the copies you make or distribute. However, 5088 1.1 mrg you may accept compensation in exchange for copies. If you 5089 1.11 mrg distribute a large enough number of copies you must also follow the 5090 1.11 mrg conditions in section 3. 5091 1.1 mrg 5092 1.1 mrg You may also lend copies, under the same conditions stated above, 5093 1.1 mrg and you may publicly display copies. 5094 1.1 mrg 5095 1.1 mrg 3. COPYING IN QUANTITY 5096 1.1 mrg 5097 1.1 mrg If you publish printed copies (or copies in media that commonly 5098 1.1 mrg have printed covers) of the Document, numbering more than 100, and 5099 1.1 mrg the Document's license notice requires Cover Texts, you must 5100 1.1 mrg enclose the copies in covers that carry, clearly and legibly, all 5101 1.1 mrg these Cover Texts: Front-Cover Texts on the front cover, and 5102 1.1 mrg Back-Cover Texts on the back cover. Both covers must also clearly 5103 1.1 mrg and legibly identify you as the publisher of these copies. The 5104 1.11 mrg front cover must present the full title with all words of the title 5105 1.11 mrg equally prominent and visible. You may add other material on the 5106 1.11 mrg covers in addition. Copying with changes limited to the covers, as 5107 1.11 mrg long as they preserve the title of the Document and satisfy these 5108 1.11 mrg conditions, can be treated as verbatim copying in other respects. 5109 1.1 mrg 5110 1.1 mrg If the required texts for either cover are too voluminous to fit 5111 1.1 mrg legibly, you should put the first ones listed (as many as fit 5112 1.1 mrg reasonably) on the actual cover, and continue the rest onto 5113 1.1 mrg adjacent pages. 5114 1.1 mrg 5115 1.1 mrg If you publish or distribute Opaque copies of the Document 5116 1.11 mrg numbering more than 100, you must either include a machine-readable 5117 1.11 mrg Transparent copy along with each Opaque copy, or state in or with 5118 1.11 mrg each Opaque copy a computer-network location from which the general 5119 1.11 mrg network-using public has access to download using public-standard 5120 1.11 mrg network protocols a complete Transparent copy of the Document, free 5121 1.11 mrg of added material. If you use the latter option, you must take 5122 1.11 mrg reasonably prudent steps, when you begin distribution of Opaque 5123 1.11 mrg copies in quantity, to ensure that this Transparent copy will 5124 1.11 mrg remain thus accessible at the stated location until at least one 5125 1.11 mrg year after the last time you distribute an Opaque copy (directly or 5126 1.11 mrg through your agents or retailers) of that edition to the public. 5127 1.1 mrg 5128 1.1 mrg It is requested, but not required, that you contact the authors of 5129 1.11 mrg the Document well before redistributing any large number of copies, 5130 1.11 mrg to give them a chance to provide you with an updated version of the 5131 1.11 mrg Document. 5132 1.1 mrg 5133 1.1 mrg 4. MODIFICATIONS 5134 1.1 mrg 5135 1.1 mrg You may copy and distribute a Modified Version of the Document 5136 1.1 mrg under the conditions of sections 2 and 3 above, provided that you 5137 1.11 mrg release the Modified Version under precisely this License, with the 5138 1.11 mrg Modified Version filling the role of the Document, thus licensing 5139 1.11 mrg distribution and modification of the Modified Version to whoever 5140 1.11 mrg possesses a copy of it. In addition, you must do these things in 5141 1.11 mrg the Modified Version: 5142 1.1 mrg 5143 1.1 mrg A. Use in the Title Page (and on the covers, if any) a title 5144 1.11 mrg distinct from that of the Document, and from those of previous 5145 1.11 mrg versions (which should, if there were any, be listed in the 5146 1.11 mrg History section of the Document). You may use the same title 5147 1.11 mrg as a previous version if the original publisher of that 5148 1.11 mrg version gives permission. 5149 1.1 mrg 5150 1.1 mrg B. List on the Title Page, as authors, one or more persons or 5151 1.1 mrg entities responsible for authorship of the modifications in 5152 1.1 mrg the Modified Version, together with at least five of the 5153 1.1 mrg principal authors of the Document (all of its principal 5154 1.1 mrg authors, if it has fewer than five), unless they release you 5155 1.1 mrg from this requirement. 5156 1.1 mrg 5157 1.1 mrg C. State on the Title page the name of the publisher of the 5158 1.1 mrg Modified Version, as the publisher. 5159 1.1 mrg 5160 1.1 mrg D. Preserve all the copyright notices of the Document. 5161 1.1 mrg 5162 1.1 mrg E. Add an appropriate copyright notice for your modifications 5163 1.1 mrg adjacent to the other copyright notices. 5164 1.1 mrg 5165 1.1 mrg F. Include, immediately after the copyright notices, a license 5166 1.1 mrg notice giving the public permission to use the Modified 5167 1.1 mrg Version under the terms of this License, in the form shown in 5168 1.1 mrg the Addendum below. 5169 1.1 mrg 5170 1.1 mrg G. Preserve in that license notice the full lists of Invariant 5171 1.1 mrg Sections and required Cover Texts given in the Document's 5172 1.1 mrg license notice. 5173 1.1 mrg 5174 1.1 mrg H. Include an unaltered copy of this License. 5175 1.1 mrg 5176 1.1 mrg I. Preserve the section Entitled "History", Preserve its Title, 5177 1.1 mrg and add to it an item stating at least the title, year, new 5178 1.11 mrg authors, and publisher of the Modified Version as given on the 5179 1.11 mrg Title Page. If there is no section Entitled "History" in the 5180 1.11 mrg Document, create one stating the title, year, authors, and 5181 1.11 mrg publisher of the Document as given on its Title Page, then add 5182 1.11 mrg an item describing the Modified Version as stated in the 5183 1.11 mrg previous sentence. 5184 1.1 mrg 5185 1.1 mrg J. Preserve the network location, if any, given in the Document 5186 1.1 mrg for public access to a Transparent copy of the Document, and 5187 1.1 mrg likewise the network locations given in the Document for 5188 1.11 mrg previous versions it was based on. These may be placed in the 5189 1.11 mrg "History" section. You may omit a network location for a work 5190 1.11 mrg that was published at least four years before the Document 5191 1.11 mrg itself, or if the original publisher of the version it refers 5192 1.11 mrg to gives permission. 5193 1.1 mrg 5194 1.1 mrg K. For any section Entitled "Acknowledgements" or "Dedications", 5195 1.11 mrg Preserve the Title of the section, and preserve in the section 5196 1.11 mrg all the substance and tone of each of the contributor 5197 1.1 mrg acknowledgements and/or dedications given therein. 5198 1.1 mrg 5199 1.11 mrg L. Preserve all the Invariant Sections of the Document, unaltered 5200 1.11 mrg in their text and in their titles. Section numbers or the 5201 1.11 mrg equivalent are not considered part of the section titles. 5202 1.1 mrg 5203 1.1 mrg M. Delete any section Entitled "Endorsements". Such a section 5204 1.1 mrg may not be included in the Modified Version. 5205 1.1 mrg 5206 1.1 mrg N. Do not retitle any existing section to be Entitled 5207 1.1 mrg "Endorsements" or to conflict in title with any Invariant 5208 1.1 mrg Section. 5209 1.1 mrg 5210 1.1 mrg O. Preserve any Warranty Disclaimers. 5211 1.1 mrg 5212 1.1 mrg If the Modified Version includes new front-matter sections or 5213 1.1 mrg appendices that qualify as Secondary Sections and contain no 5214 1.11 mrg material copied from the Document, you may at your option designate 5215 1.11 mrg some or all of these sections as invariant. To do this, add their 5216 1.11 mrg titles to the list of Invariant Sections in the Modified Version's 5217 1.11 mrg license notice. These titles must be distinct from any other 5218 1.11 mrg section titles. 5219 1.1 mrg 5220 1.1 mrg You may add a section Entitled "Endorsements", provided it contains 5221 1.1 mrg nothing but endorsements of your Modified Version by various 5222 1.1 mrg parties--for example, statements of peer review or that the text 5223 1.1 mrg has been approved by an organization as the authoritative 5224 1.1 mrg definition of a standard. 5225 1.1 mrg 5226 1.1 mrg You may add a passage of up to five words as a Front-Cover Text, 5227 1.11 mrg and a passage of up to 25 words as a Back-Cover Text, to the end of 5228 1.11 mrg the list of Cover Texts in the Modified Version. Only one passage 5229 1.11 mrg of Front-Cover Text and one of Back-Cover Text may be added by (or 5230 1.11 mrg through arrangements made by) any one entity. If the Document 5231 1.11 mrg already includes a cover text for the same cover, previously added 5232 1.11 mrg by you or by arrangement made by the same entity you are acting on 5233 1.11 mrg behalf of, you may not add another; but you may replace the old 5234 1.11 mrg one, on explicit permission from the previous publisher that added 5235 1.11 mrg the old one. 5236 1.1 mrg 5237 1.1 mrg The author(s) and publisher(s) of the Document do not by this 5238 1.1 mrg License give permission to use their names for publicity for or to 5239 1.1 mrg assert or imply endorsement of any Modified Version. 5240 1.1 mrg 5241 1.1 mrg 5. COMBINING DOCUMENTS 5242 1.1 mrg 5243 1.1 mrg You may combine the Document with other documents released under 5244 1.1 mrg this License, under the terms defined in section 4 above for 5245 1.11 mrg modified versions, provided that you include in the combination all 5246 1.11 mrg of the Invariant Sections of all of the original documents, 5247 1.1 mrg unmodified, and list them all as Invariant Sections of your 5248 1.1 mrg combined work in its license notice, and that you preserve all 5249 1.1 mrg their Warranty Disclaimers. 5250 1.1 mrg 5251 1.1 mrg The combined work need only contain one copy of this License, and 5252 1.1 mrg multiple identical Invariant Sections may be replaced with a single 5253 1.1 mrg copy. If there are multiple Invariant Sections with the same name 5254 1.1 mrg but different contents, make the title of each such section unique 5255 1.1 mrg by adding at the end of it, in parentheses, the name of the 5256 1.1 mrg original author or publisher of that section if known, or else a 5257 1.1 mrg unique number. Make the same adjustment to the section titles in 5258 1.1 mrg the list of Invariant Sections in the license notice of the 5259 1.1 mrg combined work. 5260 1.1 mrg 5261 1.1 mrg In the combination, you must combine any sections Entitled 5262 1.1 mrg "History" in the various original documents, forming one section 5263 1.1 mrg Entitled "History"; likewise combine any sections Entitled 5264 1.1 mrg "Acknowledgements", and any sections Entitled "Dedications". You 5265 1.1 mrg must delete all sections Entitled "Endorsements." 5266 1.1 mrg 5267 1.1 mrg 6. COLLECTIONS OF DOCUMENTS 5268 1.1 mrg 5269 1.1 mrg You may make a collection consisting of the Document and other 5270 1.1 mrg documents released under this License, and replace the individual 5271 1.1 mrg copies of this License in the various documents with a single copy 5272 1.1 mrg that is included in the collection, provided that you follow the 5273 1.11 mrg rules of this License for verbatim copying of each of the documents 5274 1.11 mrg in all other respects. 5275 1.1 mrg 5276 1.1 mrg You may extract a single document from such a collection, and 5277 1.1 mrg distribute it individually under this License, provided you insert 5278 1.11 mrg a copy of this License into the extracted document, and follow this 5279 1.11 mrg License in all other respects regarding verbatim copying of that 5280 1.11 mrg document. 5281 1.1 mrg 5282 1.1 mrg 7. AGGREGATION WITH INDEPENDENT WORKS 5283 1.1 mrg 5284 1.1 mrg A compilation of the Document or its derivatives with other 5285 1.11 mrg separate and independent documents or works, in or on a volume of a 5286 1.11 mrg storage or distribution medium, is called an "aggregate" if the 5287 1.1 mrg copyright resulting from the compilation is not used to limit the 5288 1.1 mrg legal rights of the compilation's users beyond what the individual 5289 1.1 mrg works permit. When the Document is included in an aggregate, this 5290 1.1 mrg License does not apply to the other works in the aggregate which 5291 1.1 mrg are not themselves derivative works of the Document. 5292 1.1 mrg 5293 1.1 mrg If the Cover Text requirement of section 3 is applicable to these 5294 1.1 mrg copies of the Document, then if the Document is less than one half 5295 1.1 mrg of the entire aggregate, the Document's Cover Texts may be placed 5296 1.1 mrg on covers that bracket the Document within the aggregate, or the 5297 1.1 mrg electronic equivalent of covers if the Document is in electronic 5298 1.1 mrg form. Otherwise they must appear on printed covers that bracket 5299 1.1 mrg the whole aggregate. 5300 1.1 mrg 5301 1.1 mrg 8. TRANSLATION 5302 1.1 mrg 5303 1.1 mrg Translation is considered a kind of modification, so you may 5304 1.1 mrg distribute translations of the Document under the terms of section 5305 1.1 mrg 4. Replacing Invariant Sections with translations requires special 5306 1.1 mrg permission from their copyright holders, but you may include 5307 1.1 mrg translations of some or all Invariant Sections in addition to the 5308 1.1 mrg original versions of these Invariant Sections. You may include a 5309 1.1 mrg translation of this License, and all the license notices in the 5310 1.1 mrg Document, and any Warranty Disclaimers, provided that you also 5311 1.1 mrg include the original English version of this License and the 5312 1.1 mrg original versions of those notices and disclaimers. In case of a 5313 1.1 mrg disagreement between the translation and the original version of 5314 1.1 mrg this License or a notice or disclaimer, the original version will 5315 1.1 mrg prevail. 5316 1.1 mrg 5317 1.1 mrg If a section in the Document is Entitled "Acknowledgements", 5318 1.1 mrg "Dedications", or "History", the requirement (section 4) to 5319 1.1 mrg Preserve its Title (section 1) will typically require changing the 5320 1.1 mrg actual title. 5321 1.1 mrg 5322 1.1 mrg 9. TERMINATION 5323 1.1 mrg 5324 1.1 mrg You may not copy, modify, sublicense, or distribute the Document 5325 1.3 mrg except as expressly provided under this License. Any attempt 5326 1.3 mrg otherwise to copy, modify, sublicense, or distribute it is void, 5327 1.3 mrg and will automatically terminate your rights under this License. 5328 1.3 mrg 5329 1.3 mrg However, if you cease all violation of this License, then your 5330 1.3 mrg license from a particular copyright holder is reinstated (a) 5331 1.11 mrg provisionally, unless and until the copyright holder explicitly and 5332 1.11 mrg finally terminates your license, and (b) permanently, if the 5333 1.3 mrg copyright holder fails to notify you of the violation by some 5334 1.3 mrg reasonable means prior to 60 days after the cessation. 5335 1.3 mrg 5336 1.3 mrg Moreover, your license from a particular copyright holder is 5337 1.3 mrg reinstated permanently if the copyright holder notifies you of the 5338 1.3 mrg violation by some reasonable means, this is the first time you have 5339 1.3 mrg received notice of violation of this License (for any work) from 5340 1.3 mrg that copyright holder, and you cure the violation prior to 30 days 5341 1.3 mrg after your receipt of the notice. 5342 1.3 mrg 5343 1.3 mrg Termination of your rights under this section does not terminate 5344 1.11 mrg the licenses of parties who have received copies or rights from you 5345 1.11 mrg under this License. If your rights have been terminated and not 5346 1.11 mrg permanently reinstated, receipt of a copy of some or all of the 5347 1.11 mrg same material does not give you any rights to use it. 5348 1.1 mrg 5349 1.11 mrg 10. FUTURE REVISIONS OF THIS LICENSE 5350 1.1 mrg 5351 1.1 mrg The Free Software Foundation may publish new, revised versions of 5352 1.1 mrg the GNU Free Documentation License from time to time. Such new 5353 1.1 mrg versions will be similar in spirit to the present version, but may 5354 1.1 mrg differ in detail to address new problems or concerns. See 5355 1.15 mrg <https://www.gnu.org/copyleft/>. 5356 1.1 mrg 5357 1.1 mrg Each version of the License is given a distinguishing version 5358 1.1 mrg number. If the Document specifies that a particular numbered 5359 1.1 mrg version of this License "or any later version" applies to it, you 5360 1.1 mrg have the option of following the terms and conditions either of 5361 1.1 mrg that specified version or of any later version that has been 5362 1.11 mrg published (not as a draft) by the Free Software Foundation. If the 5363 1.11 mrg Document does not specify a version number of this License, you may 5364 1.11 mrg choose any version ever published (not as a draft) by the Free 5365 1.11 mrg Software Foundation. If the Document specifies that a proxy can 5366 1.11 mrg decide which future versions of this License can be used, that 5367 1.3 mrg proxy's public statement of acceptance of a version permanently 5368 1.3 mrg authorizes you to choose that version for the Document. 5369 1.3 mrg 5370 1.11 mrg 11. RELICENSING 5371 1.3 mrg 5372 1.3 mrg "Massive Multiauthor Collaboration Site" (or "MMC Site") means any 5373 1.3 mrg World Wide Web server that publishes copyrightable works and also 5374 1.3 mrg provides prominent facilities for anybody to edit those works. A 5375 1.3 mrg public wiki that anybody can edit is an example of such a server. 5376 1.3 mrg A "Massive Multiauthor Collaboration" (or "MMC") contained in the 5377 1.3 mrg site means any set of copyrightable works thus published on the MMC 5378 1.3 mrg site. 5379 1.3 mrg 5380 1.3 mrg "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 5381 1.3 mrg license published by Creative Commons Corporation, a not-for-profit 5382 1.3 mrg corporation with a principal place of business in San Francisco, 5383 1.3 mrg California, as well as future copyleft versions of that license 5384 1.3 mrg published by that same organization. 5385 1.3 mrg 5386 1.3 mrg "Incorporate" means to publish or republish a Document, in whole or 5387 1.3 mrg in part, as part of another Document. 5388 1.3 mrg 5389 1.3 mrg An MMC is "eligible for relicensing" if it is licensed under this 5390 1.3 mrg License, and if all works that were first published under this 5391 1.3 mrg License somewhere other than this MMC, and subsequently 5392 1.3 mrg incorporated in whole or in part into the MMC, (1) had no cover 5393 1.3 mrg texts or invariant sections, and (2) were thus incorporated prior 5394 1.3 mrg to November 1, 2008. 5395 1.3 mrg 5396 1.3 mrg The operator of an MMC Site may republish an MMC contained in the 5397 1.3 mrg site under CC-BY-SA on the same site at any time before August 1, 5398 1.3 mrg 2009, provided the MMC is eligible for relicensing. 5399 1.3 mrg 5400 1.1 mrg ADDENDUM: How to use this License for your documents 5401 1.1 mrg ==================================================== 5402 1.1 mrg 5403 1.1 mrg To use this License in a document you have written, include a copy of 5404 1.1 mrg the License in the document and put the following copyright and license 5405 1.1 mrg notices just after the title page: 5406 1.1 mrg 5407 1.1 mrg Copyright (C) YEAR YOUR NAME. 5408 1.1 mrg Permission is granted to copy, distribute and/or modify this document 5409 1.3 mrg under the terms of the GNU Free Documentation License, Version 1.3 5410 1.1 mrg or any later version published by the Free Software Foundation; 5411 1.1 mrg with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 5412 1.1 mrg Texts. A copy of the license is included in the section entitled ``GNU 5413 1.1 mrg Free Documentation License''. 5414 1.1 mrg 5415 1.1 mrg If you have Invariant Sections, Front-Cover Texts and Back-Cover 5416 1.11 mrg Texts, replace the "with...Texts." line with this: 5417 1.1 mrg 5418 1.1 mrg with the Invariant Sections being LIST THEIR TITLES, with 5419 1.1 mrg the Front-Cover Texts being LIST, and with the Back-Cover Texts 5420 1.1 mrg being LIST. 5421 1.1 mrg 5422 1.1 mrg If you have Invariant Sections without Cover Texts, or some other 5423 1.1 mrg combination of the three, merge those two alternatives to suit the 5424 1.1 mrg situation. 5425 1.1 mrg 5426 1.1 mrg If your document contains nontrivial examples of program code, we 5427 1.11 mrg recommend releasing these examples in parallel under your choice of free 5428 1.11 mrg software license, such as the GNU General Public License, to permit 5429 1.11 mrg their use in free software. 5430 1.1 mrg 5431 1.1 mrg 5432 1.3 mrg File: libgomp.info, Node: Funding, Next: Library Index, Prev: GNU Free Documentation License, Up: Top 5433 1.1 mrg 5434 1.1 mrg Funding Free Software 5435 1.1 mrg ********************* 5436 1.1 mrg 5437 1.1 mrg If you want to have more free software a few years from now, it makes 5438 1.1 mrg sense for you to help encourage people to contribute funds for its 5439 1.1 mrg development. The most effective approach known is to encourage 5440 1.1 mrg commercial redistributors to donate. 5441 1.1 mrg 5442 1.1 mrg Users of free software systems can boost the pace of development by 5443 1.1 mrg encouraging for-a-fee distributors to donate part of their selling price 5444 1.1 mrg to free software developers--the Free Software Foundation, and others. 5445 1.1 mrg 5446 1.1 mrg The way to convince distributors to do this is to demand it and 5447 1.1 mrg expect it from them. So when you compare distributors, judge them 5448 1.1 mrg partly by how much they give to free software development. Show 5449 1.1 mrg distributors they must compete to be the one who gives the most. 5450 1.1 mrg 5451 1.1 mrg To make this approach work, you must insist on numbers that you can 5452 1.1 mrg compare, such as, "We will donate ten dollars to the Frobnitz project 5453 1.1 mrg for each disk sold." Don't be satisfied with a vague promise, such as 5454 1.1 mrg "A portion of the profits are donated," since it doesn't give a basis 5455 1.1 mrg for comparison. 5456 1.1 mrg 5457 1.1 mrg Even a precise fraction "of the profits from this disk" is not very 5458 1.1 mrg meaningful, since creative accounting and unrelated business decisions 5459 1.11 mrg can greatly alter what fraction of the sales price counts as profit. If 5460 1.11 mrg the price you pay is $50, ten percent of the profit is probably less 5461 1.1 mrg than a dollar; it might be a few cents, or nothing at all. 5462 1.1 mrg 5463 1.1 mrg Some redistributors do development work themselves. This is useful 5464 1.1 mrg too; but to keep everyone honest, you need to inquire how much they do, 5465 1.1 mrg and what kind. Some kinds of development make much more long-term 5466 1.1 mrg difference than others. For example, maintaining a separate version of 5467 1.1 mrg a program contributes very little; maintaining the standard version of a 5468 1.1 mrg program for the whole community contributes much. Easy new ports 5469 1.1 mrg contribute little, since someone else would surely do them; difficult 5470 1.11 mrg ports such as adding a new CPU to the GNU Compiler Collection contribute 5471 1.11 mrg more; major new features or packages contribute the most. 5472 1.1 mrg 5473 1.1 mrg By establishing the idea that supporting further development is "the 5474 1.1 mrg proper thing to do" when distributing free software for a fee, we can 5475 1.1 mrg assure a steady flow of resources into making more free software. 5476 1.1 mrg 5477 1.1 mrg Copyright (C) 1994 Free Software Foundation, Inc. 5478 1.1 mrg Verbatim copying and redistribution of this section is permitted 5479 1.1 mrg without royalty; alteration is not permitted. 5480 1.1 mrg 5481 1.1 mrg 5482 1.3 mrg File: libgomp.info, Node: Library Index, Prev: Funding, Up: Top 5483 1.1 mrg 5484 1.3 mrg Library Index 5485 1.3 mrg ************* 5486 1.1 mrg 5487 1.1 mrg [index] 5488 1.1 mrg * Menu: 5489 1.1 mrg 5490 1.13 mrg * acc_get_property: acc_get_property. (line 6) 5491 1.13 mrg * acc_get_property_string: acc_get_property. (line 6) 5492 1.11 mrg * Environment Variable: OMP_CANCELLATION. (line 6) 5493 1.11 mrg * Environment Variable <1>: OMP_DISPLAY_ENV. (line 6) 5494 1.11 mrg * Environment Variable <2>: OMP_DEFAULT_DEVICE. (line 6) 5495 1.11 mrg * Environment Variable <3>: OMP_DYNAMIC. (line 6) 5496 1.11 mrg * Environment Variable <4>: OMP_MAX_ACTIVE_LEVELS. (line 6) 5497 1.11 mrg * Environment Variable <5>: OMP_MAX_TASK_PRIORITY. (line 6) 5498 1.11 mrg * Environment Variable <6>: OMP_NESTED. (line 6) 5499 1.15 mrg * Environment Variable <7>: OMP_NUM_TEAMS. (line 6) 5500 1.15 mrg * Environment Variable <8>: OMP_NUM_THREADS. (line 6) 5501 1.15 mrg * Environment Variable <9>: OMP_PROC_BIND. (line 6) 5502 1.15 mrg * Environment Variable <10>: OMP_PLACES. (line 6) 5503 1.15 mrg * Environment Variable <11>: OMP_STACKSIZE. (line 6) 5504 1.15 mrg * Environment Variable <12>: OMP_SCHEDULE. (line 6) 5505 1.15 mrg * Environment Variable <13>: OMP_TARGET_OFFLOAD. (line 6) 5506 1.15 mrg * Environment Variable <14>: OMP_TEAMS_THREAD_LIMIT. 5507 1.15 mrg (line 6) 5508 1.15 mrg * Environment Variable <15>: OMP_THREAD_LIMIT. (line 6) 5509 1.15 mrg * Environment Variable <16>: OMP_WAIT_POLICY. (line 6) 5510 1.15 mrg * Environment Variable <17>: GOMP_CPU_AFFINITY. (line 6) 5511 1.15 mrg * Environment Variable <18>: GOMP_DEBUG. (line 6) 5512 1.15 mrg * Environment Variable <19>: GOMP_STACKSIZE. (line 6) 5513 1.15 mrg * Environment Variable <20>: GOMP_SPINCOUNT. (line 6) 5514 1.15 mrg * Environment Variable <21>: GOMP_RTEMS_THREAD_POOLS. 5515 1.10 mrg (line 6) 5516 1.10 mrg * FDL, GNU Free Documentation License: GNU Free Documentation License. 5517 1.7 mrg (line 6) 5518 1.11 mrg * Implementation specific setting: OMP_NESTED. (line 6) 5519 1.11 mrg * Implementation specific setting <1>: OMP_NUM_THREADS. (line 6) 5520 1.11 mrg * Implementation specific setting <2>: OMP_SCHEDULE. (line 6) 5521 1.15 mrg * Implementation specific setting <3>: OMP_TARGET_OFFLOAD. (line 6) 5522 1.15 mrg * Implementation specific setting <4>: GOMP_STACKSIZE. (line 6) 5523 1.15 mrg * Implementation specific setting <5>: GOMP_SPINCOUNT. (line 6) 5524 1.15 mrg * Implementation specific setting <6>: GOMP_RTEMS_THREAD_POOLS. 5525 1.1 mrg (line 6) 5526 1.1 mrg * Introduction: Top. (line 6) 5527 1.1 mrg 5528 1.1 mrg 5529 1.1 mrg 5530 1.1 mrg Tag Table: 5531 1.11 mrg Node: Top2083 5532 1.15 mrg Node: Enabling OpenMP4713 5533 1.15 mrg Node: OpenMP Implementation Status5519 5534 1.15 mrg Node: OpenMP 4.56074 5535 1.15 mrg Node: OpenMP 5.06250 5536 1.15 mrg Node: OpenMP 5.111120 5537 1.15 mrg Node: Runtime Library Routines14711 5538 1.15 mrg Node: omp_get_active_level18416 5539 1.15 mrg Node: omp_get_ancestor_thread_num19116 5540 1.15 mrg Node: omp_get_cancellation20046 5541 1.15 mrg Node: omp_get_default_device20860 5542 1.15 mrg Node: omp_get_device_num21539 5543 1.15 mrg Node: omp_get_dynamic22344 5544 1.15 mrg Node: omp_get_initial_device23224 5545 1.15 mrg Node: omp_get_level23968 5546 1.15 mrg Node: omp_get_max_active_levels24595 5547 1.15 mrg Node: omp_get_max_task_priority25316 5548 1.15 mrg Node: omp_get_max_teams25936 5549 1.15 mrg Node: omp_get_max_threads26630 5550 1.15 mrg Node: omp_get_nested27381 5551 1.15 mrg Node: omp_get_num_devices28989 5552 1.15 mrg Node: omp_get_num_procs29510 5553 1.15 mrg Node: omp_get_num_teams30049 5554 1.15 mrg Node: omp_get_num_threads30565 5555 1.15 mrg Node: omp_get_proc_bind31654 5556 1.15 mrg Node: omp_get_schedule32677 5557 1.15 mrg Node: omp_get_supported_active_levels33646 5558 1.15 mrg Node: omp_get_team_num34432 5559 1.15 mrg Node: omp_get_team_size34946 5560 1.15 mrg Node: omp_get_teams_thread_limit35912 5561 1.15 mrg Node: omp_get_thread_limit36674 5562 1.15 mrg Node: omp_get_thread_num37302 5563 1.15 mrg Node: omp_in_parallel38174 5564 1.15 mrg Node: omp_in_final38823 5565 1.15 mrg Node: omp_is_initial_device39497 5566 1.15 mrg Node: omp_set_default_device40190 5567 1.15 mrg Node: omp_set_dynamic40981 5568 1.15 mrg Node: omp_set_max_active_levels41867 5569 1.15 mrg Node: omp_set_nested42789 5570 1.15 mrg Node: omp_set_num_teams43984 5571 1.15 mrg Node: omp_set_num_threads44857 5572 1.15 mrg Node: omp_set_schedule45728 5573 1.15 mrg Node: omp_set_teams_thread_limit46822 5574 1.15 mrg Node: omp_init_lock47824 5575 1.15 mrg Node: omp_set_lock48487 5576 1.15 mrg Node: omp_test_lock49342 5577 1.15 mrg Node: omp_unset_lock50318 5578 1.15 mrg Node: omp_destroy_lock51249 5579 1.15 mrg Node: omp_init_nest_lock51926 5580 1.15 mrg Node: omp_set_nest_lock52661 5581 1.15 mrg Node: omp_test_nest_lock53576 5582 1.15 mrg Node: omp_unset_nest_lock54603 5583 1.15 mrg Node: omp_destroy_nest_lock55618 5584 1.15 mrg Node: omp_get_wtick56369 5585 1.15 mrg Node: omp_get_wtime56961 5586 1.15 mrg Node: omp_fulfill_event57763 5587 1.15 mrg Node: Environment Variables58784 5588 1.15 mrg Node: OMP_CANCELLATION60573 5589 1.15 mrg Node: OMP_DISPLAY_ENV61106 5590 1.15 mrg Node: OMP_DEFAULT_DEVICE61809 5591 1.15 mrg Node: OMP_DYNAMIC62589 5592 1.15 mrg Node: OMP_MAX_ACTIVE_LEVELS63185 5593 1.15 mrg Node: OMP_MAX_TASK_PRIORITY64112 5594 1.15 mrg Node: OMP_NESTED64770 5595 1.15 mrg Node: OMP_NUM_TEAMS65797 5596 1.15 mrg Node: OMP_NUM_THREADS66470 5597 1.15 mrg Node: OMP_PROC_BIND67275 5598 1.15 mrg Node: OMP_PLACES68617 5599 1.15 mrg Node: OMP_STACKSIZE71306 5600 1.15 mrg Node: OMP_SCHEDULE72130 5601 1.15 mrg Node: OMP_TARGET_OFFLOAD72830 5602 1.15 mrg Node: OMP_TEAMS_THREAD_LIMIT73792 5603 1.15 mrg Node: OMP_THREAD_LIMIT74599 5604 1.15 mrg Node: OMP_WAIT_POLICY75209 5605 1.15 mrg Node: GOMP_CPU_AFFINITY75901 5606 1.15 mrg Node: GOMP_DEBUG77631 5607 1.15 mrg Node: GOMP_STACKSIZE78138 5608 1.15 mrg Node: GOMP_SPINCOUNT78969 5609 1.15 mrg Node: GOMP_RTEMS_THREAD_POOLS80173 5610 1.15 mrg Node: Enabling OpenACC82356 5611 1.15 mrg Node: OpenACC Runtime Library Routines83257 5612 1.15 mrg Node: acc_get_num_devices87538 5613 1.15 mrg Node: acc_set_device_type88264 5614 1.15 mrg Node: acc_get_device_type89028 5615 1.15 mrg Node: acc_set_device_num90041 5616 1.15 mrg Node: acc_get_device_num90858 5617 1.15 mrg Node: acc_get_property91657 5618 1.15 mrg Node: acc_async_test93880 5619 1.15 mrg Node: acc_async_test_all94868 5620 1.15 mrg Node: acc_wait95768 5621 1.15 mrg Node: acc_wait_all96631 5622 1.15 mrg Node: acc_wait_all_async97392 5623 1.15 mrg Node: acc_wait_async98144 5624 1.15 mrg Node: acc_init98852 5625 1.15 mrg Node: acc_shutdown99497 5626 1.15 mrg Node: acc_on_device100164 5627 1.15 mrg Node: acc_malloc101168 5628 1.15 mrg Node: acc_free101667 5629 1.15 mrg Node: acc_copyin102094 5630 1.15 mrg Node: acc_present_or_copyin103681 5631 1.15 mrg Node: acc_create105459 5632 1.15 mrg Node: acc_present_or_create107091 5633 1.15 mrg Node: acc_copyout108877 5634 1.15 mrg Node: acc_delete111181 5635 1.15 mrg Node: acc_update_device113428 5636 1.15 mrg Node: acc_update_self115002 5637 1.15 mrg Node: acc_map_data116592 5638 1.15 mrg Node: acc_unmap_data117277 5639 1.15 mrg Node: acc_deviceptr117798 5640 1.15 mrg Node: acc_hostptr118368 5641 1.15 mrg Node: acc_is_present118932 5642 1.15 mrg Node: acc_memcpy_to_device120459 5643 1.15 mrg Node: acc_memcpy_from_device121122 5644 1.15 mrg Node: acc_attach121789 5645 1.15 mrg Node: acc_detach122436 5646 1.15 mrg Node: acc_get_current_cuda_device123215 5647 1.15 mrg Node: acc_get_current_cuda_context123800 5648 1.15 mrg Node: acc_get_cuda_stream124400 5649 1.15 mrg Node: acc_set_cuda_stream124991 5650 1.15 mrg Node: acc_prof_register125662 5651 1.15 mrg Node: acc_prof_unregister126221 5652 1.15 mrg Node: acc_prof_lookup126788 5653 1.15 mrg Node: acc_register_library127309 5654 1.15 mrg Node: OpenACC Environment Variables127875 5655 1.15 mrg Node: ACC_DEVICE_TYPE128447 5656 1.15 mrg Node: ACC_DEVICE_NUM128683 5657 1.15 mrg Node: ACC_PROFLIB128937 5658 1.15 mrg Node: GCC_ACC_NOTIFY129268 5659 1.15 mrg Node: CUDA Streams Usage129488 5660 1.15 mrg Ref: CUDA Streams Usage-Footnote-1131389 5661 1.15 mrg Node: OpenACC Library Interoperability131498 5662 1.15 mrg Ref: OpenACC Library Interoperability-Footnote-1137866 5663 1.15 mrg Ref: OpenACC Library Interoperability-Footnote-2138118 5664 1.15 mrg Node: OpenACC Profiling Interface138326 5665 1.15 mrg Node: The libgomp ABI148354 5666 1.15 mrg Node: Implementing MASTER construct149207 5667 1.15 mrg Node: Implementing CRITICAL construct149624 5668 1.15 mrg Node: Implementing ATOMIC construct150365 5669 1.15 mrg Node: Implementing FLUSH construct150848 5670 1.15 mrg Node: Implementing BARRIER construct151121 5671 1.15 mrg Node: Implementing THREADPRIVATE construct151392 5672 1.15 mrg Node: Implementing PRIVATE clause152047 5673 1.15 mrg Node: Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses152630 5674 1.15 mrg Node: Implementing REDUCTION clause153956 5675 1.15 mrg Node: Implementing PARALLEL construct154516 5676 1.15 mrg Node: Implementing FOR construct155775 5677 1.15 mrg Node: Implementing ORDERED construct157775 5678 1.15 mrg Node: Implementing SECTIONS construct158083 5679 1.15 mrg Node: Implementing SINGLE construct158851 5680 1.15 mrg Node: Implementing OpenACC's PARALLEL construct159565 5681 1.15 mrg Node: Reporting Bugs159825 5682 1.15 mrg Node: Copying160188 5683 1.15 mrg Node: GNU Free Documentation License197737 5684 1.15 mrg Node: Funding222862 5685 1.15 mrg Node: Library Index225388 5686 1.1 mrg 5687 1.1 mrg End Tag Table 5688
Indexes created Tue Mar 24 00:24:13 UTC 2026