libgomp/libgomp.texi

   1 \input texinfo @c -*-texinfo-*-
   2
   3 @c %**start of header
   4 @setfilename libgomp.info
   5 @settitle GNU libgomp
   6 @c %**end of header
   7
   8
   9 @copying
  10 Copyright @copyright{} 2006-2023 Free Software Foundation, Inc.
  11
  12 Permission is granted to copy, distribute and/or modify this document
  13 under the terms of the GNU Free Documentation License, Version 1.3 or
  14 any later version published by the Free Software Foundation; with the
  15 Invariant Sections being ``Funding Free Software'', the Front-Cover
  16 texts being (a) (see below), and with the Back-Cover Texts being (b)
  17 (see below).  A copy of the license is included in the section entitled
  18 ``GNU Free Documentation License''.
  19
  20 (a) The FSF's Front-Cover Text is:
  21
  22      A GNU Manual
  23
  24 (b) The FSF's Back-Cover Text is:
  25
  26      You have freedom to copy and modify this GNU Manual, like GNU
  27      software.  Copies published by the Free Software Foundation raise
  28      funds for GNU development.
  29 @end copying
  30
  31 @ifinfo
  32 @dircategory GNU Libraries
  33 @direntry
  34 * libgomp: (libgomp).          GNU Offloading and Multi Processing Runtime Library.
  35 @end direntry
  36
  37 This manual documents libgomp, the GNU Offloading and Multi Processing
  38 Runtime library.  This is the GNU implementation of the OpenMP and
  39 OpenACC APIs for parallel and accelerator programming in C/C++ and
  40 Fortran.
  41
  42 Published by the Free Software Foundation
  43 51 Franklin Street, Fifth Floor
  44 Boston, MA 02110-1301 USA
  45
  46 @insertcopying
  47 @end ifinfo
  48
  49
  50 @setchapternewpage odd
  51
  52 @titlepage
  53 @title GNU Offloading and Multi Processing Runtime Library
  54 @subtitle The GNU OpenMP and OpenACC Implementation
  55 @page
  56 @vskip 0pt plus 1filll
  57 @comment For the @value{version-GCC} Version*
  58 @sp 1
  59 Published by the Free Software Foundation @*
  60 51 Franklin Street, Fifth Floor@*
  61 Boston, MA 02110-1301, USA@*
  62 @sp 1
  63 @insertcopying
  64 @end titlepage
  65
  66 @summarycontents
  67 @contents
  68 @page
  69
  70
  71 @node Top, Enabling OpenMP
  72 @top Introduction
  73 @cindex Introduction
  74
  75 This manual documents the usage of libgomp, the GNU Offloading and
  76 Multi Processing Runtime Library.  This includes the GNU
  77 implementation of the @uref{https://www.openmp.org, OpenMP} Application
  78 Programming Interface (API) for multi-platform shared-memory parallel
  79 programming in C/C++ and Fortran, and the GNU implementation of the
  80 @uref{https://www.openacc.org, OpenACC} Application Programming
  81 Interface (API) for offloading of code to accelerator devices in C/C++
  82 and Fortran.
  83
  84 Originally, libgomp implemented the GNU OpenMP Runtime Library.  Based
  85 on this, support for OpenACC and offloading (both OpenACC and OpenMP
  86 4's target construct) has been added later on, and the library's name
  87 changed to GNU Offloading and Multi Processing Runtime Library.
  88
  89
  90
  91 @comment
  92 @comment  When you add a new menu item, please keep the right hand
  93 @comment  aligned to the same column.  Do not use tabs.  This provides
  94 @comment  better formatting.
  95 @comment
  96 @menu
  97 * Enabling OpenMP::            How to enable OpenMP for your applications.
  98 * OpenMP Implementation Status:: List of implemented features by OpenMP version
  99 * OpenMP Runtime Library Routines: Runtime Library Routines.
 100                                The OpenMP runtime application programming
 101                                interface.
 102 * OpenMP Environment Variables: Environment Variables.
 103                                Influencing OpenMP runtime behavior with
 104                                environment variables.
 105 * Enabling OpenACC::           How to enable OpenACC for your
 106                                applications.
 107 * OpenACC Runtime Library Routines:: The OpenACC runtime application
 108                                programming interface.
 109 * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with
 110                                environment variables.
 111 * CUDA Streams Usage::         Notes on the implementation of
 112                                asynchronous operations.
 113 * OpenACC Library Interoperability:: OpenACC library interoperability with the
 114                                NVIDIA CUBLAS library.
 115 * OpenACC Profiling Interface::
 116 * OpenMP-Implementation Specifics:: Notes specifics of this OpenMP
 117                                implementation
 118 * Offload-Target Specifics::   Notes on offload-target specific internals
 119 * The libgomp ABI::            Notes on the external ABI presented by libgomp.
 120 * Reporting Bugs::             How to report bugs in the GNU Offloading and
 121                                Multi Processing Runtime Library.
 122 * Copying::                    GNU general public license says
 123                                how you can copy and share libgomp.
 124 * GNU Free Documentation License::
 125                                How you can copy and share this manual.
 126 * Funding::                    How to help assure continued work for free
 127                                software.
 128 * Library Index::              Index of this documentation.
 129 @end menu
 130
 131
 132 @c ---------------------------------------------------------------------
 133 @c Enabling OpenMP
 134 @c ---------------------------------------------------------------------
 135
 136 @node Enabling OpenMP
 137 @chapter Enabling OpenMP
 138
 139 To activate the OpenMP extensions for C/C++ and Fortran, the compile-time
 140 flag @command{-fopenmp} must be specified.  This enables the OpenMP directive
 141 @code{#pragma omp} in C/C++ and @code{!$omp} directives in free form,
 142 @code{c$omp}, @code{*$omp} and @code{!$omp} directives in fixed form,
 143 @code{!$} conditional compilation sentinels in free form and @code{c$},
 144 @code{*$} and @code{!$} sentinels in fixed form, for Fortran.  The flag also
 145 arranges for automatic linking of the OpenMP runtime library
 146 (@ref{Runtime Library Routines}).
 147
 148 A complete description of all OpenMP directives may be found in the
 149 @uref{https://www.openmp.org, OpenMP Application Program Interface} manuals.
 150 See also @ref{OpenMP Implementation Status}.
 151
 152
 153 @c ---------------------------------------------------------------------
 154 @c OpenMP Implementation Status
 155 @c ---------------------------------------------------------------------
 156
 157 @node OpenMP Implementation Status
 158 @chapter OpenMP Implementation Status
 159
 160 @menu
 161 * OpenMP 4.5:: Feature completion status to 4.5 specification
 162 * OpenMP 5.0:: Feature completion status to 5.0 specification
 163 * OpenMP 5.1:: Feature completion status to 5.1 specification
 164 * OpenMP 5.2:: Feature completion status to 5.2 specification
 165 * OpenMP Technical Report 11:: Feature completion status to first 6.0 preview
 166 @end menu
 167
 168 The @code{_OPENMP} preprocessor macro and Fortran's @code{openmp_version}
 169 parameter, provided by @code{omp_lib.h} and the @code{omp_lib} module, have
 170 the value @code{201511} (i.e. OpenMP 4.5).
 171
 172 @node OpenMP 4.5
 173 @section OpenMP 4.5
 174
 175 The OpenMP 4.5 specification is fully supported.
 176
 177 @node OpenMP 5.0
 178 @section OpenMP 5.0
 179
 180 @unnumberedsubsec New features listed in Appendix B of the OpenMP specification
 181 @c This list is sorted as in OpenMP 5.1's B.3 not as in OpenMP 5.0's B.2
 182
 183 @multitable @columnfractions .60 .10 .25
 184 @headitem Description @tab Status @tab Comments
 185 @item Array shaping @tab N @tab
 186 @item Array sections with non-unit strides in C and C++ @tab N @tab
 187 @item Iterators @tab Y @tab
 188 @item @code{metadirective} directive @tab N @tab
 189 @item @code{declare variant} directive
 190       @tab P @tab @emph{simd} traits not handled correctly
 191 @item @var{target-offload-var} ICV and @code{OMP_TARGET_OFFLOAD}
 192       env variable @tab Y @tab
 193 @item Nested-parallel changes to @var{max-active-levels-var} ICV @tab Y @tab
 194 @item @code{requires} directive @tab P
 195       @tab complete but no non-host device provides @code{unified_shared_memory}
 196 @item @code{teams} construct outside an enclosing target region @tab Y @tab
 197 @item Non-rectangular loop nests @tab P
 198       @tab Full support for C/C++, partial for Fortran
 199            (@uref{https://gcc.gnu.org/PR110735,PR110735})
 200 @item @code{!=} as relational-op in canonical loop form for C/C++ @tab Y @tab
 201 @item @code{nonmonotonic} as default loop schedule modifier for worksharing-loop
 202       constructs @tab Y @tab
 203 @item Collapse of associated loops that are imperfectly nested loops @tab Y @tab
 204 @item Clauses @code{if}, @code{nontemporal} and @code{order(concurrent)} in
 205       @code{simd} construct @tab Y @tab
 206 @item @code{atomic} constructs in @code{simd} @tab Y @tab
 207 @item @code{loop} construct @tab Y @tab
 208 @item @code{order(concurrent)} clause @tab Y @tab
 209 @item @code{scan} directive and @code{in_scan} modifier for the
 210       @code{reduction} clause @tab Y @tab
 211 @item @code{in_reduction} clause on @code{task} constructs @tab Y @tab
 212 @item @code{in_reduction} clause on @code{target} constructs @tab P
 213       @tab @code{nowait} only stub
 214 @item @code{task_reduction} clause with @code{taskgroup} @tab Y @tab
 215 @item @code{task} modifier to @code{reduction} clause @tab Y @tab
 216 @item @code{affinity} clause to @code{task} construct @tab Y @tab Stub only
 217 @item @code{detach} clause to @code{task} construct @tab Y @tab
 218 @item @code{omp_fulfill_event} runtime routine @tab Y @tab
 219 @item @code{reduction} and @code{in_reduction} clauses on @code{taskloop}
 220       and @code{taskloop simd} constructs @tab Y @tab
 221 @item @code{taskloop} construct cancelable by @code{cancel} construct
 222       @tab Y @tab
 223 @item @code{mutexinoutset} @emph{dependence-type} for @code{depend} clause
 224       @tab Y @tab
 225 @item Predefined memory spaces, memory allocators, allocator traits
 226       @tab Y @tab See also @ref{Memory allocation}
 227 @item Memory management routines @tab Y @tab
 228 @item @code{allocate} directive @tab P @tab Only C, only stack variables
 229 @item @code{allocate} clause @tab P @tab Initial support
 230 @item @code{use_device_addr} clause on @code{target data} @tab Y @tab
 231 @item @code{ancestor} modifier on @code{device} clause @tab Y @tab
 232 @item Implicit declare target directive @tab Y @tab
 233 @item Discontiguous array section with @code{target update} construct
 234       @tab N @tab
 235 @item C/C++'s lvalue expressions in @code{to}, @code{from}
 236       and @code{map} clauses @tab N @tab
 237 @item C/C++'s lvalue expressions in @code{depend} clauses @tab Y @tab
 238 @item Nested @code{declare target} directive @tab Y @tab
 239 @item Combined @code{master} constructs @tab Y @tab
 240 @item @code{depend} clause on @code{taskwait} @tab Y @tab
 241 @item Weak memory ordering clauses on @code{atomic} and @code{flush} construct
 242       @tab Y @tab
 243 @item @code{hint} clause on the @code{atomic} construct @tab Y @tab Stub only
 244 @item @code{depobj} construct and depend objects  @tab Y @tab
 245 @item Lock hints were renamed to synchronization hints @tab Y @tab
 246 @item @code{conditional} modifier to @code{lastprivate} clause @tab Y @tab
 247 @item Map-order clarifications @tab P @tab
 248 @item @code{close} @emph{map-type-modifier} @tab Y @tab
 249 @item Mapping C/C++ pointer variables and to assign the address of
 250       device memory mapped by an array section @tab P @tab
 251 @item Mapping of Fortran pointer and allocatable variables, including pointer
 252       and allocatable components of variables
 253       @tab P @tab Mapping of vars with allocatable components unsupported
 254 @item @code{defaultmap} extensions @tab Y @tab
 255 @item @code{declare mapper} directive @tab N @tab
 256 @item @code{omp_get_supported_active_levels} routine @tab Y @tab
 257 @item Runtime routines and environment variables to display runtime thread
 258       affinity information @tab Y @tab
 259 @item @code{omp_pause_resource} and @code{omp_pause_resource_all} runtime
 260       routines @tab Y @tab
 261 @item @code{omp_get_device_num} runtime routine @tab Y @tab
 262 @item OMPT interface @tab N @tab
 263 @item OMPD interface @tab N @tab
 264 @end multitable
 265
 266 @unnumberedsubsec Other new OpenMP 5.0 features
 267
 268 @multitable @columnfractions .60 .10 .25
 269 @headitem Description @tab Status @tab Comments
 270 @item Supporting C++'s range-based for loop @tab Y @tab
 271 @end multitable
 272
 273
 274 @node OpenMP 5.1
 275 @section OpenMP 5.1
 276
 277 @unnumberedsubsec New features listed in Appendix B of the OpenMP specification
 278
 279 @multitable @columnfractions .60 .10 .25
 280 @headitem Description @tab Status @tab Comments
 281 @item OpenMP directive as C++ attribute specifiers @tab Y @tab
 282 @item @code{omp_all_memory} reserved locator @tab Y @tab
 283 @item @emph{target_device trait} in OpenMP Context @tab N @tab
 284 @item @code{target_device} selector set in context selectors @tab N @tab
 285 @item C/C++'s @code{declare variant} directive: elision support of
 286       preprocessed code @tab N @tab
 287 @item @code{declare variant}: new clauses @code{adjust_args} and
 288       @code{append_args} @tab N @tab
 289 @item @code{dispatch} construct @tab N @tab
 290 @item device-specific ICV settings with environment variables @tab Y @tab
 291 @item @code{assume} and @code{assumes} directives @tab Y @tab
 292 @item @code{nothing} directive @tab Y @tab
 293 @item @code{error} directive @tab Y @tab
 294 @item @code{masked} construct @tab Y @tab
 295 @item @code{scope} directive @tab Y @tab
 296 @item Loop transformation constructs @tab N @tab
 297 @item @code{strict} modifier in the @code{grainsize} and @code{num_tasks}
 298       clauses of the @code{taskloop} construct @tab Y @tab
 299 @item @code{align} clause in @code{allocate} directive @tab P
 300       @tab Only C (and only stack variables)
 301 @item @code{align} modifier in @code{allocate} clause @tab Y @tab
 302 @item @code{thread_limit} clause to @code{target} construct @tab Y @tab
 303 @item @code{has_device_addr} clause to @code{target} construct @tab Y @tab
 304 @item Iterators in @code{target update} motion clauses and @code{map}
 305       clauses @tab N @tab
 306 @item Indirect calls to the device version of a procedure or function in
 307       @code{target} regions @tab N @tab
 308 @item @code{interop} directive @tab N @tab
 309 @item @code{omp_interop_t} object support in runtime routines @tab N @tab
 310 @item @code{nowait} clause in @code{taskwait} directive @tab Y @tab
 311 @item Extensions to the @code{atomic} directive @tab Y @tab
 312 @item @code{seq_cst} clause on a @code{flush} construct @tab Y @tab
 313 @item @code{inoutset} argument to the @code{depend} clause @tab Y @tab
 314 @item @code{private} and @code{firstprivate} argument to @code{default}
 315       clause in C and C++ @tab Y @tab
 316 @item @code{present} argument to @code{defaultmap} clause @tab Y @tab
 317 @item @code{omp_set_num_teams}, @code{omp_set_teams_thread_limit},
 318       @code{omp_get_max_teams}, @code{omp_get_teams_thread_limit} runtime
 319       routines @tab Y @tab
 320 @item @code{omp_target_is_accessible} runtime routine @tab Y @tab
 321 @item @code{omp_target_memcpy_async} and @code{omp_target_memcpy_rect_async}
 322       runtime routines @tab Y @tab
 323 @item @code{omp_get_mapped_ptr} runtime routine @tab Y @tab
 324 @item @code{omp_calloc}, @code{omp_realloc}, @code{omp_aligned_alloc} and
 325       @code{omp_aligned_calloc} runtime routines @tab Y @tab
 326 @item @code{omp_alloctrait_key_t} enum: @code{omp_atv_serialized} added,
 327       @code{omp_atv_default} changed @tab Y @tab
 328 @item @code{omp_display_env} runtime routine @tab Y @tab
 329 @item @code{ompt_scope_endpoint_t} enum: @code{ompt_scope_beginend} @tab N @tab
 330 @item @code{ompt_sync_region_t} enum additions @tab N @tab
 331 @item @code{ompt_state_t} enum: @code{ompt_state_wait_barrier_implementation}
 332       and @code{ompt_state_wait_barrier_teams} @tab N @tab
 333 @item @code{ompt_callback_target_data_op_emi_t},
 334       @code{ompt_callback_target_emi_t}, @code{ompt_callback_target_map_emi_t}
 335       and @code{ompt_callback_target_submit_emi_t} @tab N @tab
 336 @item @code{ompt_callback_error_t} type @tab N @tab
 337 @item @code{OMP_PLACES} syntax extensions @tab Y @tab
 338 @item @code{OMP_NUM_TEAMS} and @code{OMP_TEAMS_THREAD_LIMIT} environment
 339       variables @tab Y @tab
 340 @end multitable
 341
 342 @unnumberedsubsec Other new OpenMP 5.1 features
 343
 344 @multitable @columnfractions .60 .10 .25
 345 @headitem Description @tab Status @tab Comments
 346 @item Support of strictly structured blocks in Fortran @tab Y @tab
 347 @item Support of structured block sequences in C/C++ @tab Y @tab
 348 @item @code{unconstrained} and @code{reproducible} modifiers on @code{order}
 349       clause @tab Y @tab
 350 @item Support @code{begin/end declare target} syntax in C/C++ @tab Y @tab
 351 @item Pointer predetermined firstprivate getting initialized
 352 to address of matching mapped list item per 5.1, Sect. 2.21.7.2 @tab N @tab
 353 @item For Fortran, diagnose placing declarative before/between @code{USE},
 354       @code{IMPORT}, and @code{IMPLICIT} as invalid @tab N @tab
 355 @item Optional comma between directive and clause in the @code{#pragma} form @tab Y @tab
 356 @item @code{indirect} clause in @code{declare target} @tab N @tab
 357 @item @code{device_type(nohost)}/@code{device_type(host)} for variables @tab N @tab
 358 @item @code{present} modifier to the @code{map}, @code{to} and @code{from}
 359       clauses @tab Y @tab
 360 @end multitable
 361
 362
 363 @node OpenMP 5.2
 364 @section OpenMP 5.2
 365
 366 @unnumberedsubsec New features listed in Appendix B of the OpenMP specification
 367
 368 @multitable @columnfractions .60 .10 .25
 369 @headitem Description @tab Status @tab Comments
 370 @item @code{omp_in_explicit_task} routine and @var{explicit-task-var} ICV
 371       @tab Y @tab
 372 @item @code{omp}/@code{ompx}/@code{omx} sentinels and @code{omp_}/@code{ompx_}
 373       namespaces @tab N/A
 374       @tab warning for @code{ompx/omx} sentinels@footnote{The @code{ompx}
 375       sentinel as C/C++ pragma and C++ attributes are warned for with
 376       @code{-Wunknown-pragmas} (implied by @code{-Wall}) and @code{-Wattributes}
 377       (enabled by default), respectively; for Fortran free-source code, there is
 378       a warning enabled by default and, for fixed-source code, the @code{omx}
 379       sentinel is warned for with with @code{-Wsurprising} (enabled by
 380       @code{-Wall}).  Unknown clauses are always rejected with an error.}
 381 @item Clauses on @code{end} directive can be on directive @tab Y @tab
 382 @item @code{destroy} clause with destroy-var argument on @code{depobj}
 383       @tab N @tab
 384 @item Deprecation of no-argument @code{destroy} clause on @code{depobj}
 385       @tab N @tab
 386 @item @code{linear} clause syntax changes and @code{step} modifier @tab Y @tab
 387 @item Deprecation of minus operator for reductions @tab N @tab
 388 @item Deprecation of separating @code{map} modifiers without comma @tab N @tab
 389 @item @code{declare mapper} with iterator and @code{present} modifiers
 390       @tab N @tab
 391 @item If a matching mapped list item is not found in the data environment, the
 392       pointer retains its original value @tab Y @tab
 393 @item New @code{enter} clause as alias for @code{to} on declare target directive
 394       @tab Y @tab
 395 @item Deprecation of @code{to} clause on declare target directive @tab N @tab
 396 @item Extended list of directives permitted in Fortran pure procedures
 397       @tab Y @tab
 398 @item New @code{allocators} directive for Fortran @tab N @tab
 399 @item Deprecation of @code{allocate} directive for Fortran
 400       allocatables/pointers @tab N @tab
 401 @item Optional paired @code{end} directive with @code{dispatch} @tab N @tab
 402 @item New @code{memspace} and @code{traits} modifiers for @code{uses_allocators}
 403       @tab N @tab
 404 @item Deprecation of traits array following the allocator_handle expression in
 405       @code{uses_allocators} @tab N @tab
 406 @item New @code{otherwise} clause as alias for @code{default} on metadirectives
 407       @tab N @tab
 408 @item Deprecation of @code{default} clause on metadirectives @tab N @tab
 409 @item Deprecation of delimited form of @code{declare target} @tab N @tab
 410 @item Reproducible semantics changed for @code{order(concurrent)} @tab N @tab
 411 @item @code{allocate} and @code{firstprivate} clauses on @code{scope}
 412       @tab Y @tab
 413 @item @code{ompt_callback_work} @tab N @tab
 414 @item Default map-type for the @code{map} clause in @code{target enter/exit data}
 415       @tab Y @tab
 416 @item New @code{doacross} clause as alias for @code{depend} with
 417       @code{source}/@code{sink} modifier @tab Y @tab
 418 @item Deprecation of @code{depend} with @code{source}/@code{sink} modifier
 419       @tab N @tab
 420 @item @code{omp_cur_iteration} keyword @tab Y @tab
 421 @end multitable
 422
 423 @unnumberedsubsec Other new OpenMP 5.2 features
 424
 425 @multitable @columnfractions .60 .10 .25
 426 @headitem Description @tab Status @tab Comments
 427 @item For Fortran, optional comma between directive and clause @tab N @tab
 428 @item Conforming device numbers and @code{omp_initial_device} and
 429       @code{omp_invalid_device} enum/PARAMETER @tab Y @tab
 430 @item Initial value of @var{default-device-var} ICV with
 431       @code{OMP_TARGET_OFFLOAD=mandatory} @tab Y @tab
 432 @item @code{all} as @emph{implicit-behavior} for @code{defaultmap} @tab Y @tab
 433 @item @emph{interop_types} in any position of the modifier list for the @code{init} clause
 434       of the @code{interop} construct @tab N @tab
 435 @end multitable
 436
 437
 438 @node OpenMP Technical Report 11
 439 @section OpenMP Technical Report 11
 440
 441 Technical Report (TR) 11 is the first preview for OpenMP 6.0.
 442
 443 @unnumberedsubsec New features listed in Appendix B of the OpenMP specification
 444 @multitable @columnfractions .60 .10 .25
 445 @item Features deprecated in versions 5.2, 5.1 and 5.0 were removed
 446       @tab N/A @tab Backward compatibility
 447 @item The @code{decl} attribute was added to the C++ attribute syntax
 448       @tab Y @tab
 449 @item @code{_ALL} suffix to the device-scope environment variables
 450       @tab P @tab Host device number wrongly accepted
 451 @item For Fortran, @emph{locator list} can be also function reference with
 452       data pointer result @tab N @tab
 453 @item Ref-count change for @code{use_device_ptr}/@code{use_device_addr}
 454       @tab N @tab
 455 @item Implicit reduction identifiers of C++ classes
 456       @tab N @tab
 457 @item Change of the @emph{map-type} property from @emph{ultimate} to
 458       @emph{default} @tab N @tab
 459 @item Concept of @emph{assumed-size arrays} in C and C++
 460       @tab N @tab
 461 @item Mapping of @emph{assumed-size arrays} in C, C++ and Fortran
 462       @tab N @tab
 463 @item @code{groupprivate} directive @tab N @tab
 464 @item @code{local} clause to declare target directive @tab N @tab
 465 @item @code{part_size} allocator trait @tab N @tab
 466 @item @code{pin_device}, @code{preferred_device} and @code{target_access}
 467       allocator traits
 468       @tab N @tab
 469 @item @code{access} allocator trait changes @tab N @tab
 470 @item Extension of @code{interop} operation of @code{append_args}, allowing all
 471       modifiers of the @code{init} clause
 472       @tab N @tab
 473 @item @code{interop} clause to @code{dispatch} @tab N @tab
 474 @item @code{apply} code to loop-transforming constructs @tab N @tab
 475 @item @code{omp_curr_progress_width} identifier @tab N @tab
 476 @item @code{safesync} clause to the @code{parallel} construct @tab N @tab
 477 @item @code{omp_get_max_progress_width} runtime routine @tab N @tab
 478 @item @code{strict} modifier keyword to @code{num_threads} @tab N @tab
 479 @item @code{memscope} clause to @code{atomic} and @code{flush} @tab N @tab
 480 @item Routines for obtaining memory spaces/allocators for shared/device memory
 481       @tab N @tab
 482 @item @code{omp_get_memspace_num_resources} routine @tab N @tab
 483 @item @code{omp_get_submemspace} routine @tab N @tab
 484 @item @code{ompt_get_buffer_limits} OMPT routine @tab N @tab
 485 @item Extension of @code{OMP_DEFAULT_DEVICE} and new
 486       @code{OMP_AVAILABLE_DEVICES} environment vars @tab N @tab
 487 @item Supporting increments with abstract names in @code{OMP_PLACES} @tab N @tab
 488 @end multitable
 489
 490 @unnumberedsubsec Other new TR 11 features
 491 @multitable @columnfractions .60 .10 .25
 492 @item Relaxed Fortran restrictions to the @code{aligned} clause @tab N @tab
 493 @item Mapping lambda captures @tab N @tab
 494 @item For Fortran, atomic compare with storing the comparison result
 495       @tab N @tab
 496 @end multitable
 497
 498
 499
 500 @c ---------------------------------------------------------------------
 501 @c OpenMP Runtime Library Routines
 502 @c ---------------------------------------------------------------------
 503
 504 @node Runtime Library Routines
 505 @chapter OpenMP Runtime Library Routines
 506
 507 The runtime routines described here are defined by Section 18 of the OpenMP
 508 specification in version 5.2.
 509
 510 @menu
 511 * Thread Team Routines::
 512 * Thread Affinity Routines::
 513 * Teams Region Routines::
 514 * Tasking Routines::
 515 @c * Resource Relinquishing Routines::
 516 * Device Information Routines::
 517 * Device Memory Routines::
 518 * Lock Routines::
 519 * Timing Routines::
 520 * Event Routine::
 521 @c * Interoperability Routines::
 522 * Memory Management Routines::
 523 @c * Tool Control Routine::
 524 @c * Environment Display Routine::
 525 @end menu
 526
 527
 528
 529 @node Thread Team Routines
 530 @section Thread Team Routines
 531
 532 Routines controlling threads in the current contention group.
 533 They have C linkage and do not throw exceptions.
 534
 535 @menu
 536 * omp_set_num_threads::         Set upper team size limit
 537 * omp_get_num_threads::         Size of the active team
 538 * omp_get_max_threads::         Maximum number of threads of parallel region
 539 * omp_get_thread_num::          Current thread ID
 540 * omp_in_parallel::             Whether a parallel region is active
 541 * omp_set_dynamic::             Enable/disable dynamic teams
 542 * omp_get_dynamic::             Dynamic teams setting
 543 * omp_get_cancellation::        Whether cancellation support is enabled
 544 * omp_set_nested::              Enable/disable nested parallel regions
 545 * omp_get_nested::              Nested parallel regions
 546 * omp_set_schedule::            Set the runtime scheduling method
 547 * omp_get_schedule::            Obtain the runtime scheduling method
 548 * omp_get_teams_thread_limit::  Maximum number of threads imposed by teams
 549 * omp_get_supported_active_levels:: Maximum number of active regions supported
 550 * omp_set_max_active_levels::   Limits the number of active parallel regions
 551 * omp_get_max_active_levels::   Current maximum number of active regions
 552 * omp_get_level::               Number of parallel regions
 553 * omp_get_ancestor_thread_num:: Ancestor thread ID
 554 * omp_get_team_size::           Number of threads in a team
 555 * omp_get_active_level::        Number of active parallel regions
 556 @end menu
 557
 558
 559
 560 @node omp_set_num_threads
 561 @subsection @code{omp_set_num_threads} -- Set upper team size limit
 562 @table @asis
 563 @item @emph{Description}:
 564 Specifies the number of threads used by default in subsequent parallel
 565 sections, if those do not specify a @code{num_threads} clause.  The
 566 argument of @code{omp_set_num_threads} shall be a positive integer.
 567
 568 @item @emph{C/C++}:
 569 @multitable @columnfractions .20 .80
 570 @item @emph{Prototype}: @tab @code{void omp_set_num_threads(int num_threads);}
 571 @end multitable
 572
 573 @item @emph{Fortran}:
 574 @multitable @columnfractions .20 .80
 575 @item @emph{Interface}: @tab @code{subroutine omp_set_num_threads(num_threads)}
 576 @item                   @tab @code{integer, intent(in) :: num_threads}
 577 @end multitable
 578
 579 @item @emph{See also}:
 580 @ref{OMP_NUM_THREADS}, @ref{omp_get_num_threads}, @ref{omp_get_max_threads}
 581
 582 @item @emph{Reference}:
 583 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.1.
 584 @end table
 585
 586
 587
 588 @node omp_get_num_threads
 589 @subsection @code{omp_get_num_threads} -- Size of the active team
 590 @table @asis
 591 @item @emph{Description}:
 592 Returns the number of threads in the current team.  In a sequential section of
 593 the program @code{omp_get_num_threads} returns 1.
 594
 595 The default team size may be initialized at startup by the
 596 @env{OMP_NUM_THREADS} environment variable.  At runtime, the size
 597 of the current team may be set either by the @code{NUM_THREADS}
 598 clause or by @code{omp_set_num_threads}.  If none of the above were
 599 used to define a specific value and @env{OMP_DYNAMIC} is disabled,
 600 one thread per CPU online is used.
 601
 602 @item @emph{C/C++}:
 603 @multitable @columnfractions .20 .80
 604 @item @emph{Prototype}: @tab @code{int omp_get_num_threads(void);}
 605 @end multitable
 606
 607 @item @emph{Fortran}:
 608 @multitable @columnfractions .20 .80
 609 @item @emph{Interface}: @tab @code{integer function omp_get_num_threads()}
 610 @end multitable
 611
 612 @item @emph{See also}:
 613 @ref{omp_get_max_threads}, @ref{omp_set_num_threads}, @ref{OMP_NUM_THREADS}
 614
 615 @item @emph{Reference}:
 616 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.2.
 617 @end table
 618
 619
 620
 621 @node omp_get_max_threads
 622 @subsection @code{omp_get_max_threads} -- Maximum number of threads of parallel region
 623 @table @asis
 624 @item @emph{Description}:
 625 Return the maximum number of threads used for the current parallel region
 626 that does not use the clause @code{num_threads}.
 627
 628 @item @emph{C/C++}:
 629 @multitable @columnfractions .20 .80
 630 @item @emph{Prototype}: @tab @code{int omp_get_max_threads(void);}
 631 @end multitable
 632
 633 @item @emph{Fortran}:
 634 @multitable @columnfractions .20 .80
 635 @item @emph{Interface}: @tab @code{integer function omp_get_max_threads()}
 636 @end multitable
 637
 638 @item @emph{See also}:
 639 @ref{omp_set_num_threads}, @ref{omp_set_dynamic}, @ref{omp_get_thread_limit}
 640
 641 @item @emph{Reference}:
 642 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.3.
 643 @end table
 644
 645
 646
 647 @node omp_get_thread_num
 648 @subsection @code{omp_get_thread_num} -- Current thread ID
 649 @table @asis
 650 @item @emph{Description}:
 651 Returns a unique thread identification number within the current team.
 652 In a sequential parts of the program, @code{omp_get_thread_num}
 653 always returns 0.  In parallel regions the return value varies
 654 from 0 to @code{omp_get_num_threads}-1 inclusive.  The return
 655 value of the primary thread of a team is always 0.
 656
 657 @item @emph{C/C++}:
 658 @multitable @columnfractions .20 .80
 659 @item @emph{Prototype}: @tab @code{int omp_get_thread_num(void);}
 660 @end multitable
 661
 662 @item @emph{Fortran}:
 663 @multitable @columnfractions .20 .80
 664 @item @emph{Interface}: @tab @code{integer function omp_get_thread_num()}
 665 @end multitable
 666
 667 @item @emph{See also}:
 668 @ref{omp_get_num_threads}, @ref{omp_get_ancestor_thread_num}
 669
 670 @item @emph{Reference}:
 671 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.4.
 672 @end table
 673
 674
 675
 676 @node omp_in_parallel
 677 @subsection @code{omp_in_parallel} -- Whether a parallel region is active
 678 @table @asis
 679 @item @emph{Description}:
 680 This function returns @code{true} if currently running in parallel,
 681 @code{false} otherwise.  Here, @code{true} and @code{false} represent
 682 their language-specific counterparts.
 683
 684 @item @emph{C/C++}:
 685 @multitable @columnfractions .20 .80
 686 @item @emph{Prototype}: @tab @code{int omp_in_parallel(void);}
 687 @end multitable
 688
 689 @item @emph{Fortran}:
 690 @multitable @columnfractions .20 .80
 691 @item @emph{Interface}: @tab @code{logical function omp_in_parallel()}
 692 @end multitable
 693
 694 @item @emph{Reference}:
 695 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.6.
 696 @end table
 697
 698
 699 @node omp_set_dynamic
 700 @subsection @code{omp_set_dynamic} -- Enable/disable dynamic teams
 701 @table @asis
 702 @item @emph{Description}:
 703 Enable or disable the dynamic adjustment of the number of threads
 704 within a team.  The function takes the language-specific equivalent
 705 of @code{true} and @code{false}, where @code{true} enables dynamic
 706 adjustment of team sizes and @code{false} disables it.
 707
 708 @item @emph{C/C++}:
 709 @multitable @columnfractions .20 .80
 710 @item @emph{Prototype}: @tab @code{void omp_set_dynamic(int dynamic_threads);}
 711 @end multitable
 712
 713 @item @emph{Fortran}:
 714 @multitable @columnfractions .20 .80
 715 @item @emph{Interface}: @tab @code{subroutine omp_set_dynamic(dynamic_threads)}
 716 @item                   @tab @code{logical, intent(in) :: dynamic_threads}
 717 @end multitable
 718
 719 @item @emph{See also}:
 720 @ref{OMP_DYNAMIC}, @ref{omp_get_dynamic}
 721
 722 @item @emph{Reference}:
 723 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.7.
 724 @end table
 725
 726
 727
 728 @node omp_get_dynamic
 729 @subsection @code{omp_get_dynamic} -- Dynamic teams setting
 730 @table @asis
 731 @item @emph{Description}:
 732 This function returns @code{true} if enabled, @code{false} otherwise.
 733 Here, @code{true} and @code{false} represent their language-specific
 734 counterparts.
 735
 736 The dynamic team setting may be initialized at startup by the
 737 @env{OMP_DYNAMIC} environment variable or at runtime using
 738 @code{omp_set_dynamic}.  If undefined, dynamic adjustment is
 739 disabled by default.
 740
 741 @item @emph{C/C++}:
 742 @multitable @columnfractions .20 .80
 743 @item @emph{Prototype}: @tab @code{int omp_get_dynamic(void);}
 744 @end multitable
 745
 746 @item @emph{Fortran}:
 747 @multitable @columnfractions .20 .80
 748 @item @emph{Interface}: @tab @code{logical function omp_get_dynamic()}
 749 @end multitable
 750
 751 @item @emph{See also}:
 752 @ref{omp_set_dynamic}, @ref{OMP_DYNAMIC}
 753
 754 @item @emph{Reference}:
 755 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.8.
 756 @end table
 757
 758
 759
 760 @node omp_get_cancellation
 761 @subsection @code{omp_get_cancellation} -- Whether cancellation support is enabled
 762 @table @asis
 763 @item @emph{Description}:
 764 This function returns @code{true} if cancellation is activated, @code{false}
 765 otherwise.  Here, @code{true} and @code{false} represent their language-specific
 766 counterparts.  Unless @env{OMP_CANCELLATION} is set true, cancellations are
 767 deactivated.
 768
 769 @item @emph{C/C++}:
 770 @multitable @columnfractions .20 .80
 771 @item @emph{Prototype}: @tab @code{int omp_get_cancellation(void);}
 772 @end multitable
 773
 774 @item @emph{Fortran}:
 775 @multitable @columnfractions .20 .80
 776 @item @emph{Interface}: @tab @code{logical function omp_get_cancellation()}
 777 @end multitable
 778
 779 @item @emph{See also}:
 780 @ref{OMP_CANCELLATION}
 781
 782 @item @emph{Reference}:
 783 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.9.
 784 @end table
 785
 786
 787
 788 @node omp_set_nested
 789 @subsection @code{omp_set_nested} -- Enable/disable nested parallel regions
 790 @table @asis
 791 @item @emph{Description}:
 792 Enable or disable nested parallel regions, i.e., whether team members
 793 are allowed to create new teams.  The function takes the language-specific
 794 equivalent of @code{true} and @code{false}, where @code{true} enables
 795 dynamic adjustment of team sizes and @code{false} disables it.
 796
 797 Enabling nested parallel regions will also set the maximum number of
 798 active nested regions to the maximum supported.  Disabling nested parallel
 799 regions will set the maximum number of active nested regions to one.
 800
 801 Note that the @code{omp_set_nested} API routine was deprecated
 802 in the OpenMP specification 5.2 in favor of @code{omp_set_max_active_levels}.
 803
 804 @item @emph{C/C++}:
 805 @multitable @columnfractions .20 .80
 806 @item @emph{Prototype}: @tab @code{void omp_set_nested(int nested);}
 807 @end multitable
 808
 809 @item @emph{Fortran}:
 810 @multitable @columnfractions .20 .80
 811 @item @emph{Interface}: @tab @code{subroutine omp_set_nested(nested)}
 812 @item                   @tab @code{logical, intent(in) :: nested}
 813 @end multitable
 814
 815 @item @emph{See also}:
 816 @ref{omp_get_nested}, @ref{omp_set_max_active_levels},
 817 @ref{OMP_MAX_ACTIVE_LEVELS}, @ref{OMP_NESTED}
 818
 819 @item @emph{Reference}:
 820 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.10.
 821 @end table
 822
 823
 824
 825 @node omp_get_nested
 826 @subsection @code{omp_get_nested} -- Nested parallel regions
 827 @table @asis
 828 @item @emph{Description}:
 829 This function returns @code{true} if nested parallel regions are
 830 enabled, @code{false} otherwise.  Here, @code{true} and @code{false}
 831 represent their language-specific counterparts.
 832
 833 The state of nested parallel regions at startup depends on several
 834 environment variables.  If @env{OMP_MAX_ACTIVE_LEVELS} is defined
 835 and is set to greater than one, then nested parallel regions will be
 836 enabled.  If not defined, then the value of the @env{OMP_NESTED}
 837 environment variable will be followed if defined.  If neither are
 838 defined, then if either @env{OMP_NUM_THREADS} or @env{OMP_PROC_BIND}
 839 are defined with a list of more than one value, then nested parallel
 840 regions are enabled.  If none of these are defined, then nested parallel
 841 regions are disabled by default.
 842
 843 Nested parallel regions can be enabled or disabled at runtime using
 844 @code{omp_set_nested}, or by setting the maximum number of nested
 845 regions with @code{omp_set_max_active_levels} to one to disable, or
 846 above one to enable.
 847
 848 Note that the @code{omp_get_nested} API routine was deprecated
 849 in the OpenMP specification 5.2 in favor of @code{omp_get_max_active_levels}.
 850
 851 @item @emph{C/C++}:
 852 @multitable @columnfractions .20 .80
 853 @item @emph{Prototype}: @tab @code{int omp_get_nested(void);}
 854 @end multitable
 855
 856 @item @emph{Fortran}:
 857 @multitable @columnfractions .20 .80
 858 @item @emph{Interface}: @tab @code{logical function omp_get_nested()}
 859 @end multitable
 860
 861 @item @emph{See also}:
 862 @ref{omp_get_max_active_levels}, @ref{omp_set_nested},
 863 @ref{OMP_MAX_ACTIVE_LEVELS}, @ref{OMP_NESTED}
 864
 865 @item @emph{Reference}:
 866 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.11.
 867 @end table
 868
 869
 870
 871 @node omp_set_schedule
 872 @subsection @code{omp_set_schedule} -- Set the runtime scheduling method
 873 @table @asis
 874 @item @emph{Description}:
 875 Sets the runtime scheduling method.  The @var{kind} argument can have the
 876 value @code{omp_sched_static}, @code{omp_sched_dynamic},
 877 @code{omp_sched_guided} or @code{omp_sched_auto}.  Except for
 878 @code{omp_sched_auto}, the chunk size is set to the value of
 879 @var{chunk_size} if positive, or to the default value if zero or negative.
 880 For @code{omp_sched_auto} the @var{chunk_size} argument is ignored.
 881
 882 @item @emph{C/C++}
 883 @multitable @columnfractions .20 .80
 884 @item @emph{Prototype}: @tab @code{void omp_set_schedule(omp_sched_t kind, int chunk_size);}
 885 @end multitable
 886
 887 @item @emph{Fortran}:
 888 @multitable @columnfractions .20 .80
 889 @item @emph{Interface}: @tab @code{subroutine omp_set_schedule(kind, chunk_size)}
 890 @item                   @tab @code{integer(kind=omp_sched_kind) kind}
 891 @item                   @tab @code{integer chunk_size}
 892 @end multitable
 893
 894 @item @emph{See also}:
 895 @ref{omp_get_schedule}
 896 @ref{OMP_SCHEDULE}
 897
 898 @item @emph{Reference}:
 899 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.12.
 900 @end table
 901
 902
 903
 904 @node omp_get_schedule
 905 @subsection @code{omp_get_schedule} -- Obtain the runtime scheduling method
 906 @table @asis
 907 @item @emph{Description}:
 908 Obtain the runtime scheduling method.  The @var{kind} argument will be
 909 set to the value @code{omp_sched_static}, @code{omp_sched_dynamic},
 910 @code{omp_sched_guided} or @code{omp_sched_auto}.  The second argument,
 911 @var{chunk_size}, is set to the chunk size.
 912
 913 @item @emph{C/C++}
 914 @multitable @columnfractions .20 .80
 915 @item @emph{Prototype}: @tab @code{void omp_get_schedule(omp_sched_t *kind, int *chunk_size);}
 916 @end multitable
 917
 918 @item @emph{Fortran}:
 919 @multitable @columnfractions .20 .80
 920 @item @emph{Interface}: @tab @code{subroutine omp_get_schedule(kind, chunk_size)}
 921 @item                   @tab @code{integer(kind=omp_sched_kind) kind}
 922 @item                   @tab @code{integer chunk_size}
 923 @end multitable
 924
 925 @item @emph{See also}:
 926 @ref{omp_set_schedule}, @ref{OMP_SCHEDULE}
 927
 928 @item @emph{Reference}:
 929 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.13.
 930 @end table
 931
 932
 933 @node omp_get_teams_thread_limit
 934 @subsection @code{omp_get_teams_thread_limit} -- Maximum number of threads imposed by teams
 935 @table @asis
 936 @item @emph{Description}:
 937 Return the maximum number of threads that will be able to participate in
 938 each team created by a teams construct.
 939
 940 @item @emph{C/C++}:
 941 @multitable @columnfractions .20 .80
 942 @item @emph{Prototype}: @tab @code{int omp_get_teams_thread_limit(void);}
 943 @end multitable
 944
 945 @item @emph{Fortran}:
 946 @multitable @columnfractions .20 .80
 947 @item @emph{Interface}: @tab @code{integer function omp_get_teams_thread_limit()}
 948 @end multitable
 949
 950 @item @emph{See also}:
 951 @ref{omp_set_teams_thread_limit}, @ref{OMP_TEAMS_THREAD_LIMIT}
 952
 953 @item @emph{Reference}:
 954 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 3.4.6.
 955 @end table
 956
 957
 958
 959 @node omp_get_supported_active_levels
 960 @subsection @code{omp_get_supported_active_levels} -- Maximum number of active regions supported
 961 @table @asis
 962 @item @emph{Description}:
 963 This function returns the maximum number of nested, active parallel regions
 964 supported by this implementation.
 965
 966 @item @emph{C/C++}
 967 @multitable @columnfractions .20 .80
 968 @item @emph{Prototype}: @tab @code{int omp_get_supported_active_levels(void);}
 969 @end multitable
 970
 971 @item @emph{Fortran}:
 972 @multitable @columnfractions .20 .80
 973 @item @emph{Interface}: @tab @code{integer function omp_get_supported_active_levels()}
 974 @end multitable
 975
 976 @item @emph{See also}:
 977 @ref{omp_get_max_active_levels}, @ref{omp_set_max_active_levels}
 978
 979 @item @emph{Reference}:
 980 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.2.15.
 981 @end table
 982
 983
 984
 985 @node omp_set_max_active_levels
 986 @subsection @code{omp_set_max_active_levels} -- Limits the number of active parallel regions
 987 @table @asis
 988 @item @emph{Description}:
 989 This function limits the maximum allowed number of nested, active
 990 parallel regions.  @var{max_levels} must be less or equal to
 991 the value returned by @code{omp_get_supported_active_levels}.
 992
 993 @item @emph{C/C++}
 994 @multitable @columnfractions .20 .80
 995 @item @emph{Prototype}: @tab @code{void omp_set_max_active_levels(int max_levels);}
 996 @end multitable
 997
 998 @item @emph{Fortran}:
 999 @multitable @columnfractions .20 .80
1000 @item @emph{Interface}: @tab @code{subroutine omp_set_max_active_levels(max_levels)}
1001 @item                   @tab @code{integer max_levels}
1002 @end multitable
1003
1004 @item @emph{See also}:
1005 @ref{omp_get_max_active_levels}, @ref{omp_get_active_level},
1006 @ref{omp_get_supported_active_levels}
1007
1008 @item @emph{Reference}:
1009 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.15.
1010 @end table
1011
1012
1013
1014 @node omp_get_max_active_levels
1015 @subsection @code{omp_get_max_active_levels} -- Current maximum number of active regions
1016 @table @asis
1017 @item @emph{Description}:
1018 This function obtains the maximum allowed number of nested, active parallel regions.
1019
1020 @item @emph{C/C++}
1021 @multitable @columnfractions .20 .80
1022 @item @emph{Prototype}: @tab @code{int omp_get_max_active_levels(void);}
1023 @end multitable
1024
1025 @item @emph{Fortran}:
1026 @multitable @columnfractions .20 .80
1027 @item @emph{Interface}: @tab @code{integer function omp_get_max_active_levels()}
1028 @end multitable
1029
1030 @item @emph{See also}:
1031 @ref{omp_set_max_active_levels}, @ref{omp_get_active_level}
1032
1033 @item @emph{Reference}:
1034 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.16.
1035 @end table
1036
1037
1038 @node omp_get_level
1039 @subsection @code{omp_get_level} -- Obtain the current nesting level
1040 @table @asis
1041 @item @emph{Description}:
1042 This function returns the nesting level for the parallel blocks,
1043 which enclose the calling call.
1044
1045 @item @emph{C/C++}
1046 @multitable @columnfractions .20 .80
1047 @item @emph{Prototype}: @tab @code{int omp_get_level(void);}
1048 @end multitable
1049
1050 @item @emph{Fortran}:
1051 @multitable @columnfractions .20 .80
1052 @item @emph{Interface}: @tab @code{integer function omp_level()}
1053 @end multitable
1054
1055 @item @emph{See also}:
1056 @ref{omp_get_active_level}
1057
1058 @item @emph{Reference}:
1059 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.17.
1060 @end table
1061
1062
1063
1064 @node omp_get_ancestor_thread_num
1065 @subsection @code{omp_get_ancestor_thread_num} -- Ancestor thread ID
1066 @table @asis
1067 @item @emph{Description}:
1068 This function returns the thread identification number for the given
1069 nesting level of the current thread.  For values of @var{level} outside
1070 zero to @code{omp_get_level} -1 is returned; if @var{level} is
1071 @code{omp_get_level} the result is identical to @code{omp_get_thread_num}.
1072
1073 @item @emph{C/C++}
1074 @multitable @columnfractions .20 .80
1075 @item @emph{Prototype}: @tab @code{int omp_get_ancestor_thread_num(int level);}
1076 @end multitable
1077
1078 @item @emph{Fortran}:
1079 @multitable @columnfractions .20 .80
1080 @item @emph{Interface}: @tab @code{integer function omp_get_ancestor_thread_num(level)}
1081 @item                   @tab @code{integer level}
1082 @end multitable
1083
1084 @item @emph{See also}:
1085 @ref{omp_get_level}, @ref{omp_get_thread_num}, @ref{omp_get_team_size}
1086
1087 @item @emph{Reference}:
1088 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.18.
1089 @end table
1090
1091
1092
1093 @node omp_get_team_size
1094 @subsection @code{omp_get_team_size} -- Number of threads in a team
1095 @table @asis
1096 @item @emph{Description}:
1097 This function returns the number of threads in a thread team to which
1098 either the current thread or its ancestor belongs.  For values of @var{level}
1099 outside zero to @code{omp_get_level}, -1 is returned; if @var{level} is zero,
1100 1 is returned, and for @code{omp_get_level}, the result is identical
1101 to @code{omp_get_num_threads}.
1102
1103 @item @emph{C/C++}:
1104 @multitable @columnfractions .20 .80
1105 @item @emph{Prototype}: @tab @code{int omp_get_team_size(int level);}
1106 @end multitable
1107
1108 @item @emph{Fortran}:
1109 @multitable @columnfractions .20 .80
1110 @item @emph{Interface}: @tab @code{integer function omp_get_team_size(level)}
1111 @item                   @tab @code{integer level}
1112 @end multitable
1113
1114 @item @emph{See also}:
1115 @ref{omp_get_num_threads}, @ref{omp_get_level}, @ref{omp_get_ancestor_thread_num}
1116
1117 @item @emph{Reference}:
1118 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.19.
1119 @end table
1120
1121
1122
1123 @node omp_get_active_level
1124 @subsection @code{omp_get_active_level} -- Number of parallel regions
1125 @table @asis
1126 @item @emph{Description}:
1127 This function returns the nesting level for the active parallel blocks,
1128 which enclose the calling call.
1129
1130 @item @emph{C/C++}
1131 @multitable @columnfractions .20 .80
1132 @item @emph{Prototype}: @tab @code{int omp_get_active_level(void);}
1133 @end multitable
1134
1135 @item @emph{Fortran}:
1136 @multitable @columnfractions .20 .80
1137 @item @emph{Interface}: @tab @code{integer function omp_get_active_level()}
1138 @end multitable
1139
1140 @item @emph{See also}:
1141 @ref{omp_get_level}, @ref{omp_get_max_active_levels}, @ref{omp_set_max_active_levels}
1142
1143 @item @emph{Reference}:
1144 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.20.
1145 @end table
1146
1147
1148
1149 @node Thread Affinity Routines
1150 @section Thread Affinity Routines
1151
1152 Routines controlling and accessing thread-affinity policies.
1153 They have C linkage and do not throw exceptions.
1154
1155 @menu
1156 * omp_get_proc_bind::           Whether threads may be moved between CPUs
1157 @c * omp_get_num_places:: <fixme>
1158 @c * omp_get_place_num_procs:: <fixme>
1159 @c * omp_get_place_proc_ids:: <fixme>
1160 @c * omp_get_place_num:: <fixme>
1161 @c * omp_get_partition_num_places:: <fixme>
1162 @c * omp_get_partition_place_nums:: <fixme>
1163 @c * omp_set_affinity_format:: <fixme>
1164 @c * omp_get_affinity_format:: <fixme>
1165 @c * omp_display_affinity:: <fixme>
1166 @c * omp_capture_affinity:: <fixme>
1167 @end menu
1168
1169
1170
1171 @node omp_get_proc_bind
1172 @subsection @code{omp_get_proc_bind} -- Whether threads may be moved between CPUs
1173 @table @asis
1174 @item @emph{Description}:
1175 This functions returns the currently active thread affinity policy, which is
1176 set via @env{OMP_PROC_BIND}.  Possible values are @code{omp_proc_bind_false},
1177 @code{omp_proc_bind_true}, @code{omp_proc_bind_primary},
1178 @code{omp_proc_bind_master}, @code{omp_proc_bind_close} and @code{omp_proc_bind_spread},
1179 where @code{omp_proc_bind_master} is an alias for @code{omp_proc_bind_primary}.
1180
1181 @item @emph{C/C++}:
1182 @multitable @columnfractions .20 .80
1183 @item @emph{Prototype}: @tab @code{omp_proc_bind_t omp_get_proc_bind(void);}
1184 @end multitable
1185
1186 @item @emph{Fortran}:
1187 @multitable @columnfractions .20 .80
1188 @item @emph{Interface}: @tab @code{integer(kind=omp_proc_bind_kind) function omp_get_proc_bind()}
1189 @end multitable
1190
1191 @item @emph{See also}:
1192 @ref{OMP_PROC_BIND}, @ref{OMP_PLACES}, @ref{GOMP_CPU_AFFINITY},
1193
1194 @item @emph{Reference}:
1195 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.22.
1196 @end table
1197
1198
1199
1200 @node Teams Region Routines
1201 @section Teams Region Routines
1202
1203 Routines controlling the league of teams that are executed in a @code{teams}
1204 region.  They have C linkage and do not throw exceptions.
1205
1206 @menu
1207 * omp_get_num_teams::           Number of teams
1208 * omp_get_team_num::            Get team number
1209 * omp_set_num_teams::           Set upper teams limit for teams region
1210 * omp_get_max_teams::           Maximum number of teams for teams region
1211 * omp_set_teams_thread_limit::  Set upper thread limit for teams construct
1212 * omp_get_thread_limit::        Maximum number of threads
1213 @end menu
1214
1215
1216
1217 @node omp_get_num_teams
1218 @subsection @code{omp_get_num_teams} -- Number of teams
1219 @table @asis
1220 @item @emph{Description}:
1221 Returns the number of teams in the current team region.
1222
1223 @item @emph{C/C++}:
1224 @multitable @columnfractions .20 .80
1225 @item @emph{Prototype}: @tab @code{int omp_get_num_teams(void);}
1226 @end multitable
1227
1228 @item @emph{Fortran}:
1229 @multitable @columnfractions .20 .80
1230 @item @emph{Interface}: @tab @code{integer function omp_get_num_teams()}
1231 @end multitable
1232
1233 @item @emph{Reference}:
1234 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.32.
1235 @end table
1236
1237
1238
1239 @node omp_get_team_num
1240 @subsection @code{omp_get_team_num} -- Get team number
1241 @table @asis
1242 @item @emph{Description}:
1243 Returns the team number of the calling thread.
1244
1245 @item @emph{C/C++}:
1246 @multitable @columnfractions .20 .80
1247 @item @emph{Prototype}: @tab @code{int omp_get_team_num(void);}
1248 @end multitable
1249
1250 @item @emph{Fortran}:
1251 @multitable @columnfractions .20 .80
1252 @item @emph{Interface}: @tab @code{integer function omp_get_team_num()}
1253 @end multitable
1254
1255 @item @emph{Reference}:
1256 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.33.
1257 @end table
1258
1259
1260
1261 @node omp_set_num_teams
1262 @subsection @code{omp_set_num_teams} -- Set upper teams limit for teams construct
1263 @table @asis
1264 @item @emph{Description}:
1265 Specifies the upper bound for number of teams created by the teams construct
1266 which does not specify a @code{num_teams} clause.  The
1267 argument of @code{omp_set_num_teams} shall be a positive integer.
1268
1269 @item @emph{C/C++}:
1270 @multitable @columnfractions .20 .80
1271 @item @emph{Prototype}: @tab @code{void omp_set_num_teams(int num_teams);}
1272 @end multitable
1273
1274 @item @emph{Fortran}:
1275 @multitable @columnfractions .20 .80
1276 @item @emph{Interface}: @tab @code{subroutine omp_set_num_teams(num_teams)}
1277 @item                   @tab @code{integer, intent(in) :: num_teams}
1278 @end multitable
1279
1280 @item @emph{See also}:
1281 @ref{OMP_NUM_TEAMS}, @ref{omp_get_num_teams}, @ref{omp_get_max_teams}
1282
1283 @item @emph{Reference}:
1284 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 3.4.3.
1285 @end table
1286
1287
1288
1289 @node omp_get_max_teams
1290 @subsection @code{omp_get_max_teams} -- Maximum number of teams of teams region
1291 @table @asis
1292 @item @emph{Description}:
1293 Return the maximum number of teams used for the teams region
1294 that does not use the clause @code{num_teams}.
1295
1296 @item @emph{C/C++}:
1297 @multitable @columnfractions .20 .80
1298 @item @emph{Prototype}: @tab @code{int omp_get_max_teams(void);}
1299 @end multitable
1300
1301 @item @emph{Fortran}:
1302 @multitable @columnfractions .20 .80
1303 @item @emph{Interface}: @tab @code{integer function omp_get_max_teams()}
1304 @end multitable
1305
1306 @item @emph{See also}:
1307 @ref{omp_set_num_teams}, @ref{omp_get_num_teams}
1308
1309 @item @emph{Reference}:
1310 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 3.4.4.
1311 @end table
1312
1313
1314
1315 @node omp_set_teams_thread_limit
1316 @subsection @code{omp_set_teams_thread_limit} -- Set upper thread limit for teams construct
1317 @table @asis
1318 @item @emph{Description}:
1319 Specifies the upper bound for number of threads that will be available
1320 for each team created by the teams construct which does not specify a
1321 @code{thread_limit} clause.  The argument of
1322 @code{omp_set_teams_thread_limit} shall be a positive integer.
1323
1324 @item @emph{C/C++}:
1325 @multitable @columnfractions .20 .80
1326 @item @emph{Prototype}: @tab @code{void omp_set_teams_thread_limit(int thread_limit);}
1327 @end multitable
1328
1329 @item @emph{Fortran}:
1330 @multitable @columnfractions .20 .80
1331 @item @emph{Interface}: @tab @code{subroutine omp_set_teams_thread_limit(thread_limit)}
1332 @item                   @tab @code{integer, intent(in) :: thread_limit}
1333 @end multitable
1334
1335 @item @emph{See also}:
1336 @ref{OMP_TEAMS_THREAD_LIMIT}, @ref{omp_get_teams_thread_limit}, @ref{omp_get_thread_limit}
1337
1338 @item @emph{Reference}:
1339 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 3.4.5.
1340 @end table
1341
1342
1343
1344 @node omp_get_thread_limit
1345 @subsection @code{omp_get_thread_limit} -- Maximum number of threads
1346 @table @asis
1347 @item @emph{Description}:
1348 Return the maximum number of threads of the program.
1349
1350 @item @emph{C/C++}:
1351 @multitable @columnfractions .20 .80
1352 @item @emph{Prototype}: @tab @code{int omp_get_thread_limit(void);}
1353 @end multitable
1354
1355 @item @emph{Fortran}:
1356 @multitable @columnfractions .20 .80
1357 @item @emph{Interface}: @tab @code{integer function omp_get_thread_limit()}
1358 @end multitable
1359
1360 @item @emph{See also}:
1361 @ref{omp_get_max_threads}, @ref{OMP_THREAD_LIMIT}
1362
1363 @item @emph{Reference}:
1364 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.14.
1365 @end table
1366
1367
1368
1369 @node Tasking Routines
1370 @section Tasking Routines
1371
1372 Routines relating to explicit tasks.
1373 They have C linkage and do not throw exceptions.
1374
1375 @menu
1376 * omp_get_max_task_priority::   Maximum task priority value that can be set
1377 * omp_in_explicit_task::        Whether a given task is an explicit task
1378 * omp_in_final::                Whether in final or included task region
1379 @end menu
1380
1381
1382
1383 @node omp_get_max_task_priority
1384 @subsection @code{omp_get_max_task_priority} -- Maximum priority value
1385 that can be set for tasks.
1386 @table @asis
1387 @item @emph{Description}:
1388 This function obtains the maximum allowed priority number for tasks.
1389
1390 @item @emph{C/C++}
1391 @multitable @columnfractions .20 .80
1392 @item @emph{Prototype}: @tab @code{int omp_get_max_task_priority(void);}
1393 @end multitable
1394
1395 @item @emph{Fortran}:
1396 @multitable @columnfractions .20 .80
1397 @item @emph{Interface}: @tab @code{integer function omp_get_max_task_priority()}
1398 @end multitable
1399
1400 @item @emph{Reference}:
1401 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.29.
1402 @end table
1403
1404
1405
1406 @node omp_in_explicit_task
1407 @subsection @code{omp_in_explicit_task} -- Whether a given task is an explicit task
1408 @table @asis
1409 @item @emph{Description}:
1410 The function returns the @var{explicit-task-var} ICV; it returns true when the
1411 encountering task was generated by a task-generating construct such as
1412 @code{target}, @code{task} or @code{taskloop}.  Otherwise, the encountering task
1413 is in an implicit task region such as generated by the implicit or explicit
1414 @code{parallel} region and @code{omp_in_explicit_task} returns false.
1415
1416 @item @emph{C/C++}
1417 @multitable @columnfractions .20 .80
1418 @item @emph{Prototype}: @tab @code{int omp_in_explicit_task(void);}
1419 @end multitable
1420
1421 @item @emph{Fortran}:
1422 @multitable @columnfractions .20 .80
1423 @item @emph{Interface}: @tab @code{logical function omp_in_explicit_task()}
1424 @end multitable
1425
1426 @item @emph{Reference}:
1427 @uref{https://www.openmp.org, OpenMP specification v5.2}, Section 18.5.2.
1428 @end table
1429
1430
1431
1432 @node omp_in_final
1433 @subsection @code{omp_in_final} -- Whether in final or included task region
1434 @table @asis
1435 @item @emph{Description}:
1436 This function returns @code{true} if currently running in a final
1437 or included task region, @code{false} otherwise.  Here, @code{true}
1438 and @code{false} represent their language-specific counterparts.
1439
1440 @item @emph{C/C++}:
1441 @multitable @columnfractions .20 .80
1442 @item @emph{Prototype}: @tab @code{int omp_in_final(void);}
1443 @end multitable
1444
1445 @item @emph{Fortran}:
1446 @multitable @columnfractions .20 .80
1447 @item @emph{Interface}: @tab @code{logical function omp_in_final()}
1448 @end multitable
1449
1450 @item @emph{Reference}:
1451 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.21.
1452 @end table
1453
1454
1455
1456 @c @node Resource Relinquishing Routines
1457 @c @section Resource Relinquishing Routines
1458 @c
1459 @c Routines releasing resources used by the OpenMP runtime.
1460 @c They have C linkage and do not throw exceptions.
1461 @c
1462 @c @menu
1463 @c * omp_pause_resource:: <fixme>
1464 @c * omp_pause_resource_all:: <fixme>
1465 @c @end menu
1466
1467 @node Device Information Routines
1468 @section Device Information Routines
1469
1470 Routines related to devices available to an OpenMP program.
1471 They have C linkage and do not throw exceptions.
1472
1473 @menu
1474 * omp_get_num_procs::           Number of processors online
1475 @c * omp_get_max_progress_width:: <fixme>/TR11
1476 * omp_set_default_device::      Set the default device for target regions
1477 * omp_get_default_device::      Get the default device for target regions
1478 * omp_get_num_devices::         Number of target devices
1479 * omp_get_device_num::          Get device that current thread is running on
1480 * omp_is_initial_device::       Whether executing on the host device
1481 * omp_get_initial_device::      Device number of host device
1482 @end menu
1483
1484
1485
1486 @node omp_get_num_procs
1487 @subsection @code{omp_get_num_procs} -- Number of processors online
1488 @table @asis
1489 @item @emph{Description}:
1490 Returns the number of processors online on that device.
1491
1492 @item @emph{C/C++}:
1493 @multitable @columnfractions .20 .80
1494 @item @emph{Prototype}: @tab @code{int omp_get_num_procs(void);}
1495 @end multitable
1496
1497 @item @emph{Fortran}:
1498 @multitable @columnfractions .20 .80
1499 @item @emph{Interface}: @tab @code{integer function omp_get_num_procs()}
1500 @end multitable
1501
1502 @item @emph{Reference}:
1503 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.5.
1504 @end table
1505
1506
1507
1508 @node omp_set_default_device
1509 @subsection @code{omp_set_default_device} -- Set the default device for target regions
1510 @table @asis
1511 @item @emph{Description}:
1512 Set the default device for target regions without device clause.  The argument
1513 shall be a nonnegative device number.
1514
1515 @item @emph{C/C++}:
1516 @multitable @columnfractions .20 .80
1517 @item @emph{Prototype}: @tab @code{void omp_set_default_device(int device_num);}
1518 @end multitable
1519
1520 @item @emph{Fortran}:
1521 @multitable @columnfractions .20 .80
1522 @item @emph{Interface}: @tab @code{subroutine omp_set_default_device(device_num)}
1523 @item                   @tab @code{integer device_num}
1524 @end multitable
1525
1526 @item @emph{See also}:
1527 @ref{OMP_DEFAULT_DEVICE}, @ref{omp_get_default_device}
1528
1529 @item @emph{Reference}:
1530 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.29.
1531 @end table
1532
1533
1534
1535 @node omp_get_default_device
1536 @subsection @code{omp_get_default_device} -- Get the default device for target regions
1537 @table @asis
1538 @item @emph{Description}:
1539 Get the default device for target regions without device clause.
1540
1541 @item @emph{C/C++}:
1542 @multitable @columnfractions .20 .80
1543 @item @emph{Prototype}: @tab @code{int omp_get_default_device(void);}
1544 @end multitable
1545
1546 @item @emph{Fortran}:
1547 @multitable @columnfractions .20 .80
1548 @item @emph{Interface}: @tab @code{integer function omp_get_default_device()}
1549 @end multitable
1550
1551 @item @emph{See also}:
1552 @ref{OMP_DEFAULT_DEVICE}, @ref{omp_set_default_device}
1553
1554 @item @emph{Reference}:
1555 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.30.
1556 @end table
1557
1558
1559
1560 @node omp_get_num_devices
1561 @subsection @code{omp_get_num_devices} -- Number of target devices
1562 @table @asis
1563 @item @emph{Description}:
1564 Returns the number of target devices.
1565
1566 @item @emph{C/C++}:
1567 @multitable @columnfractions .20 .80
1568 @item @emph{Prototype}: @tab @code{int omp_get_num_devices(void);}
1569 @end multitable
1570
1571 @item @emph{Fortran}:
1572 @multitable @columnfractions .20 .80
1573 @item @emph{Interface}: @tab @code{integer function omp_get_num_devices()}
1574 @end multitable
1575
1576 @item @emph{Reference}:
1577 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.31.
1578 @end table
1579
1580
1581
1582 @node omp_get_device_num
1583 @subsection @code{omp_get_device_num} -- Return device number of current device
1584 @table @asis
1585 @item @emph{Description}:
1586 This function returns a device number that represents the device that the
1587 current thread is executing on. For OpenMP 5.0, this must be equal to the
1588 value returned by the @code{omp_get_initial_device} function when called
1589 from the host.
1590
1591 @item @emph{C/C++}
1592 @multitable @columnfractions .20 .80
1593 @item @emph{Prototype}: @tab @code{int omp_get_device_num(void);}
1594 @end multitable
1595
1596 @item @emph{Fortran}:
1597 @multitable @columnfractions .20 .80
1598 @item @emph{Interface}: @tab @code{integer function omp_get_device_num()}
1599 @end multitable
1600
1601 @item @emph{See also}:
1602 @ref{omp_get_initial_device}
1603
1604 @item @emph{Reference}:
1605 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.2.37.
1606 @end table
1607
1608
1609
1610 @node omp_is_initial_device
1611 @subsection @code{omp_is_initial_device} -- Whether executing on the host device
1612 @table @asis
1613 @item @emph{Description}:
1614 This function returns @code{true} if currently running on the host device,
1615 @code{false} otherwise.  Here, @code{true} and @code{false} represent
1616 their language-specific counterparts.
1617
1618 @item @emph{C/C++}:
1619 @multitable @columnfractions .20 .80
1620 @item @emph{Prototype}: @tab @code{int omp_is_initial_device(void);}
1621 @end multitable
1622
1623 @item @emph{Fortran}:
1624 @multitable @columnfractions .20 .80
1625 @item @emph{Interface}: @tab @code{logical function omp_is_initial_device()}
1626 @end multitable
1627
1628 @item @emph{Reference}:
1629 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.34.
1630 @end table
1631
1632
1633
1634 @node omp_get_initial_device
1635 @subsection @code{omp_get_initial_device} -- Return device number of initial device
1636 @table @asis
1637 @item @emph{Description}:
1638 This function returns a device number that represents the host device.
1639 For OpenMP 5.1, this must be equal to the value returned by the
1640 @code{omp_get_num_devices} function.
1641
1642 @item @emph{C/C++}
1643 @multitable @columnfractions .20 .80
1644 @item @emph{Prototype}: @tab @code{int omp_get_initial_device(void);}
1645 @end multitable
1646
1647 @item @emph{Fortran}:
1648 @multitable @columnfractions .20 .80
1649 @item @emph{Interface}: @tab @code{integer function omp_get_initial_device()}
1650 @end multitable
1651
1652 @item @emph{See also}:
1653 @ref{omp_get_num_devices}
1654
1655 @item @emph{Reference}:
1656 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.35.
1657 @end table
1658
1659
1660
1661 @node Device Memory Routines
1662 @section Device Memory Routines
1663
1664 Routines related to memory allocation and managing corresponding
1665 pointers on devices. They have C linkage and do not throw exceptions.
1666
1667 @menu
1668 * omp_target_alloc:: Allocate device memory
1669 * omp_target_free:: Free device memory
1670 * omp_target_is_present:: Check whether storage is mapped
1671 @c * omp_target_is_accessible:: <fixme>
1672 @c * omp_target_memcpy:: <fixme>
1673 @c * omp_target_memcpy_rect:: <fixme>
1674 @c * omp_target_memcpy_async:: <fixme>
1675 @c * omp_target_memcpy_rect_async:: <fixme>
1676 @c * omp_target_memset:: <fixme>/TR12
1677 @c * omp_target_memset_async:: <fixme>/TR12
1678 * omp_target_associate_ptr:: Associate a device pointer with a host pointer
1679 * omp_target_disassociate_ptr:: Remove device--host pointer association
1680 * omp_get_mapped_ptr:: Return device pointer to a host pointer
1681 @end menu
1682
1683
1684
1685 @node omp_target_alloc
1686 @subsection @code{omp_target_alloc} -- Allocate device memory
1687 @table @asis
1688 @item @emph{Description}:
1689 This routine allocates @var{size} bytes of memory in the device environment
1690 associated with the device number @var{device_num}.  If successful, a device
1691 pointer is returned, otherwise a null pointer.
1692
1693 In GCC, when the device is the host or the device shares memory with the host,
1694 the memory is allocated on the host; in that case, when @var{size} is zero,
1695 either NULL or a unique pointer value that can later be successfully passed to
1696 @code{omp_target_free} is returned.  When the allocation is not performed on
1697 the host, a null pointer is returned when @var{size} is zero; in that case,
1698 additionally a diagnostic might be printed to standard error (stderr).
1699
1700 Running this routine in a @code{target} region except on the initial device
1701 is not supported.
1702
1703 @item @emph{C/C++}
1704 @multitable @columnfractions .20 .80
1705 @item @emph{Prototype}: @tab @code{void *omp_target_alloc(size_t size, int device_num)}
1706 @end multitable
1707
1708 @item @emph{Fortran}:
1709 @multitable @columnfractions .20 .80
1710 @item @emph{Interface}: @tab @code{type(c_ptr) function omp_target_alloc(size, device_num) bind(C)}
1711 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int, c_size_t}
1712 @item                   @tab @code{integer(c_size_t), value :: size}
1713 @item                   @tab @code{integer(c_int), value :: device_num}
1714 @end multitable
1715
1716 @item @emph{See also}:
1717 @ref{omp_target_free}, @ref{omp_target_associate_ptr}
1718
1719 @item @emph{Reference}:
1720 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.1
1721 @end table
1722
1723
1724
1725 @node omp_target_free
1726 @subsection @code{omp_target_free} -- Free device memory
1727 @table @asis
1728 @item @emph{Description}:
1729 This routine frees memory allocated by the @code{omp_target_alloc} routine.
1730 The @var{device_ptr} argument must be either a null pointer or a device pointer
1731 returned by @code{omp_target_alloc} for the specified @code{device_num}.  The
1732 device number @var{device_num} must be a conforming device number.
1733
1734 Running this routine in a @code{target} region except on the initial device
1735 is not supported.
1736
1737 @item @emph{C/C++}
1738 @multitable @columnfractions .20 .80
1739 @item @emph{Prototype}: @tab @code{void omp_target_free(void *device_ptr, int device_num)}
1740 @end multitable
1741
1742 @item @emph{Fortran}:
1743 @multitable @columnfractions .20 .80
1744 @item @emph{Interface}: @tab @code{subroutine omp_target_free(device_ptr, device_num) bind(C)}
1745 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int}
1746 @item                   @tab @code{type(c_ptr), value :: device_ptr}
1747 @item                   @tab @code{integer(c_int), value :: device_num}
1748 @end multitable
1749
1750 @item @emph{See also}:
1751 @ref{omp_target_alloc}, @ref{omp_target_disassociate_ptr}
1752
1753 @item @emph{Reference}:
1754 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.2
1755 @end table
1756
1757
1758
1759 @node omp_target_is_present
1760 @subsection @code{omp_target_is_present} -- Check whether storage is mapped
1761 @table @asis
1762 @item @emph{Description}:
1763 This routine tests whether storage, identified by the host pointer @var{ptr}
1764 is mapped to the device specified by @var{device_num}.  If so, it returns
1765 @emph{true} and otherwise @emph{false}.
1766
1767 In GCC, this includes self mapping such that @code{omp_target_is_present}
1768 returns @emph{true} when @var{device_num} specifies the host or when the host
1769 and the device share memory.  If @var{ptr} is a null pointer, @var{true} is
1770 returned and if @var{device_num} is an invalid device number, @var{false} is
1771 returned.
1772
1773 If those conditions do not apply, @emph{true} is returned if the association has
1774 been established by an explicit or implicit @code{map} clause, the
1775 @code{declare target} directive or a call to the @code{omp_target_associate_ptr}
1776 routine.
1777
1778 Running this routine in a @code{target} region except on the initial device
1779 is not supported.
1780
1781 @item @emph{C/C++}
1782 @multitable @columnfractions .20 .80
1783 @item @emph{Prototype}: @tab @code{int omp_target_is_present(const void *ptr,}
1784 @item                   @tab @code{                          int device_num)}
1785 @end multitable
1786
1787 @item @emph{Fortran}:
1788 @multitable @columnfractions .20 .80
1789 @item @emph{Interface}: @tab @code{integer(c_int) function omp_target_is_present(ptr, &}
1790 @item                   @tab @code{    device_num) bind(C)}
1791 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int}
1792 @item                   @tab @code{type(c_ptr), value :: ptr}
1793 @item                   @tab @code{integer(c_int), value :: device_num}
1794 @end multitable
1795
1796 @item @emph{See also}:
1797 @ref{omp_target_associate_ptr}
1798
1799 @item @emph{Reference}:
1800 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.3
1801 @end table
1802
1803
1804
1805 @node omp_target_associate_ptr
1806 @subsection @code{omp_target_associate_ptr} -- Associate a device pointer with a host pointer
1807 @table @asis
1808 @item @emph{Description}:
1809 This routine associates storage on the host with storage on a device identified
1810 by @var{device_num}.  The device pointer is usually obtained by calling
1811 @code{omp_target_alloc} or by other means (but not by using the @code{map}
1812 clauses or the @code{declare target} directive).  The host pointer should point
1813 to memory that has a storage size of at least @var{size}.
1814
1815 The @var{device_offset} parameter specifies the offset into @var{device_ptr}
1816 that is used as the base address for the device side of the mapping; the
1817 storage size should be at least @var{device_offset} plus @var{size}.
1818
1819 After the association, the host pointer can be used in a @code{map} clause and
1820 in the @code{to} and @code{from} clauses of the @code{target update} directive
1821 to transfer data between the associated pointers. The reference count of such
1822 associated storage is infinite.  The association can be removed by calling
1823 @code{omp_target_disassociate_ptr} which should be done before the lifetime
1824 of either either storage ends.
1825
1826 The routine returns nonzero (@code{EINVAL}) when the @var{device_num} invalid,
1827 for when the initial device or the associated device shares memory with the
1828 host.  @code{omp_target_associate_ptr} returns zero if @var{host_ptr} points
1829 into already associated storage that is fully inside of a previously associated
1830 memory.  Otherwise, if the association was successful zero is returned; if none
1831 of the cases above apply, nonzero (@code{EINVAL}) is returned.
1832
1833 The @code{omp_target_is_present} routine can be used to test whether
1834 associated storage for a device pointer exists.
1835
1836 Running this routine in a @code{target} region except on the initial device
1837 is not supported.
1838
1839 @item @emph{C/C++}
1840 @multitable @columnfractions .20 .80
1841 @item @emph{Prototype}: @tab @code{int omp_target_associate_ptr(const void *host_ptr,}
1842 @item                   @tab @code{                             const void *device_ptr,}
1843 @item                   @tab @code{                             size_t size,}
1844 @item                   @tab @code{                             size_t device_offset,}
1845 @item                   @tab @code{                             int device_num)}
1846 @end multitable
1847
1848 @item @emph{Fortran}:
1849 @multitable @columnfractions .20 .80
1850 @item @emph{Interface}: @tab @code{integer(c_int) function omp_target_associate_ptr(host_ptr, &}
1851 @item                   @tab @code{    device_ptr, size, device_offset, device_num) bind(C)}
1852 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int, c_size_t}
1853 @item                   @tab @code{type(c_ptr), value :: host_ptr, device_ptr}
1854 @item                   @tab @code{integer(c_size_t), value :: size, device_offset}
1855 @item                   @tab @code{integer(c_int), value :: device_num}
1856 @end multitable
1857
1858 @item @emph{See also}:
1859 @ref{omp_target_disassociate_ptr}, @ref{omp_target_is_present},
1860 @ref{omp_target_alloc}
1861
1862 @item @emph{Reference}:
1863 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.9
1864 @end table
1865
1866
1867
1868 @node omp_target_disassociate_ptr
1869 @subsection @code{omp_target_disassociate_ptr} -- Remove device--host pointer association
1870 @table @asis
1871 @item @emph{Description}:
1872 This routine removes the storage association established by calling
1873 @code{omp_target_associate_ptr} and sets the reference count to zero,
1874 even if @code{omp_target_associate_ptr} was invoked multiple times for
1875 for host pointer @code{ptr}.  If applicable, the device memory needs
1876 to be freed by the user.
1877
1878 If an associated device storage location for the @var{device_num} was
1879 found and has infinite reference count, the association is removed and
1880 zero is returned.  In all other cases, nonzero (@code{EINVAL}) is returned
1881 and no other action is taken.
1882
1883 Note that passing a host pointer where the association to the device pointer
1884 was established with the @code{declare target} directive yields undefined
1885 behavior.
1886
1887 Running this routine in a @code{target} region except on the initial device
1888 is not supported.
1889
1890 @item @emph{C/C++}
1891 @multitable @columnfractions .20 .80
1892 @item @emph{Prototype}: @tab @code{int omp_target_disassociate_ptr(const void *ptr,}
1893 @item                   @tab @code{                                int device_num)}
1894 @end multitable
1895
1896 @item @emph{Fortran}:
1897 @multitable @columnfractions .20 .80
1898 @item @emph{Interface}: @tab @code{integer(c_int) function omp_target_disassociate_ptr(ptr, &}
1899 @item                   @tab @code{    device_num) bind(C)}
1900 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int}
1901 @item                   @tab @code{type(c_ptr), value :: ptr}
1902 @item                   @tab @code{integer(c_int), value :: device_num}
1903 @end multitable
1904
1905 @item @emph{See also}:
1906 @ref{omp_target_associate_ptr}
1907
1908 @item @emph{Reference}:
1909 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.10
1910 @end table
1911
1912
1913
1914 @node omp_get_mapped_ptr
1915 @subsection @code{omp_get_mapped_ptr} -- Return device pointer to a host pointer
1916 @table @asis
1917 @item @emph{Description}:
1918 If the device number is refers to the initial device or to a device with
1919 memory accessible from the host (shared memory), the @code{omp_get_mapped_ptr}
1920 routines returnes the value of the passed @var{ptr}.  Otherwise, if associated
1921 storage to the passed host pointer @var{ptr} exists on device associated with
1922 @var{device_num}, it returns that pointer. In all other cases and in cases of
1923 an error, a null pointer is returned.
1924
1925 The association of storage location is established either via an explicit or
1926 implicit @code{map} clause, the @code{declare target} directive or the
1927 @code{omp_target_associate_ptr} routine.
1928
1929 Running this routine in a @code{target} region except on the initial device
1930 is not supported.
1931
1932 @item @emph{C/C++}
1933 @multitable @columnfractions .20 .80
1934 @item @emph{Prototype}: @tab @code{void *omp_get_mapped_ptr(const void *ptr, int device_num);}
1935 @end multitable
1936
1937 @item @emph{Fortran}:
1938 @multitable @columnfractions .20 .80
1939 @item @emph{Interface}: @tab @code{type(c_ptr) function omp_get_mapped_ptr(ptr, device_num) bind(C)}
1940 @item                   @tab @code{use, intrinsic :: iso_c_binding, only: c_ptr, c_int}
1941 @item                   @tab @code{type(c_ptr), value :: ptr}
1942 @item                   @tab @code{integer(c_int), value :: device_num}
1943 @end multitable
1944
1945 @item @emph{See also}:
1946 @ref{omp_target_associate_ptr}
1947
1948 @item @emph{Reference}:
1949 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 18.8.11
1950 @end table
1951
1952
1953
1954 @node Lock Routines
1955 @section Lock Routines
1956
1957 Initialize, set, test, unset and destroy simple and nested locks.
1958 The routines have C linkage and do not throw exceptions.
1959
1960 @menu
1961 * omp_init_lock::            Initialize simple lock
1962 * omp_init_nest_lock::       Initialize nested lock
1963 @c * omp_init_lock_with_hint:: <fixme>
1964 @c * omp_init_nest_lock_with_hint:: <fixme>
1965 * omp_destroy_lock::         Destroy simple lock
1966 * omp_destroy_nest_lock::    Destroy nested lock
1967 * omp_set_lock::             Wait for and set simple lock
1968 * omp_set_nest_lock::        Wait for and set simple lock
1969 * omp_unset_lock::           Unset simple lock
1970 * omp_unset_nest_lock::      Unset nested lock
1971 * omp_test_lock::            Test and set simple lock if available
1972 * omp_test_nest_lock::       Test and set nested lock if available
1973 @end menu
1974
1975
1976
1977 @node omp_init_lock
1978 @subsection @code{omp_init_lock} -- Initialize simple lock
1979 @table @asis
1980 @item @emph{Description}:
1981 Initialize a simple lock.  After initialization, the lock is in
1982 an unlocked state.
1983
1984 @item @emph{C/C++}:
1985 @multitable @columnfractions .20 .80
1986 @item @emph{Prototype}: @tab @code{void omp_init_lock(omp_lock_t *lock);}
1987 @end multitable
1988
1989 @item @emph{Fortran}:
1990 @multitable @columnfractions .20 .80
1991 @item @emph{Interface}: @tab @code{subroutine omp_init_lock(svar)}
1992 @item                   @tab @code{integer(omp_lock_kind), intent(out) :: svar}
1993 @end multitable
1994
1995 @item @emph{See also}:
1996 @ref{omp_destroy_lock}
1997
1998 @item @emph{Reference}:
1999 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.1.
2000 @end table
2001
2002
2003
2004 @node omp_init_nest_lock
2005 @subsection @code{omp_init_nest_lock} -- Initialize nested lock
2006 @table @asis
2007 @item @emph{Description}:
2008 Initialize a nested lock.  After initialization, the lock is in
2009 an unlocked state and the nesting count is set to zero.
2010
2011 @item @emph{C/C++}:
2012 @multitable @columnfractions .20 .80
2013 @item @emph{Prototype}: @tab @code{void omp_init_nest_lock(omp_nest_lock_t *lock);}
2014 @end multitable
2015
2016 @item @emph{Fortran}:
2017 @multitable @columnfractions .20 .80
2018 @item @emph{Interface}: @tab @code{subroutine omp_init_nest_lock(nvar)}
2019 @item                   @tab @code{integer(omp_nest_lock_kind), intent(out) :: nvar}
2020 @end multitable
2021
2022 @item @emph{See also}:
2023 @ref{omp_destroy_nest_lock}
2024
2025 @item @emph{Reference}:
2026 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.1.
2027 @end table
2028
2029
2030
2031 @node omp_destroy_lock
2032 @subsection @code{omp_destroy_lock} -- Destroy simple lock
2033 @table @asis
2034 @item @emph{Description}:
2035 Destroy a simple lock.  In order to be destroyed, a simple lock must be
2036 in the unlocked state.
2037
2038 @item @emph{C/C++}:
2039 @multitable @columnfractions .20 .80
2040 @item @emph{Prototype}: @tab @code{void omp_destroy_lock(omp_lock_t *lock);}
2041 @end multitable
2042
2043 @item @emph{Fortran}:
2044 @multitable @columnfractions .20 .80
2045 @item @emph{Interface}: @tab @code{subroutine omp_destroy_lock(svar)}
2046 @item                   @tab @code{integer(omp_lock_kind), intent(inout) :: svar}
2047 @end multitable
2048
2049 @item @emph{See also}:
2050 @ref{omp_init_lock}
2051
2052 @item @emph{Reference}:
2053 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.3.
2054 @end table
2055
2056
2057
2058 @node omp_destroy_nest_lock
2059 @subsection @code{omp_destroy_nest_lock} -- Destroy nested lock
2060 @table @asis
2061 @item @emph{Description}:
2062 Destroy a nested lock.  In order to be destroyed, a nested lock must be
2063 in the unlocked state and its nesting count must equal zero.
2064
2065 @item @emph{C/C++}:
2066 @multitable @columnfractions .20 .80
2067 @item @emph{Prototype}: @tab @code{void omp_destroy_nest_lock(omp_nest_lock_t *);}
2068 @end multitable
2069
2070 @item @emph{Fortran}:
2071 @multitable @columnfractions .20 .80
2072 @item @emph{Interface}: @tab @code{subroutine omp_destroy_nest_lock(nvar)}
2073 @item                   @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar}
2074 @end multitable
2075
2076 @item @emph{See also}:
2077 @ref{omp_init_lock}
2078
2079 @item @emph{Reference}:
2080 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.3.
2081 @end table
2082
2083
2084
2085 @node omp_set_lock
2086 @subsection @code{omp_set_lock} -- Wait for and set simple lock
2087 @table @asis
2088 @item @emph{Description}:
2089 Before setting a simple lock, the lock variable must be initialized by
2090 @code{omp_init_lock}.  The calling thread is blocked until the lock
2091 is available.  If the lock is already held by the current thread,
2092 a deadlock occurs.
2093
2094 @item @emph{C/C++}:
2095 @multitable @columnfractions .20 .80
2096 @item @emph{Prototype}: @tab @code{void omp_set_lock(omp_lock_t *lock);}
2097 @end multitable
2098
2099 @item @emph{Fortran}:
2100 @multitable @columnfractions .20 .80
2101 @item @emph{Interface}: @tab @code{subroutine omp_set_lock(svar)}
2102 @item                   @tab @code{integer(omp_lock_kind), intent(inout) :: svar}
2103 @end multitable
2104
2105 @item @emph{See also}:
2106 @ref{omp_init_lock}, @ref{omp_test_lock}, @ref{omp_unset_lock}
2107
2108 @item @emph{Reference}:
2109 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.4.
2110 @end table
2111
2112
2113
2114 @node omp_set_nest_lock
2115 @subsection @code{omp_set_nest_lock} -- Wait for and set nested lock
2116 @table @asis
2117 @item @emph{Description}:
2118 Before setting a nested lock, the lock variable must be initialized by
2119 @code{omp_init_nest_lock}.  The calling thread is blocked until the lock
2120 is available.  If the lock is already held by the current thread, the
2121 nesting count for the lock is incremented.
2122
2123 @item @emph{C/C++}:
2124 @multitable @columnfractions .20 .80
2125 @item @emph{Prototype}: @tab @code{void omp_set_nest_lock(omp_nest_lock_t *lock);}
2126 @end multitable
2127
2128 @item @emph{Fortran}:
2129 @multitable @columnfractions .20 .80
2130 @item @emph{Interface}: @tab @code{subroutine omp_set_nest_lock(nvar)}
2131 @item                   @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar}
2132 @end multitable
2133
2134 @item @emph{See also}:
2135 @ref{omp_init_nest_lock}, @ref{omp_unset_nest_lock}
2136
2137 @item @emph{Reference}:
2138 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.4.
2139 @end table
2140
2141
2142
2143 @node omp_unset_lock
2144 @subsection @code{omp_unset_lock} -- Unset simple lock
2145 @table @asis
2146 @item @emph{Description}:
2147 A simple lock about to be unset must have been locked by @code{omp_set_lock}
2148 or @code{omp_test_lock} before.  In addition, the lock must be held by the
2149 thread calling @code{omp_unset_lock}.  Then, the lock becomes unlocked.  If one
2150 or more threads attempted to set the lock before, one of them is chosen to,
2151 again, set the lock to itself.
2152
2153 @item @emph{C/C++}:
2154 @multitable @columnfractions .20 .80
2155 @item @emph{Prototype}: @tab @code{void omp_unset_lock(omp_lock_t *lock);}
2156 @end multitable
2157
2158 @item @emph{Fortran}:
2159 @multitable @columnfractions .20 .80
2160 @item @emph{Interface}: @tab @code{subroutine omp_unset_lock(svar)}
2161 @item                   @tab @code{integer(omp_lock_kind), intent(inout) :: svar}
2162 @end multitable
2163
2164 @item @emph{See also}:
2165 @ref{omp_set_lock}, @ref{omp_test_lock}
2166
2167 @item @emph{Reference}:
2168 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.5.
2169 @end table
2170
2171
2172
2173 @node omp_unset_nest_lock
2174 @subsection @code{omp_unset_nest_lock} -- Unset nested lock
2175 @table @asis
2176 @item @emph{Description}:
2177 A nested lock about to be unset must have been locked by @code{omp_set_nested_lock}
2178 or @code{omp_test_nested_lock} before.  In addition, the lock must be held by the
2179 thread calling @code{omp_unset_nested_lock}.  If the nesting count drops to zero, the
2180 lock becomes unlocked.  If one ore more threads attempted to set the lock before,
2181 one of them is chosen to, again, set the lock to itself.
2182
2183 @item @emph{C/C++}:
2184 @multitable @columnfractions .20 .80
2185 @item @emph{Prototype}: @tab @code{void omp_unset_nest_lock(omp_nest_lock_t *lock);}
2186 @end multitable
2187
2188 @item @emph{Fortran}:
2189 @multitable @columnfractions .20 .80
2190 @item @emph{Interface}: @tab @code{subroutine omp_unset_nest_lock(nvar)}
2191 @item                   @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar}
2192 @end multitable
2193
2194 @item @emph{See also}:
2195 @ref{omp_set_nest_lock}
2196
2197 @item @emph{Reference}:
2198 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.5.
2199 @end table
2200
2201
2202
2203 @node omp_test_lock
2204 @subsection @code{omp_test_lock} -- Test and set simple lock if available
2205 @table @asis
2206 @item @emph{Description}:
2207 Before setting a simple lock, the lock variable must be initialized by
2208 @code{omp_init_lock}.  Contrary to @code{omp_set_lock}, @code{omp_test_lock}
2209 does not block if the lock is not available.  This function returns
2210 @code{true} upon success, @code{false} otherwise.  Here, @code{true} and
2211 @code{false} represent their language-specific counterparts.
2212
2213 @item @emph{C/C++}:
2214 @multitable @columnfractions .20 .80
2215 @item @emph{Prototype}: @tab @code{int omp_test_lock(omp_lock_t *lock);}
2216 @end multitable
2217
2218 @item @emph{Fortran}:
2219 @multitable @columnfractions .20 .80
2220 @item @emph{Interface}: @tab @code{logical function omp_test_lock(svar)}
2221 @item                   @tab @code{integer(omp_lock_kind), intent(inout) :: svar}
2222 @end multitable
2223
2224 @item @emph{See also}:
2225 @ref{omp_init_lock}, @ref{omp_set_lock}, @ref{omp_set_lock}
2226
2227 @item @emph{Reference}:
2228 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.6.
2229 @end table
2230
2231
2232
2233 @node omp_test_nest_lock
2234 @subsection @code{omp_test_nest_lock} -- Test and set nested lock if available
2235 @table @asis
2236 @item @emph{Description}:
2237 Before setting a nested lock, the lock variable must be initialized by
2238 @code{omp_init_nest_lock}.  Contrary to @code{omp_set_nest_lock},
2239 @code{omp_test_nest_lock} does not block if the lock is not available.
2240 If the lock is already held by the current thread, the new nesting count
2241 is returned.  Otherwise, the return value equals zero.
2242
2243 @item @emph{C/C++}:
2244 @multitable @columnfractions .20 .80
2245 @item @emph{Prototype}: @tab @code{int omp_test_nest_lock(omp_nest_lock_t *lock);}
2246 @end multitable
2247
2248 @item @emph{Fortran}:
2249 @multitable @columnfractions .20 .80
2250 @item @emph{Interface}: @tab @code{logical function omp_test_nest_lock(nvar)}
2251 @item                   @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar}
2252 @end multitable
2253
2254
2255 @item @emph{See also}:
2256 @ref{omp_init_lock}, @ref{omp_set_lock}, @ref{omp_set_lock}
2257
2258 @item @emph{Reference}:
2259 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.6.
2260 @end table
2261
2262
2263
2264 @node Timing Routines
2265 @section Timing Routines
2266
2267 Portable, thread-based, wall clock timer.
2268 The routines have C linkage and do not throw exceptions.
2269
2270 @menu
2271 * omp_get_wtick::            Get timer precision.
2272 * omp_get_wtime::            Elapsed wall clock time.
2273 @end menu
2274
2275
2276
2277 @node omp_get_wtick
2278 @subsection @code{omp_get_wtick} -- Get timer precision
2279 @table @asis
2280 @item @emph{Description}:
2281 Gets the timer precision, i.e., the number of seconds between two
2282 successive clock ticks.
2283
2284 @item @emph{C/C++}:
2285 @multitable @columnfractions .20 .80
2286 @item @emph{Prototype}: @tab @code{double omp_get_wtick(void);}
2287 @end multitable
2288
2289 @item @emph{Fortran}:
2290 @multitable @columnfractions .20 .80
2291 @item @emph{Interface}: @tab @code{double precision function omp_get_wtick()}
2292 @end multitable
2293
2294 @item @emph{See also}:
2295 @ref{omp_get_wtime}
2296
2297 @item @emph{Reference}:
2298 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.4.2.
2299 @end table
2300
2301
2302
2303 @node omp_get_wtime
2304 @subsection @code{omp_get_wtime} -- Elapsed wall clock time
2305 @table @asis
2306 @item @emph{Description}:
2307 Elapsed wall clock time in seconds.  The time is measured per thread, no
2308 guarantee can be made that two distinct threads measure the same time.
2309 Time is measured from some "time in the past", which is an arbitrary time
2310 guaranteed not to change during the execution of the program.
2311
2312 @item @emph{C/C++}:
2313 @multitable @columnfractions .20 .80
2314 @item @emph{Prototype}: @tab @code{double omp_get_wtime(void);}
2315 @end multitable
2316
2317 @item @emph{Fortran}:
2318 @multitable @columnfractions .20 .80
2319 @item @emph{Interface}: @tab @code{double precision function omp_get_wtime()}
2320 @end multitable
2321
2322 @item @emph{See also}:
2323 @ref{omp_get_wtick}
2324
2325 @item @emph{Reference}:
2326 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.4.1.
2327 @end table
2328
2329
2330
2331 @node Event Routine
2332 @section Event Routine
2333
2334 Support for event objects.
2335 The routine has C linkage and do not throw exceptions.
2336
2337 @menu
2338 * omp_fulfill_event::        Fulfill and destroy an OpenMP event.
2339 @end menu
2340
2341
2342
2343 @node omp_fulfill_event
2344 @subsection @code{omp_fulfill_event} -- Fulfill and destroy an OpenMP event
2345 @table @asis
2346 @item @emph{Description}:
2347 Fulfill the event associated with the event handle argument.  Currently, it
2348 is only used to fulfill events generated by detach clauses on task
2349 constructs - the effect of fulfilling the event is to allow the task to
2350 complete.
2351
2352 The result of calling @code{omp_fulfill_event} with an event handle other
2353 than that generated by a detach clause is undefined.  Calling it with an
2354 event handle that has already been fulfilled is also undefined.
2355
2356 @item @emph{C/C++}:
2357 @multitable @columnfractions .20 .80
2358 @item @emph{Prototype}: @tab @code{void omp_fulfill_event(omp_event_handle_t event);}
2359 @end multitable
2360
2361 @item @emph{Fortran}:
2362 @multitable @columnfractions .20 .80
2363 @item @emph{Interface}: @tab @code{subroutine omp_fulfill_event(event)}
2364 @item                   @tab @code{integer (kind=omp_event_handle_kind) :: event}
2365 @end multitable
2366
2367 @item @emph{Reference}:
2368 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.5.1.
2369 @end table
2370
2371
2372
2373 @c @node Interoperability Routines
2374 @c @section Interoperability Routines
2375 @c
2376 @c Routines to obtain properties from an @code{omp_interop_t} object.
2377 @c They have C linkage and do not throw exceptions.
2378 @c
2379 @c @menu
2380 @c * omp_get_num_interop_properties:: <fixme>
2381 @c * omp_get_interop_int:: <fixme>
2382 @c * omp_get_interop_ptr:: <fixme>
2383 @c * omp_get_interop_str:: <fixme>
2384 @c * omp_get_interop_name:: <fixme>
2385 @c * omp_get_interop_type_desc:: <fixme>
2386 @c * omp_get_interop_rc_desc:: <fixme>
2387 @c @end menu
2388
2389 @node Memory Management Routines
2390 @section Memory Management Routines
2391
2392 Routines to manage and allocate memory on the current device.
2393 They have C linkage and do not throw exceptions.
2394
2395 @menu
2396 * omp_init_allocator:: Create an allocator
2397 * omp_destroy_allocator:: Destroy an allocator
2398 * omp_set_default_allocator:: Set the default allocator
2399 * omp_get_default_allocator:: Get the default allocator
2400 @c * omp_alloc:: <fixme>
2401 @c * omp_aligned_alloc:: <fixme>
2402 @c * omp_free:: <fixme>
2403 @c * omp_calloc:: <fixme>
2404 @c * omp_aligned_calloc:: <fixme>
2405 @c * omp_realloc:: <fixme>
2406 @c * omp_get_memspace_num_resources:: <fixme>/TR11
2407 @c * omp_get_submemspace:: <fixme>/TR11
2408 @end menu
2409
2410
2411
2412 @node omp_init_allocator
2413 @subsection @code{omp_init_allocator} -- Create an allocator
2414 @table @asis
2415 @item @emph{Description}:
2416 Create an allocator that uses the specified memory space and has the specified
2417 traits; if an allocator that fulfills the requirements cannot be created,
2418 @code{omp_null_allocator} is returned.
2419
2420 The predefined memory spaces and available traits can be found at
2421 @ref{OMP_ALLOCATOR}, where the trait names have to be be prefixed by
2422 @code{omp_atk_} (e.g. @code{omp_atk_pinned}) and the named trait values by
2423 @code{omp_atv_} (e.g. @code{omp_atv_true}); additionally, @code{omp_atv_default}
2424 may be used as trait value to specify that the default value should be used.
2425
2426 @item @emph{C/C++}:
2427 @multitable @columnfractions .20 .80
2428 @item @emph{Prototype}: @tab @code{omp_allocator_handle_t omp_init_allocator(}
2429 @item                   @tab @code{  omp_memspace_handle_t memspace,}
2430 @item                   @tab @code{  int ntraits,}
2431 @item                   @tab @code{  const omp_alloctrait_t traits[]);}
2432 @end multitable
2433
2434 @item @emph{Fortran}:
2435 @multitable @columnfractions .20 .80
2436 @item @emph{Interface}: @tab @code{function omp_init_allocator(memspace, ntraits, traits)}
2437 @item                   @tab @code{integer (kind=omp_allocator_handle_kind) :: omp_init_allocator}
2438 @item                   @tab @code{integer (kind=omp_memspace_handle_kind), intent(in) :: memspace}
2439 @item                   @tab @code{integer, intent(in) :: ntraits}
2440 @item                   @tab @code{type (omp_alloctrait), intent(in) :: traits(*)}
2441 @end multitable
2442
2443 @item @emph{See also}:
2444 @ref{OMP_ALLOCATOR}, @ref{Memory allocation}, @ref{omp_destroy_allocator}
2445
2446 @item @emph{Reference}:
2447 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.7.2
2448 @end table
2449
2450
2451
2452 @node omp_destroy_allocator
2453 @subsection @code{omp_destroy_allocator} -- Destroy an allocator
2454 @table @asis
2455 @item @emph{Description}:
2456 Releases all resources used by a memory allocator, which must not represent
2457 a predefined memory allocator.  Accessing memory after its allocator has been
2458 destroyed has unspecified behavior.  Passing @code{omp_null_allocator} to the
2459 routine is permitted but will have no effect.
2460
2461
2462 @item @emph{C/C++}:
2463 @multitable @columnfractions .20 .80
2464 @item @emph{Prototype}: @tab @code{void omp_destroy_allocator (omp_allocator_handle_t allocator);}
2465 @end multitable
2466
2467 @item @emph{Fortran}:
2468 @multitable @columnfractions .20 .80
2469 @item @emph{Interface}: @tab @code{subroutine omp_destroy_allocator(allocator)}
2470 @item                   @tab @code{integer (kind=omp_allocator_handle_kind), intent(in) :: allocator}
2471 @end multitable
2472
2473 @item @emph{See also}:
2474 @ref{omp_init_allocator}
2475
2476 @item @emph{Reference}:
2477 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.7.3
2478 @end table
2479
2480
2481
2482 @node omp_set_default_allocator
2483 @subsection @code{omp_set_default_allocator} -- Set the default allocator
2484 @table @asis
2485 @item @emph{Description}:
2486 Sets the default allocator that is used when no allocator has been specified
2487 in the @code{allocate} or @code{allocator} clause or if an OpenMP memory
2488 routine is invoked with the @code{omp_null_allocator} allocator.
2489
2490 @item @emph{C/C++}:
2491 @multitable @columnfractions .20 .80
2492 @item @emph{Prototype}: @tab @code{void omp_set_default_allocator(omp_allocator_handle_t allocator);}
2493 @end multitable
2494
2495 @item @emph{Fortran}:
2496 @multitable @columnfractions .20 .80
2497 @item @emph{Interface}: @tab @code{subroutine omp_set_default_allocator(allocator)}
2498 @item                   @tab @code{integer (kind=omp_allocator_handle_kind), intent(in) :: allocator}
2499 @end multitable
2500
2501 @item @emph{See also}:
2502 @ref{omp_get_default_allocator}, @ref{omp_init_allocator}, @ref{OMP_ALLOCATOR},
2503 @ref{Memory allocation}
2504
2505 @item @emph{Reference}:
2506 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.7.4
2507 @end table
2508
2509
2510
2511 @node omp_get_default_allocator
2512 @subsection @code{omp_get_default_allocator} -- Get the default allocator
2513 @table @asis
2514 @item @emph{Description}:
2515 The routine returns the default allocator that is used when no allocator has
2516 been specified in the @code{allocate} or @code{allocator} clause or if an
2517 OpenMP memory routine is invoked with the @code{omp_null_allocator} allocator.
2518
2519 @item @emph{C/C++}:
2520 @multitable @columnfractions .20 .80
2521 @item @emph{Prototype}: @tab @code{omp_allocator_handle_t omp_get_default_allocator();}
2522 @end multitable
2523
2524 @item @emph{Fortran}:
2525 @multitable @columnfractions .20 .80
2526 @item @emph{Interface}: @tab @code{function omp_get_default_allocator()}
2527 @item                   @tab @code{integer (kind=omp_allocator_handle_kind) :: omp_get_default_allocator}
2528 @end multitable
2529
2530 @item @emph{See also}:
2531 @ref{omp_set_default_allocator}, @ref{OMP_ALLOCATOR}
2532
2533 @item @emph{Reference}:
2534 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 3.7.5
2535 @end table
2536
2537
2538
2539 @c @node Tool Control Routine
2540 @c
2541 @c FIXME
2542
2543 @c @node Environment Display Routine
2544 @c @section Environment Display Routine
2545 @c
2546 @c Routine to display the OpenMP number and the initial value of ICVs.
2547 @c It has C linkage and do not throw exceptions.
2548 @c
2549 @c menu
2550 @c * omp_display_env:: <fixme>
2551 @c end menu
2552
2553 @c ---------------------------------------------------------------------
2554 @c OpenMP Environment Variables
2555 @c ---------------------------------------------------------------------
2556
2557 @node Environment Variables
2558 @chapter OpenMP Environment Variables
2559
2560 The environment variables which beginning with @env{OMP_} are defined by
2561 section 4 of the OpenMP specification in version 4.5 or in a later version
2562 of the specification, while those beginning with @env{GOMP_} are GNU extensions.
2563 Most @env{OMP_} environment variables have an associated internal control
2564 variable (ICV).
2565
2566 For any OpenMP environment variable that sets an ICV and is neither
2567 @code{OMP_DEFAULT_DEVICE} nor has global ICV scope, associated
2568 device-specific environment variables exist.  For them, the environment
2569 variable without suffix affects the host.  The suffix @code{_DEV_} followed
2570 by a non-negative device number less that the number of available devices sets
2571 the ICV for the corresponding device.  The suffix @code{_DEV} sets the ICV
2572 of all non-host devices for which a device-specific corresponding environment
2573 variable has not been set while the @code{_ALL} suffix sets the ICV of all
2574 host and non-host devices for which a more specific corresponding environment
2575 variable is not set.
2576
2577 @menu
2578 * OMP_ALLOCATOR::           Set the default allocator
2579 * OMP_AFFINITY_FORMAT::     Set the format string used for affinity display
2580 * OMP_CANCELLATION::        Set whether cancellation is activated
2581 * OMP_DISPLAY_AFFINITY::    Display thread affinity information
2582 * OMP_DISPLAY_ENV::         Show OpenMP version and environment variables
2583 * OMP_DEFAULT_DEVICE::      Set the device used in target regions
2584 * OMP_DYNAMIC::             Dynamic adjustment of threads
2585 * OMP_MAX_ACTIVE_LEVELS::   Set the maximum number of nested parallel regions
2586 * OMP_MAX_TASK_PRIORITY::   Set the maximum task priority value
2587 * OMP_NESTED::              Nested parallel regions
2588 * OMP_NUM_TEAMS::           Specifies the number of teams to use by teams region
2589 * OMP_NUM_THREADS::         Specifies the number of threads to use
2590 * OMP_PROC_BIND::           Whether threads may be moved between CPUs
2591 * OMP_PLACES::              Specifies on which CPUs the threads should be placed
2592 * OMP_STACKSIZE::           Set default thread stack size
2593 * OMP_SCHEDULE::            How threads are scheduled
2594 * OMP_TARGET_OFFLOAD::      Controls offloading behaviour
2595 * OMP_TEAMS_THREAD_LIMIT::  Set the maximum number of threads imposed by teams
2596 * OMP_THREAD_LIMIT::        Set the maximum number of threads
2597 * OMP_WAIT_POLICY::         How waiting threads are handled
2598 * GOMP_CPU_AFFINITY::       Bind threads to specific CPUs
2599 * GOMP_DEBUG::              Enable debugging output
2600 * GOMP_STACKSIZE::          Set default thread stack size
2601 * GOMP_SPINCOUNT::          Set the busy-wait spin count
2602 * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools
2603 @end menu
2604
2605
2606 @node OMP_ALLOCATOR
2607 @section @env{OMP_ALLOCATOR} -- Set the default allocator
2608 @cindex Environment Variable
2609 @table @asis
2610 @item @emph{ICV:} @var{def-allocator-var}
2611 @item @emph{Scope:} data environment
2612 @item @emph{Description}:
2613 Sets the default allocator that is used when no allocator has been specified
2614 in the @code{allocate} or @code{allocator} clause or if an OpenMP memory
2615 routine is invoked with the @code{omp_null_allocator} allocator.
2616 If unset, @code{omp_default_mem_alloc} is used.
2617
2618 The value can either be a predefined allocator or a predefined memory space
2619 or a predefined memory space followed by a colon and a comma-separated list
2620 of memory trait and value pairs, separated by @code{=}.
2621
2622 Note: The corresponding device environment variables are currently not
2623 supported.  Therefore, the non-host @var{def-allocator-var} ICVs are always
2624 initialized to @code{omp_default_mem_alloc}.  However, on all devices,
2625 the @code{omp_set_default_allocator} API routine can be used to change
2626 value.
2627
2628 @multitable @columnfractions .45 .45
2629 @headitem Predefined allocators @tab Associated predefined memory spaces
2630 @item omp_default_mem_alloc     @tab omp_default_mem_space
2631 @item omp_large_cap_mem_alloc   @tab omp_large_cap_mem_space
2632 @item omp_const_mem_alloc       @tab omp_const_mem_space
2633 @item omp_high_bw_mem_alloc     @tab omp_high_bw_mem_space
2634 @item omp_low_lat_mem_alloc     @tab omp_low_lat_mem_space
2635 @item omp_cgroup_mem_alloc      @tab --
2636 @item omp_pteam_mem_alloc       @tab --
2637 @item omp_thread_mem_alloc      @tab --
2638 @end multitable
2639
2640 The predefined allocators use the default values for the traits,
2641 as listed below.  Except that the last three allocators have the
2642 @code{access} trait set to @code{cgroup}, @code{pteam}, and
2643 @code{thread}, respectively.
2644
2645 @multitable @columnfractions .25 .40 .25
2646 @headitem Trait @tab Allowed values @tab Default value
2647 @item @code{sync_hint} @tab @code{contended}, @code{uncontended},
2648                             @code{serialized}, @code{private}
2649                        @tab @code{contended}
2650 @item @code{alignment} @tab Positive integer being a power of two
2651                        @tab 1 byte
2652 @item @code{access}    @tab @code{all}, @code{cgroup},
2653                             @code{pteam}, @code{thread}
2654                        @tab @code{all}
2655 @item @code{pool_size} @tab Positive integer
2656                        @tab See @ref{Memory allocation}
2657 @item @code{fallback}  @tab @code{default_mem_fb}, @code{null_fb},
2658                             @code{abort_fb}, @code{allocator_fb}
2659                        @tab See below
2660 @item @code{fb_data}   @tab @emph{unsupported as it needs an allocator handle}
2661                        @tab (none)
2662 @item @code{pinned}    @tab @code{true}, @code{false}
2663                        @tab @code{false}
2664 @item @code{partition} @tab @code{environment}, @code{nearest},
2665                             @code{blocked}, @code{interleaved}
2666                        @tab @code{environment}
2667 @end multitable
2668
2669 For the @code{fallback} trait, the default value is @code{null_fb} for the
2670 @code{omp_default_mem_alloc} allocator and any allocator that is associated
2671 with device memory; for all other other allocators, it is @code{default_mem_fb}
2672 by default.
2673
2674 Examples:
2675 @smallexample
2676 OMP_ALLOCATOR=omp_high_bw_mem_alloc
2677 OMP_ALLOCATOR=omp_large_cap_mem_space
2678 OMP_ALLOCATOR=omp_low_lat_mem_space:pinned=true,partition=nearest
2679 @end smallexample
2680
2681 @item @emph{See also}:
2682 @ref{Memory allocation}, @ref{omp_get_default_allocator},
2683 @ref{omp_set_default_allocator}
2684
2685 @item @emph{Reference}:
2686 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 6.21
2687 @end table
2688
2689
2690
2691 @node OMP_AFFINITY_FORMAT
2692 @section @env{OMP_AFFINITY_FORMAT} -- Set the format string used for affinity display
2693 @cindex Environment Variable
2694 @table @asis
2695 @item @emph{ICV:} @var{affinity-format-var}
2696 @item @emph{Scope:} device
2697 @item @emph{Description}:
2698 Sets the format string used when displaying OpenMP thread affinity information.
2699 Special values are output using @code{%} followed by an optional size
2700 specification and then either the single-character field type or its long
2701 name enclosed in curly braces; using @code{%%} will display a literal percent.
2702 The size specification consists of an optional @code{0.} or @code{.} followed
2703 by a positive integer, specifying the minimal width of the output.  With
2704 @code{0.} and numerical values, the output is padded with zeros on the left;
2705 with @code{.}, the output is padded by spaces on the left; otherwise, the
2706 output is padded by spaces on the right.  If unset, the value is
2707 ``@code{level %L thread %i affinity %A}''.
2708
2709 Supported field types are:
2710
2711 @multitable @columnfractions .10 .25 .60
2712 @item t @tab team_num @tab value returned by @code{omp_get_team_num}
2713 @item T @tab num_teams @tab value returned by @code{omp_get_num_teams}
2714 @item L @tab nesting_level @tab value returned by @code{omp_get_level}
2715 @item n @tab thread_num @tab value returned by @code{omp_get_thread_num}
2716 @item N @tab num_threads @tab value returned by @code{omp_get_num_threads}
2717 @item a @tab ancestor_tnum
2718       @tab value returned by
2719            @code{omp_get_ancestor_thread_num(omp_get_level()-1)}
2720 @item H @tab host @tab name of the host that executes the thread
2721 @item P @tab process_id @tab process identifier
2722 @item i @tab native_thread_id @tab native thread identifier
2723 @item A @tab thread_affinity
2724       @tab comma separated list of integer values or ranges, representing the
2725            processors on which a process might execute, subject to affinity
2726            mechanisms
2727 @end multitable
2728
2729 For instance, after setting
2730
2731 @smallexample
2732 OMP_AFFINITY_FORMAT="%0.2a!%n!%.4L!%N;%.2t;%0.2T;%@{team_num@};%@{num_teams@};%A"
2733 @end smallexample
2734
2735 with either @code{OMP_DISPLAY_AFFINITY} being set or when calling
2736 @code{omp_display_affinity} with @code{NULL} or an empty string, the program
2737 might display the following:
2738
2739 @smallexample
2740 00!0!   1!4; 0;01;0;1;0-11
2741 00!3!   1!4; 0;01;0;1;0-11
2742 00!2!   1!4; 0;01;0;1;0-11
2743 00!1!   1!4; 0;01;0;1;0-11
2744 @end smallexample
2745
2746 @item @emph{See also}:
2747 @ref{OMP_DISPLAY_AFFINITY}
2748
2749 @item @emph{Reference}:
2750 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 6.14
2751 @end table
2752
2753
2754
2755 @node OMP_CANCELLATION
2756 @section @env{OMP_CANCELLATION} -- Set whether cancellation is activated
2757 @cindex Environment Variable
2758 @table @asis
2759 @item @emph{ICV:} @var{cancel-var}
2760 @item @emph{Scope:} global
2761 @item @emph{Description}:
2762 If set to @code{TRUE}, the cancellation is activated.  If set to @code{FALSE} or
2763 if unset, cancellation is disabled and the @code{cancel} construct is ignored.
2764
2765 @item @emph{See also}:
2766 @ref{omp_get_cancellation}
2767
2768 @item @emph{Reference}:
2769 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.11
2770 @end table
2771
2772
2773
2774 @node OMP_DISPLAY_AFFINITY
2775 @section @env{OMP_DISPLAY_AFFINITY} -- Display thread affinity information
2776 @cindex Environment Variable
2777 @table @asis
2778 @item @emph{ICV:} @var{display-affinity-var}
2779 @item @emph{Scope:} global
2780 @item @emph{Description}:
2781 If set to @code{FALSE} or if unset, affinity displaying is disabled.
2782 If set to @code{TRUE}, the runtime will display affinity information about
2783 OpenMP threads in a parallel region upon entering the region and every time
2784 any change occurs.
2785
2786 @item @emph{See also}:
2787 @ref{OMP_AFFINITY_FORMAT}
2788
2789 @item @emph{Reference}:
2790 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 6.13
2791 @end table
2792
2793
2794
2795
2796 @node OMP_DISPLAY_ENV
2797 @section @env{OMP_DISPLAY_ENV} -- Show OpenMP version and environment variables
2798 @cindex Environment Variable
2799 @table @asis
2800 @item @emph{ICV:} none
2801 @item @emph{Scope:} not applicable
2802 @item @emph{Description}:
2803 If set to @code{TRUE}, the OpenMP version number and the values
2804 associated with the OpenMP environment variables are printed to @code{stderr}.
2805 If set to @code{VERBOSE}, it additionally shows the value of the environment
2806 variables which are GNU extensions.  If undefined or set to @code{FALSE},
2807 this information will not be shown.
2808
2809
2810 @item @emph{Reference}:
2811 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.12
2812 @end table
2813
2814
2815
2816 @node OMP_DEFAULT_DEVICE
2817 @section @env{OMP_DEFAULT_DEVICE} -- Set the device used in target regions
2818 @cindex Environment Variable
2819 @table @asis
2820 @item @emph{ICV:} @var{default-device-var}
2821 @item @emph{Scope:} data environment
2822 @item @emph{Description}:
2823 Set to choose the device which is used in a @code{target} region, unless the
2824 value is overridden by @code{omp_set_default_device} or by a @code{device}
2825 clause.  The value shall be the nonnegative device number. If no device with
2826 the given device number exists, the code is executed on the host.  If unset,
2827 @env{OMP_TARGET_OFFLOAD} is @code{mandatory} and no non-host devices are
2828 available, it is set to @code{omp_invalid_device}.  Otherwise, if unset,
2829 device number 0 will be used.
2830
2831
2832 @item @emph{See also}:
2833 @ref{omp_get_default_device}, @ref{omp_set_default_device},
2834
2835 @item @emph{Reference}:
2836 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.13
2837 @end table
2838
2839
2840
2841 @node OMP_DYNAMIC
2842 @section @env{OMP_DYNAMIC} -- Dynamic adjustment of threads
2843 @cindex Environment Variable
2844 @table @asis
2845 @item @emph{ICV:} @var{dyn-var}
2846 @item @emph{Scope:} global
2847 @item @emph{Description}:
2848 Enable or disable the dynamic adjustment of the number of threads
2849 within a team.  The value of this environment variable shall be
2850 @code{TRUE} or @code{FALSE}.  If undefined, dynamic adjustment is
2851 disabled by default.
2852
2853 @item @emph{See also}:
2854 @ref{omp_set_dynamic}
2855
2856 @item @emph{Reference}:
2857 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.3
2858 @end table
2859
2860
2861
2862 @node OMP_MAX_ACTIVE_LEVELS
2863 @section @env{OMP_MAX_ACTIVE_LEVELS} -- Set the maximum number of nested parallel regions
2864 @cindex Environment Variable
2865 @table @asis
2866 @item @emph{ICV:} @var{max-active-levels-var}
2867 @item @emph{Scope:} data environment
2868 @item @emph{Description}:
2869 Specifies the initial value for the maximum number of nested parallel
2870 regions.  The value of this variable shall be a positive integer.
2871 If undefined, then if @env{OMP_NESTED} is defined and set to true, or
2872 if @env{OMP_NUM_THREADS} or @env{OMP_PROC_BIND} are defined and set to
2873 a list with more than one item, the maximum number of nested parallel
2874 regions will be initialized to the largest number supported, otherwise
2875 it will be set to one.
2876
2877 @item @emph{See also}:
2878 @ref{omp_set_max_active_levels}, @ref{OMP_NESTED}, @ref{OMP_PROC_BIND},
2879 @ref{OMP_NUM_THREADS}
2880
2881
2882 @item @emph{Reference}:
2883 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.9
2884 @end table
2885
2886
2887
2888 @node OMP_MAX_TASK_PRIORITY
2889 @section @env{OMP_MAX_TASK_PRIORITY} -- Set the maximum priority
2890 number that can be set for a task.
2891 @cindex Environment Variable
2892 @table @asis
2893 @item @emph{ICV:} @var{max-task-priority-var}
2894 @item @emph{Scope:} global
2895 @item @emph{Description}:
2896 Specifies the initial value for the maximum priority value that can be
2897 set for a task.  The value of this variable shall be a non-negative
2898 integer, and zero is allowed.  If undefined, the default priority is
2899 0.
2900
2901 @item @emph{See also}:
2902 @ref{omp_get_max_task_priority}
2903
2904 @item @emph{Reference}:
2905 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.14
2906 @end table
2907
2908
2909
2910 @node OMP_NESTED
2911 @section @env{OMP_NESTED} -- Nested parallel regions
2912 @cindex Environment Variable
2913 @cindex Implementation specific setting
2914 @table @asis
2915 @item @emph{ICV:} @var{max-active-levels-var}
2916 @item @emph{Scope:} data environment
2917 @item @emph{Description}:
2918 Enable or disable nested parallel regions, i.e., whether team members
2919 are allowed to create new teams.  The value of this environment variable
2920 shall be @code{TRUE} or @code{FALSE}.  If set to @code{TRUE}, the number
2921 of maximum active nested regions supported will by default be set to the
2922 maximum supported, otherwise it will be set to one.  If
2923 @env{OMP_MAX_ACTIVE_LEVELS} is defined, its setting will override this
2924 setting.  If both are undefined, nested parallel regions are enabled if
2925 @env{OMP_NUM_THREADS} or @env{OMP_PROC_BINDS} are defined to a list with
2926 more than one item, otherwise they are disabled by default.
2927
2928 Note that the @code{OMP_NESTED} environment variable was deprecated in
2929 the OpenMP specification 5.2 in favor of @code{OMP_MAX_ACTIVE_LEVELS}.
2930
2931 @item @emph{See also}:
2932 @ref{omp_set_max_active_levels}, @ref{omp_set_nested},
2933 @ref{OMP_MAX_ACTIVE_LEVELS}
2934
2935 @item @emph{Reference}:
2936 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.6
2937 @end table
2938
2939
2940
2941 @node OMP_NUM_TEAMS
2942 @section @env{OMP_NUM_TEAMS} -- Specifies the number of teams to use by teams region
2943 @cindex Environment Variable
2944 @table @asis
2945 @item @emph{ICV:} @var{nteams-var}
2946 @item @emph{Scope:} device
2947 @item @emph{Description}:
2948 Specifies the upper bound for number of teams to use in teams regions
2949 without explicit @code{num_teams} clause.  The value of this variable shall
2950 be a positive integer.  If undefined it defaults to 0 which means
2951 implementation defined upper bound.
2952
2953 @item @emph{See also}:
2954 @ref{omp_set_num_teams}
2955
2956 @item @emph{Reference}:
2957 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 6.23
2958 @end table
2959
2960
2961
2962 @node OMP_NUM_THREADS
2963 @section @env{OMP_NUM_THREADS} -- Specifies the number of threads to use
2964 @cindex Environment Variable
2965 @cindex Implementation specific setting
2966 @table @asis
2967 @item @emph{ICV:} @var{nthreads-var}
2968 @item @emph{Scope:} data environment
2969 @item @emph{Description}:
2970 Specifies the default number of threads to use in parallel regions.  The
2971 value of this variable shall be a comma-separated list of positive integers;
2972 the value specifies the number of threads to use for the corresponding nested
2973 level.  Specifying more than one item in the list will automatically enable
2974 nesting by default.  If undefined one thread per CPU is used.
2975
2976 When a list with more than value is specified, it also affects the
2977 @var{max-active-levels-var} ICV as described in @ref{OMP_MAX_ACTIVE_LEVELS}.
2978
2979 @item @emph{See also}:
2980 @ref{omp_set_num_threads}, @ref{OMP_MAX_ACTIVE_LEVELS}
2981
2982 @item @emph{Reference}:
2983 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.2
2984 @end table
2985
2986
2987
2988 @node OMP_PROC_BIND
2989 @section @env{OMP_PROC_BIND} -- Whether threads may be moved between CPUs
2990 @cindex Environment Variable
2991 @table @asis
2992 @item @emph{ICV:} @var{bind-var}
2993 @item @emph{Scope:} data environment
2994 @item @emph{Description}:
2995 Specifies whether threads may be moved between processors.  If set to
2996 @code{TRUE}, OpenMP threads should not be moved; if set to @code{FALSE}
2997 they may be moved.  Alternatively, a comma separated list with the
2998 values @code{PRIMARY}, @code{MASTER}, @code{CLOSE} and @code{SPREAD} can
2999 be used to specify the thread affinity policy for the corresponding nesting
3000 level.  With @code{PRIMARY} and @code{MASTER} the worker threads are in the
3001 same place partition as the primary thread.  With @code{CLOSE} those are
3002 kept close to the primary thread in contiguous place partitions.  And
3003 with @code{SPREAD} a sparse distribution
3004 across the place partitions is used.  Specifying more than one item in the
3005 list will automatically enable nesting by default.
3006
3007 When a list is specified, it also affects the @var{max-active-levels-var} ICV
3008 as described in @ref{OMP_MAX_ACTIVE_LEVELS}.
3009
3010 When undefined, @env{OMP_PROC_BIND} defaults to @code{TRUE} when
3011 @env{OMP_PLACES} or @env{GOMP_CPU_AFFINITY} is set and @code{FALSE} otherwise.
3012
3013 @item @emph{See also}:
3014 @ref{omp_get_proc_bind}, @ref{GOMP_CPU_AFFINITY}, @ref{OMP_PLACES},
3015 @ref{OMP_MAX_ACTIVE_LEVELS}
3016
3017 @item @emph{Reference}:
3018 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.4
3019 @end table
3020
3021
3022
3023 @node OMP_PLACES
3024 @section @env{OMP_PLACES} -- Specifies on which CPUs the threads should be placed
3025 @cindex Environment Variable
3026 @table @asis
3027 @item @emph{ICV:} @var{place-partition-var}
3028 @item @emph{Scope:} implicit tasks
3029 @item @emph{Description}:
3030 The thread placement can be either specified using an abstract name or by an
3031 explicit list of the places.  The abstract names @code{threads}, @code{cores},
3032 @code{sockets}, @code{ll_caches} and @code{numa_domains} can be optionally
3033 followed by a positive number in parentheses, which denotes the how many places
3034 shall be created.  With @code{threads} each place corresponds to a single
3035 hardware thread; @code{cores} to a single core with the corresponding number of
3036 hardware threads; with @code{sockets} the place corresponds to a single
3037 socket; with @code{ll_caches} to a set of cores that shares the last level
3038 cache on the device; and @code{numa_domains} to a set of cores for which their
3039 closest memory on the device is the same memory and at a similar distance from
3040 the cores.  The resulting placement can be shown by setting the
3041 @env{OMP_DISPLAY_ENV} environment variable.
3042
3043 Alternatively, the placement can be specified explicitly as comma-separated
3044 list of places.  A place is specified by set of nonnegative numbers in curly
3045 braces, denoting the hardware threads.  The curly braces can be omitted
3046 when only a single number has been specified.  The hardware threads
3047 belonging to a place can either be specified as comma-separated list of
3048 nonnegative thread numbers or using an interval.  Multiple places can also be
3049 either specified by a comma-separated list of places or by an interval.  To
3050 specify an interval, a colon followed by the count is placed after
3051 the hardware thread number or the place.  Optionally, the length can be
3052 followed by a colon and the stride number -- otherwise a unit stride is
3053 assumed.  Placing an exclamation mark (@code{!}) directly before a curly
3054 brace or numbers inside the curly braces (excluding intervals) will
3055 exclude those hardware threads.
3056
3057 For instance, the following specifies the same places list:
3058 @code{"@{0,1,2@}, @{3,4,6@}, @{7,8,9@}, @{10,11,12@}"};
3059 @code{"@{0:3@}, @{3:3@}, @{7:3@}, @{10:3@}"}; and @code{"@{0:2@}:4:3"}.
3060
3061 If @env{OMP_PLACES} and @env{GOMP_CPU_AFFINITY} are unset and
3062 @env{OMP_PROC_BIND} is either unset or @code{false}, threads may be moved
3063 between CPUs following no placement policy.
3064
3065 @item @emph{See also}:
3066 @ref{OMP_PROC_BIND}, @ref{GOMP_CPU_AFFINITY}, @ref{omp_get_proc_bind},
3067 @ref{OMP_DISPLAY_ENV}
3068
3069 @item @emph{Reference}:
3070 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.5
3071 @end table
3072
3073
3074
3075 @node OMP_STACKSIZE
3076 @section @env{OMP_STACKSIZE} -- Set default thread stack size
3077 @cindex Environment Variable
3078 @table @asis
3079 @item @emph{ICV:} @var{stacksize-var}
3080 @item @emph{Scope:} device
3081 @item @emph{Description}:
3082 Set the default thread stack size in kilobytes, unless the number
3083 is suffixed by @code{B}, @code{K}, @code{M} or @code{G}, in which
3084 case the size is, respectively, in bytes, kilobytes, megabytes
3085 or gigabytes.  This is different from @code{pthread_attr_setstacksize}
3086 which gets the number of bytes as an argument.  If the stack size cannot
3087 be set due to system constraints, an error is reported and the initial
3088 stack size is left unchanged.  If undefined, the stack size is system
3089 dependent.
3090
3091 @item @emph{See also}:
3092 @ref{GOMP_STACKSIZE}
3093
3094 @item @emph{Reference}:
3095 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.7
3096 @end table
3097
3098
3099
3100 @node OMP_SCHEDULE
3101 @section @env{OMP_SCHEDULE} -- How threads are scheduled
3102 @cindex Environment Variable
3103 @cindex Implementation specific setting
3104 @table @asis
3105 @item @emph{ICV:} @var{run-sched-var}
3106 @item @emph{Scope:} data environment
3107 @item @emph{Description}:
3108 Allows to specify @code{schedule type} and @code{chunk size}.
3109 The value of the variable shall have the form: @code{type[,chunk]} where
3110 @code{type} is one of @code{static}, @code{dynamic}, @code{guided} or @code{auto}
3111 The optional @code{chunk} size shall be a positive integer.  If undefined,
3112 dynamic scheduling and a chunk size of 1 is used.
3113
3114 @item @emph{See also}:
3115 @ref{omp_set_schedule}
3116
3117 @item @emph{Reference}:
3118 @uref{https://www.openmp.org, OpenMP specification v4.5}, Sections 2.7.1.1 and 4.1
3119 @end table
3120
3121
3122
3123 @node OMP_TARGET_OFFLOAD
3124 @section @env{OMP_TARGET_OFFLOAD} -- Controls offloading behaviour
3125 @cindex Environment Variable
3126 @cindex Implementation specific setting
3127 @table @asis
3128 @item @emph{ICV:} @var{target-offload-var}
3129 @item @emph{Scope:} global
3130 @item @emph{Description}:
3131 Specifies the behaviour with regard to offloading code to a device.  This
3132 variable can be set to one of three values - @code{MANDATORY}, @code{DISABLED}
3133 or @code{DEFAULT}.
3134
3135 If set to @code{MANDATORY}, the program will terminate with an error if
3136 the offload device is not present or is not supported.  If set to
3137 @code{DISABLED}, then offloading is disabled and all code will run on the
3138 host. If set to @code{DEFAULT}, the program will try offloading to the
3139 device first, then fall back to running code on the host if it cannot.
3140
3141 If undefined, then the program will behave as if @code{DEFAULT} was set.
3142
3143 @item @emph{Reference}:
3144 @uref{https://www.openmp.org, OpenMP specification v5.0}, Section 6.17
3145 @end table
3146
3147
3148
3149 @node OMP_TEAMS_THREAD_LIMIT
3150 @section @env{OMP_TEAMS_THREAD_LIMIT} -- Set the maximum number of threads imposed by teams
3151 @cindex Environment Variable
3152 @table @asis
3153 @item @emph{ICV:} @var{teams-thread-limit-var}
3154 @item @emph{Scope:} device
3155 @item @emph{Description}:
3156 Specifies an upper bound for the number of threads to use by each contention
3157 group created by a teams construct without explicit @code{thread_limit}
3158 clause.  The value of this variable shall be a positive integer.  If undefined,
3159 the value of 0 is used which stands for an implementation defined upper
3160 limit.
3161
3162 @item @emph{See also}:
3163 @ref{OMP_THREAD_LIMIT}, @ref{omp_set_teams_thread_limit}
3164
3165 @item @emph{Reference}:
3166 @uref{https://www.openmp.org, OpenMP specification v5.1}, Section 6.24
3167 @end table
3168
3169
3170
3171 @node OMP_THREAD_LIMIT
3172 @section @env{OMP_THREAD_LIMIT} -- Set the maximum number of threads
3173 @cindex Environment Variable
3174 @table @asis
3175 @item @emph{ICV:} @var{thread-limit-var}
3176 @item @emph{Scope:} data environment
3177 @item @emph{Description}:
3178 Specifies the number of threads to use for the whole program.  The
3179 value of this variable shall be a positive integer.  If undefined,
3180 the number of threads is not limited.
3181
3182 @item @emph{See also}:
3183 @ref{OMP_NUM_THREADS}, @ref{omp_get_thread_limit}
3184
3185 @item @emph{Reference}:
3186 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.10
3187 @end table
3188
3189
3190
3191 @node OMP_WAIT_POLICY
3192 @section @env{OMP_WAIT_POLICY} -- How waiting threads are handled
3193 @cindex Environment Variable
3194 @table @asis
3195 @item @emph{Description}:
3196 Specifies whether waiting threads should be active or passive.  If
3197 the value is @code{PASSIVE}, waiting threads should not consume CPU
3198 power while waiting; while the value is @code{ACTIVE} specifies that
3199 they should.  If undefined, threads wait actively for a short time
3200 before waiting passively.
3201
3202 @item @emph{See also}:
3203 @ref{GOMP_SPINCOUNT}
3204
3205 @item @emph{Reference}:
3206 @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.8
3207 @end table
3208
3209
3210
3211 @node GOMP_CPU_AFFINITY
3212 @section @env{GOMP_CPU_AFFINITY} -- Bind threads to specific CPUs
3213 @cindex Environment Variable
3214 @table @asis
3215 @item @emph{Description}:
3216 Binds threads to specific CPUs.  The variable should contain a space-separated
3217 or comma-separated list of CPUs.  This list may contain different kinds of
3218 entries: either single CPU numbers in any order, a range of CPUs (M-N)
3219 or a range with some stride (M-N:S).  CPU numbers are zero based.  For example,
3220 @code{GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"} will bind the initial thread
3221 to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to
3222 CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 10, 12,
3223 and 14 respectively and then start assigning back from the beginning of
3224 the list.  @code{GOMP_CPU_AFFINITY=0} binds all threads to CPU 0.
3225
3226 There is no libgomp library routine to determine whether a CPU affinity
3227 specification is in effect.  As a workaround, language-specific library
3228 functions, e.g., @code{getenv} in C or @code{GET_ENVIRONMENT_VARIABLE} in
3229 Fortran, may be used to query the setting of the @code{GOMP_CPU_AFFINITY}
3230 environment variable.  A defined CPU affinity on startup cannot be changed
3231 or disabled during the runtime of the application.
3232
3233 If both @env{GOMP_CPU_AFFINITY} and @env{OMP_PROC_BIND} are set,
3234 @env{OMP_PROC_BIND} has a higher precedence.  If neither has been set and
3235 @env{OMP_PROC_BIND} is unset, or when @env{OMP_PROC_BIND} is set to
3236 @code{FALSE}, the host system will handle the assignment of threads to CPUs.
3237
3238 @item @emph{See also}:
3239 @ref{OMP_PLACES}, @ref{OMP_PROC_BIND}
3240 @end table
3241
3242
3243
3244 @node GOMP_DEBUG
3245 @section @env{GOMP_DEBUG} -- Enable debugging output
3246 @cindex Environment Variable
3247 @table @asis
3248 @item @emph{Description}:
3249 Enable debugging output.  The variable should be set to @code{0}
3250 (disabled, also the default if not set), or @code{1} (enabled).
3251
3252 If enabled, some debugging output will be printed during execution.
3253 This is currently not specified in more detail, and subject to change.
3254 @end table
3255
3256
3257
3258 @node GOMP_STACKSIZE
3259 @section @env{GOMP_STACKSIZE} -- Set default thread stack size
3260 @cindex Environment Variable
3261 @cindex Implementation specific setting
3262 @table @asis
3263 @item @emph{Description}:
3264 Set the default thread stack size in kilobytes.  This is different from
3265 @code{pthread_attr_setstacksize} which gets the number of bytes as an
3266 argument.  If the stack size cannot be set due to system constraints, an
3267 error is reported and the initial stack size is left unchanged.  If undefined,
3268 the stack size is system dependent.
3269
3270 @item @emph{See also}:
3271 @ref{OMP_STACKSIZE}
3272
3273 @item @emph{Reference}:
3274 @uref{https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html,
3275 GCC Patches Mailinglist},
3276 @uref{https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html,
3277 GCC Patches Mailinglist}
3278 @end table
3279
3280
3281
3282 @node GOMP_SPINCOUNT
3283 @section @env{GOMP_SPINCOUNT} -- Set the busy-wait spin count
3284 @cindex Environment Variable
3285 @cindex Implementation specific setting
3286 @table @asis
3287 @item @emph{Description}:
3288 Determines how long a threads waits actively with consuming CPU power
3289 before waiting passively without consuming CPU power.  The value may be
3290 either @code{INFINITE}, @code{INFINITY} to always wait actively or an
3291 integer which gives the number of spins of the busy-wait loop.  The
3292 integer may optionally be followed by the following suffixes acting
3293 as multiplication factors: @code{k} (kilo, thousand), @code{M} (mega,
3294 million), @code{G} (giga, billion), or @code{T} (tera, trillion).
3295 If undefined, 0 is used when @env{OMP_WAIT_POLICY} is @code{PASSIVE},
3296 300,000 is used when @env{OMP_WAIT_POLICY} is undefined and
3297 30 billion is used when @env{OMP_WAIT_POLICY} is @code{ACTIVE}.
3298 If there are more OpenMP threads than available CPUs, 1000 and 100
3299 spins are used for @env{OMP_WAIT_POLICY} being @code{ACTIVE} or
3300 undefined, respectively; unless the @env{GOMP_SPINCOUNT} is lower
3301 or @env{OMP_WAIT_POLICY} is @code{PASSIVE}.
3302
3303 @item @emph{See also}:
3304 @ref{OMP_WAIT_POLICY}
3305 @end table
3306
3307
3308
3309 @node GOMP_RTEMS_THREAD_POOLS
3310 @section @env{GOMP_RTEMS_THREAD_POOLS} -- Set the RTEMS specific thread pools
3311 @cindex Environment Variable
3312 @cindex Implementation specific setting
3313 @table @asis
3314 @item @emph{Description}:
3315 This environment variable is only used on the RTEMS real-time operating system.
3316 It determines the scheduler instance specific thread pools.  The format for
3317 @env{GOMP_RTEMS_THREAD_POOLS} is a list of optional
3318 @code{<thread-pool-count>[$<priority>]@@<scheduler-name>} configurations
3319 separated by @code{:} where:
3320 @itemize @bullet
3321 @item @code{<thread-pool-count>} is the thread pool count for this scheduler
3322 instance.
3323 @item @code{$<priority>} is an optional priority for the worker threads of a
3324 thread pool according to @code{pthread_setschedparam}.  In case a priority
3325 value is omitted, then a worker thread will inherit the priority of the OpenMP
3326 primary thread that created it.  The priority of the worker thread is not
3327 changed after creation, even if a new OpenMP primary thread using the worker has
3328 a different priority.
3329 @item @code{@@<scheduler-name>} is the scheduler instance name according to the
3330 RTEMS application configuration.
3331 @end itemize
3332 In case no thread pool configuration is specified for a scheduler instance,
3333 then each OpenMP primary thread of this scheduler instance will use its own
3334 dynamically allocated thread pool.  To limit the worker thread count of the
3335 thread pools, each OpenMP primary thread must call @code{omp_set_num_threads}.
3336 @item @emph{Example}:
3337 Lets suppose we have three scheduler instances @code{IO}, @code{WRK0}, and
3338 @code{WRK1} with @env{GOMP_RTEMS_THREAD_POOLS} set to
3339 @code{"1@@WRK0:3$4@@WRK1"}.  Then there are no thread pool restrictions for
3340 scheduler instance @code{IO}.  In the scheduler instance @code{WRK0} there is
3341 one thread pool available.  Since no priority is specified for this scheduler
3342 instance, the worker thread inherits the priority of the OpenMP primary thread
3343 that created it.  In the scheduler instance @code{WRK1} there are three thread
3344 pools available and their worker threads run at priority four.
3345 @end table
3346
3347
3348
3349 @c ---------------------------------------------------------------------
3350 @c Enabling OpenACC
3351 @c ---------------------------------------------------------------------
3352
3353 @node Enabling OpenACC
3354 @chapter Enabling OpenACC
3355
3356 To activate the OpenACC extensions for C/C++ and Fortran, the compile-time
3357 flag @option{-fopenacc} must be specified.  This enables the OpenACC directive
3358 @code{#pragma acc} in C/C++ and @code{!$acc} directives in free form,
3359 @code{c$acc}, @code{*$acc} and @code{!$acc} directives in fixed form,
3360 @code{!$} conditional compilation sentinels in free form and @code{c$},
3361 @code{*$} and @code{!$} sentinels in fixed form, for Fortran.  The flag also
3362 arranges for automatic linking of the OpenACC runtime library
3363 (@ref{OpenACC Runtime Library Routines}).
3364
3365 See @uref{https://gcc.gnu.org/wiki/OpenACC} for more information.
3366
3367 A complete description of all OpenACC directives accepted may be found in
3368 the @uref{https://www.openacc.org, OpenACC} Application Programming
3369 Interface manual, version 2.6.
3370
3371
3372
3373 @c ---------------------------------------------------------------------
3374 @c OpenACC Runtime Library Routines
3375 @c ---------------------------------------------------------------------
3376
3377 @node OpenACC Runtime Library Routines
3378 @chapter OpenACC Runtime Library Routines
3379
3380 The runtime routines described here are defined by section 3 of the OpenACC
3381 specifications in version 2.6.
3382 They have C linkage, and do not throw exceptions.
3383 Generally, they are available only for the host, with the exception of
3384 @code{acc_on_device}, which is available for both the host and the
3385 acceleration device.
3386
3387 @menu
3388 * acc_get_num_devices::         Get number of devices for the given device
3389                                 type.
3390 * acc_set_device_type::         Set type of device accelerator to use.
3391 * acc_get_device_type::         Get type of device accelerator to be used.
3392 * acc_set_device_num::          Set device number to use.
3393 * acc_get_device_num::          Get device number to be used.
3394 * acc_get_property::            Get device property.
3395 * acc_async_test::              Tests for completion of a specific asynchronous
3396                                 operation.
3397 * acc_async_test_all::          Tests for completion of all asynchronous
3398                                 operations.
3399 * acc_wait::                    Wait for completion of a specific asynchronous
3400                                 operation.
3401 * acc_wait_all::                Waits for completion of all asynchronous
3402                                 operations.
3403 * acc_wait_all_async::          Wait for completion of all asynchronous
3404                                 operations.
3405 * acc_wait_async::              Wait for completion of asynchronous operations.
3406 * acc_init::                    Initialize runtime for a specific device type.
3407 * acc_shutdown::                Shuts down the runtime for a specific device
3408                                 type.
3409 * acc_on_device::               Whether executing on a particular device
3410 * acc_malloc::                  Allocate device memory.
3411 * acc_free::                    Free device memory.
3412 * acc_copyin::                  Allocate device memory and copy host memory to
3413                                 it.
3414 * acc_present_or_copyin::       If the data is not present on the device,
3415                                 allocate device memory and copy from host
3416                                 memory.
3417 * acc_create::                  Allocate device memory and map it to host
3418                                 memory.
3419 * acc_present_or_create::       If the data is not present on the device,
3420                                 allocate device memory and map it to host
3421                                 memory.
3422 * acc_copyout::                 Copy device memory to host memory.
3423 * acc_delete::                  Free device memory.
3424 * acc_update_device::           Update device memory from mapped host memory.
3425 * acc_update_self::             Update host memory from mapped device memory.
3426 * acc_map_data::                Map previously allocated device memory to host
3427                                 memory.
3428 * acc_unmap_data::              Unmap device memory from host memory.
3429 * acc_deviceptr::               Get device pointer associated with specific
3430                                 host address.
3431 * acc_hostptr::                 Get host pointer associated with specific
3432                                 device address.
3433 * acc_is_present::              Indicate whether host variable / array is
3434                                 present on device.
3435 * acc_memcpy_to_device::        Copy host memory to device memory.
3436 * acc_memcpy_from_device::      Copy device memory to host memory.
3437 * acc_attach::                  Let device pointer point to device-pointer target.
3438 * acc_detach::                  Let device pointer point to host-pointer target.
3439
3440 API routines for target platforms.
3441
3442 * acc_get_current_cuda_device:: Get CUDA device handle.
3443 * acc_get_current_cuda_context::Get CUDA context handle.
3444 * acc_get_cuda_stream::         Get CUDA stream handle.
3445 * acc_set_cuda_stream::         Set CUDA stream handle.
3446
3447 API routines for the OpenACC Profiling Interface.
3448
3449 * acc_prof_register::           Register callbacks.
3450 * acc_prof_unregister::         Unregister callbacks.
3451 * acc_prof_lookup::             Obtain inquiry functions.
3452 * acc_register_library::        Library registration.
3453 @end menu
3454
3455
3456
3457 @node acc_get_num_devices
3458 @section @code{acc_get_num_devices} -- Get number of devices for given device type
3459 @table @asis
3460 @item @emph{Description}
3461 This function returns a value indicating the number of devices available
3462 for the device type specified in @var{devicetype}.
3463
3464 @item @emph{C/C++}:
3465 @multitable @columnfractions .20 .80
3466 @item @emph{Prototype}: @tab @code{int acc_get_num_devices(acc_device_t devicetype);}
3467 @end multitable
3468
3469 @item @emph{Fortran}:
3470 @multitable @columnfractions .20 .80
3471 @item @emph{Interface}: @tab @code{integer function acc_get_num_devices(devicetype)}
3472 @item                  @tab @code{integer(kind=acc_device_kind) devicetype}
3473 @end multitable
3474
3475 @item @emph{Reference}:
3476 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3477 3.2.1.
3478 @end table
3479
3480
3481
3482 @node acc_set_device_type
3483 @section @code{acc_set_device_type} -- Set type of device accelerator to use.
3484 @table @asis
3485 @item @emph{Description}
3486 This function indicates to the runtime library which device type, specified
3487 in @var{devicetype}, to use when executing a parallel or kernels region.
3488
3489 @item @emph{C/C++}:
3490 @multitable @columnfractions .20 .80
3491 @item @emph{Prototype}: @tab @code{acc_set_device_type(acc_device_t devicetype);}
3492 @end multitable
3493
3494 @item @emph{Fortran}:
3495 @multitable @columnfractions .20 .80
3496 @item @emph{Interface}: @tab @code{subroutine acc_set_device_type(devicetype)}
3497 @item                   @tab @code{integer(kind=acc_device_kind) devicetype}
3498 @end multitable
3499
3500 @item @emph{Reference}:
3501 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3502 3.2.2.
3503 @end table
3504
3505
3506
3507 @node acc_get_device_type
3508 @section @code{acc_get_device_type} -- Get type of device accelerator to be used.
3509 @table @asis
3510 @item @emph{Description}
3511 This function returns what device type will be used when executing a
3512 parallel or kernels region.
3513
3514 This function returns @code{acc_device_none} if
3515 @code{acc_get_device_type} is called from
3516 @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end}
3517 callbacks of the OpenACC Profiling Interface (@ref{OpenACC Profiling
3518 Interface}), that is, if the device is currently being initialized.
3519
3520 @item @emph{C/C++}:
3521 @multitable @columnfractions .20 .80
3522 @item @emph{Prototype}: @tab @code{acc_device_t acc_get_device_type(void);}
3523 @end multitable
3524
3525 @item @emph{Fortran}:
3526 @multitable @columnfractions .20 .80
3527 @item @emph{Interface}: @tab @code{function acc_get_device_type(void)}
3528 @item                  @tab @code{integer(kind=acc_device_kind) acc_get_device_type}
3529 @end multitable
3530
3531 @item @emph{Reference}:
3532 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3533 3.2.3.
3534 @end table
3535
3536
3537
3538 @node acc_set_device_num
3539 @section @code{acc_set_device_num} -- Set device number to use.
3540 @table @asis
3541 @item @emph{Description}
3542 This function will indicate to the runtime which device number,
3543 specified by @var{devicenum}, associated with the specified device
3544 type @var{devicetype}.
3545
3546 @item @emph{C/C++}:
3547 @multitable @columnfractions .20 .80
3548 @item @emph{Prototype}: @tab @code{acc_set_device_num(int devicenum, acc_device_t devicetype);}
3549 @end multitable
3550
3551 @item @emph{Fortran}:
3552 @multitable @columnfractions .20 .80
3553 @item @emph{Interface}: @tab @code{subroutine acc_set_device_num(devicenum, devicetype)}
3554 @item                   @tab @code{integer devicenum}
3555 @item                   @tab @code{integer(kind=acc_device_kind) devicetype}
3556 @end multitable
3557
3558 @item @emph{Reference}:
3559 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3560 3.2.4.
3561 @end table
3562
3563
3564
3565 @node acc_get_device_num
3566 @section @code{acc_get_device_num} -- Get device number to be used.
3567 @table @asis
3568 @item @emph{Description}
3569 This function returns which device number associated with the specified device
3570 type @var{devicetype}, will be used when executing a parallel or kernels
3571 region.
3572
3573 @item @emph{C/C++}:
3574 @multitable @columnfractions .20 .80
3575 @item @emph{Prototype}: @tab @code{int acc_get_device_num(acc_device_t devicetype);}
3576 @end multitable
3577
3578 @item @emph{Fortran}:
3579 @multitable @columnfractions .20 .80
3580 @item @emph{Interface}: @tab @code{function acc_get_device_num(devicetype)}
3581 @item                   @tab @code{integer(kind=acc_device_kind) devicetype}
3582 @item                   @tab @code{integer acc_get_device_num}
3583 @end multitable
3584
3585 @item @emph{Reference}:
3586 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3587 3.2.5.
3588 @end table
3589
3590
3591
3592 @node acc_get_property
3593 @section @code{acc_get_property} -- Get device property.
3594 @cindex acc_get_property
3595 @cindex acc_get_property_string
3596 @table @asis
3597 @item @emph{Description}
3598 These routines return the value of the specified @var{property} for the
3599 device being queried according to @var{devicenum} and @var{devicetype}.
3600 Integer-valued and string-valued properties are returned by
3601 @code{acc_get_property} and @code{acc_get_property_string} respectively.
3602 The Fortran @code{acc_get_property_string} subroutine returns the string
3603 retrieved in its fourth argument while the remaining entry points are
3604 functions, which pass the return value as their result.
3605
3606 Note for Fortran, only: the OpenACC technical committee corrected and, hence,
3607 modified the interface introduced in OpenACC 2.6.  The kind-value parameter
3608 @code{acc_device_property} has been renamed to @code{acc_device_property_kind}
3609 for consistency and the return type of the @code{acc_get_property} function is
3610 now a @code{c_size_t} integer instead of a @code{acc_device_property} integer.
3611 The parameter @code{acc_device_property} will continue to be provided,
3612 but might be removed in a future version of GCC.
3613
3614 @item @emph{C/C++}:
3615 @multitable @columnfractions .20 .80
3616 @item @emph{Prototype}: @tab @code{size_t acc_get_property(int devicenum, acc_device_t devicetype, acc_device_property_t property);}
3617 @item @emph{Prototype}: @tab @code{const char *acc_get_property_string(int devicenum, acc_device_t devicetype, acc_device_property_t property);}
3618 @end multitable
3619
3620 @item @emph{Fortran}:
3621 @multitable @columnfractions .20 .80
3622 @item @emph{Interface}: @tab @code{function acc_get_property(devicenum, devicetype, property)}
3623 @item @emph{Interface}: @tab @code{subroutine acc_get_property_string(devicenum, devicetype, property, string)}
3624 @item                   @tab @code{use ISO_C_Binding, only: c_size_t}
3625 @item                   @tab @code{integer devicenum}
3626 @item                   @tab @code{integer(kind=acc_device_kind) devicetype}
3627 @item                   @tab @code{integer(kind=acc_device_property_kind) property}
3628 @item                   @tab @code{integer(kind=c_size_t) acc_get_property}
3629 @item                   @tab @code{character(*) string}
3630 @end multitable
3631
3632 @item @emph{Reference}:
3633 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3634 3.2.6.
3635 @end table
3636
3637
3638
3639 @node acc_async_test
3640 @section @code{acc_async_test} -- Test for completion of a specific asynchronous operation.
3641 @table @asis
3642 @item @emph{Description}
3643 This function tests for completion of the asynchronous operation specified
3644 in @var{arg}. In C/C++, a non-zero value will be returned to indicate
3645 the specified asynchronous operation has completed. While Fortran will return
3646 a @code{true}. If the asynchronous operation has not completed, C/C++ returns
3647 a zero and Fortran returns a @code{false}.
3648
3649 @item @emph{C/C++}:
3650 @multitable @columnfractions .20 .80
3651 @item @emph{Prototype}: @tab @code{int acc_async_test(int arg);}
3652 @end multitable
3653
3654 @item @emph{Fortran}:
3655 @multitable @columnfractions .20 .80
3656 @item @emph{Interface}: @tab @code{function acc_async_test(arg)}
3657 @item                   @tab @code{integer(kind=acc_handle_kind) arg}
3658 @item                   @tab @code{logical acc_async_test}
3659 @end multitable
3660
3661 @item @emph{Reference}:
3662 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3663 3.2.9.
3664 @end table
3665
3666
3667
3668 @node acc_async_test_all
3669 @section @code{acc_async_test_all} -- Tests for completion of all asynchronous operations.
3670 @table @asis
3671 @item @emph{Description}
3672 This function tests for completion of all asynchronous operations.
3673 In C/C++, a non-zero value will be returned to indicate all asynchronous
3674 operations have completed. While Fortran will return a @code{true}. If
3675 any asynchronous operation has not completed, C/C++ returns a zero and
3676 Fortran returns a @code{false}.
3677
3678 @item @emph{C/C++}:
3679 @multitable @columnfractions .20 .80
3680 @item @emph{Prototype}: @tab @code{int acc_async_test_all(void);}
3681 @end multitable
3682
3683 @item @emph{Fortran}:
3684 @multitable @columnfractions .20 .80
3685 @item @emph{Interface}: @tab @code{function acc_async_test()}
3686 @item                   @tab @code{logical acc_get_device_num}
3687 @end multitable
3688
3689 @item @emph{Reference}:
3690 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3691 3.2.10.
3692 @end table
3693
3694
3695
3696 @node acc_wait
3697 @section @code{acc_wait} -- Wait for completion of a specific asynchronous operation.
3698 @table @asis
3699 @item @emph{Description}
3700 This function waits for completion of the asynchronous operation
3701 specified in @var{arg}.
3702
3703 @item @emph{C/C++}:
3704 @multitable @columnfractions .20 .80
3705 @item @emph{Prototype}: @tab @code{acc_wait(arg);}
3706 @item @emph{Prototype (OpenACC 1.0 compatibility)}: @tab @code{acc_async_wait(arg);}
3707 @end multitable
3708
3709 @item @emph{Fortran}:
3710 @multitable @columnfractions .20 .80
3711 @item @emph{Interface}: @tab @code{subroutine acc_wait(arg)}
3712 @item                   @tab @code{integer(acc_handle_kind) arg}
3713 @item @emph{Interface (OpenACC 1.0 compatibility)}: @tab @code{subroutine acc_async_wait(arg)}
3714 @item                                               @tab @code{integer(acc_handle_kind) arg}
3715 @end multitable
3716
3717 @item @emph{Reference}:
3718 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3719 3.2.11.
3720 @end table
3721
3722
3723
3724 @node acc_wait_all
3725 @section @code{acc_wait_all} -- Waits for completion of all asynchronous operations.
3726 @table @asis
3727 @item @emph{Description}
3728 This function waits for the completion of all asynchronous operations.
3729
3730 @item @emph{C/C++}:
3731 @multitable @columnfractions .20 .80
3732 @item @emph{Prototype}: @tab @code{acc_wait_all(void);}
3733 @item @emph{Prototype (OpenACC 1.0 compatibility)}: @tab @code{acc_async_wait_all(void);}
3734 @end multitable
3735
3736 @item @emph{Fortran}:
3737 @multitable @columnfractions .20 .80
3738 @item @emph{Interface}: @tab @code{subroutine acc_wait_all()}
3739 @item @emph{Interface (OpenACC 1.0 compatibility)}: @tab @code{subroutine acc_async_wait_all()}
3740 @end multitable
3741
3742 @item @emph{Reference}:
3743 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3744 3.2.13.
3745 @end table
3746
3747
3748
3749 @node acc_wait_all_async
3750 @section @code{acc_wait_all_async} -- Wait for completion of all asynchronous operations.
3751 @table @asis
3752 @item @emph{Description}
3753 This function enqueues a wait operation on the queue @var{async} for any
3754 and all asynchronous operations that have been previously enqueued on
3755 any queue.
3756
3757 @item @emph{C/C++}:
3758 @multitable @columnfractions .20 .80
3759 @item @emph{Prototype}: @tab @code{acc_wait_all_async(int async);}
3760 @end multitable
3761
3762 @item @emph{Fortran}:
3763 @multitable @columnfractions .20 .80
3764 @item @emph{Interface}: @tab @code{subroutine acc_wait_all_async(async)}
3765 @item                   @tab @code{integer(acc_handle_kind) async}
3766 @end multitable
3767
3768 @item @emph{Reference}:
3769 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3770 3.2.14.
3771 @end table
3772
3773
3774
3775 @node acc_wait_async
3776 @section @code{acc_wait_async} -- Wait for completion of asynchronous operations.
3777 @table @asis
3778 @item @emph{Description}
3779 This function enqueues a wait operation on queue @var{async} for any and all
3780 asynchronous operations enqueued on queue @var{arg}.
3781
3782 @item @emph{C/C++}:
3783 @multitable @columnfractions .20 .80
3784 @item @emph{Prototype}: @tab @code{acc_wait_async(int arg, int async);}
3785 @end multitable
3786
3787 @item @emph{Fortran}:
3788 @multitable @columnfractions .20 .80
3789 @item @emph{Interface}: @tab @code{subroutine acc_wait_async(arg, async)}
3790 @item                   @tab @code{integer(acc_handle_kind) arg, async}
3791 @end multitable
3792
3793 @item @emph{Reference}:
3794 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3795 3.2.12.
3796 @end table
3797
3798
3799
3800 @node acc_init
3801 @section @code{acc_init} -- Initialize runtime for a specific device type.
3802 @table @asis
3803 @item @emph{Description}
3804 This function initializes the runtime for the device type specified in
3805 @var{devicetype}.
3806
3807 @item @emph{C/C++}:
3808 @multitable @columnfractions .20 .80
3809 @item @emph{Prototype}: @tab @code{acc_init(acc_device_t devicetype);}
3810 @end multitable
3811
3812 @item @emph{Fortran}:
3813 @multitable @columnfractions .20 .80
3814 @item @emph{Interface}: @tab @code{subroutine acc_init(devicetype)}
3815 @item                   @tab @code{integer(acc_device_kind) devicetype}
3816 @end multitable
3817
3818 @item @emph{Reference}:
3819 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3820 3.2.7.
3821 @end table
3822
3823
3824
3825 @node acc_shutdown
3826 @section @code{acc_shutdown} -- Shuts down the runtime for a specific device type.
3827 @table @asis
3828 @item @emph{Description}
3829 This function shuts down the runtime for the device type specified in
3830 @var{devicetype}.
3831
3832 @item @emph{C/C++}:
3833 @multitable @columnfractions .20 .80
3834 @item @emph{Prototype}: @tab @code{acc_shutdown(acc_device_t devicetype);}
3835 @end multitable
3836
3837 @item @emph{Fortran}:
3838 @multitable @columnfractions .20 .80
3839 @item @emph{Interface}: @tab @code{subroutine acc_shutdown(devicetype)}
3840 @item                   @tab @code{integer(acc_device_kind) devicetype}
3841 @end multitable
3842
3843 @item @emph{Reference}:
3844 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3845 3.2.8.
3846 @end table
3847
3848
3849
3850 @node acc_on_device
3851 @section @code{acc_on_device} -- Whether executing on a particular device
3852 @table @asis
3853 @item @emph{Description}:
3854 This function returns whether the program is executing on a particular
3855 device specified in @var{devicetype}. In C/C++ a non-zero value is
3856 returned to indicate the device is executing on the specified device type.
3857 In Fortran, @code{true} will be returned. If the program is not executing
3858 on the specified device type C/C++ will return a zero, while Fortran will
3859 return @code{false}.
3860
3861 @item @emph{C/C++}:
3862 @multitable @columnfractions .20 .80
3863 @item @emph{Prototype}: @tab @code{acc_on_device(acc_device_t devicetype);}
3864 @end multitable
3865
3866 @item @emph{Fortran}:
3867 @multitable @columnfractions .20 .80
3868 @item @emph{Interface}: @tab @code{function acc_on_device(devicetype)}
3869 @item                   @tab @code{integer(acc_device_kind) devicetype}
3870 @item                   @tab @code{logical acc_on_device}
3871 @end multitable
3872
3873
3874 @item @emph{Reference}:
3875 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3876 3.2.17.
3877 @end table
3878
3879
3880
3881 @node acc_malloc
3882 @section @code{acc_malloc} -- Allocate device memory.
3883 @table @asis
3884 @item @emph{Description}
3885 This function allocates @var{len} bytes of device memory. It returns
3886 the device address of the allocated memory.
3887
3888 @item @emph{C/C++}:
3889 @multitable @columnfractions .20 .80
3890 @item @emph{Prototype}: @tab @code{d_void* acc_malloc(size_t len);}
3891 @end multitable
3892
3893 @item @emph{Reference}:
3894 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3895 3.2.18.
3896 @end table
3897
3898
3899
3900 @node acc_free
3901 @section @code{acc_free} -- Free device memory.
3902 @table @asis
3903 @item @emph{Description}
3904 Free previously allocated device memory at the device address @code{a}.
3905
3906 @item @emph{C/C++}:
3907 @multitable @columnfractions .20 .80
3908 @item @emph{Prototype}: @tab @code{acc_free(d_void *a);}
3909 @end multitable
3910
3911 @item @emph{Reference}:
3912 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3913 3.2.19.
3914 @end table
3915
3916
3917
3918 @node acc_copyin
3919 @section @code{acc_copyin} -- Allocate device memory and copy host memory to it.
3920 @table @asis
3921 @item @emph{Description}
3922 In C/C++, this function allocates @var{len} bytes of device memory
3923 and maps it to the specified host address in @var{a}. The device
3924 address of the newly allocated device memory is returned.
3925
3926 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
3927 a contiguous array section. The second form @var{a} specifies a
3928 variable or array element and @var{len} specifies the length in bytes.
3929
3930 @item @emph{C/C++}:
3931 @multitable @columnfractions .20 .80
3932 @item @emph{Prototype}: @tab @code{void *acc_copyin(h_void *a, size_t len);}
3933 @item @emph{Prototype}: @tab @code{void *acc_copyin_async(h_void *a, size_t len, int async);}
3934 @end multitable
3935
3936 @item @emph{Fortran}:
3937 @multitable @columnfractions .20 .80
3938 @item @emph{Interface}: @tab @code{subroutine acc_copyin(a)}
3939 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3940 @item @emph{Interface}: @tab @code{subroutine acc_copyin(a, len)}
3941 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3942 @item                   @tab @code{integer len}
3943 @item @emph{Interface}: @tab @code{subroutine acc_copyin_async(a, async)}
3944 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3945 @item                   @tab @code{integer(acc_handle_kind) :: async}
3946 @item @emph{Interface}: @tab @code{subroutine acc_copyin_async(a, len, async)}
3947 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3948 @item                   @tab @code{integer len}
3949 @item                   @tab @code{integer(acc_handle_kind) :: async}
3950 @end multitable
3951
3952 @item @emph{Reference}:
3953 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3954 3.2.20.
3955 @end table
3956
3957
3958
3959 @node acc_present_or_copyin
3960 @section @code{acc_present_or_copyin} -- If the data is not present on the device, allocate device memory and copy from host memory.
3961 @table @asis
3962 @item @emph{Description}
3963 This function tests if the host data specified by @var{a} and of length
3964 @var{len} is present or not. If it is not present, then device memory
3965 will be allocated and the host memory copied. The device address of
3966 the newly allocated device memory is returned.
3967
3968 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
3969 a contiguous array section. The second form @var{a} specifies a variable or
3970 array element and @var{len} specifies the length in bytes.
3971
3972 Note that @code{acc_present_or_copyin} and @code{acc_pcopyin} exist for
3973 backward compatibility with OpenACC 2.0; use @ref{acc_copyin} instead.
3974
3975 @item @emph{C/C++}:
3976 @multitable @columnfractions .20 .80
3977 @item @emph{Prototype}: @tab @code{void *acc_present_or_copyin(h_void *a, size_t len);}
3978 @item @emph{Prototype}: @tab @code{void *acc_pcopyin(h_void *a, size_t len);}
3979 @end multitable
3980
3981 @item @emph{Fortran}:
3982 @multitable @columnfractions .20 .80
3983 @item @emph{Interface}: @tab @code{subroutine acc_present_or_copyin(a)}
3984 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3985 @item @emph{Interface}: @tab @code{subroutine acc_present_or_copyin(a, len)}
3986 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3987 @item                   @tab @code{integer len}
3988 @item @emph{Interface}: @tab @code{subroutine acc_pcopyin(a)}
3989 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3990 @item @emph{Interface}: @tab @code{subroutine acc_pcopyin(a, len)}
3991 @item                   @tab @code{type, dimension(:[,:]...) :: a}
3992 @item                   @tab @code{integer len}
3993 @end multitable
3994
3995 @item @emph{Reference}:
3996 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
3997 3.2.20.
3998 @end table
3999
4000
4001
4002 @node acc_create
4003 @section @code{acc_create} -- Allocate device memory and map it to host memory.
4004 @table @asis
4005 @item @emph{Description}
4006 This function allocates device memory and maps it to host memory specified
4007 by the host address @var{a} with a length of @var{len} bytes. In C/C++,
4008 the function returns the device address of the allocated device memory.
4009
4010 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4011 a contiguous array section. The second form @var{a} specifies a variable or
4012 array element and @var{len} specifies the length in bytes.
4013
4014 @item @emph{C/C++}:
4015 @multitable @columnfractions .20 .80
4016 @item @emph{Prototype}: @tab @code{void *acc_create(h_void *a, size_t len);}
4017 @item @emph{Prototype}: @tab @code{void *acc_create_async(h_void *a, size_t len, int async);}
4018 @end multitable
4019
4020 @item @emph{Fortran}:
4021 @multitable @columnfractions .20 .80
4022 @item @emph{Interface}: @tab @code{subroutine acc_create(a)}
4023 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4024 @item @emph{Interface}: @tab @code{subroutine acc_create(a, len)}
4025 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4026 @item                   @tab @code{integer len}
4027 @item @emph{Interface}: @tab @code{subroutine acc_create_async(a, async)}
4028 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4029 @item                   @tab @code{integer(acc_handle_kind) :: async}
4030 @item @emph{Interface}: @tab @code{subroutine acc_create_async(a, len, async)}
4031 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4032 @item                   @tab @code{integer len}
4033 @item                   @tab @code{integer(acc_handle_kind) :: async}
4034 @end multitable
4035
4036 @item @emph{Reference}:
4037 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4038 3.2.21.
4039 @end table
4040
4041
4042
4043 @node acc_present_or_create
4044 @section @code{acc_present_or_create} -- If the data is not present on the device, allocate device memory and map it to host memory.
4045 @table @asis
4046 @item @emph{Description}
4047 This function tests if the host data specified by @var{a} and of length
4048 @var{len} is present or not. If it is not present, then device memory
4049 will be allocated and mapped to host memory. In C/C++, the device address
4050 of the newly allocated device memory is returned.
4051
4052 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4053 a contiguous array section. The second form @var{a} specifies a variable or
4054 array element and @var{len} specifies the length in bytes.
4055
4056 Note that @code{acc_present_or_create} and @code{acc_pcreate} exist for
4057 backward compatibility with OpenACC 2.0; use @ref{acc_create} instead.
4058
4059 @item @emph{C/C++}:
4060 @multitable @columnfractions .20 .80
4061 @item @emph{Prototype}: @tab @code{void *acc_present_or_create(h_void *a, size_t len)}
4062 @item @emph{Prototype}: @tab @code{void *acc_pcreate(h_void *a, size_t len)}
4063 @end multitable
4064
4065 @item @emph{Fortran}:
4066 @multitable @columnfractions .20 .80
4067 @item @emph{Interface}: @tab @code{subroutine acc_present_or_create(a)}
4068 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4069 @item @emph{Interface}: @tab @code{subroutine acc_present_or_create(a, len)}
4070 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4071 @item                   @tab @code{integer len}
4072 @item @emph{Interface}: @tab @code{subroutine acc_pcreate(a)}
4073 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4074 @item @emph{Interface}: @tab @code{subroutine acc_pcreate(a, len)}
4075 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4076 @item                   @tab @code{integer len}
4077 @end multitable
4078
4079 @item @emph{Reference}:
4080 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4081 3.2.21.
4082 @end table
4083
4084
4085
4086 @node acc_copyout
4087 @section @code{acc_copyout} -- Copy device memory to host memory.
4088 @table @asis
4089 @item @emph{Description}
4090 This function copies mapped device memory to host memory which is specified
4091 by host address @var{a} for a length @var{len} bytes in C/C++.
4092
4093 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4094 a contiguous array section. The second form @var{a} specifies a variable or
4095 array element and @var{len} specifies the length in bytes.
4096
4097 @item @emph{C/C++}:
4098 @multitable @columnfractions .20 .80
4099 @item @emph{Prototype}: @tab @code{acc_copyout(h_void *a, size_t len);}
4100 @item @emph{Prototype}: @tab @code{acc_copyout_async(h_void *a, size_t len, int async);}
4101 @item @emph{Prototype}: @tab @code{acc_copyout_finalize(h_void *a, size_t len);}
4102 @item @emph{Prototype}: @tab @code{acc_copyout_finalize_async(h_void *a, size_t len, int async);}
4103 @end multitable
4104
4105 @item @emph{Fortran}:
4106 @multitable @columnfractions .20 .80
4107 @item @emph{Interface}: @tab @code{subroutine acc_copyout(a)}
4108 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4109 @item @emph{Interface}: @tab @code{subroutine acc_copyout(a, len)}
4110 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4111 @item                   @tab @code{integer len}
4112 @item @emph{Interface}: @tab @code{subroutine acc_copyout_async(a, async)}
4113 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4114 @item                   @tab @code{integer(acc_handle_kind) :: async}
4115 @item @emph{Interface}: @tab @code{subroutine acc_copyout_async(a, len, async)}
4116 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4117 @item                   @tab @code{integer len}
4118 @item                   @tab @code{integer(acc_handle_kind) :: async}
4119 @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize(a)}
4120 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4121 @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize(a, len)}
4122 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4123 @item                   @tab @code{integer len}
4124 @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize_async(a, async)}
4125 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4126 @item                   @tab @code{integer(acc_handle_kind) :: async}
4127 @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize_async(a, len, async)}
4128 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4129 @item                   @tab @code{integer len}
4130 @item                   @tab @code{integer(acc_handle_kind) :: async}
4131 @end multitable
4132
4133 @item @emph{Reference}:
4134 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4135 3.2.22.
4136 @end table
4137
4138
4139
4140 @node acc_delete
4141 @section @code{acc_delete} -- Free device memory.
4142 @table @asis
4143 @item @emph{Description}
4144 This function frees previously allocated device memory specified by
4145 the device address @var{a} and the length of @var{len} bytes.
4146
4147 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4148 a contiguous array section. The second form @var{a} specifies a variable or
4149 array element and @var{len} specifies the length in bytes.
4150
4151 @item @emph{C/C++}:
4152 @multitable @columnfractions .20 .80
4153 @item @emph{Prototype}: @tab @code{acc_delete(h_void *a, size_t len);}
4154 @item @emph{Prototype}: @tab @code{acc_delete_async(h_void *a, size_t len, int async);}
4155 @item @emph{Prototype}: @tab @code{acc_delete_finalize(h_void *a, size_t len);}
4156 @item @emph{Prototype}: @tab @code{acc_delete_finalize_async(h_void *a, size_t len, int async);}
4157 @end multitable
4158
4159 @item @emph{Fortran}:
4160 @multitable @columnfractions .20 .80
4161 @item @emph{Interface}: @tab @code{subroutine acc_delete(a)}
4162 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4163 @item @emph{Interface}: @tab @code{subroutine acc_delete(a, len)}
4164 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4165 @item                   @tab @code{integer len}
4166 @item @emph{Interface}: @tab @code{subroutine acc_delete_async(a, async)}
4167 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4168 @item                   @tab @code{integer(acc_handle_kind) :: async}
4169 @item @emph{Interface}: @tab @code{subroutine acc_delete_async(a, len, async)}
4170 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4171 @item                   @tab @code{integer len}
4172 @item                   @tab @code{integer(acc_handle_kind) :: async}
4173 @item @emph{Interface}: @tab @code{subroutine acc_delete_finalize(a)}
4174 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4175 @item @emph{Interface}: @tab @code{subroutine acc_delete_finalize(a, len)}
4176 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4177 @item                   @tab @code{integer len}
4178 @item @emph{Interface}: @tab @code{subroutine acc_delete_async_finalize(a, async)}
4179 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4180 @item                   @tab @code{integer(acc_handle_kind) :: async}
4181 @item @emph{Interface}: @tab @code{subroutine acc_delete_async_finalize(a, len, async)}
4182 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4183 @item                   @tab @code{integer len}
4184 @item                   @tab @code{integer(acc_handle_kind) :: async}
4185 @end multitable
4186
4187 @item @emph{Reference}:
4188 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4189 3.2.23.
4190 @end table
4191
4192
4193
4194 @node acc_update_device
4195 @section @code{acc_update_device} -- Update device memory from mapped host memory.
4196 @table @asis
4197 @item @emph{Description}
4198 This function updates the device copy from the previously mapped host memory.
4199 The host memory is specified with the host address @var{a} and a length of
4200 @var{len} bytes.
4201
4202 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4203 a contiguous array section. The second form @var{a} specifies a variable or
4204 array element and @var{len} specifies the length in bytes.
4205
4206 @item @emph{C/C++}:
4207 @multitable @columnfractions .20 .80
4208 @item @emph{Prototype}: @tab @code{acc_update_device(h_void *a, size_t len);}
4209 @item @emph{Prototype}: @tab @code{acc_update_device(h_void *a, size_t len, async);}
4210 @end multitable
4211
4212 @item @emph{Fortran}:
4213 @multitable @columnfractions .20 .80
4214 @item @emph{Interface}: @tab @code{subroutine acc_update_device(a)}
4215 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4216 @item @emph{Interface}: @tab @code{subroutine acc_update_device(a, len)}
4217 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4218 @item                   @tab @code{integer len}
4219 @item @emph{Interface}: @tab @code{subroutine acc_update_device_async(a, async)}
4220 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4221 @item                   @tab @code{integer(acc_handle_kind) :: async}
4222 @item @emph{Interface}: @tab @code{subroutine acc_update_device_async(a, len, async)}
4223 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4224 @item                   @tab @code{integer len}
4225 @item                   @tab @code{integer(acc_handle_kind) :: async}
4226 @end multitable
4227
4228 @item @emph{Reference}:
4229 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4230 3.2.24.
4231 @end table
4232
4233
4234
4235 @node acc_update_self
4236 @section @code{acc_update_self} -- Update host memory from mapped device memory.
4237 @table @asis
4238 @item @emph{Description}
4239 This function updates the host copy from the previously mapped device memory.
4240 The host memory is specified with the host address @var{a} and a length of
4241 @var{len} bytes.
4242
4243 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4244 a contiguous array section. The second form @var{a} specifies a variable or
4245 array element and @var{len} specifies the length in bytes.
4246
4247 @item @emph{C/C++}:
4248 @multitable @columnfractions .20 .80
4249 @item @emph{Prototype}: @tab @code{acc_update_self(h_void *a, size_t len);}
4250 @item @emph{Prototype}: @tab @code{acc_update_self_async(h_void *a, size_t len, int async);}
4251 @end multitable
4252
4253 @item @emph{Fortran}:
4254 @multitable @columnfractions .20 .80
4255 @item @emph{Interface}: @tab @code{subroutine acc_update_self(a)}
4256 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4257 @item @emph{Interface}: @tab @code{subroutine acc_update_self(a, len)}
4258 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4259 @item                   @tab @code{integer len}
4260 @item @emph{Interface}: @tab @code{subroutine acc_update_self_async(a, async)}
4261 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4262 @item                   @tab @code{integer(acc_handle_kind) :: async}
4263 @item @emph{Interface}: @tab @code{subroutine acc_update_self_async(a, len, async)}
4264 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4265 @item                   @tab @code{integer len}
4266 @item                   @tab @code{integer(acc_handle_kind) :: async}
4267 @end multitable
4268
4269 @item @emph{Reference}:
4270 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4271 3.2.25.
4272 @end table
4273
4274
4275
4276 @node acc_map_data
4277 @section @code{acc_map_data} -- Map previously allocated device memory to host memory.
4278 @table @asis
4279 @item @emph{Description}
4280 This function maps previously allocated device and host memory. The device
4281 memory is specified with the device address @var{d}. The host memory is
4282 specified with the host address @var{h} and a length of @var{len}.
4283
4284 @item @emph{C/C++}:
4285 @multitable @columnfractions .20 .80
4286 @item @emph{Prototype}: @tab @code{acc_map_data(h_void *h, d_void *d, size_t len);}
4287 @end multitable
4288
4289 @item @emph{Reference}:
4290 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4291 3.2.26.
4292 @end table
4293
4294
4295
4296 @node acc_unmap_data
4297 @section @code{acc_unmap_data} -- Unmap device memory from host memory.
4298 @table @asis
4299 @item @emph{Description}
4300 This function unmaps previously mapped device and host memory. The latter
4301 specified by @var{h}.
4302
4303 @item @emph{C/C++}:
4304 @multitable @columnfractions .20 .80
4305 @item @emph{Prototype}: @tab @code{acc_unmap_data(h_void *h);}
4306 @end multitable
4307
4308 @item @emph{Reference}:
4309 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4310 3.2.27.
4311 @end table
4312
4313
4314
4315 @node acc_deviceptr
4316 @section @code{acc_deviceptr} -- Get device pointer associated with specific host address.
4317 @table @asis
4318 @item @emph{Description}
4319 This function returns the device address that has been mapped to the
4320 host address specified by @var{h}.
4321
4322 @item @emph{C/C++}:
4323 @multitable @columnfractions .20 .80
4324 @item @emph{Prototype}: @tab @code{void *acc_deviceptr(h_void *h);}
4325 @end multitable
4326
4327 @item @emph{Reference}:
4328 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4329 3.2.28.
4330 @end table
4331
4332
4333
4334 @node acc_hostptr
4335 @section @code{acc_hostptr} -- Get host pointer associated with specific device address.
4336 @table @asis
4337 @item @emph{Description}
4338 This function returns the host address that has been mapped to the
4339 device address specified by @var{d}.
4340
4341 @item @emph{C/C++}:
4342 @multitable @columnfractions .20 .80
4343 @item @emph{Prototype}: @tab @code{void *acc_hostptr(d_void *d);}
4344 @end multitable
4345
4346 @item @emph{Reference}:
4347 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4348 3.2.29.
4349 @end table
4350
4351
4352
4353 @node acc_is_present
4354 @section @code{acc_is_present} -- Indicate whether host variable / array is present on device.
4355 @table @asis
4356 @item @emph{Description}
4357 This function indicates whether the specified host address in @var{a} and a
4358 length of @var{len} bytes is present on the device. In C/C++, a non-zero
4359 value is returned to indicate the presence of the mapped memory on the
4360 device. A zero is returned to indicate the memory is not mapped on the
4361 device.
4362
4363 In Fortran, two (2) forms are supported. In the first form, @var{a} specifies
4364 a contiguous array section. The second form @var{a} specifies a variable or
4365 array element and @var{len} specifies the length in bytes. If the host
4366 memory is mapped to device memory, then a @code{true} is returned. Otherwise,
4367 a @code{false} is return to indicate the mapped memory is not present.
4368
4369 @item @emph{C/C++}:
4370 @multitable @columnfractions .20 .80
4371 @item @emph{Prototype}: @tab @code{int acc_is_present(h_void *a, size_t len);}
4372 @end multitable
4373
4374 @item @emph{Fortran}:
4375 @multitable @columnfractions .20 .80
4376 @item @emph{Interface}: @tab @code{function acc_is_present(a)}
4377 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4378 @item                   @tab @code{logical acc_is_present}
4379 @item @emph{Interface}: @tab @code{function acc_is_present(a, len)}
4380 @item                   @tab @code{type, dimension(:[,:]...) :: a}
4381 @item                   @tab @code{integer len}
4382 @item                   @tab @code{logical acc_is_present}
4383 @end multitable
4384
4385 @item @emph{Reference}:
4386 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4387 3.2.30.
4388 @end table
4389
4390
4391
4392 @node acc_memcpy_to_device
4393 @section @code{acc_memcpy_to_device} -- Copy host memory to device memory.
4394 @table @asis
4395 @item @emph{Description}
4396 This function copies host memory specified by host address of @var{src} to
4397 device memory specified by the device address @var{dest} for a length of
4398 @var{bytes} bytes.
4399
4400 @item @emph{C/C++}:
4401 @multitable @columnfractions .20 .80
4402 @item @emph{Prototype}: @tab @code{acc_memcpy_to_device(d_void *dest, h_void *src, size_t bytes);}
4403 @end multitable
4404
4405 @item @emph{Reference}:
4406 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4407 3.2.31.
4408 @end table
4409
4410
4411
4412 @node acc_memcpy_from_device
4413 @section @code{acc_memcpy_from_device} -- Copy device memory to host memory.
4414 @table @asis
4415 @item @emph{Description}
4416 This function copies host memory specified by host address of @var{src} from
4417 device memory specified by the device address @var{dest} for a length of
4418 @var{bytes} bytes.
4419
4420 @item @emph{C/C++}:
4421 @multitable @columnfractions .20 .80
4422 @item @emph{Prototype}: @tab @code{acc_memcpy_from_device(d_void *dest, h_void *src, size_t bytes);}
4423 @end multitable
4424
4425 @item @emph{Reference}:
4426 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4427 3.2.32.
4428 @end table
4429
4430
4431
4432 @node acc_attach
4433 @section @code{acc_attach} -- Let device pointer point to device-pointer target.
4434 @table @asis
4435 @item @emph{Description}
4436 This function updates a pointer on the device from pointing to a host-pointer
4437 address to pointing to the corresponding device data.
4438
4439 @item @emph{C/C++}:
4440 @multitable @columnfractions .20 .80
4441 @item @emph{Prototype}: @tab @code{acc_attach(h_void **ptr);}
4442 @item @emph{Prototype}: @tab @code{acc_attach_async(h_void **ptr, int async);}
4443 @end multitable
4444
4445 @item @emph{Reference}:
4446 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4447 3.2.34.
4448 @end table
4449
4450
4451
4452 @node acc_detach
4453 @section @code{acc_detach} -- Let device pointer point to host-pointer target.
4454 @table @asis
4455 @item @emph{Description}
4456 This function updates a pointer on the device from pointing to a device-pointer
4457 address to pointing to the corresponding host data.
4458
4459 @item @emph{C/C++}:
4460 @multitable @columnfractions .20 .80
4461 @item @emph{Prototype}: @tab @code{acc_detach(h_void **ptr);}
4462 @item @emph{Prototype}: @tab @code{acc_detach_async(h_void **ptr, int async);}
4463 @item @emph{Prototype}: @tab @code{acc_detach_finalize(h_void **ptr);}
4464 @item @emph{Prototype}: @tab @code{acc_detach_finalize_async(h_void **ptr, int async);}
4465 @end multitable
4466
4467 @item @emph{Reference}:
4468 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4469 3.2.35.
4470 @end table
4471
4472
4473
4474 @node acc_get_current_cuda_device
4475 @section @code{acc_get_current_cuda_device} -- Get CUDA device handle.
4476 @table @asis
4477 @item @emph{Description}
4478 This function returns the CUDA device handle. This handle is the same
4479 as used by the CUDA Runtime or Driver API's.
4480
4481 @item @emph{C/C++}:
4482 @multitable @columnfractions .20 .80
4483 @item @emph{Prototype}: @tab @code{void *acc_get_current_cuda_device(void);}
4484 @end multitable
4485
4486 @item @emph{Reference}:
4487 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4488 A.2.1.1.
4489 @end table
4490
4491
4492
4493 @node acc_get_current_cuda_context
4494 @section @code{acc_get_current_cuda_context} -- Get CUDA context handle.
4495 @table @asis
4496 @item @emph{Description}
4497 This function returns the CUDA context handle. This handle is the same
4498 as used by the CUDA Runtime or Driver API's.
4499
4500 @item @emph{C/C++}:
4501 @multitable @columnfractions .20 .80
4502 @item @emph{Prototype}: @tab @code{void *acc_get_current_cuda_context(void);}
4503 @end multitable
4504
4505 @item @emph{Reference}:
4506 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4507 A.2.1.2.
4508 @end table
4509
4510
4511
4512 @node acc_get_cuda_stream
4513 @section @code{acc_get_cuda_stream} -- Get CUDA stream handle.
4514 @table @asis
4515 @item @emph{Description}
4516 This function returns the CUDA stream handle for the queue @var{async}.
4517 This handle is the same as used by the CUDA Runtime or Driver API's.
4518
4519 @item @emph{C/C++}:
4520 @multitable @columnfractions .20 .80
4521 @item @emph{Prototype}: @tab @code{void *acc_get_cuda_stream(int async);}
4522 @end multitable
4523
4524 @item @emph{Reference}:
4525 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4526 A.2.1.3.
4527 @end table
4528
4529
4530
4531 @node acc_set_cuda_stream
4532 @section @code{acc_set_cuda_stream} -- Set CUDA stream handle.
4533 @table @asis
4534 @item @emph{Description}
4535 This function associates the stream handle specified by @var{stream} with
4536 the queue @var{async}.
4537
4538 This cannot be used to change the stream handle associated with
4539 @code{acc_async_sync}.
4540
4541 The return value is not specified.
4542
4543 @item @emph{C/C++}:
4544 @multitable @columnfractions .20 .80
4545 @item @emph{Prototype}: @tab @code{int acc_set_cuda_stream(int async, void *stream);}
4546 @end multitable
4547
4548 @item @emph{Reference}:
4549 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4550 A.2.1.4.
4551 @end table
4552
4553
4554
4555 @node acc_prof_register
4556 @section @code{acc_prof_register} -- Register callbacks.
4557 @table @asis
4558 @item @emph{Description}:
4559 This function registers callbacks.
4560
4561 @item @emph{C/C++}:
4562 @multitable @columnfractions .20 .80
4563 @item @emph{Prototype}: @tab @code{void acc_prof_register (acc_event_t, acc_prof_callback, acc_register_t);}
4564 @end multitable
4565
4566 @item @emph{See also}:
4567 @ref{OpenACC Profiling Interface}
4568
4569 @item @emph{Reference}:
4570 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4571 5.3.
4572 @end table
4573
4574
4575
4576 @node acc_prof_unregister
4577 @section @code{acc_prof_unregister} -- Unregister callbacks.
4578 @table @asis
4579 @item @emph{Description}:
4580 This function unregisters callbacks.
4581
4582 @item @emph{C/C++}:
4583 @multitable @columnfractions .20 .80
4584 @item @emph{Prototype}: @tab @code{void acc_prof_unregister (acc_event_t, acc_prof_callback, acc_register_t);}
4585 @end multitable
4586
4587 @item @emph{See also}:
4588 @ref{OpenACC Profiling Interface}
4589
4590 @item @emph{Reference}:
4591 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4592 5.3.
4593 @end table
4594
4595
4596
4597 @node acc_prof_lookup
4598 @section @code{acc_prof_lookup} -- Obtain inquiry functions.
4599 @table @asis
4600 @item @emph{Description}:
4601 Function to obtain inquiry functions.
4602
4603 @item @emph{C/C++}:
4604 @multitable @columnfractions .20 .80
4605 @item @emph{Prototype}: @tab @code{acc_query_fn acc_prof_lookup (const char *);}
4606 @end multitable
4607
4608 @item @emph{See also}:
4609 @ref{OpenACC Profiling Interface}
4610
4611 @item @emph{Reference}:
4612 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4613 5.3.
4614 @end table
4615
4616
4617
4618 @node acc_register_library
4619 @section @code{acc_register_library} -- Library registration.
4620 @table @asis
4621 @item @emph{Description}:
4622 Function for library registration.
4623
4624 @item @emph{C/C++}:
4625 @multitable @columnfractions .20 .80
4626 @item @emph{Prototype}: @tab @code{void acc_register_library (acc_prof_reg, acc_prof_reg, acc_prof_lookup_func);}
4627 @end multitable
4628
4629 @item @emph{See also}:
4630 @ref{OpenACC Profiling Interface}, @ref{ACC_PROFLIB}
4631
4632 @item @emph{Reference}:
4633 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4634 5.3.
4635 @end table
4636
4637
4638
4639 @c ---------------------------------------------------------------------
4640 @c OpenACC Environment Variables
4641 @c ---------------------------------------------------------------------
4642
4643 @node OpenACC Environment Variables
4644 @chapter OpenACC Environment Variables
4645
4646 The variables @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM}
4647 are defined by section 4 of the OpenACC specification in version 2.0.
4648 The variable @env{ACC_PROFLIB}
4649 is defined by section 4 of the OpenACC specification in version 2.6.
4650 The variable @env{GCC_ACC_NOTIFY} is used for diagnostic purposes.
4651
4652 @menu
4653 * ACC_DEVICE_TYPE::
4654 * ACC_DEVICE_NUM::
4655 * ACC_PROFLIB::
4656 * GCC_ACC_NOTIFY::
4657 @end menu
4658
4659
4660
4661 @node ACC_DEVICE_TYPE
4662 @section @code{ACC_DEVICE_TYPE}
4663 @table @asis
4664 @item @emph{Reference}:
4665 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4666 4.1.
4667 @end table
4668
4669
4670
4671 @node ACC_DEVICE_NUM
4672 @section @code{ACC_DEVICE_NUM}
4673 @table @asis
4674 @item @emph{Reference}:
4675 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4676 4.2.
4677 @end table
4678
4679
4680
4681 @node ACC_PROFLIB
4682 @section @code{ACC_PROFLIB}
4683 @table @asis
4684 @item @emph{See also}:
4685 @ref{acc_register_library}, @ref{OpenACC Profiling Interface}
4686
4687 @item @emph{Reference}:
4688 @uref{https://www.openacc.org, OpenACC specification v2.6}, section
4689 4.3.
4690 @end table
4691
4692
4693
4694 @node GCC_ACC_NOTIFY
4695 @section @code{GCC_ACC_NOTIFY}
4696 @table @asis
4697 @item @emph{Description}:
4698 Print debug information pertaining to the accelerator.
4699 @end table
4700
4701
4702
4703 @c ---------------------------------------------------------------------
4704 @c CUDA Streams Usage
4705 @c ---------------------------------------------------------------------
4706
4707 @node CUDA Streams Usage
4708 @chapter CUDA Streams Usage
4709
4710 This applies to the @code{nvptx} plugin only.
4711
4712 The library provides elements that perform asynchronous movement of
4713 data and asynchronous operation of computing constructs.  This
4714 asynchronous functionality is implemented by making use of CUDA
4715 streams@footnote{See "Stream Management" in "CUDA Driver API",
4716 TRM-06703-001, Version 5.5, for additional information}.
4717
4718 The primary means by that the asynchronous functionality is accessed
4719 is through the use of those OpenACC directives which make use of the
4720 @code{async} and @code{wait} clauses.  When the @code{async} clause is
4721 first used with a directive, it creates a CUDA stream.  If an
4722 @code{async-argument} is used with the @code{async} clause, then the
4723 stream is associated with the specified @code{async-argument}.
4724
4725 Following the creation of an association between a CUDA stream and the
4726 @code{async-argument} of an @code{async} clause, both the @code{wait}
4727 clause and the @code{wait} directive can be used.  When either the
4728 clause or directive is used after stream creation, it creates a
4729 rendezvous point whereby execution waits until all operations
4730 associated with the @code{async-argument}, that is, stream, have
4731 completed.
4732
4733 Normally, the management of the streams that are created as a result of
4734 using the @code{async} clause, is done without any intervention by the
4735 caller.  This implies the association between the @code{async-argument}
4736 and the CUDA stream will be maintained for the lifetime of the program.
4737 However, this association can be changed through the use of the library
4738 function @code{acc_set_cuda_stream}.  When the function
4739 @code{acc_set_cuda_stream} is called, the CUDA stream that was
4740 originally associated with the @code{async} clause will be destroyed.
4741 Caution should be taken when changing the association as subsequent
4742 references to the @code{async-argument} refer to a different
4743 CUDA stream.
4744
4745
4746
4747 @c ---------------------------------------------------------------------
4748 @c OpenACC Library Interoperability
4749 @c ---------------------------------------------------------------------
4750
4751 @node OpenACC Library Interoperability
4752 @chapter OpenACC Library Interoperability
4753
4754 @section Introduction
4755
4756 The OpenACC library uses the CUDA Driver API, and may interact with
4757 programs that use the Runtime library directly, or another library
4758 based on the Runtime library, e.g., CUBLAS@footnote{See section 2.26,
4759 "Interactions with the CUDA Driver API" in
4760 "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU
4761 Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5,
4762 for additional information on library interoperability.}.
4763 This chapter describes the use cases and what changes are
4764 required in order to use both the OpenACC library and the CUBLAS and Runtime
4765 libraries within a program.
4766
4767 @section First invocation: NVIDIA CUBLAS library API
4768
4769 In this first use case (see below), a function in the CUBLAS library is called
4770 prior to any of the functions in the OpenACC library. More specifically, the
4771 function @code{cublasCreate()}.
4772
4773 When invoked, the function initializes the library and allocates the
4774 hardware resources on the host and the device on behalf of the caller. Once
4775 the initialization and allocation has completed, a handle is returned to the
4776 caller. The OpenACC library also requires initialization and allocation of
4777 hardware resources. Since the CUBLAS library has already allocated the
4778 hardware resources for the device, all that is left to do is to initialize
4779 the OpenACC library and acquire the hardware resources on the host.
4780
4781 Prior to calling the OpenACC function that initializes the library and
4782 allocate the host hardware resources, you need to acquire the device number
4783 that was allocated during the call to @code{cublasCreate()}. The invoking of the
4784 runtime library function @code{cudaGetDevice()} accomplishes this. Once
4785 acquired, the device number is passed along with the device type as
4786 parameters to the OpenACC library function @code{acc_set_device_num()}.
4787
4788 Once the call to @code{acc_set_device_num()} has completed, the OpenACC
4789 library uses the  context that was created during the call to
4790 @code{cublasCreate()}. In other words, both libraries will be sharing the
4791 same context.
4792
4793 @smallexample
4794     /* Create the handle */
4795     s = cublasCreate(&h);
4796     if (s != CUBLAS_STATUS_SUCCESS)
4797     @{
4798         fprintf(stderr, "cublasCreate failed %d\n", s);
4799         exit(EXIT_FAILURE);
4800     @}
4801
4802     /* Get the device number */
4803     e = cudaGetDevice(&dev);
4804     if (e != cudaSuccess)
4805     @{
4806         fprintf(stderr, "cudaGetDevice failed %d\n", e);
4807         exit(EXIT_FAILURE);
4808     @}
4809
4810     /* Initialize OpenACC library and use device 'dev' */
4811     acc_set_device_num(dev, acc_device_nvidia);
4812
4813 @end smallexample
4814 @center Use Case 1
4815
4816 @section First invocation: OpenACC library API
4817
4818 In this second use case (see below), a function in the OpenACC library is
4819 called prior to any of the functions in the CUBLAS library. More specifically,
4820 the function @code{acc_set_device_num()}.
4821
4822 In the use case presented here, the function @code{acc_set_device_num()}
4823 is used to both initialize the OpenACC library and allocate the hardware
4824 resources on the host and the device. In the call to the function, the
4825 call parameters specify which device to use and what device
4826 type to use, i.e., @code{acc_device_nvidia}. It should be noted that this
4827 is but one method to initialize the OpenACC library and allocate the
4828 appropriate hardware resources. Other methods are available through the
4829 use of environment variables and these will be discussed in the next section.
4830
4831 Once the call to @code{acc_set_device_num()} has completed, other OpenACC
4832 functions can be called as seen with multiple calls being made to
4833 @code{acc_copyin()}. In addition, calls can be made to functions in the
4834 CUBLAS library. In the use case a call to @code{cublasCreate()} is made
4835 subsequent to the calls to @code{acc_copyin()}.
4836 As seen in the previous use case, a call to @code{cublasCreate()}
4837 initializes the CUBLAS library and allocates the hardware resources on the
4838 host and the device.  However, since the device has already been allocated,
4839 @code{cublasCreate()} will only initialize the CUBLAS library and allocate
4840 the appropriate hardware resources on the host. The context that was created
4841 as part of the OpenACC initialization is shared with the CUBLAS library,
4842 similarly to the first use case.
4843
4844 @smallexample
4845     dev = 0;
4846
4847     acc_set_device_num(dev, acc_device_nvidia);
4848
4849     /* Copy the first set to the device */
4850     d_X = acc_copyin(&h_X[0], N * sizeof (float));
4851     if (d_X == NULL)
4852     @{
4853         fprintf(stderr, "copyin error h_X\n");
4854         exit(EXIT_FAILURE);
4855     @}
4856
4857     /* Copy the second set to the device */
4858     d_Y = acc_copyin(&h_Y1[0], N * sizeof (float));
4859     if (d_Y == NULL)
4860     @{
4861         fprintf(stderr, "copyin error h_Y1\n");
4862         exit(EXIT_FAILURE);
4863     @}
4864
4865     /* Create the handle */
4866     s = cublasCreate(&h);
4867     if (s != CUBLAS_STATUS_SUCCESS)
4868     @{
4869         fprintf(stderr, "cublasCreate failed %d\n", s);
4870         exit(EXIT_FAILURE);
4871     @}
4872
4873     /* Perform saxpy using CUBLAS library function */
4874     s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1);
4875     if (s != CUBLAS_STATUS_SUCCESS)
4876     @{
4877         fprintf(stderr, "cublasSaxpy failed %d\n", s);
4878         exit(EXIT_FAILURE);
4879     @}
4880
4881     /* Copy the results from the device */
4882     acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float));
4883
4884 @end smallexample
4885 @center Use Case 2
4886
4887 @section OpenACC library and environment variables
4888
4889 There are two environment variables associated with the OpenACC library
4890 that may be used to control the device type and device number:
4891 @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM}, respectively. These two
4892 environment variables can be used as an alternative to calling
4893 @code{acc_set_device_num()}. As seen in the second use case, the device
4894 type and device number were specified using @code{acc_set_device_num()}.
4895 If however, the aforementioned environment variables were set, then the
4896 call to @code{acc_set_device_num()} would not be required.
4897
4898
4899 The use of the environment variables is only relevant when an OpenACC function
4900 is called prior to a call to @code{cudaCreate()}. If @code{cudaCreate()}
4901 is called prior to a call to an OpenACC function, then you must call
4902 @code{acc_set_device_num()}@footnote{More complete information
4903 about @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM} can be found in
4904 sections 4.1 and 4.2 of the @uref{https://www.openacc.org, OpenACC}
4905 Application Programming Interface”, Version 2.6.}
4906
4907
4908
4909 @c ---------------------------------------------------------------------
4910 @c OpenACC Profiling Interface
4911 @c ---------------------------------------------------------------------
4912
4913 @node OpenACC Profiling Interface
4914 @chapter OpenACC Profiling Interface
4915
4916 @section Implementation Status and Implementation-Defined Behavior
4917
4918 We're implementing the OpenACC Profiling Interface as defined by the
4919 OpenACC 2.6 specification.  We're clarifying some aspects here as
4920 @emph{implementation-defined behavior}, while they're still under
4921 discussion within the OpenACC Technical Committee.
4922
4923 This implementation is tuned to keep the performance impact as low as
4924 possible for the (very common) case that the Profiling Interface is
4925 not enabled.  This is relevant, as the Profiling Interface affects all
4926 the @emph{hot} code paths (in the target code, not in the offloaded
4927 code).  Users of the OpenACC Profiling Interface can be expected to
4928 understand that performance will be impacted to some degree once the
4929 Profiling Interface has gotten enabled: for example, because of the
4930 @emph{runtime} (libgomp) calling into a third-party @emph{library} for
4931 every event that has been registered.
4932
4933 We're not yet accounting for the fact that @cite{OpenACC events may
4934 occur during event processing}.
4935 We just handle one case specially, as required by CUDA 9.0
4936 @command{nvprof}, that @code{acc_get_device_type}
4937 (@ref{acc_get_device_type})) may be called from
4938 @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end}
4939 callbacks.
4940
4941 We're not yet implementing initialization via a
4942 @code{acc_register_library} function that is either statically linked
4943 in, or dynamically via @env{LD_PRELOAD}.
4944 Initialization via @code{acc_register_library} functions dynamically
4945 loaded via the @env{ACC_PROFLIB} environment variable does work, as
4946 does directly calling @code{acc_prof_register},
4947 @code{acc_prof_unregister}, @code{acc_prof_lookup}.
4948
4949 As currently there are no inquiry functions defined, calls to
4950 @code{acc_prof_lookup} will always return @code{NULL}.
4951
4952 There aren't separate @emph{start}, @emph{stop} events defined for the
4953 event types @code{acc_ev_create}, @code{acc_ev_delete},
4954 @code{acc_ev_alloc}, @code{acc_ev_free}.  It's not clear if these
4955 should be triggered before or after the actual device-specific call is
4956 made.  We trigger them after.
4957
4958 Remarks about data provided to callbacks:
4959
4960 @table @asis
4961
4962 @item @code{acc_prof_info.event_type}
4963 It's not clear if for @emph{nested} event callbacks (for example,
4964 @code{acc_ev_enqueue_launch_start} as part of a parent compute
4965 construct), this should be set for the nested event
4966 (@code{acc_ev_enqueue_launch_start}), or if the value of the parent
4967 construct should remain (@code{acc_ev_compute_construct_start}).  In
4968 this implementation, the value will generally correspond to the
4969 innermost nested event type.
4970
4971 @item @code{acc_prof_info.device_type}
4972 @itemize
4973
4974 @item
4975 For @code{acc_ev_compute_construct_start}, and in presence of an
4976 @code{if} clause with @emph{false} argument, this will still refer to
4977 the offloading device type.
4978 It's not clear if that's the expected behavior.
4979
4980 @item
4981 Complementary to the item before, for
4982 @code{acc_ev_compute_construct_end}, this is set to
4983 @code{acc_device_host} in presence of an @code{if} clause with
4984 @emph{false} argument.
4985 It's not clear if that's the expected behavior.
4986
4987 @end itemize
4988
4989 @item @code{acc_prof_info.thread_id}
4990 Always @code{-1}; not yet implemented.
4991
4992 @item @code{acc_prof_info.async}
4993 @itemize
4994
4995 @item
4996 Not yet implemented correctly for
4997 @code{acc_ev_compute_construct_start}.
4998
4999 @item
5000 In a compute construct, for host-fallback
5001 execution/@code{acc_device_host} it will always be
5002 @code{acc_async_sync}.
5003 It's not clear if that's the expected behavior.
5004
5005 @item
5006 For @code{acc_ev_device_init_start} and @code{acc_ev_device_init_end},
5007 it will always be @code{acc_async_sync}.
5008 It's not clear if that's the expected behavior.
5009
5010 @end itemize
5011
5012 @item @code{acc_prof_info.async_queue}
5013 There is no @cite{limited number of asynchronous queues} in libgomp.
5014 This will always have the same value as @code{acc_prof_info.async}.
5015
5016 @item @code{acc_prof_info.src_file}
5017 Always @code{NULL}; not yet implemented.
5018
5019 @item @code{acc_prof_info.func_name}
5020 Always @code{NULL}; not yet implemented.
5021
5022 @item @code{acc_prof_info.line_no}
5023 Always @code{-1}; not yet implemented.
5024
5025 @item @code{acc_prof_info.end_line_no}
5026 Always @code{-1}; not yet implemented.
5027
5028 @item @code{acc_prof_info.func_line_no}
5029 Always @code{-1}; not yet implemented.
5030
5031 @item @code{acc_prof_info.func_end_line_no}
5032 Always @code{-1}; not yet implemented.
5033
5034 @item @code{acc_event_info.event_type}, @code{acc_event_info.*.event_type}
5035 Relating to @code{acc_prof_info.event_type} discussed above, in this
5036 implementation, this will always be the same value as
5037 @code{acc_prof_info.event_type}.
5038
5039 @item @code{acc_event_info.*.parent_construct}
5040 @itemize
5041
5042 @item
5043 Will be @code{acc_construct_parallel} for all OpenACC compute
5044 constructs as well as many OpenACC Runtime API calls; should be the
5045 one matching the actual construct, or
5046 @code{acc_construct_runtime_api}, respectively.
5047
5048 @item
5049 Will be @code{acc_construct_enter_data} or
5050 @code{acc_construct_exit_data} when processing variable mappings
5051 specified in OpenACC @emph{declare} directives; should be
5052 @code{acc_construct_declare}.
5053
5054 @item
5055 For implicit @code{acc_ev_device_init_start},
5056 @code{acc_ev_device_init_end}, and explicit as well as implicit
5057 @code{acc_ev_alloc}, @code{acc_ev_free},
5058 @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end},
5059 @code{acc_ev_enqueue_download_start}, and
5060 @code{acc_ev_enqueue_download_end}, will be
5061 @code{acc_construct_parallel}; should reflect the real parent
5062 construct.
5063
5064 @end itemize
5065
5066 @item @code{acc_event_info.*.implicit}
5067 For @code{acc_ev_alloc}, @code{acc_ev_free},
5068 @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end},
5069 @code{acc_ev_enqueue_download_start}, and
5070 @code{acc_ev_enqueue_download_end}, this currently will be @code{1}
5071 also for explicit usage.
5072
5073 @item @code{acc_event_info.data_event.var_name}
5074 Always @code{NULL}; not yet implemented.
5075
5076 @item @code{acc_event_info.data_event.host_ptr}
5077 For @code{acc_ev_alloc}, and @code{acc_ev_free}, this is always
5078 @code{NULL}.
5079
5080 @item @code{typedef union acc_api_info}
5081 @dots{} as printed in @cite{5.2.3. Third Argument: API-Specific
5082 Information}.  This should obviously be @code{typedef @emph{struct}
5083 acc_api_info}.
5084
5085 @item @code{acc_api_info.device_api}
5086 Possibly not yet implemented correctly for
5087 @code{acc_ev_compute_construct_start},
5088 @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end}:
5089 will always be @code{acc_device_api_none} for these event types.
5090 For @code{acc_ev_enter_data_start}, it will be
5091 @code{acc_device_api_none} in some cases.
5092
5093 @item @code{acc_api_info.device_type}
5094 Always the same as @code{acc_prof_info.device_type}.
5095
5096 @item @code{acc_api_info.vendor}
5097 Always @code{-1}; not yet implemented.
5098
5099 @item @code{acc_api_info.device_handle}
5100 Always @code{NULL}; not yet implemented.
5101
5102 @item @code{acc_api_info.context_handle}
5103 Always @code{NULL}; not yet implemented.
5104
5105 @item @code{acc_api_info.async_handle}
5106 Always @code{NULL}; not yet implemented.
5107
5108 @end table
5109
5110 Remarks about certain event types:
5111
5112 @table @asis
5113
5114 @item @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end}
5115 @itemize
5116
5117 @item
5118 @c See 'DEVICE_INIT_INSIDE_COMPUTE_CONSTRUCT' in
5119 @c 'libgomp.oacc-c-c++-common/acc_prof-kernels-1.c',
5120 @c 'libgomp.oacc-c-c++-common/acc_prof-parallel-1.c'.
5121 When a compute construct triggers implicit
5122 @code{acc_ev_device_init_start} and @code{acc_ev_device_init_end}
5123 events, they currently aren't @emph{nested within} the corresponding
5124 @code{acc_ev_compute_construct_start} and
5125 @code{acc_ev_compute_construct_end}, but they're currently observed
5126 @emph{before} @code{acc_ev_compute_construct_start}.
5127 It's not clear what to do: the standard asks us provide a lot of
5128 details to the @code{acc_ev_compute_construct_start} callback, without
5129 (implicitly) initializing a device before?
5130
5131 @item
5132 Callbacks for these event types will not be invoked for calls to the
5133 @code{acc_set_device_type} and @code{acc_set_device_num} functions.
5134 It's not clear if they should be.
5135
5136 @end itemize
5137
5138 @item @code{acc_ev_enter_data_start}, @code{acc_ev_enter_data_end}, @code{acc_ev_exit_data_start}, @code{acc_ev_exit_data_end}
5139 @itemize
5140
5141 @item
5142 Callbacks for these event types will also be invoked for OpenACC
5143 @emph{host_data} constructs.
5144 It's not clear if they should be.
5145
5146 @item
5147 Callbacks for these event types will also be invoked when processing
5148 variable mappings specified in OpenACC @emph{declare} directives.
5149 It's not clear if they should be.
5150
5151 @end itemize
5152
5153 @end table
5154
5155 Callbacks for the following event types will be invoked, but dispatch
5156 and information provided therein has not yet been thoroughly reviewed:
5157
5158 @itemize
5159 @item @code{acc_ev_alloc}
5160 @item @code{acc_ev_free}
5161 @item @code{acc_ev_update_start}, @code{acc_ev_update_end}
5162 @item @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end}
5163 @item @code{acc_ev_enqueue_download_start}, @code{acc_ev_enqueue_download_end}
5164 @end itemize
5165
5166 During device initialization, and finalization, respectively,
5167 callbacks for the following event types will not yet be invoked:
5168
5169 @itemize
5170 @item @code{acc_ev_alloc}
5171 @item @code{acc_ev_free}
5172 @end itemize
5173
5174 Callbacks for the following event types have not yet been implemented,
5175 so currently won't be invoked:
5176
5177 @itemize
5178 @item @code{acc_ev_device_shutdown_start}, @code{acc_ev_device_shutdown_end}
5179 @item @code{acc_ev_runtime_shutdown}
5180 @item @code{acc_ev_create}, @code{acc_ev_delete}
5181 @item @code{acc_ev_wait_start}, @code{acc_ev_wait_end}
5182 @end itemize
5183
5184 For the following runtime library functions, not all expected
5185 callbacks will be invoked (mostly concerning implicit device
5186 initialization):
5187
5188 @itemize
5189 @item @code{acc_get_num_devices}
5190 @item @code{acc_set_device_type}
5191 @item @code{acc_get_device_type}
5192 @item @code{acc_set_device_num}
5193 @item @code{acc_get_device_num}
5194 @item @code{acc_init}
5195 @item @code{acc_shutdown}
5196 @end itemize
5197
5198 Aside from implicit device initialization, for the following runtime
5199 library functions, no callbacks will be invoked for shared-memory
5200 offloading devices (it's not clear if they should be):
5201
5202 @itemize
5203 @item @code{acc_malloc}
5204 @item @code{acc_free}
5205 @item @code{acc_copyin}, @code{acc_present_or_copyin}, @code{acc_copyin_async}
5206 @item @code{acc_create}, @code{acc_present_or_create}, @code{acc_create_async}
5207 @item @code{acc_copyout}, @code{acc_copyout_async}, @code{acc_copyout_finalize}, @code{acc_copyout_finalize_async}
5208 @item @code{acc_delete}, @code{acc_delete_async}, @code{acc_delete_finalize}, @code{acc_delete_finalize_async}
5209 @item @code{acc_update_device}, @code{acc_update_device_async}
5210 @item @code{acc_update_self}, @code{acc_update_self_async}
5211 @item @code{acc_map_data}, @code{acc_unmap_data}
5212 @item @code{acc_memcpy_to_device}, @code{acc_memcpy_to_device_async}
5213 @item @code{acc_memcpy_from_device}, @code{acc_memcpy_from_device_async}
5214 @end itemize
5215
5216 @c ---------------------------------------------------------------------
5217 @c OpenMP-Implementation Specifics
5218 @c ---------------------------------------------------------------------
5219
5220 @node OpenMP-Implementation Specifics
5221 @chapter OpenMP-Implementation Specifics
5222
5223 @menu
5224 * Implementation-defined ICV Initialization::
5225 * OpenMP Context Selectors::
5226 * Memory allocation::
5227 @end menu
5228
5229 @node Implementation-defined ICV Initialization
5230 @section Implementation-defined ICV Initialization
5231 @cindex Implementation specific setting
5232
5233 @multitable @columnfractions .30 .70
5234 @item @var{affinity-format-var} @tab See @ref{OMP_AFFINITY_FORMAT}.
5235 @item @var{def-allocator-var} @tab See @ref{OMP_ALLOCATOR}.
5236 @item @var{max-active-levels-var} @tab See @ref{OMP_MAX_ACTIVE_LEVELS}.
5237 @item @var{dyn-var} @tab See @ref{OMP_DYNAMIC}.
5238 @item @var{nthreads-var} @tab See @ref{OMP_NUM_THREADS}.
5239 @item @var{num-devices-var} @tab Number of non-host devices found
5240 by GCC's run-time library
5241 @item @var{num-procs-var} @tab The number of CPU cores on the
5242 initial device, except that affinity settings might lead to a
5243 smaller number.  On non-host devices, the value of the
5244 @var{nthreads-var} ICV.
5245 @item @var{place-partition-var} @tab See @ref{OMP_PLACES}.
5246 @item @var{run-sched-var} @tab See @ref{OMP_SCHEDULE}.
5247 @item @var{stacksize-var} @tab See @ref{OMP_STACKSIZE}.
5248 @item @var{thread-limit-var} @tab See @ref{OMP_TEAMS_THREAD_LIMIT}
5249 @item @var{wait-policy-var} @tab See @ref{OMP_WAIT_POLICY} and
5250 @ref{GOMP_SPINCOUNT}
5251 @end multitable
5252
5253 @node OpenMP Context Selectors
5254 @section OpenMP Context Selectors
5255
5256 @code{vendor} is always @code{gnu}. References are to the GCC manual.
5257
5258 @c NOTE: Only the following selectors have been implemented. To add
5259 @c additional traits for target architecture, TARGET_OMP_DEVICE_KIND_ARCH_ISA
5260 @c has to be implemented; cf. also PR target/105640.
5261 @c For offload devices, add *additionally* gcc/config/*/t-omp-device.
5262
5263 For the host compiler, @code{kind} always matches @code{host}; for the
5264 offloading architectures AMD GCN and Nvidia PTX, @code{kind} always matches
5265 @code{gpu}.  For the x86 family of computers, AMD GCN and Nvidia PTX
5266 the following traits are supported in addition; while OpenMP is supported
5267 on more architectures, GCC currently does not match any @code{arch} or
5268 @code{isa} traits for those.
5269
5270 @multitable @columnfractions .65 .30
5271 @headitem @code{arch} @tab @code{isa}
5272 @item @code{x86}, @code{x86_64}, @code{i386}, @code{i486},
5273       @code{i586}, @code{i686}, @code{ia32}
5274       @tab See @code{-m...} flags in ``x86 Options'' (without @code{-m})
5275 @item @code{amdgcn}, @code{gcn}
5276       @tab See @code{-march=} in ``AMD GCN Options''@footnote{Additionally,
5277       @code{gfx803} is supported as an alias for @code{fiji}.}
5278 @item @code{nvptx}
5279       @tab See @code{-march=} in ``Nvidia PTX Options''
5280 @end multitable
5281
5282 @node Memory allocation
5283 @section Memory allocation
5284
5285 For the available predefined allocators and, as applicable, their associated
5286 predefined memory spaces and for the available traits and their default values,
5287 see @ref{OMP_ALLOCATOR}.  Predefined allocators without an associated memory
5288 space use the @code{omp_default_mem_space} memory space.
5289
5290 For the memory spaces, the following applies:
5291 @itemize
5292 @item @code{omp_default_mem_space} is supported
5293 @item @code{omp_const_mem_space} maps to @code{omp_default_mem_space}
5294 @item @code{omp_low_lat_mem_space} maps to @code{omp_default_mem_space}
5295 @item @code{omp_large_cap_mem_space} maps to @code{omp_default_mem_space},
5296       unless the memkind library is available
5297 @item @code{omp_high_bw_mem_space} maps to @code{omp_default_mem_space},
5298       unless the memkind library is available
5299 @end itemize
5300
5301 On Linux systems, where the @uref{https://github.com/memkind/memkind, memkind
5302 library} (@code{libmemkind.so.0}) is available at runtime, it is used when
5303 creating memory allocators requesting
5304
5305 @itemize
5306 @item the memory space @code{omp_high_bw_mem_space}
5307 @item the memory space @code{omp_large_cap_mem_space}
5308 @item the @code{partition} trait @code{interleaved}; note that for
5309       @code{omp_large_cap_mem_space} the allocation will not be interleaved
5310 @end itemize
5311
5312 On Linux systems, where the @uref{https://github.com/numactl/numactl, numa
5313 library} (@code{libnuma.so.1}) is available at runtime, it used when creating
5314 memory allocators requesting
5315
5316 @itemize
5317 @item the @code{partition} trait @code{nearest}, except when both the
5318 libmemkind library is available and the memory space is either
5319 @code{omp_large_cap_mem_space} or @code{omp_high_bw_mem_space}
5320 @end itemize
5321
5322 Note that the numa library will round up the allocation size to a multiple of
5323 the system page size; therefore, consider using it only with large data or
5324 by sharing allocations via the @code{pool_size} trait.  Furthermore, the Linux
5325 kernel does not guarantee that an allocation will always be on the nearest NUMA
5326 node nor that after reallocation the same node will be used.  Note additionally
5327 that, on Linux, the default setting of the memory placement policy is to use the
5328 current node; therefore, unless the memory placement policy has been overridden,
5329 the @code{partition} trait @code{environment} (the default) will be effectively
5330 a @code{nearest} allocation.
5331
5332 Additional notes regarding the traits:
5333 @itemize
5334 @item The @code{pinned} trait is unsupported.
5335 @item The default for the @code{pool_size} trait is no pool and for every
5336       (re)allocation the associated library routine is called, which might
5337       internally use a memory pool.
5338 @item For the @code{partition} trait, the partition part size will be the same
5339       as the requested size (i.e. @code{interleaved} or @code{blocked} has no
5340       effect), except for @code{interleaved} when the memkind library is
5341       available.  Furthermore, for @code{nearest} and unless the numa library
5342       is available, the memory might not be on the same NUMA node as thread
5343       that allocated the memory; on Linux, this is in particular the case when
5344       the memory placement policy is set to preferred.
5345 @item The @code{access} trait has no effect such that memory is always
5346       accessible by all threads.
5347 @item The @code{sync_hint} trait has no effect.
5348 @end itemize
5349
5350 @c ---------------------------------------------------------------------
5351 @c Offload-Target Specifics
5352 @c ---------------------------------------------------------------------
5353
5354 @node Offload-Target Specifics
5355 @chapter Offload-Target Specifics
5356
5357 The following sections present notes on the offload-target specifics
5358
5359 @menu
5360 * AMD Radeon::
5361 * nvptx::
5362 @end menu
5363
5364 @node AMD Radeon
5365 @section AMD Radeon (GCN)
5366
5367 On the hardware side, there is the hierarchy (fine to coarse):
5368 @itemize
5369 @item work item (thread)
5370 @item wavefront
5371 @item work group
5372 @item compute unit (CU)
5373 @end itemize
5374
5375 All OpenMP and OpenACC levels are used, i.e.
5376 @itemize
5377 @item OpenMP's simd and OpenACC's vector map to work items (thread)
5378 @item OpenMP's threads (``parallel'') and OpenACC's workers map
5379       to wavefronts
5380 @item OpenMP's teams and OpenACC's gang use a threadpool with the
5381       size of the number of teams or gangs, respectively.
5382 @end itemize
5383
5384 The used sizes are
5385 @itemize
5386 @item Number of teams is the specified @code{num_teams} (OpenMP) or
5387       @code{num_gangs} (OpenACC) or otherwise the number of CU. It is limited
5388       by two times the number of CU.
5389 @item Number of wavefronts is 4 for gfx900 and 16 otherwise;
5390       @code{num_threads} (OpenMP) and @code{num_workers} (OpenACC)
5391       overrides this if smaller.
5392 @item The wavefront has 102 scalars and 64 vectors
5393 @item Number of workitems is always 64
5394 @item The hardware permits maximally 40 workgroups/CU and
5395       16 wavefronts/workgroup up to a limit of 40 wavefronts in total per CU.
5396 @item 80 scalars registers and 24 vector registers in non-kernel functions
5397       (the chosen procedure-calling API).
5398 @item For the kernel itself: as many as register pressure demands (number of
5399       teams and number of threads, scaled down if registers are exhausted)
5400 @end itemize
5401
5402 The implementation remark:
5403 @itemize
5404 @item I/O within OpenMP target regions and OpenACC parallel/kernels is supported
5405       using the C library @code{printf} functions and the Fortran
5406       @code{print}/@code{write} statements.
5407 @item Reverse offload regions (i.e. @code{target} regions with
5408       @code{device(ancestor:1)}) are processed serially per @code{target} region
5409       such that the next reverse offload region is only executed after the previous
5410       one returned.
5411 @item OpenMP code that has a @code{requires} directive with
5412       @code{unified_shared_memory} will remove any GCN device from the list of
5413       available devices (``host fallback'').
5414 @item The available stack size can be changed using the @code{GCN_STACK_SIZE}
5415       environment variable; the default is 32 kiB per thread.
5416 @end itemize
5417
5418
5419
5420 @node nvptx
5421 @section nvptx
5422
5423 On the hardware side, there is the hierarchy (fine to coarse):
5424 @itemize
5425 @item thread
5426 @item warp
5427 @item thread block
5428 @item streaming multiprocessor
5429 @end itemize
5430
5431 All OpenMP and OpenACC levels are used, i.e.
5432 @itemize
5433 @item OpenMP's simd and OpenACC's vector map to threads
5434 @item OpenMP's threads (``parallel'') and OpenACC's workers map to warps
5435 @item OpenMP's teams and OpenACC's gang use a threadpool with the
5436       size of the number of teams or gangs, respectively.
5437 @end itemize
5438
5439 The used sizes are
5440 @itemize
5441 @item The @code{warp_size} is always 32
5442 @item CUDA kernel launched: @code{dim=@{#teams,1,1@}, blocks=@{#threads,warp_size,1@}}.
5443 @item The number of teams is limited by the number of blocks the device can
5444       host simultaneously.
5445 @end itemize
5446
5447 Additional information can be obtained by setting the environment variable to
5448 @code{GOMP_DEBUG=1} (very verbose; grep for @code{kernel.*launch} for launch
5449 parameters).
5450
5451 GCC generates generic PTX ISA code, which is just-in-time compiled by CUDA,
5452 which caches the JIT in the user's directory (see CUDA documentation; can be
5453 tuned by the environment variables @code{CUDA_CACHE_@{DISABLE,MAXSIZE,PATH@}}.
5454
5455 Note: While PTX ISA is generic, the @code{-mptx=} and @code{-march=} commandline
5456 options still affect the used PTX ISA code and, thus, the requirements on
5457 CUDA version and hardware.
5458
5459 The implementation remark:
5460 @itemize
5461 @item I/O within OpenMP target regions and OpenACC parallel/kernels is supported
5462       using the C library @code{printf} functions. Note that the Fortran
5463       @code{print}/@code{write} statements are not supported, yet.
5464 @item Compilation OpenMP code that contains @code{requires reverse_offload}
5465       requires at least @code{-march=sm_35}, compiling for @code{-march=sm_30}
5466       is not supported.
5467 @item For code containing reverse offload (i.e. @code{target} regions with
5468       @code{device(ancestor:1)}), there is a slight performance penalty
5469       for @emph{all} target regions, consisting mostly of shutdown delay
5470       Per device, reverse offload regions are processed serially such that
5471       the next reverse offload region is only executed after the previous
5472       one returned.
5473 @item OpenMP code that has a @code{requires} directive with
5474       @code{unified_shared_memory} will remove any nvptx device from the
5475       list of available devices (``host fallback'').
5476 @item The default per-warp stack size is 128 kiB; see also @code{-msoft-stack}
5477       in the GCC manual.
5478 @item The OpenMP routines @code{omp_target_memcpy_rect} and
5479       @code{omp_target_memcpy_rect_async} and the @code{target update}
5480       directive for non-contiguous list items will use the 2D and 3D
5481       memory-copy functions of the CUDA library.  Higher dimensions will
5482       call those functions in a loop and are therefore supported.
5483 @end itemize
5484
5485
5486 @c ---------------------------------------------------------------------
5487 @c The libgomp ABI
5488 @c ---------------------------------------------------------------------
5489
5490 @node The libgomp ABI
5491 @chapter The libgomp ABI
5492
5493 The following sections present notes on the external ABI as
5494 presented by libgomp.  Only maintainers should need them.
5495
5496 @menu
5497 * Implementing MASTER construct::
5498 * Implementing CRITICAL construct::
5499 * Implementing ATOMIC construct::
5500 * Implementing FLUSH construct::
5501 * Implementing BARRIER construct::
5502 * Implementing THREADPRIVATE construct::
5503 * Implementing PRIVATE clause::
5504 * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses::
5505 * Implementing REDUCTION clause::
5506 * Implementing PARALLEL construct::
5507 * Implementing FOR construct::
5508 * Implementing ORDERED construct::
5509 * Implementing SECTIONS construct::
5510 * Implementing SINGLE construct::
5511 * Implementing OpenACC's PARALLEL construct::
5512 @end menu
5513
5514
5515 @node Implementing MASTER construct
5516 @section Implementing MASTER construct
5517
5518 @smallexample
5519 if (omp_get_thread_num () == 0)
5520   block
5521 @end smallexample
5522
5523 Alternately, we generate two copies of the parallel subfunction
5524 and only include this in the version run by the primary thread.
5525 Surely this is not worthwhile though...
5526
5527
5528
5529 @node Implementing CRITICAL construct
5530 @section Implementing CRITICAL construct
5531
5532 Without a specified name,
5533
5534 @smallexample
5535   void GOMP_critical_start (void);
5536   void GOMP_critical_end (void);
5537 @end smallexample
5538
5539 so that we don't get COPY relocations from libgomp to the main
5540 application.
5541
5542 With a specified name, use omp_set_lock and omp_unset_lock with
5543 name being transformed into a variable declared like
5544
5545 @smallexample
5546   omp_lock_t gomp_critical_user_<name> __attribute__((common))
5547 @end smallexample
5548
5549 Ideally the ABI would specify that all zero is a valid unlocked
5550 state, and so we wouldn't need to initialize this at
5551 startup.
5552
5553
5554
5555 @node Implementing ATOMIC construct
5556 @section Implementing ATOMIC construct
5557
5558 The target should implement the @code{__sync} builtins.
5559
5560 Failing that we could add
5561
5562 @smallexample
5563   void GOMP_atomic_enter (void)
5564   void GOMP_atomic_exit (void)
5565 @end smallexample
5566
5567 which reuses the regular lock code, but with yet another lock
5568 object private to the library.
5569
5570
5571
5572 @node Implementing FLUSH construct
5573 @section Implementing FLUSH construct
5574
5575 Expands to the @code{__sync_synchronize} builtin.
5576
5577
5578
5579 @node Implementing BARRIER construct
5580 @section Implementing BARRIER construct
5581
5582 @smallexample
5583   void GOMP_barrier (void)
5584 @end smallexample
5585
5586
5587 @node Implementing THREADPRIVATE construct
5588 @section Implementing THREADPRIVATE construct
5589
5590 In _most_ cases we can map this directly to @code{__thread}.  Except
5591 that OMP allows constructors for C++ objects.  We can either
5592 refuse to support this (how often is it used?) or we can
5593 implement something akin to .ctors.
5594
5595 Even more ideally, this ctor feature is handled by extensions
5596 to the main pthreads library.  Failing that, we can have a set
5597 of entry points to register ctor functions to be called.
5598
5599
5600
5601 @node Implementing PRIVATE clause
5602 @section Implementing PRIVATE clause
5603
5604 In association with a PARALLEL, or within the lexical extent
5605 of a PARALLEL block, the variable becomes a local variable in
5606 the parallel subfunction.
5607
5608 In association with FOR or SECTIONS blocks, create a new
5609 automatic variable within the current function.  This preserves
5610 the semantic of new variable creation.
5611
5612
5613
5614 @node Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses
5615 @section Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses
5616
5617 This seems simple enough for PARALLEL blocks.  Create a private
5618 struct for communicating between the parent and subfunction.
5619 In the parent, copy in values for scalar and "small" structs;
5620 copy in addresses for others TREE_ADDRESSABLE types.  In the
5621 subfunction, copy the value into the local variable.
5622
5623 It is not clear what to do with bare FOR or SECTION blocks.
5624 The only thing I can figure is that we do something like:
5625
5626 @smallexample
5627 #pragma omp for firstprivate(x) lastprivate(y)
5628 for (int i = 0; i < n; ++i)
5629   body;
5630 @end smallexample
5631
5632 which becomes
5633
5634 @smallexample
5635 @{
5636   int x = x, y;
5637
5638   // for stuff
5639
5640   if (i == n)
5641     y = y;
5642 @}
5643 @end smallexample
5644
5645 where the "x=x" and "y=y" assignments actually have different
5646 uids for the two variables, i.e. not something you could write
5647 directly in C.  Presumably this only makes sense if the "outer"
5648 x and y are global variables.
5649
5650 COPYPRIVATE would work the same way, except the structure
5651 broadcast would have to happen via SINGLE machinery instead.
5652
5653
5654
5655 @node Implementing REDUCTION clause
5656 @section Implementing REDUCTION clause
5657
5658 The private struct mentioned in the previous section should have
5659 a pointer to an array of the type of the variable, indexed by the
5660 thread's @var{team_id}.  The thread stores its final value into the
5661 array, and after the barrier, the primary thread iterates over the
5662 array to collect the values.
5663
5664
5665 @node Implementing PARALLEL construct
5666 @section Implementing PARALLEL construct
5667
5668 @smallexample
5669   #pragma omp parallel
5670   @{
5671     body;
5672   @}
5673 @end smallexample
5674
5675 becomes
5676
5677 @smallexample
5678   void subfunction (void *data)
5679   @{
5680     use data;
5681     body;
5682   @}
5683
5684   setup data;
5685   GOMP_parallel_start (subfunction, &data, num_threads);
5686   subfunction (&data);
5687   GOMP_parallel_end ();
5688 @end smallexample
5689
5690 @smallexample
5691   void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads)
5692 @end smallexample
5693
5694 The @var{FN} argument is the subfunction to be run in parallel.
5695
5696 The @var{DATA} argument is a pointer to a structure used to
5697 communicate data in and out of the subfunction, as discussed
5698 above with respect to FIRSTPRIVATE et al.
5699
5700 The @var{NUM_THREADS} argument is 1 if an IF clause is present
5701 and false, or the value of the NUM_THREADS clause, if
5702 present, or 0.
5703
5704 The function needs to create the appropriate number of
5705 threads and/or launch them from the dock.  It needs to
5706 create the team structure and assign team ids.
5707
5708 @smallexample
5709   void GOMP_parallel_end (void)
5710 @end smallexample
5711
5712 Tears down the team and returns us to the previous @code{omp_in_parallel()} state.
5713
5714
5715
5716 @node Implementing FOR construct
5717 @section Implementing FOR construct
5718
5719 @smallexample
5720   #pragma omp parallel for
5721   for (i = lb; i <= ub; i++)
5722     body;
5723 @end smallexample
5724
5725 becomes
5726
5727 @smallexample
5728   void subfunction (void *data)
5729   @{
5730     long _s0, _e0;
5731     while (GOMP_loop_static_next (&_s0, &_e0))
5732     @{
5733       long _e1 = _e0, i;
5734       for (i = _s0; i < _e1; i++)
5735         body;
5736     @}
5737     GOMP_loop_end_nowait ();
5738   @}
5739
5740   GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0);
5741   subfunction (NULL);
5742   GOMP_parallel_end ();
5743 @end smallexample
5744
5745 @smallexample
5746   #pragma omp for schedule(runtime)
5747   for (i = 0; i < n; i++)
5748     body;
5749 @end smallexample
5750
5751 becomes
5752
5753 @smallexample
5754   @{
5755     long i, _s0, _e0;
5756     if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0))
5757       do @{
5758         long _e1 = _e0;
5759         for (i = _s0, i < _e0; i++)
5760           body;
5761       @} while (GOMP_loop_runtime_next (&_s0, _&e0));
5762     GOMP_loop_end ();
5763   @}
5764 @end smallexample
5765
5766 Note that while it looks like there is trickiness to propagating
5767 a non-constant STEP, there isn't really.  We're explicitly allowed
5768 to evaluate it as many times as we want, and any variables involved
5769 should automatically be handled as PRIVATE or SHARED like any other
5770 variables.  So the expression should remain evaluable in the
5771 subfunction.  We can also pull it into a local variable if we like,
5772 but since its supposed to remain unchanged, we can also not if we like.
5773
5774 If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be
5775 able to get away with no work-sharing context at all, since we can
5776 simply perform the arithmetic directly in each thread to divide up
5777 the iterations.  Which would mean that we wouldn't need to call any
5778 of these routines.
5779
5780 There are separate routines for handling loops with an ORDERED
5781 clause.  Bookkeeping for that is non-trivial...
5782
5783
5784
5785 @node Implementing ORDERED construct
5786 @section Implementing ORDERED construct
5787
5788 @smallexample
5789   void GOMP_ordered_start (void)
5790   void GOMP_ordered_end (void)
5791 @end smallexample
5792
5793
5794
5795 @node Implementing SECTIONS construct
5796 @section Implementing SECTIONS construct
5797
5798 A block as
5799
5800 @smallexample
5801   #pragma omp sections
5802   @{
5803     #pragma omp section
5804     stmt1;
5805     #pragma omp section
5806     stmt2;
5807     #pragma omp section
5808     stmt3;
5809   @}
5810 @end smallexample
5811
5812 becomes
5813
5814 @smallexample
5815   for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ())
5816     switch (i)
5817       @{
5818       case 1:
5819         stmt1;
5820         break;
5821       case 2:
5822         stmt2;
5823         break;
5824       case 3:
5825         stmt3;
5826         break;
5827       @}
5828   GOMP_barrier ();
5829 @end smallexample
5830
5831
5832 @node Implementing SINGLE construct
5833 @section Implementing SINGLE construct
5834
5835 A block like
5836
5837 @smallexample
5838   #pragma omp single
5839   @{
5840     body;
5841   @}
5842 @end smallexample
5843
5844 becomes
5845
5846 @smallexample
5847   if (GOMP_single_start ())
5848     body;
5849   GOMP_barrier ();
5850 @end smallexample
5851
5852 while
5853
5854 @smallexample
5855   #pragma omp single copyprivate(x)
5856     body;
5857 @end smallexample
5858
5859 becomes
5860
5861 @smallexample
5862   datap = GOMP_single_copy_start ();
5863   if (datap == NULL)
5864     @{
5865       body;
5866       data.x = x;
5867       GOMP_single_copy_end (&data);
5868     @}
5869   else
5870     x = datap->x;
5871   GOMP_barrier ();
5872 @end smallexample
5873
5874
5875
5876 @node Implementing OpenACC's PARALLEL construct
5877 @section Implementing OpenACC's PARALLEL construct
5878
5879 @smallexample
5880   void GOACC_parallel ()
5881 @end smallexample
5882
5883
5884
5885 @c ---------------------------------------------------------------------
5886 @c Reporting Bugs
5887 @c ---------------------------------------------------------------------
5888
5889 @node Reporting Bugs
5890 @chapter Reporting Bugs
5891
5892 Bugs in the GNU Offloading and Multi Processing Runtime Library should
5893 be reported via @uref{https://gcc.gnu.org/bugzilla/, Bugzilla}.  Please add
5894 "openacc", or "openmp", or both to the keywords field in the bug
5895 report, as appropriate.
5896
5897
5898
5899 @c ---------------------------------------------------------------------
5900 @c GNU General Public License
5901 @c ---------------------------------------------------------------------
5902
5903 @include gpl_v3.texi
5904
5905
5906
5907 @c ---------------------------------------------------------------------
5908 @c GNU Free Documentation License
5909 @c ---------------------------------------------------------------------
5910
5911 @include fdl.texi
5912
5913
5914
5915 @c ---------------------------------------------------------------------
5916 @c Funding Free Software
5917 @c ---------------------------------------------------------------------
5918
5919 @include funding.texi
5920
5921 @c ---------------------------------------------------------------------
5922 @c Index
5923 @c ---------------------------------------------------------------------
5924
5925 @node Library Index
5926 @unnumbered Library Index
5927
5928 @printindex cp
5929
5930 @bye