]>
Commit | Line | Data |
---|---|---|
3721b9e1 DF |
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 | |
7e7065b9 | 10 | Copyright @copyright{} 2006-2020 Free Software Foundation, Inc. |
3721b9e1 DF |
11 | |
12 | Permission is granted to copy, distribute and/or modify this document | |
07a67d6a | 13 | under the terms of the GNU Free Documentation License, Version 1.3 or |
3721b9e1 | 14 | any later version published by the Free Software Foundation; with the |
70b1e376 | 15 | Invariant Sections being ``Funding Free Software'', the Front-Cover |
3721b9e1 DF |
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 | |
f1f3453e | 34 | * libgomp: (libgomp). GNU Offloading and Multi Processing Runtime Library. |
3721b9e1 DF |
35 | @end direntry |
36 | ||
f1f3453e | 37 | This manual documents libgomp, the GNU Offloading and Multi Processing |
41dbbb37 TS |
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. | |
3721b9e1 DF |
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 | |
f1f3453e | 53 | @title GNU Offloading and Multi Processing Runtime Library |
41dbbb37 | 54 | @subtitle The GNU OpenMP and OpenACC Implementation |
3721b9e1 DF |
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 | |
72 | @top Introduction | |
73 | @cindex Introduction | |
74 | ||
f1f3453e | 75 | This manual documents the usage of libgomp, the GNU Offloading and |
41dbbb37 | 76 | Multi Processing Runtime Library. This includes the GNU |
1a6d1d24 | 77 | implementation of the @uref{https://www.openmp.org, OpenMP} Application |
41dbbb37 TS |
78 | Programming Interface (API) for multi-platform shared-memory parallel |
79 | programming in C/C++ and Fortran, and the GNU implementation of the | |
9651fbaf | 80 | @uref{https://www.openacc.org, OpenACC} Application Programming |
41dbbb37 TS |
81 | Interface (API) for offloading of code to accelerator devices in C/C++ |
82 | and Fortran. | |
3721b9e1 | 83 | |
41dbbb37 TS |
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. | |
f1f3453e | 88 | |
3721b9e1 DF |
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. | |
4102bda6 TS |
98 | * OpenMP Runtime Library Routines: Runtime Library Routines. |
99 | The OpenMP runtime application programming | |
3721b9e1 | 100 | interface. |
4102bda6 TS |
101 | * OpenMP Environment Variables: Environment Variables. |
102 | Influencing OpenMP runtime behavior with | |
103 | environment variables. | |
cdf6119d JN |
104 | * Enabling OpenACC:: How to enable OpenACC for your |
105 | applications. | |
106 | * OpenACC Runtime Library Routines:: The OpenACC runtime application | |
107 | programming interface. | |
108 | * OpenACC Environment Variables:: Influencing OpenACC runtime behavior with | |
109 | environment variables. | |
110 | * CUDA Streams Usage:: Notes on the implementation of | |
111 | asynchronous operations. | |
112 | * OpenACC Library Interoperability:: OpenACC library interoperability with the | |
113 | NVIDIA CUBLAS library. | |
5fae049d | 114 | * OpenACC Profiling Interface:: |
3721b9e1 | 115 | * The libgomp ABI:: Notes on the external ABI presented by libgomp. |
f1f3453e TS |
116 | * Reporting Bugs:: How to report bugs in the GNU Offloading and |
117 | Multi Processing Runtime Library. | |
3721b9e1 DF |
118 | * Copying:: GNU general public license says |
119 | how you can copy and share libgomp. | |
120 | * GNU Free Documentation License:: | |
121 | How you can copy and share this manual. | |
122 | * Funding:: How to help assure continued work for free | |
123 | software. | |
3d3949df | 124 | * Library Index:: Index of this documentation. |
3721b9e1 DF |
125 | @end menu |
126 | ||
127 | ||
128 | @c --------------------------------------------------------------------- | |
129 | @c Enabling OpenMP | |
130 | @c --------------------------------------------------------------------- | |
131 | ||
132 | @node Enabling OpenMP | |
133 | @chapter Enabling OpenMP | |
134 | ||
135 | To activate the OpenMP extensions for C/C++ and Fortran, the compile-time | |
83fd6c5b | 136 | flag @command{-fopenmp} must be specified. This enables the OpenMP directive |
3721b9e1 DF |
137 | @code{#pragma omp} in C/C++ and @code{!$omp} directives in free form, |
138 | @code{c$omp}, @code{*$omp} and @code{!$omp} directives in fixed form, | |
139 | @code{!$} conditional compilation sentinels in free form and @code{c$}, | |
83fd6c5b | 140 | @code{*$} and @code{!$} sentinels in fixed form, for Fortran. The flag also |
3721b9e1 DF |
141 | arranges for automatic linking of the OpenMP runtime library |
142 | (@ref{Runtime Library Routines}). | |
143 | ||
144 | A complete description of all OpenMP directives accepted may be found in | |
1a6d1d24 | 145 | the @uref{https://www.openmp.org, OpenMP Application Program Interface} manual, |
00b9bd52 | 146 | version 4.5. |
3721b9e1 DF |
147 | |
148 | ||
149 | @c --------------------------------------------------------------------- | |
4102bda6 | 150 | @c OpenMP Runtime Library Routines |
3721b9e1 DF |
151 | @c --------------------------------------------------------------------- |
152 | ||
153 | @node Runtime Library Routines | |
4102bda6 | 154 | @chapter OpenMP Runtime Library Routines |
3721b9e1 | 155 | |
83fd6c5b | 156 | The runtime routines described here are defined by Section 3 of the OpenMP |
00b9bd52 | 157 | specification in version 4.5. The routines are structured in following |
5c6ed53a | 158 | three parts: |
3721b9e1 | 159 | |
72832460 | 160 | @menu |
83fd6c5b TB |
161 | Control threads, processors and the parallel environment. They have C |
162 | linkage, and do not throw exceptions. | |
f5745bed | 163 | |
5c6ed53a TB |
164 | * omp_get_active_level:: Number of active parallel regions |
165 | * omp_get_ancestor_thread_num:: Ancestor thread ID | |
83fd6c5b TB |
166 | * omp_get_cancellation:: Whether cancellation support is enabled |
167 | * omp_get_default_device:: Get the default device for target regions | |
5c6ed53a TB |
168 | * omp_get_dynamic:: Dynamic teams setting |
169 | * omp_get_level:: Number of parallel regions | |
6a2ba183 | 170 | * omp_get_max_active_levels:: Maximum number of active regions |
d9a6bd32 | 171 | * omp_get_max_task_priority:: Maximum task priority value that can be set |
6a2ba183 | 172 | * omp_get_max_threads:: Maximum number of threads of parallel region |
5c6ed53a | 173 | * omp_get_nested:: Nested parallel regions |
83fd6c5b | 174 | * omp_get_num_devices:: Number of target devices |
5c6ed53a | 175 | * omp_get_num_procs:: Number of processors online |
83fd6c5b | 176 | * omp_get_num_teams:: Number of teams |
5c6ed53a | 177 | * omp_get_num_threads:: Size of the active team |
83fd6c5b | 178 | * omp_get_proc_bind:: Whether theads may be moved between CPUs |
5c6ed53a | 179 | * omp_get_schedule:: Obtain the runtime scheduling method |
83fd6c5b | 180 | * omp_get_team_num:: Get team number |
5c6ed53a | 181 | * omp_get_team_size:: Number of threads in a team |
6a2ba183 | 182 | * omp_get_thread_limit:: Maximum number of threads |
5c6ed53a TB |
183 | * omp_get_thread_num:: Current thread ID |
184 | * omp_in_parallel:: Whether a parallel region is active | |
20906c66 | 185 | * omp_in_final:: Whether in final or included task region |
83fd6c5b TB |
186 | * omp_is_initial_device:: Whether executing on the host device |
187 | * omp_set_default_device:: Set the default device for target regions | |
5c6ed53a TB |
188 | * omp_set_dynamic:: Enable/disable dynamic teams |
189 | * omp_set_max_active_levels:: Limits the number of active parallel regions | |
190 | * omp_set_nested:: Enable/disable nested parallel regions | |
191 | * omp_set_num_threads:: Set upper team size limit | |
192 | * omp_set_schedule:: Set the runtime scheduling method | |
3721b9e1 DF |
193 | |
194 | Initialize, set, test, unset and destroy simple and nested locks. | |
195 | ||
3721b9e1 DF |
196 | * omp_init_lock:: Initialize simple lock |
197 | * omp_set_lock:: Wait for and set simple lock | |
198 | * omp_test_lock:: Test and set simple lock if available | |
199 | * omp_unset_lock:: Unset simple lock | |
200 | * omp_destroy_lock:: Destroy simple lock | |
201 | * omp_init_nest_lock:: Initialize nested lock | |
202 | * omp_set_nest_lock:: Wait for and set simple lock | |
203 | * omp_test_nest_lock:: Test and set nested lock if available | |
204 | * omp_unset_nest_lock:: Unset nested lock | |
205 | * omp_destroy_nest_lock:: Destroy nested lock | |
3721b9e1 DF |
206 | |
207 | Portable, thread-based, wall clock timer. | |
208 | ||
3721b9e1 DF |
209 | * omp_get_wtick:: Get timer precision. |
210 | * omp_get_wtime:: Elapsed wall clock time. | |
211 | @end menu | |
212 | ||
5c6ed53a TB |
213 | |
214 | ||
215 | @node omp_get_active_level | |
216 | @section @code{omp_get_active_level} -- Number of parallel regions | |
217 | @table @asis | |
218 | @item @emph{Description}: | |
219 | This function returns the nesting level for the active parallel blocks, | |
220 | which enclose the calling call. | |
221 | ||
222 | @item @emph{C/C++} | |
223 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 224 | @item @emph{Prototype}: @tab @code{int omp_get_active_level(void);} |
5c6ed53a TB |
225 | @end multitable |
226 | ||
227 | @item @emph{Fortran}: | |
228 | @multitable @columnfractions .20 .80 | |
acb5c916 | 229 | @item @emph{Interface}: @tab @code{integer function omp_get_active_level()} |
5c6ed53a TB |
230 | @end multitable |
231 | ||
232 | @item @emph{See also}: | |
233 | @ref{omp_get_level}, @ref{omp_get_max_active_levels}, @ref{omp_set_max_active_levels} | |
234 | ||
235 | @item @emph{Reference}: | |
1a6d1d24 | 236 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.20. |
5c6ed53a TB |
237 | @end table |
238 | ||
239 | ||
240 | ||
241 | @node omp_get_ancestor_thread_num | |
242 | @section @code{omp_get_ancestor_thread_num} -- Ancestor thread ID | |
243 | @table @asis | |
244 | @item @emph{Description}: | |
245 | This function returns the thread identification number for the given | |
83fd6c5b | 246 | nesting level of the current thread. For values of @var{level} outside |
5c6ed53a TB |
247 | zero to @code{omp_get_level} -1 is returned; if @var{level} is |
248 | @code{omp_get_level} the result is identical to @code{omp_get_thread_num}. | |
249 | ||
250 | @item @emph{C/C++} | |
251 | @multitable @columnfractions .20 .80 | |
252 | @item @emph{Prototype}: @tab @code{int omp_get_ancestor_thread_num(int level);} | |
253 | @end multitable | |
254 | ||
255 | @item @emph{Fortran}: | |
256 | @multitable @columnfractions .20 .80 | |
acb5c916 | 257 | @item @emph{Interface}: @tab @code{integer function omp_get_ancestor_thread_num(level)} |
5c6ed53a TB |
258 | @item @tab @code{integer level} |
259 | @end multitable | |
260 | ||
261 | @item @emph{See also}: | |
262 | @ref{omp_get_level}, @ref{omp_get_thread_num}, @ref{omp_get_team_size} | |
263 | ||
264 | @item @emph{Reference}: | |
1a6d1d24 | 265 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.18. |
83fd6c5b TB |
266 | @end table |
267 | ||
268 | ||
269 | ||
270 | @node omp_get_cancellation | |
271 | @section @code{omp_get_cancellation} -- Whether cancellation support is enabled | |
272 | @table @asis | |
273 | @item @emph{Description}: | |
274 | This function returns @code{true} if cancellation is activated, @code{false} | |
275 | otherwise. Here, @code{true} and @code{false} represent their language-specific | |
276 | counterparts. Unless @env{OMP_CANCELLATION} is set true, cancellations are | |
277 | deactivated. | |
278 | ||
279 | @item @emph{C/C++}: | |
280 | @multitable @columnfractions .20 .80 | |
281 | @item @emph{Prototype}: @tab @code{int omp_get_cancellation(void);} | |
282 | @end multitable | |
283 | ||
284 | @item @emph{Fortran}: | |
285 | @multitable @columnfractions .20 .80 | |
286 | @item @emph{Interface}: @tab @code{logical function omp_get_cancellation()} | |
287 | @end multitable | |
288 | ||
289 | @item @emph{See also}: | |
290 | @ref{OMP_CANCELLATION} | |
291 | ||
292 | @item @emph{Reference}: | |
1a6d1d24 | 293 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.9. |
83fd6c5b TB |
294 | @end table |
295 | ||
296 | ||
297 | ||
298 | @node omp_get_default_device | |
299 | @section @code{omp_get_default_device} -- Get the default device for target regions | |
300 | @table @asis | |
301 | @item @emph{Description}: | |
302 | Get the default device for target regions without device clause. | |
303 | ||
304 | @item @emph{C/C++}: | |
305 | @multitable @columnfractions .20 .80 | |
306 | @item @emph{Prototype}: @tab @code{int omp_get_default_device(void);} | |
307 | @end multitable | |
308 | ||
309 | @item @emph{Fortran}: | |
310 | @multitable @columnfractions .20 .80 | |
311 | @item @emph{Interface}: @tab @code{integer function omp_get_default_device()} | |
312 | @end multitable | |
313 | ||
314 | @item @emph{See also}: | |
315 | @ref{OMP_DEFAULT_DEVICE}, @ref{omp_set_default_device} | |
316 | ||
317 | @item @emph{Reference}: | |
1a6d1d24 | 318 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.30. |
5c6ed53a TB |
319 | @end table |
320 | ||
321 | ||
322 | ||
3721b9e1 DF |
323 | @node omp_get_dynamic |
324 | @section @code{omp_get_dynamic} -- Dynamic teams setting | |
325 | @table @asis | |
326 | @item @emph{Description}: | |
327 | This function returns @code{true} if enabled, @code{false} otherwise. | |
328 | Here, @code{true} and @code{false} represent their language-specific | |
329 | counterparts. | |
330 | ||
14734fc7 | 331 | The dynamic team setting may be initialized at startup by the |
83fd6c5b TB |
332 | @env{OMP_DYNAMIC} environment variable or at runtime using |
333 | @code{omp_set_dynamic}. If undefined, dynamic adjustment is | |
14734fc7 DF |
334 | disabled by default. |
335 | ||
3721b9e1 DF |
336 | @item @emph{C/C++}: |
337 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 338 | @item @emph{Prototype}: @tab @code{int omp_get_dynamic(void);} |
3721b9e1 DF |
339 | @end multitable |
340 | ||
341 | @item @emph{Fortran}: | |
342 | @multitable @columnfractions .20 .80 | |
343 | @item @emph{Interface}: @tab @code{logical function omp_get_dynamic()} | |
344 | @end multitable | |
345 | ||
346 | @item @emph{See also}: | |
14734fc7 | 347 | @ref{omp_set_dynamic}, @ref{OMP_DYNAMIC} |
3721b9e1 DF |
348 | |
349 | @item @emph{Reference}: | |
1a6d1d24 | 350 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.8. |
5c6ed53a TB |
351 | @end table |
352 | ||
353 | ||
354 | ||
355 | @node omp_get_level | |
356 | @section @code{omp_get_level} -- Obtain the current nesting level | |
357 | @table @asis | |
358 | @item @emph{Description}: | |
359 | This function returns the nesting level for the parallel blocks, | |
360 | which enclose the calling call. | |
361 | ||
362 | @item @emph{C/C++} | |
363 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 364 | @item @emph{Prototype}: @tab @code{int omp_get_level(void);} |
5c6ed53a TB |
365 | @end multitable |
366 | ||
367 | @item @emph{Fortran}: | |
368 | @multitable @columnfractions .20 .80 | |
acb5c916 | 369 | @item @emph{Interface}: @tab @code{integer function omp_level()} |
5c6ed53a TB |
370 | @end multitable |
371 | ||
372 | @item @emph{See also}: | |
373 | @ref{omp_get_active_level} | |
374 | ||
375 | @item @emph{Reference}: | |
1a6d1d24 | 376 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.17. |
5c6ed53a TB |
377 | @end table |
378 | ||
379 | ||
380 | ||
381 | @node omp_get_max_active_levels | |
6a2ba183 | 382 | @section @code{omp_get_max_active_levels} -- Maximum number of active regions |
5c6ed53a TB |
383 | @table @asis |
384 | @item @emph{Description}: | |
6a2ba183 | 385 | This function obtains the maximum allowed number of nested, active parallel regions. |
5c6ed53a TB |
386 | |
387 | @item @emph{C/C++} | |
388 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 389 | @item @emph{Prototype}: @tab @code{int omp_get_max_active_levels(void);} |
5c6ed53a TB |
390 | @end multitable |
391 | ||
392 | @item @emph{Fortran}: | |
393 | @multitable @columnfractions .20 .80 | |
acb5c916 | 394 | @item @emph{Interface}: @tab @code{integer function omp_get_max_active_levels()} |
5c6ed53a TB |
395 | @end multitable |
396 | ||
397 | @item @emph{See also}: | |
398 | @ref{omp_set_max_active_levels}, @ref{omp_get_active_level} | |
399 | ||
400 | @item @emph{Reference}: | |
1a6d1d24 | 401 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.16. |
3721b9e1 DF |
402 | @end table |
403 | ||
404 | ||
d9a6bd32 JJ |
405 | @node omp_get_max_task_priority |
406 | @section @code{omp_get_max_task_priority} -- Maximum priority value | |
407 | that can be set for tasks. | |
408 | @table @asis | |
409 | @item @emph{Description}: | |
410 | This function obtains the maximum allowed priority number for tasks. | |
411 | ||
412 | @item @emph{C/C++} | |
413 | @multitable @columnfractions .20 .80 | |
414 | @item @emph{Prototype}: @tab @code{int omp_get_max_task_priority(void);} | |
415 | @end multitable | |
416 | ||
417 | @item @emph{Fortran}: | |
418 | @multitable @columnfractions .20 .80 | |
419 | @item @emph{Interface}: @tab @code{integer function omp_get_max_task_priority()} | |
420 | @end multitable | |
421 | ||
422 | @item @emph{Reference}: | |
1a6d1d24 | 423 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.29. |
d9a6bd32 JJ |
424 | @end table |
425 | ||
3721b9e1 DF |
426 | |
427 | @node omp_get_max_threads | |
6a2ba183 | 428 | @section @code{omp_get_max_threads} -- Maximum number of threads of parallel region |
3721b9e1 DF |
429 | @table @asis |
430 | @item @emph{Description}: | |
6a2ba183 | 431 | Return the maximum number of threads used for the current parallel region |
5c6ed53a | 432 | that does not use the clause @code{num_threads}. |
3721b9e1 DF |
433 | |
434 | @item @emph{C/C++}: | |
435 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 436 | @item @emph{Prototype}: @tab @code{int omp_get_max_threads(void);} |
3721b9e1 DF |
437 | @end multitable |
438 | ||
439 | @item @emph{Fortran}: | |
440 | @multitable @columnfractions .20 .80 | |
441 | @item @emph{Interface}: @tab @code{integer function omp_get_max_threads()} | |
442 | @end multitable | |
443 | ||
444 | @item @emph{See also}: | |
5c6ed53a | 445 | @ref{omp_set_num_threads}, @ref{omp_set_dynamic}, @ref{omp_get_thread_limit} |
3721b9e1 DF |
446 | |
447 | @item @emph{Reference}: | |
1a6d1d24 | 448 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.3. |
3721b9e1 DF |
449 | @end table |
450 | ||
451 | ||
452 | ||
453 | @node omp_get_nested | |
454 | @section @code{omp_get_nested} -- Nested parallel regions | |
455 | @table @asis | |
456 | @item @emph{Description}: | |
457 | This function returns @code{true} if nested parallel regions are | |
83fd6c5b | 458 | enabled, @code{false} otherwise. Here, @code{true} and @code{false} |
3721b9e1 DF |
459 | represent their language-specific counterparts. |
460 | ||
14734fc7 | 461 | Nested parallel regions may be initialized at startup by the |
83fd6c5b TB |
462 | @env{OMP_NESTED} environment variable or at runtime using |
463 | @code{omp_set_nested}. If undefined, nested parallel regions are | |
14734fc7 DF |
464 | disabled by default. |
465 | ||
3721b9e1 DF |
466 | @item @emph{C/C++}: |
467 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 468 | @item @emph{Prototype}: @tab @code{int omp_get_nested(void);} |
3721b9e1 DF |
469 | @end multitable |
470 | ||
471 | @item @emph{Fortran}: | |
472 | @multitable @columnfractions .20 .80 | |
87350d4a | 473 | @item @emph{Interface}: @tab @code{logical function omp_get_nested()} |
3721b9e1 DF |
474 | @end multitable |
475 | ||
476 | @item @emph{See also}: | |
14734fc7 | 477 | @ref{omp_set_nested}, @ref{OMP_NESTED} |
3721b9e1 DF |
478 | |
479 | @item @emph{Reference}: | |
1a6d1d24 | 480 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.11. |
83fd6c5b TB |
481 | @end table |
482 | ||
483 | ||
484 | ||
485 | @node omp_get_num_devices | |
486 | @section @code{omp_get_num_devices} -- Number of target devices | |
487 | @table @asis | |
488 | @item @emph{Description}: | |
489 | Returns the number of target devices. | |
490 | ||
491 | @item @emph{C/C++}: | |
492 | @multitable @columnfractions .20 .80 | |
493 | @item @emph{Prototype}: @tab @code{int omp_get_num_devices(void);} | |
494 | @end multitable | |
495 | ||
496 | @item @emph{Fortran}: | |
497 | @multitable @columnfractions .20 .80 | |
498 | @item @emph{Interface}: @tab @code{integer function omp_get_num_devices()} | |
499 | @end multitable | |
500 | ||
501 | @item @emph{Reference}: | |
1a6d1d24 | 502 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.31. |
3721b9e1 DF |
503 | @end table |
504 | ||
505 | ||
506 | ||
507 | @node omp_get_num_procs | |
508 | @section @code{omp_get_num_procs} -- Number of processors online | |
509 | @table @asis | |
510 | @item @emph{Description}: | |
83fd6c5b | 511 | Returns the number of processors online on that device. |
3721b9e1 DF |
512 | |
513 | @item @emph{C/C++}: | |
514 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 515 | @item @emph{Prototype}: @tab @code{int omp_get_num_procs(void);} |
3721b9e1 DF |
516 | @end multitable |
517 | ||
518 | @item @emph{Fortran}: | |
519 | @multitable @columnfractions .20 .80 | |
520 | @item @emph{Interface}: @tab @code{integer function omp_get_num_procs()} | |
521 | @end multitable | |
522 | ||
523 | @item @emph{Reference}: | |
1a6d1d24 | 524 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.5. |
83fd6c5b TB |
525 | @end table |
526 | ||
527 | ||
528 | ||
529 | @node omp_get_num_teams | |
530 | @section @code{omp_get_num_teams} -- Number of teams | |
531 | @table @asis | |
532 | @item @emph{Description}: | |
533 | Returns the number of teams in the current team region. | |
534 | ||
535 | @item @emph{C/C++}: | |
536 | @multitable @columnfractions .20 .80 | |
537 | @item @emph{Prototype}: @tab @code{int omp_get_num_teams(void);} | |
538 | @end multitable | |
539 | ||
540 | @item @emph{Fortran}: | |
541 | @multitable @columnfractions .20 .80 | |
542 | @item @emph{Interface}: @tab @code{integer function omp_get_num_teams()} | |
543 | @end multitable | |
544 | ||
545 | @item @emph{Reference}: | |
1a6d1d24 | 546 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.32. |
3721b9e1 DF |
547 | @end table |
548 | ||
549 | ||
550 | ||
551 | @node omp_get_num_threads | |
552 | @section @code{omp_get_num_threads} -- Size of the active team | |
553 | @table @asis | |
554 | @item @emph{Description}: | |
83fd6c5b | 555 | Returns the number of threads in the current team. In a sequential section of |
3721b9e1 DF |
556 | the program @code{omp_get_num_threads} returns 1. |
557 | ||
14734fc7 | 558 | The default team size may be initialized at startup by the |
83fd6c5b | 559 | @env{OMP_NUM_THREADS} environment variable. At runtime, the size |
14734fc7 | 560 | of the current team may be set either by the @code{NUM_THREADS} |
83fd6c5b TB |
561 | clause or by @code{omp_set_num_threads}. If none of the above were |
562 | used to define a specific value and @env{OMP_DYNAMIC} is disabled, | |
14734fc7 DF |
563 | one thread per CPU online is used. |
564 | ||
3721b9e1 DF |
565 | @item @emph{C/C++}: |
566 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 567 | @item @emph{Prototype}: @tab @code{int omp_get_num_threads(void);} |
3721b9e1 DF |
568 | @end multitable |
569 | ||
570 | @item @emph{Fortran}: | |
571 | @multitable @columnfractions .20 .80 | |
572 | @item @emph{Interface}: @tab @code{integer function omp_get_num_threads()} | |
573 | @end multitable | |
574 | ||
575 | @item @emph{See also}: | |
576 | @ref{omp_get_max_threads}, @ref{omp_set_num_threads}, @ref{OMP_NUM_THREADS} | |
577 | ||
578 | @item @emph{Reference}: | |
1a6d1d24 | 579 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.2. |
83fd6c5b TB |
580 | @end table |
581 | ||
582 | ||
583 | ||
584 | @node omp_get_proc_bind | |
585 | @section @code{omp_get_proc_bind} -- Whether theads may be moved between CPUs | |
586 | @table @asis | |
587 | @item @emph{Description}: | |
588 | This functions returns the currently active thread affinity policy, which is | |
589 | set via @env{OMP_PROC_BIND}. Possible values are @code{omp_proc_bind_false}, | |
590 | @code{omp_proc_bind_true}, @code{omp_proc_bind_master}, | |
591 | @code{omp_proc_bind_close} and @code{omp_proc_bind_spread}. | |
592 | ||
593 | @item @emph{C/C++}: | |
594 | @multitable @columnfractions .20 .80 | |
595 | @item @emph{Prototype}: @tab @code{omp_proc_bind_t omp_get_proc_bind(void);} | |
596 | @end multitable | |
597 | ||
598 | @item @emph{Fortran}: | |
599 | @multitable @columnfractions .20 .80 | |
600 | @item @emph{Interface}: @tab @code{integer(kind=omp_proc_bind_kind) function omp_get_proc_bind()} | |
601 | @end multitable | |
602 | ||
603 | @item @emph{See also}: | |
604 | @ref{OMP_PROC_BIND}, @ref{OMP_PLACES}, @ref{GOMP_CPU_AFFINITY}, | |
605 | ||
606 | @item @emph{Reference}: | |
1a6d1d24 | 607 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.22. |
5c6ed53a TB |
608 | @end table |
609 | ||
610 | ||
611 | ||
612 | @node omp_get_schedule | |
613 | @section @code{omp_get_schedule} -- Obtain the runtime scheduling method | |
614 | @table @asis | |
615 | @item @emph{Description}: | |
83fd6c5b | 616 | Obtain the runtime scheduling method. The @var{kind} argument will be |
5c6ed53a | 617 | set to the value @code{omp_sched_static}, @code{omp_sched_dynamic}, |
83fd6c5b | 618 | @code{omp_sched_guided} or @code{omp_sched_auto}. The second argument, |
d9a6bd32 | 619 | @var{chunk_size}, is set to the chunk size. |
5c6ed53a TB |
620 | |
621 | @item @emph{C/C++} | |
622 | @multitable @columnfractions .20 .80 | |
d9a6bd32 | 623 | @item @emph{Prototype}: @tab @code{void omp_get_schedule(omp_sched_t *kind, int *chunk_size);} |
5c6ed53a TB |
624 | @end multitable |
625 | ||
626 | @item @emph{Fortran}: | |
627 | @multitable @columnfractions .20 .80 | |
d9a6bd32 | 628 | @item @emph{Interface}: @tab @code{subroutine omp_get_schedule(kind, chunk_size)} |
5c6ed53a | 629 | @item @tab @code{integer(kind=omp_sched_kind) kind} |
d9a6bd32 | 630 | @item @tab @code{integer chunk_size} |
5c6ed53a TB |
631 | @end multitable |
632 | ||
633 | @item @emph{See also}: | |
634 | @ref{omp_set_schedule}, @ref{OMP_SCHEDULE} | |
635 | ||
636 | @item @emph{Reference}: | |
1a6d1d24 | 637 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.13. |
83fd6c5b TB |
638 | @end table |
639 | ||
640 | ||
641 | ||
642 | @node omp_get_team_num | |
643 | @section @code{omp_get_team_num} -- Get team number | |
644 | @table @asis | |
645 | @item @emph{Description}: | |
646 | Returns the team number of the calling thread. | |
647 | ||
648 | @item @emph{C/C++}: | |
649 | @multitable @columnfractions .20 .80 | |
650 | @item @emph{Prototype}: @tab @code{int omp_get_team_num(void);} | |
651 | @end multitable | |
652 | ||
653 | @item @emph{Fortran}: | |
654 | @multitable @columnfractions .20 .80 | |
655 | @item @emph{Interface}: @tab @code{integer function omp_get_team_num()} | |
656 | @end multitable | |
657 | ||
658 | @item @emph{Reference}: | |
1a6d1d24 | 659 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.33. |
5c6ed53a TB |
660 | @end table |
661 | ||
662 | ||
663 | ||
664 | @node omp_get_team_size | |
665 | @section @code{omp_get_team_size} -- Number of threads in a team | |
666 | @table @asis | |
667 | @item @emph{Description}: | |
668 | This function returns the number of threads in a thread team to which | |
83fd6c5b | 669 | either the current thread or its ancestor belongs. For values of @var{level} |
6a2ba183 AH |
670 | outside zero to @code{omp_get_level}, -1 is returned; if @var{level} is zero, |
671 | 1 is returned, and for @code{omp_get_level}, the result is identical | |
5c6ed53a TB |
672 | to @code{omp_get_num_threads}. |
673 | ||
674 | @item @emph{C/C++}: | |
675 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 676 | @item @emph{Prototype}: @tab @code{int omp_get_team_size(int level);} |
5c6ed53a TB |
677 | @end multitable |
678 | ||
679 | @item @emph{Fortran}: | |
680 | @multitable @columnfractions .20 .80 | |
681 | @item @emph{Interface}: @tab @code{integer function omp_get_team_size(level)} | |
682 | @item @tab @code{integer level} | |
683 | @end multitable | |
684 | ||
685 | @item @emph{See also}: | |
686 | @ref{omp_get_num_threads}, @ref{omp_get_level}, @ref{omp_get_ancestor_thread_num} | |
687 | ||
688 | @item @emph{Reference}: | |
1a6d1d24 | 689 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.19. |
5c6ed53a TB |
690 | @end table |
691 | ||
692 | ||
693 | ||
694 | @node omp_get_thread_limit | |
6a2ba183 | 695 | @section @code{omp_get_thread_limit} -- Maximum number of threads |
5c6ed53a TB |
696 | @table @asis |
697 | @item @emph{Description}: | |
6a2ba183 | 698 | Return the maximum number of threads of the program. |
5c6ed53a TB |
699 | |
700 | @item @emph{C/C++}: | |
701 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 702 | @item @emph{Prototype}: @tab @code{int omp_get_thread_limit(void);} |
5c6ed53a TB |
703 | @end multitable |
704 | ||
705 | @item @emph{Fortran}: | |
706 | @multitable @columnfractions .20 .80 | |
707 | @item @emph{Interface}: @tab @code{integer function omp_get_thread_limit()} | |
708 | @end multitable | |
709 | ||
710 | @item @emph{See also}: | |
711 | @ref{omp_get_max_threads}, @ref{OMP_THREAD_LIMIT} | |
712 | ||
713 | @item @emph{Reference}: | |
1a6d1d24 | 714 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.14. |
3721b9e1 DF |
715 | @end table |
716 | ||
717 | ||
718 | ||
83fd6c5b | 719 | @node omp_get_thread_num |
3721b9e1 DF |
720 | @section @code{omp_get_thread_num} -- Current thread ID |
721 | @table @asis | |
722 | @item @emph{Description}: | |
6a2ba183 | 723 | Returns a unique thread identification number within the current team. |
5c6ed53a | 724 | In a sequential parts of the program, @code{omp_get_thread_num} |
83fd6c5b TB |
725 | always returns 0. In parallel regions the return value varies |
726 | from 0 to @code{omp_get_num_threads}-1 inclusive. The return | |
3721b9e1 DF |
727 | value of the master thread of a team is always 0. |
728 | ||
729 | @item @emph{C/C++}: | |
730 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 731 | @item @emph{Prototype}: @tab @code{int omp_get_thread_num(void);} |
3721b9e1 DF |
732 | @end multitable |
733 | ||
734 | @item @emph{Fortran}: | |
735 | @multitable @columnfractions .20 .80 | |
736 | @item @emph{Interface}: @tab @code{integer function omp_get_thread_num()} | |
737 | @end multitable | |
738 | ||
739 | @item @emph{See also}: | |
5c6ed53a | 740 | @ref{omp_get_num_threads}, @ref{omp_get_ancestor_thread_num} |
3721b9e1 DF |
741 | |
742 | @item @emph{Reference}: | |
1a6d1d24 | 743 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.4. |
3721b9e1 DF |
744 | @end table |
745 | ||
746 | ||
747 | ||
748 | @node omp_in_parallel | |
749 | @section @code{omp_in_parallel} -- Whether a parallel region is active | |
750 | @table @asis | |
751 | @item @emph{Description}: | |
83fd6c5b TB |
752 | This function returns @code{true} if currently running in parallel, |
753 | @code{false} otherwise. Here, @code{true} and @code{false} represent | |
3721b9e1 DF |
754 | their language-specific counterparts. |
755 | ||
756 | @item @emph{C/C++}: | |
757 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 758 | @item @emph{Prototype}: @tab @code{int omp_in_parallel(void);} |
3721b9e1 DF |
759 | @end multitable |
760 | ||
761 | @item @emph{Fortran}: | |
762 | @multitable @columnfractions .20 .80 | |
763 | @item @emph{Interface}: @tab @code{logical function omp_in_parallel()} | |
764 | @end multitable | |
765 | ||
766 | @item @emph{Reference}: | |
1a6d1d24 | 767 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.6. |
20906c66 JJ |
768 | @end table |
769 | ||
770 | ||
771 | @node omp_in_final | |
772 | @section @code{omp_in_final} -- Whether in final or included task region | |
773 | @table @asis | |
774 | @item @emph{Description}: | |
775 | This function returns @code{true} if currently running in a final | |
83fd6c5b | 776 | or included task region, @code{false} otherwise. Here, @code{true} |
20906c66 JJ |
777 | and @code{false} represent their language-specific counterparts. |
778 | ||
779 | @item @emph{C/C++}: | |
780 | @multitable @columnfractions .20 .80 | |
781 | @item @emph{Prototype}: @tab @code{int omp_in_final(void);} | |
782 | @end multitable | |
783 | ||
784 | @item @emph{Fortran}: | |
785 | @multitable @columnfractions .20 .80 | |
786 | @item @emph{Interface}: @tab @code{logical function omp_in_final()} | |
787 | @end multitable | |
788 | ||
789 | @item @emph{Reference}: | |
1a6d1d24 | 790 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.21. |
3721b9e1 DF |
791 | @end table |
792 | ||
793 | ||
83fd6c5b TB |
794 | |
795 | @node omp_is_initial_device | |
796 | @section @code{omp_is_initial_device} -- Whether executing on the host device | |
797 | @table @asis | |
798 | @item @emph{Description}: | |
799 | This function returns @code{true} if currently running on the host device, | |
800 | @code{false} otherwise. Here, @code{true} and @code{false} represent | |
801 | their language-specific counterparts. | |
802 | ||
803 | @item @emph{C/C++}: | |
804 | @multitable @columnfractions .20 .80 | |
805 | @item @emph{Prototype}: @tab @code{int omp_is_initial_device(void);} | |
806 | @end multitable | |
807 | ||
808 | @item @emph{Fortran}: | |
809 | @multitable @columnfractions .20 .80 | |
810 | @item @emph{Interface}: @tab @code{logical function omp_is_initial_device()} | |
811 | @end multitable | |
812 | ||
813 | @item @emph{Reference}: | |
1a6d1d24 | 814 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.34. |
83fd6c5b TB |
815 | @end table |
816 | ||
817 | ||
818 | ||
819 | @node omp_set_default_device | |
820 | @section @code{omp_set_default_device} -- Set the default device for target regions | |
821 | @table @asis | |
822 | @item @emph{Description}: | |
823 | Set the default device for target regions without device clause. The argument | |
824 | shall be a nonnegative device number. | |
825 | ||
826 | @item @emph{C/C++}: | |
827 | @multitable @columnfractions .20 .80 | |
828 | @item @emph{Prototype}: @tab @code{void omp_set_default_device(int device_num);} | |
829 | @end multitable | |
830 | ||
831 | @item @emph{Fortran}: | |
832 | @multitable @columnfractions .20 .80 | |
833 | @item @emph{Interface}: @tab @code{subroutine omp_set_default_device(device_num)} | |
834 | @item @tab @code{integer device_num} | |
835 | @end multitable | |
836 | ||
837 | @item @emph{See also}: | |
838 | @ref{OMP_DEFAULT_DEVICE}, @ref{omp_get_default_device} | |
839 | ||
840 | @item @emph{Reference}: | |
1a6d1d24 | 841 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.29. |
83fd6c5b TB |
842 | @end table |
843 | ||
844 | ||
845 | ||
3721b9e1 DF |
846 | @node omp_set_dynamic |
847 | @section @code{omp_set_dynamic} -- Enable/disable dynamic teams | |
848 | @table @asis | |
849 | @item @emph{Description}: | |
850 | Enable or disable the dynamic adjustment of the number of threads | |
83fd6c5b | 851 | within a team. The function takes the language-specific equivalent |
3721b9e1 DF |
852 | of @code{true} and @code{false}, where @code{true} enables dynamic |
853 | adjustment of team sizes and @code{false} disables it. | |
854 | ||
855 | @item @emph{C/C++}: | |
856 | @multitable @columnfractions .20 .80 | |
4fed6b25 | 857 | @item @emph{Prototype}: @tab @code{void omp_set_dynamic(int dynamic_threads);} |
3721b9e1 DF |
858 | @end multitable |
859 | ||
860 | @item @emph{Fortran}: | |
861 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
862 | @item @emph{Interface}: @tab @code{subroutine omp_set_dynamic(dynamic_threads)} |
863 | @item @tab @code{logical, intent(in) :: dynamic_threads} | |
3721b9e1 DF |
864 | @end multitable |
865 | ||
866 | @item @emph{See also}: | |
867 | @ref{OMP_DYNAMIC}, @ref{omp_get_dynamic} | |
868 | ||
869 | @item @emph{Reference}: | |
1a6d1d24 | 870 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.7. |
5c6ed53a TB |
871 | @end table |
872 | ||
873 | ||
874 | ||
875 | @node omp_set_max_active_levels | |
876 | @section @code{omp_set_max_active_levels} -- Limits the number of active parallel regions | |
877 | @table @asis | |
878 | @item @emph{Description}: | |
6a2ba183 AH |
879 | This function limits the maximum allowed number of nested, active |
880 | parallel regions. | |
5c6ed53a TB |
881 | |
882 | @item @emph{C/C++} | |
883 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 884 | @item @emph{Prototype}: @tab @code{void omp_set_max_active_levels(int max_levels);} |
5c6ed53a TB |
885 | @end multitable |
886 | ||
887 | @item @emph{Fortran}: | |
888 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 889 | @item @emph{Interface}: @tab @code{subroutine omp_set_max_active_levels(max_levels)} |
5c6ed53a TB |
890 | @item @tab @code{integer max_levels} |
891 | @end multitable | |
892 | ||
893 | @item @emph{See also}: | |
894 | @ref{omp_get_max_active_levels}, @ref{omp_get_active_level} | |
895 | ||
896 | @item @emph{Reference}: | |
1a6d1d24 | 897 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.15. |
3721b9e1 DF |
898 | @end table |
899 | ||
900 | ||
901 | ||
902 | @node omp_set_nested | |
903 | @section @code{omp_set_nested} -- Enable/disable nested parallel regions | |
904 | @table @asis | |
905 | @item @emph{Description}: | |
f1b0882e | 906 | Enable or disable nested parallel regions, i.e., whether team members |
83fd6c5b | 907 | are allowed to create new teams. The function takes the language-specific |
3721b9e1 DF |
908 | equivalent of @code{true} and @code{false}, where @code{true} enables |
909 | dynamic adjustment of team sizes and @code{false} disables it. | |
910 | ||
911 | @item @emph{C/C++}: | |
912 | @multitable @columnfractions .20 .80 | |
4fed6b25 | 913 | @item @emph{Prototype}: @tab @code{void omp_set_nested(int nested);} |
3721b9e1 DF |
914 | @end multitable |
915 | ||
916 | @item @emph{Fortran}: | |
917 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
918 | @item @emph{Interface}: @tab @code{subroutine omp_set_nested(nested)} |
919 | @item @tab @code{logical, intent(in) :: nested} | |
3721b9e1 DF |
920 | @end multitable |
921 | ||
922 | @item @emph{See also}: | |
923 | @ref{OMP_NESTED}, @ref{omp_get_nested} | |
924 | ||
925 | @item @emph{Reference}: | |
1a6d1d24 | 926 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.10. |
3721b9e1 DF |
927 | @end table |
928 | ||
929 | ||
930 | ||
931 | @node omp_set_num_threads | |
932 | @section @code{omp_set_num_threads} -- Set upper team size limit | |
933 | @table @asis | |
934 | @item @emph{Description}: | |
935 | Specifies the number of threads used by default in subsequent parallel | |
83fd6c5b TB |
936 | sections, if those do not specify a @code{num_threads} clause. The |
937 | argument of @code{omp_set_num_threads} shall be a positive integer. | |
3721b9e1 | 938 | |
3721b9e1 DF |
939 | @item @emph{C/C++}: |
940 | @multitable @columnfractions .20 .80 | |
4fed6b25 | 941 | @item @emph{Prototype}: @tab @code{void omp_set_num_threads(int num_threads);} |
3721b9e1 DF |
942 | @end multitable |
943 | ||
944 | @item @emph{Fortran}: | |
945 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
946 | @item @emph{Interface}: @tab @code{subroutine omp_set_num_threads(num_threads)} |
947 | @item @tab @code{integer, intent(in) :: num_threads} | |
3721b9e1 DF |
948 | @end multitable |
949 | ||
950 | @item @emph{See also}: | |
951 | @ref{OMP_NUM_THREADS}, @ref{omp_get_num_threads}, @ref{omp_get_max_threads} | |
952 | ||
953 | @item @emph{Reference}: | |
1a6d1d24 | 954 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.1. |
5c6ed53a TB |
955 | @end table |
956 | ||
957 | ||
958 | ||
959 | @node omp_set_schedule | |
960 | @section @code{omp_set_schedule} -- Set the runtime scheduling method | |
961 | @table @asis | |
962 | @item @emph{Description}: | |
83fd6c5b | 963 | Sets the runtime scheduling method. The @var{kind} argument can have the |
5c6ed53a | 964 | value @code{omp_sched_static}, @code{omp_sched_dynamic}, |
83fd6c5b | 965 | @code{omp_sched_guided} or @code{omp_sched_auto}. Except for |
5c6ed53a | 966 | @code{omp_sched_auto}, the chunk size is set to the value of |
d9a6bd32 JJ |
967 | @var{chunk_size} if positive, or to the default value if zero or negative. |
968 | For @code{omp_sched_auto} the @var{chunk_size} argument is ignored. | |
5c6ed53a TB |
969 | |
970 | @item @emph{C/C++} | |
971 | @multitable @columnfractions .20 .80 | |
d9a6bd32 | 972 | @item @emph{Prototype}: @tab @code{void omp_set_schedule(omp_sched_t kind, int chunk_size);} |
5c6ed53a TB |
973 | @end multitable |
974 | ||
975 | @item @emph{Fortran}: | |
976 | @multitable @columnfractions .20 .80 | |
d9a6bd32 | 977 | @item @emph{Interface}: @tab @code{subroutine omp_set_schedule(kind, chunk_size)} |
5c6ed53a | 978 | @item @tab @code{integer(kind=omp_sched_kind) kind} |
d9a6bd32 | 979 | @item @tab @code{integer chunk_size} |
5c6ed53a TB |
980 | @end multitable |
981 | ||
982 | @item @emph{See also}: | |
983 | @ref{omp_get_schedule} | |
984 | @ref{OMP_SCHEDULE} | |
985 | ||
986 | @item @emph{Reference}: | |
1a6d1d24 | 987 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.2.12. |
3721b9e1 DF |
988 | @end table |
989 | ||
990 | ||
991 | ||
992 | @node omp_init_lock | |
993 | @section @code{omp_init_lock} -- Initialize simple lock | |
994 | @table @asis | |
995 | @item @emph{Description}: | |
83fd6c5b | 996 | Initialize a simple lock. After initialization, the lock is in |
3721b9e1 DF |
997 | an unlocked state. |
998 | ||
999 | @item @emph{C/C++}: | |
1000 | @multitable @columnfractions .20 .80 | |
1001 | @item @emph{Prototype}: @tab @code{void omp_init_lock(omp_lock_t *lock);} | |
1002 | @end multitable | |
1003 | ||
1004 | @item @emph{Fortran}: | |
1005 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1006 | @item @emph{Interface}: @tab @code{subroutine omp_init_lock(svar)} |
1007 | @item @tab @code{integer(omp_lock_kind), intent(out) :: svar} | |
3721b9e1 DF |
1008 | @end multitable |
1009 | ||
1010 | @item @emph{See also}: | |
1011 | @ref{omp_destroy_lock} | |
1012 | ||
1013 | @item @emph{Reference}: | |
1a6d1d24 | 1014 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.1. |
3721b9e1 DF |
1015 | @end table |
1016 | ||
1017 | ||
1018 | ||
1019 | @node omp_set_lock | |
1020 | @section @code{omp_set_lock} -- Wait for and set simple lock | |
1021 | @table @asis | |
1022 | @item @emph{Description}: | |
1023 | Before setting a simple lock, the lock variable must be initialized by | |
83fd6c5b TB |
1024 | @code{omp_init_lock}. The calling thread is blocked until the lock |
1025 | is available. If the lock is already held by the current thread, | |
3721b9e1 DF |
1026 | a deadlock occurs. |
1027 | ||
1028 | @item @emph{C/C++}: | |
1029 | @multitable @columnfractions .20 .80 | |
1030 | @item @emph{Prototype}: @tab @code{void omp_set_lock(omp_lock_t *lock);} | |
1031 | @end multitable | |
1032 | ||
1033 | @item @emph{Fortran}: | |
1034 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1035 | @item @emph{Interface}: @tab @code{subroutine omp_set_lock(svar)} |
1036 | @item @tab @code{integer(omp_lock_kind), intent(inout) :: svar} | |
3721b9e1 DF |
1037 | @end multitable |
1038 | ||
1039 | @item @emph{See also}: | |
1040 | @ref{omp_init_lock}, @ref{omp_test_lock}, @ref{omp_unset_lock} | |
1041 | ||
1042 | @item @emph{Reference}: | |
1a6d1d24 | 1043 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.4. |
3721b9e1 DF |
1044 | @end table |
1045 | ||
1046 | ||
1047 | ||
1048 | @node omp_test_lock | |
1049 | @section @code{omp_test_lock} -- Test and set simple lock if available | |
1050 | @table @asis | |
1051 | @item @emph{Description}: | |
1052 | Before setting a simple lock, the lock variable must be initialized by | |
83fd6c5b TB |
1053 | @code{omp_init_lock}. Contrary to @code{omp_set_lock}, @code{omp_test_lock} |
1054 | does not block if the lock is not available. This function returns | |
1055 | @code{true} upon success, @code{false} otherwise. Here, @code{true} and | |
3721b9e1 DF |
1056 | @code{false} represent their language-specific counterparts. |
1057 | ||
1058 | @item @emph{C/C++}: | |
1059 | @multitable @columnfractions .20 .80 | |
1060 | @item @emph{Prototype}: @tab @code{int omp_test_lock(omp_lock_t *lock);} | |
1061 | @end multitable | |
1062 | ||
1063 | @item @emph{Fortran}: | |
1064 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1065 | @item @emph{Interface}: @tab @code{logical function omp_test_lock(svar)} |
1066 | @item @tab @code{integer(omp_lock_kind), intent(inout) :: svar} | |
3721b9e1 DF |
1067 | @end multitable |
1068 | ||
1069 | @item @emph{See also}: | |
1070 | @ref{omp_init_lock}, @ref{omp_set_lock}, @ref{omp_set_lock} | |
1071 | ||
1072 | @item @emph{Reference}: | |
1a6d1d24 | 1073 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.6. |
3721b9e1 DF |
1074 | @end table |
1075 | ||
1076 | ||
1077 | ||
1078 | @node omp_unset_lock | |
1079 | @section @code{omp_unset_lock} -- Unset simple lock | |
1080 | @table @asis | |
1081 | @item @emph{Description}: | |
1082 | A simple lock about to be unset must have been locked by @code{omp_set_lock} | |
83fd6c5b TB |
1083 | or @code{omp_test_lock} before. In addition, the lock must be held by the |
1084 | thread calling @code{omp_unset_lock}. Then, the lock becomes unlocked. If one | |
1085 | or more threads attempted to set the lock before, one of them is chosen to, | |
20906c66 | 1086 | again, set the lock to itself. |
3721b9e1 DF |
1087 | |
1088 | @item @emph{C/C++}: | |
1089 | @multitable @columnfractions .20 .80 | |
1090 | @item @emph{Prototype}: @tab @code{void omp_unset_lock(omp_lock_t *lock);} | |
1091 | @end multitable | |
1092 | ||
1093 | @item @emph{Fortran}: | |
1094 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1095 | @item @emph{Interface}: @tab @code{subroutine omp_unset_lock(svar)} |
1096 | @item @tab @code{integer(omp_lock_kind), intent(inout) :: svar} | |
3721b9e1 DF |
1097 | @end multitable |
1098 | ||
1099 | @item @emph{See also}: | |
1100 | @ref{omp_set_lock}, @ref{omp_test_lock} | |
1101 | ||
1102 | @item @emph{Reference}: | |
1a6d1d24 | 1103 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.5. |
3721b9e1 DF |
1104 | @end table |
1105 | ||
1106 | ||
1107 | ||
1108 | @node omp_destroy_lock | |
1109 | @section @code{omp_destroy_lock} -- Destroy simple lock | |
1110 | @table @asis | |
1111 | @item @emph{Description}: | |
83fd6c5b | 1112 | Destroy a simple lock. In order to be destroyed, a simple lock must be |
3721b9e1 DF |
1113 | in the unlocked state. |
1114 | ||
1115 | @item @emph{C/C++}: | |
1116 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 1117 | @item @emph{Prototype}: @tab @code{void omp_destroy_lock(omp_lock_t *lock);} |
3721b9e1 DF |
1118 | @end multitable |
1119 | ||
1120 | @item @emph{Fortran}: | |
1121 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1122 | @item @emph{Interface}: @tab @code{subroutine omp_destroy_lock(svar)} |
1123 | @item @tab @code{integer(omp_lock_kind), intent(inout) :: svar} | |
3721b9e1 DF |
1124 | @end multitable |
1125 | ||
1126 | @item @emph{See also}: | |
1127 | @ref{omp_init_lock} | |
1128 | ||
1129 | @item @emph{Reference}: | |
1a6d1d24 | 1130 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.3. |
3721b9e1 DF |
1131 | @end table |
1132 | ||
1133 | ||
1134 | ||
1135 | @node omp_init_nest_lock | |
1136 | @section @code{omp_init_nest_lock} -- Initialize nested lock | |
1137 | @table @asis | |
1138 | @item @emph{Description}: | |
83fd6c5b | 1139 | Initialize a nested lock. After initialization, the lock is in |
3721b9e1 DF |
1140 | an unlocked state and the nesting count is set to zero. |
1141 | ||
1142 | @item @emph{C/C++}: | |
1143 | @multitable @columnfractions .20 .80 | |
1144 | @item @emph{Prototype}: @tab @code{void omp_init_nest_lock(omp_nest_lock_t *lock);} | |
1145 | @end multitable | |
1146 | ||
1147 | @item @emph{Fortran}: | |
1148 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1149 | @item @emph{Interface}: @tab @code{subroutine omp_init_nest_lock(nvar)} |
1150 | @item @tab @code{integer(omp_nest_lock_kind), intent(out) :: nvar} | |
3721b9e1 DF |
1151 | @end multitable |
1152 | ||
1153 | @item @emph{See also}: | |
1154 | @ref{omp_destroy_nest_lock} | |
1155 | ||
1156 | @item @emph{Reference}: | |
1a6d1d24 | 1157 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.1. |
3721b9e1 DF |
1158 | @end table |
1159 | ||
1160 | ||
1161 | @node omp_set_nest_lock | |
6a2ba183 | 1162 | @section @code{omp_set_nest_lock} -- Wait for and set nested lock |
3721b9e1 DF |
1163 | @table @asis |
1164 | @item @emph{Description}: | |
1165 | Before setting a nested lock, the lock variable must be initialized by | |
83fd6c5b TB |
1166 | @code{omp_init_nest_lock}. The calling thread is blocked until the lock |
1167 | is available. If the lock is already held by the current thread, the | |
20906c66 | 1168 | nesting count for the lock is incremented. |
3721b9e1 DF |
1169 | |
1170 | @item @emph{C/C++}: | |
1171 | @multitable @columnfractions .20 .80 | |
1172 | @item @emph{Prototype}: @tab @code{void omp_set_nest_lock(omp_nest_lock_t *lock);} | |
1173 | @end multitable | |
1174 | ||
1175 | @item @emph{Fortran}: | |
1176 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1177 | @item @emph{Interface}: @tab @code{subroutine omp_set_nest_lock(nvar)} |
1178 | @item @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar} | |
3721b9e1 DF |
1179 | @end multitable |
1180 | ||
1181 | @item @emph{See also}: | |
1182 | @ref{omp_init_nest_lock}, @ref{omp_unset_nest_lock} | |
1183 | ||
1184 | @item @emph{Reference}: | |
1a6d1d24 | 1185 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.4. |
3721b9e1 DF |
1186 | @end table |
1187 | ||
1188 | ||
1189 | ||
1190 | @node omp_test_nest_lock | |
1191 | @section @code{omp_test_nest_lock} -- Test and set nested lock if available | |
1192 | @table @asis | |
1193 | @item @emph{Description}: | |
1194 | Before setting a nested lock, the lock variable must be initialized by | |
83fd6c5b | 1195 | @code{omp_init_nest_lock}. Contrary to @code{omp_set_nest_lock}, |
3721b9e1 DF |
1196 | @code{omp_test_nest_lock} does not block if the lock is not available. |
1197 | If the lock is already held by the current thread, the new nesting count | |
83fd6c5b | 1198 | is returned. Otherwise, the return value equals zero. |
3721b9e1 DF |
1199 | |
1200 | @item @emph{C/C++}: | |
1201 | @multitable @columnfractions .20 .80 | |
1202 | @item @emph{Prototype}: @tab @code{int omp_test_nest_lock(omp_nest_lock_t *lock);} | |
1203 | @end multitable | |
1204 | ||
1205 | @item @emph{Fortran}: | |
1206 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1207 | @item @emph{Interface}: @tab @code{logical function omp_test_nest_lock(nvar)} |
1208 | @item @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar} | |
3721b9e1 DF |
1209 | @end multitable |
1210 | ||
1211 | ||
1212 | @item @emph{See also}: | |
1213 | @ref{omp_init_lock}, @ref{omp_set_lock}, @ref{omp_set_lock} | |
1214 | ||
1215 | @item @emph{Reference}: | |
1a6d1d24 | 1216 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.6. |
3721b9e1 DF |
1217 | @end table |
1218 | ||
1219 | ||
1220 | ||
1221 | @node omp_unset_nest_lock | |
1222 | @section @code{omp_unset_nest_lock} -- Unset nested lock | |
1223 | @table @asis | |
1224 | @item @emph{Description}: | |
1225 | A nested lock about to be unset must have been locked by @code{omp_set_nested_lock} | |
83fd6c5b TB |
1226 | or @code{omp_test_nested_lock} before. In addition, the lock must be held by the |
1227 | thread calling @code{omp_unset_nested_lock}. If the nesting count drops to zero, the | |
1228 | lock becomes unlocked. If one ore more threads attempted to set the lock before, | |
20906c66 | 1229 | one of them is chosen to, again, set the lock to itself. |
3721b9e1 DF |
1230 | |
1231 | @item @emph{C/C++}: | |
1232 | @multitable @columnfractions .20 .80 | |
1233 | @item @emph{Prototype}: @tab @code{void omp_unset_nest_lock(omp_nest_lock_t *lock);} | |
1234 | @end multitable | |
1235 | ||
1236 | @item @emph{Fortran}: | |
1237 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1238 | @item @emph{Interface}: @tab @code{subroutine omp_unset_nest_lock(nvar)} |
1239 | @item @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar} | |
3721b9e1 DF |
1240 | @end multitable |
1241 | ||
1242 | @item @emph{See also}: | |
1243 | @ref{omp_set_nest_lock} | |
1244 | ||
1245 | @item @emph{Reference}: | |
1a6d1d24 | 1246 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.5. |
3721b9e1 DF |
1247 | @end table |
1248 | ||
1249 | ||
1250 | ||
1251 | @node omp_destroy_nest_lock | |
1252 | @section @code{omp_destroy_nest_lock} -- Destroy nested lock | |
1253 | @table @asis | |
1254 | @item @emph{Description}: | |
83fd6c5b | 1255 | Destroy a nested lock. In order to be destroyed, a nested lock must be |
3721b9e1 DF |
1256 | in the unlocked state and its nesting count must equal zero. |
1257 | ||
1258 | @item @emph{C/C++}: | |
1259 | @multitable @columnfractions .20 .80 | |
1260 | @item @emph{Prototype}: @tab @code{void omp_destroy_nest_lock(omp_nest_lock_t *);} | |
1261 | @end multitable | |
1262 | ||
1263 | @item @emph{Fortran}: | |
1264 | @multitable @columnfractions .20 .80 | |
4fed6b25 TB |
1265 | @item @emph{Interface}: @tab @code{subroutine omp_destroy_nest_lock(nvar)} |
1266 | @item @tab @code{integer(omp_nest_lock_kind), intent(inout) :: nvar} | |
3721b9e1 DF |
1267 | @end multitable |
1268 | ||
1269 | @item @emph{See also}: | |
1270 | @ref{omp_init_lock} | |
1271 | ||
1272 | @item @emph{Reference}: | |
1a6d1d24 | 1273 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.3.3. |
3721b9e1 DF |
1274 | @end table |
1275 | ||
1276 | ||
1277 | ||
1278 | @node omp_get_wtick | |
1279 | @section @code{omp_get_wtick} -- Get timer precision | |
1280 | @table @asis | |
1281 | @item @emph{Description}: | |
f1b0882e | 1282 | Gets the timer precision, i.e., the number of seconds between two |
3721b9e1 DF |
1283 | successive clock ticks. |
1284 | ||
1285 | @item @emph{C/C++}: | |
1286 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 1287 | @item @emph{Prototype}: @tab @code{double omp_get_wtick(void);} |
3721b9e1 DF |
1288 | @end multitable |
1289 | ||
1290 | @item @emph{Fortran}: | |
1291 | @multitable @columnfractions .20 .80 | |
1292 | @item @emph{Interface}: @tab @code{double precision function omp_get_wtick()} | |
1293 | @end multitable | |
1294 | ||
1295 | @item @emph{See also}: | |
1296 | @ref{omp_get_wtime} | |
1297 | ||
1298 | @item @emph{Reference}: | |
1a6d1d24 | 1299 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.4.2. |
3721b9e1 DF |
1300 | @end table |
1301 | ||
1302 | ||
1303 | ||
1304 | @node omp_get_wtime | |
1305 | @section @code{omp_get_wtime} -- Elapsed wall clock time | |
1306 | @table @asis | |
1307 | @item @emph{Description}: | |
83fd6c5b | 1308 | Elapsed wall clock time in seconds. The time is measured per thread, no |
6a2ba183 | 1309 | guarantee can be made that two distinct threads measure the same time. |
21e1e594 JJ |
1310 | Time is measured from some "time in the past", which is an arbitrary time |
1311 | guaranteed not to change during the execution of the program. | |
3721b9e1 DF |
1312 | |
1313 | @item @emph{C/C++}: | |
1314 | @multitable @columnfractions .20 .80 | |
6a2ba183 | 1315 | @item @emph{Prototype}: @tab @code{double omp_get_wtime(void);} |
3721b9e1 DF |
1316 | @end multitable |
1317 | ||
1318 | @item @emph{Fortran}: | |
1319 | @multitable @columnfractions .20 .80 | |
1320 | @item @emph{Interface}: @tab @code{double precision function omp_get_wtime()} | |
1321 | @end multitable | |
1322 | ||
1323 | @item @emph{See also}: | |
1324 | @ref{omp_get_wtick} | |
1325 | ||
1326 | @item @emph{Reference}: | |
1a6d1d24 | 1327 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 3.4.1. |
3721b9e1 DF |
1328 | @end table |
1329 | ||
1330 | ||
1331 | ||
1332 | @c --------------------------------------------------------------------- | |
4102bda6 | 1333 | @c OpenMP Environment Variables |
3721b9e1 DF |
1334 | @c --------------------------------------------------------------------- |
1335 | ||
1336 | @node Environment Variables | |
4102bda6 | 1337 | @chapter OpenMP Environment Variables |
3721b9e1 | 1338 | |
acf0174b | 1339 | The environment variables which beginning with @env{OMP_} are defined by |
00b9bd52 | 1340 | section 4 of the OpenMP specification in version 4.5, while those |
acf0174b | 1341 | beginning with @env{GOMP_} are GNU extensions. |
3721b9e1 DF |
1342 | |
1343 | @menu | |
06441dd5 SH |
1344 | * OMP_CANCELLATION:: Set whether cancellation is activated |
1345 | * OMP_DISPLAY_ENV:: Show OpenMP version and environment variables | |
1346 | * OMP_DEFAULT_DEVICE:: Set the device used in target regions | |
1347 | * OMP_DYNAMIC:: Dynamic adjustment of threads | |
1348 | * OMP_MAX_ACTIVE_LEVELS:: Set the maximum number of nested parallel regions | |
d9a6bd32 | 1349 | * OMP_MAX_TASK_PRIORITY:: Set the maximum task priority value |
06441dd5 SH |
1350 | * OMP_NESTED:: Nested parallel regions |
1351 | * OMP_NUM_THREADS:: Specifies the number of threads to use | |
1352 | * OMP_PROC_BIND:: Whether theads may be moved between CPUs | |
1353 | * OMP_PLACES:: Specifies on which CPUs the theads should be placed | |
1354 | * OMP_STACKSIZE:: Set default thread stack size | |
1355 | * OMP_SCHEDULE:: How threads are scheduled | |
1356 | * OMP_THREAD_LIMIT:: Set the maximum number of threads | |
1357 | * OMP_WAIT_POLICY:: How waiting threads are handled | |
1358 | * GOMP_CPU_AFFINITY:: Bind threads to specific CPUs | |
1359 | * GOMP_DEBUG:: Enable debugging output | |
1360 | * GOMP_STACKSIZE:: Set default thread stack size | |
1361 | * GOMP_SPINCOUNT:: Set the busy-wait spin count | |
1362 | * GOMP_RTEMS_THREAD_POOLS:: Set the RTEMS specific thread pools | |
3721b9e1 DF |
1363 | @end menu |
1364 | ||
1365 | ||
83fd6c5b TB |
1366 | @node OMP_CANCELLATION |
1367 | @section @env{OMP_CANCELLATION} -- Set whether cancellation is activated | |
1368 | @cindex Environment Variable | |
1369 | @table @asis | |
1370 | @item @emph{Description}: | |
1371 | If set to @code{TRUE}, the cancellation is activated. If set to @code{FALSE} or | |
1372 | if unset, cancellation is disabled and the @code{cancel} construct is ignored. | |
1373 | ||
1374 | @item @emph{See also}: | |
1375 | @ref{omp_get_cancellation} | |
1376 | ||
1377 | @item @emph{Reference}: | |
1a6d1d24 | 1378 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.11 |
83fd6c5b TB |
1379 | @end table |
1380 | ||
1381 | ||
1382 | ||
1383 | @node OMP_DISPLAY_ENV | |
1384 | @section @env{OMP_DISPLAY_ENV} -- Show OpenMP version and environment variables | |
1385 | @cindex Environment Variable | |
1386 | @table @asis | |
1387 | @item @emph{Description}: | |
1388 | If set to @code{TRUE}, the OpenMP version number and the values | |
1389 | associated with the OpenMP environment variables are printed to @code{stderr}. | |
1390 | If set to @code{VERBOSE}, it additionally shows the value of the environment | |
1391 | variables which are GNU extensions. If undefined or set to @code{FALSE}, | |
1392 | this information will not be shown. | |
1393 | ||
1394 | ||
1395 | @item @emph{Reference}: | |
1a6d1d24 | 1396 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.12 |
83fd6c5b TB |
1397 | @end table |
1398 | ||
1399 | ||
1400 | ||
1401 | @node OMP_DEFAULT_DEVICE | |
1402 | @section @env{OMP_DEFAULT_DEVICE} -- Set the device used in target regions | |
1403 | @cindex Environment Variable | |
1404 | @table @asis | |
1405 | @item @emph{Description}: | |
1406 | Set to choose the device which is used in a @code{target} region, unless the | |
1407 | value is overridden by @code{omp_set_default_device} or by a @code{device} | |
1408 | clause. The value shall be the nonnegative device number. If no device with | |
1409 | the given device number exists, the code is executed on the host. If unset, | |
1410 | device number 0 will be used. | |
1411 | ||
1412 | ||
1413 | @item @emph{See also}: | |
1414 | @ref{omp_get_default_device}, @ref{omp_set_default_device}, | |
1415 | ||
1416 | @item @emph{Reference}: | |
1a6d1d24 | 1417 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.13 |
83fd6c5b TB |
1418 | @end table |
1419 | ||
1420 | ||
1421 | ||
3721b9e1 DF |
1422 | @node OMP_DYNAMIC |
1423 | @section @env{OMP_DYNAMIC} -- Dynamic adjustment of threads | |
1424 | @cindex Environment Variable | |
1425 | @table @asis | |
1426 | @item @emph{Description}: | |
1427 | Enable or disable the dynamic adjustment of the number of threads | |
83fd6c5b TB |
1428 | within a team. The value of this environment variable shall be |
1429 | @code{TRUE} or @code{FALSE}. If undefined, dynamic adjustment is | |
7c2b7f45 | 1430 | disabled by default. |
3721b9e1 DF |
1431 | |
1432 | @item @emph{See also}: | |
1433 | @ref{omp_set_dynamic} | |
1434 | ||
1435 | @item @emph{Reference}: | |
1a6d1d24 | 1436 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.3 |
5c6ed53a TB |
1437 | @end table |
1438 | ||
1439 | ||
1440 | ||
1441 | @node OMP_MAX_ACTIVE_LEVELS | |
6a2ba183 | 1442 | @section @env{OMP_MAX_ACTIVE_LEVELS} -- Set the maximum number of nested parallel regions |
5c6ed53a TB |
1443 | @cindex Environment Variable |
1444 | @table @asis | |
1445 | @item @emph{Description}: | |
6a2ba183 | 1446 | Specifies the initial value for the maximum number of nested parallel |
83fd6c5b | 1447 | regions. The value of this variable shall be a positive integer. |
5c6ed53a TB |
1448 | If undefined, the number of active levels is unlimited. |
1449 | ||
1450 | @item @emph{See also}: | |
1451 | @ref{omp_set_max_active_levels} | |
1452 | ||
1453 | @item @emph{Reference}: | |
1a6d1d24 | 1454 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.9 |
3721b9e1 DF |
1455 | @end table |
1456 | ||
1457 | ||
1458 | ||
d9a6bd32 JJ |
1459 | @node OMP_MAX_TASK_PRIORITY |
1460 | @section @env{OMP_MAX_TASK_PRIORITY} -- Set the maximum priority | |
1461 | number that can be set for a task. | |
1462 | @cindex Environment Variable | |
1463 | @table @asis | |
1464 | @item @emph{Description}: | |
1465 | Specifies the initial value for the maximum priority value that can be | |
1466 | set for a task. The value of this variable shall be a non-negative | |
1467 | integer, and zero is allowed. If undefined, the default priority is | |
1468 | 0. | |
1469 | ||
1470 | @item @emph{See also}: | |
1471 | @ref{omp_get_max_task_priority} | |
1472 | ||
1473 | @item @emph{Reference}: | |
1a6d1d24 | 1474 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.14 |
d9a6bd32 JJ |
1475 | @end table |
1476 | ||
1477 | ||
1478 | ||
3721b9e1 DF |
1479 | @node OMP_NESTED |
1480 | @section @env{OMP_NESTED} -- Nested parallel regions | |
1481 | @cindex Environment Variable | |
14734fc7 | 1482 | @cindex Implementation specific setting |
3721b9e1 DF |
1483 | @table @asis |
1484 | @item @emph{Description}: | |
f1b0882e | 1485 | Enable or disable nested parallel regions, i.e., whether team members |
83fd6c5b TB |
1486 | are allowed to create new teams. The value of this environment variable |
1487 | shall be @code{TRUE} or @code{FALSE}. If undefined, nested parallel | |
7c2b7f45 | 1488 | regions are disabled by default. |
3721b9e1 DF |
1489 | |
1490 | @item @emph{See also}: | |
1491 | @ref{omp_set_nested} | |
1492 | ||
1493 | @item @emph{Reference}: | |
1a6d1d24 | 1494 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.6 |
3721b9e1 DF |
1495 | @end table |
1496 | ||
1497 | ||
1498 | ||
1499 | @node OMP_NUM_THREADS | |
1500 | @section @env{OMP_NUM_THREADS} -- Specifies the number of threads to use | |
1501 | @cindex Environment Variable | |
14734fc7 | 1502 | @cindex Implementation specific setting |
3721b9e1 DF |
1503 | @table @asis |
1504 | @item @emph{Description}: | |
83fd6c5b | 1505 | Specifies the default number of threads to use in parallel regions. The |
20906c66 JJ |
1506 | value of this variable shall be a comma-separated list of positive integers; |
1507 | the value specified the number of threads to use for the corresponding nested | |
83fd6c5b | 1508 | level. If undefined one thread per CPU is used. |
3721b9e1 DF |
1509 | |
1510 | @item @emph{See also}: | |
1511 | @ref{omp_set_num_threads} | |
1512 | ||
1513 | @item @emph{Reference}: | |
1a6d1d24 | 1514 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.2 |
83fd6c5b TB |
1515 | @end table |
1516 | ||
1517 | ||
1518 | ||
72832460 UB |
1519 | @node OMP_PROC_BIND |
1520 | @section @env{OMP_PROC_BIND} -- Whether theads may be moved between CPUs | |
1521 | @cindex Environment Variable | |
1522 | @table @asis | |
1523 | @item @emph{Description}: | |
1524 | Specifies whether threads may be moved between processors. If set to | |
1525 | @code{TRUE}, OpenMP theads should not be moved; if set to @code{FALSE} | |
1526 | they may be moved. Alternatively, a comma separated list with the | |
1527 | values @code{MASTER}, @code{CLOSE} and @code{SPREAD} can be used to specify | |
1528 | the thread affinity policy for the corresponding nesting level. With | |
1529 | @code{MASTER} the worker threads are in the same place partition as the | |
1530 | master thread. With @code{CLOSE} those are kept close to the master thread | |
1531 | in contiguous place partitions. And with @code{SPREAD} a sparse distribution | |
1532 | across the place partitions is used. | |
1533 | ||
1534 | When undefined, @env{OMP_PROC_BIND} defaults to @code{TRUE} when | |
1535 | @env{OMP_PLACES} or @env{GOMP_CPU_AFFINITY} is set and @code{FALSE} otherwise. | |
1536 | ||
1537 | @item @emph{See also}: | |
1538 | @ref{OMP_PLACES}, @ref{GOMP_CPU_AFFINITY}, @ref{omp_get_proc_bind} | |
1539 | ||
1540 | @item @emph{Reference}: | |
1a6d1d24 | 1541 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.4 |
72832460 UB |
1542 | @end table |
1543 | ||
1544 | ||
1545 | ||
83fd6c5b TB |
1546 | @node OMP_PLACES |
1547 | @section @env{OMP_PLACES} -- Specifies on which CPUs the theads should be placed | |
1548 | @cindex Environment Variable | |
1549 | @table @asis | |
1550 | @item @emph{Description}: | |
1551 | The thread placement can be either specified using an abstract name or by an | |
1552 | explicit list of the places. The abstract names @code{threads}, @code{cores} | |
1553 | and @code{sockets} can be optionally followed by a positive number in | |
1554 | parentheses, which denotes the how many places shall be created. With | |
1555 | @code{threads} each place corresponds to a single hardware thread; @code{cores} | |
1556 | to a single core with the corresponding number of hardware threads; and with | |
1557 | @code{sockets} the place corresponds to a single socket. The resulting | |
1558 | placement can be shown by setting the @env{OMP_DISPLAY_ENV} environment | |
1559 | variable. | |
1560 | ||
1561 | Alternatively, the placement can be specified explicitly as comma-separated | |
1562 | list of places. A place is specified by set of nonnegative numbers in curly | |
1563 | braces, denoting the denoting the hardware threads. The hardware threads | |
1564 | belonging to a place can either be specified as comma-separated list of | |
1565 | nonnegative thread numbers or using an interval. Multiple places can also be | |
1566 | either specified by a comma-separated list of places or by an interval. To | |
1567 | specify an interval, a colon followed by the count is placed after after | |
1568 | the hardware thread number or the place. Optionally, the length can be | |
1569 | followed by a colon and the stride number -- otherwise a unit stride is | |
1570 | assumed. For instance, the following specifies the same places list: | |
1571 | @code{"@{0,1,2@}, @{3,4,6@}, @{7,8,9@}, @{10,11,12@}"}; | |
1572 | @code{"@{0:3@}, @{3:3@}, @{7:3@}, @{10:3@}"}; and @code{"@{0:2@}:4:3"}. | |
1573 | ||
1574 | If @env{OMP_PLACES} and @env{GOMP_CPU_AFFINITY} are unset and | |
1575 | @env{OMP_PROC_BIND} is either unset or @code{false}, threads may be moved | |
1576 | between CPUs following no placement policy. | |
1577 | ||
1578 | @item @emph{See also}: | |
1579 | @ref{OMP_PROC_BIND}, @ref{GOMP_CPU_AFFINITY}, @ref{omp_get_proc_bind}, | |
1580 | @ref{OMP_DISPLAY_ENV} | |
1581 | ||
1582 | @item @emph{Reference}: | |
1a6d1d24 | 1583 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.5 |
83fd6c5b TB |
1584 | @end table |
1585 | ||
1586 | ||
1587 | ||
72832460 UB |
1588 | @node OMP_STACKSIZE |
1589 | @section @env{OMP_STACKSIZE} -- Set default thread stack size | |
83fd6c5b TB |
1590 | @cindex Environment Variable |
1591 | @table @asis | |
1592 | @item @emph{Description}: | |
72832460 UB |
1593 | Set the default thread stack size in kilobytes, unless the number |
1594 | is suffixed by @code{B}, @code{K}, @code{M} or @code{G}, in which | |
1595 | case the size is, respectively, in bytes, kilobytes, megabytes | |
1596 | or gigabytes. This is different from @code{pthread_attr_setstacksize} | |
1597 | which gets the number of bytes as an argument. If the stack size cannot | |
1598 | be set due to system constraints, an error is reported and the initial | |
1599 | stack size is left unchanged. If undefined, the stack size is system | |
1600 | dependent. | |
83fd6c5b | 1601 | |
72832460 | 1602 | @item @emph{Reference}: |
1a6d1d24 | 1603 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.7 |
3721b9e1 DF |
1604 | @end table |
1605 | ||
1606 | ||
1607 | ||
1608 | @node OMP_SCHEDULE | |
1609 | @section @env{OMP_SCHEDULE} -- How threads are scheduled | |
1610 | @cindex Environment Variable | |
14734fc7 | 1611 | @cindex Implementation specific setting |
3721b9e1 DF |
1612 | @table @asis |
1613 | @item @emph{Description}: | |
1614 | Allows to specify @code{schedule type} and @code{chunk size}. | |
1615 | The value of the variable shall have the form: @code{type[,chunk]} where | |
5c6ed53a | 1616 | @code{type} is one of @code{static}, @code{dynamic}, @code{guided} or @code{auto} |
83fd6c5b | 1617 | The optional @code{chunk} size shall be a positive integer. If undefined, |
7c2b7f45 | 1618 | dynamic scheduling and a chunk size of 1 is used. |
3721b9e1 | 1619 | |
5c6ed53a TB |
1620 | @item @emph{See also}: |
1621 | @ref{omp_set_schedule} | |
1622 | ||
1623 | @item @emph{Reference}: | |
1a6d1d24 | 1624 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Sections 2.7.1.1 and 4.1 |
5c6ed53a TB |
1625 | @end table |
1626 | ||
1627 | ||
1628 | ||
5c6ed53a | 1629 | @node OMP_THREAD_LIMIT |
6a2ba183 | 1630 | @section @env{OMP_THREAD_LIMIT} -- Set the maximum number of threads |
5c6ed53a TB |
1631 | @cindex Environment Variable |
1632 | @table @asis | |
1633 | @item @emph{Description}: | |
83fd6c5b TB |
1634 | Specifies the number of threads to use for the whole program. The |
1635 | value of this variable shall be a positive integer. If undefined, | |
5c6ed53a TB |
1636 | the number of threads is not limited. |
1637 | ||
1638 | @item @emph{See also}: | |
83fd6c5b | 1639 | @ref{OMP_NUM_THREADS}, @ref{omp_get_thread_limit} |
5c6ed53a TB |
1640 | |
1641 | @item @emph{Reference}: | |
1a6d1d24 | 1642 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.10 |
5c6ed53a TB |
1643 | @end table |
1644 | ||
1645 | ||
1646 | ||
1647 | @node OMP_WAIT_POLICY | |
1648 | @section @env{OMP_WAIT_POLICY} -- How waiting threads are handled | |
1649 | @cindex Environment Variable | |
1650 | @table @asis | |
1651 | @item @emph{Description}: | |
83fd6c5b | 1652 | Specifies whether waiting threads should be active or passive. If |
5c6ed53a TB |
1653 | the value is @code{PASSIVE}, waiting threads should not consume CPU |
1654 | power while waiting; while the value is @code{ACTIVE} specifies that | |
83fd6c5b | 1655 | they should. If undefined, threads wait actively for a short time |
acf0174b JJ |
1656 | before waiting passively. |
1657 | ||
1658 | @item @emph{See also}: | |
1659 | @ref{GOMP_SPINCOUNT} | |
5c6ed53a TB |
1660 | |
1661 | @item @emph{Reference}: | |
1a6d1d24 | 1662 | @uref{https://www.openmp.org, OpenMP specification v4.5}, Section 4.8 |
3721b9e1 DF |
1663 | @end table |
1664 | ||
1665 | ||
1666 | ||
1667 | @node GOMP_CPU_AFFINITY | |
1668 | @section @env{GOMP_CPU_AFFINITY} -- Bind threads to specific CPUs | |
1669 | @cindex Environment Variable | |
1670 | @table @asis | |
1671 | @item @emph{Description}: | |
83fd6c5b TB |
1672 | Binds threads to specific CPUs. The variable should contain a space-separated |
1673 | or comma-separated list of CPUs. This list may contain different kinds of | |
06785a48 | 1674 | entries: either single CPU numbers in any order, a range of CPUs (M-N) |
83fd6c5b | 1675 | or a range with some stride (M-N:S). CPU numbers are zero based. For example, |
06785a48 DF |
1676 | @code{GOMP_CPU_AFFINITY="0 3 1-2 4-15:2"} will bind the initial thread |
1677 | to CPU 0, the second to CPU 3, the third to CPU 1, the fourth to | |
1678 | CPU 2, the fifth to CPU 4, the sixth through tenth to CPUs 6, 8, 10, 12, | |
1679 | and 14 respectively and then start assigning back from the beginning of | |
6a2ba183 | 1680 | the list. @code{GOMP_CPU_AFFINITY=0} binds all threads to CPU 0. |
06785a48 | 1681 | |
f1f3453e | 1682 | There is no libgomp library routine to determine whether a CPU affinity |
83fd6c5b | 1683 | specification is in effect. As a workaround, language-specific library |
06785a48 DF |
1684 | functions, e.g., @code{getenv} in C or @code{GET_ENVIRONMENT_VARIABLE} in |
1685 | Fortran, may be used to query the setting of the @code{GOMP_CPU_AFFINITY} | |
83fd6c5b | 1686 | environment variable. A defined CPU affinity on startup cannot be changed |
06785a48 DF |
1687 | or disabled during the runtime of the application. |
1688 | ||
83fd6c5b TB |
1689 | If both @env{GOMP_CPU_AFFINITY} and @env{OMP_PROC_BIND} are set, |
1690 | @env{OMP_PROC_BIND} has a higher precedence. If neither has been set and | |
1691 | @env{OMP_PROC_BIND} is unset, or when @env{OMP_PROC_BIND} is set to | |
1692 | @code{FALSE}, the host system will handle the assignment of threads to CPUs. | |
20906c66 JJ |
1693 | |
1694 | @item @emph{See also}: | |
83fd6c5b | 1695 | @ref{OMP_PLACES}, @ref{OMP_PROC_BIND} |
3721b9e1 DF |
1696 | @end table |
1697 | ||
1698 | ||
1699 | ||
41dbbb37 TS |
1700 | @node GOMP_DEBUG |
1701 | @section @env{GOMP_DEBUG} -- Enable debugging output | |
1702 | @cindex Environment Variable | |
1703 | @table @asis | |
1704 | @item @emph{Description}: | |
1705 | Enable debugging output. The variable should be set to @code{0} | |
1706 | (disabled, also the default if not set), or @code{1} (enabled). | |
1707 | ||
1708 | If enabled, some debugging output will be printed during execution. | |
1709 | This is currently not specified in more detail, and subject to change. | |
1710 | @end table | |
1711 | ||
1712 | ||
1713 | ||
3721b9e1 DF |
1714 | @node GOMP_STACKSIZE |
1715 | @section @env{GOMP_STACKSIZE} -- Set default thread stack size | |
1716 | @cindex Environment Variable | |
14734fc7 | 1717 | @cindex Implementation specific setting |
3721b9e1 DF |
1718 | @table @asis |
1719 | @item @emph{Description}: | |
83fd6c5b | 1720 | Set the default thread stack size in kilobytes. This is different from |
5c6ed53a | 1721 | @code{pthread_attr_setstacksize} which gets the number of bytes as an |
83fd6c5b TB |
1722 | argument. If the stack size cannot be set due to system constraints, an |
1723 | error is reported and the initial stack size is left unchanged. If undefined, | |
7c2b7f45 | 1724 | the stack size is system dependent. |
3721b9e1 | 1725 | |
5c6ed53a | 1726 | @item @emph{See also}: |
0024f1af | 1727 | @ref{OMP_STACKSIZE} |
5c6ed53a | 1728 | |
3721b9e1 | 1729 | @item @emph{Reference}: |
c1030b5c | 1730 | @uref{https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00493.html, |
3721b9e1 | 1731 | GCC Patches Mailinglist}, |
c1030b5c | 1732 | @uref{https://gcc.gnu.org/ml/gcc-patches/2006-06/msg00496.html, |
3721b9e1 DF |
1733 | GCC Patches Mailinglist} |
1734 | @end table | |
1735 | ||
1736 | ||
1737 | ||
acf0174b JJ |
1738 | @node GOMP_SPINCOUNT |
1739 | @section @env{GOMP_SPINCOUNT} -- Set the busy-wait spin count | |
1740 | @cindex Environment Variable | |
1741 | @cindex Implementation specific setting | |
1742 | @table @asis | |
1743 | @item @emph{Description}: | |
1744 | Determines how long a threads waits actively with consuming CPU power | |
83fd6c5b | 1745 | before waiting passively without consuming CPU power. The value may be |
acf0174b | 1746 | either @code{INFINITE}, @code{INFINITY} to always wait actively or an |
83fd6c5b | 1747 | integer which gives the number of spins of the busy-wait loop. The |
acf0174b JJ |
1748 | integer may optionally be followed by the following suffixes acting |
1749 | as multiplication factors: @code{k} (kilo, thousand), @code{M} (mega, | |
1750 | million), @code{G} (giga, billion), or @code{T} (tera, trillion). | |
1751 | If undefined, 0 is used when @env{OMP_WAIT_POLICY} is @code{PASSIVE}, | |
1752 | 300,000 is used when @env{OMP_WAIT_POLICY} is undefined and | |
1753 | 30 billion is used when @env{OMP_WAIT_POLICY} is @code{ACTIVE}. | |
1754 | If there are more OpenMP threads than available CPUs, 1000 and 100 | |
1755 | spins are used for @env{OMP_WAIT_POLICY} being @code{ACTIVE} or | |
1756 | undefined, respectively; unless the @env{GOMP_SPINCOUNT} is lower | |
1757 | or @env{OMP_WAIT_POLICY} is @code{PASSIVE}. | |
1758 | ||
1759 | @item @emph{See also}: | |
1760 | @ref{OMP_WAIT_POLICY} | |
1761 | @end table | |
1762 | ||
1763 | ||
1764 | ||
06441dd5 SH |
1765 | @node GOMP_RTEMS_THREAD_POOLS |
1766 | @section @env{GOMP_RTEMS_THREAD_POOLS} -- Set the RTEMS specific thread pools | |
1767 | @cindex Environment Variable | |
1768 | @cindex Implementation specific setting | |
1769 | @table @asis | |
1770 | @item @emph{Description}: | |
1771 | This environment variable is only used on the RTEMS real-time operating system. | |
1772 | It determines the scheduler instance specific thread pools. The format for | |
1773 | @env{GOMP_RTEMS_THREAD_POOLS} is a list of optional | |
1774 | @code{<thread-pool-count>[$<priority>]@@<scheduler-name>} configurations | |
1775 | separated by @code{:} where: | |
1776 | @itemize @bullet | |
1777 | @item @code{<thread-pool-count>} is the thread pool count for this scheduler | |
1778 | instance. | |
1779 | @item @code{$<priority>} is an optional priority for the worker threads of a | |
1780 | thread pool according to @code{pthread_setschedparam}. In case a priority | |
1781 | value is omitted, then a worker thread will inherit the priority of the OpenMP | |
1782 | master thread that created it. The priority of the worker thread is not | |
1783 | changed after creation, even if a new OpenMP master thread using the worker has | |
1784 | a different priority. | |
1785 | @item @code{@@<scheduler-name>} is the scheduler instance name according to the | |
1786 | RTEMS application configuration. | |
1787 | @end itemize | |
1788 | In case no thread pool configuration is specified for a scheduler instance, | |
1789 | then each OpenMP master thread of this scheduler instance will use its own | |
1790 | dynamically allocated thread pool. To limit the worker thread count of the | |
1791 | thread pools, each OpenMP master thread must call @code{omp_set_num_threads}. | |
1792 | @item @emph{Example}: | |
1793 | Lets suppose we have three scheduler instances @code{IO}, @code{WRK0}, and | |
1794 | @code{WRK1} with @env{GOMP_RTEMS_THREAD_POOLS} set to | |
1795 | @code{"1@@WRK0:3$4@@WRK1"}. Then there are no thread pool restrictions for | |
1796 | scheduler instance @code{IO}. In the scheduler instance @code{WRK0} there is | |
1797 | one thread pool available. Since no priority is specified for this scheduler | |
1798 | instance, the worker thread inherits the priority of the OpenMP master thread | |
1799 | that created it. In the scheduler instance @code{WRK1} there are three thread | |
1800 | pools available and their worker threads run at priority four. | |
1801 | @end table | |
1802 | ||
1803 | ||
1804 | ||
cdf6119d JN |
1805 | @c --------------------------------------------------------------------- |
1806 | @c Enabling OpenACC | |
1807 | @c --------------------------------------------------------------------- | |
1808 | ||
1809 | @node Enabling OpenACC | |
1810 | @chapter Enabling OpenACC | |
1811 | ||
1812 | To activate the OpenACC extensions for C/C++ and Fortran, the compile-time | |
1813 | flag @option{-fopenacc} must be specified. This enables the OpenACC directive | |
c1030b5c | 1814 | @code{#pragma acc} in C/C++ and @code{!$acc} directives in free form, |
cdf6119d JN |
1815 | @code{c$acc}, @code{*$acc} and @code{!$acc} directives in fixed form, |
1816 | @code{!$} conditional compilation sentinels in free form and @code{c$}, | |
1817 | @code{*$} and @code{!$} sentinels in fixed form, for Fortran. The flag also | |
1818 | arranges for automatic linking of the OpenACC runtime library | |
1819 | (@ref{OpenACC Runtime Library Routines}). | |
1820 | ||
8d1a1cb1 TB |
1821 | See @uref{https://gcc.gnu.org/wiki/OpenACC} for more information. |
1822 | ||
cdf6119d | 1823 | A complete description of all OpenACC directives accepted may be found in |
9651fbaf | 1824 | the @uref{https://www.openacc.org, OpenACC} Application Programming |
e464fc90 | 1825 | Interface manual, version 2.6. |
cdf6119d | 1826 | |
cdf6119d JN |
1827 | |
1828 | ||
1829 | @c --------------------------------------------------------------------- | |
1830 | @c OpenACC Runtime Library Routines | |
1831 | @c --------------------------------------------------------------------- | |
1832 | ||
1833 | @node OpenACC Runtime Library Routines | |
1834 | @chapter OpenACC Runtime Library Routines | |
1835 | ||
1836 | The runtime routines described here are defined by section 3 of the OpenACC | |
e464fc90 | 1837 | specifications in version 2.6. |
cdf6119d JN |
1838 | They have C linkage, and do not throw exceptions. |
1839 | Generally, they are available only for the host, with the exception of | |
1840 | @code{acc_on_device}, which is available for both the host and the | |
1841 | acceleration device. | |
1842 | ||
1843 | @menu | |
1844 | * acc_get_num_devices:: Get number of devices for the given device | |
1845 | type. | |
1846 | * acc_set_device_type:: Set type of device accelerator to use. | |
1847 | * acc_get_device_type:: Get type of device accelerator to be used. | |
1848 | * acc_set_device_num:: Set device number to use. | |
1849 | * acc_get_device_num:: Get device number to be used. | |
6c84c8bf | 1850 | * acc_get_property:: Get device property. |
cdf6119d JN |
1851 | * acc_async_test:: Tests for completion of a specific asynchronous |
1852 | operation. | |
c1030b5c | 1853 | * acc_async_test_all:: Tests for completion of all asynchronous |
cdf6119d JN |
1854 | operations. |
1855 | * acc_wait:: Wait for completion of a specific asynchronous | |
1856 | operation. | |
c1030b5c | 1857 | * acc_wait_all:: Waits for completion of all asynchronous |
cdf6119d JN |
1858 | operations. |
1859 | * acc_wait_all_async:: Wait for completion of all asynchronous | |
1860 | operations. | |
1861 | * acc_wait_async:: Wait for completion of asynchronous operations. | |
1862 | * acc_init:: Initialize runtime for a specific device type. | |
1863 | * acc_shutdown:: Shuts down the runtime for a specific device | |
1864 | type. | |
1865 | * acc_on_device:: Whether executing on a particular device | |
1866 | * acc_malloc:: Allocate device memory. | |
1867 | * acc_free:: Free device memory. | |
1868 | * acc_copyin:: Allocate device memory and copy host memory to | |
1869 | it. | |
1870 | * acc_present_or_copyin:: If the data is not present on the device, | |
1871 | allocate device memory and copy from host | |
1872 | memory. | |
1873 | * acc_create:: Allocate device memory and map it to host | |
1874 | memory. | |
1875 | * acc_present_or_create:: If the data is not present on the device, | |
1876 | allocate device memory and map it to host | |
1877 | memory. | |
1878 | * acc_copyout:: Copy device memory to host memory. | |
1879 | * acc_delete:: Free device memory. | |
1880 | * acc_update_device:: Update device memory from mapped host memory. | |
1881 | * acc_update_self:: Update host memory from mapped device memory. | |
1882 | * acc_map_data:: Map previously allocated device memory to host | |
1883 | memory. | |
1884 | * acc_unmap_data:: Unmap device memory from host memory. | |
1885 | * acc_deviceptr:: Get device pointer associated with specific | |
1886 | host address. | |
1887 | * acc_hostptr:: Get host pointer associated with specific | |
1888 | device address. | |
93d90219 | 1889 | * acc_is_present:: Indicate whether host variable / array is |
cdf6119d JN |
1890 | present on device. |
1891 | * acc_memcpy_to_device:: Copy host memory to device memory. | |
1892 | * acc_memcpy_from_device:: Copy device memory to host memory. | |
e464fc90 TB |
1893 | * acc_attach:: Let device pointer point to device-pointer target. |
1894 | * acc_detach:: Let device pointer point to host-pointer target. | |
cdf6119d JN |
1895 | |
1896 | API routines for target platforms. | |
1897 | ||
1898 | * acc_get_current_cuda_device:: Get CUDA device handle. | |
1899 | * acc_get_current_cuda_context::Get CUDA context handle. | |
1900 | * acc_get_cuda_stream:: Get CUDA stream handle. | |
1901 | * acc_set_cuda_stream:: Set CUDA stream handle. | |
5fae049d TS |
1902 | |
1903 | API routines for the OpenACC Profiling Interface. | |
1904 | ||
1905 | * acc_prof_register:: Register callbacks. | |
1906 | * acc_prof_unregister:: Unregister callbacks. | |
1907 | * acc_prof_lookup:: Obtain inquiry functions. | |
1908 | * acc_register_library:: Library registration. | |
cdf6119d JN |
1909 | @end menu |
1910 | ||
1911 | ||
1912 | ||
1913 | @node acc_get_num_devices | |
1914 | @section @code{acc_get_num_devices} -- Get number of devices for given device type | |
1915 | @table @asis | |
1916 | @item @emph{Description} | |
1917 | This function returns a value indicating the number of devices available | |
1918 | for the device type specified in @var{devicetype}. | |
1919 | ||
1920 | @item @emph{C/C++}: | |
1921 | @multitable @columnfractions .20 .80 | |
1922 | @item @emph{Prototype}: @tab @code{int acc_get_num_devices(acc_device_t devicetype);} | |
1923 | @end multitable | |
1924 | ||
1925 | @item @emph{Fortran}: | |
1926 | @multitable @columnfractions .20 .80 | |
1927 | @item @emph{Interface}: @tab @code{integer function acc_get_num_devices(devicetype)} | |
1928 | @item @tab @code{integer(kind=acc_device_kind) devicetype} | |
1929 | @end multitable | |
1930 | ||
1931 | @item @emph{Reference}: | |
e464fc90 | 1932 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
1933 | 3.2.1. |
1934 | @end table | |
1935 | ||
1936 | ||
1937 | ||
1938 | @node acc_set_device_type | |
1939 | @section @code{acc_set_device_type} -- Set type of device accelerator to use. | |
1940 | @table @asis | |
1941 | @item @emph{Description} | |
c1030b5c | 1942 | This function indicates to the runtime library which device type, specified |
cdf6119d JN |
1943 | in @var{devicetype}, to use when executing a parallel or kernels region. |
1944 | ||
1945 | @item @emph{C/C++}: | |
1946 | @multitable @columnfractions .20 .80 | |
1947 | @item @emph{Prototype}: @tab @code{acc_set_device_type(acc_device_t devicetype);} | |
1948 | @end multitable | |
1949 | ||
1950 | @item @emph{Fortran}: | |
1951 | @multitable @columnfractions .20 .80 | |
1952 | @item @emph{Interface}: @tab @code{subroutine acc_set_device_type(devicetype)} | |
1953 | @item @tab @code{integer(kind=acc_device_kind) devicetype} | |
1954 | @end multitable | |
1955 | ||
1956 | @item @emph{Reference}: | |
e464fc90 | 1957 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
1958 | 3.2.2. |
1959 | @end table | |
1960 | ||
1961 | ||
1962 | ||
1963 | @node acc_get_device_type | |
1964 | @section @code{acc_get_device_type} -- Get type of device accelerator to be used. | |
1965 | @table @asis | |
1966 | @item @emph{Description} | |
1967 | This function returns what device type will be used when executing a | |
1968 | parallel or kernels region. | |
1969 | ||
1970 | @item @emph{C/C++}: | |
1971 | @multitable @columnfractions .20 .80 | |
1972 | @item @emph{Prototype}: @tab @code{acc_device_t acc_get_device_type(void);} | |
1973 | @end multitable | |
1974 | ||
1975 | @item @emph{Fortran}: | |
1976 | @multitable @columnfractions .20 .80 | |
1977 | @item @emph{Interface}: @tab @code{function acc_get_device_type(void)} | |
1978 | @item @tab @code{integer(kind=acc_device_kind) acc_get_device_type} | |
1979 | @end multitable | |
1980 | ||
1981 | @item @emph{Reference}: | |
e464fc90 | 1982 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
1983 | 3.2.3. |
1984 | @end table | |
1985 | ||
1986 | ||
1987 | ||
1988 | @node acc_set_device_num | |
1989 | @section @code{acc_set_device_num} -- Set device number to use. | |
1990 | @table @asis | |
1991 | @item @emph{Description} | |
1992 | This function will indicate to the runtime which device number, | |
8d1a1cb1 | 1993 | specified by @var{devicenum}, associated with the specified device |
cdf6119d JN |
1994 | type @var{devicetype}. |
1995 | ||
1996 | @item @emph{C/C++}: | |
1997 | @multitable @columnfractions .20 .80 | |
8d1a1cb1 | 1998 | @item @emph{Prototype}: @tab @code{acc_set_device_num(int devicenum, acc_device_t devicetype);} |
cdf6119d JN |
1999 | @end multitable |
2000 | ||
2001 | @item @emph{Fortran}: | |
2002 | @multitable @columnfractions .20 .80 | |
2003 | @item @emph{Interface}: @tab @code{subroutine acc_set_device_num(devicenum, devicetype)} | |
2004 | @item @tab @code{integer devicenum} | |
2005 | @item @tab @code{integer(kind=acc_device_kind) devicetype} | |
2006 | @end multitable | |
2007 | ||
2008 | @item @emph{Reference}: | |
e464fc90 | 2009 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
2010 | 3.2.4. |
2011 | @end table | |
2012 | ||
2013 | ||
2014 | ||
2015 | @node acc_get_device_num | |
2016 | @section @code{acc_get_device_num} -- Get device number to be used. | |
2017 | @table @asis | |
2018 | @item @emph{Description} | |
2019 | This function returns which device number associated with the specified device | |
2020 | type @var{devicetype}, will be used when executing a parallel or kernels | |
2021 | region. | |
2022 | ||
2023 | @item @emph{C/C++}: | |
2024 | @multitable @columnfractions .20 .80 | |
2025 | @item @emph{Prototype}: @tab @code{int acc_get_device_num(acc_device_t devicetype);} | |
2026 | @end multitable | |
2027 | ||
2028 | @item @emph{Fortran}: | |
2029 | @multitable @columnfractions .20 .80 | |
2030 | @item @emph{Interface}: @tab @code{function acc_get_device_num(devicetype)} | |
2031 | @item @tab @code{integer(kind=acc_device_kind) devicetype} | |
2032 | @item @tab @code{integer acc_get_device_num} | |
2033 | @end multitable | |
2034 | ||
2035 | @item @emph{Reference}: | |
e464fc90 | 2036 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
2037 | 3.2.5. |
2038 | @end table | |
2039 | ||
2040 | ||
2041 | ||
6c84c8bf MR |
2042 | @node acc_get_property |
2043 | @section @code{acc_get_property} -- Get device property. | |
2044 | @cindex acc_get_property | |
2045 | @cindex acc_get_property_string | |
2046 | @table @asis | |
2047 | @item @emph{Description} | |
2048 | These routines return the value of the specified @var{property} for the | |
2049 | device being queried according to @var{devicenum} and @var{devicetype}. | |
2050 | Integer-valued and string-valued properties are returned by | |
2051 | @code{acc_get_property} and @code{acc_get_property_string} respectively. | |
2052 | The Fortran @code{acc_get_property_string} subroutine returns the string | |
2053 | retrieved in its fourth argument while the remaining entry points are | |
2054 | functions, which pass the return value as their result. | |
2055 | ||
8d1a1cb1 TB |
2056 | Note for Fortran, only: the OpenACC technical committee corrected and, hence, |
2057 | modified the interface introduced in OpenACC 2.6. The kind-value parameter | |
2058 | @code{acc_device_property} has been renamed to @code{acc_device_property_kind} | |
2059 | for consistency and the return type of the @code{acc_get_property} function is | |
2060 | now a @code{c_size_t} integer instead of a @code{acc_device_property} integer. | |
2061 | The parameter @code{acc_device_property} will continue to be provided, | |
2062 | but might be removed in a future version of GCC. | |
2063 | ||
6c84c8bf MR |
2064 | @item @emph{C/C++}: |
2065 | @multitable @columnfractions .20 .80 | |
2066 | @item @emph{Prototype}: @tab @code{size_t acc_get_property(int devicenum, acc_device_t devicetype, acc_device_property_t property);} | |
2067 | @item @emph{Prototype}: @tab @code{const char *acc_get_property_string(int devicenum, acc_device_t devicetype, acc_device_property_t property);} | |
2068 | @end multitable | |
2069 | ||
2070 | @item @emph{Fortran}: | |
2071 | @multitable @columnfractions .20 .80 | |
2072 | @item @emph{Interface}: @tab @code{function acc_get_property(devicenum, devicetype, property)} | |
2073 | @item @emph{Interface}: @tab @code{subroutine acc_get_property_string(devicenum, devicetype, property, string)} | |
8d1a1cb1 | 2074 | @item @tab @code{use ISO_C_Binding, only: c_size_t} |
6c84c8bf MR |
2075 | @item @tab @code{integer devicenum} |
2076 | @item @tab @code{integer(kind=acc_device_kind) devicetype} | |
8d1a1cb1 TB |
2077 | @item @tab @code{integer(kind=acc_device_property_kind) property} |
2078 | @item @tab @code{integer(kind=c_size_t) acc_get_property} | |
6c84c8bf MR |
2079 | @item @tab @code{character(*) string} |
2080 | @end multitable | |
2081 | ||
2082 | @item @emph{Reference}: | |
2083 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
2084 | 3.2.6. | |
2085 | @end table | |
2086 | ||
2087 | ||
2088 | ||
cdf6119d JN |
2089 | @node acc_async_test |
2090 | @section @code{acc_async_test} -- Test for completion of a specific asynchronous operation. | |
2091 | @table @asis | |
2092 | @item @emph{Description} | |
93d90219 | 2093 | This function tests for completion of the asynchronous operation specified |
cdf6119d JN |
2094 | in @var{arg}. In C/C++, a non-zero value will be returned to indicate |
2095 | the specified asynchronous operation has completed. While Fortran will return | |
93d90219 | 2096 | a @code{true}. If the asynchronous operation has not completed, C/C++ returns |
cdf6119d JN |
2097 | a zero and Fortran returns a @code{false}. |
2098 | ||
2099 | @item @emph{C/C++}: | |
2100 | @multitable @columnfractions .20 .80 | |
2101 | @item @emph{Prototype}: @tab @code{int acc_async_test(int arg);} | |
2102 | @end multitable | |
2103 | ||
2104 | @item @emph{Fortran}: | |
2105 | @multitable @columnfractions .20 .80 | |
2106 | @item @emph{Interface}: @tab @code{function acc_async_test(arg)} | |
2107 | @item @tab @code{integer(kind=acc_handle_kind) arg} | |
2108 | @item @tab @code{logical acc_async_test} | |
2109 | @end multitable | |
2110 | ||
2111 | @item @emph{Reference}: | |
e464fc90 TB |
2112 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2113 | 3.2.9. | |
cdf6119d JN |
2114 | @end table |
2115 | ||
2116 | ||
2117 | ||
2118 | @node acc_async_test_all | |
2119 | @section @code{acc_async_test_all} -- Tests for completion of all asynchronous operations. | |
2120 | @table @asis | |
2121 | @item @emph{Description} | |
93d90219 | 2122 | This function tests for completion of all asynchronous operations. |
cdf6119d JN |
2123 | In C/C++, a non-zero value will be returned to indicate all asynchronous |
2124 | operations have completed. While Fortran will return a @code{true}. If | |
2125 | any asynchronous operation has not completed, C/C++ returns a zero and | |
2126 | Fortran returns a @code{false}. | |
2127 | ||
2128 | @item @emph{C/C++}: | |
2129 | @multitable @columnfractions .20 .80 | |
2130 | @item @emph{Prototype}: @tab @code{int acc_async_test_all(void);} | |
2131 | @end multitable | |
2132 | ||
2133 | @item @emph{Fortran}: | |
2134 | @multitable @columnfractions .20 .80 | |
2135 | @item @emph{Interface}: @tab @code{function acc_async_test()} | |
2136 | @item @tab @code{logical acc_get_device_num} | |
2137 | @end multitable | |
2138 | ||
2139 | @item @emph{Reference}: | |
e464fc90 TB |
2140 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2141 | 3.2.10. | |
cdf6119d JN |
2142 | @end table |
2143 | ||
2144 | ||
2145 | ||
2146 | @node acc_wait | |
2147 | @section @code{acc_wait} -- Wait for completion of a specific asynchronous operation. | |
2148 | @table @asis | |
2149 | @item @emph{Description} | |
2150 | This function waits for completion of the asynchronous operation | |
2151 | specified in @var{arg}. | |
2152 | ||
2153 | @item @emph{C/C++}: | |
2154 | @multitable @columnfractions .20 .80 | |
2155 | @item @emph{Prototype}: @tab @code{acc_wait(arg);} | |
7ce64403 | 2156 | @item @emph{Prototype (OpenACC 1.0 compatibility)}: @tab @code{acc_async_wait(arg);} |
cdf6119d JN |
2157 | @end multitable |
2158 | ||
2159 | @item @emph{Fortran}: | |
2160 | @multitable @columnfractions .20 .80 | |
2161 | @item @emph{Interface}: @tab @code{subroutine acc_wait(arg)} | |
2162 | @item @tab @code{integer(acc_handle_kind) arg} | |
7ce64403 TS |
2163 | @item @emph{Interface (OpenACC 1.0 compatibility)}: @tab @code{subroutine acc_async_wait(arg)} |
2164 | @item @tab @code{integer(acc_handle_kind) arg} | |
cdf6119d JN |
2165 | @end multitable |
2166 | ||
2167 | @item @emph{Reference}: | |
e464fc90 TB |
2168 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2169 | 3.2.11. | |
cdf6119d JN |
2170 | @end table |
2171 | ||
2172 | ||
2173 | ||
2174 | @node acc_wait_all | |
2175 | @section @code{acc_wait_all} -- Waits for completion of all asynchronous operations. | |
2176 | @table @asis | |
2177 | @item @emph{Description} | |
2178 | This function waits for the completion of all asynchronous operations. | |
2179 | ||
2180 | @item @emph{C/C++}: | |
2181 | @multitable @columnfractions .20 .80 | |
2182 | @item @emph{Prototype}: @tab @code{acc_wait_all(void);} | |
7ce64403 | 2183 | @item @emph{Prototype (OpenACC 1.0 compatibility)}: @tab @code{acc_async_wait_all(void);} |
cdf6119d JN |
2184 | @end multitable |
2185 | ||
2186 | @item @emph{Fortran}: | |
2187 | @multitable @columnfractions .20 .80 | |
7ce64403 TS |
2188 | @item @emph{Interface}: @tab @code{subroutine acc_wait_all()} |
2189 | @item @emph{Interface (OpenACC 1.0 compatibility)}: @tab @code{subroutine acc_async_wait_all()} | |
cdf6119d JN |
2190 | @end multitable |
2191 | ||
2192 | @item @emph{Reference}: | |
e464fc90 TB |
2193 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2194 | 3.2.13. | |
cdf6119d JN |
2195 | @end table |
2196 | ||
2197 | ||
2198 | ||
2199 | @node acc_wait_all_async | |
2200 | @section @code{acc_wait_all_async} -- Wait for completion of all asynchronous operations. | |
2201 | @table @asis | |
2202 | @item @emph{Description} | |
2203 | This function enqueues a wait operation on the queue @var{async} for any | |
2204 | and all asynchronous operations that have been previously enqueued on | |
2205 | any queue. | |
2206 | ||
2207 | @item @emph{C/C++}: | |
2208 | @multitable @columnfractions .20 .80 | |
2209 | @item @emph{Prototype}: @tab @code{acc_wait_all_async(int async);} | |
2210 | @end multitable | |
2211 | ||
2212 | @item @emph{Fortran}: | |
2213 | @multitable @columnfractions .20 .80 | |
2214 | @item @emph{Interface}: @tab @code{subroutine acc_wait_all_async(async)} | |
2215 | @item @tab @code{integer(acc_handle_kind) async} | |
2216 | @end multitable | |
2217 | ||
2218 | @item @emph{Reference}: | |
e464fc90 TB |
2219 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2220 | 3.2.14. | |
cdf6119d JN |
2221 | @end table |
2222 | ||
2223 | ||
2224 | ||
2225 | @node acc_wait_async | |
2226 | @section @code{acc_wait_async} -- Wait for completion of asynchronous operations. | |
2227 | @table @asis | |
2228 | @item @emph{Description} | |
2229 | This function enqueues a wait operation on queue @var{async} for any and all | |
2230 | asynchronous operations enqueued on queue @var{arg}. | |
2231 | ||
2232 | @item @emph{C/C++}: | |
2233 | @multitable @columnfractions .20 .80 | |
2234 | @item @emph{Prototype}: @tab @code{acc_wait_async(int arg, int async);} | |
2235 | @end multitable | |
2236 | ||
2237 | @item @emph{Fortran}: | |
2238 | @multitable @columnfractions .20 .80 | |
2239 | @item @emph{Interface}: @tab @code{subroutine acc_wait_async(arg, async)} | |
2240 | @item @tab @code{integer(acc_handle_kind) arg, async} | |
2241 | @end multitable | |
2242 | ||
2243 | @item @emph{Reference}: | |
e464fc90 TB |
2244 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2245 | 3.2.12. | |
cdf6119d JN |
2246 | @end table |
2247 | ||
2248 | ||
2249 | ||
2250 | @node acc_init | |
2251 | @section @code{acc_init} -- Initialize runtime for a specific device type. | |
2252 | @table @asis | |
2253 | @item @emph{Description} | |
2254 | This function initializes the runtime for the device type specified in | |
2255 | @var{devicetype}. | |
2256 | ||
2257 | @item @emph{C/C++}: | |
2258 | @multitable @columnfractions .20 .80 | |
2259 | @item @emph{Prototype}: @tab @code{acc_init(acc_device_t devicetype);} | |
2260 | @end multitable | |
2261 | ||
2262 | @item @emph{Fortran}: | |
2263 | @multitable @columnfractions .20 .80 | |
2264 | @item @emph{Interface}: @tab @code{subroutine acc_init(devicetype)} | |
2265 | @item @tab @code{integer(acc_device_kind) devicetype} | |
2266 | @end multitable | |
2267 | ||
2268 | @item @emph{Reference}: | |
e464fc90 TB |
2269 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2270 | 3.2.7. | |
cdf6119d JN |
2271 | @end table |
2272 | ||
2273 | ||
2274 | ||
2275 | @node acc_shutdown | |
2276 | @section @code{acc_shutdown} -- Shuts down the runtime for a specific device type. | |
2277 | @table @asis | |
2278 | @item @emph{Description} | |
2279 | This function shuts down the runtime for the device type specified in | |
2280 | @var{devicetype}. | |
2281 | ||
2282 | @item @emph{C/C++}: | |
2283 | @multitable @columnfractions .20 .80 | |
2284 | @item @emph{Prototype}: @tab @code{acc_shutdown(acc_device_t devicetype);} | |
2285 | @end multitable | |
2286 | ||
2287 | @item @emph{Fortran}: | |
2288 | @multitable @columnfractions .20 .80 | |
2289 | @item @emph{Interface}: @tab @code{subroutine acc_shutdown(devicetype)} | |
2290 | @item @tab @code{integer(acc_device_kind) devicetype} | |
2291 | @end multitable | |
2292 | ||
2293 | @item @emph{Reference}: | |
e464fc90 TB |
2294 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2295 | 3.2.8. | |
cdf6119d JN |
2296 | @end table |
2297 | ||
2298 | ||
2299 | ||
2300 | @node acc_on_device | |
2301 | @section @code{acc_on_device} -- Whether executing on a particular device | |
2302 | @table @asis | |
2303 | @item @emph{Description}: | |
2304 | This function returns whether the program is executing on a particular | |
2305 | device specified in @var{devicetype}. In C/C++ a non-zero value is | |
93d90219 | 2306 | returned to indicate the device is executing on the specified device type. |
cdf6119d JN |
2307 | In Fortran, @code{true} will be returned. If the program is not executing |
2308 | on the specified device type C/C++ will return a zero, while Fortran will | |
2309 | return @code{false}. | |
2310 | ||
2311 | @item @emph{C/C++}: | |
2312 | @multitable @columnfractions .20 .80 | |
2313 | @item @emph{Prototype}: @tab @code{acc_on_device(acc_device_t devicetype);} | |
2314 | @end multitable | |
2315 | ||
2316 | @item @emph{Fortran}: | |
2317 | @multitable @columnfractions .20 .80 | |
2318 | @item @emph{Interface}: @tab @code{function acc_on_device(devicetype)} | |
2319 | @item @tab @code{integer(acc_device_kind) devicetype} | |
2320 | @item @tab @code{logical acc_on_device} | |
2321 | @end multitable | |
2322 | ||
2323 | ||
2324 | @item @emph{Reference}: | |
e464fc90 TB |
2325 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2326 | 3.2.17. | |
cdf6119d JN |
2327 | @end table |
2328 | ||
2329 | ||
2330 | ||
2331 | @node acc_malloc | |
2332 | @section @code{acc_malloc} -- Allocate device memory. | |
2333 | @table @asis | |
2334 | @item @emph{Description} | |
2335 | This function allocates @var{len} bytes of device memory. It returns | |
2336 | the device address of the allocated memory. | |
2337 | ||
2338 | @item @emph{C/C++}: | |
2339 | @multitable @columnfractions .20 .80 | |
2340 | @item @emph{Prototype}: @tab @code{d_void* acc_malloc(size_t len);} | |
2341 | @end multitable | |
2342 | ||
2343 | @item @emph{Reference}: | |
e464fc90 TB |
2344 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2345 | 3.2.18. | |
cdf6119d JN |
2346 | @end table |
2347 | ||
2348 | ||
2349 | ||
2350 | @node acc_free | |
2351 | @section @code{acc_free} -- Free device memory. | |
2352 | @table @asis | |
2353 | @item @emph{Description} | |
2354 | Free previously allocated device memory at the device address @code{a}. | |
2355 | ||
2356 | @item @emph{C/C++}: | |
2357 | @multitable @columnfractions .20 .80 | |
2358 | @item @emph{Prototype}: @tab @code{acc_free(d_void *a);} | |
2359 | @end multitable | |
2360 | ||
2361 | @item @emph{Reference}: | |
e464fc90 TB |
2362 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2363 | 3.2.19. | |
cdf6119d JN |
2364 | @end table |
2365 | ||
2366 | ||
2367 | ||
2368 | @node acc_copyin | |
2369 | @section @code{acc_copyin} -- Allocate device memory and copy host memory to it. | |
2370 | @table @asis | |
2371 | @item @emph{Description} | |
2372 | In C/C++, this function allocates @var{len} bytes of device memory | |
2373 | and maps it to the specified host address in @var{a}. The device | |
2374 | address of the newly allocated device memory is returned. | |
2375 | ||
2376 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2377 | a contiguous array section. The second form @var{a} specifies a | |
2378 | variable or array element and @var{len} specifies the length in bytes. | |
2379 | ||
2380 | @item @emph{C/C++}: | |
2381 | @multitable @columnfractions .20 .80 | |
2382 | @item @emph{Prototype}: @tab @code{void *acc_copyin(h_void *a, size_t len);} | |
e464fc90 | 2383 | @item @emph{Prototype}: @tab @code{void *acc_copyin_async(h_void *a, size_t len, int async);} |
cdf6119d JN |
2384 | @end multitable |
2385 | ||
2386 | @item @emph{Fortran}: | |
2387 | @multitable @columnfractions .20 .80 | |
2388 | @item @emph{Interface}: @tab @code{subroutine acc_copyin(a)} | |
2389 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2390 | @item @emph{Interface}: @tab @code{subroutine acc_copyin(a, len)} | |
2391 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2392 | @item @tab @code{integer len} | |
e464fc90 TB |
2393 | @item @emph{Interface}: @tab @code{subroutine acc_copyin_async(a, async)} |
2394 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2395 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2396 | @item @emph{Interface}: @tab @code{subroutine acc_copyin_async(a, len, async)} | |
2397 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2398 | @item @tab @code{integer len} | |
2399 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2400 | @end multitable |
2401 | ||
2402 | @item @emph{Reference}: | |
e464fc90 TB |
2403 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2404 | 3.2.20. | |
cdf6119d JN |
2405 | @end table |
2406 | ||
2407 | ||
2408 | ||
2409 | @node acc_present_or_copyin | |
2410 | @section @code{acc_present_or_copyin} -- If the data is not present on the device, allocate device memory and copy from host memory. | |
2411 | @table @asis | |
2412 | @item @emph{Description} | |
c1030b5c | 2413 | This function tests if the host data specified by @var{a} and of length |
cdf6119d JN |
2414 | @var{len} is present or not. If it is not present, then device memory |
2415 | will be allocated and the host memory copied. The device address of | |
2416 | the newly allocated device memory is returned. | |
2417 | ||
2418 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2419 | a contiguous array section. The second form @var{a} specifies a variable or | |
2420 | array element and @var{len} specifies the length in bytes. | |
2421 | ||
e464fc90 TB |
2422 | Note that @code{acc_present_or_copyin} and @code{acc_pcopyin} exist for |
2423 | backward compatibility with OpenACC 2.0; use @ref{acc_copyin} instead. | |
2424 | ||
cdf6119d JN |
2425 | @item @emph{C/C++}: |
2426 | @multitable @columnfractions .20 .80 | |
2427 | @item @emph{Prototype}: @tab @code{void *acc_present_or_copyin(h_void *a, size_t len);} | |
2428 | @item @emph{Prototype}: @tab @code{void *acc_pcopyin(h_void *a, size_t len);} | |
2429 | @end multitable | |
2430 | ||
2431 | @item @emph{Fortran}: | |
2432 | @multitable @columnfractions .20 .80 | |
2433 | @item @emph{Interface}: @tab @code{subroutine acc_present_or_copyin(a)} | |
2434 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2435 | @item @emph{Interface}: @tab @code{subroutine acc_present_or_copyin(a, len)} | |
2436 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2437 | @item @tab @code{integer len} | |
2438 | @item @emph{Interface}: @tab @code{subroutine acc_pcopyin(a)} | |
2439 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2440 | @item @emph{Interface}: @tab @code{subroutine acc_pcopyin(a, len)} | |
2441 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2442 | @item @tab @code{integer len} | |
2443 | @end multitable | |
2444 | ||
2445 | @item @emph{Reference}: | |
e464fc90 TB |
2446 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2447 | 3.2.20. | |
cdf6119d JN |
2448 | @end table |
2449 | ||
2450 | ||
2451 | ||
2452 | @node acc_create | |
2453 | @section @code{acc_create} -- Allocate device memory and map it to host memory. | |
2454 | @table @asis | |
2455 | @item @emph{Description} | |
2456 | This function allocates device memory and maps it to host memory specified | |
2457 | by the host address @var{a} with a length of @var{len} bytes. In C/C++, | |
2458 | the function returns the device address of the allocated device memory. | |
2459 | ||
2460 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2461 | a contiguous array section. The second form @var{a} specifies a variable or | |
2462 | array element and @var{len} specifies the length in bytes. | |
2463 | ||
2464 | @item @emph{C/C++}: | |
2465 | @multitable @columnfractions .20 .80 | |
2466 | @item @emph{Prototype}: @tab @code{void *acc_create(h_void *a, size_t len);} | |
e464fc90 | 2467 | @item @emph{Prototype}: @tab @code{void *acc_create_async(h_void *a, size_t len, int async);} |
cdf6119d JN |
2468 | @end multitable |
2469 | ||
2470 | @item @emph{Fortran}: | |
2471 | @multitable @columnfractions .20 .80 | |
2472 | @item @emph{Interface}: @tab @code{subroutine acc_create(a)} | |
2473 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2474 | @item @emph{Interface}: @tab @code{subroutine acc_create(a, len)} | |
2475 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2476 | @item @tab @code{integer len} | |
e464fc90 TB |
2477 | @item @emph{Interface}: @tab @code{subroutine acc_create_async(a, async)} |
2478 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2479 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2480 | @item @emph{Interface}: @tab @code{subroutine acc_create_async(a, len, async)} | |
2481 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2482 | @item @tab @code{integer len} | |
2483 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2484 | @end multitable |
2485 | ||
2486 | @item @emph{Reference}: | |
e464fc90 TB |
2487 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2488 | 3.2.21. | |
cdf6119d JN |
2489 | @end table |
2490 | ||
2491 | ||
2492 | ||
2493 | @node acc_present_or_create | |
2494 | @section @code{acc_present_or_create} -- If the data is not present on the device, allocate device memory and map it to host memory. | |
2495 | @table @asis | |
2496 | @item @emph{Description} | |
c1030b5c | 2497 | This function tests if the host data specified by @var{a} and of length |
cdf6119d JN |
2498 | @var{len} is present or not. If it is not present, then device memory |
2499 | will be allocated and mapped to host memory. In C/C++, the device address | |
2500 | of the newly allocated device memory is returned. | |
2501 | ||
2502 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2503 | a contiguous array section. The second form @var{a} specifies a variable or | |
2504 | array element and @var{len} specifies the length in bytes. | |
2505 | ||
e464fc90 TB |
2506 | Note that @code{acc_present_or_create} and @code{acc_pcreate} exist for |
2507 | backward compatibility with OpenACC 2.0; use @ref{acc_create} instead. | |
cdf6119d JN |
2508 | |
2509 | @item @emph{C/C++}: | |
2510 | @multitable @columnfractions .20 .80 | |
2511 | @item @emph{Prototype}: @tab @code{void *acc_present_or_create(h_void *a, size_t len)} | |
2512 | @item @emph{Prototype}: @tab @code{void *acc_pcreate(h_void *a, size_t len)} | |
2513 | @end multitable | |
2514 | ||
2515 | @item @emph{Fortran}: | |
2516 | @multitable @columnfractions .20 .80 | |
2517 | @item @emph{Interface}: @tab @code{subroutine acc_present_or_create(a)} | |
2518 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2519 | @item @emph{Interface}: @tab @code{subroutine acc_present_or_create(a, len)} | |
2520 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2521 | @item @tab @code{integer len} | |
2522 | @item @emph{Interface}: @tab @code{subroutine acc_pcreate(a)} | |
2523 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2524 | @item @emph{Interface}: @tab @code{subroutine acc_pcreate(a, len)} | |
2525 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2526 | @item @tab @code{integer len} | |
2527 | @end multitable | |
2528 | ||
2529 | @item @emph{Reference}: | |
e464fc90 TB |
2530 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2531 | 3.2.21. | |
cdf6119d JN |
2532 | @end table |
2533 | ||
2534 | ||
2535 | ||
2536 | @node acc_copyout | |
2537 | @section @code{acc_copyout} -- Copy device memory to host memory. | |
2538 | @table @asis | |
2539 | @item @emph{Description} | |
2540 | This function copies mapped device memory to host memory which is specified | |
2541 | by host address @var{a} for a length @var{len} bytes in C/C++. | |
2542 | ||
2543 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2544 | a contiguous array section. The second form @var{a} specifies a variable or | |
2545 | array element and @var{len} specifies the length in bytes. | |
2546 | ||
2547 | @item @emph{C/C++}: | |
2548 | @multitable @columnfractions .20 .80 | |
2549 | @item @emph{Prototype}: @tab @code{acc_copyout(h_void *a, size_t len);} | |
e464fc90 TB |
2550 | @item @emph{Prototype}: @tab @code{acc_copyout_async(h_void *a, size_t len, int async);} |
2551 | @item @emph{Prototype}: @tab @code{acc_copyout_finalize(h_void *a, size_t len);} | |
2552 | @item @emph{Prototype}: @tab @code{acc_copyout_finalize_async(h_void *a, size_t len, int async);} | |
cdf6119d JN |
2553 | @end multitable |
2554 | ||
2555 | @item @emph{Fortran}: | |
2556 | @multitable @columnfractions .20 .80 | |
2557 | @item @emph{Interface}: @tab @code{subroutine acc_copyout(a)} | |
2558 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2559 | @item @emph{Interface}: @tab @code{subroutine acc_copyout(a, len)} | |
2560 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2561 | @item @tab @code{integer len} | |
e464fc90 TB |
2562 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_async(a, async)} |
2563 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2564 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2565 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_async(a, len, async)} | |
2566 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2567 | @item @tab @code{integer len} | |
2568 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2569 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize(a)} | |
2570 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2571 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize(a, len)} | |
2572 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2573 | @item @tab @code{integer len} | |
2574 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize_async(a, async)} | |
2575 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2576 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2577 | @item @emph{Interface}: @tab @code{subroutine acc_copyout_finalize_async(a, len, async)} | |
2578 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2579 | @item @tab @code{integer len} | |
2580 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2581 | @end multitable |
2582 | ||
2583 | @item @emph{Reference}: | |
e464fc90 TB |
2584 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2585 | 3.2.22. | |
cdf6119d JN |
2586 | @end table |
2587 | ||
2588 | ||
2589 | ||
2590 | @node acc_delete | |
2591 | @section @code{acc_delete} -- Free device memory. | |
2592 | @table @asis | |
2593 | @item @emph{Description} | |
2594 | This function frees previously allocated device memory specified by | |
2595 | the device address @var{a} and the length of @var{len} bytes. | |
2596 | ||
2597 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2598 | a contiguous array section. The second form @var{a} specifies a variable or | |
2599 | array element and @var{len} specifies the length in bytes. | |
2600 | ||
2601 | @item @emph{C/C++}: | |
2602 | @multitable @columnfractions .20 .80 | |
2603 | @item @emph{Prototype}: @tab @code{acc_delete(h_void *a, size_t len);} | |
e464fc90 TB |
2604 | @item @emph{Prototype}: @tab @code{acc_delete_async(h_void *a, size_t len, int async);} |
2605 | @item @emph{Prototype}: @tab @code{acc_delete_finalize(h_void *a, size_t len);} | |
2606 | @item @emph{Prototype}: @tab @code{acc_delete_finalize_async(h_void *a, size_t len, int async);} | |
cdf6119d JN |
2607 | @end multitable |
2608 | ||
2609 | @item @emph{Fortran}: | |
2610 | @multitable @columnfractions .20 .80 | |
2611 | @item @emph{Interface}: @tab @code{subroutine acc_delete(a)} | |
2612 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2613 | @item @emph{Interface}: @tab @code{subroutine acc_delete(a, len)} | |
2614 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2615 | @item @tab @code{integer len} | |
e464fc90 TB |
2616 | @item @emph{Interface}: @tab @code{subroutine acc_delete_async(a, async)} |
2617 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2618 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2619 | @item @emph{Interface}: @tab @code{subroutine acc_delete_async(a, len, async)} | |
2620 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2621 | @item @tab @code{integer len} | |
2622 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2623 | @item @emph{Interface}: @tab @code{subroutine acc_delete_finalize(a)} | |
2624 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2625 | @item @emph{Interface}: @tab @code{subroutine acc_delete_finalize(a, len)} | |
2626 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2627 | @item @tab @code{integer len} | |
2628 | @item @emph{Interface}: @tab @code{subroutine acc_delete_async_finalize(a, async)} | |
2629 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2630 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2631 | @item @emph{Interface}: @tab @code{subroutine acc_delete_async_finalize(a, len, async)} | |
2632 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2633 | @item @tab @code{integer len} | |
2634 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2635 | @end multitable |
2636 | ||
2637 | @item @emph{Reference}: | |
e464fc90 TB |
2638 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2639 | 3.2.23. | |
cdf6119d JN |
2640 | @end table |
2641 | ||
2642 | ||
2643 | ||
2644 | @node acc_update_device | |
2645 | @section @code{acc_update_device} -- Update device memory from mapped host memory. | |
2646 | @table @asis | |
2647 | @item @emph{Description} | |
2648 | This function updates the device copy from the previously mapped host memory. | |
2649 | The host memory is specified with the host address @var{a} and a length of | |
2650 | @var{len} bytes. | |
2651 | ||
2652 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2653 | a contiguous array section. The second form @var{a} specifies a variable or | |
2654 | array element and @var{len} specifies the length in bytes. | |
2655 | ||
2656 | @item @emph{C/C++}: | |
2657 | @multitable @columnfractions .20 .80 | |
2658 | @item @emph{Prototype}: @tab @code{acc_update_device(h_void *a, size_t len);} | |
e464fc90 | 2659 | @item @emph{Prototype}: @tab @code{acc_update_device(h_void *a, size_t len, async);} |
cdf6119d JN |
2660 | @end multitable |
2661 | ||
2662 | @item @emph{Fortran}: | |
2663 | @multitable @columnfractions .20 .80 | |
2664 | @item @emph{Interface}: @tab @code{subroutine acc_update_device(a)} | |
2665 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2666 | @item @emph{Interface}: @tab @code{subroutine acc_update_device(a, len)} | |
2667 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2668 | @item @tab @code{integer len} | |
e464fc90 TB |
2669 | @item @emph{Interface}: @tab @code{subroutine acc_update_device_async(a, async)} |
2670 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2671 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2672 | @item @emph{Interface}: @tab @code{subroutine acc_update_device_async(a, len, async)} | |
2673 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2674 | @item @tab @code{integer len} | |
2675 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2676 | @end multitable |
2677 | ||
2678 | @item @emph{Reference}: | |
e464fc90 TB |
2679 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2680 | 3.2.24. | |
cdf6119d JN |
2681 | @end table |
2682 | ||
2683 | ||
2684 | ||
2685 | @node acc_update_self | |
2686 | @section @code{acc_update_self} -- Update host memory from mapped device memory. | |
2687 | @table @asis | |
2688 | @item @emph{Description} | |
2689 | This function updates the host copy from the previously mapped device memory. | |
2690 | The host memory is specified with the host address @var{a} and a length of | |
2691 | @var{len} bytes. | |
2692 | ||
2693 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2694 | a contiguous array section. The second form @var{a} specifies a variable or | |
2695 | array element and @var{len} specifies the length in bytes. | |
2696 | ||
2697 | @item @emph{C/C++}: | |
2698 | @multitable @columnfractions .20 .80 | |
2699 | @item @emph{Prototype}: @tab @code{acc_update_self(h_void *a, size_t len);} | |
e464fc90 | 2700 | @item @emph{Prototype}: @tab @code{acc_update_self_async(h_void *a, size_t len, int async);} |
cdf6119d JN |
2701 | @end multitable |
2702 | ||
2703 | @item @emph{Fortran}: | |
2704 | @multitable @columnfractions .20 .80 | |
2705 | @item @emph{Interface}: @tab @code{subroutine acc_update_self(a)} | |
2706 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2707 | @item @emph{Interface}: @tab @code{subroutine acc_update_self(a, len)} | |
2708 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2709 | @item @tab @code{integer len} | |
e464fc90 TB |
2710 | @item @emph{Interface}: @tab @code{subroutine acc_update_self_async(a, async)} |
2711 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2712 | @item @tab @code{integer(acc_handle_kind) :: async} | |
2713 | @item @emph{Interface}: @tab @code{subroutine acc_update_self_async(a, len, async)} | |
2714 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2715 | @item @tab @code{integer len} | |
2716 | @item @tab @code{integer(acc_handle_kind) :: async} | |
cdf6119d JN |
2717 | @end multitable |
2718 | ||
2719 | @item @emph{Reference}: | |
e464fc90 TB |
2720 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2721 | 3.2.25. | |
cdf6119d JN |
2722 | @end table |
2723 | ||
2724 | ||
2725 | ||
2726 | @node acc_map_data | |
2727 | @section @code{acc_map_data} -- Map previously allocated device memory to host memory. | |
2728 | @table @asis | |
2729 | @item @emph{Description} | |
2730 | This function maps previously allocated device and host memory. The device | |
2731 | memory is specified with the device address @var{d}. The host memory is | |
2732 | specified with the host address @var{h} and a length of @var{len}. | |
2733 | ||
2734 | @item @emph{C/C++}: | |
2735 | @multitable @columnfractions .20 .80 | |
2736 | @item @emph{Prototype}: @tab @code{acc_map_data(h_void *h, d_void *d, size_t len);} | |
2737 | @end multitable | |
2738 | ||
2739 | @item @emph{Reference}: | |
e464fc90 TB |
2740 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2741 | 3.2.26. | |
cdf6119d JN |
2742 | @end table |
2743 | ||
2744 | ||
2745 | ||
2746 | @node acc_unmap_data | |
2747 | @section @code{acc_unmap_data} -- Unmap device memory from host memory. | |
2748 | @table @asis | |
2749 | @item @emph{Description} | |
2750 | This function unmaps previously mapped device and host memory. The latter | |
2751 | specified by @var{h}. | |
2752 | ||
2753 | @item @emph{C/C++}: | |
2754 | @multitable @columnfractions .20 .80 | |
2755 | @item @emph{Prototype}: @tab @code{acc_unmap_data(h_void *h);} | |
2756 | @end multitable | |
2757 | ||
2758 | @item @emph{Reference}: | |
e464fc90 TB |
2759 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2760 | 3.2.27. | |
cdf6119d JN |
2761 | @end table |
2762 | ||
2763 | ||
2764 | ||
2765 | @node acc_deviceptr | |
2766 | @section @code{acc_deviceptr} -- Get device pointer associated with specific host address. | |
2767 | @table @asis | |
2768 | @item @emph{Description} | |
2769 | This function returns the device address that has been mapped to the | |
2770 | host address specified by @var{h}. | |
2771 | ||
2772 | @item @emph{C/C++}: | |
2773 | @multitable @columnfractions .20 .80 | |
2774 | @item @emph{Prototype}: @tab @code{void *acc_deviceptr(h_void *h);} | |
2775 | @end multitable | |
2776 | ||
2777 | @item @emph{Reference}: | |
e464fc90 TB |
2778 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2779 | 3.2.28. | |
cdf6119d JN |
2780 | @end table |
2781 | ||
2782 | ||
2783 | ||
2784 | @node acc_hostptr | |
2785 | @section @code{acc_hostptr} -- Get host pointer associated with specific device address. | |
2786 | @table @asis | |
2787 | @item @emph{Description} | |
2788 | This function returns the host address that has been mapped to the | |
2789 | device address specified by @var{d}. | |
2790 | ||
2791 | @item @emph{C/C++}: | |
2792 | @multitable @columnfractions .20 .80 | |
2793 | @item @emph{Prototype}: @tab @code{void *acc_hostptr(d_void *d);} | |
2794 | @end multitable | |
2795 | ||
2796 | @item @emph{Reference}: | |
e464fc90 TB |
2797 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2798 | 3.2.29. | |
cdf6119d JN |
2799 | @end table |
2800 | ||
2801 | ||
2802 | ||
2803 | @node acc_is_present | |
2804 | @section @code{acc_is_present} -- Indicate whether host variable / array is present on device. | |
2805 | @table @asis | |
2806 | @item @emph{Description} | |
2807 | This function indicates whether the specified host address in @var{a} and a | |
2808 | length of @var{len} bytes is present on the device. In C/C++, a non-zero | |
2809 | value is returned to indicate the presence of the mapped memory on the | |
2810 | device. A zero is returned to indicate the memory is not mapped on the | |
2811 | device. | |
2812 | ||
2813 | In Fortran, two (2) forms are supported. In the first form, @var{a} specifies | |
2814 | a contiguous array section. The second form @var{a} specifies a variable or | |
2815 | array element and @var{len} specifies the length in bytes. If the host | |
2816 | memory is mapped to device memory, then a @code{true} is returned. Otherwise, | |
2817 | a @code{false} is return to indicate the mapped memory is not present. | |
2818 | ||
2819 | @item @emph{C/C++}: | |
2820 | @multitable @columnfractions .20 .80 | |
2821 | @item @emph{Prototype}: @tab @code{int acc_is_present(h_void *a, size_t len);} | |
2822 | @end multitable | |
2823 | ||
2824 | @item @emph{Fortran}: | |
2825 | @multitable @columnfractions .20 .80 | |
2826 | @item @emph{Interface}: @tab @code{function acc_is_present(a)} | |
2827 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2828 | @item @tab @code{logical acc_is_present} | |
2829 | @item @emph{Interface}: @tab @code{function acc_is_present(a, len)} | |
2830 | @item @tab @code{type, dimension(:[,:]...) :: a} | |
2831 | @item @tab @code{integer len} | |
2832 | @item @tab @code{logical acc_is_present} | |
2833 | @end multitable | |
2834 | ||
2835 | @item @emph{Reference}: | |
e464fc90 TB |
2836 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2837 | 3.2.30. | |
cdf6119d JN |
2838 | @end table |
2839 | ||
2840 | ||
2841 | ||
2842 | @node acc_memcpy_to_device | |
2843 | @section @code{acc_memcpy_to_device} -- Copy host memory to device memory. | |
2844 | @table @asis | |
2845 | @item @emph{Description} | |
2846 | This function copies host memory specified by host address of @var{src} to | |
2847 | device memory specified by the device address @var{dest} for a length of | |
2848 | @var{bytes} bytes. | |
2849 | ||
2850 | @item @emph{C/C++}: | |
2851 | @multitable @columnfractions .20 .80 | |
2852 | @item @emph{Prototype}: @tab @code{acc_memcpy_to_device(d_void *dest, h_void *src, size_t bytes);} | |
2853 | @end multitable | |
2854 | ||
2855 | @item @emph{Reference}: | |
e464fc90 TB |
2856 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2857 | 3.2.31. | |
cdf6119d JN |
2858 | @end table |
2859 | ||
2860 | ||
2861 | ||
2862 | @node acc_memcpy_from_device | |
2863 | @section @code{acc_memcpy_from_device} -- Copy device memory to host memory. | |
2864 | @table @asis | |
2865 | @item @emph{Description} | |
2866 | This function copies host memory specified by host address of @var{src} from | |
2867 | device memory specified by the device address @var{dest} for a length of | |
2868 | @var{bytes} bytes. | |
2869 | ||
2870 | @item @emph{C/C++}: | |
2871 | @multitable @columnfractions .20 .80 | |
2872 | @item @emph{Prototype}: @tab @code{acc_memcpy_from_device(d_void *dest, h_void *src, size_t bytes);} | |
2873 | @end multitable | |
2874 | ||
2875 | @item @emph{Reference}: | |
e464fc90 TB |
2876 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
2877 | 3.2.32. | |
2878 | @end table | |
2879 | ||
2880 | ||
2881 | ||
2882 | @node acc_attach | |
2883 | @section @code{acc_attach} -- Let device pointer point to device-pointer target. | |
2884 | @table @asis | |
2885 | @item @emph{Description} | |
2886 | This function updates a pointer on the device from pointing to a host-pointer | |
2887 | address to pointing to the corresponding device data. | |
2888 | ||
2889 | @item @emph{C/C++}: | |
2890 | @multitable @columnfractions .20 .80 | |
2891 | @item @emph{Prototype}: @tab @code{acc_attach(h_void **ptr);} | |
2892 | @item @emph{Prototype}: @tab @code{acc_attach_async(h_void **ptr, int async);} | |
2893 | @end multitable | |
2894 | ||
2895 | @item @emph{Reference}: | |
2896 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
2897 | 3.2.34. | |
2898 | @end table | |
2899 | ||
2900 | ||
2901 | ||
2902 | @node acc_detach | |
2903 | @section @code{acc_detach} -- Let device pointer point to host-pointer target. | |
2904 | @table @asis | |
2905 | @item @emph{Description} | |
2906 | This function updates a pointer on the device from pointing to a device-pointer | |
2907 | address to pointing to the corresponding host data. | |
2908 | ||
2909 | @item @emph{C/C++}: | |
2910 | @multitable @columnfractions .20 .80 | |
2911 | @item @emph{Prototype}: @tab @code{acc_detach(h_void **ptr);} | |
2912 | @item @emph{Prototype}: @tab @code{acc_detach_async(h_void **ptr, int async);} | |
2913 | @item @emph{Prototype}: @tab @code{acc_detach_finalize(h_void **ptr);} | |
2914 | @item @emph{Prototype}: @tab @code{acc_detach_finalize_async(h_void **ptr, int async);} | |
2915 | @end multitable | |
2916 | ||
2917 | @item @emph{Reference}: | |
2918 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
2919 | 3.2.35. | |
cdf6119d JN |
2920 | @end table |
2921 | ||
2922 | ||
2923 | ||
2924 | @node acc_get_current_cuda_device | |
2925 | @section @code{acc_get_current_cuda_device} -- Get CUDA device handle. | |
2926 | @table @asis | |
2927 | @item @emph{Description} | |
2928 | This function returns the CUDA device handle. This handle is the same | |
2929 | as used by the CUDA Runtime or Driver API's. | |
2930 | ||
2931 | @item @emph{C/C++}: | |
2932 | @multitable @columnfractions .20 .80 | |
2933 | @item @emph{Prototype}: @tab @code{void *acc_get_current_cuda_device(void);} | |
2934 | @end multitable | |
2935 | ||
2936 | @item @emph{Reference}: | |
e464fc90 | 2937 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
2938 | A.2.1.1. |
2939 | @end table | |
2940 | ||
2941 | ||
2942 | ||
2943 | @node acc_get_current_cuda_context | |
2944 | @section @code{acc_get_current_cuda_context} -- Get CUDA context handle. | |
2945 | @table @asis | |
2946 | @item @emph{Description} | |
2947 | This function returns the CUDA context handle. This handle is the same | |
2948 | as used by the CUDA Runtime or Driver API's. | |
2949 | ||
2950 | @item @emph{C/C++}: | |
2951 | @multitable @columnfractions .20 .80 | |
18c247cc | 2952 | @item @emph{Prototype}: @tab @code{void *acc_get_current_cuda_context(void);} |
cdf6119d JN |
2953 | @end multitable |
2954 | ||
2955 | @item @emph{Reference}: | |
e464fc90 | 2956 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
2957 | A.2.1.2. |
2958 | @end table | |
2959 | ||
2960 | ||
2961 | ||
2962 | @node acc_get_cuda_stream | |
2963 | @section @code{acc_get_cuda_stream} -- Get CUDA stream handle. | |
2964 | @table @asis | |
2965 | @item @emph{Description} | |
18c247cc TS |
2966 | This function returns the CUDA stream handle for the queue @var{async}. |
2967 | This handle is the same as used by the CUDA Runtime or Driver API's. | |
cdf6119d JN |
2968 | |
2969 | @item @emph{C/C++}: | |
2970 | @multitable @columnfractions .20 .80 | |
18c247cc | 2971 | @item @emph{Prototype}: @tab @code{void *acc_get_cuda_stream(int async);} |
cdf6119d JN |
2972 | @end multitable |
2973 | ||
2974 | @item @emph{Reference}: | |
e464fc90 | 2975 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
2976 | A.2.1.3. |
2977 | @end table | |
2978 | ||
2979 | ||
2980 | ||
2981 | @node acc_set_cuda_stream | |
2982 | @section @code{acc_set_cuda_stream} -- Set CUDA stream handle. | |
2983 | @table @asis | |
2984 | @item @emph{Description} | |
2985 | This function associates the stream handle specified by @var{stream} with | |
18c247cc TS |
2986 | the queue @var{async}. |
2987 | ||
2988 | This cannot be used to change the stream handle associated with | |
2989 | @code{acc_async_sync}. | |
2990 | ||
2991 | The return value is not specified. | |
cdf6119d JN |
2992 | |
2993 | @item @emph{C/C++}: | |
2994 | @multitable @columnfractions .20 .80 | |
18c247cc | 2995 | @item @emph{Prototype}: @tab @code{int acc_set_cuda_stream(int async, void *stream);} |
cdf6119d JN |
2996 | @end multitable |
2997 | ||
2998 | @item @emph{Reference}: | |
e464fc90 | 2999 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
3000 | A.2.1.4. |
3001 | @end table | |
3002 | ||
3003 | ||
3004 | ||
5fae049d TS |
3005 | @node acc_prof_register |
3006 | @section @code{acc_prof_register} -- Register callbacks. | |
3007 | @table @asis | |
3008 | @item @emph{Description}: | |
3009 | This function registers callbacks. | |
3010 | ||
3011 | @item @emph{C/C++}: | |
3012 | @multitable @columnfractions .20 .80 | |
3013 | @item @emph{Prototype}: @tab @code{void acc_prof_register (acc_event_t, acc_prof_callback, acc_register_t);} | |
3014 | @end multitable | |
3015 | ||
3016 | @item @emph{See also}: | |
3017 | @ref{OpenACC Profiling Interface} | |
3018 | ||
3019 | @item @emph{Reference}: | |
3020 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
3021 | 5.3. | |
3022 | @end table | |
3023 | ||
3024 | ||
3025 | ||
3026 | @node acc_prof_unregister | |
3027 | @section @code{acc_prof_unregister} -- Unregister callbacks. | |
3028 | @table @asis | |
3029 | @item @emph{Description}: | |
3030 | This function unregisters callbacks. | |
3031 | ||
3032 | @item @emph{C/C++}: | |
3033 | @multitable @columnfractions .20 .80 | |
3034 | @item @emph{Prototype}: @tab @code{void acc_prof_unregister (acc_event_t, acc_prof_callback, acc_register_t);} | |
3035 | @end multitable | |
3036 | ||
3037 | @item @emph{See also}: | |
3038 | @ref{OpenACC Profiling Interface} | |
3039 | ||
3040 | @item @emph{Reference}: | |
3041 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
3042 | 5.3. | |
3043 | @end table | |
3044 | ||
3045 | ||
3046 | ||
3047 | @node acc_prof_lookup | |
3048 | @section @code{acc_prof_lookup} -- Obtain inquiry functions. | |
3049 | @table @asis | |
3050 | @item @emph{Description}: | |
3051 | Function to obtain inquiry functions. | |
3052 | ||
3053 | @item @emph{C/C++}: | |
3054 | @multitable @columnfractions .20 .80 | |
3055 | @item @emph{Prototype}: @tab @code{acc_query_fn acc_prof_lookup (const char *);} | |
3056 | @end multitable | |
3057 | ||
3058 | @item @emph{See also}: | |
3059 | @ref{OpenACC Profiling Interface} | |
3060 | ||
3061 | @item @emph{Reference}: | |
3062 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
3063 | 5.3. | |
3064 | @end table | |
3065 | ||
3066 | ||
3067 | ||
3068 | @node acc_register_library | |
3069 | @section @code{acc_register_library} -- Library registration. | |
3070 | @table @asis | |
3071 | @item @emph{Description}: | |
3072 | Function for library registration. | |
3073 | ||
3074 | @item @emph{C/C++}: | |
3075 | @multitable @columnfractions .20 .80 | |
3076 | @item @emph{Prototype}: @tab @code{void acc_register_library (acc_prof_reg, acc_prof_reg, acc_prof_lookup_func);} | |
3077 | @end multitable | |
3078 | ||
3079 | @item @emph{See also}: | |
3080 | @ref{OpenACC Profiling Interface}, @ref{ACC_PROFLIB} | |
3081 | ||
3082 | @item @emph{Reference}: | |
3083 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
3084 | 5.3. | |
3085 | @end table | |
3086 | ||
3087 | ||
3088 | ||
cdf6119d JN |
3089 | @c --------------------------------------------------------------------- |
3090 | @c OpenACC Environment Variables | |
3091 | @c --------------------------------------------------------------------- | |
3092 | ||
3093 | @node OpenACC Environment Variables | |
3094 | @chapter OpenACC Environment Variables | |
3095 | ||
3096 | The variables @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM} | |
3097 | are defined by section 4 of the OpenACC specification in version 2.0. | |
5fae049d TS |
3098 | The variable @env{ACC_PROFLIB} |
3099 | is defined by section 4 of the OpenACC specification in version 2.6. | |
cdf6119d JN |
3100 | The variable @env{GCC_ACC_NOTIFY} is used for diagnostic purposes. |
3101 | ||
3102 | @menu | |
3103 | * ACC_DEVICE_TYPE:: | |
3104 | * ACC_DEVICE_NUM:: | |
5fae049d | 3105 | * ACC_PROFLIB:: |
cdf6119d JN |
3106 | * GCC_ACC_NOTIFY:: |
3107 | @end menu | |
3108 | ||
3109 | ||
3110 | ||
3111 | @node ACC_DEVICE_TYPE | |
3112 | @section @code{ACC_DEVICE_TYPE} | |
3113 | @table @asis | |
3114 | @item @emph{Reference}: | |
e464fc90 | 3115 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
3116 | 4.1. |
3117 | @end table | |
3118 | ||
3119 | ||
3120 | ||
3121 | @node ACC_DEVICE_NUM | |
3122 | @section @code{ACC_DEVICE_NUM} | |
3123 | @table @asis | |
3124 | @item @emph{Reference}: | |
e464fc90 | 3125 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section |
cdf6119d JN |
3126 | 4.2. |
3127 | @end table | |
3128 | ||
3129 | ||
3130 | ||
5fae049d TS |
3131 | @node ACC_PROFLIB |
3132 | @section @code{ACC_PROFLIB} | |
3133 | @table @asis | |
3134 | @item @emph{See also}: | |
3135 | @ref{acc_register_library}, @ref{OpenACC Profiling Interface} | |
3136 | ||
3137 | @item @emph{Reference}: | |
3138 | @uref{https://www.openacc.org, OpenACC specification v2.6}, section | |
3139 | 4.3. | |
3140 | @end table | |
3141 | ||
3142 | ||
3143 | ||
cdf6119d JN |
3144 | @node GCC_ACC_NOTIFY |
3145 | @section @code{GCC_ACC_NOTIFY} | |
3146 | @table @asis | |
3147 | @item @emph{Description}: | |
3148 | Print debug information pertaining to the accelerator. | |
3149 | @end table | |
3150 | ||
3151 | ||
3152 | ||
3153 | @c --------------------------------------------------------------------- | |
3154 | @c CUDA Streams Usage | |
3155 | @c --------------------------------------------------------------------- | |
3156 | ||
3157 | @node CUDA Streams Usage | |
3158 | @chapter CUDA Streams Usage | |
3159 | ||
3160 | This applies to the @code{nvptx} plugin only. | |
3161 | ||
3162 | The library provides elements that perform asynchronous movement of | |
3163 | data and asynchronous operation of computing constructs. This | |
3164 | asynchronous functionality is implemented by making use of CUDA | |
3165 | streams@footnote{See "Stream Management" in "CUDA Driver API", | |
3166 | TRM-06703-001, Version 5.5, for additional information}. | |
3167 | ||
c1030b5c | 3168 | The primary means by that the asynchronous functionality is accessed |
cdf6119d JN |
3169 | is through the use of those OpenACC directives which make use of the |
3170 | @code{async} and @code{wait} clauses. When the @code{async} clause is | |
3171 | first used with a directive, it creates a CUDA stream. If an | |
3172 | @code{async-argument} is used with the @code{async} clause, then the | |
3173 | stream is associated with the specified @code{async-argument}. | |
3174 | ||
3175 | Following the creation of an association between a CUDA stream and the | |
3176 | @code{async-argument} of an @code{async} clause, both the @code{wait} | |
3177 | clause and the @code{wait} directive can be used. When either the | |
3178 | clause or directive is used after stream creation, it creates a | |
3179 | rendezvous point whereby execution waits until all operations | |
3180 | associated with the @code{async-argument}, that is, stream, have | |
3181 | completed. | |
3182 | ||
3183 | Normally, the management of the streams that are created as a result of | |
3184 | using the @code{async} clause, is done without any intervention by the | |
3185 | caller. This implies the association between the @code{async-argument} | |
3186 | and the CUDA stream will be maintained for the lifetime of the program. | |
3187 | However, this association can be changed through the use of the library | |
3188 | function @code{acc_set_cuda_stream}. When the function | |
3189 | @code{acc_set_cuda_stream} is called, the CUDA stream that was | |
3190 | originally associated with the @code{async} clause will be destroyed. | |
3191 | Caution should be taken when changing the association as subsequent | |
3192 | references to the @code{async-argument} refer to a different | |
3193 | CUDA stream. | |
3194 | ||
3195 | ||
3196 | ||
3197 | @c --------------------------------------------------------------------- | |
3198 | @c OpenACC Library Interoperability | |
3199 | @c --------------------------------------------------------------------- | |
3200 | ||
3201 | @node OpenACC Library Interoperability | |
3202 | @chapter OpenACC Library Interoperability | |
3203 | ||
3204 | @section Introduction | |
3205 | ||
3206 | The OpenACC library uses the CUDA Driver API, and may interact with | |
3207 | programs that use the Runtime library directly, or another library | |
3208 | based on the Runtime library, e.g., CUBLAS@footnote{See section 2.26, | |
3209 | "Interactions with the CUDA Driver API" in | |
3210 | "CUDA Runtime API", Version 5.5, and section 2.27, "VDPAU | |
3211 | Interoperability", in "CUDA Driver API", TRM-06703-001, Version 5.5, | |
3212 | for additional information on library interoperability.}. | |
3213 | This chapter describes the use cases and what changes are | |
3214 | required in order to use both the OpenACC library and the CUBLAS and Runtime | |
3215 | libraries within a program. | |
3216 | ||
3217 | @section First invocation: NVIDIA CUBLAS library API | |
3218 | ||
3219 | In this first use case (see below), a function in the CUBLAS library is called | |
3220 | prior to any of the functions in the OpenACC library. More specifically, the | |
3221 | function @code{cublasCreate()}. | |
3222 | ||
3223 | When invoked, the function initializes the library and allocates the | |
3224 | hardware resources on the host and the device on behalf of the caller. Once | |
3225 | the initialization and allocation has completed, a handle is returned to the | |
3226 | caller. The OpenACC library also requires initialization and allocation of | |
3227 | hardware resources. Since the CUBLAS library has already allocated the | |
3228 | hardware resources for the device, all that is left to do is to initialize | |
3229 | the OpenACC library and acquire the hardware resources on the host. | |
3230 | ||
3231 | Prior to calling the OpenACC function that initializes the library and | |
3232 | allocate the host hardware resources, you need to acquire the device number | |
3233 | that was allocated during the call to @code{cublasCreate()}. The invoking of the | |
3234 | runtime library function @code{cudaGetDevice()} accomplishes this. Once | |
3235 | acquired, the device number is passed along with the device type as | |
3236 | parameters to the OpenACC library function @code{acc_set_device_num()}. | |
3237 | ||
3238 | Once the call to @code{acc_set_device_num()} has completed, the OpenACC | |
3239 | library uses the context that was created during the call to | |
3240 | @code{cublasCreate()}. In other words, both libraries will be sharing the | |
3241 | same context. | |
3242 | ||
3243 | @smallexample | |
3244 | /* Create the handle */ | |
3245 | s = cublasCreate(&h); | |
3246 | if (s != CUBLAS_STATUS_SUCCESS) | |
3247 | @{ | |
3248 | fprintf(stderr, "cublasCreate failed %d\n", s); | |
3249 | exit(EXIT_FAILURE); | |
3250 | @} | |
3251 | ||
3252 | /* Get the device number */ | |
3253 | e = cudaGetDevice(&dev); | |
3254 | if (e != cudaSuccess) | |
3255 | @{ | |
3256 | fprintf(stderr, "cudaGetDevice failed %d\n", e); | |
3257 | exit(EXIT_FAILURE); | |
3258 | @} | |
3259 | ||
3260 | /* Initialize OpenACC library and use device 'dev' */ | |
3261 | acc_set_device_num(dev, acc_device_nvidia); | |
3262 | ||
3263 | @end smallexample | |
3264 | @center Use Case 1 | |
3265 | ||
3266 | @section First invocation: OpenACC library API | |
3267 | ||
3268 | In this second use case (see below), a function in the OpenACC library is | |
3269 | called prior to any of the functions in the CUBLAS library. More specificially, | |
3270 | the function @code{acc_set_device_num()}. | |
3271 | ||
3272 | In the use case presented here, the function @code{acc_set_device_num()} | |
3273 | is used to both initialize the OpenACC library and allocate the hardware | |
3274 | resources on the host and the device. In the call to the function, the | |
3275 | call parameters specify which device to use and what device | |
3276 | type to use, i.e., @code{acc_device_nvidia}. It should be noted that this | |
3277 | is but one method to initialize the OpenACC library and allocate the | |
3278 | appropriate hardware resources. Other methods are available through the | |
3279 | use of environment variables and these will be discussed in the next section. | |
3280 | ||
3281 | Once the call to @code{acc_set_device_num()} has completed, other OpenACC | |
3282 | functions can be called as seen with multiple calls being made to | |
3283 | @code{acc_copyin()}. In addition, calls can be made to functions in the | |
3284 | CUBLAS library. In the use case a call to @code{cublasCreate()} is made | |
3285 | subsequent to the calls to @code{acc_copyin()}. | |
3286 | As seen in the previous use case, a call to @code{cublasCreate()} | |
3287 | initializes the CUBLAS library and allocates the hardware resources on the | |
3288 | host and the device. However, since the device has already been allocated, | |
3289 | @code{cublasCreate()} will only initialize the CUBLAS library and allocate | |
3290 | the appropriate hardware resources on the host. The context that was created | |
3291 | as part of the OpenACC initialization is shared with the CUBLAS library, | |
3292 | similarly to the first use case. | |
3293 | ||
3294 | @smallexample | |
3295 | dev = 0; | |
3296 | ||
3297 | acc_set_device_num(dev, acc_device_nvidia); | |
3298 | ||
3299 | /* Copy the first set to the device */ | |
3300 | d_X = acc_copyin(&h_X[0], N * sizeof (float)); | |
3301 | if (d_X == NULL) | |
3302 | @{ | |
3303 | fprintf(stderr, "copyin error h_X\n"); | |
3304 | exit(EXIT_FAILURE); | |
3305 | @} | |
3306 | ||
3307 | /* Copy the second set to the device */ | |
3308 | d_Y = acc_copyin(&h_Y1[0], N * sizeof (float)); | |
3309 | if (d_Y == NULL) | |
3310 | @{ | |
3311 | fprintf(stderr, "copyin error h_Y1\n"); | |
3312 | exit(EXIT_FAILURE); | |
3313 | @} | |
3314 | ||
3315 | /* Create the handle */ | |
3316 | s = cublasCreate(&h); | |
3317 | if (s != CUBLAS_STATUS_SUCCESS) | |
3318 | @{ | |
3319 | fprintf(stderr, "cublasCreate failed %d\n", s); | |
3320 | exit(EXIT_FAILURE); | |
3321 | @} | |
3322 | ||
3323 | /* Perform saxpy using CUBLAS library function */ | |
3324 | s = cublasSaxpy(h, N, &alpha, d_X, 1, d_Y, 1); | |
3325 | if (s != CUBLAS_STATUS_SUCCESS) | |
3326 | @{ | |
3327 | fprintf(stderr, "cublasSaxpy failed %d\n", s); | |
3328 | exit(EXIT_FAILURE); | |
3329 | @} | |
3330 | ||
3331 | /* Copy the results from the device */ | |
3332 | acc_memcpy_from_device(&h_Y1[0], d_Y, N * sizeof (float)); | |
3333 | ||
3334 | @end smallexample | |
3335 | @center Use Case 2 | |
3336 | ||
3337 | @section OpenACC library and environment variables | |
3338 | ||
3339 | There are two environment variables associated with the OpenACC library | |
3340 | that may be used to control the device type and device number: | |
8d1a1cb1 TB |
3341 | @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM}, respectively. These two |
3342 | environment variables can be used as an alternative to calling | |
cdf6119d JN |
3343 | @code{acc_set_device_num()}. As seen in the second use case, the device |
3344 | type and device number were specified using @code{acc_set_device_num()}. | |
3345 | If however, the aforementioned environment variables were set, then the | |
3346 | call to @code{acc_set_device_num()} would not be required. | |
3347 | ||
3348 | ||
3349 | The use of the environment variables is only relevant when an OpenACC function | |
3350 | is called prior to a call to @code{cudaCreate()}. If @code{cudaCreate()} | |
3351 | is called prior to a call to an OpenACC function, then you must call | |
3352 | @code{acc_set_device_num()}@footnote{More complete information | |
3353 | about @env{ACC_DEVICE_TYPE} and @env{ACC_DEVICE_NUM} can be found in | |
9651fbaf | 3354 | sections 4.1 and 4.2 of the @uref{https://www.openacc.org, OpenACC} |
e464fc90 | 3355 | Application Programming Interface”, Version 2.6.} |
cdf6119d JN |
3356 | |
3357 | ||
3358 | ||
5fae049d TS |
3359 | @c --------------------------------------------------------------------- |
3360 | @c OpenACC Profiling Interface | |
3361 | @c --------------------------------------------------------------------- | |
3362 | ||
3363 | @node OpenACC Profiling Interface | |
3364 | @chapter OpenACC Profiling Interface | |
3365 | ||
3366 | @section Implementation Status and Implementation-Defined Behavior | |
3367 | ||
3368 | We're implementing the OpenACC Profiling Interface as defined by the | |
3369 | OpenACC 2.6 specification. We're clarifying some aspects here as | |
3370 | @emph{implementation-defined behavior}, while they're still under | |
3371 | discussion within the OpenACC Technical Committee. | |
3372 | ||
3373 | This implementation is tuned to keep the performance impact as low as | |
3374 | possible for the (very common) case that the Profiling Interface is | |
3375 | not enabled. This is relevant, as the Profiling Interface affects all | |
3376 | the @emph{hot} code paths (in the target code, not in the offloaded | |
3377 | code). Users of the OpenACC Profiling Interface can be expected to | |
3378 | understand that performance will be impacted to some degree once the | |
3379 | Profiling Interface has gotten enabled: for example, because of the | |
3380 | @emph{runtime} (libgomp) calling into a third-party @emph{library} for | |
3381 | every event that has been registered. | |
3382 | ||
3383 | We're not yet accounting for the fact that @cite{OpenACC events may | |
3384 | occur during event processing}. | |
3385 | ||
3386 | We're not yet implementing initialization via a | |
3387 | @code{acc_register_library} function that is either statically linked | |
3388 | in, or dynamically via @env{LD_PRELOAD}. | |
3389 | Initialization via @code{acc_register_library} functions dynamically | |
3390 | loaded via the @env{ACC_PROFLIB} environment variable does work, as | |
3391 | does directly calling @code{acc_prof_register}, | |
3392 | @code{acc_prof_unregister}, @code{acc_prof_lookup}. | |
3393 | ||
3394 | As currently there are no inquiry functions defined, calls to | |
3395 | @code{acc_prof_lookup} will always return @code{NULL}. | |
3396 | ||
3397 | There aren't separate @emph{start}, @emph{stop} events defined for the | |
3398 | event types @code{acc_ev_create}, @code{acc_ev_delete}, | |
3399 | @code{acc_ev_alloc}, @code{acc_ev_free}. It's not clear if these | |
3400 | should be triggered before or after the actual device-specific call is | |
3401 | made. We trigger them after. | |
3402 | ||
3403 | Remarks about data provided to callbacks: | |
3404 | ||
3405 | @table @asis | |
3406 | ||
3407 | @item @code{acc_prof_info.event_type} | |
3408 | It's not clear if for @emph{nested} event callbacks (for example, | |
3409 | @code{acc_ev_enqueue_launch_start} as part of a parent compute | |
3410 | construct), this should be set for the nested event | |
3411 | (@code{acc_ev_enqueue_launch_start}), or if the value of the parent | |
3412 | construct should remain (@code{acc_ev_compute_construct_start}). In | |
3413 | this implementation, the value will generally correspond to the | |
3414 | innermost nested event type. | |
3415 | ||
3416 | @item @code{acc_prof_info.device_type} | |
3417 | @itemize | |
3418 | ||
3419 | @item | |
3420 | For @code{acc_ev_compute_construct_start}, and in presence of an | |
3421 | @code{if} clause with @emph{false} argument, this will still refer to | |
3422 | the offloading device type. | |
3423 | It's not clear if that's the expected behavior. | |
3424 | ||
3425 | @item | |
3426 | Complementary to the item before, for | |
3427 | @code{acc_ev_compute_construct_end}, this is set to | |
3428 | @code{acc_device_host} in presence of an @code{if} clause with | |
3429 | @emph{false} argument. | |
3430 | It's not clear if that's the expected behavior. | |
3431 | ||
3432 | @end itemize | |
3433 | ||
3434 | @item @code{acc_prof_info.thread_id} | |
3435 | Always @code{-1}; not yet implemented. | |
3436 | ||
3437 | @item @code{acc_prof_info.async} | |
3438 | @itemize | |
3439 | ||
3440 | @item | |
3441 | Not yet implemented correctly for | |
3442 | @code{acc_ev_compute_construct_start}. | |
3443 | ||
3444 | @item | |
3445 | In a compute construct, for host-fallback | |
3446 | execution/@code{acc_device_host} it will always be | |
3447 | @code{acc_async_sync}. | |
3448 | It's not clear if that's the expected behavior. | |
3449 | ||
3450 | @item | |
3451 | For @code{acc_ev_device_init_start} and @code{acc_ev_device_init_end}, | |
3452 | it will always be @code{acc_async_sync}. | |
3453 | It's not clear if that's the expected behavior. | |
3454 | ||
3455 | @end itemize | |
3456 | ||
3457 | @item @code{acc_prof_info.async_queue} | |
3458 | There is no @cite{limited number of asynchronous queues} in libgomp. | |
3459 | This will always have the same value as @code{acc_prof_info.async}. | |
3460 | ||
3461 | @item @code{acc_prof_info.src_file} | |
3462 | Always @code{NULL}; not yet implemented. | |
3463 | ||
3464 | @item @code{acc_prof_info.func_name} | |
3465 | Always @code{NULL}; not yet implemented. | |
3466 | ||
3467 | @item @code{acc_prof_info.line_no} | |
3468 | Always @code{-1}; not yet implemented. | |
3469 | ||
3470 | @item @code{acc_prof_info.end_line_no} | |
3471 | Always @code{-1}; not yet implemented. | |
3472 | ||
3473 | @item @code{acc_prof_info.func_line_no} | |
3474 | Always @code{-1}; not yet implemented. | |
3475 | ||
3476 | @item @code{acc_prof_info.func_end_line_no} | |
3477 | Always @code{-1}; not yet implemented. | |
3478 | ||
3479 | @item @code{acc_event_info.event_type}, @code{acc_event_info.*.event_type} | |
3480 | Relating to @code{acc_prof_info.event_type} discussed above, in this | |
3481 | implementation, this will always be the same value as | |
3482 | @code{acc_prof_info.event_type}. | |
3483 | ||
3484 | @item @code{acc_event_info.*.parent_construct} | |
3485 | @itemize | |
3486 | ||
3487 | @item | |
3488 | Will be @code{acc_construct_parallel} for all OpenACC compute | |
3489 | constructs as well as many OpenACC Runtime API calls; should be the | |
3490 | one matching the actual construct, or | |
3491 | @code{acc_construct_runtime_api}, respectively. | |
3492 | ||
3493 | @item | |
3494 | Will be @code{acc_construct_enter_data} or | |
3495 | @code{acc_construct_exit_data} when processing variable mappings | |
3496 | specified in OpenACC @emph{declare} directives; should be | |
3497 | @code{acc_construct_declare}. | |
3498 | ||
3499 | @item | |
3500 | For implicit @code{acc_ev_device_init_start}, | |
3501 | @code{acc_ev_device_init_end}, and explicit as well as implicit | |
3502 | @code{acc_ev_alloc}, @code{acc_ev_free}, | |
3503 | @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end}, | |
3504 | @code{acc_ev_enqueue_download_start}, and | |
3505 | @code{acc_ev_enqueue_download_end}, will be | |
3506 | @code{acc_construct_parallel}; should reflect the real parent | |
3507 | construct. | |
3508 | ||
3509 | @end itemize | |
3510 | ||
3511 | @item @code{acc_event_info.*.implicit} | |
3512 | For @code{acc_ev_alloc}, @code{acc_ev_free}, | |
3513 | @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end}, | |
3514 | @code{acc_ev_enqueue_download_start}, and | |
3515 | @code{acc_ev_enqueue_download_end}, this currently will be @code{1} | |
3516 | also for explicit usage. | |
3517 | ||
3518 | @item @code{acc_event_info.data_event.var_name} | |
3519 | Always @code{NULL}; not yet implemented. | |
3520 | ||
3521 | @item @code{acc_event_info.data_event.host_ptr} | |
3522 | For @code{acc_ev_alloc}, and @code{acc_ev_free}, this is always | |
3523 | @code{NULL}. | |
3524 | ||
3525 | @item @code{typedef union acc_api_info} | |
3526 | @dots{} as printed in @cite{5.2.3. Third Argument: API-Specific | |
3527 | Information}. This should obviously be @code{typedef @emph{struct} | |
3528 | acc_api_info}. | |
3529 | ||
3530 | @item @code{acc_api_info.device_api} | |
3531 | Possibly not yet implemented correctly for | |
3532 | @code{acc_ev_compute_construct_start}, | |
3533 | @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end}: | |
3534 | will always be @code{acc_device_api_none} for these event types. | |
3535 | For @code{acc_ev_enter_data_start}, it will be | |
3536 | @code{acc_device_api_none} in some cases. | |
3537 | ||
3538 | @item @code{acc_api_info.device_type} | |
3539 | Always the same as @code{acc_prof_info.device_type}. | |
3540 | ||
3541 | @item @code{acc_api_info.vendor} | |
3542 | Always @code{-1}; not yet implemented. | |
3543 | ||
3544 | @item @code{acc_api_info.device_handle} | |
3545 | Always @code{NULL}; not yet implemented. | |
3546 | ||
3547 | @item @code{acc_api_info.context_handle} | |
3548 | Always @code{NULL}; not yet implemented. | |
3549 | ||
3550 | @item @code{acc_api_info.async_handle} | |
3551 | Always @code{NULL}; not yet implemented. | |
3552 | ||
3553 | @end table | |
3554 | ||
3555 | Remarks about certain event types: | |
3556 | ||
3557 | @table @asis | |
3558 | ||
3559 | @item @code{acc_ev_device_init_start}, @code{acc_ev_device_init_end} | |
3560 | @itemize | |
3561 | ||
3562 | @item | |
3563 | @c See 'DEVICE_INIT_INSIDE_COMPUTE_CONSTRUCT' in | |
3564 | @c 'libgomp.oacc-c-c++-common/acc_prof-kernels-1.c', | |
3565 | @c 'libgomp.oacc-c-c++-common/acc_prof-parallel-1.c'. | |
3566 | Whan a compute construct triggers implicit | |
3567 | @code{acc_ev_device_init_start} and @code{acc_ev_device_init_end} | |
3568 | events, they currently aren't @emph{nested within} the corresponding | |
3569 | @code{acc_ev_compute_construct_start} and | |
3570 | @code{acc_ev_compute_construct_end}, but they're currently observed | |
3571 | @emph{before} @code{acc_ev_compute_construct_start}. | |
3572 | It's not clear what to do: the standard asks us provide a lot of | |
3573 | details to the @code{acc_ev_compute_construct_start} callback, without | |
3574 | (implicitly) initializing a device before? | |
3575 | ||
3576 | @item | |
3577 | Callbacks for these event types will not be invoked for calls to the | |
3578 | @code{acc_set_device_type} and @code{acc_set_device_num} functions. | |
3579 | It's not clear if they should be. | |
3580 | ||
3581 | @end itemize | |
3582 | ||
3583 | @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} | |
3584 | @itemize | |
3585 | ||
3586 | @item | |
3587 | Callbacks for these event types will also be invoked for OpenACC | |
3588 | @emph{host_data} constructs. | |
3589 | It's not clear if they should be. | |
3590 | ||
3591 | @item | |
3592 | Callbacks for these event types will also be invoked when processing | |
3593 | variable mappings specified in OpenACC @emph{declare} directives. | |
3594 | It's not clear if they should be. | |
3595 | ||
3596 | @end itemize | |
3597 | ||
3598 | @end table | |
3599 | ||
3600 | Callbacks for the following event types will be invoked, but dispatch | |
3601 | and information provided therein has not yet been thoroughly reviewed: | |
3602 | ||
3603 | @itemize | |
3604 | @item @code{acc_ev_alloc} | |
3605 | @item @code{acc_ev_free} | |
3606 | @item @code{acc_ev_update_start}, @code{acc_ev_update_end} | |
3607 | @item @code{acc_ev_enqueue_upload_start}, @code{acc_ev_enqueue_upload_end} | |
3608 | @item @code{acc_ev_enqueue_download_start}, @code{acc_ev_enqueue_download_end} | |
3609 | @end itemize | |
3610 | ||
3611 | During device initialization, and finalization, respectively, | |
3612 | callbacks for the following event types will not yet be invoked: | |
3613 | ||
3614 | @itemize | |
3615 | @item @code{acc_ev_alloc} | |
3616 | @item @code{acc_ev_free} | |
3617 | @end itemize | |
3618 | ||
3619 | Callbacks for the following event types have not yet been implemented, | |
3620 | so currently won't be invoked: | |
3621 | ||
3622 | @itemize | |
3623 | @item @code{acc_ev_device_shutdown_start}, @code{acc_ev_device_shutdown_end} | |
3624 | @item @code{acc_ev_runtime_shutdown} | |
3625 | @item @code{acc_ev_create}, @code{acc_ev_delete} | |
3626 | @item @code{acc_ev_wait_start}, @code{acc_ev_wait_end} | |
3627 | @end itemize | |
3628 | ||
3629 | For the following runtime library functions, not all expected | |
3630 | callbacks will be invoked (mostly concerning implicit device | |
3631 | initialization): | |
3632 | ||
3633 | @itemize | |
3634 | @item @code{acc_get_num_devices} | |
3635 | @item @code{acc_set_device_type} | |
3636 | @item @code{acc_get_device_type} | |
3637 | @item @code{acc_set_device_num} | |
3638 | @item @code{acc_get_device_num} | |
3639 | @item @code{acc_init} | |
3640 | @item @code{acc_shutdown} | |
3641 | @end itemize | |
3642 | ||
3643 | Aside from implicit device initialization, for the following runtime | |
3644 | library functions, no callbacks will be invoked for shared-memory | |
3645 | offloading devices (it's not clear if they should be): | |
3646 | ||
3647 | @itemize | |
3648 | @item @code{acc_malloc} | |
3649 | @item @code{acc_free} | |
3650 | @item @code{acc_copyin}, @code{acc_present_or_copyin}, @code{acc_copyin_async} | |
3651 | @item @code{acc_create}, @code{acc_present_or_create}, @code{acc_create_async} | |
3652 | @item @code{acc_copyout}, @code{acc_copyout_async}, @code{acc_copyout_finalize}, @code{acc_copyout_finalize_async} | |
3653 | @item @code{acc_delete}, @code{acc_delete_async}, @code{acc_delete_finalize}, @code{acc_delete_finalize_async} | |
3654 | @item @code{acc_update_device}, @code{acc_update_device_async} | |
3655 | @item @code{acc_update_self}, @code{acc_update_self_async} | |
3656 | @item @code{acc_map_data}, @code{acc_unmap_data} | |
3657 | @item @code{acc_memcpy_to_device}, @code{acc_memcpy_to_device_async} | |
3658 | @item @code{acc_memcpy_from_device}, @code{acc_memcpy_from_device_async} | |
3659 | @end itemize | |
3660 | ||
3661 | ||
3662 | ||
3721b9e1 DF |
3663 | @c --------------------------------------------------------------------- |
3664 | @c The libgomp ABI | |
3665 | @c --------------------------------------------------------------------- | |
3666 | ||
3667 | @node The libgomp ABI | |
3668 | @chapter The libgomp ABI | |
3669 | ||
3670 | The following sections present notes on the external ABI as | |
6a2ba183 | 3671 | presented by libgomp. Only maintainers should need them. |
3721b9e1 DF |
3672 | |
3673 | @menu | |
3674 | * Implementing MASTER construct:: | |
3675 | * Implementing CRITICAL construct:: | |
3676 | * Implementing ATOMIC construct:: | |
3677 | * Implementing FLUSH construct:: | |
3678 | * Implementing BARRIER construct:: | |
3679 | * Implementing THREADPRIVATE construct:: | |
3680 | * Implementing PRIVATE clause:: | |
3681 | * Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses:: | |
3682 | * Implementing REDUCTION clause:: | |
3683 | * Implementing PARALLEL construct:: | |
3684 | * Implementing FOR construct:: | |
3685 | * Implementing ORDERED construct:: | |
3686 | * Implementing SECTIONS construct:: | |
3687 | * Implementing SINGLE construct:: | |
cdf6119d | 3688 | * Implementing OpenACC's PARALLEL construct:: |
3721b9e1 DF |
3689 | @end menu |
3690 | ||
3691 | ||
3692 | @node Implementing MASTER construct | |
3693 | @section Implementing MASTER construct | |
3694 | ||
3695 | @smallexample | |
3696 | if (omp_get_thread_num () == 0) | |
3697 | block | |
3698 | @end smallexample | |
3699 | ||
3700 | Alternately, we generate two copies of the parallel subfunction | |
3701 | and only include this in the version run by the master thread. | |
6a2ba183 | 3702 | Surely this is not worthwhile though... |
3721b9e1 DF |
3703 | |
3704 | ||
3705 | ||
3706 | @node Implementing CRITICAL construct | |
3707 | @section Implementing CRITICAL construct | |
3708 | ||
3709 | Without a specified name, | |
3710 | ||
3711 | @smallexample | |
3712 | void GOMP_critical_start (void); | |
3713 | void GOMP_critical_end (void); | |
3714 | @end smallexample | |
3715 | ||
3716 | so that we don't get COPY relocations from libgomp to the main | |
3717 | application. | |
3718 | ||
3719 | With a specified name, use omp_set_lock and omp_unset_lock with | |
3720 | name being transformed into a variable declared like | |
3721 | ||
3722 | @smallexample | |
3723 | omp_lock_t gomp_critical_user_<name> __attribute__((common)) | |
3724 | @end smallexample | |
3725 | ||
3726 | Ideally the ABI would specify that all zero is a valid unlocked | |
6a2ba183 | 3727 | state, and so we wouldn't need to initialize this at |
3721b9e1 DF |
3728 | startup. |
3729 | ||
3730 | ||
3731 | ||
3732 | @node Implementing ATOMIC construct | |
3733 | @section Implementing ATOMIC construct | |
3734 | ||
3735 | The target should implement the @code{__sync} builtins. | |
3736 | ||
3737 | Failing that we could add | |
3738 | ||
3739 | @smallexample | |
3740 | void GOMP_atomic_enter (void) | |
3741 | void GOMP_atomic_exit (void) | |
3742 | @end smallexample | |
3743 | ||
3744 | which reuses the regular lock code, but with yet another lock | |
3745 | object private to the library. | |
3746 | ||
3747 | ||
3748 | ||
3749 | @node Implementing FLUSH construct | |
3750 | @section Implementing FLUSH construct | |
3751 | ||
3752 | Expands to the @code{__sync_synchronize} builtin. | |
3753 | ||
3754 | ||
3755 | ||
3756 | @node Implementing BARRIER construct | |
3757 | @section Implementing BARRIER construct | |
3758 | ||
3759 | @smallexample | |
3760 | void GOMP_barrier (void) | |
3761 | @end smallexample | |
3762 | ||
3763 | ||
3764 | @node Implementing THREADPRIVATE construct | |
3765 | @section Implementing THREADPRIVATE construct | |
3766 | ||
3767 | In _most_ cases we can map this directly to @code{__thread}. Except | |
3768 | that OMP allows constructors for C++ objects. We can either | |
3769 | refuse to support this (how often is it used?) or we can | |
3770 | implement something akin to .ctors. | |
3771 | ||
3772 | Even more ideally, this ctor feature is handled by extensions | |
3773 | to the main pthreads library. Failing that, we can have a set | |
3774 | of entry points to register ctor functions to be called. | |
3775 | ||
3776 | ||
3777 | ||
3778 | @node Implementing PRIVATE clause | |
3779 | @section Implementing PRIVATE clause | |
3780 | ||
3781 | In association with a PARALLEL, or within the lexical extent | |
3782 | of a PARALLEL block, the variable becomes a local variable in | |
3783 | the parallel subfunction. | |
3784 | ||
3785 | In association with FOR or SECTIONS blocks, create a new | |
3786 | automatic variable within the current function. This preserves | |
3787 | the semantic of new variable creation. | |
3788 | ||
3789 | ||
3790 | ||
3791 | @node Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses | |
3792 | @section Implementing FIRSTPRIVATE LASTPRIVATE COPYIN and COPYPRIVATE clauses | |
3793 | ||
6a2ba183 AH |
3794 | This seems simple enough for PARALLEL blocks. Create a private |
3795 | struct for communicating between the parent and subfunction. | |
3721b9e1 DF |
3796 | In the parent, copy in values for scalar and "small" structs; |
3797 | copy in addresses for others TREE_ADDRESSABLE types. In the | |
3798 | subfunction, copy the value into the local variable. | |
3799 | ||
6a2ba183 AH |
3800 | It is not clear what to do with bare FOR or SECTION blocks. |
3801 | The only thing I can figure is that we do something like: | |
3721b9e1 DF |
3802 | |
3803 | @smallexample | |
3804 | #pragma omp for firstprivate(x) lastprivate(y) | |
3805 | for (int i = 0; i < n; ++i) | |
3806 | body; | |
3807 | @end smallexample | |
3808 | ||
3809 | which becomes | |
3810 | ||
3811 | @smallexample | |
3812 | @{ | |
3813 | int x = x, y; | |
3814 | ||
3815 | // for stuff | |
3816 | ||
3817 | if (i == n) | |
3818 | y = y; | |
3819 | @} | |
3820 | @end smallexample | |
3821 | ||
3822 | where the "x=x" and "y=y" assignments actually have different | |
3823 | uids for the two variables, i.e. not something you could write | |
3824 | directly in C. Presumably this only makes sense if the "outer" | |
3825 | x and y are global variables. | |
3826 | ||
3827 | COPYPRIVATE would work the same way, except the structure | |
3828 | broadcast would have to happen via SINGLE machinery instead. | |
3829 | ||
3830 | ||
3831 | ||
3832 | @node Implementing REDUCTION clause | |
3833 | @section Implementing REDUCTION clause | |
3834 | ||
3835 | The private struct mentioned in the previous section should have | |
3836 | a pointer to an array of the type of the variable, indexed by the | |
3837 | thread's @var{team_id}. The thread stores its final value into the | |
6a2ba183 | 3838 | array, and after the barrier, the master thread iterates over the |
3721b9e1 DF |
3839 | array to collect the values. |
3840 | ||
3841 | ||
3842 | @node Implementing PARALLEL construct | |
3843 | @section Implementing PARALLEL construct | |
3844 | ||
3845 | @smallexample | |
3846 | #pragma omp parallel | |
3847 | @{ | |
3848 | body; | |
3849 | @} | |
3850 | @end smallexample | |
3851 | ||
3852 | becomes | |
3853 | ||
3854 | @smallexample | |
3855 | void subfunction (void *data) | |
3856 | @{ | |
3857 | use data; | |
3858 | body; | |
3859 | @} | |
3860 | ||
3861 | setup data; | |
3862 | GOMP_parallel_start (subfunction, &data, num_threads); | |
3863 | subfunction (&data); | |
3864 | GOMP_parallel_end (); | |
3865 | @end smallexample | |
3866 | ||
3867 | @smallexample | |
3868 | void GOMP_parallel_start (void (*fn)(void *), void *data, unsigned num_threads) | |
3869 | @end smallexample | |
3870 | ||
3871 | The @var{FN} argument is the subfunction to be run in parallel. | |
3872 | ||
3873 | The @var{DATA} argument is a pointer to a structure used to | |
3874 | communicate data in and out of the subfunction, as discussed | |
f1b0882e | 3875 | above with respect to FIRSTPRIVATE et al. |
3721b9e1 DF |
3876 | |
3877 | The @var{NUM_THREADS} argument is 1 if an IF clause is present | |
3878 | and false, or the value of the NUM_THREADS clause, if | |
3879 | present, or 0. | |
3880 | ||
3881 | The function needs to create the appropriate number of | |
3882 | threads and/or launch them from the dock. It needs to | |
3883 | create the team structure and assign team ids. | |
3884 | ||
3885 | @smallexample | |
3886 | void GOMP_parallel_end (void) | |
3887 | @end smallexample | |
3888 | ||
3889 | Tears down the team and returns us to the previous @code{omp_in_parallel()} state. | |
3890 | ||
3891 | ||
3892 | ||
3893 | @node Implementing FOR construct | |
3894 | @section Implementing FOR construct | |
3895 | ||
3896 | @smallexample | |
3897 | #pragma omp parallel for | |
3898 | for (i = lb; i <= ub; i++) | |
3899 | body; | |
3900 | @end smallexample | |
3901 | ||
3902 | becomes | |
3903 | ||
3904 | @smallexample | |
3905 | void subfunction (void *data) | |
3906 | @{ | |
3907 | long _s0, _e0; | |
3908 | while (GOMP_loop_static_next (&_s0, &_e0)) | |
3909 | @{ | |
3910 | long _e1 = _e0, i; | |
3911 | for (i = _s0; i < _e1; i++) | |
3912 | body; | |
3913 | @} | |
3914 | GOMP_loop_end_nowait (); | |
3915 | @} | |
3916 | ||
3917 | GOMP_parallel_loop_static (subfunction, NULL, 0, lb, ub+1, 1, 0); | |
3918 | subfunction (NULL); | |
3919 | GOMP_parallel_end (); | |
3920 | @end smallexample | |
3921 | ||
3922 | @smallexample | |
3923 | #pragma omp for schedule(runtime) | |
3924 | for (i = 0; i < n; i++) | |
3925 | body; | |
3926 | @end smallexample | |
3927 | ||
3928 | becomes | |
3929 | ||
3930 | @smallexample | |
3931 | @{ | |
3932 | long i, _s0, _e0; | |
3933 | if (GOMP_loop_runtime_start (0, n, 1, &_s0, &_e0)) | |
3934 | do @{ | |
3935 | long _e1 = _e0; | |
3936 | for (i = _s0, i < _e0; i++) | |
3937 | body; | |
3938 | @} while (GOMP_loop_runtime_next (&_s0, _&e0)); | |
3939 | GOMP_loop_end (); | |
3940 | @} | |
3941 | @end smallexample | |
3942 | ||
6a2ba183 | 3943 | Note that while it looks like there is trickiness to propagating |
3721b9e1 DF |
3944 | a non-constant STEP, there isn't really. We're explicitly allowed |
3945 | to evaluate it as many times as we want, and any variables involved | |
3946 | should automatically be handled as PRIVATE or SHARED like any other | |
3947 | variables. So the expression should remain evaluable in the | |
3948 | subfunction. We can also pull it into a local variable if we like, | |
3949 | but since its supposed to remain unchanged, we can also not if we like. | |
3950 | ||
3951 | If we have SCHEDULE(STATIC), and no ORDERED, then we ought to be | |
3952 | able to get away with no work-sharing context at all, since we can | |
3953 | simply perform the arithmetic directly in each thread to divide up | |
3954 | the iterations. Which would mean that we wouldn't need to call any | |
3955 | of these routines. | |
3956 | ||
3957 | There are separate routines for handling loops with an ORDERED | |
3958 | clause. Bookkeeping for that is non-trivial... | |
3959 | ||
3960 | ||
3961 | ||
3962 | @node Implementing ORDERED construct | |
3963 | @section Implementing ORDERED construct | |
3964 | ||
3965 | @smallexample | |
3966 | void GOMP_ordered_start (void) | |
3967 | void GOMP_ordered_end (void) | |
3968 | @end smallexample | |
3969 | ||
3970 | ||
3971 | ||
3972 | @node Implementing SECTIONS construct | |
3973 | @section Implementing SECTIONS construct | |
3974 | ||
3975 | A block as | |
3976 | ||
3977 | @smallexample | |
3978 | #pragma omp sections | |
3979 | @{ | |
3980 | #pragma omp section | |
3981 | stmt1; | |
3982 | #pragma omp section | |
3983 | stmt2; | |
3984 | #pragma omp section | |
3985 | stmt3; | |
3986 | @} | |
3987 | @end smallexample | |
3988 | ||
3989 | becomes | |
3990 | ||
3991 | @smallexample | |
3992 | for (i = GOMP_sections_start (3); i != 0; i = GOMP_sections_next ()) | |
3993 | switch (i) | |
3994 | @{ | |
3995 | case 1: | |
3996 | stmt1; | |
3997 | break; | |
3998 | case 2: | |
3999 | stmt2; | |
4000 | break; | |
4001 | case 3: | |
4002 | stmt3; | |
4003 | break; | |
4004 | @} | |
4005 | GOMP_barrier (); | |
4006 | @end smallexample | |
4007 | ||
4008 | ||
4009 | @node Implementing SINGLE construct | |
4010 | @section Implementing SINGLE construct | |
4011 | ||
4012 | A block like | |
4013 | ||
4014 | @smallexample | |
4015 | #pragma omp single | |
4016 | @{ | |
4017 | body; | |
4018 | @} | |
4019 | @end smallexample | |
4020 | ||
4021 | becomes | |
4022 | ||
4023 | @smallexample | |
4024 | if (GOMP_single_start ()) | |
4025 | body; | |
4026 | GOMP_barrier (); | |
4027 | @end smallexample | |
4028 | ||
4029 | while | |
4030 | ||
4031 | @smallexample | |
4032 | #pragma omp single copyprivate(x) | |
4033 | body; | |
4034 | @end smallexample | |
4035 | ||
4036 | becomes | |
4037 | ||
4038 | @smallexample | |
4039 | datap = GOMP_single_copy_start (); | |
4040 | if (datap == NULL) | |
4041 | @{ | |
4042 | body; | |
4043 | data.x = x; | |
4044 | GOMP_single_copy_end (&data); | |
4045 | @} | |
4046 | else | |
4047 | x = datap->x; | |
4048 | GOMP_barrier (); | |
4049 | @end smallexample | |
4050 | ||
4051 | ||
4052 | ||
cdf6119d JN |
4053 | @node Implementing OpenACC's PARALLEL construct |
4054 | @section Implementing OpenACC's PARALLEL construct | |
4055 | ||
4056 | @smallexample | |
4057 | void GOACC_parallel () | |
4058 | @end smallexample | |
4059 | ||
4060 | ||
4061 | ||
3721b9e1 | 4062 | @c --------------------------------------------------------------------- |
f1f3453e | 4063 | @c Reporting Bugs |
3721b9e1 DF |
4064 | @c --------------------------------------------------------------------- |
4065 | ||
4066 | @node Reporting Bugs | |
4067 | @chapter Reporting Bugs | |
4068 | ||
f1f3453e | 4069 | Bugs in the GNU Offloading and Multi Processing Runtime Library should |
c1030b5c | 4070 | be reported via @uref{https://gcc.gnu.org/bugzilla/, Bugzilla}. Please add |
41dbbb37 TS |
4071 | "openacc", or "openmp", or both to the keywords field in the bug |
4072 | report, as appropriate. | |
3721b9e1 DF |
4073 | |
4074 | ||
4075 | ||
4076 | @c --------------------------------------------------------------------- | |
4077 | @c GNU General Public License | |
4078 | @c --------------------------------------------------------------------- | |
4079 | ||
e6fdc918 | 4080 | @include gpl_v3.texi |
3721b9e1 DF |
4081 | |
4082 | ||
4083 | ||
4084 | @c --------------------------------------------------------------------- | |
4085 | @c GNU Free Documentation License | |
4086 | @c --------------------------------------------------------------------- | |
4087 | ||
4088 | @include fdl.texi | |
4089 | ||
4090 | ||
4091 | ||
4092 | @c --------------------------------------------------------------------- | |
4093 | @c Funding Free Software | |
4094 | @c --------------------------------------------------------------------- | |
4095 | ||
4096 | @include funding.texi | |
4097 | ||
4098 | @c --------------------------------------------------------------------- | |
4099 | @c Index | |
4100 | @c --------------------------------------------------------------------- | |
4101 | ||
3d3949df SL |
4102 | @node Library Index |
4103 | @unnumbered Library Index | |
3721b9e1 DF |
4104 | |
4105 | @printindex cp | |
4106 | ||
4107 | @bye |