]>
Commit | Line | Data |
---|---|---|
c2ba9709 JS |
1 | // -*- C++ -*- |
2 | ||
748086b7 | 3 | // Copyright (C) 2007, 2008, 2009 Free Software Foundation, Inc. |
c2ba9709 JS |
4 | // |
5 | // This file is part of the GNU ISO C++ Library. This library is free | |
6 | // software; you can redistribute it and/or modify it under the terms | |
7 | // of the GNU General Public License as published by the Free Software | |
748086b7 | 8 | // Foundation; either version 3, or (at your option) any later |
c2ba9709 JS |
9 | // version. |
10 | ||
11 | // This library is distributed in the hope that it will be useful, but | |
12 | // WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | // General Public License for more details. | |
15 | ||
748086b7 JJ |
16 | // Under Section 7 of GPL version 3, you are granted additional |
17 | // permissions described in the GCC Runtime Library Exception, version | |
18 | // 3.1, as published by the Free Software Foundation. | |
19 | ||
20 | // You should have received a copy of the GNU General Public License and | |
21 | // a copy of the GCC Runtime Library Exception along with this program; | |
22 | // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see | |
23 | // <http://www.gnu.org/licenses/>. | |
c2ba9709 JS |
24 | |
25 | /** @file parallel/omp_loop.h | |
26 | * @brief Parallelization of embarrassingly parallel execution by | |
27 | * means of an OpenMP for loop. | |
28 | * This file is a GNU parallel extension to the Standard C++ Library. | |
29 | */ | |
30 | ||
31 | // Written by Felix Putze. | |
32 | ||
33 | #ifndef _GLIBCXX_PARALLEL_OMP_LOOP_H | |
34 | #define _GLIBCXX_PARALLEL_OMP_LOOP_H 1 | |
35 | ||
36 | #include <omp.h> | |
37 | ||
38 | #include <parallel/settings.h> | |
39 | #include <parallel/basic_iterator.h> | |
e683ee2a | 40 | #include <parallel/base.h> |
c2ba9709 JS |
41 | |
42 | namespace __gnu_parallel | |
43 | { | |
e683ee2a JS |
44 | /** @brief Embarrassingly parallel algorithm for random access |
45 | * iterators, using an OpenMP for loop. | |
46 | * | |
1acba85b JS |
47 | * @param __begin Begin iterator of element __sequence. |
48 | * @param __end End iterator of element __sequence. | |
49 | * @param __o User-supplied functor (comparator, predicate, adding | |
e683ee2a | 50 | * functor, etc.). |
1acba85b | 51 | * @param __f Functor to "process" an element with __op (depends on |
e683ee2a | 52 | * desired functionality, e. g. for std::for_each(), ...). |
1acba85b JS |
53 | * @param __r Functor to "add" a single __result to the already |
54 | * processed __elements (depends on functionality). | |
55 | * @param __base Base value for reduction. | |
56 | * @param __output Pointer to position where final result is written to | |
57 | * @param __bound Maximum number of elements processed (e. g. for | |
e683ee2a JS |
58 | * std::count_n()). |
59 | * @return User-supplied functor (that may contain a part of the result). | |
60 | */ | |
1acba85b JS |
61 | template<typename _RAIter, |
62 | typename _Op, | |
63 | typename _Fu, | |
64 | typename _Red, | |
65 | typename _Result> | |
66 | _Op | |
67 | for_each_template_random_access_omp_loop(_RAIter __begin, | |
68 | _RAIter __end, | |
69 | _Op __o, _Fu& __f, _Red __r, _Result __base, | |
70 | _Result& __output, | |
5817ff8e | 71 | typename std::iterator_traits |
1acba85b JS |
72 | <_RAIter>:: |
73 | difference_type __bound) | |
c2ba9709 | 74 | { |
e683ee2a | 75 | typedef typename |
1acba85b JS |
76 | std::iterator_traits<_RAIter>::difference_type |
77 | _DifferenceType; | |
c2ba9709 | 78 | |
1acba85b JS |
79 | _DifferenceType __length = __end - __begin; |
80 | _ThreadIndex __num_threads = | |
81 | __gnu_parallel::min<_DifferenceType>(__get_max_threads(), __length); | |
c2ba9709 | 82 | |
1acba85b | 83 | _Result *__thread_results; |
e683ee2a | 84 | |
1acba85b | 85 | # pragma omp parallel num_threads(__num_threads) |
c2ba9709 | 86 | { |
e683ee2a JS |
87 | # pragma omp single |
88 | { | |
1acba85b JS |
89 | __num_threads = omp_get_num_threads(); |
90 | __thread_results = new _Result[__num_threads]; | |
e683ee2a | 91 | |
1acba85b JS |
92 | for (_ThreadIndex __i = 0; __i < __num_threads; ++__i) |
93 | __thread_results[__i] = _Result(); | |
e683ee2a JS |
94 | } |
95 | ||
1acba85b | 96 | _ThreadIndex __iam = omp_get_thread_num(); |
e683ee2a | 97 | |
ee1b5fc5 | 98 | # pragma omp for schedule(dynamic, _Settings::get().workstealing_chunk_size) |
1acba85b JS |
99 | for (_DifferenceType __pos = 0; __pos < __length; ++__pos) |
100 | __thread_results[__iam] = | |
101 | __r(__thread_results[__iam], __f(__o, __begin+__pos)); | |
e683ee2a | 102 | } //parallel |
c2ba9709 | 103 | |
1acba85b JS |
104 | for (_ThreadIndex __i = 0; __i < __num_threads; ++__i) |
105 | __output = __r(__output, __thread_results[__i]); | |
c2ba9709 | 106 | |
1acba85b | 107 | delete [] __thread_results; |
c2ba9709 JS |
108 | |
109 | // Points to last element processed (needed as return value for | |
110 | // some algorithms like transform). | |
1acba85b | 111 | __f.finish_iterator = __begin + __length; |
c2ba9709 | 112 | |
1acba85b | 113 | return __o; |
c2ba9709 | 114 | } |
e683ee2a | 115 | |
c2ba9709 JS |
116 | } // end namespace |
117 | ||
cbcd1e45 | 118 | #endif /* _GLIBCXX_PARALLEL_OMP_LOOP_H */ |