<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/> <title>libstdc++: par_loop.h Source File</title> <link href="tabs.css" rel="stylesheet" type="text/css"/> <link href="navtree.css" rel="stylesheet" type="text/css"/> <script type="text/javascript" src="jquery.js"></script> <script type="text/javascript" src="navtree.js"></script> <script type="text/javascript" src="resize.js"></script> <script type="text/javascript"> $(document).ready(initResizable); </script> <link href="doxygen.css" rel="stylesheet" type="text/css"/> </head> <body> <!-- Generated by Doxygen 1.7.4 --> <div id="top"> <div id="titlearea"> <table cellspacing="0" cellpadding="0"> <tbody> <tr style="height: 56px;"> <td style="padding-left: 0.5em;"> <div id="projectname">libstdc++</div> </td> </tr> </tbody> </table> </div> </div> <div id="side-nav" class="ui-resizable side-nav-resizable"> <div id="nav-tree"> <div id="nav-tree-contents"> </div> </div> <div id="splitbar" style="-moz-user-select:none;" class="ui-resizable-handle"> </div> </div> <script type="text/javascript"> initNavTree('a00970.html',''); </script> <div id="doc-content"> <div class="header"> <div class="headertitle"> <div class="title">par_loop.h</div> </div> </div> <div class="contents"> <a href="a00970.html">Go to the documentation of this file.</a><div class="fragment"><pre class="fragment"><a name="l00001"></a>00001 <span class="comment">// -*- C++ -*-</span> <a name="l00002"></a>00002 <a name="l00003"></a>00003 <span class="comment">// Copyright (C) 2007, 2008, 2009 Free Software Foundation, Inc.</span> <a name="l00004"></a>00004 <span class="comment">//</span> <a name="l00005"></a>00005 <span class="comment">// This file is part of the GNU ISO C++ Library. This library is free</span> <a name="l00006"></a>00006 <span class="comment">// software; you can redistribute it and/or modify it under the terms</span> <a name="l00007"></a>00007 <span class="comment">// of the GNU General Public License as published by the Free Software</span> <a name="l00008"></a>00008 <span class="comment">// Foundation; either version 3, or (at your option) any later</span> <a name="l00009"></a>00009 <span class="comment">// version.</span> <a name="l00010"></a>00010 <a name="l00011"></a>00011 <span class="comment">// This library is distributed in the hope that it will be useful, but</span> <a name="l00012"></a>00012 <span class="comment">// WITHOUT ANY WARRANTY; without even the implied warranty of</span> <a name="l00013"></a>00013 <span class="comment">// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU</span> <a name="l00014"></a>00014 <span class="comment">// General Public License for more details.</span> <a name="l00015"></a>00015 <a name="l00016"></a>00016 <span class="comment">// Under Section 7 of GPL version 3, you are granted additional</span> <a name="l00017"></a>00017 <span class="comment">// permissions described in the GCC Runtime Library Exception, version</span> <a name="l00018"></a>00018 <span class="comment">// 3.1, as published by the Free Software Foundation.</span> <a name="l00019"></a>00019 <a name="l00020"></a>00020 <span class="comment">// You should have received a copy of the GNU General Public License and</span> <a name="l00021"></a>00021 <span class="comment">// a copy of the GCC Runtime Library Exception along with this program;</span> <a name="l00022"></a>00022 <span class="comment">// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see</span> <a name="l00023"></a>00023 <span class="comment">// <http://www.gnu.org/licenses/>.</span> <a name="l00024"></a>00024 <span class="comment"></span> <a name="l00025"></a>00025 <span class="comment">/** @file parallel/par_loop.h</span> <a name="l00026"></a>00026 <span class="comment"> * @brief Parallelization of embarrassingly parallel execution by</span> <a name="l00027"></a>00027 <span class="comment"> * means of equal splitting.</span> <a name="l00028"></a>00028 <span class="comment"> * This file is a GNU parallel extension to the Standard C++ Library.</span> <a name="l00029"></a>00029 <span class="comment"> */</span> <a name="l00030"></a>00030 <a name="l00031"></a>00031 <span class="comment">// Written by Felix Putze.</span> <a name="l00032"></a>00032 <a name="l00033"></a>00033 <span class="preprocessor">#ifndef _GLIBCXX_PARALLEL_PAR_LOOP_H</span> <a name="l00034"></a>00034 <span class="preprocessor"></span><span class="preprocessor">#define _GLIBCXX_PARALLEL_PAR_LOOP_H 1</span> <a name="l00035"></a>00035 <span class="preprocessor"></span> <a name="l00036"></a>00036 <span class="preprocessor">#include <omp.h></span> <a name="l00037"></a>00037 <span class="preprocessor">#include <<a class="code" href="a01029.html" title="Runtime settings and tuning parameters, heuristics to decide whether to use parallelized algorithms...">parallel/settings.h</a>></span> <a name="l00038"></a>00038 <span class="preprocessor">#include <<a class="code" href="a00764.html" title="Sequential helper functions. This file is a GNU parallel extension to the Standard C++ Library...">parallel/base.h</a>></span> <a name="l00039"></a>00039 <span class="preprocessor">#include <<a class="code" href="a00861.html">parallel/equally_split.h</a>></span> <a name="l00040"></a>00040 <a name="l00041"></a>00041 <span class="keyword">namespace </span>__gnu_parallel <a name="l00042"></a>00042 {<span class="comment"></span> <a name="l00043"></a>00043 <span class="comment"> /** @brief Embarrassingly parallel algorithm for random access</span> <a name="l00044"></a>00044 <span class="comment"> * iterators, using hand-crafted parallelization by equal splitting</span> <a name="l00045"></a>00045 <span class="comment"> * the work.</span> <a name="l00046"></a>00046 <span class="comment"> *</span> <a name="l00047"></a>00047 <span class="comment"> * @param __begin Begin iterator of element sequence.</span> <a name="l00048"></a>00048 <span class="comment"> * @param __end End iterator of element sequence.</span> <a name="l00049"></a>00049 <span class="comment"> * @param __o User-supplied functor (comparator, predicate, adding</span> <a name="l00050"></a>00050 <span class="comment"> * functor, ...)</span> <a name="l00051"></a>00051 <span class="comment"> * @param __f Functor to "process" an element with __op (depends on</span> <a name="l00052"></a>00052 <span class="comment"> * desired functionality, e. g. for std::for_each(), ...).</span> <a name="l00053"></a>00053 <span class="comment"> * @param __r Functor to "add" a single __result to the already</span> <a name="l00054"></a>00054 <span class="comment"> * processed elements (depends on functionality).</span> <a name="l00055"></a>00055 <span class="comment"> * @param __base Base value for reduction.</span> <a name="l00056"></a>00056 <span class="comment"> * @param __output Pointer to position where final result is written to</span> <a name="l00057"></a>00057 <span class="comment"> * @param __bound Maximum number of elements processed (e. g. for</span> <a name="l00058"></a>00058 <span class="comment"> * std::count_n()).</span> <a name="l00059"></a>00059 <span class="comment"> * @return User-supplied functor (that may contain a part of the result).</span> <a name="l00060"></a>00060 <span class="comment"> */</span> <a name="l00061"></a>00061 <span class="keyword">template</span><<span class="keyword">typename</span> _RAIter, <a name="l00062"></a>00062 <span class="keyword">typename</span> _Op, <a name="l00063"></a>00063 <span class="keyword">typename</span> _Fu, <a name="l00064"></a>00064 <span class="keyword">typename</span> _Red, <a name="l00065"></a>00065 <span class="keyword">typename</span> _Result> <a name="l00066"></a>00066 _Op <a name="l00067"></a><a class="code" href="a01131.html#a0453e79d37dc4bff76695e07d8a72f31">00067</a> <a class="code" href="a01131.html#a0453e79d37dc4bff76695e07d8a72f31" title="Embarrassingly parallel algorithm for random access iterators, using hand-crafted parallelization by ...">__for_each_template_random_access_ed</a>(_RAIter __begin, _RAIter __end, <a name="l00068"></a>00068 _Op __o, _Fu& __f, _Red __r, <a name="l00069"></a>00069 _Result <a class="code" href="a01129.html#a1eb258935ccc1f18d8b8423cf079c353">__base</a>, _Result& __output, <a name="l00070"></a>00070 <span class="keyword">typename</span> std::iterator_traits<_RAIter>::difference_type __bound) <a name="l00071"></a>00071 { <a name="l00072"></a>00072 <span class="keyword">typedef</span> std::iterator_traits<_RAIter> _TraitsType; <a name="l00073"></a>00073 <span class="keyword">typedef</span> <span class="keyword">typename</span> _TraitsType::difference_type _DifferenceType; <a name="l00074"></a>00074 <span class="keyword">const</span> _DifferenceType __length = __end - __begin; <a name="l00075"></a>00075 _Result *__thread_results; <a name="l00076"></a>00076 <span class="keywordtype">bool</span>* __constructed; <a name="l00077"></a>00077 <a name="l00078"></a>00078 <a class="code" href="a01131.html#a05e502e51bfc3233671730f74a44dc6a" title="Unsigned integer to index a thread number. The maximum thread number (for each processor) must fit in...">_ThreadIndex</a> __num_threads = __gnu_parallel::min<_DifferenceType> <a name="l00079"></a>00079 (__get_max_threads(), __length); <a name="l00080"></a>00080 <a name="l00081"></a>00081 <span class="preprocessor"># pragma omp parallel num_threads(__num_threads)</span> <a name="l00082"></a>00082 <span class="preprocessor"></span> { <a name="l00083"></a>00083 <span class="preprocessor"># pragma omp single</span> <a name="l00084"></a>00084 <span class="preprocessor"></span> { <a name="l00085"></a>00085 __num_threads = omp_get_num_threads(); <a name="l00086"></a>00086 __thread_results = <span class="keyword">static_cast<</span>_Result*<span class="keyword">></span> <a name="l00087"></a>00087 (::operator <span class="keyword">new</span>(__num_threads * <span class="keyword">sizeof</span>(_Result))); <a name="l00088"></a>00088 __constructed = <span class="keyword">new</span> <span class="keywordtype">bool</span>[__num_threads]; <a name="l00089"></a>00089 } <a name="l00090"></a>00090 <a name="l00091"></a>00091 <a class="code" href="a01131.html#a05e502e51bfc3233671730f74a44dc6a" title="Unsigned integer to index a thread number. The maximum thread number (for each processor) must fit in...">_ThreadIndex</a> __iam = omp_get_thread_num(); <a name="l00092"></a>00092 <a name="l00093"></a>00093 <span class="comment">// Neutral element.</span> <a name="l00094"></a>00094 _Result* __reduct; <a name="l00095"></a>00095 <a name="l00096"></a>00096 _DifferenceType <a name="l00097"></a>00097 __start = <a class="code" href="a01131.html#af5fa80a30211cf4845899deada4e5b5e" title="function to split a sequence into parts of almost equal size.">equally_split_point</a>(__length, __num_threads, __iam), <a name="l00098"></a>00098 __stop = <a class="code" href="a01131.html#af5fa80a30211cf4845899deada4e5b5e" title="function to split a sequence into parts of almost equal size.">equally_split_point</a>(__length, __num_threads, __iam + 1); <a name="l00099"></a>00099 <a name="l00100"></a>00100 <span class="keywordflow">if</span> (__start < __stop) <a name="l00101"></a>00101 { <a name="l00102"></a>00102 __reduct = <span class="keyword">new</span> _Result(__f(__o, __begin + __start)); <a name="l00103"></a>00103 ++__start; <a name="l00104"></a>00104 __constructed[__iam] = <span class="keyword">true</span>; <a name="l00105"></a>00105 } <a name="l00106"></a>00106 <span class="keywordflow">else</span> <a name="l00107"></a>00107 __constructed[__iam] = <span class="keyword">false</span>; <a name="l00108"></a>00108 <a name="l00109"></a>00109 <span class="keywordflow">for</span> (; __start < __stop; ++__start) <a name="l00110"></a>00110 *__reduct = __r(*__reduct, __f(__o, __begin + __start)); <a name="l00111"></a>00111 <a name="l00112"></a>00112 <span class="keywordflow">if</span> (__constructed[__iam]) <a name="l00113"></a>00113 { <a name="l00114"></a>00114 ::new(&__thread_results[__iam]) _Result(*__reduct); <a name="l00115"></a>00115 <span class="keyword">delete</span> __reduct; <a name="l00116"></a>00116 } <a name="l00117"></a>00117 } <span class="comment">//parallel</span> <a name="l00118"></a>00118 <a name="l00119"></a>00119 <span class="keywordflow">for</span> (<a class="code" href="a01131.html#a05e502e51bfc3233671730f74a44dc6a" title="Unsigned integer to index a thread number. The maximum thread number (for each processor) must fit in...">_ThreadIndex</a> __i = 0; __i < __num_threads; ++__i) <a name="l00120"></a>00120 <span class="keywordflow">if</span> (__constructed[__i]) <a name="l00121"></a>00121 { <a name="l00122"></a>00122 __output = __r(__output, __thread_results[__i]); <a name="l00123"></a>00123 __thread_results[__i].~_Result(); <a name="l00124"></a>00124 } <a name="l00125"></a>00125 <a name="l00126"></a>00126 <span class="comment">// Points to last element processed (needed as return value for</span> <a name="l00127"></a>00127 <span class="comment">// some algorithms like transform).</span> <a name="l00128"></a>00128 __f._M_finish_iterator = __begin + __length; <a name="l00129"></a>00129 <a name="l00130"></a>00130 ::operator <span class="keyword">delete</span>(__thread_results); <a name="l00131"></a>00131 <a name="l00132"></a>00132 <span class="keyword">delete</span>[] __constructed; <a name="l00133"></a>00133 <a name="l00134"></a>00134 <span class="keywordflow">return</span> __o; <a name="l00135"></a>00135 } <a name="l00136"></a>00136 <a name="l00137"></a>00137 } <span class="comment">// end namespace</span> <a name="l00138"></a>00138 <a name="l00139"></a>00139 <span class="preprocessor">#endif </span><span class="comment">/* _GLIBCXX_PARALLEL_PAR_LOOP_H */</span> </pre></div></div> </div> <div id="nav-path" class="navpath"> <ul> <li class="navelem"><a class="el" href="a00970.html">par_loop.h</a> </li> <li class="footer">Generated by  <a href="http://www.doxygen.org/index.html"> <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.7.4 </li> </ul> </div> </body> </html>