<!-- HTML header for doxygen 1.8.14-->
<!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"/>
<meta http-equiv="X-UA-Compatible" content="IE=9"/>
<meta name="generator" content="Doxygen 1.8.11"/>
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<title>Caffe2 - C++ API: caffe2/core/net_singlethread_async_gpu.cc Source File</title>
<link href="tabs.css" rel="stylesheet" type="text/css"/>
<link rel="icon" href="/static/favicon.png" type="image/x-icon">
<script type="text/javascript" src="jquery.js"></script>
<script type="text/javascript" src="dynsections.js"></script>
<link href="search/search.css" rel="stylesheet" type="text/css"/>
<script type="text/javascript" src="search/searchdata.js"></script>
<script type="text/javascript" src="search/search.js"></script>
<script type="text/javascript">
$(document).ready(function() { init_search(); });
</script>
<link href="stylesheet.css" rel="stylesheet" type="text/css" />
<link href="main.css" rel="stylesheet" type="text/css"/>
</head>
<body>
<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
<div id="titlearea">
<table cellspacing="0" cellpadding="0">
<tbody>
<tr style="height: 56px;">
<td id="projectlogo" width="56"><a href="/"><img alt="Logo" src="Caffe2-with-name-55-tall.png"/></a></td>
<td id="projectalign" style="padding-left: 0.5em;">
<div id="projectname">Caffe2 - C++ API
</div>
<div id="projectbrief">A deep learning, cross platform ML framework</div>
</td>
</tr>
</tbody>
</table>
</div>
<!-- end header part -->
<!-- Generated by Doxygen 1.8.11 -->
<script type="text/javascript">
var searchBox = new SearchBox("searchBox", "search",false,'Search');
</script>
<div id="navrow1" class="tabs">
<ul class="tablist">
<li><a href="pages.html"><span>Related Pages</span></a></li>
<li><a href="modules.html"><span>Modules</span></a></li>
<li><a href="annotated.html"><span>Data Structures</span></a></li>
<li class="current"><a href="files.html"><span>Files</span></a></li>
<li><a href="/doxygen-c/html/classes.html"><span>C++ API</span></a></li>
<li><a href="/doxygen-python/html/annotated.html"><span>Python API</span></a></li>
<li><a href="https://github.com/caffe2/caffe2"><span>GitHub</span></a></li>
<li>
<div id="MSearchBox" class="MSearchBoxInactive">
<span class="left">
<img id="MSearchSelect" src="search/mag_sel.png"
onmouseover="return searchBox.OnSearchSelectShow()"
onmouseout="return searchBox.OnSearchSelectHide()"
alt=""/>
<input type="text" id="MSearchField" value="Search" accesskey="S"
onfocus="searchBox.OnSearchFieldFocus(true)"
onblur="searchBox.OnSearchFieldFocus(false)"
onkeyup="searchBox.OnSearchFieldChange(event)"/>
</span><span class="right">
<a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
</span>
</div>
</li>
</ul>
</div>
<div id="navrow2" class="tabs2">
<ul class="tablist">
<li><a href="files.html"><span>File List</span></a></li>
<li><a href="globals.html"><span>Globals</span></a></li>
</ul>
</div>
<!-- window showing the filter options -->
<div id="MSearchSelectWindow"
onmouseover="return searchBox.OnSearchSelectShow()"
onmouseout="return searchBox.OnSearchSelectHide()"
onkeydown="return searchBox.OnSearchSelectKey(event)">
</div>
<!-- iframe showing the search results (closed by default) -->
<div id="MSearchResultsWindow">
<iframe src="javascript:void(0)" frameborder="0"
name="MSearchResults" id="MSearchResults">
</iframe>
</div>
<div id="nav-path" class="navpath">
<ul>
<li class="navelem"><a class="el" href="dir_20697b8f204bdfcab31e6b1a416f3ab8.html">caffe2</a></li><li class="navelem"><a class="el" href="dir_78eec69ac3a4b32ad49d9e5fc7146850.html">core</a></li> </ul>
</div>
</div><!-- top -->
<div class="header">
<div class="headertitle">
<div class="title">net_singlethread_async_gpu.cc</div> </div>
</div><!--header-->
<div class="contents">
<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno"> 1</span> <span class="preprocessor">#include <condition_variable></span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span> <span class="preprocessor">#include <mutex></span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="preprocessor">#include <stack></span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span> </div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span> <span class="preprocessor">#if !defined(_MSC_VER) && !defined(__APPLE__)</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span> <span class="preprocessor">#include <sched.h></span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span> </div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span> <span class="preprocessor">#include "caffe2/core/context_gpu.h"</span></div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span> <span class="preprocessor">#include "caffe2/core/net_simple.h"</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span> <span class="preprocessor">#include "caffe2/core/operator.h"</span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span> <span class="preprocessor">#include "caffe2/proto/caffe2.pb.h"</span></div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span> </div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span> <span class="keyword">namespace </span><a class="code" href="namespacecaffe2.html">caffe2</a> {</div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span> </div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span> <span class="keyword">namespace </span>gpu_single_thread {</div><div class="line"><a name="l00017"></a><span class="lineno"> 17</span> </div><div class="line"><a name="l00018"></a><span class="lineno"><a class="line" href="structcaffe2_1_1gpu__single__thread_1_1_task.html"> 18</a></span> <span class="keyword">struct </span><a class="code" href="structcaffe2_1_1gpu__single__thread_1_1_task.html">Task</a> {</div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span>  std::vector<std::unique_ptr<OperatorBase>>* ops_;</div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span>  std::condition_variable* cv_;</div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span>  std::mutex* mtx_;</div><div class="line"><a name="l00022"></a><span class="lineno"> 22</span>  <span class="keywordtype">int</span> stream_id_;</div><div class="line"><a name="l00023"></a><span class="lineno"> 23</span>  <span class="keywordtype">bool</span> done_ = <span class="keyword">false</span>;</div><div class="line"><a name="l00024"></a><span class="lineno"> 24</span> };</div><div class="line"><a name="l00025"></a><span class="lineno"> 25</span> </div><div class="line"><a name="l00026"></a><span class="lineno"><a class="line" href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html"> 26</a></span> <span class="keyword">class </span><a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html">GPUExecutor</a> {</div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span>  <span class="keyword">public</span>:</div><div class="line"><a name="l00028"></a><span class="lineno"> 28</span>  <span class="keyword">explicit</span> <a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html">GPUExecutor</a>(<span class="keywordtype">int</span> gpu_id) : gpu_id_(gpu_id) {}</div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span> </div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span>  ~<a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html">GPUExecutor</a>() {</div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span>  queue_.NoMoreJobs();</div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span>  thread_.join();</div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span>  }</div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span> </div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span>  <span class="keywordtype">void</span> RunJob(<a class="code" href="structcaffe2_1_1gpu__single__thread_1_1_task.html">Task</a>* task) {</div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span>  queue_.Push(task);</div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span>  }</div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span> </div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span>  <span class="keywordtype">void</span> start() {</div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span>  thread_ = std::thread(&GPUExecutor::WorkerFunction, <span class="keyword">this</span>);</div><div class="line"><a name="l00041"></a><span class="lineno"> 41</span>  }</div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span> </div><div class="line"><a name="l00043"></a><span class="lineno"> 43</span>  <span class="keyword">static</span> std::shared_ptr<GPUExecutor> Get(<span class="keywordtype">int</span> gpu);</div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span>  <span class="keyword">static</span> <span class="keywordtype">void</span> Release(<span class="keywordtype">int</span> gpu);</div><div class="line"><a name="l00045"></a><span class="lineno"> 45</span> </div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span>  <span class="keyword">private</span>:</div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span>  <span class="keywordtype">void</span> set_affinity();</div><div class="line"><a name="l00048"></a><span class="lineno"> 48</span>  <span class="keywordtype">void</span> WorkerFunction();</div><div class="line"><a name="l00049"></a><span class="lineno"> 49</span> </div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span>  std::thread thread_;</div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span>  <span class="keywordtype">int</span> gpu_id_;</div><div class="line"><a name="l00052"></a><span class="lineno"> 52</span>  <a class="code" href="classcaffe2_1_1_simple_queue.html">SimpleQueue<Task*></a> queue_;</div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span>  <span class="keyword">static</span> std::shared_ptr<GPUExecutor> executors_[CAFFE2_COMPILE_TIME_MAX_GPUS];</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>  <span class="keyword">static</span> std::mutex gpu_mtx_[CAFFE2_COMPILE_TIME_MAX_GPUS];</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span> };</div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span> </div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span> std::shared_ptr<GPUExecutor></div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span>  GPUExecutor::executors_[CAFFE2_COMPILE_TIME_MAX_GPUS];</div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span> std::mutex GPUExecutor::gpu_mtx_[CAFFE2_COMPILE_TIME_MAX_GPUS];</div><div class="line"><a name="l00060"></a><span class="lineno"> 60</span> </div><div class="line"><a name="l00061"></a><span class="lineno"> 61</span> std::shared_ptr<GPUExecutor> GPUExecutor::Get(<span class="keywordtype">int</span> gpu) {</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span>  std::lock_guard<std::mutex> grd(gpu_mtx_[gpu]);</div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span>  <span class="keywordflow">if</span> (!executors_[gpu].<span class="keyword">get</span>()) {</div><div class="line"><a name="l00064"></a><span class="lineno"> 64</span>  executors_[gpu].reset(<span class="keyword">new</span> <a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html">GPUExecutor</a>(gpu));</div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  executors_[gpu].get()->start();</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  }</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span>  <span class="keywordflow">return</span> executors_[gpu];</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span> }</div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span> </div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span> <span class="keywordtype">void</span> GPUExecutor::Release(<span class="keywordtype">int</span> gpu) {</div><div class="line"><a name="l00071"></a><span class="lineno"> 71</span>  std::lock_guard<std::mutex> grd(gpu_mtx_[gpu]);</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span>  <span class="keywordflow">if</span> (executors_[gpu].use_count() == 1) {</div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  executors_[gpu].reset();</div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  }</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span> }</div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span> </div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span> <span class="keywordtype">void</span> GPUExecutor::set_affinity() {</div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span> <span class="comment">// TODO: find a Windows-compatible affinity setting approach.</span></div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span> <span class="comment">// Currently, set_affinity has no effect in Windows. The code is still</span></div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span> <span class="comment">// correct with possible slowdowns.</span></div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span> <span class="preprocessor">#if !defined(_MSC_VER) && !defined(__APPLE__)</span></div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  <span class="comment">/* Set CPU affinity */</span></div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>  <span class="keywordtype">int</span> num_cores = std::thread::hardware_concurrency();</div><div class="line"><a name="l00084"></a><span class="lineno"> 84</span>  <span class="keywordflow">if</span> (num_cores > 0) {</div><div class="line"><a name="l00085"></a><span class="lineno"> 85</span>  cpu_set_t mask;</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  CPU_ZERO(&mask);</div><div class="line"><a name="l00087"></a><span class="lineno"> 87</span> </div><div class="line"><a name="l00088"></a><span class="lineno"> 88</span>  CPU_SET(gpu_id_ % num_cores, &mask);</div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  <span class="keywordflow">if</span> (sched_setaffinity(0, <span class="keyword">sizeof</span>(cpu_set_t), &mask)) {</div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>  LOG(WARNING) << <span class="stringliteral">"Could not set CPU affinity"</span>;</div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  }</div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  }</div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00094"></a><span class="lineno"> 94</span> }</div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span> </div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span> <span class="comment">// Worker that takes list of operators from the queue</span></div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span> <span class="comment">// and executes them.</span></div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span> <span class="keywordtype">void</span> GPUExecutor::WorkerFunction() {</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span>  <span class="keywordtype">int</span> stream_id_seq = 0;</div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  std::stack<int> streams;</div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>  set_affinity();</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span> </div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span>  <span class="keywordflow">while</span> (<span class="keyword">true</span>) {</div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span>  <a class="code" href="structcaffe2_1_1gpu__single__thread_1_1_task.html">Task</a>* task = <span class="keyword">nullptr</span>;</div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span>  vector<Task*> task_batch;</div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span> </div><div class="line"><a name="l00107"></a><span class="lineno"> 107</span>  <span class="keywordflow">if</span> (!queue_.Pop(&task)) {</div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span>  <span class="keywordflow">return</span>;</div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>  }</div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span>  <span class="keywordtype">int</span> num_tasks = 1 + queue_.size();</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span> </div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span>  <span class="comment">// Grab all tasks currently in queue so we can run them in parallel</span></div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>  <span class="comment">// Since we have only one producer, we know this does not block</span></div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span> </div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  <span class="comment">// TODO: launch ops in "zig-zag" manner so that we can start multiple</span></div><div class="line"><a name="l00116"></a><span class="lineno"> 116</span>  <span class="comment">// streams as simultaneously as possible</span></div><div class="line"><a name="l00117"></a><span class="lineno"> 117</span>  <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = num_tasks - 1; i >= 0; i--) {</div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  assert(task != <span class="keyword">nullptr</span>);</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  <span class="keywordflow">if</span> (streams.empty()) {</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  task->stream_id_ = stream_id_seq++;</div><div class="line"><a name="l00121"></a><span class="lineno"> 121</span>  } <span class="keywordflow">else</span> {</div><div class="line"><a name="l00122"></a><span class="lineno"> 122</span>  task->stream_id_ = streams.top();</div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  streams.pop();</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>  }</div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span> </div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <span class="keywordflow">for</span> (<span class="keyword">auto</span>& op : *task->ops_) {</div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>  op->RunAsync(task->stream_id_);</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>  }</div><div class="line"><a name="l00129"></a><span class="lineno"> 129</span>  task_batch.push_back(task);</div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span> </div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>  <span class="comment">// Get the next one</span></div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  <span class="keywordflow">if</span> (i > 0) {</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  <span class="keywordflow">if</span> (!queue_.Pop(&task)) {</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span>  <span class="keywordflow">return</span>;</div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>  }</div><div class="line"><a name="l00136"></a><span class="lineno"> 136</span>  }</div><div class="line"><a name="l00137"></a><span class="lineno"> 137</span>  }</div><div class="line"><a name="l00138"></a><span class="lineno"> 138</span> </div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span>  <span class="comment">// Wait for the currently executing streams</span></div><div class="line"><a name="l00140"></a><span class="lineno"> 140</span>  <span class="keywordflow">for</span> (<span class="keyword">auto</span>& pendtask : task_batch) {</div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span>  cudaStream_t stream =</div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span>  CUDAContext::cuda_stream(gpu_id_, pendtask->stream_id_);</div><div class="line"><a name="l00143"></a><span class="lineno"> 143</span>  CUDA_ENFORCE(cudaStreamSynchronize(stream));</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  streams.push(pendtask->stream_id_);</div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span>  std::unique_lock<std::mutex> lk(*pendtask->mtx_);</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>  pendtask->done_ = <span class="keyword">true</span>;</div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  pendtask->cv_->notify_one();</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>  }</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span>  }</div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span> }</div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span> </div><div class="line"><a name="l00152"></a><span class="lineno"><a class="line" href="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net.html"> 152</a></span> <span class="keyword">class </span><a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net.html">SingleThreadAsyncNet</a> : <span class="keyword">public</span> <a class="code" href="classcaffe2_1_1_simple_net.html">SimpleNet</a> {</div><div class="line"><a name="l00153"></a><span class="lineno"> 153</span>  <span class="keyword">public</span>:</div><div class="line"><a name="l00154"></a><span class="lineno"> 154</span>  <span class="keyword">using</span> SimpleNet::SimpleNet;</div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span> </div><div class="line"><a name="l00156"></a><span class="lineno"> 156</span>  ~<a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net.html">SingleThreadAsyncNet</a>() {</div><div class="line"><a name="l00157"></a><span class="lineno"> 157</span>  <span class="keywordflow">if</span> (executor_.get()) {</div><div class="line"><a name="l00158"></a><span class="lineno"> 158</span>  <span class="comment">// Explicitly reset my holding of the exeuctor so it can be</span></div><div class="line"><a name="l00159"></a><span class="lineno"> 159</span>  <span class="comment">// killed.</span></div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span>  executor_.reset();</div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span>  GPUExecutor::Release(gpu_id_);</div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  }</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  }</div><div class="line"><a name="l00164"></a><span class="lineno"> 164</span> </div><div class="line"><a name="l00165"></a><span class="lineno"> 165</span>  <span class="keywordtype">bool</span> Run()<span class="keyword"> override </span>{</div><div class="line"><a name="l00166"></a><span class="lineno"> 166</span>  <span class="keywordflow">if</span> (!executor_.get()) {</div><div class="line"><a name="l00167"></a><span class="lineno"> 167</span>  initialize();</div><div class="line"><a name="l00168"></a><span class="lineno"> 168</span>  }</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span> </div><div class="line"><a name="l00170"></a><span class="lineno"> 170</span>  <span class="comment">// Dispatch jobs to the gpu-specific executor thread</span></div><div class="line"><a name="l00171"></a><span class="lineno"> 171</span>  std::unique_lock<std::mutex> lk(mutex_);</div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  <a class="code" href="structcaffe2_1_1gpu__single__thread_1_1_task.html">Task</a> t;</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span>  t.ops_ = &operators_;</div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>  t.cv_ = &cv_;</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span>  t.mtx_ = &mutex_;</div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span>  t.done_ = <span class="keyword">false</span>;</div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>  executor_.get()->RunJob(&t);</div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span> </div><div class="line"><a name="l00179"></a><span class="lineno"> 179</span>  <span class="keywordflow">while</span> (!t.done_) {</div><div class="line"><a name="l00180"></a><span class="lineno"> 180</span>  cv_.wait(lk);</div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  }</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span> </div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</div><div class="line"><a name="l00184"></a><span class="lineno"> 184</span>  }</div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span> </div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>  <span class="keyword">private</span>:</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span>  std::condition_variable cv_;</div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span>  std::mutex mutex_;</div><div class="line"><a name="l00189"></a><span class="lineno"> 189</span> </div><div class="line"><a name="l00190"></a><span class="lineno"> 190</span>  <span class="keywordtype">void</span> initialize() {</div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  std::lock_guard<std::mutex> grd(mutex_);</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span> </div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  <span class="comment">/* Check the gpu id of this net and check that only one</span></div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span> <span class="comment"> GPU has operators on this net */</span></div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span>  gpu_id_ = (-1);</div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>  <span class="keywordflow">for</span> (<span class="keyword">auto</span>& op : operators_) {</div><div class="line"><a name="l00197"></a><span class="lineno"> 197</span>  <span class="keywordflow">if</span> (op->device_option().device_type() == CUDA) {</div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  <span class="keywordflow">if</span> (gpu_id_ < 0) {</div><div class="line"><a name="l00199"></a><span class="lineno"> 199</span>  gpu_id_ = op->device_option().cuda_gpu_id();</div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  } <span class="keywordflow">else</span> {</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  CAFFE_ENFORCE_EQ(</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span>  gpu_id_,</div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span>  op->device_option().cuda_gpu_id(),</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  <span class="stringliteral">"One net can only have operators for one GPU"</span>);</div><div class="line"><a name="l00205"></a><span class="lineno"> 205</span>  }</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span>  }</div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span>  }</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>  executor_ = GPUExecutor::Get(gpu_id_);</div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>  }</div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span> </div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span>  <span class="keywordtype">int</span> gpu_id_;</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>  std::shared_ptr<GPUExecutor> executor_;</div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span> };</div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span> </div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span> REGISTER_NET(singlethread_async, <a class="code" href="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net.html">SingleThreadAsyncNet</a>)</div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span> </div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span> } <span class="comment">// namespace gpu_single_thread</span></div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span> } <span class="comment">// namespace caffe2</span></div><div class="ttc" id="classcaffe2_1_1_simple_net_html"><div class="ttname"><a href="classcaffe2_1_1_simple_net.html">caffe2::SimpleNet</a></div><div class="ttdef"><b>Definition:</b> <a href="net__simple_8h_source.html#l00019">net_simple.h:19</a></div></div>
<div class="ttc" id="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor_html"><div class="ttname"><a href="classcaffe2_1_1gpu__single__thread_1_1_g_p_u_executor.html">caffe2::gpu_single_thread::GPUExecutor</a></div><div class="ttdef"><b>Definition:</b> <a href="net__singlethread__async__gpu_8cc_source.html#l00026">net_singlethread_async_gpu.cc:26</a></div></div>
<div class="ttc" id="namespacecaffe2_html"><div class="ttname"><a href="namespacecaffe2.html">caffe2</a></div><div class="ttdoc">A global dictionary that holds information about what Caffe2 modules have been loaded in the current ...</div><div class="ttdef"><b>Definition:</b> <a href="convert__encoded__to__raw__leveldb_8cc_source.html#l00047">convert_encoded_to_raw_leveldb.cc:47</a></div></div>
<div class="ttc" id="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net_html"><div class="ttname"><a href="classcaffe2_1_1gpu__single__thread_1_1_single_thread_async_net.html">caffe2::gpu_single_thread::SingleThreadAsyncNet</a></div><div class="ttdef"><b>Definition:</b> <a href="net__singlethread__async__gpu_8cc_source.html#l00152">net_singlethread_async_gpu.cc:152</a></div></div>
<div class="ttc" id="structcaffe2_1_1gpu__single__thread_1_1_task_html"><div class="ttname"><a href="structcaffe2_1_1gpu__single__thread_1_1_task.html">caffe2::gpu_single_thread::Task</a></div><div class="ttdef"><b>Definition:</b> <a href="net__singlethread__async__gpu_8cc_source.html#l00018">net_singlethread_async_gpu.cc:18</a></div></div>
<div class="ttc" id="classcaffe2_1_1_simple_queue_html"><div class="ttname"><a href="classcaffe2_1_1_simple_queue.html">caffe2::SimpleQueue</a></div><div class="ttdef"><b>Definition:</b> <a href="simple__queue_8h_source.html#l00022">simple_queue.h:22</a></div></div>
</div><!-- fragment --></div><!-- contents -->
<!-- HTML footer for doxygen 1.8.14-->
<!-- start footer part -->
<hr class="footer"/><address class="footer"><small>
Generated on Thu Apr 19 2018 13:03:50 for Caffe2 - C++ API by  <a href="http://www.doxygen.org/index.html">
<img class="footer" src="doxygen.png" alt="doxygen"/>
</a> 1.8.11
</small></address>
<div class="footerContainer">
<div id="footer_wrap" class="wrapper footerWrapper">
<div class="footerBlocks">
<div id="fb_oss" class="footerSection fbOpenSourceFooter">
<svg class="facebookOSSLogoSvg" viewBox="0 0 1133.9 1133.9" x="0px" y="0px" height=50 width=50>
<g>
<path class="logoRing outerRing" d="M 498.3 3.7 c 153.6 88.9 307.3 177.7 461.1 266.2 c 7.6 4.4 10.3 9.1 10.3 17.8 c -0.3 179.1 -0.2 358.3 0 537.4 c 0 8.1 -2.4 12.8 -9.7 17.1 c -154.5 88.9 -308.8 178.1 -462.9 267.5 c -9 5.2 -15.5 5.3 -24.6 0.1 c -153.9 -89.2 -307.9 -178 -462.1 -266.8 C 3 838.8 0 833.9 0 825.1 c 0.3 -179.1 0.2 -358.3 0 -537.4 c 0 -8.6 2.6 -13.6 10.2 -18 C 164.4 180.9 318.4 92 472.4 3 C 477 -1.5 494.3 -0.7 498.3 3.7 Z M 48.8 555.3 c 0 79.9 0.2 159.9 -0.2 239.8 c -0.1 10 3 15.6 11.7 20.6 c 137.2 78.8 274.2 157.8 411 237.3 c 9.9 5.7 17 5.7 26.8 0.1 c 137.5 -79.8 275.2 -159.2 412.9 -238.5 c 7.4 -4.3 10.5 -8.9 10.5 -17.8 c -0.3 -160.2 -0.3 -320.5 0 -480.7 c 0 -8.8 -2.8 -13.6 -10.3 -18 C 772.1 218 633.1 137.8 494.2 57.4 c -6.5 -3.8 -11.5 -4.5 -18.5 -0.5 C 336.8 137.4 197.9 217.7 58.8 297.7 c -7.7 4.4 -10.2 9.2 -10.2 17.9 C 48.9 395.5 48.8 475.4 48.8 555.3 Z" />
<path class="logoRing middleRing" d="M 184.4 555.9 c 0 -33.3 -1 -66.7 0.3 -100 c 1.9 -48 24.1 -86 64.7 -110.9 c 54.8 -33.6 110.7 -65.5 167 -96.6 c 45.7 -25.2 92.9 -24.7 138.6 1 c 54.4 30.6 108.7 61.5 162.2 93.7 c 44 26.5 67.3 66.8 68 118.4 c 0.9 63.2 0.9 126.5 0 189.7 c -0.7 50.6 -23.4 90.7 -66.6 116.9 c -55 33.4 -110.8 65.4 -167.1 96.5 c -43.4 24 -89 24.2 -132.3 0.5 c -57.5 -31.3 -114.2 -64 -170 -98.3 c -41 -25.1 -62.9 -63.7 -64.5 -112.2 C 183.5 621.9 184.3 588.9 184.4 555.9 Z M 232.9 556.3 c 0 29.5 0.5 59.1 -0.1 88.6 c -0.8 39.2 16.9 67.1 50.2 86.2 c 51.2 29.4 102.2 59.2 153.4 88.4 c 31.4 17.9 63.6 18.3 95 0.6 c 53.7 -30.3 107.1 -61.2 160.3 -92.5 c 29.7 -17.5 45 -44.5 45.3 -78.8 c 0.6 -61.7 0.5 -123.5 0 -185.2 c -0.3 -34.4 -15.3 -61.5 -44.9 -79 C 637.7 352.6 583 320.8 527.9 290 c -27.5 -15.4 -57.2 -16.1 -84.7 -0.7 c -56.9 31.6 -113.4 64 -169.1 97.6 c -26.4 15.9 -40.7 41.3 -41.1 72.9 C 232.6 491.9 232.9 524.1 232.9 556.3 Z" />
<path class="logoRing innerRing" d="M 484.9 424.4 c 69.8 -2.8 133.2 57.8 132.6 132 C 617 630 558.5 688.7 484.9 689.1 c -75.1 0.4 -132.6 -63.6 -132.7 -132.7 C 352.1 485 413.4 421.5 484.9 424.4 Z M 401.3 556.7 c -3.4 37.2 30.5 83.6 83 84.1 c 46.6 0.4 84.8 -37.6 84.9 -84 c 0.1 -46.6 -37.2 -84.4 -84.2 -84.6 C 432.2 472.1 397.9 518.3 401.3 556.7 Z" />
</g>
</svg>
<h2>Facebook Open Source</h2>
</div>
<div class="footerSection">
<a class="footerLink" href="https://code.facebook.com/projects/" target="_blank">Open Source Projects</a>
<a class="footerLink" href="https://github.com/facebook/" target="_blank">GitHub</a>
<a class="footerLink" href="https://twitter.com/fbOpenSource" target="_blank">Twitter</a>
</div>
<div class="footerSection rightAlign">
<a class="footerLink" href="https://github.com/caffe2/caffe2" target="_blank">Contribute to this project on GitHub</a>
</div>
</div>
</div>
</div>
<script type="text/javascript" src="/js/jekyll-link-anchors.js"></script>
<script>
(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
})(window,document,'script','//www.google-analytics.com/analytics.js','ga');
ga('create', '{{ site.gacode }}', 'auto');
ga('send', 'pageview');
</script>
</body>
</html>