Skip to content

F
faiss
Switch branch/tag
Find file
BlameHistoryPermalink
Transpose_8cuh_source.html 29.2 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 
<!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 1.8.5"/>
<title>Faiss: /tmp/faiss/gpu/utils/Transpose.cuh Source File</title>
<link href="tabs.css" rel="stylesheet" type="text/css"/>
<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/search.js"></script>
<script type="text/javascript">
  $(document).ready(function() { searchBox.OnSelectItem(0); });
</script>
<link href="doxygen.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 style="padding-left: 0.5em;">
   <div id="projectname">Faiss
   </div>
  </td>
 </tr>
 </tbody>
</table>
</div>
<!-- end header part -->
<!-- Generated by Doxygen 1.8.5 -->
<script type="text/javascript">
var searchBox = new SearchBox("searchBox", "search",false,'Search');
</script>
  <div id="navrow1" class="tabs">
    <ul class="tablist">
      <li><a href="index.html"><span>Main&#160;Page</span></a></li>
      <li><a href="namespaces.html"><span>Namespaces</span></a></li>
      <li><a href="annotated.html"><span>Classes</span></a></li>
      <li class="current"><a href="files.html"><span>Files</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&#160;List</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)">
<a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(0)"><span class="SelectionMark">&#160;</span>All</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(1)"><span class="SelectionMark">&#160;</span>Classes</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(2)"><span class="SelectionMark">&#160;</span>Namespaces</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(3)"><span class="SelectionMark">&#160;</span>Functions</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(4)"><span class="SelectionMark">&#160;</span>Variables</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(5)"><span class="SelectionMark">&#160;</span>Typedefs</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(6)"><span class="SelectionMark">&#160;</span>Enumerations</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(7)"><span class="SelectionMark">&#160;</span>Enumerator</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(8)"><span class="SelectionMark">&#160;</span>Friends</a></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_af0e5773a0d3761542ab6067d117c294.html">gpu</a></li><li class="navelem"><a class="el" href="dir_4b23abee0b4e5d71a28bc7475a9e97e1.html">utils</a></li>  </ul>
</div>
</div><!-- top -->
<div class="header">
  <div class="headertitle">
<div class="title">Transpose.cuh</div>  </div>
</div><!--header-->
<div class="contents">
<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="comment">/**</span></div>
<div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;<span class="comment"> * Copyright (c) 2015-present, Facebook, Inc.</span></div>
<div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="comment"> * All rights reserved.</span></div>
<div class="line"><a name="l00004"></a><span class="lineno">    4</span>&#160;<span class="comment"> *</span></div>
<div class="line"><a name="l00005"></a><span class="lineno">    5</span>&#160;<span class="comment"> * This source code is licensed under the BSD+Patents license found in the</span></div>
<div class="line"><a name="l00006"></a><span class="lineno">    6</span>&#160;<span class="comment"> * LICENSE file in the root directory of this source tree.</span></div>
<div class="line"><a name="l00007"></a><span class="lineno">    7</span>&#160;<span class="comment"> */</span></div>
<div class="line"><a name="l00008"></a><span class="lineno">    8</span>&#160;</div>
<div class="line"><a name="l00009"></a><span class="lineno">    9</span>&#160;</div>
<div class="line"><a name="l00010"></a><span class="lineno">   10</span>&#160;<span class="preprocessor">#pragma once</span></div>
<div class="line"><a name="l00011"></a><span class="lineno">   11</span>&#160;<span class="preprocessor"></span></div>
<div class="line"><a name="l00012"></a><span class="lineno">   12</span>&#160;<span class="preprocessor">#include &quot;../../FaissAssert.h&quot;</span></div>
<div class="line"><a name="l00013"></a><span class="lineno">   13</span>&#160;<span class="preprocessor">#include &quot;Tensor.cuh&quot;</span></div>
<div class="line"><a name="l00014"></a><span class="lineno">   14</span>&#160;<span class="preprocessor">#include &quot;DeviceUtils.h&quot;</span></div>
<div class="line"><a name="l00015"></a><span class="lineno">   15</span>&#160;<span class="preprocessor">#include &lt;cuda.h&gt;</span></div>
<div class="line"><a name="l00016"></a><span class="lineno">   16</span>&#160;</div>
<div class="line"><a name="l00017"></a><span class="lineno">   17</span>&#160;<span class="preprocessor">#include &lt;stdio.h&gt;</span></div>
<div class="line"><a name="l00018"></a><span class="lineno">   18</span>&#160;</div>
<div class="line"><a name="l00019"></a><span class="lineno">   19</span>&#160;<span class="keyword">namespace </span>faiss { <span class="keyword">namespace </span>gpu {</div>
<div class="line"><a name="l00020"></a><span class="lineno">   20</span>&#160;</div>
<div class="line"><a name="l00021"></a><span class="lineno">   21</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keyword">typename</span> IndexT&gt;</div>
<div class="line"><a name="l00022"></a><span class="lineno"><a class="line" href="structfaiss_1_1gpu_1_1TensorInfo.html">   22</a></span>&#160;<span class="keyword">struct </span><a class="code" href="structfaiss_1_1gpu_1_1TensorInfo.html">TensorInfo</a> {</div>
<div class="line"><a name="l00023"></a><span class="lineno">   23</span>&#160;  <span class="keyword">static</span> constexpr <span class="keywordtype">int</span> kMaxDims = 8;</div>
<div class="line"><a name="l00024"></a><span class="lineno">   24</span>&#160;</div>
<div class="line"><a name="l00025"></a><span class="lineno">   25</span>&#160;  T* data;</div>
<div class="line"><a name="l00026"></a><span class="lineno">   26</span>&#160;  IndexT sizes[kMaxDims];</div>
<div class="line"><a name="l00027"></a><span class="lineno">   27</span>&#160;  IndexT strides[kMaxDims];</div>
<div class="line"><a name="l00028"></a><span class="lineno">   28</span>&#160;  <span class="keywordtype">int</span> dims;</div>
<div class="line"><a name="l00029"></a><span class="lineno">   29</span>&#160;};</div>
<div class="line"><a name="l00030"></a><span class="lineno">   30</span>&#160;</div>
<div class="line"><a name="l00031"></a><span class="lineno">   31</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keyword">typename</span> IndexT, <span class="keywordtype">int</span> Dim&gt;</div>
<div class="line"><a name="l00032"></a><span class="lineno"><a class="line" href="structfaiss_1_1gpu_1_1TensorInfoOffset.html">   32</a></span>&#160;<span class="keyword">struct </span><a class="code" href="structfaiss_1_1gpu_1_1TensorInfoOffset.html">TensorInfoOffset</a> {</div>
<div class="line"><a name="l00033"></a><span class="lineno">   33</span>&#160;  __device__ <span class="keyword">inline</span> <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <span class="keyword">get</span>(<span class="keyword">const</span> <a class="code" href="structfaiss_1_1gpu_1_1TensorInfo.html">TensorInfo&lt;T, IndexT&gt;</a>&amp; info,</div>
<div class="line"><a name="l00034"></a><span class="lineno">   34</span>&#160;                                            IndexT linearId) {</div>
<div class="line"><a name="l00035"></a><span class="lineno">   35</span>&#160;    IndexT offset = 0;</div>
<div class="line"><a name="l00036"></a><span class="lineno">   36</span>&#160;</div>
<div class="line"><a name="l00037"></a><span class="lineno">   37</span>&#160;<span class="preprocessor">#pragma unroll</span></div>
<div class="line"><a name="l00038"></a><span class="lineno">   38</span>&#160;<span class="preprocessor"></span>    <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = Dim - 1; i &gt;= 0; --i) {</div>
<div class="line"><a name="l00039"></a><span class="lineno">   39</span>&#160;      IndexT curDimIndex = linearId % info.sizes[i];</div>
<div class="line"><a name="l00040"></a><span class="lineno">   40</span>&#160;      IndexT curDimOffset = curDimIndex * info.strides[i];</div>
<div class="line"><a name="l00041"></a><span class="lineno">   41</span>&#160;</div>
<div class="line"><a name="l00042"></a><span class="lineno">   42</span>&#160;      offset += curDimOffset;</div>
<div class="line"><a name="l00043"></a><span class="lineno">   43</span>&#160;</div>
<div class="line"><a name="l00044"></a><span class="lineno">   44</span>&#160;      <span class="keywordflow">if</span> (i &gt; 0) {</div>
<div class="line"><a name="l00045"></a><span class="lineno">   45</span>&#160;        linearId /= info.sizes[i];</div>
<div class="line"><a name="l00046"></a><span class="lineno">   46</span>&#160;      }</div>
<div class="line"><a name="l00047"></a><span class="lineno">   47</span>&#160;    }</div>
<div class="line"><a name="l00048"></a><span class="lineno">   48</span>&#160;</div>
<div class="line"><a name="l00049"></a><span class="lineno">   49</span>&#160;    <span class="keywordflow">return</span> offset;</div>
<div class="line"><a name="l00050"></a><span class="lineno">   50</span>&#160;  }</div>
<div class="line"><a name="l00051"></a><span class="lineno">   51</span>&#160;};</div>
<div class="line"><a name="l00052"></a><span class="lineno">   52</span>&#160;</div>
<div class="line"><a name="l00053"></a><span class="lineno">   53</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keyword">typename</span> IndexT&gt;</div>
<div class="line"><a name="l00054"></a><span class="lineno"><a class="line" href="structfaiss_1_1gpu_1_1TensorInfoOffset_3_01T_00_01IndexT_00-1_01_4.html">   54</a></span>&#160;<span class="keyword">struct </span><a class="code" href="structfaiss_1_1gpu_1_1TensorInfoOffset.html">TensorInfoOffset</a>&lt;T, IndexT, -1&gt; {</div>
<div class="line"><a name="l00055"></a><span class="lineno">   55</span>&#160;  __device__ <span class="keyword">inline</span> <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <span class="keyword">get</span>(<span class="keyword">const</span> <a class="code" href="structfaiss_1_1gpu_1_1TensorInfo.html">TensorInfo&lt;T, IndexT&gt;</a>&amp; info,</div>
<div class="line"><a name="l00056"></a><span class="lineno">   56</span>&#160;                                            IndexT linearId) {</div>
<div class="line"><a name="l00057"></a><span class="lineno">   57</span>&#160;    <span class="keywordflow">return</span> linearId;</div>
<div class="line"><a name="l00058"></a><span class="lineno">   58</span>&#160;  }</div>
<div class="line"><a name="l00059"></a><span class="lineno">   59</span>&#160;};</div>
<div class="line"><a name="l00060"></a><span class="lineno">   60</span>&#160;</div>
<div class="line"><a name="l00061"></a><span class="lineno">   61</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keyword">typename</span> IndexT, <span class="keywordtype">int</span> Dim&gt;</div>
<div class="line"><a name="l00062"></a><span class="lineno">   62</span>&#160;<a class="code" href="structfaiss_1_1gpu_1_1TensorInfo.html">TensorInfo&lt;T, IndexT&gt;</a> getTensorInfo(<span class="keyword">const</span> <a class="code" href="classfaiss_1_1gpu_1_1Tensor.html">Tensor&lt;T, Dim, true&gt;</a>&amp; t) {</div>
<div class="line"><a name="l00063"></a><span class="lineno">   63</span>&#160;  <a class="code" href="structfaiss_1_1gpu_1_1TensorInfo.html">TensorInfo&lt;T, IndexT&gt;</a> info;</div>
<div class="line"><a name="l00064"></a><span class="lineno">   64</span>&#160;</div>
<div class="line"><a name="l00065"></a><span class="lineno">   65</span>&#160;  <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 0; i &lt; Dim; ++i) {</div>
<div class="line"><a name="l00066"></a><span class="lineno">   66</span>&#160;    info.sizes[i] = (IndexT) t.<a class="code" href="classfaiss_1_1gpu_1_1Tensor.html#a6699c311648457f257afa340c61f417c">getSize</a>(i);</div>
<div class="line"><a name="l00067"></a><span class="lineno">   67</span>&#160;    info.strides[i] = (IndexT) t.<a class="code" href="classfaiss_1_1gpu_1_1Tensor.html#a0b8bba630f7a1fa217f90b20d298420a">getStride</a>(i);</div>
<div class="line"><a name="l00068"></a><span class="lineno">   68</span>&#160;  }</div>
<div class="line"><a name="l00069"></a><span class="lineno">   69</span>&#160;</div>
<div class="line"><a name="l00070"></a><span class="lineno">   70</span>&#160;  info.data = t.<a class="code" href="classfaiss_1_1gpu_1_1Tensor.html#a50411ce4d0fa32ef715e3321b6e33212">data</a>();</div>
<div class="line"><a name="l00071"></a><span class="lineno">   71</span>&#160;  info.dims = Dim;</div>
<div class="line"><a name="l00072"></a><span class="lineno">   72</span>&#160;</div>
<div class="line"><a name="l00073"></a><span class="lineno">   73</span>&#160;  <span class="keywordflow">return</span> info;</div>
<div class="line"><a name="l00074"></a><span class="lineno">   74</span>&#160;}</div>
<div class="line"><a name="l00075"></a><span class="lineno">   75</span>&#160;</div>
<div class="line"><a name="l00076"></a><span class="lineno">   76</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keyword">typename</span> IndexT, <span class="keywordtype">int</span> DimInput, <span class="keywordtype">int</span> DimOutput&gt;</div>
<div class="line"><a name="l00077"></a><span class="lineno">   77</span>&#160;__global__ <span class="keywordtype">void</span> transposeAny(TensorInfo&lt;T, IndexT&gt; input,</div>
<div class="line"><a name="l00078"></a><span class="lineno">   78</span>&#160;                             TensorInfo&lt;T, IndexT&gt; output,</div>
<div class="line"><a name="l00079"></a><span class="lineno">   79</span>&#160;                             IndexT totalSize) {</div>
<div class="line"><a name="l00080"></a><span class="lineno">   80</span>&#160;  <span class="keywordflow">for</span> (IndexT i = blockIdx.x * blockDim.x + threadIdx.x;</div>
<div class="line"><a name="l00081"></a><span class="lineno">   81</span>&#160;       i &lt; totalSize;</div>
<div class="line"><a name="l00082"></a><span class="lineno">   82</span>&#160;       i += gridDim.x + blockDim.x) {</div>
<div class="line"><a name="l00083"></a><span class="lineno">   83</span>&#160;    <span class="keyword">auto</span> inputOffset = TensorInfoOffset&lt;T, IndexT, DimInput&gt;::get(input, i);</div>
<div class="line"><a name="l00084"></a><span class="lineno">   84</span>&#160;    <span class="keyword">auto</span> outputOffset = TensorInfoOffset&lt;T, IndexT, DimOutput&gt;::get(output, i);</div>
<div class="line"><a name="l00085"></a><span class="lineno">   85</span>&#160;</div>
<div class="line"><a name="l00086"></a><span class="lineno">   86</span>&#160;<span class="preprocessor">#if __CUDA_ARCH__ &gt;= 350</span></div>
<div class="line"><a name="l00087"></a><span class="lineno">   87</span>&#160;<span class="preprocessor"></span>    output.data[outputOffset] = __ldg(&amp;input.data[inputOffset]);</div>
<div class="line"><a name="l00088"></a><span class="lineno">   88</span>&#160;<span class="preprocessor">#else</span></div>
<div class="line"><a name="l00089"></a><span class="lineno">   89</span>&#160;<span class="preprocessor"></span>    output.data[outputOffset] = input.data[inputOffset];</div>
<div class="line"><a name="l00090"></a><span class="lineno">   90</span>&#160;<span class="preprocessor">#endif</span></div>
<div class="line"><a name="l00091"></a><span class="lineno">   91</span>&#160;<span class="preprocessor"></span>  }</div>
<div class="line"><a name="l00092"></a><span class="lineno">   92</span>&#160;}</div>
<div class="line"><a name="l00093"></a><span class="lineno">   93</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00094"></a><span class="lineno">   94</span>&#160;<span class="comment">/// Performs an out-of-place transposition between any two dimensions.</span></div>
<div class="line"><a name="l00095"></a><span class="lineno">   95</span>&#160;<span class="comment">/// Best performance is if the transposed dimensions are not</span></div>
<div class="line"><a name="l00096"></a><span class="lineno">   96</span>&#160;<span class="comment">/// innermost, since the reads and writes will be coalesced.</span></div>
<div class="line"><a name="l00097"></a><span class="lineno">   97</span>&#160;<span class="comment">/// Could include a shared memory transposition if the dimensions</span></div>
<div class="line"><a name="l00098"></a><span class="lineno">   98</span>&#160;<span class="comment">/// being transposed are innermost, but would require support for</span></div>
<div class="line"><a name="l00099"></a><span class="lineno">   99</span>&#160;<span class="comment">/// arbitrary rectangular matrices.</span></div>
<div class="line"><a name="l00100"></a><span class="lineno">  100</span>&#160;<span class="comment">/// This linearized implementation seems to perform well enough,</span></div>
<div class="line"><a name="l00101"></a><span class="lineno">  101</span>&#160;<span class="comment">/// especially for cases that we care about (outer dimension</span></div>
<div class="line"><a name="l00102"></a><span class="lineno">  102</span>&#160;<span class="comment">/// transpositions).</span></div>
<div class="line"><a name="l00103"></a><span class="lineno">  103</span>&#160;<span class="comment"></span><span class="keyword">template</span> &lt;<span class="keyword">typename</span> T, <span class="keywordtype">int</span> Dim&gt;</div>
<div class="line"><a name="l00104"></a><span class="lineno">  104</span>&#160;<span class="keywordtype">void</span> runTransposeAny(Tensor&lt;T, Dim, true&gt;&amp; in,</div>
<div class="line"><a name="l00105"></a><span class="lineno">  105</span>&#160;                     <span class="keywordtype">int</span> dim1, <span class="keywordtype">int</span> dim2,</div>
<div class="line"><a name="l00106"></a><span class="lineno">  106</span>&#160;                     Tensor&lt;T, Dim, true&gt;&amp; out,</div>
<div class="line"><a name="l00107"></a><span class="lineno">  107</span>&#160;                     cudaStream_t stream) {</div>
<div class="line"><a name="l00108"></a><span class="lineno">  108</span>&#160;  static_assert(Dim &lt;= TensorInfo&lt;T, unsigned int&gt;::kMaxDims,</div>
<div class="line"><a name="l00109"></a><span class="lineno">  109</span>&#160;                <span class="stringliteral">&quot;too many dimensions&quot;</span>);</div>
<div class="line"><a name="l00110"></a><span class="lineno">  110</span>&#160;</div>
<div class="line"><a name="l00111"></a><span class="lineno">  111</span>&#160;  FAISS_ASSERT(dim1 != dim2);</div>
<div class="line"><a name="l00112"></a><span class="lineno">  112</span>&#160;  FAISS_ASSERT(dim1 &lt; Dim &amp;&amp; dim2 &lt; Dim);</div>
<div class="line"><a name="l00113"></a><span class="lineno">  113</span>&#160;</div>
<div class="line"><a name="l00114"></a><span class="lineno">  114</span>&#160;  <span class="keywordtype">int</span> outSize[Dim];</div>
<div class="line"><a name="l00115"></a><span class="lineno">  115</span>&#160;</div>
<div class="line"><a name="l00116"></a><span class="lineno">  116</span>&#160;  <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 0; i &lt; Dim; ++i) {</div>
<div class="line"><a name="l00117"></a><span class="lineno">  117</span>&#160;    outSize[i] = in.getSize(i);</div>
<div class="line"><a name="l00118"></a><span class="lineno">  118</span>&#160;  }</div>
<div class="line"><a name="l00119"></a><span class="lineno">  119</span>&#160;</div>
<div class="line"><a name="l00120"></a><span class="lineno">  120</span>&#160;  std::swap(outSize[dim1], outSize[dim2]);</div>
<div class="line"><a name="l00121"></a><span class="lineno">  121</span>&#160;</div>
<div class="line"><a name="l00122"></a><span class="lineno">  122</span>&#160;  <span class="keywordflow">for</span> (<span class="keywordtype">int</span> i = 0; i &lt; Dim; ++i) {</div>
<div class="line"><a name="l00123"></a><span class="lineno">  123</span>&#160;    FAISS_ASSERT(out.getSize(i) == outSize[i]);</div>
<div class="line"><a name="l00124"></a><span class="lineno">  124</span>&#160;  }</div>
<div class="line"><a name="l00125"></a><span class="lineno">  125</span>&#160;</div>
<div class="line"><a name="l00126"></a><span class="lineno">  126</span>&#160;  <span class="keywordtype">size_t</span> totalSize = in.numElements();</div>
<div class="line"><a name="l00127"></a><span class="lineno">  127</span>&#160;  <span class="keywordtype">size_t</span> block = std::min((<span class="keywordtype">size_t</span>) getMaxThreadsCurrentDevice(), totalSize);</div>
<div class="line"><a name="l00128"></a><span class="lineno">  128</span>&#160;</div>
<div class="line"><a name="l00129"></a><span class="lineno">  129</span>&#160;  <span class="keywordflow">if</span> (totalSize &lt;= (<span class="keywordtype">size_t</span>) std::numeric_limits&lt;int&gt;::max()) {</div>
<div class="line"><a name="l00130"></a><span class="lineno">  130</span>&#160;    <span class="comment">// div/mod seems faster with unsigned types</span></div>
<div class="line"><a name="l00131"></a><span class="lineno">  131</span>&#160;    <span class="keyword">auto</span> inInfo = getTensorInfo&lt;T, unsigned int, Dim&gt;(in);</div>
<div class="line"><a name="l00132"></a><span class="lineno">  132</span>&#160;    <span class="keyword">auto</span> outInfo = getTensorInfo&lt;T, unsigned int, Dim&gt;(out);</div>
<div class="line"><a name="l00133"></a><span class="lineno">  133</span>&#160;</div>
<div class="line"><a name="l00134"></a><span class="lineno">  134</span>&#160;    std::swap(inInfo.sizes[dim1], inInfo.sizes[dim2]);</div>
<div class="line"><a name="l00135"></a><span class="lineno">  135</span>&#160;    std::swap(inInfo.strides[dim1], inInfo.strides[dim2]);</div>
<div class="line"><a name="l00136"></a><span class="lineno">  136</span>&#160;</div>
<div class="line"><a name="l00137"></a><span class="lineno">  137</span>&#160;    <span class="keyword">auto</span> grid = std::min(utils::divUp(totalSize, block), (<span class="keywordtype">size_t</span>) 4096);</div>
<div class="line"><a name="l00138"></a><span class="lineno">  138</span>&#160;</div>
<div class="line"><a name="l00139"></a><span class="lineno">  139</span>&#160;    transposeAny&lt;T, <span class="keywordtype">unsigned</span> int, Dim, -1&gt;</div>
<div class="line"><a name="l00140"></a><span class="lineno">  140</span>&#160;      &lt;&lt;&lt;grid, block, 0, stream&gt;&gt;&gt;(inInfo, outInfo, totalSize);</div>
<div class="line"><a name="l00141"></a><span class="lineno">  141</span>&#160;  } <span class="keywordflow">else</span> {</div>
<div class="line"><a name="l00142"></a><span class="lineno">  142</span>&#160;    <span class="keyword">auto</span> inInfo = getTensorInfo&lt;T, unsigned long, Dim&gt;(in);</div>
<div class="line"><a name="l00143"></a><span class="lineno">  143</span>&#160;    <span class="keyword">auto</span> outInfo = getTensorInfo&lt;T, unsigned long, Dim&gt;(out);</div>
<div class="line"><a name="l00144"></a><span class="lineno">  144</span>&#160;</div>
<div class="line"><a name="l00145"></a><span class="lineno">  145</span>&#160;    std::swap(inInfo.sizes[dim1], inInfo.sizes[dim2]);</div>
<div class="line"><a name="l00146"></a><span class="lineno">  146</span>&#160;    std::swap(inInfo.strides[dim1], inInfo.strides[dim2]);</div>
<div class="line"><a name="l00147"></a><span class="lineno">  147</span>&#160;</div>
<div class="line"><a name="l00148"></a><span class="lineno">  148</span>&#160;    <span class="keyword">auto</span> grid = std::min(utils::divUp(totalSize, block), (<span class="keywordtype">size_t</span>) 4096);</div>
<div class="line"><a name="l00149"></a><span class="lineno">  149</span>&#160;</div>
<div class="line"><a name="l00150"></a><span class="lineno">  150</span>&#160;    transposeAny&lt;T, <span class="keywordtype">unsigned</span> long, Dim, -1&gt;</div>
<div class="line"><a name="l00151"></a><span class="lineno">  151</span>&#160;      &lt;&lt;&lt;grid, block, 0, stream&gt;&gt;&gt;(inInfo, outInfo, totalSize);</div>
<div class="line"><a name="l00152"></a><span class="lineno">  152</span>&#160;  }</div>
<div class="line"><a name="l00153"></a><span class="lineno">  153</span>&#160;  CUDA_TEST_ERROR();</div>
<div class="line"><a name="l00154"></a><span class="lineno">  154</span>&#160;}</div>
<div class="line"><a name="l00155"></a><span class="lineno">  155</span>&#160;</div>
<div class="line"><a name="l00156"></a><span class="lineno">  156</span>&#160;} } <span class="comment">// namespace</span></div>
<div class="ttc" id="structfaiss_1_1gpu_1_1TensorInfo_html"><div class="ttname"><a href="structfaiss_1_1gpu_1_1TensorInfo.html">faiss::gpu::TensorInfo</a></div><div class="ttdef"><b>Definition:</b> <a href="Transpose_8cuh_source.html#l00022">Transpose.cuh:22</a></div></div>
<div class="ttc" id="classfaiss_1_1gpu_1_1Tensor_html_a6699c311648457f257afa340c61f417c"><div class="ttname"><a href="classfaiss_1_1gpu_1_1Tensor.html#a6699c311648457f257afa340c61f417c">faiss::gpu::Tensor::getSize</a></div><div class="ttdeci">__host__ __device__ IndexT getSize(int i) const </div><div class="ttdef"><b>Definition:</b> <a href="Tensor_8cuh_source.html#l00223">Tensor.cuh:223</a></div></div>
<div class="ttc" id="classfaiss_1_1gpu_1_1Tensor_html_a50411ce4d0fa32ef715e3321b6e33212"><div class="ttname"><a href="classfaiss_1_1gpu_1_1Tensor.html#a50411ce4d0fa32ef715e3321b6e33212">faiss::gpu::Tensor::data</a></div><div class="ttdeci">__host__ __device__ DataPtrType data()</div><div class="ttdoc">Returns a raw pointer to the start of our data. </div><div class="ttdef"><b>Definition:</b> <a href="Tensor_8cuh_source.html#l00175">Tensor.cuh:175</a></div></div>
<div class="ttc" id="classfaiss_1_1gpu_1_1Tensor_html"><div class="ttname"><a href="classfaiss_1_1gpu_1_1Tensor.html">faiss::gpu::Tensor</a></div><div class="ttdoc">Our tensor type. </div><div class="ttdef"><b>Definition:</b> <a href="Tensor_8cuh_source.html#l00029">Tensor.cuh:29</a></div></div>
<div class="ttc" id="classfaiss_1_1gpu_1_1Tensor_html_a0b8bba630f7a1fa217f90b20d298420a"><div class="ttname"><a href="classfaiss_1_1gpu_1_1Tensor.html#a0b8bba630f7a1fa217f90b20d298420a">faiss::gpu::Tensor::getStride</a></div><div class="ttdeci">__host__ __device__ IndexT getStride(int i) const </div><div class="ttdef"><b>Definition:</b> <a href="Tensor_8cuh_source.html#l00229">Tensor.cuh:229</a></div></div>
<div class="ttc" id="structfaiss_1_1gpu_1_1TensorInfoOffset_html"><div class="ttname"><a href="structfaiss_1_1gpu_1_1TensorInfoOffset.html">faiss::gpu::TensorInfoOffset</a></div><div class="ttdef"><b>Definition:</b> <a href="Transpose_8cuh_source.html#l00032">Transpose.cuh:32</a></div></div>
</div><!-- fragment --></div><!-- contents -->
<!-- start footer part -->
<hr class="footer"/><address class="footer"><small>
Generated by &#160;<a href="http://www.doxygen.org/index.html">
<img class="footer" src="doxygen.png" alt="doxygen"/>
</a> 1.8.5
</small></address>
</body>
</html>