cutlass/docs/device_2tensor__compare_8h_...

118 lines
38 KiB
HTML

<!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"/>
<title>CUTLASS: tensor_compare.h 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/searchdata.js"></script>
<script type="text/javascript" src="search/search.js"></script>
<script type="text/javascript">
$(document).ready(function() { init_search(); });
</script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
extensions: ["tex2jax.js"],
jax: ["input/TeX","output/HTML-CSS"],
});
</script><script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></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 id="projectlogo"><img alt="Logo" src="cutlass-logo-small.png"/></td>
<td id="projectalign" style="padding-left: 0.5em;">
<div id="projectname">CUTLASS
</div>
<div id="projectbrief">CUDA Templates for Linear Algebra Subroutines and Solvers</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="index.html"><span>Main&#160;Page</span></a></li>
<li><a href="modules.html"><span>Modules</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>
<li><a href="globals.html"><span>File&#160;Members</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_4eeb864c4eec08c7d6b9d3b0352cfdde.html">tools</a></li><li class="navelem"><a class="el" href="dir_88de82f9e8d739a2f42f92d95f0d7933.html">util</a></li><li class="navelem"><a class="el" href="dir_7e9e609009df72bf6226de354e72c328.html">include</a></li><li class="navelem"><a class="el" href="dir_ade2f6ff57439d30f4164e14e54bcf30.html">cutlass</a></li><li class="navelem"><a class="el" href="dir_ff60863f958a43c892071bb1f8a4c81a.html">util</a></li><li class="navelem"><a class="el" href="dir_01de8928c960cafb028e5f164701e1de.html">reference</a></li><li class="navelem"><a class="el" href="dir_ebbbb6f6f10686db77ac27d0af6d8201.html">device</a></li> </ul>
</div>
</div><!-- top -->
<div class="header">
<div class="headertitle">
<div class="title">device/tensor_compare.h</div> </div>
</div><!--header-->
<div class="contents">
<a href="device_2tensor__compare_8h.html">Go to the documentation of this file.</a><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) 2017-2019, NVIDIA CORPORATION. All rights reserved.</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span>&#160;<span class="comment"> *</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span>&#160;<span class="comment"> * Redistribution and use in source and binary forms, with or without modification, are permitted</span></div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span>&#160;<span class="comment"> * provided that the following conditions are met:</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span>&#160;<span class="comment"> * * Redistributions of source code must retain the above copyright notice, this list of</span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span>&#160;<span class="comment"> * conditions and the following disclaimer.</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span>&#160;<span class="comment"> * * Redistributions in binary form must reproduce the above copyright notice, this list of</span></div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span>&#160;<span class="comment"> * conditions and the following disclaimer in the documentation and/or other materials</span></div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span>&#160;<span class="comment"> * provided with the distribution.</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span>&#160;<span class="comment"> * * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used</span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span>&#160;<span class="comment"> * to endorse or promote products derived from this software without specific prior written</span></div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span>&#160;<span class="comment"> * permission.</span></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span>&#160;<span class="comment"> *</span></div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span>&#160;<span class="comment"> * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS &quot;AS IS&quot; AND ANY EXPRESS OR</span></div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span>&#160;<span class="comment"> * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND</span></div><div class="line"><a name="l00017"></a><span class="lineno"> 17</span>&#160;<span class="comment"> * FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE</span></div><div class="line"><a name="l00018"></a><span class="lineno"> 18</span>&#160;<span class="comment"> * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,</span></div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span>&#160;<span class="comment"> * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;</span></div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span>&#160;<span class="comment"> * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,</span></div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span>&#160;<span class="comment"> * STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE</span></div><div class="line"><a name="l00022"></a><span class="lineno"> 22</span>&#160;<span class="comment"> * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.</span></div><div class="line"><a name="l00023"></a><span class="lineno"> 23</span>&#160;<span class="comment"> *</span></div><div class="line"><a name="l00024"></a><span class="lineno"> 24</span>&#160;<span class="comment"> **************************************************************************************************/</span></div><div class="line"><a name="l00025"></a><span class="lineno"> 25</span>&#160;<span class="comment">/* \file</span></div><div class="line"><a name="l00026"></a><span class="lineno"> 26</span>&#160;<span class="comment"> \brief Defines host-side elementwise operations on TensorView.</span></div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span>&#160;<span class="comment">*/</span></div><div class="line"><a name="l00028"></a><span class="lineno"> 28</span>&#160;</div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span>&#160;<span class="preprocessor">#pragma once</span></div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span>&#160;<span class="comment">// Standard Library includes</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span>&#160;<span class="preprocessor">#include &lt;utility&gt;</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span>&#160;</div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span>&#160;<span class="comment">// Cutlass includes</span></div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span>&#160;<span class="preprocessor">#include &quot;<a class="code" href="cutlass_8h.html">cutlass/cutlass.h</a>&quot;</span></div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span>&#160;<span class="preprocessor">#include &quot;<a class="code" href="relatively__equal_8h.html">cutlass/relatively_equal.h</a>&quot;</span></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">#include &quot;<a class="code" href="distribution_8h.html">cutlass/util/distribution.h</a>&quot;</span></div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span>&#160;</div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span>&#160;<span class="preprocessor">#include &quot;<a class="code" href="device_2tensor__foreach_8h.html">tensor_foreach.h</a>&quot;</span></div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span>&#160;</div><div class="line"><a name="l00041"></a><span class="lineno"> 41</span>&#160;<span class="keyword">namespace </span><a class="code" href="namespacecutlass.html">cutlass</a> {</div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span>&#160;<span class="keyword">namespace </span>reference {</div><div class="line"><a name="l00043"></a><span class="lineno"> 43</span>&#160;<span class="keyword">namespace </span>device {</div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span>&#160;</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;<span class="keyword">namespace </span>kernel {</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="keyword">template</span> &lt;<span class="keyword">typename</span> Element&gt;</div><div class="line"><a name="l00050"></a><span class="lineno"><a class="line" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a4595ede72eddace3c973c7f0f74b001d"> 50</a></span>&#160;__global__ <span class="keywordtype">void</span> <a class="code" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a4595ede72eddace3c973c7f0f74b001d">BlockCompareEqual</a>(</div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span>&#160; <span class="keywordtype">int</span> *equal, </div><div class="line"><a name="l00052"></a><span class="lineno"> 52</span>&#160; Element <span class="keyword">const</span> *ptr_A,</div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span>&#160; Element <span class="keyword">const</span> *ptr_B,</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>&#160; <span class="keywordtype">size_t</span> capacity) {</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span>&#160;</div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span>&#160; <span class="keywordtype">size_t</span> idx = threadIdx.x + blockDim.x * blockIdx.x;</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>&#160;</div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span>&#160; <span class="keywordflow">for</span> (; idx &lt; capacity; idx += gridDim.x * blockDim.x) {</div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span>&#160; <span class="keywordflow">if</span> (ptr_A[idx] != ptr_B[idx]) {</div><div class="line"><a name="l00060"></a><span class="lineno"> 60</span>&#160; *equal = 0;</div><div class="line"><a name="l00061"></a><span class="lineno"> 61</span>&#160; <span class="keywordflow">return</span>;</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span>&#160; }</div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span>&#160; }</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;</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> Element&gt;</div><div class="line"><a name="l00067"></a><span class="lineno"><a class="line" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a6da13fb683d56d6973af0a97a4023677"> 67</a></span>&#160;__global__ <span class="keywordtype">void</span> <a class="code" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a6da13fb683d56d6973af0a97a4023677">BlockCompareRelativelyEqual</a>(</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>&#160; <span class="keywordtype">int</span> *equal, </div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>&#160; Element <span class="keyword">const</span> *ptr_A,</div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span>&#160; Element <span class="keyword">const</span> *ptr_B,</div><div class="line"><a name="l00071"></a><span class="lineno"> 71</span>&#160; <span class="keywordtype">size_t</span> capacity,</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span>&#160; Element epsilon,</div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>&#160; Element nonzero_floor) {</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; <span class="keywordtype">size_t</span> idx = threadIdx.x + blockDim.x * blockIdx.x;</div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span>&#160;</div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span>&#160; <span class="keywordflow">for</span> (; idx &lt; capacity; idx += gridDim.x * blockDim.x) {</div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>&#160;</div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span>&#160; Element a = ptr_A[idx];</div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span>&#160; Element b = ptr_B[idx];</div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>&#160;</div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>&#160; <span class="keywordflow">if</span> (!<a class="code" href="namespacecutlass.html#aebc97e25721b38692159a6d50d545fe0">relatively_equal</a>(a, b, epsilon, nonzero_floor)) {</div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>&#160; *equal = 0;</div><div class="line"><a name="l00084"></a><span class="lineno"> 84</span>&#160; <span class="keywordflow">return</span>;</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; }</div><div class="line"><a name="l00087"></a><span class="lineno"> 87</span>&#160;}</div><div class="line"><a name="l00088"></a><span class="lineno"> 88</span>&#160;</div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>&#160;} <span class="comment">// namespace kernel</span></div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>&#160;</div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>&#160;</div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span>&#160;</div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> Element&gt;</div><div class="line"><a name="l00096"></a><span class="lineno"><a class="line" href="namespacecutlass_1_1reference_1_1device.html#aad19927d67f15b89e66560cb77f2a813"> 96</a></span>&#160;<span class="keywordtype">bool</span> <a class="code" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a4595ede72eddace3c973c7f0f74b001d">BlockCompareEqual</a>(</div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span>&#160; Element <span class="keyword">const</span> *ptr_A,</div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>&#160; Element <span class="keyword">const</span> *ptr_B,</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span>&#160; <span class="keywordtype">size_t</span> capacity,</div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>&#160; <span class="keywordtype">int</span> grid_size = 0, </div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>&#160; <span class="keywordtype">int</span> block_size = 0) {</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>&#160;</div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span>&#160; <span class="keywordtype">int</span> equal_flag = 1;</div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span>&#160; <span class="keywordtype">int</span> *device_equal_flag = <span class="keyword">nullptr</span>;</div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span>&#160;</div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span>&#160; <span class="keywordflow">if</span> (cudaMalloc((<span class="keywordtype">void</span> **)&amp;device_equal_flag, <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>)) != cudaSuccess) {</div><div class="line"><a name="l00107"></a><span class="lineno"> 107</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to allocate device flag.&quot;</span>);</div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span>&#160; }</div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>&#160;</div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span>&#160; <span class="keywordflow">if</span> (cudaMemcpy(</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>&#160; device_equal_flag, </div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span>&#160; &amp;equal_flag, </div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>&#160; <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>), </div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>&#160; cudaMemcpyHostToDevice) != cudaSuccess) {</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">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to copy equality flag to device.&quot;</span>);</div><div class="line"><a name="l00117"></a><span class="lineno"> 117</span>&#160; }</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; <span class="keywordflow">if</span> (!grid_size || !block_size) {</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>&#160;</div><div class="line"><a name="l00121"></a><span class="lineno"> 121</span>&#160; <span class="comment">// if grid_size or block_size are zero, query occupancy using the CUDA Occupancy API</span></div><div class="line"><a name="l00122"></a><span class="lineno"> 122</span>&#160; cudaError_t result = cudaOccupancyMaxPotentialBlockSize(</div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>&#160; &amp;grid_size,</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>&#160; &amp;block_size,</div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span>&#160; reinterpret_cast&lt;void const *&gt;(kernel::BlockCompareEqual&lt;Element&gt;));</div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>&#160;</div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>&#160; <span class="keywordflow">if</span> (result != cudaSuccess) {</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to query occupancy.&quot;</span>);</div><div class="line"><a name="l00129"></a><span class="lineno"> 129</span>&#160; }</div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span>&#160;</div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>&#160; <span class="comment">// Limit block size. This has the effect of increasing the number of items processed by a</span></div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>&#160; <span class="comment">// single thread and reduces the impact of initialization overhead.</span></div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>&#160; block_size = (block_size &lt; 128 ? block_size : 128);</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span>&#160; }</div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>&#160;</div><div class="line"><a name="l00136"></a><span class="lineno"> 136</span>&#160; dim3 grid(grid_size, 1, 1);</div><div class="line"><a name="l00137"></a><span class="lineno"> 137</span>&#160; dim3 block(block_size, 1, 1);</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; kernel::BlockCompareEqual&lt;Element&gt;&lt;&lt;&lt; grid, block &gt;&gt;&gt;(device_equal_flag, ptr_A, ptr_B, capacity);</div><div class="line"><a name="l00140"></a><span class="lineno"> 140</span>&#160;</div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span>&#160; <span class="keywordflow">if</span> (cudaMemcpy(</div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span>&#160; &amp;equal_flag, </div><div class="line"><a name="l00143"></a><span class="lineno"> 143</span>&#160; device_equal_flag,</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>&#160; <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>), </div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span>&#160; cudaMemcpyDeviceToHost) != cudaSuccess) {</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>&#160; </div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>&#160; cudaFree(device_equal_flag);</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>&#160;</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to copy equality flag from device.&quot;</span>);</div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span>&#160; }</div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span>&#160;</div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span>&#160; cudaFree(device_equal_flag);</div><div class="line"><a name="l00153"></a><span class="lineno"> 153</span>&#160;</div><div class="line"><a name="l00154"></a><span class="lineno"> 154</span>&#160; <span class="keywordflow">return</span> equal_flag;</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;</div><div class="line"><a name="l00158"></a><span class="lineno"> 158</span>&#160;</div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span>&#160;<span class="keyword">template</span> &lt;<span class="keyword">typename</span> Element&gt;</div><div class="line"><a name="l00161"></a><span class="lineno"><a class="line" href="namespacecutlass_1_1reference_1_1device.html#a286d24a9faabc0be18f96e1069dca23e"> 161</a></span>&#160;<span class="keywordtype">bool</span> <a class="code" href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a6da13fb683d56d6973af0a97a4023677">BlockCompareRelativelyEqual</a>(</div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>&#160; Element <span class="keyword">const</span> *ptr_A,</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>&#160; Element <span class="keyword">const</span> *ptr_B,</div><div class="line"><a name="l00164"></a><span class="lineno"> 164</span>&#160; <span class="keywordtype">size_t</span> capacity,</div><div class="line"><a name="l00165"></a><span class="lineno"> 165</span>&#160; Element epsilon,</div><div class="line"><a name="l00166"></a><span class="lineno"> 166</span>&#160; Element nonzero_floor,</div><div class="line"><a name="l00167"></a><span class="lineno"> 167</span>&#160; <span class="keywordtype">int</span> grid_size = 0, </div><div class="line"><a name="l00168"></a><span class="lineno"> 168</span>&#160; <span class="keywordtype">int</span> block_size = 0) {</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>&#160;</div><div class="line"><a name="l00170"></a><span class="lineno"> 170</span>&#160; <span class="keywordtype">int</span> equal_flag = 1;</div><div class="line"><a name="l00171"></a><span class="lineno"> 171</span>&#160; <span class="keywordtype">int</span> *device_equal_flag = <span class="keyword">nullptr</span>;</div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>&#160;</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span>&#160; <span class="keywordflow">if</span> (cudaMalloc((<span class="keywordtype">void</span> **)&amp;device_equal_flag, <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>)) != cudaSuccess) {</div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to allocate device flag.&quot;</span>);</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span>&#160; }</div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span>&#160;</div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>&#160; <span class="keywordflow">if</span> (cudaMemcpy(</div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>&#160; device_equal_flag, </div><div class="line"><a name="l00179"></a><span class="lineno"> 179</span>&#160; &amp;equal_flag, </div><div class="line"><a name="l00180"></a><span class="lineno"> 180</span>&#160; <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>), </div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>&#160; cudaMemcpyHostToDevice) != cudaSuccess) {</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>&#160;</div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to copy equality flag to device.&quot;</span>);</div><div class="line"><a name="l00184"></a><span class="lineno"> 184</span>&#160; }</div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span>&#160;</div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>&#160; <span class="keywordflow">if</span> (!grid_size || !block_size) {</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span>&#160;</div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span>&#160; <span class="comment">// if grid_size or block_size are zero, query occupancy using the CUDA Occupancy API</span></div><div class="line"><a name="l00189"></a><span class="lineno"> 189</span>&#160; cudaError_t result = cudaOccupancyMaxPotentialBlockSize(</div><div class="line"><a name="l00190"></a><span class="lineno"> 190</span>&#160; &amp;grid_size,</div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>&#160; &amp;block_size,</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>&#160; reinterpret_cast&lt;void const *&gt;(kernel::BlockCompareRelativelyEqual&lt;Element&gt;));</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>&#160;</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span>&#160; <span class="keywordflow">if</span> (result != cudaSuccess) {</div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to query occupancy.&quot;</span>);</div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>&#160; }</div><div class="line"><a name="l00197"></a><span class="lineno"> 197</span>&#160;</div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>&#160; <span class="comment">// Limit block size. This has the effect of increasing the number of items processed by a</span></div><div class="line"><a name="l00199"></a><span class="lineno"> 199</span>&#160; <span class="comment">// single thread and reduces the impact of initialization overhead.</span></div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>&#160; block_size = (block_size &lt; 128 ? block_size : 128);</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>&#160; }</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span>&#160;</div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span>&#160; dim3 grid(grid_size, 1, 1);</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>&#160; dim3 block(block_size, 1, 1);</div><div class="line"><a name="l00205"></a><span class="lineno"> 205</span>&#160;</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span>&#160; kernel::BlockCompareRelativelyEqual&lt;Element&gt;&lt;&lt;&lt; grid, block &gt;&gt;&gt;(</div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span>&#160; device_equal_flag, </div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>&#160; ptr_A, </div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>&#160; ptr_B, </div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span>&#160; capacity, </div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span>&#160; epsilon, </div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>&#160; nonzero_floor</div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span>&#160; );</div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span>&#160;</div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span>&#160; <span class="keywordflow">if</span> (cudaMemcpy(</div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span>&#160; &amp;equal_flag, </div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>&#160; device_equal_flag,</div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>&#160; <span class="keyword">sizeof</span>(<span class="keywordtype">int</span>), </div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span>&#160; cudaMemcpyDeviceToHost) != cudaSuccess) {</div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span>&#160; </div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>&#160; cudaFree(device_equal_flag);</div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span>&#160;</div><div class="line"><a name="l00223"></a><span class="lineno"> 223</span>&#160; <span class="keywordflow">throw</span> std::runtime_error(<span class="stringliteral">&quot;Failed to copy equality flag from device.&quot;</span>);</div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span>&#160; }</div><div class="line"><a name="l00225"></a><span class="lineno"> 225</span>&#160;</div><div class="line"><a name="l00226"></a><span class="lineno"> 226</span>&#160; cudaFree(device_equal_flag);</div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>&#160;</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span>&#160; <span class="keywordflow">return</span> equal_flag;</div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>&#160;}</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span>&#160;</div><div class="line"><a name="l00232"></a><span class="lineno"> 232</span>&#160;</div><div class="line"><a name="l00233"></a><span class="lineno"> 233</span>&#160;} <span class="comment">// device</span></div><div class="line"><a name="l00234"></a><span class="lineno"> 234</span>&#160;} <span class="comment">// reference</span></div><div class="line"><a name="l00235"></a><span class="lineno"> 235</span>&#160;} <span class="comment">// cutlass</span></div><div class="ttc" id="namespacecutlass_html"><div class="ttname"><a href="namespacecutlass.html">cutlass</a></div><div class="ttdef"><b>Definition:</b> aligned_buffer.h:35</div></div>
<div class="ttc" id="namespacecutlass_1_1reference_1_1device_1_1kernel_html_a6da13fb683d56d6973af0a97a4023677"><div class="ttname"><a href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a6da13fb683d56d6973af0a97a4023677">cutlass::reference::device::kernel::BlockCompareRelativelyEqual</a></div><div class="ttdeci">__global__ void BlockCompareRelativelyEqual(int *equal, Element const *ptr_A, Element const *ptr_B, size_t capacity, Element epsilon, Element nonzero_floor)</div><div class="ttdef"><b>Definition:</b> device/tensor_compare.h:67</div></div>
<div class="ttc" id="namespacecutlass_html_aebc97e25721b38692159a6d50d545fe0"><div class="ttname"><a href="namespacecutlass.html#aebc97e25721b38692159a6d50d545fe0">cutlass::relatively_equal</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE bool relatively_equal(T a, T b, T epsilon, T nonzero_floor)</div></div>
<div class="ttc" id="device_2tensor__foreach_8h_html"><div class="ttname"><a href="device_2tensor__foreach_8h.html">tensor_foreach.h</a></div></div>
<div class="ttc" id="relatively__equal_8h_html"><div class="ttname"><a href="relatively__equal_8h.html">relatively_equal.h</a></div></div>
<div class="ttc" id="namespacecutlass_1_1reference_1_1device_1_1kernel_html_a4595ede72eddace3c973c7f0f74b001d"><div class="ttname"><a href="namespacecutlass_1_1reference_1_1device_1_1kernel.html#a4595ede72eddace3c973c7f0f74b001d">cutlass::reference::device::kernel::BlockCompareEqual</a></div><div class="ttdeci">__global__ void BlockCompareEqual(int *equal, Element const *ptr_A, Element const *ptr_B, size_t capacity)</div><div class="ttdef"><b>Definition:</b> device/tensor_compare.h:50</div></div>
<div class="ttc" id="distribution_8h_html"><div class="ttname"><a href="distribution_8h.html">distribution.h</a></div><div class="ttdoc">This header contains a class to parametrize a statistical distribution function. </div></div>
<div class="ttc" id="cutlass_8h_html"><div class="ttname"><a href="cutlass_8h.html">cutlass.h</a></div><div class="ttdoc">Basic include for CUTLASS. </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> 1.8.11
</small></address>
</body>
</html>