| <!-- HTML header for doxygen 1.8.15--> |
| <!-- Remember to use version doxygen 1.8.15 +--> |
| <!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 1.8.15"/> |
| <meta name="robots" content="NOINDEX, NOFOLLOW" /> <!-- Prevent indexing by search engines --> |
| <title>Compute Library: src/graph/mutators/NodeFusionMutator.cpp 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="navtree.css" rel="stylesheet" type="text/css"/> |
| <script type="text/javascript" src="resize.js"></script> |
| <script type="text/javascript" src="navtreedata.js"></script> |
| <script type="text/javascript" src="navtree.js"></script> |
| <script type="text/javascript"> |
| /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */ |
| $(document).ready(initResizable); |
| /* @license-end */</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/x-mathjax-config"> |
| MathJax.Hub.Config({ |
| extensions: ["tex2jax.js"], |
| jax: ["input/TeX","output/HTML-CSS"], |
| }); |
| </script><script type="text/javascript" async="async" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script> |
| <link href="doxygen.css" rel="stylesheet" type="text/css" /> |
| <link href="stylesheet.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;"> |
| <img alt="Compute Library" src="https://raw.githubusercontent.com/ARM-software/ComputeLibrary/gh-pages/ACL_logo.png" style="max-width: 100%;margin-top: 15px;margin-left: 10px"/> |
| <td style="padding-left: 0.5em;"> |
| <div id="projectname"> |
|  <span id="projectnumber">19.08</span> |
| </div> |
| </td> |
| </tr> |
| </tbody> |
| </table> |
| </div> |
| <!-- end header part --> |
| <!-- Generated by Doxygen 1.8.15 --> |
| <script type="text/javascript"> |
| /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */ |
| var searchBox = new SearchBox("searchBox", "search",false,'Search'); |
| /* @license-end */ |
| </script> |
| <script type="text/javascript" src="menudata.js"></script> |
| <script type="text/javascript" src="menu.js"></script> |
| <script type="text/javascript"> |
| /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */ |
| $(function() { |
| initMenu('',true,false,'search.php','Search'); |
| $(document).ready(function() { init_search(); }); |
| }); |
| /* @license-end */</script> |
| <div id="main-nav"></div> |
| </div><!-- top --> |
| <div id="side-nav" class="ui-resizable side-nav-resizable"> |
| <div id="nav-tree"> |
| <div id="nav-tree-contents"> |
| <div id="nav-sync" class="sync"></div> |
| </div> |
| </div> |
| <div id="splitbar" style="-moz-user-select:none;" |
| class="ui-resizable-handle"> |
| </div> |
| </div> |
| <script type="text/javascript"> |
| /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */ |
| $(document).ready(function(){initNavTree('_node_fusion_mutator_8cpp_source.xhtml','');}); |
| /* @license-end */ |
| </script> |
| <div id="doc-content"> |
| <!-- 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 class="header"> |
| <div class="headertitle"> |
| <div class="title">NodeFusionMutator.cpp</div> </div> |
| </div><!--header--> |
| <div class="contents"> |
| <a href="_node_fusion_mutator_8cpp.xhtml">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno"> 1</span> <span class="comment">/*</span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span> <span class="comment"> * Copyright (c) 2018-2019 ARM Limited.</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment"> *</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span> <span class="comment"> * SPDX-License-Identifier: MIT</span></div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span> <span class="comment"> *</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span> <span class="comment"> * Permission is hereby granted, free of charge, to any person obtaining a copy</span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span> <span class="comment"> * of this software and associated documentation files (the "Software"), to</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span> <span class="comment"> * deal in the Software without restriction, including without limitation the</span></div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span> <span class="comment"> * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or</span></div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span> <span class="comment"> * sell copies of the Software, and to permit persons to whom the Software is</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span> <span class="comment"> * furnished to do so, subject to the following conditions:</span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span> <span class="comment"> *</span></div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span> <span class="comment"> * The above copyright notice and this permission notice shall be included in all</span></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span> <span class="comment"> * copies or substantial portions of the Software.</span></div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span> <span class="comment"> *</span></div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span> <span class="comment"> * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR</span></div><div class="line"><a name="l00017"></a><span class="lineno"> 17</span> <span class="comment"> * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,</span></div><div class="line"><a name="l00018"></a><span class="lineno"> 18</span> <span class="comment"> * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE</span></div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span> <span class="comment"> * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER</span></div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span> <span class="comment"> * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,</span></div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span> <span class="comment"> * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE</span></div><div class="line"><a name="l00022"></a><span class="lineno"> 22</span> <span class="comment"> * SOFTWARE.</span></div><div class="line"><a name="l00023"></a><span class="lineno"> 23</span> <span class="comment"> */</span></div><div class="line"><a name="l00024"></a><span class="lineno"> 24</span> <span class="preprocessor">#include "<a class="code" href="_node_fusion_mutator_8h.xhtml">arm_compute/graph/mutators/NodeFusionMutator.h</a>"</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"> 26</span> <span class="preprocessor">#include "<a class="code" href="_graph_builder_8h.xhtml">arm_compute/graph/GraphBuilder.h</a>"</span></div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span> <span class="preprocessor">#include "<a class="code" href="graph_2_logger_8h.xhtml">arm_compute/graph/Logger.h</a>"</span></div><div class="line"><a name="l00028"></a><span class="lineno"> 28</span> <span class="preprocessor">#include "<a class="code" href="arm__compute_2graph_2_utils_8h.xhtml">arm_compute/graph/Utils.h</a>"</span></div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span> <span class="preprocessor">#include "<a class="code" href="_backend_registry_8h.xhtml">arm_compute/graph/backends/BackendRegistry.h</a>"</span></div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span> <span class="preprocessor">#include "<a class="code" href="_fused_convolution_batch_normalization_node_8h.xhtml">arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h</a>"</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span> <span class="preprocessor">#include "<a class="code" href="_nodes_8h.xhtml">arm_compute/graph/nodes/Nodes.h</a>"</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span> </div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span> <span class="preprocessor">#include "<a class="code" href="_cast_8h.xhtml">arm_compute/core/utils/misc/Cast.h</a>"</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="preprocessor">#include <set></span></div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span> </div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span> <span class="keyword">namespace </span><a class="code" href="namespacearm__compute.xhtml">arm_compute</a></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="keyword">namespace </span>graph</div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span> {</div><div class="line"><a name="l00041"></a><span class="lineno"> 41</span> <span class="keyword">namespace </span>detail</div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span> {</div><div class="line"><a name="l00043"></a><span class="lineno"><a class="line" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a501484d2b5f0213bbede4f44471c148b"> 43</a></span> <span class="keywordtype">void</span> <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a501484d2b5f0213bbede4f44471c148b">fuse_convolution_with_batch_normalization</a>(<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml">Graph</a> &g, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml">Edge</a> *output_edge)</div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span> {</div><div class="line"><a name="l00045"></a><span class="lineno"> 45</span>  <a class="code" href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a>(output_edge == <span class="keyword">nullptr</span>);</div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span> </div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span>  <span class="keyword">auto</span> *conv_node = arm_compute::utils::cast::polymorphic_downcast<ConvolutionLayerNode *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#acef46a426a99b126a412e361125f2ce9">producer</a>());</div><div class="line"><a name="l00048"></a><span class="lineno"> 48</span>  <span class="keyword">auto</span> *bn_node = arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a592a2c76b00960964a3f6f2ef792a7f0">consumer</a>());</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>  <span class="comment">// Not fusing if number of groups is greater than 1</span></div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span>  <span class="keywordflow">if</span>(conv_node->num_groups() > 1)</div><div class="line"><a name="l00052"></a><span class="lineno"> 52</span>  {</div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span>  <span class="keywordflow">return</span>;</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>  }</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>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Fusing convolution node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">producer_id</a>()</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>  << <span class="stringliteral">" with BatchNormalization Layer node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#ad00e584f78c622e5ee9ec9613f6d6633">consumer_id</a>() << std::endl);</div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span> </div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span>  <span class="comment">// Prevent fusion if fused node has an output accessor</span></div><div class="line"><a name="l00060"></a><span class="lineno"> 60</span>  <span class="keywordflow">if</span>(conv_node->output(0)->accessor() == <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00061"></a><span class="lineno"> 61</span>  {</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span>  <span class="keyword">const</span> <a class="code" href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35d">Target</a> assigned_target = conv_node->assigned_target();</div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span> </div><div class="line"><a name="l00064"></a><span class="lineno"> 64</span>  <span class="comment">// Extract conv inputs</span></div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  <span class="keyword">const</span> <span class="keyword">auto</span> conv_input_id = conv_node->input_edge(0)->producer_id();</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  <span class="keyword">const</span> <span class="keyword">auto</span> conv_weights_id = conv_node->input_edge(1)->producer_id();</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a00525ff582f16038a1d3819aa44a23a3">conv_info</a> = conv_node->convolution_info();</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  <span class="keyword">const</span> <span class="keyword">auto</span> conv_method = conv_node->convolution_method();</div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a2270b3e1d20651d2d8341c858c890830">num_groups</a> = conv_node->num_groups();</div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a1f8aca235c095df227e7444f6b237eb1">act_info</a> = bn_node->fused_activation();</div><div class="line"><a name="l00071"></a><span class="lineno"> 71</span>  <a class="code" href="namespacearm__compute_1_1graph.xhtml#ac85a46f3ebd3ab09f576a994ac2dce11">FastMathHint</a> fast_math_hint = conv_node->fast_math_hint();</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span> </div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  <span class="comment">// Extract bn inputs</span></div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_mean_id = bn_node->input_edge(1)->producer_id();</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_var_id = bn_node->input_edge(2)->producer_id();</div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_beta_id = bn_node->input_edge(3)->producer_id();</div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_gamma_id = bn_node->input_edge(4)->producer_id();</div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="_asymm_helpers_8cpp.xhtml#a552dc3787d7ea1675f3e4e8993501d58">epsilon</a> = bn_node->epsilon();</div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span> </div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span>  <span class="comment">// Create the fused node</span></div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>  <span class="keyword">const</span> <a class="code" href="namespacearm__compute_1_1graph.xhtml#a56952a8591a9d8481450ae707ae355e7">NodeID</a> fused_id = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a424221c5dfc8fe79f88a677d1b3b4494">add_node</a><<a class="code" href="classarm__compute_1_1graph_1_1_fused_convolution_batch_normalization_node.xhtml">FusedConvolutionBatchNormalizationNode</a>>(<a class="code" href="_asymm_helpers_8cpp.xhtml#a552dc3787d7ea1675f3e4e8993501d58">epsilon</a>, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a00525ff582f16038a1d3819aa44a23a3">conv_info</a>, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a2270b3e1d20651d2d8341c858c890830">num_groups</a>, conv_method, fast_math_hint, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a1f8aca235c095df227e7444f6b237eb1">act_info</a>);</div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span> </div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>  <span class="keywordflow">if</span>(conv_node->input_edge(2) != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00084"></a><span class="lineno"> 84</span>  {</div><div class="line"><a name="l00085"></a><span class="lineno"> 85</span>  <span class="keyword">auto</span> conv_bias_id = conv_node-><a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml#a6bca7473aa08cb0ecba36cb5dda2badf">input_edge</a>(2)-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">producer_id</a>();</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(conv_bias_id, 0, fused_id, 2);</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> </div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  <span class="comment">// Add connections from the conv/batch_norm inputs to the fused node</span></div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(conv_input_id, 0, fused_id, 0);</div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(conv_weights_id, 0, fused_id, 1);</div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_mean_id, 0, fused_id, 3);</div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_var_id, 0, fused_id, 4);</div><div class="line"><a name="l00094"></a><span class="lineno"> 94</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_beta_id, 0, fused_id, 5);</div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_gamma_id, 0, fused_id, 6);</div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span> </div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span>  <span class="keyword">auto</span> fused_node = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#af8baf1f3da6d42a94d0569395ece882a">node</a>(fused_id);</div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>  std::vector<NodeIdxPair> bn_driving_nodes = <a class="code" href="namespacearm__compute_1_1graph.xhtml#a634230f98a5918f214e47d913c452d3b">get_driving_nodes</a>(*bn_node);</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span> </div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  <span class="comment">// Extract batch normalization node accessor if any</span></div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>  <span class="keyword">auto</span> bn_node_accessor = bn_node->output(0)->extract_accessor();</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>  <span class="keyword">auto</span> bn_node_name = bn_node->name();</div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span> </div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span>  <span class="comment">// Remove batch normalization node</span></div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">remove_node</a>(bn_node->id());</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="comment">// Get driving nodes of batch normalization node</span></div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span>  <span class="keywordflow">for</span>(<span class="keyword">auto</span> &driving_node : bn_driving_nodes)</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>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(fused_id, 0, driving_node.node_id, driving_node.index);</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>  <a class="code" href="namespacearm__compute_1_1graph.xhtml#a36fd3cfa2e1d33e59e1d3e95664d8b9c">configure_tensor</a>(fused_node->output(0));</div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span>  }</div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>  <span class="comment">// Update fused node outputs</span></div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>  fused_node->output(0)->set_accessor(std::move(bn_node_accessor));</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  fused_node->set_assigned_target(assigned_target);</div><div class="line"><a name="l00116"></a><span class="lineno"> 116</span>  fused_node->set_common_node_parameters(<a class="code" href="structarm__compute_1_1graph_1_1_node_params.xhtml">NodeParams</a>{ conv_node-><a class="code" href="structarm__compute_1_1graph_1_1_node_params.xhtml#a9b45b3e13bd9167aab02e17e08916231">name</a>() + <span class="stringliteral">"+"</span> + bn_node_name, assigned_target });</div><div class="line"><a name="l00117"></a><span class="lineno"> 117</span> </div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  <span class="comment">// Remove convolution node</span></div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">remove_node</a>(conv_node->id());</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  }</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>  {</div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Prevented fusion of convolution with batch normalization due to the presence of an output accessor\n"</span>);</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> </div><div class="line"><a name="l00127"></a><span class="lineno"><a class="line" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a07811ee0f5c19da2658a647e6d7dc2fa"> 127</a></span> <span class="keywordtype">void</span> <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a07811ee0f5c19da2658a647e6d7dc2fa">fuse_depthwise_convolution_with_batch_normalization</a>(<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml">Graph</a> &g, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml">Edge</a> *output_edge)</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>  <a class="code" href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a>(output_edge == <span class="keyword">nullptr</span>);</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="keyword">auto</span> *depth_conv_node = arm_compute::utils::cast::polymorphic_downcast<DepthwiseConvolutionLayerNode *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#acef46a426a99b126a412e361125f2ce9">producer</a>());</div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  <span class="keyword">auto</span> *bn_node = arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a592a2c76b00960964a3f6f2ef792a7f0">consumer</a>());</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span> </div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Fusing depthwise convolution node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">producer_id</a>()</div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>  << <span class="stringliteral">" with BatchNormalization Layer node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#ad00e584f78c622e5ee9ec9613f6d6633">consumer_id</a>() << std::endl);</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>  <span class="comment">// Prevent fusion if fused node has an output accessor</span></div><div class="line"><a name="l00138"></a><span class="lineno"> 138</span>  <span class="keywordflow">if</span>(depth_conv_node->output(0)->accessor() == <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span>  {</div><div class="line"><a name="l00140"></a><span class="lineno"> 140</span>  <span class="keyword">const</span> <a class="code" href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35d">Target</a> assigned_target = depth_conv_node->assigned_target();</div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span> </div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span>  <span class="comment">// Extract conv inputs</span></div><div class="line"><a name="l00143"></a><span class="lineno"> 143</span>  <span class="keyword">const</span> <span class="keyword">auto</span> depth_conv_input_id = depth_conv_node->input_edge(0)->producer_id();</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  <span class="keyword">const</span> <span class="keyword">auto</span> conv_weights_id = depth_conv_node->input_edge(1)->producer_id();</div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a00525ff582f16038a1d3819aa44a23a3">conv_info</a> = depth_conv_node->convolution_info();</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>  <span class="keyword">const</span> <span class="keyword">auto</span> depth_conv_method = depth_conv_node->depthwise_convolution_method();</div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  <span class="keyword">const</span> <span class="keyword">auto</span> depth_multiplier = depth_conv_node->depth_multiplier();</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a1f8aca235c095df227e7444f6b237eb1">act_info</a> = bn_node->fused_activation();</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>  <span class="comment">// Extract bn inputs</span></div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_mean_id = bn_node->input_edge(1)->producer_id();</div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_var_id = bn_node->input_edge(2)->producer_id();</div><div class="line"><a name="l00153"></a><span class="lineno"> 153</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_beta_id = bn_node->input_edge(3)->producer_id();</div><div class="line"><a name="l00154"></a><span class="lineno"> 154</span>  <span class="keyword">const</span> <span class="keyword">auto</span> bn_gamma_id = bn_node->input_edge(4)->producer_id();</div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span>  <span class="keyword">const</span> <span class="keyword">auto</span> <a class="code" href="_asymm_helpers_8cpp.xhtml#a552dc3787d7ea1675f3e4e8993501d58">epsilon</a> = bn_node->epsilon();</div><div class="line"><a name="l00156"></a><span class="lineno"> 156</span> </div><div class="line"><a name="l00157"></a><span class="lineno"> 157</span>  <span class="comment">// Create the fused node</span></div><div class="line"><a name="l00158"></a><span class="lineno"> 158</span>  <span class="keyword">const</span> <a class="code" href="namespacearm__compute_1_1graph.xhtml#a56952a8591a9d8481450ae707ae355e7">NodeID</a> fused_id = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a424221c5dfc8fe79f88a677d1b3b4494">add_node</a><<a class="code" href="classarm__compute_1_1graph_1_1_fused_depthwise_convolution_batch_normalization_node.xhtml">FusedDepthwiseConvolutionBatchNormalizationNode</a>>(<a class="code" href="_asymm_helpers_8cpp.xhtml#a552dc3787d7ea1675f3e4e8993501d58">epsilon</a>, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a00525ff582f16038a1d3819aa44a23a3">conv_info</a>, depth_multiplier, depth_conv_method, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a1f8aca235c095df227e7444f6b237eb1">act_info</a>);</div><div class="line"><a name="l00159"></a><span class="lineno"> 159</span> </div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span>  <span class="keywordflow">if</span>(depth_conv_node->input_edge(2) != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span>  {</div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  <span class="keyword">const</span> <span class="keyword">auto</span> conv_bias_id = depth_conv_node-><a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml#a6bca7473aa08cb0ecba36cb5dda2badf">input_edge</a>(2)-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">producer_id</a>();</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(conv_bias_id, 0, fused_id, 2);</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> </div><div class="line"><a name="l00166"></a><span class="lineno"> 166</span>  <span class="comment">// Add connections from the conv/batch_norm inputs to the fused node</span></div><div class="line"><a name="l00167"></a><span class="lineno"> 167</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(depth_conv_input_id, 0, fused_id, 0);</div><div class="line"><a name="l00168"></a><span class="lineno"> 168</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(conv_weights_id, 0, fused_id, 1);</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_mean_id, 0, fused_id, 3);</div><div class="line"><a name="l00170"></a><span class="lineno"> 170</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_var_id, 0, fused_id, 4);</div><div class="line"><a name="l00171"></a><span class="lineno"> 171</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_beta_id, 0, fused_id, 5);</div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(bn_gamma_id, 0, fused_id, 6);</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span> </div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>  <span class="keyword">auto</span> fused_node = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#af8baf1f3da6d42a94d0569395ece882a">node</a>(fused_id);</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span>  std::vector<NodeIdxPair> bn_driving_nodes = <a class="code" href="namespacearm__compute_1_1graph.xhtml#a634230f98a5918f214e47d913c452d3b">get_driving_nodes</a>(*bn_node);</div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span> </div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>  <span class="comment">// Extract batch normalization node accessor if any</span></div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>  <span class="keyword">auto</span> bn_node_accessor = bn_node->output(0)->extract_accessor();</div><div class="line"><a name="l00179"></a><span class="lineno"> 179</span>  <span class="keyword">auto</span> bn_node_name = bn_node->name();</div><div class="line"><a name="l00180"></a><span class="lineno"> 180</span> </div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  <span class="comment">// Remove batch normalization node</span></div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">remove_node</a>(bn_node->id());</div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span> </div><div class="line"><a name="l00184"></a><span class="lineno"> 184</span>  <span class="comment">// Get driving nodes of batch normalization node</span></div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span>  <span class="keywordflow">for</span>(<span class="keyword">auto</span> &driving_node : bn_driving_nodes)</div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>  {</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(fused_id, 0, driving_node.node_id, driving_node.index);</div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span>  <a class="code" href="namespacearm__compute_1_1graph.xhtml#a36fd3cfa2e1d33e59e1d3e95664d8b9c">configure_tensor</a>(fused_node->output(0));</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="comment">// Update fused node outputs</span></div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  fused_node->output(0)->set_accessor(std::move(bn_node_accessor));</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  fused_node->set_assigned_target(assigned_target);</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  fused_node->set_common_node_parameters(<a class="code" href="structarm__compute_1_1graph_1_1_node_params.xhtml">NodeParams</a>{ depth_conv_node-><a class="code" href="structarm__compute_1_1graph_1_1_node_params.xhtml#a9b45b3e13bd9167aab02e17e08916231">name</a>() + <span class="stringliteral">"+"</span> + bn_node_name, assigned_target });</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span> </div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span>  <span class="comment">// Remove convolution node</span></div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">remove_node</a>(depth_conv_node->id());</div><div class="line"><a name="l00197"></a><span class="lineno"> 197</span>  }</div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00199"></a><span class="lineno"> 199</span>  {</div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Prevented fusion of depthwise convolution with batch normalization due to the presence of an output accessor\n"</span>);</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  }</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span> }</div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span> </div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span> <span class="keyword">template</span> <<span class="keyword">typename</span> N></div><div class="line"><a name="l00205"></a><span class="lineno"><a class="line" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a36b5b67e26dfdfbedf39e6703ef059bd"> 205</a></span> <span class="keywordtype">void</span> <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a36b5b67e26dfdfbedf39e6703ef059bd">fuse_node_with_activation</a>(<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml">Graph</a> &g, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml">Edge</a> *output_edge, <span class="keyword">const</span> std::set<Activation> &supported_fused_activations)</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>  <a class="code" href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a>(output_edge == <span class="keyword">nullptr</span>);</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span> </div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>  <span class="keyword">auto</span> *n_node = arm_compute::utils::cast::polymorphic_downcast<N *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#acef46a426a99b126a412e361125f2ce9">producer</a>());</div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span>  <span class="keyword">auto</span> *act_node = arm_compute::utils::cast::polymorphic_downcast<ActivationLayerNode *>(output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a592a2c76b00960964a3f6f2ef792a7f0">consumer</a>());</div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span> </div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>  <a class="code" href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a>(act_node->output(0) == <span class="keyword">nullptr</span> || n_node->output(0) == <span class="keyword">nullptr</span>);</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>  <span class="comment">// Check if activation is supported for fusion</span></div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span>  <span class="keywordflow">if</span>(supported_fused_activations.count(act_node->activation_info().activation()) == 0)</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="keywordflow">return</span>;</div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  }</div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span> </div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Fusing node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">producer_id</a>()</div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  << <span class="stringliteral">" with Activation Layer node with ID : "</span> << output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#ad00e584f78c622e5ee9ec9613f6d6633">consumer_id</a>() << std::endl);</div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span> </div><div class="line"><a name="l00223"></a><span class="lineno"> 223</span>  <span class="comment">// Prevent fusion if fused node has an output accessor</span></div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span>  <span class="keywordflow">if</span>(n_node->output(0)->accessor() == <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00225"></a><span class="lineno"> 225</span>  {</div><div class="line"><a name="l00226"></a><span class="lineno"> 226</span>  <span class="comment">// Get driving nodes of activation node</span></div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  std::vector<NodeIdxPair> act_driving_nodes = <a class="code" href="namespacearm__compute_1_1graph.xhtml#a634230f98a5918f214e47d913c452d3b">get_driving_nodes</a>(*act_node);</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span> </div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  <span class="comment">// Set activation info to fused node</span></div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span>  n_node->set_fused_activation(act_node->activation_info());</div><div class="line"><a name="l00231"></a><span class="lineno"> 231</span> </div><div class="line"><a name="l00232"></a><span class="lineno"> 232</span>  <span class="comment">// Extract activation node accessor if any</span></div><div class="line"><a name="l00233"></a><span class="lineno"> 233</span>  <span class="keyword">auto</span> act_node_accessor = act_node->output(0)->extract_accessor();</div><div class="line"><a name="l00234"></a><span class="lineno"> 234</span> </div><div class="line"><a name="l00235"></a><span class="lineno"> 235</span>  <span class="comment">// Remove activation node</span></div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">remove_node</a>(act_node->id());</div><div class="line"><a name="l00237"></a><span class="lineno"> 237</span> </div><div class="line"><a name="l00238"></a><span class="lineno"> 238</span>  <span class="comment">// Update fused node outputs</span></div><div class="line"><a name="l00239"></a><span class="lineno"> 239</span>  <span class="keywordflow">for</span>(<span class="keyword">auto</span> &driving_node : act_driving_nodes)</div><div class="line"><a name="l00240"></a><span class="lineno"> 240</span>  {</div><div class="line"><a name="l00241"></a><span class="lineno"> 241</span>  g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">add_connection</a>(n_node->id(), 0, driving_node.node_id, driving_node.index);</div><div class="line"><a name="l00242"></a><span class="lineno"> 242</span>  }</div><div class="line"><a name="l00243"></a><span class="lineno"> 243</span> </div><div class="line"><a name="l00244"></a><span class="lineno"> 244</span>  <span class="comment">// Update accessor to fused node</span></div><div class="line"><a name="l00245"></a><span class="lineno"> 245</span>  n_node->output(0)->set_accessor(std::move(act_node_accessor));</div><div class="line"><a name="l00246"></a><span class="lineno"> 246</span>  }</div><div class="line"><a name="l00247"></a><span class="lineno"> 247</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00248"></a><span class="lineno"> 248</span>  {</div><div class="line"><a name="l00249"></a><span class="lineno"> 249</span>  <a class="code" href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a>(<span class="stringliteral">"Prevented fusion of node with activation due to the presence of an output accessor\n"</span>);</div><div class="line"><a name="l00250"></a><span class="lineno"> 250</span>  }</div><div class="line"><a name="l00251"></a><span class="lineno"> 251</span> }</div><div class="line"><a name="l00252"></a><span class="lineno"> 252</span> </div><div class="line"><a name="l00253"></a><span class="lineno"> 253</span> <span class="keyword">template</span> <<span class="keyword">typename</span> N1, <span class="keyword">typename</span> N2, <span class="keyword">typename</span> F, <span class="keyword">typename</span>... Args></div><div class="line"><a name="l00254"></a><span class="lineno"><a class="line" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a1f9900d626af4230ff61f851e8d5eab5"> 254</a></span> <span class="keywordtype">void</span> <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a1f9900d626af4230ff61f851e8d5eab5">fuse_layer</a>(<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml">Graph</a> &g, std::function<<span class="keywordtype">bool</span>(<a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml">INode</a> &)> <span class="keyword">const</span> &prec, <span class="keyword">const</span> F fuse_fcn, Args &&... optional_arguments)</div><div class="line"><a name="l00255"></a><span class="lineno"> 255</span> {</div><div class="line"><a name="l00256"></a><span class="lineno"> 256</span>  <span class="comment">// Not interested in the order of nodes</span></div><div class="line"><a name="l00257"></a><span class="lineno"> 257</span>  <span class="keywordflow">for</span>(<span class="keyword">auto</span> &node : g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a8a979250f01a5edba059a02748b10ea3">nodes</a>())</div><div class="line"><a name="l00258"></a><span class="lineno"> 258</span>  {</div><div class="line"><a name="l00259"></a><span class="lineno"> 259</span>  <span class="comment">// Check if the node is of type N and not a branching node</span></div><div class="line"><a name="l00260"></a><span class="lineno"> 260</span>  <span class="keywordflow">if</span>(node && node->type() == N1::node_type && node->output_edges().size() == 1)</div><div class="line"><a name="l00261"></a><span class="lineno"> 261</span>  {</div><div class="line"><a name="l00262"></a><span class="lineno"> 262</span>  <span class="keyword">const</span> <span class="keyword">auto</span> output_edge_id = *node->output_edges().begin();</div><div class="line"><a name="l00263"></a><span class="lineno"> 263</span>  <span class="keyword">const</span> <span class="keyword">auto</span> output_edge = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a0ea47db036b1c140af002ee1494dcb6f">edge</a>(output_edge_id);</div><div class="line"><a name="l00264"></a><span class="lineno"> 264</span> </div><div class="line"><a name="l00265"></a><span class="lineno"> 265</span>  <span class="comment">// Check if following node is an activation layer node</span></div><div class="line"><a name="l00266"></a><span class="lineno"> 266</span>  <span class="keywordflow">if</span>((output_edge != <span class="keyword">nullptr</span>) && (output_edge->consumer() != <span class="keyword">nullptr</span>) && (output_edge->consumer()->type() == N2::node_type) && prec(*output_edge->producer()))</div><div class="line"><a name="l00267"></a><span class="lineno"> 267</span>  {</div><div class="line"><a name="l00268"></a><span class="lineno"> 268</span>  fuse_fcn(g, output_edge, optional_arguments...);</div><div class="line"><a name="l00269"></a><span class="lineno"> 269</span>  }</div><div class="line"><a name="l00270"></a><span class="lineno"> 270</span>  }</div><div class="line"><a name="l00271"></a><span class="lineno"> 271</span>  }</div><div class="line"><a name="l00272"></a><span class="lineno"> 272</span> }</div><div class="line"><a name="l00273"></a><span class="lineno"> 273</span> } <span class="comment">// namespace detail</span></div><div class="line"><a name="l00274"></a><span class="lineno"> 274</span> </div><div class="line"><a name="l00275"></a><span class="lineno"><a class="line" href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#a0ee2c922a45715b33e30f83f8c005b68"> 275</a></span> <span class="keyword">const</span> <span class="keywordtype">char</span> *<a class="code" href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#a0ee2c922a45715b33e30f83f8c005b68">NodeFusionMutator::name</a>()</div><div class="line"><a name="l00276"></a><span class="lineno"> 276</span> {</div><div class="line"><a name="l00277"></a><span class="lineno"> 277</span>  <span class="keywordflow">return</span> <span class="stringliteral">"NodeFusionMutator"</span>;</div><div class="line"><a name="l00278"></a><span class="lineno"> 278</span> }</div><div class="line"><a name="l00279"></a><span class="lineno"> 279</span> </div><div class="line"><a name="l00280"></a><span class="lineno"><a class="line" href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#af6705a49326de235df4e9b6028bdcdce"> 280</a></span> <span class="keywordtype">void</span> <a class="code" href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#af6705a49326de235df4e9b6028bdcdce">NodeFusionMutator::mutate</a>(<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml">Graph</a> &g)</div><div class="line"><a name="l00281"></a><span class="lineno"> 281</span> {</div><div class="line"><a name="l00282"></a><span class="lineno"> 282</span>  <span class="comment">// Supported activations when fusing</span></div><div class="line"><a name="l00283"></a><span class="lineno"> 283</span>  <span class="keyword">const</span> std::set<Activation> supported_fused_activations = { <a class="code" href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaad346bb4679d29be241279f15d7795c1c">Activation::RELU</a>, <a class="code" href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaacc516ab03b98f1c908ddf6ed4a7c45e9">Activation::BOUNDED_RELU</a>, <a class="code" href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaaab1d4411a9e7f5e82002512cddfdc33a">Activation::LU_BOUNDED_RELU</a> };</div><div class="line"><a name="l00284"></a><span class="lineno"> 284</span> </div><div class="line"><a name="l00285"></a><span class="lineno"> 285</span>  <span class="comment">// Preconditions</span></div><div class="line"><a name="l00286"></a><span class="lineno"> 286</span>  <span class="keyword">auto</span> empty_prec = [](<a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml">INode</a> &)</div><div class="line"><a name="l00287"></a><span class="lineno"> 287</span>  {</div><div class="line"><a name="l00288"></a><span class="lineno"> 288</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</div><div class="line"><a name="l00289"></a><span class="lineno"> 289</span>  };</div><div class="line"><a name="l00290"></a><span class="lineno"> 290</span>  <span class="keyword">auto</span> qs8_prec = [&g](<a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml">INode</a> & n)</div><div class="line"><a name="l00291"></a><span class="lineno"> 291</span>  {</div><div class="line"><a name="l00292"></a><span class="lineno"> 292</span>  <a class="code" href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a>(n.output(0) == <span class="keyword">nullptr</span>);</div><div class="line"><a name="l00293"></a><span class="lineno"> 293</span> </div><div class="line"><a name="l00294"></a><span class="lineno"> 294</span>  <span class="keyword">const</span> <span class="keyword">auto</span> output_edge_id = *n.output_edges().begin();</div><div class="line"><a name="l00295"></a><span class="lineno"> 295</span>  <span class="keyword">const</span> <span class="keyword">auto</span> output_edge = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a0ea47db036b1c140af002ee1494dcb6f">edge</a>(output_edge_id);</div><div class="line"><a name="l00296"></a><span class="lineno"> 296</span>  <span class="comment">// To perform fusion the two nodes must have same output quantization information</span></div><div class="line"><a name="l00297"></a><span class="lineno"> 297</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> same_qinfo = n.output(0)->desc().quant_info == output_edge-><a class="code" href="classarm__compute_1_1graph_1_1_edge.xhtml#acef46a426a99b126a412e361125f2ce9">producer</a>()-><a class="code" href="classarm__compute_1_1graph_1_1_i_node.xhtml#a27369471d9b2ba47746e1e923f585b9f">output</a>(0)-><a class="code" href="classarm__compute_1_1graph_1_1_tensor.xhtml#a0d3608f94078b90ab1ff9e9465d4ed3b">desc</a>().<a class="code" href="structarm__compute_1_1graph_1_1_tensor_descriptor.xhtml#a65bc62ff84efcb7e4a410600480a4dc9">quant_info</a>;</div><div class="line"><a name="l00298"></a><span class="lineno"> 298</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> output_qasymm8 = n.output(0)->desc().data_type == <a class="code" href="namespacearm__compute.xhtml#ad8ed01ff3ff33333d8e19db4d2818bb6af14462d71aa842202c3e4b272c7ec924">DataType::QASYMM8</a>;</div><div class="line"><a name="l00299"></a><span class="lineno"> 299</span> </div><div class="line"><a name="l00300"></a><span class="lineno"> 300</span>  <span class="keywordflow">return</span> (output_qasymm8 && same_qinfo) || !output_qasymm8;</div><div class="line"><a name="l00301"></a><span class="lineno"> 301</span>  };</div><div class="line"><a name="l00302"></a><span class="lineno"> 302</span> </div><div class="line"><a name="l00303"></a><span class="lineno"> 303</span>  <a class="code" href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35d">Target</a> target = g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a8a979250f01a5edba059a02748b10ea3">nodes</a>()[0].get()->output(0)->desc().target;</div><div class="line"><a name="l00304"></a><span class="lineno"> 304</span> </div><div class="line"><a name="l00305"></a><span class="lineno"> 305</span>  <span class="comment">// Fusion mutations</span></div><div class="line"><a name="l00306"></a><span class="lineno"> 306</span>  detail::fuse_layer<BatchNormalizationLayerNode, ActivationLayerNode>(g, empty_prec, detail::fuse_node_with_activation<BatchNormalizationLayerNode>, supported_fused_activations);</div><div class="line"><a name="l00307"></a><span class="lineno"> 307</span>  detail::fuse_layer<ConvolutionLayerNode, ActivationLayerNode>(g, empty_prec, detail::fuse_node_with_activation<ConvolutionLayerNode>, supported_fused_activations);</div><div class="line"><a name="l00308"></a><span class="lineno"> 308</span>  detail::fuse_layer<DepthwiseConvolutionLayerNode, ActivationLayerNode>(g, qs8_prec, detail::fuse_node_with_activation<DepthwiseConvolutionLayerNode>, supported_fused_activations);</div><div class="line"><a name="l00309"></a><span class="lineno"> 309</span> </div><div class="line"><a name="l00310"></a><span class="lineno"> 310</span>  <span class="comment">// Currently fuse batch normalization brings performance uplift only on OpenCL with FP32 data type</span></div><div class="line"><a name="l00311"></a><span class="lineno"> 311</span>  <span class="comment">// TODO (COMPMID-2524): Fuse batch normalization with convolution and depthwise convolution at graph level for NEON - FP32</span></div><div class="line"><a name="l00312"></a><span class="lineno"> 312</span>  <span class="comment">// TODO (COMPMID-2581): Fuse batch normalization with convolution and depthwise convolution at graph level for OpenCL - FP16</span></div><div class="line"><a name="l00313"></a><span class="lineno"> 313</span>  <span class="keywordflow">if</span>(target == <a class="code" href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35da5bc574a47246f122016869b32a6aa6f0">Target::CL</a> && (g.<a class="code" href="classarm__compute_1_1graph_1_1_graph.xhtml#a8a979250f01a5edba059a02748b10ea3">nodes</a>()[0].get()->output(0)->desc().data_type == <a class="code" href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58a44ad4ef5a76e6aa6fb3e3fa079a54fda">DataType::F32</a>))</div><div class="line"><a name="l00314"></a><span class="lineno"> 314</span>  {</div><div class="line"><a name="l00315"></a><span class="lineno"> 315</span>  <span class="comment">//Depthwise Convolution and Batch Normalization Fusion active only for CL</span></div><div class="line"><a name="l00316"></a><span class="lineno"> 316</span>  detail::fuse_layer<ConvolutionLayerNode, BatchNormalizationLayerNode>(g, empty_prec, <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a501484d2b5f0213bbede4f44471c148b">detail::fuse_convolution_with_batch_normalization</a>);</div><div class="line"><a name="l00317"></a><span class="lineno"> 317</span>  detail::fuse_layer<DepthwiseConvolutionLayerNode, BatchNormalizationLayerNode>(g, empty_prec, <a class="code" href="namespacearm__compute_1_1graph_1_1detail.xhtml#a07811ee0f5c19da2658a647e6d7dc2fa">detail::fuse_depthwise_convolution_with_batch_normalization</a>);</div><div class="line"><a name="l00318"></a><span class="lineno"> 318</span>  }</div><div class="line"><a name="l00319"></a><span class="lineno"> 319</span> }</div><div class="line"><a name="l00320"></a><span class="lineno"> 320</span> } <span class="comment">// namespace graph</span></div><div class="line"><a name="l00321"></a><span class="lineno"> 321</span> } <span class="comment">// namespace arm_compute</span></div><div class="ttc" id="classarm__compute_1_1graph_1_1_i_node_xhtml_a6bca7473aa08cb0ecba36cb5dda2badf"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_i_node.xhtml#a6bca7473aa08cb0ecba36cb5dda2badf">arm_compute::graph::INode::input_edge</a></div><div class="ttdeci">Edge * input_edge(size_t idx) const</div><div class="ttdoc">Returns the edge of a given input of the node.</div><div class="ttdef"><b>Definition:</b> <a href="_i_node_8cpp_source.xhtml#l00171">INode.cpp:171</a></div></div> |
| <div class="ttc" id="structarm__compute_1_1graph_1_1_node_params_xhtml"><div class="ttname"><a href="structarm__compute_1_1graph_1_1_node_params.xhtml">arm_compute::graph::NodeParams</a></div><div class="ttdoc">Common node parameters.</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2graph_2_types_8h_source.xhtml#l00191">Types.h:191</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_a36fd3cfa2e1d33e59e1d3e95664d8b9c"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#a36fd3cfa2e1d33e59e1d3e95664d8b9c">arm_compute::graph::configure_tensor</a></div><div class="ttdeci">void configure_tensor(Tensor *tensor)</div><div class="ttdoc">Configures tensor.</div><div class="ttdef"><b>Definition:</b> <a href="src_2graph_2_utils_8cpp_source.xhtml#l00174">Utils.cpp:174</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_edge_xhtml_a592a2c76b00960964a3f6f2ef792a7f0"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_edge.xhtml#a592a2c76b00960964a3f6f2ef792a7f0">arm_compute::graph::Edge::consumer</a></div><div class="ttdeci">INode * consumer() const</div><div class="ttdoc">Returns consumer node.</div><div class="ttdef"><b>Definition:</b> <a href="_edge_8h_source.xhtml#l00092">Edge.h:92</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_fused_depthwise_convolution_batch_normalization_node_xhtml"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_fused_depthwise_convolution_batch_normalization_node.xhtml">arm_compute::graph::FusedDepthwiseConvolutionBatchNormalizationNode</a></div><div class="ttdoc">Fused Depthwise Convolution Batch Normalization node.</div><div class="ttdef"><b>Definition:</b> <a href="_fused_depthwise_convolution_batch_normalization_node_8h_source.xhtml#l00034">FusedDepthwiseConvolutionBatchNormalizationNode.h:34</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_a634230f98a5918f214e47d913c452d3b"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#a634230f98a5918f214e47d913c452d3b">arm_compute::graph::get_driving_nodes</a></div><div class="ttdeci">std::vector< NodeIdxPair > get_driving_nodes(const INode &node)</div><div class="ttdoc">Get the list of driving nodes of a given node.</div><div class="ttdef"><b>Definition:</b> <a href="src_2graph_2_utils_8cpp_source.xhtml#l00154">Utils.cpp:154</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a1f8aca235c095df227e7444f6b237eb1"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a1f8aca235c095df227e7444f6b237eb1">arm_compute::test::validation::act_info</a></div><div class="ttdeci">act_info</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_convolution_layer_8cpp_source.xhtml#l00175">ConvolutionLayer.cpp:175</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_activation_layer_info_xhtml_a56297e0f7b215eea46c818cb7528d9eaad346bb4679d29be241279f15d7795c1c"><div class="ttname"><a href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaad346bb4679d29be241279f15d7795c1c">arm_compute::ActivationLayerInfo::ActivationFunction::RELU</a></div><div class="ttdoc">Rectifier ( )</div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_a424221c5dfc8fe79f88a677d1b3b4494"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#a424221c5dfc8fe79f88a677d1b3b4494">arm_compute::graph::Graph::add_node</a></div><div class="ttdeci">NodeID add_node(Ts &&... args)</div><div class="ttdoc">Adds a node to the graph.</div><div class="ttdef"><b>Definition:</b> <a href="graph_2_graph_8h_source.xhtml#l00232">Graph.h:232</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a00525ff582f16038a1d3819aa44a23a3"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a00525ff582f16038a1d3819aa44a23a3">arm_compute::test::validation::conv_info</a></div><div class="ttdeci">conv_info</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_winograd_8cpp_source.xhtml#l00599">Winograd.cpp:599</a></div></div> |
| <div class="ttc" id="_asymm_helpers_8cpp_xhtml_a552dc3787d7ea1675f3e4e8993501d58"><div class="ttname"><a href="_asymm_helpers_8cpp.xhtml#a552dc3787d7ea1675f3e4e8993501d58">epsilon</a></div><div class="ttdeci">constexpr float epsilon</div><div class="ttdef"><b>Definition:</b> <a href="_asymm_helpers_8cpp_source.xhtml#l00033">AsymmHelpers.cpp:33</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_ab4e88c89b3b7ea1735996cc4def22d58a44ad4ef5a76e6aa6fb3e3fa079a54fda"><div class="ttname"><a href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58a44ad4ef5a76e6aa6fb3e3fa079a54fda">arm_compute::Format::F32</a></div><div class="ttdoc">1 channel, 1 F32 per channel</div></div> |
| <div class="ttc" id="_fused_convolution_batch_normalization_node_8h_xhtml"><div class="ttname"><a href="_fused_convolution_batch_normalization_node_8h.xhtml">FusedConvolutionBatchNormalizationNode.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_1_1detail_xhtml_a501484d2b5f0213bbede4f44471c148b"><div class="ttname"><a href="namespacearm__compute_1_1graph_1_1detail.xhtml#a501484d2b5f0213bbede4f44471c148b">arm_compute::graph::detail::fuse_convolution_with_batch_normalization</a></div><div class="ttdeci">void fuse_convolution_with_batch_normalization(Graph &g, const Edge *output_edge)</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00043">NodeFusionMutator.cpp:43</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a54a6080c9f4df1f908e57a9bbb46f5da"><div class="ttname"><a href="_error_8h.xhtml#a54a6080c9f4df1f908e57a9bbb46f5da">ARM_COMPUTE_ERROR_ON</a></div><div class="ttdeci">#define ARM_COMPUTE_ERROR_ON(cond)</div><div class="ttdoc">If the condition is true then an error message is printed and an exception thrown.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00337">Error.h:337</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_1_1detail_xhtml_a07811ee0f5c19da2658a647e6d7dc2fa"><div class="ttname"><a href="namespacearm__compute_1_1graph_1_1detail.xhtml#a07811ee0f5c19da2658a647e6d7dc2fa">arm_compute::graph::detail::fuse_depthwise_convolution_with_batch_normalization</a></div><div class="ttdeci">void fuse_depthwise_convolution_with_batch_normalization(Graph &g, const Edge *output_edge)</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00127">NodeFusionMutator.cpp:127</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml"><div class="ttname"><a href="namespacearm__compute.xhtml">arm_compute</a></div><div class="ttdoc">Copyright (c) 2017-2018 ARM Limited.</div><div class="ttdef"><b>Definition:</b> <a href="00__introduction_8dox_source.xhtml#l00024">00_introduction.dox:24</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_fused_convolution_batch_normalization_node_xhtml"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_fused_convolution_batch_normalization_node.xhtml">arm_compute::graph::FusedConvolutionBatchNormalizationNode</a></div><div class="ttdoc">Batch Normalization node.</div><div class="ttdef"><b>Definition:</b> <a href="_fused_convolution_batch_normalization_node_8h_source.xhtml#l00034">FusedConvolutionBatchNormalizationNode.h:34</a></div></div> |
| <div class="ttc" id="_cast_8h_xhtml"><div class="ttname"><a href="_cast_8h.xhtml">Cast.h</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_tensor_xhtml_a0d3608f94078b90ab1ff9e9465d4ed3b"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_tensor.xhtml#a0d3608f94078b90ab1ff9e9465d4ed3b">arm_compute::graph::Tensor::desc</a></div><div class="ttdeci">TensorDescriptor & desc()</div><div class="ttdoc">TensorInfo metadata accessor.</div><div class="ttdef"><b>Definition:</b> <a href="graph_2_tensor_8cpp_source.xhtml#l00040">Tensor.cpp:40</a></div></div> |
| <div class="ttc" id="structarm__compute_1_1graph_1_1_tensor_descriptor_xhtml_a65bc62ff84efcb7e4a410600480a4dc9"><div class="ttname"><a href="structarm__compute_1_1graph_1_1_tensor_descriptor.xhtml#a65bc62ff84efcb7e4a410600480a4dc9">arm_compute::graph::TensorDescriptor::quant_info</a></div><div class="ttdeci">QuantizationInfo quant_info</div><div class="ttdoc">Quantization info.</div><div class="ttdef"><b>Definition:</b> <a href="_tensor_descriptor_8h_source.xhtml#l00112">TensorDescriptor.h:112</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_i_node_xhtml"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_i_node.xhtml">arm_compute::graph::INode</a></div><div class="ttdoc">Node interface.</div><div class="ttdef"><b>Definition:</b> <a href="_i_node_8h_source.xhtml#l00045">INode.h:45</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_i_node_xhtml_a27369471d9b2ba47746e1e923f585b9f"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_i_node.xhtml#a27369471d9b2ba47746e1e923f585b9f">arm_compute::graph::INode::output</a></div><div class="ttdeci">Tensor * output(size_t idx) const</div><div class="ttdoc">Returns the tensor of a given output of the node.</div><div class="ttdef"><b>Definition:</b> <a href="_i_node_8cpp_source.xhtml#l00158">INode.cpp:158</a></div></div> |
| <div class="ttc" id="arm__compute_2graph_2_utils_8h_xhtml"><div class="ttname"><a href="arm__compute_2graph_2_utils_8h.xhtml">Utils.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_ad8ed01ff3ff33333d8e19db4d2818bb6af14462d71aa842202c3e4b272c7ec924"><div class="ttname"><a href="namespacearm__compute.xhtml#ad8ed01ff3ff33333d8e19db4d2818bb6af14462d71aa842202c3e4b272c7ec924">arm_compute::DataType::QASYMM8</a></div><div class="ttdoc">quantized, asymmetric fixed-point 8-bit number</div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a2270b3e1d20651d2d8341c858c890830"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a2270b3e1d20651d2d8341c858c890830">arm_compute::test::validation::num_groups</a></div><div class="ttdeci">const unsigned int num_groups</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_o_n_2_im2_col_8cpp_source.xhtml#l00148">Im2Col.cpp:148</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_a09f8d22de4cd2a2881730ad58096c7c1"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#a09f8d22de4cd2a2881730ad58096c7c1">arm_compute::graph::Graph::remove_node</a></div><div class="ttdeci">bool remove_node(NodeID nid)</div><div class="ttdoc">Remove the node with the given ID.</div><div class="ttdef"><b>Definition:</b> <a href="_graph_8cpp_source.xhtml#l00035">Graph.cpp:35</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_edge_xhtml_a73b54dbd7dc72560e805067f2e68207c"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_edge.xhtml#a73b54dbd7dc72560e805067f2e68207c">arm_compute::graph::Edge::producer_id</a></div><div class="ttdeci">NodeID producer_id() const</div><div class="ttdoc">Returns producer node id.</div><div class="ttdef"><b>Definition:</b> <a href="_edge_8h_source.xhtml#l00068">Edge.h:68</a></div></div> |
| <div class="ttc" id="graph_2_logger_8h_xhtml"><div class="ttname"><a href="graph_2_logger_8h.xhtml">Logger.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_1_1detail_xhtml_a1f9900d626af4230ff61f851e8d5eab5"><div class="ttname"><a href="namespacearm__compute_1_1graph_1_1detail.xhtml#a1f9900d626af4230ff61f851e8d5eab5">arm_compute::graph::detail::fuse_layer</a></div><div class="ttdeci">void fuse_layer(Graph &g, std::function< bool(INode &)> const &prec, const F fuse_fcn, Args &&... optional_arguments)</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00254">NodeFusionMutator.cpp:254</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_a538f789bf074c367457a6f8f32b83d2d"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#a538f789bf074c367457a6f8f32b83d2d">arm_compute::graph::Graph::add_connection</a></div><div class="ttdeci">EdgeID add_connection(NodeID source, size_t source_idx, NodeID sink, size_t sink_idx)</div><div class="ttdoc">Adds a connection between two nodes.</div><div class="ttdef"><b>Definition:</b> <a href="_graph_8cpp_source.xhtml#l00069">Graph.cpp:69</a></div></div> |
| <div class="ttc" id="_node_fusion_mutator_8h_xhtml"><div class="ttname"><a href="_node_fusion_mutator_8h.xhtml">NodeFusionMutator.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_a31488d29805a596498c0234ae392d35d"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35d">arm_compute::graph::Target</a></div><div class="ttdeci">Target</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2graph_2_types_8h_source.xhtml#l00089">Types.h:89</a></div></div> |
| <div class="ttc" id="_graph_builder_8h_xhtml"><div class="ttname"><a href="_graph_builder_8h.xhtml">GraphBuilder.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_ac85a46f3ebd3ab09f576a994ac2dce11"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#ac85a46f3ebd3ab09f576a994ac2dce11">arm_compute::graph::FastMathHint</a></div><div class="ttdeci">FastMathHint</div><div class="ttdoc">Enable or disable fast math for Convolution layer.</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2graph_2_types_8h_source.xhtml#l00123">Types.h:123</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_activation_layer_info_xhtml_a56297e0f7b215eea46c818cb7528d9eaaab1d4411a9e7f5e82002512cddfdc33a"><div class="ttname"><a href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaaab1d4411a9e7f5e82002512cddfdc33a">arm_compute::ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU</a></div><div class="ttdoc">Lower and Upper Bounded Rectifier ( )</div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml">arm_compute::graph::Graph</a></div><div class="ttdoc">Graph class.</div><div class="ttdef"><b>Definition:</b> <a href="graph_2_graph_8h_source.xhtml#l00050">Graph.h:50</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_a56952a8591a9d8481450ae707ae355e7"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#a56952a8591a9d8481450ae707ae355e7">arm_compute::graph::NodeID</a></div><div class="ttdeci">unsigned int NodeID</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2graph_2_types_8h_source.xhtml#l00064">Types.h:64</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_a8a979250f01a5edba059a02748b10ea3"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#a8a979250f01a5edba059a02748b10ea3">arm_compute::graph::Graph::nodes</a></div><div class="ttdeci">const std::vector< NodeID > & nodes(NodeType type)</div><div class="ttdoc">Returns graph input nodes.</div><div class="ttdef"><b>Definition:</b> <a href="_graph_8cpp_source.xhtml#l00174">Graph.cpp:174</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_node_fusion_mutator_xhtml_a0ee2c922a45715b33e30f83f8c005b68"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#a0ee2c922a45715b33e30f83f8c005b68">arm_compute::graph::NodeFusionMutator::name</a></div><div class="ttdeci">const char * name() override</div><div class="ttdoc">Returns mutator name.</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00275">NodeFusionMutator.cpp:275</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_edge_xhtml"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_edge.xhtml">arm_compute::graph::Edge</a></div><div class="ttdoc">Graph Edge.</div><div class="ttdef"><b>Definition:</b> <a href="_edge_8h_source.xhtml#l00039">Edge.h:39</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_activation_layer_info_xhtml_a56297e0f7b215eea46c818cb7528d9eaacc516ab03b98f1c908ddf6ed4a7c45e9"><div class="ttname"><a href="classarm__compute_1_1_activation_layer_info.xhtml#a56297e0f7b215eea46c818cb7528d9eaacc516ab03b98f1c908ddf6ed4a7c45e9">arm_compute::ActivationLayerInfo::ActivationFunction::BOUNDED_RELU</a></div><div class="ttdoc">Upper Bounded Rectifier ( )</div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_af8baf1f3da6d42a94d0569395ece882a"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#af8baf1f3da6d42a94d0569395ece882a">arm_compute::graph::Graph::node</a></div><div class="ttdeci">const INode * node(NodeID id) const</div><div class="ttdoc">Get node object given its id.</div><div class="ttdef"><b>Definition:</b> <a href="_graph_8cpp_source.xhtml#l00204">Graph.cpp:204</a></div></div> |
| <div class="ttc" id="structarm__compute_1_1graph_1_1_node_params_xhtml_a9b45b3e13bd9167aab02e17e08916231"><div class="ttname"><a href="structarm__compute_1_1graph_1_1_node_params.xhtml#a9b45b3e13bd9167aab02e17e08916231">arm_compute::graph::NodeParams::name</a></div><div class="ttdeci">std::string name</div><div class="ttdoc">Node name.</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2graph_2_types_8h_source.xhtml#l00193">Types.h:193</a></div></div> |
| <div class="ttc" id="graph_2_logger_8h_xhtml_a300d153929a99c7b571d4cda3f7987a5"><div class="ttname"><a href="graph_2_logger_8h.xhtml#a300d153929a99c7b571d4cda3f7987a5">ARM_COMPUTE_LOG_GRAPH_VERBOSE</a></div><div class="ttdeci">#define ARM_COMPUTE_LOG_GRAPH_VERBOSE(x)</div><div class="ttdef"><b>Definition:</b> <a href="graph_2_logger_8h_source.xhtml#l00050">Logger.h:50</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_graph_xhtml_a0ea47db036b1c140af002ee1494dcb6f"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_graph.xhtml#a0ea47db036b1c140af002ee1494dcb6f">arm_compute::graph::Graph::edge</a></div><div class="ttdeci">const Edge * edge(EdgeID id) const</div><div class="ttdoc">Get edge object given its id.</div><div class="ttdef"><b>Definition:</b> <a href="_graph_8cpp_source.xhtml#l00214">Graph.cpp:214</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_edge_xhtml_ad00e584f78c622e5ee9ec9613f6d6633"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_edge.xhtml#ad00e584f78c622e5ee9ec9613f6d6633">arm_compute::graph::Edge::consumer_id</a></div><div class="ttdeci">NodeID consumer_id() const</div><div class="ttdoc">Returns sink node id.</div><div class="ttdef"><b>Definition:</b> <a href="_edge_8h_source.xhtml#l00076">Edge.h:76</a></div></div> |
| <div class="ttc" id="_nodes_8h_xhtml"><div class="ttname"><a href="_nodes_8h.xhtml">Nodes.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_1_1detail_xhtml_a36b5b67e26dfdfbedf39e6703ef059bd"><div class="ttname"><a href="namespacearm__compute_1_1graph_1_1detail.xhtml#a36b5b67e26dfdfbedf39e6703ef059bd">arm_compute::graph::detail::fuse_node_with_activation</a></div><div class="ttdeci">void fuse_node_with_activation(Graph &g, const Edge *output_edge, const std::set< Activation > &supported_fused_activations)</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00205">NodeFusionMutator.cpp:205</a></div></div> |
| <div class="ttc" id="_backend_registry_8h_xhtml"><div class="ttname"><a href="_backend_registry_8h.xhtml">BackendRegistry.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1graph_xhtml_a31488d29805a596498c0234ae392d35da5bc574a47246f122016869b32a6aa6f0"><div class="ttname"><a href="namespacearm__compute_1_1graph.xhtml#a31488d29805a596498c0234ae392d35da5bc574a47246f122016869b32a6aa6f0">arm_compute::graph::Target::CL</a></div><div class="ttdoc">OpenCL capable target device.</div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_edge_xhtml_acef46a426a99b126a412e361125f2ce9"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_edge.xhtml#acef46a426a99b126a412e361125f2ce9">arm_compute::graph::Edge::producer</a></div><div class="ttdeci">INode * producer() const</div><div class="ttdoc">Returns producer node.</div><div class="ttdef"><b>Definition:</b> <a href="_edge_8h_source.xhtml#l00084">Edge.h:84</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1graph_1_1_node_fusion_mutator_xhtml_af6705a49326de235df4e9b6028bdcdce"><div class="ttname"><a href="classarm__compute_1_1graph_1_1_node_fusion_mutator.xhtml#af6705a49326de235df4e9b6028bdcdce">arm_compute::graph::NodeFusionMutator::mutate</a></div><div class="ttdeci">virtual void mutate(Graph &g) override</div><div class="ttdoc">Walk the graph and perform a specific mutation.</div><div class="ttdef"><b>Definition:</b> <a href="_node_fusion_mutator_8cpp_source.xhtml#l00280">NodeFusionMutator.cpp:280</a></div></div> |
| </div><!-- fragment --></div><!-- contents --> |
| </div><!-- doc-content --> |
| <!-- start footer part --> |
| <div id="nav-path" class="navpath"><!-- id is needed for treeview function! --> |
| <ul> |
| <li class="navelem"><a class="el" href="dir_68267d1309a1af8e8297ef4c3efbcdba.xhtml">src</a></li><li class="navelem"><a class="el" href="dir_d182250f0c231765fc862e69db210731.xhtml">graph</a></li><li class="navelem"><a class="el" href="dir_1b9e532a4c8623825d945e964c6e4c7f.xhtml">mutators</a></li><li class="navelem"><a class="el" href="_node_fusion_mutator_8cpp.xhtml">NodeFusionMutator.cpp</a></li> |
| <li class="footer">Generated on Mon Sep 2 2019 11:47:25 for Compute Library by |
| <a href="http://www.doxygen.org/index.html"> |
| <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.15 </li> |
| </ul> |
| </div> |
| </body> |
| </html> |
