| <!-- 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/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.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('_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_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">NEGEMMLowpOffsetContributionOutputStageKernel.cpp</div> </div> |
| </div><!--header--> |
| <div class="contents"> |
| <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_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) 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="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8h.xhtml">arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.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="_access_window_static_8h.xhtml">arm_compute/core/AccessWindowStatic.h</a>"</span></div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span> <span class="preprocessor">#include "<a class="code" href="_error_8h.xhtml">arm_compute/core/Error.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_2core_2_helpers_8h.xhtml">arm_compute/core/Helpers.h</a>"</span></div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span> <span class="preprocessor">#include "<a class="code" href="_i_tensor_8h.xhtml">arm_compute/core/ITensor.h</a>"</span></div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span> <span class="preprocessor">#include "<a class="code" href="_n_e_asymm_8h.xhtml">arm_compute/core/NEON/NEAsymm.h</a>"</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span> <span class="preprocessor">#include "<a class="code" href="wrapper_8h.xhtml">arm_compute/core/NEON/wrapper/wrapper.h</a>"</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span> <span class="preprocessor">#include "<a class="code" href="_tensor_info_8h.xhtml">arm_compute/core/TensorInfo.h</a>"</span></div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span> <span class="preprocessor">#include "<a class="code" href="arm__compute_2core_2_types_8h.xhtml">arm_compute/core/Types.h</a>"</span></div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span> <span class="preprocessor">#include "<a class="code" href="arm__compute_2core_2_utils_8h.xhtml">arm_compute/core/Utils.h</a>"</span></div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span> <span class="preprocessor">#include "<a class="code" href="_validate_8h.xhtml">arm_compute/core/Validate.h</a>"</span></div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span> <span class="preprocessor">#include "<a class="code" href="_window_8h.xhtml">arm_compute/core/Window.h</a>"</span></div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span> </div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span> <span class="preprocessor">#include <arm_neon.h></span></div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span> <span class="preprocessor">#include <cstddef></span></div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span> <span class="preprocessor">#include <cstdint></span></div><div class="line"><a name="l00041"></a><span class="lineno"> 41</span> <span class="preprocessor">#include <map></span></div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span> </div><div class="line"><a name="l00043"></a><span class="lineno"> 43</span> <span class="keyword">namespace </span><a class="code" href="namespacearm__compute.xhtml">arm_compute</a></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> <span class="keyword">class </span>Coordinates;</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">namespace</span></div><div class="line"><a name="l00048"></a><span class="lineno"> 48</span> {</div><div class="line"><a name="l00049"></a><span class="lineno"> 49</span> <span class="keyword">inline</span> int32x4x4_t load_results_input(<span class="keyword">const</span> Iterator &mm_result_it, int32_t x)</div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span> {</div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span>  <span class="keywordflow">return</span></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>  {</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>  vld1q_s32(reinterpret_cast<const int32_t *>(mm_result_it.ptr()) + x + 0),</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span>  vld1q_s32(reinterpret_cast<const int32_t *>(mm_result_it.ptr()) + x + 4),</div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span>  vld1q_s32(reinterpret_cast<const int32_t *>(mm_result_it.ptr()) + x + 8),</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>  vld1q_s32(reinterpret_cast<const int32_t *>(mm_result_it.ptr()) + x + 12)</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>  };</div><div class="line"><a name="l00060"></a><span class="lineno"> 60</span> }</div><div class="line"><a name="l00061"></a><span class="lineno"> 61</span> </div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span> <span class="keyword">inline</span> int32x4x4_t load(<span class="keyword">const</span> int32_t *ptr, int32_t x)</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="keywordflow">return</span></div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  {</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  {</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span>  vld1q_s32(ptr + x + 0),</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  vld1q_s32(ptr + x + 4),</div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>  vld1q_s32(ptr + x + 8),</div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span>  vld1q_s32(ptr + x + 12)</div><div class="line"><a name="l00071"></a><span class="lineno"> 71</span>  }</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> }</div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span> </div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span> <span class="keyword">inline</span> int32x4x4_t get_a_offset(<span class="keyword">const</span> int32_t *vector_sum_col_ptr, int32_t a_offset, int32_t x)</div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span> {</div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span>  int32x4x4_t a_offset_term_s32 = load(vector_sum_col_ptr, x);</div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span> </div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span>  a_offset_term_s32.val[0] = vmulq_n_s32(a_offset_term_s32.val[0], a_offset);</div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span>  a_offset_term_s32.val[1] = vmulq_n_s32(a_offset_term_s32.val[1], a_offset);</div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>  a_offset_term_s32.val[2] = vmulq_n_s32(a_offset_term_s32.val[2], a_offset);</div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  a_offset_term_s32.val[3] = vmulq_n_s32(a_offset_term_s32.val[3], a_offset);</div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>  <span class="keywordflow">return</span> a_offset_term_s32;</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> </div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span> <span class="keyword">inline</span> int32x4_t get_b_offset(<span class="keyword">const</span> int32_t *vector_sum_row_ptr, int32_t b_offset)</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>  int32x4_t b_offset_term_s32 = vld1q_dup_s32(vector_sum_row_ptr);</div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  b_offset_term_s32 = vmulq_n_s32(b_offset_term_s32, b_offset);</div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>  <span class="keywordflow">return</span> b_offset_term_s32;</div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span> }</div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span> </div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span> <span class="keyword">inline</span> int32x4x4_t get_k_offset(int32_t k_offset)</div><div class="line"><a name="l00094"></a><span class="lineno"> 94</span> {</div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  <span class="keywordflow">return</span></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>  {</div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>  vdupq_n_s32(k_offset),</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span>  vdupq_n_s32(k_offset),</div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  vdupq_n_s32(k_offset),</div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>  vdupq_n_s32(k_offset)</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>  }</div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span>  };</div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span> }</div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span> </div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span> <span class="keyword">template</span> <<span class="keywordtype">bool</span> is_bounded_relu></div><div class="line"><a name="l00107"></a><span class="lineno"> 107</span> <span class="keyword">inline</span> uint8x16_t finalize_quantization_floating_point(int32x4x4_t &in_s32, int32x4_t result_shift_s32, uint8x16_t min_u8, uint8x16_t max_u8)</div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span> {</div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>  <span class="keyword">const</span> <span class="keyword">static</span> int32x4_t zero_s32 = vdupq_n_s32(0);</div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span> </div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>  <span class="comment">// Shift final result (negative value shift right)</span></div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span>  in_s32.val[0] = vshlq_s32(in_s32.val[0], result_shift_s32);</div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>  in_s32.val[1] = vshlq_s32(in_s32.val[1], result_shift_s32);</div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>  in_s32.val[2] = vshlq_s32(in_s32.val[2], result_shift_s32);</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  in_s32.val[3] = vshlq_s32(in_s32.val[3], result_shift_s32);</div><div class="line"><a name="l00116"></a><span class="lineno"> 116</span> </div><div class="line"><a name="l00117"></a><span class="lineno"> 117</span>  <span class="comment">// Saturate negative values</span></div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  in_s32.val[0] = vmaxq_s32(in_s32.val[0], zero_s32);</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  in_s32.val[1] = vmaxq_s32(in_s32.val[1], zero_s32);</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  in_s32.val[2] = vmaxq_s32(in_s32.val[2], zero_s32);</div><div class="line"><a name="l00121"></a><span class="lineno"> 121</span>  in_s32.val[3] = vmaxq_s32(in_s32.val[3], zero_s32);</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>  <span class="comment">// Convert S32 to S16</span></div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>  <span class="keyword">const</span> int16x8x2_t in_s16 =</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"> 127</span>  vcombine_s16(vqmovn_s32(in_s32.val[0]), vqmovn_s32(in_s32.val[1])),</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>  vcombine_s16(vqmovn_s32(in_s32.val[2]), vqmovn_s32(in_s32.val[3]))</div><div class="line"><a name="l00129"></a><span class="lineno"> 129</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> </div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  <span class="comment">// Convert S16 to U8</span></div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  uint8x16_t out_u8 = vcombine_u8(vqmovun_s16(in_s16.val[0]), vqmovun_s16(in_s16.val[1]));</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span> </div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>  <span class="keywordflow">if</span>(is_bounded_relu)</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>  out_u8 = vmaxq_u8(out_u8, min_u8);</div><div class="line"><a name="l00138"></a><span class="lineno"> 138</span>  out_u8 = vminq_u8(out_u8, max_u8);</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> </div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span>  <span class="keywordflow">return</span> out_u8;</div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span> }</div><div class="line"><a name="l00143"></a><span class="lineno"> 143</span> </div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span> <span class="keyword">inline</span> Window get_win_vector_sum(<span class="keyword">const</span> Window &window)</div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span> {</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>  Window win_vector_sum(window);</div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  win_vector_sum.set(<a class="code" href="classarm__compute_1_1_window.xhtml#ad2d402364fa822b0b7775081291eeca9">Window::DimY</a>, Window::Dimension(0, 0, 0));</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>  win_vector_sum.set(<a class="code" href="classarm__compute_1_1_window.xhtml#a893d17b56b9abc4423ce26e9a24ac5dc">Window::DimZ</a>, Window::Dimension(0, 0, 0));</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span>  <span class="keywordflow">return</span> win_vector_sum;</div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span> }</div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span> </div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span> <span class="keyword">inline</span> Iterator get_vector_sum_col_it(<span class="keyword">const</span> Window &window, <span class="keyword">const</span> ITensor *vector_sum_col)</div><div class="line"><a name="l00153"></a><span class="lineno"> 153</span> {</div><div class="line"><a name="l00154"></a><span class="lineno"> 154</span>  Iterator vector_sum_col_it(vector_sum_col, get_win_vector_sum(window));</div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span>  <span class="keywordflow">return</span> vector_sum_col_it;</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> </div><div class="line"><a name="l00158"></a><span class="lineno"> 158</span> <span class="keyword">inline</span> Iterator get_vector_sum_row_it(<span class="keyword">const</span> Window &window, <span class="keyword">const</span> ITensor *vector_sum_row)</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>  Window win_vector_sum_row = get_win_vector_sum(window);</div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span>  win_vector_sum_row.set(<a class="code" href="classarm__compute_1_1_window.xhtml#aa96e81276ee4f87ab386cd05a5539a7d">Window::DimX</a>, Window::Dimension(0, 0, 0));</div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  Iterator vector_sum_row_it(vector_sum_row, win_vector_sum_row);</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  <span class="keywordflow">return</span> vector_sum_row_it;</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="keyword">inline</span> Iterator get_bias_it(<span class="keyword">const</span> Window &window, <span class="keyword">const</span> ITensor *<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>)</div><div class="line"><a name="l00167"></a><span class="lineno"> 167</span> {</div><div class="line"><a name="l00168"></a><span class="lineno"> 168</span>  Window win_bias(window);</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>  win_bias.set(<a class="code" href="classarm__compute_1_1_window.xhtml#ad2d402364fa822b0b7775081291eeca9">Window::DimY</a>, Window::Dimension(0, 1, 1));</div><div class="line"><a name="l00170"></a><span class="lineno"> 170</span>  win_bias.set(<a class="code" href="classarm__compute_1_1_window.xhtml#a893d17b56b9abc4423ce26e9a24ac5dc">Window::DimZ</a>, Window::Dimension(0, 1, 1));</div><div class="line"><a name="l00171"></a><span class="lineno"> 171</span>  Iterator bias_it(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, win_bias);</div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  <span class="keywordflow">return</span> bias_it;</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> </div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span> <span class="keyword">inline</span> int32x4x4_t add_s32(int32x4x4_t a, int32x4_t <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>)</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="keywordflow">return</span></div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>  {</div><div class="line"><a name="l00179"></a><span class="lineno"> 179</span>  {</div><div class="line"><a name="l00180"></a><span class="lineno"> 180</span>  vaddq_s32(a.val[0], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>),</div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  vaddq_s32(a.val[1], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>),</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>  vaddq_s32(a.val[2], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>),</div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>  vaddq_s32(a.val[3], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>)</div><div class="line"><a name="l00184"></a><span class="lineno"> 184</span>  }</div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span>  };</div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span> }</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span> </div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span> <span class="keyword">inline</span> int32x4x4_t add_s32(int32x4x4_t a, int32x4x4_t <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>)</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="keywordflow">return</span></div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  {</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  {</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  vaddq_s32(a.val[0], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>.val[0]),</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span>  vaddq_s32(a.val[1], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>.val[1]),</div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span>  vaddq_s32(a.val[2], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>.val[2]),</div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>  vaddq_s32(a.val[3], <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">b</a>.val[3])</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>  };</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> </div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span> <span class="keyword">inline</span> int32x4x4_t mul_s32(int32x4x4_t &a, int32_t mul_scalar)</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>  <span class="keywordflow">return</span></div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  {</div><div class="line"><a name="l00205"></a><span class="lineno"> 205</span>  {</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span>  vmulq_n_s32(a.val[0], mul_scalar),</div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span>  vmulq_n_s32(a.val[1], mul_scalar),</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>  vmulq_n_s32(a.val[2], mul_scalar),</div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>  vmulq_n_s32(a.val[3], mul_scalar)</div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span>  }</div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span>  };</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</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="keyword">template</span> <<span class="keywordtype">bool</span> has_a_offset, <span class="keywordtype">bool</span> has_b_offset, <span class="keywordtype">bool</span> has_bias, <span class="keywordtype">bool</span> is_bounded_relu, <span class="keywordtype">bool</span> is_fixed_po<span class="keywordtype">int</span>></div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span> <span class="keyword">inline</span> <span class="keywordtype">void</span> run_offset_contribution_output_stage_window(<span class="keyword">const</span> int32_t *vector_sum_col_ptr, <span class="keyword">const</span> int32_t *vector_sum_row_ptr, <span class="keyword">const</span> int32_t *bias_ptr, Iterator mm_result_it, Iterator out_it,</div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span>  <span class="keyword">const</span> int32x4_t result_offset_s32, <span class="keyword">const</span> int32x4_t result_shift_s32, uint8x16_t min_u8, uint8x16_t max_u8,</div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>  int32_t a_offset, int32_t b_offset, int32_t k_offset,</div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  GEMMLowpOutputStageInfo output_stage, <span class="keywordtype">int</span> window_step_x, <span class="keywordtype">int</span> window_start_x, <span class="keywordtype">int</span> window_end_x)</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>  int32x4x4_t offset_term_s32 = { 0, 0, 0, 0 };</div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  <span class="keywordflow">if</span>(!is_fixed_point)</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">// Combine quantization offset with other offsets.</span></div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span>  offset_term_s32 = add_s32(offset_term_s32, result_offset_s32);</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="keywordflow">if</span>(has_a_offset && has_b_offset)</div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  {</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span>  offset_term_s32 = add_s32(offset_term_s32, get_k_offset(k_offset));</div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  }</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span>  <span class="keywordflow">if</span>(has_b_offset)</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>  offset_term_s32 = add_s32(offset_term_s32, get_b_offset(vector_sum_row_ptr, b_offset));</div><div class="line"><a name="l00233"></a><span class="lineno"> 233</span>  }</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="keywordtype">int</span> x = window_start_x;</div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span>  <span class="keywordflow">for</span>(; x <= (window_end_x - window_step_x); x += window_step_x)</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>  int32x4x4_t in_s32 = load_results_input(mm_result_it, x);</div><div class="line"><a name="l00239"></a><span class="lineno"> 239</span> </div><div class="line"><a name="l00240"></a><span class="lineno"> 240</span>  <span class="keywordflow">if</span>(has_a_offset)</div><div class="line"><a name="l00241"></a><span class="lineno"> 241</span>  {</div><div class="line"><a name="l00242"></a><span class="lineno"> 242</span>  in_s32 = add_s32(in_s32, get_a_offset(vector_sum_col_ptr, a_offset, x));</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="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a9aeced5a5128f60a31ea3e327a45ee21">has_bias</a>)</div><div class="line"><a name="l00245"></a><span class="lineno"> 245</span>  {</div><div class="line"><a name="l00246"></a><span class="lineno"> 246</span>  in_s32 = add_s32(in_s32, load(bias_ptr, x));</div><div class="line"><a name="l00247"></a><span class="lineno"> 247</span>  }</div><div class="line"><a name="l00248"></a><span class="lineno"> 248</span>  <span class="keywordflow">if</span>(!is_fixed_point || has_b_offset)</div><div class="line"><a name="l00249"></a><span class="lineno"> 249</span>  {</div><div class="line"><a name="l00250"></a><span class="lineno"> 250</span>  in_s32 = add_s32(in_s32, offset_term_s32);</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>  <span class="keywordflow">if</span>(!is_fixed_point)</div><div class="line"><a name="l00253"></a><span class="lineno"> 253</span>  {</div><div class="line"><a name="l00254"></a><span class="lineno"> 254</span>  in_s32 = mul_s32(in_s32, output_stage.gemmlowp_multiplier);</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> </div><div class="line"><a name="l00257"></a><span class="lineno"> 257</span>  <span class="keywordflow">if</span>(is_fixed_point)</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>  vst1q_u8(out_it.ptr() + x, finalize_quantization<is_bounded_relu>(in_s32, output_stage.gemmlowp_multiplier, output_stage.gemmlowp_shift, result_offset_s32, min_u8, max_u8));</div><div class="line"><a name="l00260"></a><span class="lineno"> 260</span>  }</div><div class="line"><a name="l00261"></a><span class="lineno"> 261</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00262"></a><span class="lineno"> 262</span>  {</div><div class="line"><a name="l00263"></a><span class="lineno"> 263</span>  vst1q_u8(out_it.ptr() + x, finalize_quantization_floating_point<is_bounded_relu>(in_s32, result_shift_s32, min_u8, max_u8));</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>  }</div><div class="line"><a name="l00266"></a><span class="lineno"> 266</span>  <span class="comment">// Compute left-over elements</span></div><div class="line"><a name="l00267"></a><span class="lineno"> 267</span>  <span class="keywordflow">for</span>(; x < window_end_x; ++x)</div><div class="line"><a name="l00268"></a><span class="lineno"> 268</span>  {</div><div class="line"><a name="l00269"></a><span class="lineno"> 269</span>  int32_t in_value = *(reinterpret_cast<const int32_t *>(mm_result_it.ptr()) + x) + <a class="code" href="namespacearm__compute_1_1wrapper.xhtml#aa16ace001ab8287faa46d6962f369219">wrapper::vgetlane</a>(offset_term_s32.val[0], 0);</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>  <span class="keywordflow">if</span>(has_a_offset)</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>  in_value += (*(vector_sum_col_ptr + x) * a_offset);</div><div class="line"><a name="l00274"></a><span class="lineno"> 274</span>  }</div><div class="line"><a name="l00275"></a><span class="lineno"> 275</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a9aeced5a5128f60a31ea3e327a45ee21">has_bias</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>  in_value += *(bias_ptr + x);</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"> 280</span>  <span class="keywordflow">if</span>(is_fixed_point)</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">// Finalize and store the result</span></div><div class="line"><a name="l00283"></a><span class="lineno"> 283</span>  *(out_it.ptr() + x) = finalize_quantization<is_bounded_relu>(in_value, output_stage.gemmlowp_multiplier, output_stage.gemmlowp_shift,</div><div class="line"><a name="l00284"></a><span class="lineno"> 284</span>  output_stage.gemmlowp_offset, static_cast<uint8_t>(output_stage.gemmlowp_min_bound), static_cast<uint8_t>(output_stage.gemmlowp_max_bound));</div><div class="line"><a name="l00285"></a><span class="lineno"> 285</span>  }</div><div class="line"><a name="l00286"></a><span class="lineno"> 286</span>  <span class="keywordflow">else</span></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="comment">// Finalize quantization</span></div><div class="line"><a name="l00289"></a><span class="lineno"> 289</span>  in_value = (in_value * output_stage.gemmlowp_multiplier) >> output_stage.gemmlowp_shift;</div><div class="line"><a name="l00290"></a><span class="lineno"> 290</span> </div><div class="line"><a name="l00291"></a><span class="lineno"> 291</span>  <span class="comment">// Bound and store the result</span></div><div class="line"><a name="l00292"></a><span class="lineno"> 292</span>  <span class="keywordflow">if</span>(is_bounded_relu)</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>  in_value = static_cast<uint8_t>(std::max<int32_t>(output_stage.gemmlowp_min_bound, std::min<int32_t>(output_stage.gemmlowp_max_bound, in_value)));</div><div class="line"><a name="l00295"></a><span class="lineno"> 295</span>  }</div><div class="line"><a name="l00296"></a><span class="lineno"> 296</span>  *(out_it.ptr() + x) = static_cast<uint8_t>(std::max<int32_t>(0, std::min<int32_t>(255, in_value)));</div><div class="line"><a name="l00297"></a><span class="lineno"> 297</span>  }</div><div class="line"><a name="l00298"></a><span class="lineno"> 298</span>  }</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> </div><div class="line"><a name="l00301"></a><span class="lineno"> 301</span> <span class="keyword">template</span> <<span class="keywordtype">bool</span> is_gemm3d, <span class="keywordtype">bool</span> is_bounded_relu, <span class="keywordtype">bool</span> is_fixed_po<span class="keywordtype">int</span>></div><div class="line"><a name="l00302"></a><span class="lineno"> 302</span> <span class="keywordtype">void</span> run_offset_contribution_output_stage(<span class="keyword">const</span> Window &window,</div><div class="line"><a name="l00303"></a><span class="lineno"> 303</span>  <span class="keyword">const</span> ITensor *mm_result, <span class="keyword">const</span> ITensor *vector_sum_col, <span class="keyword">const</span> ITensor *vector_sum_row, <span class="keyword">const</span> ITensor *<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, ITensor *output,</div><div class="line"><a name="l00304"></a><span class="lineno"> 304</span>  int32_t a_offset, int32_t b_offset, int32_t k_offset, <span class="keywordtype">bool</span> slide_vector_sum_col,</div><div class="line"><a name="l00305"></a><span class="lineno"> 305</span>  GEMMLowpOutputStageInfo output_stage)</div><div class="line"><a name="l00306"></a><span class="lineno"> 306</span> {</div><div class="line"><a name="l00307"></a><span class="lineno"> 307</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> height_input = is_gemm3d ? mm_result->info()->dimension(1) : 0;</div><div class="line"><a name="l00308"></a><span class="lineno"> 308</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> depth_input = is_gemm3d ? mm_result->info()->dimension(2) : 1;</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="keyword">const</span> int32x4_t result_offset_s32 = vdupq_n_s32(output_stage.gemmlowp_offset);</div><div class="line"><a name="l00311"></a><span class="lineno"> 311</span>  <span class="keyword">const</span> int32x4_t result_shift_s32 = vdupq_n_s32(is_fixed_point ? output_stage.gemmlowp_shift : -output_stage.gemmlowp_shift);</div><div class="line"><a name="l00312"></a><span class="lineno"> 312</span>  <span class="keyword">const</span> uint8x16_t min_u8 = vdupq_n_u8(static_cast<uint8_t>(output_stage.gemmlowp_min_bound));</div><div class="line"><a name="l00313"></a><span class="lineno"> 313</span>  <span class="keyword">const</span> uint8x16_t max_u8 = vdupq_n_u8(static_cast<uint8_t>(output_stage.gemmlowp_max_bound));</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="keyword">const</span> <span class="keywordtype">int</span> window_step_x = 16;</div><div class="line"><a name="l00316"></a><span class="lineno"> 316</span>  <span class="keyword">const</span> <span class="keyword">auto</span> window_start_x = static_cast<int>(window.x().start());</div><div class="line"><a name="l00317"></a><span class="lineno"> 317</span>  <span class="keyword">const</span> <span class="keyword">auto</span> window_end_x = static_cast<int>(window.x().end());</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>  Window win(window);</div><div class="line"><a name="l00320"></a><span class="lineno"> 320</span>  win.set(<a class="code" href="classarm__compute_1_1_window.xhtml#aa96e81276ee4f87ab386cd05a5539a7d">Window::DimX</a>, Window::Dimension(0, 1, 1));</div><div class="line"><a name="l00321"></a><span class="lineno"> 321</span> </div><div class="line"><a name="l00322"></a><span class="lineno"> 322</span>  Window collapsed_window = win.collapse_if_possible(win, <a class="code" href="classarm__compute_1_1_window.xhtml#a893d17b56b9abc4423ce26e9a24ac5dc">Window::DimZ</a>);</div><div class="line"><a name="l00323"></a><span class="lineno"> 323</span> </div><div class="line"><a name="l00324"></a><span class="lineno"> 324</span>  Iterator mm_result_it(mm_result, win);</div><div class="line"><a name="l00325"></a><span class="lineno"> 325</span>  Iterator out_it(output, win);</div><div class="line"><a name="l00326"></a><span class="lineno"> 326</span> </div><div class="line"><a name="l00327"></a><span class="lineno"> 327</span>  <span class="keywordflow">if</span>((a_offset != 0) && (b_offset != 0))</div><div class="line"><a name="l00328"></a><span class="lineno"> 328</span>  {</div><div class="line"><a name="l00329"></a><span class="lineno"> 329</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(vector_sum_col);</div><div class="line"><a name="l00330"></a><span class="lineno"> 330</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(vector_sum_row);</div><div class="line"><a name="l00331"></a><span class="lineno"> 331</span> </div><div class="line"><a name="l00332"></a><span class="lineno"> 332</span>  Iterator vector_sum_col_it = get_vector_sum_col_it(collapsed_window, vector_sum_col);</div><div class="line"><a name="l00333"></a><span class="lineno"> 333</span>  Iterator vector_sum_row_it = get_vector_sum_row_it(collapsed_window, vector_sum_row);</div><div class="line"><a name="l00334"></a><span class="lineno"> 334</span> </div><div class="line"><a name="l00335"></a><span class="lineno"> 335</span>  <span class="keyword">const</span> <span class="keywordtype">size_t</span> sum_row_stride_y = vector_sum_row->info()->strides_in_bytes().y();</div><div class="line"><a name="l00336"></a><span class="lineno"> 336</span> </div><div class="line"><a name="l00337"></a><span class="lineno"> 337</span>  <span class="comment">// Offset in case vector_sum_col is batched</span></div><div class="line"><a name="l00338"></a><span class="lineno"> 338</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> vector_sum_col_batch_offset = slide_vector_sum_col ? vector_sum_col->info()->strides_in_bytes().z() : 0;</div><div class="line"><a name="l00339"></a><span class="lineno"> 339</span> </div><div class="line"><a name="l00340"></a><span class="lineno"> 340</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00341"></a><span class="lineno"> 341</span>  {</div><div class="line"><a name="l00342"></a><span class="lineno"> 342</span>  Iterator bias_it = get_bias_it(collapsed_window, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>);</div><div class="line"><a name="l00343"></a><span class="lineno"> 343</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00344"></a><span class="lineno"> 344</span>  {</div><div class="line"><a name="l00345"></a><span class="lineno"> 345</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00346"></a><span class="lineno"> 346</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_col_ptr = reinterpret_cast<const int32_t *>(vector_sum_col_it.ptr() + batch_id * vector_sum_col_batch_offset);</div><div class="line"><a name="l00347"></a><span class="lineno"> 347</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_row_ptr = reinterpret_cast<const int32_t *>(vector_sum_row_it.ptr() + batch_id * sum_row_stride_y)</div><div class="line"><a name="l00348"></a><span class="lineno"> 348</span>  + <span class="keywordtype">id</span>.y() + (<span class="keywordtype">id</span>.z() % depth_input) * height_input;</div><div class="line"><a name="l00349"></a><span class="lineno"> 349</span>  run_offset_contribution_output_stage_window<true, true, true, is_bounded_relu, is_fixed_point>(vector_sum_col_ptr, vector_sum_row_ptr, reinterpret_cast<const int32_t *>(bias_it.ptr()), mm_result_it,</div><div class="line"><a name="l00350"></a><span class="lineno"> 350</span>  out_it,</div><div class="line"><a name="l00351"></a><span class="lineno"> 351</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00352"></a><span class="lineno"> 352</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00353"></a><span class="lineno"> 353</span>  },</div><div class="line"><a name="l00354"></a><span class="lineno"> 354</span>  vector_sum_col_it, vector_sum_row_it, bias_it, mm_result_it, out_it);</div><div class="line"><a name="l00355"></a><span class="lineno"> 355</span>  }</div><div class="line"><a name="l00356"></a><span class="lineno"> 356</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00357"></a><span class="lineno"> 357</span>  {</div><div class="line"><a name="l00358"></a><span class="lineno"> 358</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00359"></a><span class="lineno"> 359</span>  {</div><div class="line"><a name="l00360"></a><span class="lineno"> 360</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00361"></a><span class="lineno"> 361</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_col_ptr = reinterpret_cast<const int32_t *>(vector_sum_col_it.ptr() + batch_id * vector_sum_col_batch_offset);</div><div class="line"><a name="l00362"></a><span class="lineno"> 362</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_row_ptr = reinterpret_cast<const int32_t *>(vector_sum_row_it.ptr() + batch_id * sum_row_stride_y)</div><div class="line"><a name="l00363"></a><span class="lineno"> 363</span>  + <span class="keywordtype">id</span>.y() + (<span class="keywordtype">id</span>.z() % depth_input) * height_input;</div><div class="line"><a name="l00364"></a><span class="lineno"> 364</span>  run_offset_contribution_output_stage_window<true, true, false, is_bounded_relu, is_fixed_point>(vector_sum_col_ptr, vector_sum_row_ptr, <span class="keyword">nullptr</span>, mm_result_it, out_it,</div><div class="line"><a name="l00365"></a><span class="lineno"> 365</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00366"></a><span class="lineno"> 366</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00367"></a><span class="lineno"> 367</span>  },</div><div class="line"><a name="l00368"></a><span class="lineno"> 368</span>  vector_sum_col_it, vector_sum_row_it, mm_result_it, out_it);</div><div class="line"><a name="l00369"></a><span class="lineno"> 369</span>  }</div><div class="line"><a name="l00370"></a><span class="lineno"> 370</span>  }</div><div class="line"><a name="l00371"></a><span class="lineno"> 371</span>  <span class="keywordflow">else</span> <span class="keywordflow">if</span>((a_offset == 0) && (b_offset != 0))</div><div class="line"><a name="l00372"></a><span class="lineno"> 372</span>  {</div><div class="line"><a name="l00373"></a><span class="lineno"> 373</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(vector_sum_row);</div><div class="line"><a name="l00374"></a><span class="lineno"> 374</span> </div><div class="line"><a name="l00375"></a><span class="lineno"> 375</span>  Iterator vector_sum_row_it = get_vector_sum_row_it(collapsed_window, vector_sum_row);</div><div class="line"><a name="l00376"></a><span class="lineno"> 376</span> </div><div class="line"><a name="l00377"></a><span class="lineno"> 377</span>  <span class="keyword">const</span> <span class="keywordtype">size_t</span> sum_row_stride_y = vector_sum_row->info()->strides_in_bytes().y();</div><div class="line"><a name="l00378"></a><span class="lineno"> 378</span> </div><div class="line"><a name="l00379"></a><span class="lineno"> 379</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00380"></a><span class="lineno"> 380</span>  {</div><div class="line"><a name="l00381"></a><span class="lineno"> 381</span>  Iterator bias_it = get_bias_it(collapsed_window, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>);</div><div class="line"><a name="l00382"></a><span class="lineno"> 382</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00383"></a><span class="lineno"> 383</span>  {</div><div class="line"><a name="l00384"></a><span class="lineno"> 384</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00385"></a><span class="lineno"> 385</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_row_ptr = reinterpret_cast<const int32_t *>(vector_sum_row_it.ptr() + batch_id * sum_row_stride_y)</div><div class="line"><a name="l00386"></a><span class="lineno"> 386</span>  + <span class="keywordtype">id</span>.y() + (<span class="keywordtype">id</span>.z() % depth_input) * height_input;</div><div class="line"><a name="l00387"></a><span class="lineno"> 387</span>  run_offset_contribution_output_stage_window<false, true, true, is_bounded_relu, is_fixed_point>(<span class="keyword">nullptr</span>, vector_sum_row_ptr, reinterpret_cast<const int32_t *>(bias_it.ptr()), mm_result_it, out_it,</div><div class="line"><a name="l00388"></a><span class="lineno"> 388</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00389"></a><span class="lineno"> 389</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00390"></a><span class="lineno"> 390</span>  },</div><div class="line"><a name="l00391"></a><span class="lineno"> 391</span>  vector_sum_row_it, bias_it, mm_result_it, out_it);</div><div class="line"><a name="l00392"></a><span class="lineno"> 392</span>  }</div><div class="line"><a name="l00393"></a><span class="lineno"> 393</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00394"></a><span class="lineno"> 394</span>  {</div><div class="line"><a name="l00395"></a><span class="lineno"> 395</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00396"></a><span class="lineno"> 396</span>  {</div><div class="line"><a name="l00397"></a><span class="lineno"> 397</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00398"></a><span class="lineno"> 398</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_row_ptr = reinterpret_cast<const int32_t *>(vector_sum_row_it.ptr() + batch_id * sum_row_stride_y)</div><div class="line"><a name="l00399"></a><span class="lineno"> 399</span>  + <span class="keywordtype">id</span>.y() + (<span class="keywordtype">id</span>.z() % depth_input) * height_input;</div><div class="line"><a name="l00400"></a><span class="lineno"> 400</span>  run_offset_contribution_output_stage_window<false, true, false, is_bounded_relu, is_fixed_point>(<span class="keyword">nullptr</span>, vector_sum_row_ptr, <span class="keyword">nullptr</span>, mm_result_it, out_it,</div><div class="line"><a name="l00401"></a><span class="lineno"> 401</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00402"></a><span class="lineno"> 402</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00403"></a><span class="lineno"> 403</span>  },</div><div class="line"><a name="l00404"></a><span class="lineno"> 404</span>  vector_sum_row_it, mm_result_it, out_it);</div><div class="line"><a name="l00405"></a><span class="lineno"> 405</span>  }</div><div class="line"><a name="l00406"></a><span class="lineno"> 406</span>  }</div><div class="line"><a name="l00407"></a><span class="lineno"> 407</span>  <span class="keywordflow">else</span> <span class="keywordflow">if</span>((a_offset != 0) && (b_offset == 0))</div><div class="line"><a name="l00408"></a><span class="lineno"> 408</span>  {</div><div class="line"><a name="l00409"></a><span class="lineno"> 409</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(vector_sum_col);</div><div class="line"><a name="l00410"></a><span class="lineno"> 410</span> </div><div class="line"><a name="l00411"></a><span class="lineno"> 411</span>  Iterator vector_sum_col_it = get_vector_sum_col_it(collapsed_window, vector_sum_col);</div><div class="line"><a name="l00412"></a><span class="lineno"> 412</span> </div><div class="line"><a name="l00413"></a><span class="lineno"> 413</span>  <span class="comment">// Offset in case vector_sum_col is batched</span></div><div class="line"><a name="l00414"></a><span class="lineno"> 414</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> vector_sum_col_batch_offset = slide_vector_sum_col ? vector_sum_col->info()->strides_in_bytes().z() : 0;</div><div class="line"><a name="l00415"></a><span class="lineno"> 415</span> </div><div class="line"><a name="l00416"></a><span class="lineno"> 416</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00417"></a><span class="lineno"> 417</span>  {</div><div class="line"><a name="l00418"></a><span class="lineno"> 418</span>  Iterator bias_it = get_bias_it(collapsed_window, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>);</div><div class="line"><a name="l00419"></a><span class="lineno"> 419</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00420"></a><span class="lineno"> 420</span>  {</div><div class="line"><a name="l00421"></a><span class="lineno"> 421</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00422"></a><span class="lineno"> 422</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_col_ptr = reinterpret_cast<const int32_t *>(vector_sum_col_it.ptr() + batch_id * vector_sum_col_batch_offset);</div><div class="line"><a name="l00423"></a><span class="lineno"> 423</span>  run_offset_contribution_output_stage_window<true, false, true, is_bounded_relu, is_fixed_point>(vector_sum_col_ptr, <span class="keyword">nullptr</span>, reinterpret_cast<const int32_t *>(bias_it.ptr()), mm_result_it, out_it,</div><div class="line"><a name="l00424"></a><span class="lineno"> 424</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00425"></a><span class="lineno"> 425</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00426"></a><span class="lineno"> 426</span>  },</div><div class="line"><a name="l00427"></a><span class="lineno"> 427</span>  vector_sum_col_it, bias_it, mm_result_it, out_it);</div><div class="line"><a name="l00428"></a><span class="lineno"> 428</span>  }</div><div class="line"><a name="l00429"></a><span class="lineno"> 429</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00430"></a><span class="lineno"> 430</span>  {</div><div class="line"><a name="l00431"></a><span class="lineno"> 431</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates & <span class="keywordtype">id</span>)</div><div class="line"><a name="l00432"></a><span class="lineno"> 432</span>  {</div><div class="line"><a name="l00433"></a><span class="lineno"> 433</span>  <span class="keyword">const</span> <span class="keywordtype">int</span> batch_id = <span class="keywordtype">id</span>.z() / depth_input;</div><div class="line"><a name="l00434"></a><span class="lineno"> 434</span>  <span class="keyword">const</span> <span class="keyword">auto</span> vector_sum_col_ptr = reinterpret_cast<const int32_t *>(vector_sum_col_it.ptr() + batch_id * vector_sum_col_batch_offset);</div><div class="line"><a name="l00435"></a><span class="lineno"> 435</span>  run_offset_contribution_output_stage_window<true, false, false, is_bounded_relu, is_fixed_point>(vector_sum_col_ptr, <span class="keyword">nullptr</span>, <span class="keyword">nullptr</span>, mm_result_it, out_it,</div><div class="line"><a name="l00436"></a><span class="lineno"> 436</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00437"></a><span class="lineno"> 437</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00438"></a><span class="lineno"> 438</span>  },</div><div class="line"><a name="l00439"></a><span class="lineno"> 439</span>  vector_sum_col_it, mm_result_it, out_it);</div><div class="line"><a name="l00440"></a><span class="lineno"> 440</span>  }</div><div class="line"><a name="l00441"></a><span class="lineno"> 441</span>  }</div><div class="line"><a name="l00442"></a><span class="lineno"> 442</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00443"></a><span class="lineno"> 443</span>  {</div><div class="line"><a name="l00444"></a><span class="lineno"> 444</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00445"></a><span class="lineno"> 445</span>  {</div><div class="line"><a name="l00446"></a><span class="lineno"> 446</span>  Iterator bias_it = get_bias_it(collapsed_window, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>);</div><div class="line"><a name="l00447"></a><span class="lineno"> 447</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates &)</div><div class="line"><a name="l00448"></a><span class="lineno"> 448</span>  {</div><div class="line"><a name="l00449"></a><span class="lineno"> 449</span>  run_offset_contribution_output_stage_window<false, false, true, is_bounded_relu, is_fixed_point>(<span class="keyword">nullptr</span>, <span class="keyword">nullptr</span>, reinterpret_cast<const int32_t *>(bias_it.ptr()), mm_result_it, out_it,</div><div class="line"><a name="l00450"></a><span class="lineno"> 450</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00451"></a><span class="lineno"> 451</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00452"></a><span class="lineno"> 452</span>  },</div><div class="line"><a name="l00453"></a><span class="lineno"> 453</span>  bias_it, mm_result_it, out_it);</div><div class="line"><a name="l00454"></a><span class="lineno"> 454</span>  }</div><div class="line"><a name="l00455"></a><span class="lineno"> 455</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00456"></a><span class="lineno"> 456</span>  {</div><div class="line"><a name="l00457"></a><span class="lineno"> 457</span>  <a class="code" href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">execute_window_loop</a>(collapsed_window, [&](<span class="keyword">const</span> Coordinates &)</div><div class="line"><a name="l00458"></a><span class="lineno"> 458</span>  {</div><div class="line"><a name="l00459"></a><span class="lineno"> 459</span>  run_offset_contribution_output_stage_window<false, false, false, is_bounded_relu, is_fixed_point>(<span class="keyword">nullptr</span>, <span class="keyword">nullptr</span>, <span class="keyword">nullptr</span>, mm_result_it, out_it,</div><div class="line"><a name="l00460"></a><span class="lineno"> 460</span>  result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset,</div><div class="line"><a name="l00461"></a><span class="lineno"> 461</span>  output_stage, window_step_x, window_start_x, window_end_x);</div><div class="line"><a name="l00462"></a><span class="lineno"> 462</span>  },</div><div class="line"><a name="l00463"></a><span class="lineno"> 463</span>  mm_result_it, out_it);</div><div class="line"><a name="l00464"></a><span class="lineno"> 464</span>  }</div><div class="line"><a name="l00465"></a><span class="lineno"> 465</span>  <span class="keywordflow">return</span>;</div><div class="line"><a name="l00466"></a><span class="lineno"> 466</span>  }</div><div class="line"><a name="l00467"></a><span class="lineno"> 467</span> }</div><div class="line"><a name="l00468"></a><span class="lineno"> 468</span> </div><div class="line"><a name="l00469"></a><span class="lineno"> 469</span> Status validate_arguments(<span class="keyword">const</span> ITensorInfo *mm_result, <span class="keyword">const</span> ITensorInfo *vector_sum_col, <span class="keyword">const</span> ITensorInfo *vector_sum_row, <span class="keyword">const</span> ITensorInfo *<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, <span class="keyword">const</span> ITensorInfo *output,</div><div class="line"><a name="l00470"></a><span class="lineno"> 470</span>  int32_t a_offset, int32_t b_offset, GEMMLowpOutputStageInfo output_stage)</div><div class="line"><a name="l00471"></a><span class="lineno"> 471</span> {</div><div class="line"><a name="l00472"></a><span class="lineno"> 472</span>  <a class="code" href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a>(mm_result, 1, <a class="code" href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c">DataType::S32</a>);</div><div class="line"><a name="l00473"></a><span class="lineno"> 473</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(output_stage.gemmlowp_max_bound > 255);</div><div class="line"><a name="l00474"></a><span class="lineno"> 474</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(output_stage.gemmlowp_min_bound < 0 || output_stage.gemmlowp_min_bound > output_stage.gemmlowp_max_bound);</div><div class="line"><a name="l00475"></a><span class="lineno"> 475</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(output_stage.type != <a class="code" href="namespacearm__compute.xhtml#a5558e2cc22f7f4771653d992c8ad8864a079e2ddc95b344b5cb0188bed9a80d8b">GEMMLowpOutputStageType::QUANTIZE_DOWN</a> && output_stage.type != <a class="code" href="namespacearm__compute.xhtml#a5558e2cc22f7f4771653d992c8ad8864ab300cae200f67712c1eb9234e28158ca">GEMMLowpOutputStageType::QUANTIZE_DOWN_FIXEDPOINT</a>);</div><div class="line"><a name="l00476"></a><span class="lineno"> 476</span> </div><div class="line"><a name="l00477"></a><span class="lineno"> 477</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span>)</div><div class="line"><a name="l00478"></a><span class="lineno"> 478</span>  {</div><div class="line"><a name="l00479"></a><span class="lineno"> 479</span>  <a class="code" href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, 1, <a class="code" href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c">DataType::S32</a>);</div><div class="line"><a name="l00480"></a><span class="lineno"> 480</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>->num_dimensions() > 1);</div><div class="line"><a name="l00481"></a><span class="lineno"> 481</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(mm_result->dimension(0) != <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>->dimension(0));</div><div class="line"><a name="l00482"></a><span class="lineno"> 482</span>  }</div><div class="line"><a name="l00483"></a><span class="lineno"> 483</span> </div><div class="line"><a name="l00484"></a><span class="lineno"> 484</span>  <span class="comment">// If a_offset == 0, vector_sum_col can be a nullptr</span></div><div class="line"><a name="l00485"></a><span class="lineno"> 485</span>  <span class="keywordflow">if</span>(a_offset != 0)</div><div class="line"><a name="l00486"></a><span class="lineno"> 486</span>  {</div><div class="line"><a name="l00487"></a><span class="lineno"> 487</span>  <a class="code" href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a>(vector_sum_col, 1, <a class="code" href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c">DataType::S32</a>);</div><div class="line"><a name="l00488"></a><span class="lineno"> 488</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(vector_sum_col->dimension(0) != mm_result->dimension(0));</div><div class="line"><a name="l00489"></a><span class="lineno"> 489</span>  }</div><div class="line"><a name="l00490"></a><span class="lineno"> 490</span> </div><div class="line"><a name="l00491"></a><span class="lineno"> 491</span>  <span class="comment">// If b_offset == 0, vector_sum_row can be a nullptr</span></div><div class="line"><a name="l00492"></a><span class="lineno"> 492</span>  <span class="keywordflow">if</span>(b_offset != 0)</div><div class="line"><a name="l00493"></a><span class="lineno"> 493</span>  {</div><div class="line"><a name="l00494"></a><span class="lineno"> 494</span>  <a class="code" href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a>(vector_sum_row, 1, <a class="code" href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c">DataType::S32</a>);</div><div class="line"><a name="l00495"></a><span class="lineno"> 495</span> </div><div class="line"><a name="l00496"></a><span class="lineno"> 496</span>  <span class="comment">// Check if input is a 3D reinterpretation</span></div><div class="line"><a name="l00497"></a><span class="lineno"> 497</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> reinterpret_as_3d = mm_result->num_dimensions() > 1 && mm_result->tensor_shape().y() != vector_sum_row->tensor_shape().x();</div><div class="line"><a name="l00498"></a><span class="lineno"> 498</span> </div><div class="line"><a name="l00499"></a><span class="lineno"> 499</span>  <span class="comment">// Validate input</span></div><div class="line"><a name="l00500"></a><span class="lineno"> 500</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(reinterpret_as_3d && vector_sum_row->dimension(0) != (mm_result->dimension(1) * mm_result->dimension(2)));</div><div class="line"><a name="l00501"></a><span class="lineno"> 501</span>  <a class="code" href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a>(!reinterpret_as_3d && vector_sum_row->dimension(0) != mm_result->dimension(1));</div><div class="line"><a name="l00502"></a><span class="lineno"> 502</span> </div><div class="line"><a name="l00503"></a><span class="lineno"> 503</span>  TensorShape <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a7fc93f37dac131a1a40b7921f9df3a9a">output_shape</a> = output->tensor_shape();</div><div class="line"><a name="l00504"></a><span class="lineno"> 504</span>  <span class="keywordflow">if</span>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a7fc93f37dac131a1a40b7921f9df3a9a">output_shape</a>.num_dimensions() > 1)</div><div class="line"><a name="l00505"></a><span class="lineno"> 505</span>  {</div><div class="line"><a name="l00506"></a><span class="lineno"> 506</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> output_batch_idx = reinterpret_as_3d ? 3 : 2;</div><div class="line"><a name="l00507"></a><span class="lineno"> 507</span> </div><div class="line"><a name="l00508"></a><span class="lineno"> 508</span>  TensorShape vector_sum_row_shape = vector_sum_row->tensor_shape();</div><div class="line"><a name="l00509"></a><span class="lineno"> 509</span>  vector_sum_row_shape.collapse_from(1);</div><div class="line"><a name="l00510"></a><span class="lineno"> 510</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a7fc93f37dac131a1a40b7921f9df3a9a">output_shape</a>.collapse_from(output_batch_idx);</div><div class="line"><a name="l00511"></a><span class="lineno"> 511</span> </div><div class="line"><a name="l00512"></a><span class="lineno"> 512</span>  <a class="code" href="_error_8h.xhtml#a86084036bd3851575ef871ad5bf079a7">ARM_COMPUTE_RETURN_ERROR_ON_MSG</a>(vector_sum_row_shape[1] != <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a7fc93f37dac131a1a40b7921f9df3a9a">output_shape</a>[output_batch_idx],</div><div class="line"><a name="l00513"></a><span class="lineno"> 513</span>  <span class="stringliteral">"mm_result tensor must have the same number of batches of output tensor"</span>);</div><div class="line"><a name="l00514"></a><span class="lineno"> 514</span> </div><div class="line"><a name="l00515"></a><span class="lineno"> 515</span>  <span class="keywordflow">if</span>(a_offset != 0)</div><div class="line"><a name="l00516"></a><span class="lineno"> 516</span>  {</div><div class="line"><a name="l00517"></a><span class="lineno"> 517</span>  TensorShape vector_sum_col_shape = vector_sum_col->tensor_shape();</div><div class="line"><a name="l00518"></a><span class="lineno"> 518</span>  vector_sum_col_shape.collapse_from(1);</div><div class="line"><a name="l00519"></a><span class="lineno"> 519</span> </div><div class="line"><a name="l00520"></a><span class="lineno"> 520</span>  <a class="code" href="_error_8h.xhtml#a86084036bd3851575ef871ad5bf079a7">ARM_COMPUTE_RETURN_ERROR_ON_MSG</a>(vector_sum_col_shape[1] != 1 && vector_sum_col_shape[1] != vector_sum_row_shape[1],</div><div class="line"><a name="l00521"></a><span class="lineno"> 521</span>  <span class="stringliteral">"vector_sum_col tensor must have the same number of batches of vector_sum_row_shape or the number of batches must be set to 1"</span>);</div><div class="line"><a name="l00522"></a><span class="lineno"> 522</span>  }</div><div class="line"><a name="l00523"></a><span class="lineno"> 523</span>  }</div><div class="line"><a name="l00524"></a><span class="lineno"> 524</span>  }</div><div class="line"><a name="l00525"></a><span class="lineno"> 525</span> </div><div class="line"><a name="l00526"></a><span class="lineno"> 526</span>  <span class="keywordflow">if</span>(output->total_size() != 0)</div><div class="line"><a name="l00527"></a><span class="lineno"> 527</span>  {</div><div class="line"><a name="l00528"></a><span class="lineno"> 528</span>  <a class="code" href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a>(output, 1, <a class="code" href="namespacearm__compute.xhtml#ad8ed01ff3ff33333d8e19db4d2818bb6af14462d71aa842202c3e4b272c7ec924">DataType::QASYMM8</a>);</div><div class="line"><a name="l00529"></a><span class="lineno"> 529</span>  <a class="code" href="_validate_8h.xhtml#a27e4638546c88b8916f967e6e54480a9">ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES</a>(mm_result, output);</div><div class="line"><a name="l00530"></a><span class="lineno"> 530</span>  }</div><div class="line"><a name="l00531"></a><span class="lineno"> 531</span> </div><div class="line"><a name="l00532"></a><span class="lineno"> 532</span>  <span class="keywordflow">return</span> Status{};</div><div class="line"><a name="l00533"></a><span class="lineno"> 533</span> }</div><div class="line"><a name="l00534"></a><span class="lineno"> 534</span> </div><div class="line"><a name="l00535"></a><span class="lineno"> 535</span> std::pair<Status, Window> <a class="code" href="namespacearm__compute.xhtml#ad6630777dc2d315531f1e0b02491051f">validate_and_configure_window</a>(ITensorInfo *mm_result, ITensorInfo *output)</div><div class="line"><a name="l00536"></a><span class="lineno"> 536</span> {</div><div class="line"><a name="l00537"></a><span class="lineno"> 537</span>  <span class="comment">// Output auto inizialitation if not yet initialized</span></div><div class="line"><a name="l00538"></a><span class="lineno"> 538</span>  <a class="code" href="namespacearm__compute.xhtml#a47be6fa38308d0003c25b60b7dbc45ce">auto_init_if_empty</a>(*output, mm_result->clone()->set_data_type(<a class="code" href="namespacearm__compute.xhtml#ad8ed01ff3ff33333d8e19db4d2818bb6af14462d71aa842202c3e4b272c7ec924">DataType::QASYMM8</a>));</div><div class="line"><a name="l00539"></a><span class="lineno"> 539</span> </div><div class="line"><a name="l00540"></a><span class="lineno"> 540</span>  <span class="comment">// Configure kernel window</span></div><div class="line"><a name="l00541"></a><span class="lineno"> 541</span>  Window win = <a class="code" href="namespacearm__compute.xhtml#ab7980fa5ee693e3282a76da047a1c3b5">calculate_max_window</a>(*mm_result, Steps());</div><div class="line"><a name="l00542"></a><span class="lineno"> 542</span> </div><div class="line"><a name="l00543"></a><span class="lineno"> 543</span>  <span class="comment">// Note: This kernel performs 16 elements per iteration.</span></div><div class="line"><a name="l00544"></a><span class="lineno"> 544</span>  <span class="comment">// However, since we use a left-over for loop, we cannot have any read or write out of memory</span></div><div class="line"><a name="l00545"></a><span class="lineno"> 545</span>  <span class="comment">// For this reason num_elems_processed_per_iteration is 1 and so update_window_and_padding() can be skipped</span></div><div class="line"><a name="l00546"></a><span class="lineno"> 546</span>  Coordinates coord;</div><div class="line"><a name="l00547"></a><span class="lineno"> 547</span>  coord.set_num_dimensions(output->num_dimensions());</div><div class="line"><a name="l00548"></a><span class="lineno"> 548</span>  output->set_valid_region(ValidRegion(coord, output->tensor_shape()));</div><div class="line"><a name="l00549"></a><span class="lineno"> 549</span> </div><div class="line"><a name="l00550"></a><span class="lineno"> 550</span>  <span class="keywordflow">return</span> std::make_pair(Status{}, win);</div><div class="line"><a name="l00551"></a><span class="lineno"> 551</span> }</div><div class="line"><a name="l00552"></a><span class="lineno"> 552</span> </div><div class="line"><a name="l00553"></a><span class="lineno"> 553</span> <a class="code" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#adb79bb3ad15444f00b55b35f1d6e16b7">NEGEMMLowpOffsetContributionOutputStageKernel::NEGEMMLowpOffsetContributionOutputStageFunction</a></div><div class="line"><a name="l00554"></a><span class="lineno"> 554</span> get_configured_function(<span class="keyword">const</span> ITensor *mm_result, <span class="keyword">const</span> ITensor *vector_sum_row, GEMMLowpOutputStageInfo output_stage)</div><div class="line"><a name="l00555"></a><span class="lineno"> 555</span> {</div><div class="line"><a name="l00556"></a><span class="lineno"> 556</span>  <span class="keyword">static</span> std::map<uint8_t, NEGEMMLowpOffsetContributionOutputStageKernel::NEGEMMLowpOffsetContributionOutputStageFunction> map_function =</div><div class="line"><a name="l00557"></a><span class="lineno"> 557</span>  {</div><div class="line"><a name="l00558"></a><span class="lineno"> 558</span>  { 0, &run_offset_contribution_output_stage<false, false, false> },</div><div class="line"><a name="l00559"></a><span class="lineno"> 559</span>  { 1, &run_offset_contribution_output_stage<true, false, false> },</div><div class="line"><a name="l00560"></a><span class="lineno"> 560</span>  { 2, &run_offset_contribution_output_stage<false, true, false> },</div><div class="line"><a name="l00561"></a><span class="lineno"> 561</span>  { 3, &run_offset_contribution_output_stage<true, true, false> },</div><div class="line"><a name="l00562"></a><span class="lineno"> 562</span>  { 4, &run_offset_contribution_output_stage<false, false, true> },</div><div class="line"><a name="l00563"></a><span class="lineno"> 563</span>  { 5, &run_offset_contribution_output_stage<true, false, true> },</div><div class="line"><a name="l00564"></a><span class="lineno"> 564</span>  { 6, &run_offset_contribution_output_stage<false, true, true> },</div><div class="line"><a name="l00565"></a><span class="lineno"> 565</span>  { 7, &run_offset_contribution_output_stage<true, true, true> }</div><div class="line"><a name="l00566"></a><span class="lineno"> 566</span>  };</div><div class="line"><a name="l00567"></a><span class="lineno"> 567</span> </div><div class="line"><a name="l00568"></a><span class="lineno"> 568</span>  <span class="comment">// Check if input is a 3D reinterpretation</span></div><div class="line"><a name="l00569"></a><span class="lineno"> 569</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> reinterpret_as_3d = vector_sum_row != <span class="keyword">nullptr</span></div><div class="line"><a name="l00570"></a><span class="lineno"> 570</span>  && mm_result->info()->num_dimensions() > 1</div><div class="line"><a name="l00571"></a><span class="lineno"> 571</span>  && mm_result->info()->tensor_shape().y() != vector_sum_row->info()->tensor_shape().x();</div><div class="line"><a name="l00572"></a><span class="lineno"> 572</span> </div><div class="line"><a name="l00573"></a><span class="lineno"> 573</span>  <span class="comment">// Check if we need to clamp the result using min and max</span></div><div class="line"><a name="l00574"></a><span class="lineno"> 574</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> is_bounded_relu = ((output_stage.gemmlowp_min_bound != output_stage.gemmlowp_max_bound)</div><div class="line"><a name="l00575"></a><span class="lineno"> 575</span>  && !(output_stage.gemmlowp_min_bound == 0 && output_stage.gemmlowp_max_bound == 255));</div><div class="line"><a name="l00576"></a><span class="lineno"> 576</span> </div><div class="line"><a name="l00577"></a><span class="lineno"> 577</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> is_fixed_point = output_stage.type != <a class="code" href="namespacearm__compute.xhtml#a5558e2cc22f7f4771653d992c8ad8864a079e2ddc95b344b5cb0188bed9a80d8b">GEMMLowpOutputStageType::QUANTIZE_DOWN</a>;</div><div class="line"><a name="l00578"></a><span class="lineno"> 578</span> </div><div class="line"><a name="l00579"></a><span class="lineno"> 579</span>  <span class="comment">// key acts as a bitset, setting the first bit on reinterpret_as_3d,</span></div><div class="line"><a name="l00580"></a><span class="lineno"> 580</span>  <span class="comment">// the second on is_bounded_relu, and the third on is_fixed_point.</span></div><div class="line"><a name="l00581"></a><span class="lineno"> 581</span>  uint8_t key = (reinterpret_as_3d ? 1UL : 0UL) | ((is_bounded_relu ? 1UL : 0UL) << 1) | ((is_fixed_point ? 1UL : 0UL) << 2);</div><div class="line"><a name="l00582"></a><span class="lineno"> 582</span>  <span class="keywordflow">return</span> map_function.find(key)->second;</div><div class="line"><a name="l00583"></a><span class="lineno"> 583</span> }</div><div class="line"><a name="l00584"></a><span class="lineno"> 584</span> } <span class="comment">// namespace</span></div><div class="line"><a name="l00585"></a><span class="lineno"> 585</span> </div><div class="line"><a name="l00586"></a><span class="lineno"><a class="line" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a2851f631e9660a4dd9644cb749282723"> 586</a></span> <a class="code" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a2851f631e9660a4dd9644cb749282723">NEGEMMLowpOffsetContributionOutputStageKernel::NEGEMMLowpOffsetContributionOutputStageKernel</a>()</div><div class="line"><a name="l00587"></a><span class="lineno"> 587</span>  : _function(nullptr), _vector_sum_col(nullptr), _vector_sum_row(nullptr), _bias(nullptr), _mm_result(nullptr), _output(nullptr), _a_offset(0), _b_offset(0), _k_offset(0), _slide_vector_sum_col(true),</div><div class="line"><a name="l00588"></a><span class="lineno"> 588</span>  _output_stage(<a class="code" href="structarm__compute_1_1_g_e_m_m_lowp_output_stage_info.xhtml">GEMMLowpOutputStageInfo</a>())</div><div class="line"><a name="l00589"></a><span class="lineno"> 589</span> </div><div class="line"><a name="l00590"></a><span class="lineno"> 590</span> {</div><div class="line"><a name="l00591"></a><span class="lineno"> 591</span> }</div><div class="line"><a name="l00592"></a><span class="lineno"> 592</span> </div><div class="line"><a name="l00593"></a><span class="lineno"><a class="line" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a97ebe5c0444a53d58d9b9f079ebe2d0f"> 593</a></span> <span class="keywordtype">void</span> <a class="code" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a97ebe5c0444a53d58d9b9f079ebe2d0f">NEGEMMLowpOffsetContributionOutputStageKernel::configure</a>(<span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor.xhtml">ITensor</a> *mm_result, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor.xhtml">ITensor</a> *vector_sum_col,</div><div class="line"><a name="l00594"></a><span class="lineno"> 594</span>  <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor.xhtml">ITensor</a> *vector_sum_row, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor.xhtml">ITensor</a> *<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, <a class="code" href="classarm__compute_1_1_i_tensor.xhtml">ITensor</a> *output, int32_t k,</div><div class="line"><a name="l00595"></a><span class="lineno"> 595</span>  int32_t a_offset, int32_t b_offset, <a class="code" href="structarm__compute_1_1_g_e_m_m_lowp_output_stage_info.xhtml">GEMMLowpOutputStageInfo</a> output_stage)</div><div class="line"><a name="l00596"></a><span class="lineno"> 596</span> {</div><div class="line"><a name="l00597"></a><span class="lineno"> 597</span>  <span class="comment">// Perform validate step</span></div><div class="line"><a name="l00598"></a><span class="lineno"> 598</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(mm_result, output);</div><div class="line"><a name="l00599"></a><span class="lineno"> 599</span> </div><div class="line"><a name="l00600"></a><span class="lineno"> 600</span>  <a class="code" href="_error_8h.xhtml#a938dcd406ce611ef5345ad2531cdb948">ARM_COMPUTE_ERROR_THROW_ON</a>(validate_arguments(mm_result-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>(),</div><div class="line"><a name="l00601"></a><span class="lineno"> 601</span>  vector_sum_col != <span class="keyword">nullptr</span> ? vector_sum_col-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>() : <span class="keyword">nullptr</span>, <span class="comment">// NOLINT</span></div><div class="line"><a name="l00602"></a><span class="lineno"> 602</span>  vector_sum_row != <span class="keyword">nullptr</span> ? vector_sum_row-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>() : <span class="keyword">nullptr</span>, <span class="comment">// NOLINT</span></div><div class="line"><a name="l00603"></a><span class="lineno"> 603</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a> != <span class="keyword">nullptr</span> ? <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>-><a class="code" href="classarm__compute_1_1_c_l_tensor.xhtml#ad45f0c01a0713dfb6bd7232c7f396fc4">info</a>() : <span class="keyword">nullptr</span>, <span class="comment">// NOLINT</span></div><div class="line"><a name="l00604"></a><span class="lineno"> 604</span>  output-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>(), a_offset, b_offset, output_stage)); <span class="comment">// NOLINT</span></div><div class="line"><a name="l00605"></a><span class="lineno"> 605</span> </div><div class="line"><a name="l00606"></a><span class="lineno"> 606</span>  _vector_sum_col = vector_sum_col;</div><div class="line"><a name="l00607"></a><span class="lineno"> 607</span>  _vector_sum_row = vector_sum_row;</div><div class="line"><a name="l00608"></a><span class="lineno"> 608</span>  _bias = <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>;</div><div class="line"><a name="l00609"></a><span class="lineno"> 609</span>  _mm_result = mm_result;</div><div class="line"><a name="l00610"></a><span class="lineno"> 610</span>  _output = output;</div><div class="line"><a name="l00611"></a><span class="lineno"> 611</span>  _a_offset = a_offset;</div><div class="line"><a name="l00612"></a><span class="lineno"> 612</span>  _b_offset = b_offset;</div><div class="line"><a name="l00613"></a><span class="lineno"> 613</span>  _k_offset = a_offset * b_offset * k;</div><div class="line"><a name="l00614"></a><span class="lineno"> 614</span>  _output_stage = output_stage;</div><div class="line"><a name="l00615"></a><span class="lineno"> 615</span> </div><div class="line"><a name="l00616"></a><span class="lineno"> 616</span>  <span class="comment">// If a_offset == 0, vector_sum_col can be a nullptr</span></div><div class="line"><a name="l00617"></a><span class="lineno"> 617</span>  <span class="keywordflow">if</span>(a_offset != 0)</div><div class="line"><a name="l00618"></a><span class="lineno"> 618</span>  {</div><div class="line"><a name="l00619"></a><span class="lineno"> 619</span>  <span class="comment">// Check if vector_sum_col_shape should be slidden or not</span></div><div class="line"><a name="l00620"></a><span class="lineno"> 620</span>  <span class="comment">// Don't slide vector_sum_col_shape along the y dimension if vector_sum_col_shape has just 1 dimension and vector_sum_row_shape more than 1</span></div><div class="line"><a name="l00621"></a><span class="lineno"> 621</span>  <span class="comment">// This scenario can happen when the the matrix multiplication is used to perform a convolution operation</span></div><div class="line"><a name="l00622"></a><span class="lineno"> 622</span>  _slide_vector_sum_col = vector_sum_col-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>()-><a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml#a7c66505457d00ece3aa4b34cab80757d">tensor_shape</a>().<a class="code" href="classarm__compute_1_1_dimensions.xhtml#a80a5f2d6e3a697c9aad893a3b4242615">num_dimensions</a>() > 1;</div><div class="line"><a name="l00623"></a><span class="lineno"> 623</span>  }</div><div class="line"><a name="l00624"></a><span class="lineno"> 624</span> </div><div class="line"><a name="l00625"></a><span class="lineno"> 625</span>  <span class="comment">// Configure kernel window</span></div><div class="line"><a name="l00626"></a><span class="lineno"> 626</span>  <span class="keyword">auto</span> win_config = <a class="code" href="namespacearm__compute.xhtml#ad6630777dc2d315531f1e0b02491051f">validate_and_configure_window</a>(mm_result-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>(), output-><a class="code" href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">info</a>());</div><div class="line"><a name="l00627"></a><span class="lineno"> 627</span>  <a class="code" href="_error_8h.xhtml#a938dcd406ce611ef5345ad2531cdb948">ARM_COMPUTE_ERROR_THROW_ON</a>(win_config.first);</div><div class="line"><a name="l00628"></a><span class="lineno"> 628</span>  INEKernel::configure(win_config.second);</div><div class="line"><a name="l00629"></a><span class="lineno"> 629</span> </div><div class="line"><a name="l00630"></a><span class="lineno"> 630</span>  _function = get_configured_function(mm_result, vector_sum_row, output_stage);</div><div class="line"><a name="l00631"></a><span class="lineno"> 631</span> }</div><div class="line"><a name="l00632"></a><span class="lineno"> 632</span> </div><div class="line"><a name="l00633"></a><span class="lineno"><a class="line" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a6296f2754011b2221b343ccedfc0ba35"> 633</a></span> <a class="code" href="classarm__compute_1_1_status.xhtml">Status</a> <a class="code" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a6296f2754011b2221b343ccedfc0ba35">NEGEMMLowpOffsetContributionOutputStageKernel::validate</a>(<span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml">ITensorInfo</a> *mm_result, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml">ITensorInfo</a> *vector_sum_col,</div><div class="line"><a name="l00634"></a><span class="lineno"> 634</span>  <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml">ITensorInfo</a> *vector_sum_row, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml">ITensorInfo</a> *<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, <span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_i_tensor_info.xhtml">ITensorInfo</a> *output,</div><div class="line"><a name="l00635"></a><span class="lineno"> 635</span>  int32_t a_offset, int32_t b_offset, <a class="code" href="structarm__compute_1_1_g_e_m_m_lowp_output_stage_info.xhtml">GEMMLowpOutputStageInfo</a> output_stage)</div><div class="line"><a name="l00636"></a><span class="lineno"> 636</span> {</div><div class="line"><a name="l00637"></a><span class="lineno"> 637</span>  <a class="code" href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a>(mm_result, output);</div><div class="line"><a name="l00638"></a><span class="lineno"> 638</span>  <a class="code" href="_error_8h.xhtml#a8a1e1c105f0bdaf37db408c7cfcb77a4">ARM_COMPUTE_RETURN_ON_ERROR</a>(validate_arguments(mm_result, vector_sum_col, vector_sum_row, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">bias</a>, output, a_offset, b_offset, output_stage));</div><div class="line"><a name="l00639"></a><span class="lineno"> 639</span>  <a class="code" href="_error_8h.xhtml#a8a1e1c105f0bdaf37db408c7cfcb77a4">ARM_COMPUTE_RETURN_ON_ERROR</a>(<a class="code" href="namespacearm__compute.xhtml#ad6630777dc2d315531f1e0b02491051f">validate_and_configure_window</a>(mm_result-><a class="code" href="classarm__compute_1_1misc_1_1_i_cloneable.xhtml#a4d10e5012a872e7f78f2b539b673049d">clone</a>().get(), output-><a class="code" href="classarm__compute_1_1misc_1_1_i_cloneable.xhtml#a4d10e5012a872e7f78f2b539b673049d">clone</a>().get()).first);</div><div class="line"><a name="l00640"></a><span class="lineno"> 640</span>  <span class="keywordflow">return</span> <a class="code" href="classarm__compute_1_1_status.xhtml">Status</a>{};</div><div class="line"><a name="l00641"></a><span class="lineno"> 641</span> }</div><div class="line"><a name="l00642"></a><span class="lineno"> 642</span> </div><div class="line"><a name="l00643"></a><span class="lineno"><a class="line" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a112b35dd205c62ea6ed1447ef226da82"> 643</a></span> <span class="keywordtype">void</span> <a class="code" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a112b35dd205c62ea6ed1447ef226da82">NEGEMMLowpOffsetContributionOutputStageKernel::run</a>(<span class="keyword">const</span> <a class="code" href="classarm__compute_1_1_window.xhtml">Window</a> &window, <span class="keyword">const</span> <a class="code" href="structarm__compute_1_1_thread_info.xhtml">ThreadInfo</a> &<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a4f4125dba5283887b34f889b1c615c0c">info</a>)</div><div class="line"><a name="l00644"></a><span class="lineno"> 644</span> {</div><div class="line"><a name="l00645"></a><span class="lineno"> 645</span>  <a class="code" href="_error_8h.xhtml#a6dc630a6ae9cc063b3924bcea8dee9d6">ARM_COMPUTE_UNUSED</a>(<a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a4f4125dba5283887b34f889b1c615c0c">info</a>);</div><div class="line"><a name="l00646"></a><span class="lineno"> 646</span>  <a class="code" href="_validate_8h.xhtml#a1b35b0d258183cf9ef36adf684d0b88c">ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL</a>(<span class="keyword">this</span>);</div><div class="line"><a name="l00647"></a><span class="lineno"> 647</span>  <a class="code" href="_validate_8h.xhtml#a6eb9ce82815fe429250189da7592ba75">ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW</a>(<a class="code" href="classarm__compute_1_1_i_kernel.xhtml#ad34a46f53686c12a5c5e717cc9617fb6">INEKernel::window</a>(), <a class="code" href="classarm__compute_1_1_i_kernel.xhtml#ad34a46f53686c12a5c5e717cc9617fb6">window</a>);</div><div class="line"><a name="l00648"></a><span class="lineno"> 648</span>  _function(<a class="code" href="classarm__compute_1_1_i_kernel.xhtml#ad34a46f53686c12a5c5e717cc9617fb6">window</a>, _mm_result, _vector_sum_col, _vector_sum_row, _bias, _output, _a_offset, _b_offset, _k_offset, _slide_vector_sum_col, _output_stage);</div><div class="line"><a name="l00649"></a><span class="lineno"> 649</span> }</div><div class="line"><a name="l00650"></a><span class="lineno"> 650</span> </div><div class="line"><a name="l00651"></a><span class="lineno"> 651</span> } <span class="comment">// namespace arm_compute</span></div><div class="ttc" id="classarm__compute_1_1_i_kernel_xhtml_ad34a46f53686c12a5c5e717cc9617fb6"><div class="ttname"><a href="classarm__compute_1_1_i_kernel.xhtml#ad34a46f53686c12a5c5e717cc9617fb6">arm_compute::IKernel::window</a></div><div class="ttdeci">const Window & window() const</div><div class="ttdoc">The maximum window the kernel can be executed on.</div><div class="ttdef"><b>Definition:</b> <a href="_i_kernel_8cpp_source.xhtml#l00028">IKernel.cpp:28</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_a5558e2cc22f7f4771653d992c8ad8864ab300cae200f67712c1eb9234e28158ca"><div class="ttname"><a href="namespacearm__compute.xhtml#a5558e2cc22f7f4771653d992c8ad8864ab300cae200f67712c1eb9234e28158ca">arm_compute::GEMMLowpOutputStageType::QUANTIZE_DOWN_FIXEDPOINT</a></div><div class="ttdoc">Quantize to uint8 using a fixed point multiplication.</div></div> |
| <div class="ttc" id="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_xhtml_a112b35dd205c62ea6ed1447ef226da82"><div class="ttname"><a href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a112b35dd205c62ea6ed1447ef226da82">arm_compute::NEGEMMLowpOffsetContributionOutputStageKernel::run</a></div><div class="ttdeci">void run(const Window &window, const ThreadInfo &info) override</div><div class="ttdoc">Execute the kernel on the passed window.</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8cpp_source.xhtml#l00643">NEGEMMLowpOffsetContributionOutputStageKernel.cpp:643</a></div></div> |
| <div class="ttc" id="_i_tensor_8h_xhtml"><div class="ttname"><a href="_i_tensor_8h.xhtml">ITensor.h</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_c_l_tensor_xhtml_ad45f0c01a0713dfb6bd7232c7f396fc4"><div class="ttname"><a href="classarm__compute_1_1_c_l_tensor.xhtml#ad45f0c01a0713dfb6bd7232c7f396fc4">arm_compute::CLTensor::info</a></div><div class="ttdeci">TensorInfo * info() const override</div><div class="ttdoc">Interface to be implemented by the child class to return the tensor's metadata.</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_tensor_8cpp_source.xhtml#l00035">CLTensor.cpp:35</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_aa76b4a6e74940dabc5b7fc6b2dab3545"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#aa76b4a6e74940dabc5b7fc6b2dab3545">arm_compute::test::validation::b</a></div><div class="ttdeci">SimpleTensor< float > b</div><div class="ttdef"><b>Definition:</b> <a href="_c_p_p_2_d_f_t_8cpp_source.xhtml#l00157">DFT.cpp:157</a></div></div> |
| <div class="ttc" id="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8h_xhtml"><div class="ttname"><a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8h.xhtml">NEGEMMLowpOffsetContributionOutputStageKernel.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_ad6630777dc2d315531f1e0b02491051f"><div class="ttname"><a href="namespacearm__compute.xhtml#ad6630777dc2d315531f1e0b02491051f">arm_compute::validate_and_configure_window</a></div><div class="ttdeci">std::pair< Status, Window > validate_and_configure_window(ITensorInfo *input, ITensorInfo *weights, ITensorInfo *biases, ITensorInfo *output, const PadStrideInfo &conv_info, unsigned int depth_multiplier, const Size2D &dilation)</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_depthwise_convolution_layer_native_kernel_8cpp_source.xhtml#l00221">NEDepthwiseConvolutionLayerNativeKernel.cpp:221</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_xhtml_a2851f631e9660a4dd9644cb749282723"><div class="ttname"><a href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a2851f631e9660a4dd9644cb749282723">arm_compute::NEGEMMLowpOffsetContributionOutputStageKernel::NEGEMMLowpOffsetContributionOutputStageKernel</a></div><div class="ttdeci">NEGEMMLowpOffsetContributionOutputStageKernel()</div><div class="ttdoc">Constructor.</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8cpp_source.xhtml#l00586">NEGEMMLowpOffsetContributionOutputStageKernel.cpp:586</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a8a1e1c105f0bdaf37db408c7cfcb77a4"><div class="ttname"><a href="_error_8h.xhtml#a8a1e1c105f0bdaf37db408c7cfcb77a4">ARM_COMPUTE_RETURN_ON_ERROR</a></div><div class="ttdeci">#define ARM_COMPUTE_RETURN_ON_ERROR(status)</div><div class="ttdoc">Checks if a status contains an error and returns it.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00193">Error.h:193</a></div></div> |
| <div class="ttc" id="_window_8h_xhtml"><div class="ttname"><a href="_window_8h.xhtml">Window.h</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml_ae7eed178dac535c6e727061b1f5bc6eb"><div class="ttname"><a href="_validate_8h.xhtml#ae7eed178dac535c6e727061b1f5bc6eb">ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN</a></div><div class="ttdeci">#define ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)</div><div class="ttdef"><b>Definition:</b> <a href="_validate_8h_source.xhtml#l00791">Validate.h:791</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_i_tensor_info_xhtml"><div class="ttname"><a href="classarm__compute_1_1_i_tensor_info.xhtml">arm_compute::ITensorInfo</a></div><div class="ttdoc">Store the tensor's metadata.</div><div class="ttdef"><b>Definition:</b> <a href="_i_tensor_info_8h_source.xhtml#l00040">ITensorInfo.h:40</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a938dcd406ce611ef5345ad2531cdb948"><div class="ttname"><a href="_error_8h.xhtml#a938dcd406ce611ef5345ad2531cdb948">ARM_COMPUTE_ERROR_THROW_ON</a></div><div class="ttdeci">#define ARM_COMPUTE_ERROR_THROW_ON(status)</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00327">Error.h:327</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_status_xhtml"><div class="ttname"><a href="classarm__compute_1_1_status.xhtml">arm_compute::Status</a></div><div class="ttdoc">Status class.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00052">Error.h:52</a></div></div> |
| <div class="ttc" id="_tensor_info_8h_xhtml"><div class="ttname"><a href="_tensor_info_8h.xhtml">TensorInfo.h</a></div></div> |
| <div class="ttc" id="_n_e_asymm_8h_xhtml"><div class="ttname"><a href="_n_e_asymm_8h.xhtml">NEAsymm.h</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a206d6e247e0957ac3dee45d27756fc25"><div class="ttname"><a href="_error_8h.xhtml#a206d6e247e0957ac3dee45d27756fc25">ARM_COMPUTE_RETURN_ERROR_ON</a></div><div class="ttdeci">#define ARM_COMPUTE_RETURN_ERROR_ON(cond)</div><div class="ttdoc">If the condition is true, an error is returned.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00244">Error.h:244</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_i_tensor_xhtml"><div class="ttname"><a href="classarm__compute_1_1_i_tensor.xhtml">arm_compute::ITensor</a></div><div class="ttdoc">Interface for NEON tensor.</div><div class="ttdef"><b>Definition:</b> <a href="_i_tensor_8h_source.xhtml#l00036">ITensor.h:36</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_ab7980fa5ee693e3282a76da047a1c3b5"><div class="ttname"><a href="namespacearm__compute.xhtml#ab7980fa5ee693e3282a76da047a1c3b5">arm_compute::calculate_max_window</a></div><div class="ttdeci">Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps=Steps(), bool skip_border=false, BorderSize border_size=BorderSize())</div><div class="ttdoc">Calculate the maximum window for a given tensor shape and border setting.</div><div class="ttdef"><b>Definition:</b> <a href="src_2core_2_helpers_8cpp_source.xhtml#l00028">Helpers.cpp:28</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="namespacearm__compute_xhtml_a47be6fa38308d0003c25b60b7dbc45ce"><div class="ttname"><a href="namespacearm__compute.xhtml#a47be6fa38308d0003c25b60b7dbc45ce">arm_compute::auto_init_if_empty</a></div><div class="ttdeci">bool auto_init_if_empty(ITensorInfo &info, const TensorShape &shape, int num_channels, DataType data_type, QuantizationInfo quantization_info=QuantizationInfo())</div><div class="ttdoc">Auto initialize the tensor info (shape, number of channels and data type) if the current assignment i...</div><div class="ttdef"><b>Definition:</b> <a href="_helpers_8inl_source.xhtml#l00201">Helpers.inl:201</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c"><div class="ttname"><a href="namespacearm__compute.xhtml#ab4e88c89b3b7ea1735996cc4def22d58aa1e28eee0339658d39a8b4d325b56e9c">arm_compute::Format::S32</a></div><div class="ttdoc">1 channel, 1 S32 per channel</div></div> |
| <div class="ttc" id="arm__compute_2core_2_utils_8h_xhtml"><div class="ttname"><a href="arm__compute_2core_2_utils_8h.xhtml">Utils.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1wrapper_xhtml_aa16ace001ab8287faa46d6962f369219"><div class="ttname"><a href="namespacearm__compute_1_1wrapper.xhtml#aa16ace001ab8287faa46d6962f369219">arm_compute::wrapper::vgetlane</a></div><div class="ttdeci">uint8_t vgetlane(const uint8x8_t vector, const unsigned int lane)</div><div class="ttdef"><b>Definition:</b> <a href="getlane_8h_source.xhtml#l00091">getlane.h:91</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_window_xhtml_aa96e81276ee4f87ab386cd05a5539a7d"><div class="ttname"><a href="classarm__compute_1_1_window.xhtml#aa96e81276ee4f87ab386cd05a5539a7d">arm_compute::Window::DimX</a></div><div class="ttdeci">static constexpr size_t DimX</div><div class="ttdoc">Alias for dimension 0 also known as X dimension.</div><div class="ttdef"><b>Definition:</b> <a href="_window_8h_source.xhtml#l00043">Window.h:43</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a6dc630a6ae9cc063b3924bcea8dee9d6"><div class="ttname"><a href="_error_8h.xhtml#a6dc630a6ae9cc063b3924bcea8dee9d6">ARM_COMPUTE_UNUSED</a></div><div class="ttdeci">#define ARM_COMPUTE_UNUSED(...)</div><div class="ttdoc">To avoid unused variables warnings.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00160">Error.h:160</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml_a27e4638546c88b8916f967e6e54480a9"><div class="ttname"><a href="_validate_8h.xhtml#a27e4638546c88b8916f967e6e54480a9">ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES</a></div><div class="ttdeci">#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(...)</div><div class="ttdef"><b>Definition:</b> <a href="_validate_8h_source.xhtml#l00443">Validate.h:443</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_i_tensor_info_xhtml_a7c66505457d00ece3aa4b34cab80757d"><div class="ttname"><a href="classarm__compute_1_1_i_tensor_info.xhtml#a7c66505457d00ece3aa4b34cab80757d">arm_compute::ITensorInfo::tensor_shape</a></div><div class="ttdeci">virtual const TensorShape & tensor_shape() const =0</div><div class="ttdoc">Size for each dimension of the tensor.</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_a7fc93f37dac131a1a40b7921f9df3a9a"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a7fc93f37dac131a1a40b7921f9df3a9a">arm_compute::test::validation::output_shape</a></div><div class="ttdeci">output_shape</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_convolution_layer_8cpp_source.xhtml#l00174">ConvolutionLayer.cpp:174</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml_a86084036bd3851575ef871ad5bf079a7"><div class="ttname"><a href="_error_8h.xhtml#a86084036bd3851575ef871ad5bf079a7">ARM_COMPUTE_RETURN_ERROR_ON_MSG</a></div><div class="ttdeci">#define ARM_COMPUTE_RETURN_ERROR_ON_MSG(cond,...)</div><div class="ttdoc">If the condition is true, an error is returned.</div><div class="ttdef"><b>Definition:</b> <a href="_error_8h_source.xhtml#l00214">Error.h:214</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1misc_1_1_i_cloneable_xhtml_a4d10e5012a872e7f78f2b539b673049d"><div class="ttname"><a href="classarm__compute_1_1misc_1_1_i_cloneable.xhtml#a4d10e5012a872e7f78f2b539b673049d">arm_compute::misc::ICloneable::clone</a></div><div class="ttdeci">virtual std::unique_ptr< T > clone() const =0</div><div class="ttdoc">Provide a clone of the current object of class T.</div></div> |
| <div class="ttc" id="structarm__compute_1_1_g_e_m_m_lowp_output_stage_info_xhtml"><div class="ttname"><a href="structarm__compute_1_1_g_e_m_m_lowp_output_stage_info.xhtml">arm_compute::GEMMLowpOutputStageInfo</a></div><div class="ttdoc">GEMMLowp output stage info.</div><div class="ttdef"><b>Definition:</b> <a href="arm__compute_2core_2_types_8h_source.xhtml#l01845">Types.h:1845</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_i_tensor_xhtml_a0e95dc1e53c361348314873b168ae237"><div class="ttname"><a href="classarm__compute_1_1_i_tensor.xhtml#a0e95dc1e53c361348314873b168ae237">arm_compute::ITensor::info</a></div><div class="ttdeci">virtual ITensorInfo * info() const =0</div><div class="ttdoc">Interface to be implemented by the child class to return the tensor's metadata.</div></div> |
| <div class="ttc" id="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_xhtml_a97ebe5c0444a53d58d9b9f079ebe2d0f"><div class="ttname"><a href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a97ebe5c0444a53d58d9b9f079ebe2d0f">arm_compute::NEGEMMLowpOffsetContributionOutputStageKernel::configure</a></div><div class="ttdeci">void configure(const ITensor *mm_result, const ITensor *vector_sum_col, const ITensor *vector_sum_row, const ITensor *bias, ITensor *output, int32_t k, int32_t a_offset, int32_t b_offset, GEMMLowpOutputStageInfo output_stage)</div><div class="ttdoc">Initialise the kernel's input and output.</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8cpp_source.xhtml#l00593">NEGEMMLowpOffsetContributionOutputStageKernel.cpp:593</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a9aeced5a5128f60a31ea3e327a45ee21"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a9aeced5a5128f60a31ea3e327a45ee21">arm_compute::test::validation::has_bias</a></div><div class="ttdeci">const bool has_bias</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_o_n_2_im2_col_8cpp_source.xhtml#l00147">Im2Col.cpp:147</a></div></div> |
| <div class="ttc" id="_error_8h_xhtml"><div class="ttname"><a href="_error_8h.xhtml">Error.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a3a77be8aebd8e00522b32061d46ccdbd"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a3a77be8aebd8e00522b32061d46ccdbd">arm_compute::test::validation::bias</a></div><div class="ttdeci">CLTensor bias</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_convolution_layer_8cpp_source.xhtml#l00181">ConvolutionLayer.cpp:181</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_a5558e2cc22f7f4771653d992c8ad8864a079e2ddc95b344b5cb0188bed9a80d8b"><div class="ttname"><a href="namespacearm__compute.xhtml#a5558e2cc22f7f4771653d992c8ad8864a079e2ddc95b344b5cb0188bed9a80d8b">arm_compute::GEMMLowpOutputStageType::QUANTIZE_DOWN</a></div><div class="ttdoc">Quantize to uint8 using an integer multiplication.</div></div> |
| <div class="ttc" id="classarm__compute_1_1_window_xhtml_ad2d402364fa822b0b7775081291eeca9"><div class="ttname"><a href="classarm__compute_1_1_window.xhtml#ad2d402364fa822b0b7775081291eeca9">arm_compute::Window::DimY</a></div><div class="ttdeci">static constexpr size_t DimY</div><div class="ttdoc">Alias for dimension 1 also known as Y dimension.</div><div class="ttdef"><b>Definition:</b> <a href="_window_8h_source.xhtml#l00045">Window.h:45</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml_a921b705e9e3e0fe928928447869e62a5"><div class="ttname"><a href="_validate_8h.xhtml#a921b705e9e3e0fe928928447869e62a5">ARM_COMPUTE_ERROR_ON_NULLPTR</a></div><div class="ttdeci">#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)</div><div class="ttdef"><b>Definition:</b> <a href="_validate_8h_source.xhtml#l00161">Validate.h:161</a></div></div> |
| <div class="ttc" id="structarm__compute_1_1_thread_info_xhtml"><div class="ttname"><a href="structarm__compute_1_1_thread_info.xhtml">arm_compute::ThreadInfo</a></div><div class="ttdoc">Information about executing thread and CPU.</div><div class="ttdef"><b>Definition:</b> <a href="_c_p_p_types_8h_source.xhtml#l00225">CPPTypes.h:225</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_window_xhtml_a893d17b56b9abc4423ce26e9a24ac5dc"><div class="ttname"><a href="classarm__compute_1_1_window.xhtml#a893d17b56b9abc4423ce26e9a24ac5dc">arm_compute::Window::DimZ</a></div><div class="ttdeci">static constexpr size_t DimZ</div><div class="ttdoc">Alias for dimension 2 also known as Z dimension.</div><div class="ttdef"><b>Definition:</b> <a href="_window_8h_source.xhtml#l00047">Window.h:47</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_dimensions_xhtml_a80a5f2d6e3a697c9aad893a3b4242615"><div class="ttname"><a href="classarm__compute_1_1_dimensions.xhtml#a80a5f2d6e3a697c9aad893a3b4242615">arm_compute::Dimensions::num_dimensions</a></div><div class="ttdeci">unsigned int num_dimensions() const</div><div class="ttdoc">Returns the effective dimensionality of the tensor.</div><div class="ttdef"><b>Definition:</b> <a href="_dimensions_8h_source.xhtml#l00122">Dimensions.h:122</a></div></div> |
| <div class="ttc" id="arm__compute_2core_2_helpers_8h_xhtml"><div class="ttname"><a href="arm__compute_2core_2_helpers_8h.xhtml">Helpers.h</a></div></div> |
| <div class="ttc" id="namespacearm__compute_xhtml_a5002bf7ec46d52971f9526e94172cfee"><div class="ttname"><a href="namespacearm__compute.xhtml#a5002bf7ec46d52971f9526e94172cfee">arm_compute::execute_window_loop</a></div><div class="ttdeci">void execute_window_loop(const Window &w, L &&lambda_function, Ts &&... iterators)</div><div class="ttdoc">Iterate through the passed window, automatically adjusting the iterators and calling the lambda_funct...</div><div class="ttdef"><b>Definition:</b> <a href="_helpers_8inl_source.xhtml#l00122">Helpers.inl:122</a></div></div> |
| <div class="ttc" id="_access_window_static_8h_xhtml"><div class="ttname"><a href="_access_window_static_8h.xhtml">AccessWindowStatic.h</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_xhtml_adb79bb3ad15444f00b55b35f1d6e16b7"><div class="ttname"><a href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#adb79bb3ad15444f00b55b35f1d6e16b7">arm_compute::NEGEMMLowpOffsetContributionOutputStageKernel::NEGEMMLowpOffsetContributionOutputStageFunction</a></div><div class="ttdeci">std::function< void(const Window, const ITensor *, const ITensor *, const ITensor *, const ITensor *, ITensor *, int32_t, int32_t, int32_t, bool, GEMMLowpOutputStageInfo)> NEGEMMLowpOffsetContributionOutputStageFunction</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8h_source.xhtml#l00118">NEGEMMLowpOffsetContributionOutputStageKernel.h:118</a></div></div> |
| <div class="ttc" id="wrapper_8h_xhtml"><div class="ttname"><a href="wrapper_8h.xhtml">wrapper.h</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_xhtml_a6296f2754011b2221b343ccedfc0ba35"><div class="ttname"><a href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel.xhtml#a6296f2754011b2221b343ccedfc0ba35">arm_compute::NEGEMMLowpOffsetContributionOutputStageKernel::validate</a></div><div class="ttdeci">static Status validate(const ITensorInfo *mm_result, const ITensorInfo *vector_sum_col, const ITensorInfo *vector_sum_row, const ITensorInfo *bias, const ITensorInfo *output, int32_t a_offset, int32_t b_offset, GEMMLowpOutputStageInfo output_stage)</div><div class="ttdoc">Static function to check if given info will lead to a valid configuration of NEGEMMLowpOffsetContribu...</div><div class="ttdef"><b>Definition:</b> <a href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8cpp_source.xhtml#l00633">NEGEMMLowpOffsetContributionOutputStageKernel.cpp:633</a></div></div> |
| <div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a4f4125dba5283887b34f889b1c615c0c"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a4f4125dba5283887b34f889b1c615c0c">arm_compute::test::validation::info</a></div><div class="ttdeci">info</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_convolution_layer_8cpp_source.xhtml#l00174">ConvolutionLayer.cpp:174</a></div></div> |
| <div class="ttc" id="arm__compute_2core_2_types_8h_xhtml"><div class="ttname"><a href="arm__compute_2core_2_types_8h.xhtml">Types.h</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml_a6eb9ce82815fe429250189da7592ba75"><div class="ttname"><a href="_validate_8h.xhtml#a6eb9ce82815fe429250189da7592ba75">ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW</a></div><div class="ttdeci">#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)</div><div class="ttdef"><b>Definition:</b> <a href="_validate_8h_source.xhtml#l00205">Validate.h:205</a></div></div> |
| <div class="ttc" id="classarm__compute_1_1_window_xhtml"><div class="ttname"><a href="classarm__compute_1_1_window.xhtml">arm_compute::Window</a></div><div class="ttdoc">Describe a multidimensional execution window.</div><div class="ttdef"><b>Definition:</b> <a href="_window_8h_source.xhtml#l00039">Window.h:39</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml_a1b35b0d258183cf9ef36adf684d0b88c"><div class="ttname"><a href="_validate_8h.xhtml#a1b35b0d258183cf9ef36adf684d0b88c">ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL</a></div><div class="ttdeci">#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)</div><div class="ttdef"><b>Definition:</b> <a href="_validate_8h_source.xhtml#l00940">Validate.h:940</a></div></div> |
| <div class="ttc" id="_validate_8h_xhtml"><div class="ttname"><a href="_validate_8h.xhtml">Validate.h</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_aebb8dcc11953d78e620bbef0b9e2183.xhtml">core</a></li><li class="navelem"><a class="el" href="dir_d91e1966cc57df4f49d1e07be7697713.xhtml">NEON</a></li><li class="navelem"><a class="el" href="dir_1ce938f580a6f304b99685e2c1dd19db.xhtml">kernels</a></li><li class="navelem"><a class="el" href="_n_e_g_e_m_m_lowp_offset_contribution_output_stage_kernel_8cpp.xhtml">NEGEMMLowpOffsetContributionOutputStageKernel.cpp</a></li> |
| <li class="footer">Generated on Mon Sep 2 2019 11:47:24 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> |
