1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
|
<!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 1.8.13"/>
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<title>ArmNN: src/backends/reference/workloads/ConvImpl.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">
$(document).ready(initResizable);
</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>
<link href="doxygen.css" rel="stylesheet" type="text/css" />
</head>
<body>
<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
<div id="titlearea">
<table cellspacing="0" cellpadding="0">
<tbody>
<tr style="height: 56px;">
<td id="projectalign" style="padding-left: 0.5em;">
<div id="projectname">ArmNN
 <span id="projectnumber">NotReleased</span>
</div>
</td>
</tr>
</tbody>
</table>
</div>
<!-- end header part -->
<!-- Generated by Doxygen 1.8.13 -->
<script type="text/javascript">
var searchBox = new SearchBox("searchBox", "search",false,'Search');
</script>
<script type="text/javascript" src="menudata.js"></script>
<script type="text/javascript" src="menu.js"></script>
<script type="text/javascript">
$(function() {
initMenu('',true,false,'search.php','Search');
$(document).ready(function() { init_search(); });
});
</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">
$(document).ready(function(){initNavTree('_conv_impl_8cpp_source.html','');});
</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">ConvImpl.cpp</div> </div>
</div><!--header-->
<div class="contents">
<a href="_conv_impl_8cpp.html">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 © 2017 Arm Ltd. All rights reserved.</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment">// SPDX-License-Identifier: MIT</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span> <span class="comment">//</span></div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span> </div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span> <span class="preprocessor">#include "<a class="code" href="_conv_impl_8hpp.html">ConvImpl.hpp</a>"</span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span> </div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span> <span class="preprocessor">#include <boost/assert.hpp></span></div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span> </div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span> <span class="preprocessor">#include <cmath></span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span> <span class="preprocessor">#include <limits></span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span> </div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span> <span class="keyword">namespace </span><a class="code" href="namespacearmnn.html">armnn</a></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span> {</div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span> </div><div class="line"><a name="l00016"></a><span class="lineno"><a class="line" href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#a1ec4f749fbbdacbe2cc7e8895d5b65ed"> 16</a></span> <a class="code" href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#a1ec4f749fbbdacbe2cc7e8895d5b65ed">QuantizedMultiplierSmallerThanOne::QuantizedMultiplierSmallerThanOne</a>(<span class="keywordtype">float</span> multiplier)</div><div class="line"><a name="l00017"></a><span class="lineno"> 17</span> {</div><div class="line"><a name="l00018"></a><span class="lineno"> 18</span>  BOOST_ASSERT(multiplier >= 0.0f && multiplier < 1.0f);</div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span>  <span class="keywordflow">if</span> (multiplier == 0.0f)</div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span>  {</div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span>  m_Multiplier = 0;</div><div class="line"><a name="l00022"></a><span class="lineno"> 22</span>  m_RightShift = 0;</div><div class="line"><a name="l00023"></a><span class="lineno"> 23</span>  }</div><div class="line"><a name="l00024"></a><span class="lineno"> 24</span>  <span class="keywordflow">else</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="keyword">const</span> <span class="keywordtype">double</span> q = std::frexp(multiplier, &m_RightShift);</div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span>  m_RightShift = -m_RightShift;</div><div class="line"><a name="l00028"></a><span class="lineno"> 28</span>  int64_t qFixed = <span class="keyword">static_cast<</span>int64_t<span class="keyword">></span>(std::round(q * (1ll << 31)));</div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span>  BOOST_ASSERT(qFixed <= (1ll << 31));</div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span>  <span class="keywordflow">if</span> (qFixed == (1ll << 31))</div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span>  {</div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span>  qFixed /= 2;</div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span>  --m_RightShift;</div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span>  }</div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span>  BOOST_ASSERT(m_RightShift >= 0);</div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span>  BOOST_ASSERT(qFixed <= std::numeric_limits<int32_t>::max());</div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span>  m_Multiplier = <span class="keyword">static_cast<</span>int32_t<span class="keyword">></span>(qFixed);</div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span>  }</div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span> }</div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span> </div><div class="line"><a name="l00041"></a><span class="lineno"><a class="line" href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#ac04834893e568dfea4642ecb575faed8"> 41</a></span> int32_t <a class="code" href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#ac04834893e568dfea4642ecb575faed8">QuantizedMultiplierSmallerThanOne::operator*</a>(int32_t rhs)<span class="keyword"> const</span></div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span> <span class="keyword"></span>{</div><div class="line"><a name="l00043"></a><span class="lineno"> 43</span>  int32_t x = SaturatingRoundingDoublingHighMul(rhs, m_Multiplier);</div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span>  <span class="keywordflow">return</span> RoundingDivideByPOT(x, m_RightShift);</div><div class="line"><a name="l00045"></a><span class="lineno"> 45</span> }</div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span> </div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span> int32_t QuantizedMultiplierSmallerThanOne::SaturatingRoundingDoublingHighMul(int32_t a, int32_t b)</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="comment">// Check for overflow.</span></div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span>  <span class="keywordflow">if</span> (a == b && a == std::numeric_limits<int32_t>::min())</div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span>  {</div><div class="line"><a name="l00052"></a><span class="lineno"> 52</span>  <span class="keywordflow">return</span> std::numeric_limits<int32_t>::max();</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>  int64_t a_64(a);</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span>  int64_t b_64(b);</div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span>  int64_t ab_64 = a_64 * b_64;</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>  int32_t nudge = ab_64 >= 0 ? (1 << 30) : (1 - (1 << 30));</div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span>  int32_t ab_x2_high32 = <span class="keyword">static_cast<</span>std::int32_t<span class="keyword">></span>((ab_64 + nudge) / (1ll << 31));</div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span>  <span class="keywordflow">return</span> ab_x2_high32;</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> int32_t QuantizedMultiplierSmallerThanOne::RoundingDivideByPOT(int32_t x, <span class="keywordtype">int</span> exponent)</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>  BOOST_ASSERT(exponent >= 0 && exponent <= 31);</div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  int32_t mask = (1 << exponent) - 1;</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  int32_t remainder = x & mask;</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span>  int32_t threshold = (mask >> 1) + (x < 0 ? 1 : 0);</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  <span class="keywordflow">return</span> (x >> exponent) + (remainder > threshold ? 1 : 0);</div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span> }</div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span> </div><div class="line"><a name="l00071"></a><span class="lineno"><a class="line" href="namespacearmnn.html#af98115cd07776d3fa8424434d2a7a897"> 71</a></span> <span class="keywordtype">void</span> <a class="code" href="namespacearmnn.html#af98115cd07776d3fa8424434d2a7a897">Convolve</a>(<span class="keyword">const</span> <a class="code" href="classarmnn_1_1_tensor_shape.html">TensorShape</a>& rInputShape,</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span>  <a class="code" href="classarmnn_1_1_decoder.html">Decoder<float></a>& rInputDecoder,</div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  <span class="keyword">const</span> <a class="code" href="classarmnn_1_1_tensor_shape.html">TensorShape</a>& rOutputShape,</div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  <a class="code" href="classarmnn_1_1_encoder.html">Encoder<float></a>& rOutputEncoder,</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span>  <span class="keyword">const</span> <a class="code" href="classarmnn_1_1_tensor_shape.html">TensorShape</a>& rFilterShape,</div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span>  <a class="code" href="classarmnn_1_1_decoder.html">Decoder<float></a>& rFilterDecoder,</div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span>  <span class="keywordtype">bool</span> biasEnabled,</div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>  <a class="code" href="classarmnn_1_1_decoder.html">Decoder<float></a>* pBiasDecoder,</div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span>  <a class="code" href="namespacearmnn.html#ad1d5cce2d9e9a5d61c243e5c989112e0">DataLayout</a> dataLayout,</div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> paddingTop,</div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> paddingLeft,</div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xStride,</div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> yStride,</div><div class="line"><a name="l00084"></a><span class="lineno"> 84</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xDilation,</div><div class="line"><a name="l00085"></a><span class="lineno"> 85</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> yDilation,</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  <span class="keywordtype">bool</span> depthwise)</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>  <span class="keywordflow">if</span> (biasEnabled && !pBiasDecoder)</div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  {</div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>  <span class="keywordflow">throw</span> <a class="code" href="classarmnn_1_1_invalid_argument_exception.html">InvalidArgumentException</a>(<span class="stringliteral">"Bias is enabled but the bias data is invalid"</span>);</div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  }</div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  <span class="keyword">const</span> <a class="code" href="classarmnn_utils_1_1_data_layout_indexed.html">armnnUtils::DataLayoutIndexed</a> dataLayoutIndexed(dataLayout);</div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span> </div><div class="line"><a name="l00094"></a><span class="lineno"> 94</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> channelsIndex = dataLayoutIndexed.<a class="code" href="classarmnn_utils_1_1_data_layout_indexed.html#a861b2621ee46e4b63379988b360b8cd9">GetChannelsIndex</a>();</div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> heightIndex = dataLayoutIndexed.<a class="code" href="classarmnn_utils_1_1_data_layout_indexed.html#a61c00316c443adc233c24e85c6c5b740">GetHeightIndex</a>();</div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> widthIndex = dataLayoutIndexed.<a class="code" href="classarmnn_utils_1_1_data_layout_indexed.html#a414e6f95548e6f7a01d5028b55ad3941">GetWidthIndex</a>();</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>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> depthMultiplier = depthwise ? rFilterShape[0] : 1;</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> inputChannels = depthwise ? rFilterShape[1] : rFilterShape[channelsIndex];</div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> outputChannels = depthwise ? inputChannels * depthMultiplier : rFilterShape[0];</div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span> </div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> batchSize = rOutputShape[0];</div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> outputHeight = rOutputShape[heightIndex];</div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> outputWidth = rOutputShape[widthIndex];</div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> inputHeight = rInputShape[heightIndex];</div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> inputWidth = rInputShape[widthIndex];</div><div class="line"><a name="l00107"></a><span class="lineno"> 107</span> </div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> filterHeight = depthwise ? rFilterShape[2] : rFilterShape[heightIndex];</div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> filterWidth = depthwise ? rFilterShape[3] : rFilterShape[widthIndex];</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="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> batchIdx = 0; batchIdx < batchSize; batchIdx++)</div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span>  {</div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> cOutput = 0; cOutput < outputChannels; cOutput++)</div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>  {</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> yOutput = 0; yOutput < outputHeight; yOutput++)</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="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xOutput = 0; xOutput < outputWidth; xOutput++)</div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  {</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  <span class="comment">// This loop goes over each output element.</span></div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  <span class="keywordtype">float</span> sum = 0.0f;</div><div class="line"><a name="l00121"></a><span class="lineno"> 121</span> </div><div class="line"><a name="l00122"></a><span class="lineno"> 122</span>  <span class="comment">// For depthwise, each output channel corresponds to exactly one input channel.</span></div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  <span class="comment">// For normal, must loop over each input channel.</span></div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> cInput = 0; cInput < (depthwise ? 1 : inputChannels); cInput++)</div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span>  {</div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> depthwiseMultiplierIdx = 0;</div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>  <span class="keywordflow">if</span> (depthwise)</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>  {</div><div class="line"><a name="l00129"></a><span class="lineno"> 129</span>  cInput = cOutput / depthMultiplier;</div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span>  depthwiseMultiplierIdx = cOutput % depthMultiplier;</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> </div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> yFilter = 0; yFilter < filterHeight; yFilter++)</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">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xFilter = 0; xFilter < filterWidth; xFilter++)</div><div class="line"><a name="l00136"></a><span class="lineno"> 136</span>  {</div><div class="line"><a name="l00137"></a><span class="lineno"> 137</span>  <span class="comment">// This loop goes over each input element for each output element.</span></div><div class="line"><a name="l00138"></a><span class="lineno"> 138</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> filterIndex = 0;</div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span> </div><div class="line"><a name="l00140"></a><span class="lineno"> 140</span>  <span class="comment">// Since dimensionality of kernel depends on depthwiseness, so does index.</span></div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span>  <span class="keywordflow">if</span> (depthwise)</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>  filterIndex = depthwiseMultiplierIdx * filterWidth * filterHeight * inputChannels +</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  cInput * filterWidth * filterHeight +</div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span>  yFilter * filterWidth +</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>  xFilter;</div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  }</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span>  {</div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span>  <span class="comment">// Keep this implementation, as using DataLayoutIndexed::GetIndex causes great</span></div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span>  <span class="comment">// performance regression.</span></div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span>  <span class="keywordflow">if</span> (dataLayout == <a class="code" href="namespacearmnn.html#ad1d5cce2d9e9a5d61c243e5c989112e0ad066db54b89b0912e7e7c6da51e2da51">DataLayout::NHWC</a>)</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>  filterIndex = cOutput * filterHeight * filterWidth * inputChannels +</div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span>  yFilter * filterWidth * inputChannels +</div><div class="line"><a name="l00156"></a><span class="lineno"> 156</span>  xFilter * inputChannels +</div><div class="line"><a name="l00157"></a><span class="lineno"> 157</span>  cInput;</div><div class="line"><a name="l00158"></a><span class="lineno"> 158</span>  }</div><div class="line"><a name="l00159"></a><span class="lineno"> 159</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span>  {</div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span>  filterIndex = cOutput * filterWidth * filterHeight * inputChannels +</div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  cInput * filterWidth * filterHeight +</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  yFilter * filterWidth +</div><div class="line"><a name="l00164"></a><span class="lineno"> 164</span>  xFilter;</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>  }</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>  rFilterDecoder.<a class="code" href="classarmnn_1_1_base_iterator.html#a1ec75b077d774dbfebf3662e8e4363c9">SetIndex</a>(filterIndex, cOutput);</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>  <span class="keywordtype">float</span> filterValue = rFilterDecoder.<a class="code" href="classarmnn_1_1_decoder.html#ac729108381e2340bea12877971713ecb">Get</a>();</div><div class="line"><a name="l00170"></a><span class="lineno"> 170</span> </div><div class="line"><a name="l00171"></a><span class="lineno"> 171</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> yInput = yOutput * yStride + yFilter * yDilation;</div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xInput = xOutput * xStride + xFilter * xDilation;</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span> </div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>  <span class="keywordtype">float</span> inputValue;</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span> </div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span>  <span class="comment">// Check if we're in the padding.</span></div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>  <span class="keywordflow">if</span> (yInput < paddingTop || yInput >= inputHeight + paddingTop ||</div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>  xInput < paddingLeft || xInput >= inputWidth + paddingLeft )</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>  inputValue = 0.0f;</div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  }</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>  {</div><div class="line"><a name="l00184"></a><span class="lineno"> 184</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> inputIndex = 0;</div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span> </div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>  <span class="comment">// Keep this implementation, as using DataLayoutIndexed::GetIndex causes great</span></div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span>  <span class="comment">// performance regression.</span></div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span>  <span class="keywordflow">if</span> (dataLayout == <a class="code" href="namespacearmnn.html#ad1d5cce2d9e9a5d61c243e5c989112e0ad066db54b89b0912e7e7c6da51e2da51">DataLayout::NHWC</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>  inputIndex = batchIdx * inputHeight * inputWidth * inputChannels +</div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  (yInput - paddingTop) * inputWidth * inputChannels +</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  (xInput - paddingLeft) * inputChannels +</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  cInput;</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span>  }</div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span>  <span class="keywordflow">else</span></div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>  {</div><div class="line"><a name="l00197"></a><span class="lineno"> 197</span>  inputIndex = batchIdx * inputWidth * inputHeight * inputChannels +</div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  inputWidth * inputHeight * cInput +</div><div class="line"><a name="l00199"></a><span class="lineno"> 199</span>  inputWidth * (yInput - paddingTop) +</div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  xInput - paddingLeft;</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  }</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span> </div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span>  rInputDecoder[inputIndex];</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  inputValue = rInputDecoder.<a class="code" href="classarmnn_1_1_decoder.html#ac729108381e2340bea12877971713ecb">Get</a>();</div><div class="line"><a name="l00205"></a><span class="lineno"> 205</span>  }</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span> </div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span>  sum += filterValue * inputValue;</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>  }</div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>  }</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>  <span class="keywordflow">if</span> (biasEnabled)</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>  (*pBiasDecoder).SetIndex(cOutput, cOutput);</div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span>  sum += pBiasDecoder-><a class="code" href="classarmnn_1_1_decoder.html#ac729108381e2340bea12877971713ecb">Get</a>();</div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span>  }</div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span> </div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> outIdx = dataLayoutIndexed.<a class="code" href="classarmnn_utils_1_1_data_layout_indexed.html#a1e25d8623da985a43597b5756c73b206">GetIndex</a>(rOutputShape, batchIdx, cOutput, yOutput, xOutput);</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>  rOutputEncoder[outIdx];</div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  rOutputEncoder.<a class="code" href="classarmnn_1_1_encoder.html#ae3b62b846a9c239f332830b9e36030eb">Set</a>(sum);</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>  }</div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span>  }</div><div class="line"><a name="l00225"></a><span class="lineno"> 225</span>  }</div><div class="line"><a name="l00226"></a><span class="lineno"> 226</span> }</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> } <span class="comment">// namespace armnn</span></div><div class="ttc" id="classarmnn_utils_1_1_data_layout_indexed_html_a61c00316c443adc233c24e85c6c5b740"><div class="ttname"><a href="classarmnn_utils_1_1_data_layout_indexed.html#a61c00316c443adc233c24e85c6c5b740">armnnUtils::DataLayoutIndexed::GetHeightIndex</a></div><div class="ttdeci">unsigned int GetHeightIndex() const</div><div class="ttdef"><b>Definition:</b> <a href="_data_layout_indexed_8hpp_source.html#l00024">DataLayoutIndexed.hpp:24</a></div></div>
<div class="ttc" id="namespacearmnn_html_af98115cd07776d3fa8424434d2a7a897"><div class="ttname"><a href="namespacearmnn.html#af98115cd07776d3fa8424434d2a7a897">armnn::Convolve</a></div><div class="ttdeci">void Convolve(const TensorShape &rInputShape, Decoder< float > &rInputDecoder, const TensorShape &rOutputShape, Encoder< float > &rOutputEncoder, const TensorShape &rFilterShape, Decoder< float > &rFilterDecoder, bool biasEnabled, Decoder< float > *pBiasDecoder, DataLayout dataLayout, unsigned int paddingTop, unsigned int paddingLeft, unsigned int xStride, unsigned int yStride, unsigned int xDilation, unsigned int yDilation, bool depthwise)</div><div class="ttdef"><b>Definition:</b> <a href="_conv_impl_8cpp_source.html#l00071">ConvImpl.cpp:71</a></div></div>
<div class="ttc" id="classarmnn_1_1_decoder_html"><div class="ttname"><a href="classarmnn_1_1_decoder.html">armnn::Decoder< float ></a></div></div>
<div class="ttc" id="classarmnn_1_1_invalid_argument_exception_html"><div class="ttname"><a href="classarmnn_1_1_invalid_argument_exception.html">armnn::InvalidArgumentException</a></div><div class="ttdef"><b>Definition:</b> <a href="_exceptions_8hpp_source.html#l00080">Exceptions.hpp:80</a></div></div>
<div class="ttc" id="classarmnn_1_1_base_iterator_html_a1ec75b077d774dbfebf3662e8e4363c9"><div class="ttname"><a href="classarmnn_1_1_base_iterator.html#a1ec75b077d774dbfebf3662e8e4363c9">armnn::BaseIterator::SetIndex</a></div><div class="ttdeci">virtual BaseIterator & SetIndex(unsigned int index, unsigned int axisIndex=0)=0</div></div>
<div class="ttc" id="classarmnn_utils_1_1_data_layout_indexed_html_a414e6f95548e6f7a01d5028b55ad3941"><div class="ttname"><a href="classarmnn_utils_1_1_data_layout_indexed.html#a414e6f95548e6f7a01d5028b55ad3941">armnnUtils::DataLayoutIndexed::GetWidthIndex</a></div><div class="ttdeci">unsigned int GetWidthIndex() const</div><div class="ttdef"><b>Definition:</b> <a href="_data_layout_indexed_8hpp_source.html#l00025">DataLayoutIndexed.hpp:25</a></div></div>
<div class="ttc" id="_conv_impl_8hpp_html"><div class="ttname"><a href="_conv_impl_8hpp.html">ConvImpl.hpp</a></div></div>
<div class="ttc" id="classarmnn_1_1_encoder_html"><div class="ttname"><a href="classarmnn_1_1_encoder.html">armnn::Encoder< float ></a></div></div>
<div class="ttc" id="namespacearmnn_html"><div class="ttname"><a href="namespacearmnn.html">armnn</a></div><div class="ttdef"><b>Definition:</b> <a href="_backend_helper_8hpp_source.html#l00011">BackendHelper.hpp:11</a></div></div>
<div class="ttc" id="classarmnn_utils_1_1_data_layout_indexed_html_a861b2621ee46e4b63379988b360b8cd9"><div class="ttname"><a href="classarmnn_utils_1_1_data_layout_indexed.html#a861b2621ee46e4b63379988b360b8cd9">armnnUtils::DataLayoutIndexed::GetChannelsIndex</a></div><div class="ttdeci">unsigned int GetChannelsIndex() const</div><div class="ttdef"><b>Definition:</b> <a href="_data_layout_indexed_8hpp_source.html#l00023">DataLayoutIndexed.hpp:23</a></div></div>
<div class="ttc" id="namespacearmnn_html_ad1d5cce2d9e9a5d61c243e5c989112e0"><div class="ttname"><a href="namespacearmnn.html#ad1d5cce2d9e9a5d61c243e5c989112e0">armnn::DataLayout</a></div><div class="ttdeci">DataLayout</div><div class="ttdef"><b>Definition:</b> <a href="_types_8hpp_source.html#l00048">Types.hpp:48</a></div></div>
<div class="ttc" id="classarmnn_1_1_tensor_shape_html"><div class="ttname"><a href="classarmnn_1_1_tensor_shape.html">armnn::TensorShape</a></div><div class="ttdef"><b>Definition:</b> <a href="_tensor_8hpp_source.html#l00020">Tensor.hpp:20</a></div></div>
<div class="ttc" id="classarmnn_1_1_decoder_html_ac729108381e2340bea12877971713ecb"><div class="ttname"><a href="classarmnn_1_1_decoder.html#ac729108381e2340bea12877971713ecb">armnn::Decoder::Get</a></div><div class="ttdeci">virtual IType Get() const =0</div></div>
<div class="ttc" id="structarmnn_1_1_quantized_multiplier_smaller_than_one_html_a1ec4f749fbbdacbe2cc7e8895d5b65ed"><div class="ttname"><a href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#a1ec4f749fbbdacbe2cc7e8895d5b65ed">armnn::QuantizedMultiplierSmallerThanOne::QuantizedMultiplierSmallerThanOne</a></div><div class="ttdeci">QuantizedMultiplierSmallerThanOne(float multiplier)</div><div class="ttdef"><b>Definition:</b> <a href="_conv_impl_8cpp_source.html#l00016">ConvImpl.cpp:16</a></div></div>
<div class="ttc" id="classarmnn_utils_1_1_data_layout_indexed_html"><div class="ttname"><a href="classarmnn_utils_1_1_data_layout_indexed.html">armnnUtils::DataLayoutIndexed</a></div><div class="ttdef"><b>Definition:</b> <a href="_data_layout_indexed_8hpp_source.html#l00017">DataLayoutIndexed.hpp:17</a></div></div>
<div class="ttc" id="structarmnn_1_1_quantized_multiplier_smaller_than_one_html_ac04834893e568dfea4642ecb575faed8"><div class="ttname"><a href="structarmnn_1_1_quantized_multiplier_smaller_than_one.html#ac04834893e568dfea4642ecb575faed8">armnn::QuantizedMultiplierSmallerThanOne::operator*</a></div><div class="ttdeci">int32_t operator*(int32_t rhs) const</div><div class="ttdoc">The implementation of this function is adapted from Android NN&#39;s MultiplyByQuantizedMultiplierSmaller...</div><div class="ttdef"><b>Definition:</b> <a href="_conv_impl_8cpp_source.html#l00041">ConvImpl.cpp:41</a></div></div>
<div class="ttc" id="classarmnn_1_1_encoder_html_ae3b62b846a9c239f332830b9e36030eb"><div class="ttname"><a href="classarmnn_1_1_encoder.html#ae3b62b846a9c239f332830b9e36030eb">armnn::Encoder::Set</a></div><div class="ttdeci">virtual void Set(IType right)=0</div></div>
<div class="ttc" id="classarmnn_utils_1_1_data_layout_indexed_html_a1e25d8623da985a43597b5756c73b206"><div class="ttname"><a href="classarmnn_utils_1_1_data_layout_indexed.html#a1e25d8623da985a43597b5756c73b206">armnnUtils::DataLayoutIndexed::GetIndex</a></div><div class="ttdeci">unsigned int GetIndex(const armnn::TensorShape &shape, unsigned int batchIndex, unsigned int channelIndex, unsigned int heightIndex, unsigned int widthIndex) const</div><div class="ttdef"><b>Definition:</b> <a href="_data_layout_indexed_8hpp_source.html#l00027">DataLayoutIndexed.hpp:27</a></div></div>
<div class="ttc" id="namespacearmnn_html_ad1d5cce2d9e9a5d61c243e5c989112e0ad066db54b89b0912e7e7c6da51e2da51"><div class="ttname"><a href="namespacearmnn.html#ad1d5cce2d9e9a5d61c243e5c989112e0ad066db54b89b0912e7e7c6da51e2da51">armnn::DataLayout::NHWC</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.html">src</a></li><li class="navelem"><a class="el" href="dir_0f3cdec46afbc61a1ded8e1687c9c9a0.html">backends</a></li><li class="navelem"><a class="el" href="dir_efae4012d0e357ebeaba7d02491d70e5.html">reference</a></li><li class="navelem"><a class="el" href="dir_d2f3b8e2e64df3181ebe92efcc0a3012.html">workloads</a></li><li class="navelem"><a class="el" href="_conv_impl_8cpp.html">ConvImpl.cpp</a></li>
<li class="footer">Generated on Fri Mar 13 2020 16:06:58 for ArmNN by
<a href="http://www.doxygen.org/index.html">
<img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.13 </li>
</ul>
</div>
</body>
</html>
|
