-
Notifications
You must be signed in to change notification settings - Fork 14
/
Copy patha00044_source.html
124 lines (122 loc) · 34.6 KB
/
a00044_source.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
<!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 1.8.14"/>
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<title>tesseract: /usr/src/tesseract-ocr.master/src/arch/intsimdmatrix.h Source File</title>
<link href="tabs.css" rel="stylesheet" type="text/css"/>
<script type="text/javascript" src="jquery.js"></script>
<script type="text/javascript" src="dynsections.js"></script>
<link href="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>
<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">tesseract
 <span id="projectnumber">4.0.0-1-g2a2b</span>
</div>
</td>
</tr>
</tbody>
</table>
</div>
<!-- end header part -->
<!-- Generated by Doxygen 1.8.14 -->
<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('a00044_source.html','');});
/* @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">intsimdmatrix.h</div> </div>
</div><!--header-->
<div class="contents">
<a href="a00044.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">// File: intsimdmatrix.h</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment">// Description: Base class for 8-bit int SIMD matrix multipliers.</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span> <span class="comment">// Author: Ray Smith</span></div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span> <span class="comment">// Created: Tue Aug 15 07:37:20 PST 2017</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span> <span class="comment">//</span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span> <span class="comment">// (C) Copyright 2017, Google Inc.</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span> <span class="comment">// Licensed under the Apache License, Version 2.0 (the "License");</span></div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span> <span class="comment">// you may not use this file except in compliance with the License.</span></div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span> <span class="comment">// You may obtain a copy of the License at</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span> <span class="comment">// http://www.apache.org/licenses/LICENSE-2.0</span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span> <span class="comment">// Unless required by applicable law or agreed to in writing, software</span></div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span> <span class="comment">// distributed under the License is distributed on an "AS IS" BASIS,</span></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span> <span class="comment">// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.</span></div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span> <span class="comment">// See the License for the specific language governing permissions and</span></div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span> <span class="comment">// limitations under the License.</span></div><div class="line"><a name="l00018"></a><span class="lineno"> 18</span> <span class="comment"></span></div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span> <span class="preprocessor">#ifndef TESSERACT_ARCH_INTSIMDMATRIX_H_</span></div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span> <span class="preprocessor">#define TESSERACT_ARCH_INTSIMDMATRIX_H_</span></div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span> </div><div class="line"><a name="l00022"></a><span class="lineno"> 22</span> <span class="preprocessor">#include <cstdint></span></div><div class="line"><a name="l00023"></a><span class="lineno"> 23</span> <span class="preprocessor">#include <vector></span></div><div class="line"><a name="l00024"></a><span class="lineno"> 24</span> </div><div class="line"><a name="l00025"></a><span class="lineno"><a class="line" href="a02222.html"> 25</a></span> <span class="keyword">template</span> <<span class="keyword">class</span> T> <span class="keyword">class </span><a class="code" href="a02222.html">GENERIC_2D_ARRAY</a>;</div><div class="line"><a name="l00026"></a><span class="lineno"> 26</span> <span class="keyword">template</span> <<span class="keyword">typename</span> T> <span class="keyword">class </span><a class="code" href="a02182.html">GenericVector</a>;</div><div class="line"><a name="l00027"></a><span class="lineno"> 27</span> </div><div class="line"><a name="l00028"></a><span class="lineno"> 28</span> <span class="keyword">namespace </span><a class="code" href="a01629.html">tesseract</a> {</div><div class="line"><a name="l00029"></a><span class="lineno"> 29</span> </div><div class="line"><a name="l00030"></a><span class="lineno"> 30</span> <span class="comment">// Base class for a SIMD function to multiply a matrix by a vector, with sources</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span> <span class="comment">// of 8-bit signed integer, and result in a double, after appropriate scaling.</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span> <span class="comment">// Assumes a specific method of multiplication that can be applied to any size</span></div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span> <span class="comment">// and number of SIMD registers as follows:</span></div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span> <span class="comment">// int32_t results are computed with num_outputs_per_register_ in each of</span></div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span> <span class="comment">// max_output_registers_ result registers, repeatedly until it would make too</span></div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span> <span class="comment">// many results, then the number of registers is halved, and so-on down to a</span></div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span> <span class="comment">// single result register. The last calculation only outputs the required number</span></div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span> <span class="comment">// of results instead of writing beyond the bounds. Eg: matrix has 75 outputs,</span></div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span> <span class="comment">// num_outputs_per_register_ = 4, and max_output_registers_ = 8,</span></div><div class="line"><a name="l00040"></a><span class="lineno"> 40</span> <span class="comment">// Step 1: 8x4=32 results are computed,</span></div><div class="line"><a name="l00041"></a><span class="lineno"> 41</span> <span class="comment">// Step 2: 8x4=32 again, total 64,</span></div><div class="line"><a name="l00042"></a><span class="lineno"> 42</span> <span class="comment">// Step 3: 2x4=8 (since 8x4 is too many, so is 4x4), total 72,</span></div><div class="line"><a name="l00043"></a><span class="lineno"> 43</span> <span class="comment">// Step 4: 1x3, total 75.</span></div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span> <span class="comment">// Each step above is computed using a PartialFunc, which runs over the input</span></div><div class="line"><a name="l00045"></a><span class="lineno"> 45</span> <span class="comment">// vector once. The input is read one registerful of num_inputs_per_register_</span></div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span> <span class="comment">// at a time (presumably 4x num_outputs_per_register_ since they are int8_t)</span></div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span> <span class="comment">// so the inputs MUST BE PADDED to a multiple of num_inputs_per_register_.</span></div><div class="line"><a name="l00048"></a><span class="lineno"> 48</span> <span class="comment">// Since it is slow (on Intel at least) to horizontally add in a register,</span></div><div class="line"><a name="l00049"></a><span class="lineno"> 49</span> <span class="comment">// provision is made to process num_inputs_per_group_ inputs at a time, with</span></div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span> <span class="comment">// the group being replicated num_input_groups_ times and multiplied by a</span></div><div class="line"><a name="l00051"></a><span class="lineno"> 51</span> <span class="comment">// num_inputs_per_group_ by num_input_groups_ rectangle of the weights matrix.</span></div><div class="line"><a name="l00052"></a><span class="lineno"> 52</span> <span class="comment">// This is most convenient if num_inputs_per_group_ is 4, and the product</span></div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span> <span class="comment">// sign-extends and sums 8x8=16 bit results to 32 bits, adding 4 adjacent</span></div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span> <span class="comment">// results in the process, but it doesn't have to be implemented that way.</span></div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span> <span class="comment">// The weights are re-ordered by Init() to be used sequentially by the above</span></div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span> <span class="comment">// algorithm, followed by the biases, so they can be added at the end.</span></div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span> <span class="comment">// The base class computes the base C++ implementation.</span></div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span> <span class="comment">// NOTE that, although the subclasses execute on different SIMD hardware, no</span></div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span> <span class="comment">// virtual methods are needed, as the constructor sets up everything that</span></div><div class="line"><a name="l00060"></a><span class="lineno"> 60</span> <span class="comment">// is required to allow the base class implementation to do all the work.</span></div><div class="line"><a name="l00061"></a><span class="lineno"><a class="line" href="a02226.html"> 61</a></span> <span class="keyword">class </span><a class="code" href="a02226.html">IntSimdMatrix</a> {</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span>  <span class="keyword">public</span>:</div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span>  <span class="comment">// Constructor should set the data members to indicate the sizes.</span></div><div class="line"><a name="l00064"></a><span class="lineno"> 64</span>  <span class="comment">// NOTE: Base constructor public only for test purposes.</span></div><div class="line"><a name="l00065"></a><span class="lineno"><a class="line" href="a02226.html#a12c832ad24fd4c1581259b7a2743625d"> 65</a></span>  <a class="code" href="a02226.html#a12c832ad24fd4c1581259b7a2743625d">IntSimdMatrix</a>()</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  : <a class="code" href="a02226.html#a8760aa1d64a18e46cc77698f2381177c">num_outputs_per_register_</a>(1),</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span>  <a class="code" href="a02226.html#a6c071f8e8104b7f4ff744622fa21edc1">max_output_registers_</a>(1),</div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  <a class="code" href="a02226.html#ad17ed5688109967a28d46853f5babbab">num_inputs_per_register_</a>(1),</div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>  <a class="code" href="a02226.html#a23256c8836310012b8a80e61f9dd5a35">num_inputs_per_group_</a>(1),</div><div class="line"><a name="l00070"></a><span class="lineno"> 70</span>  <a class="code" href="a02226.html#a42ae28ff1dbacd64628913946684ca30">num_input_groups_</a>(1) {}</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>  <span class="comment">// Factory makes and returns an IntSimdMatrix (sub)class of the best</span></div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  <span class="comment">// available type for the current architecture.</span></div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  <span class="keyword">static</span> <a class="code" href="a02226.html">IntSimdMatrix</a>* <a class="code" href="a02226.html#a8b7dc48f425a7d108df0f5f489c92cb7">GetFastestMultiplier</a>();</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span> </div><div class="line"><a name="l00076"></a><span class="lineno"> 76</span>  <span class="comment">// Computes a reshaped copy of the weight matrix w. If there are no</span></div><div class="line"><a name="l00077"></a><span class="lineno"> 77</span>  <span class="comment">// partial_funcs_, it does nothing.</span></div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>  <span class="keywordtype">void</span> <a class="code" href="a02226.html#a48e759a483e6ce9574f0e05d7a7eb4b6">Init</a>(<span class="keyword">const</span> <a class="code" href="a02222.html">GENERIC_2D_ARRAY<int8_t></a>& w);</div><div class="line"><a name="l00079"></a><span class="lineno"> 79</span> </div><div class="line"><a name="l00080"></a><span class="lineno"> 80</span>  <span class="comment">// Rounds the size up to a multiple of the input register size (in int8_t).</span></div><div class="line"><a name="l00081"></a><span class="lineno"><a class="line" href="a02226.html#aa1f04190c56fc632d8da1a35caf496a5"> 81</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#aa1f04190c56fc632d8da1a35caf496a5">RoundInputs</a>(<span class="keywordtype">int</span> size)<span class="keyword"> const </span>{</div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  <span class="keywordflow">return</span> <a class="code" href="a02226.html#acf9f1a5093e7fa9de9c107b543404d0f">Roundup</a>(size, <a class="code" href="a02226.html#ad17ed5688109967a28d46853f5babbab">num_inputs_per_register_</a>);</div><div class="line"><a name="l00083"></a><span class="lineno"> 83</span>  }</div><div class="line"><a name="l00084"></a><span class="lineno"> 84</span>  <span class="comment">// Rounds the size up to a multiple of the output register size (in int32_t).</span></div><div class="line"><a name="l00085"></a><span class="lineno"><a class="line" href="a02226.html#a74a99972e32fe1659f839791bb7dde49"> 85</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#a74a99972e32fe1659f839791bb7dde49">RoundOutputs</a>(<span class="keywordtype">int</span> size)<span class="keyword"> const </span>{</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  <span class="keywordflow">return</span> <a class="code" href="a02226.html#acf9f1a5093e7fa9de9c107b543404d0f">Roundup</a>(size, <a class="code" href="a02226.html#a8760aa1d64a18e46cc77698f2381177c">num_outputs_per_register_</a>);</div><div class="line"><a name="l00087"></a><span class="lineno"> 87</span>  }</div><div class="line"><a name="l00088"></a><span class="lineno"> 88</span> </div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  <span class="comment">// Computes matrix.vector v = Wu.</span></div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span>  <span class="comment">// u is of size W.dim2() - 1 and the output v is of size W.dim1().</span></div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  <span class="comment">// u is imagined to have an extra element at the end with value 1, to</span></div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  <span class="comment">// implement the bias, but it doesn't actually have it.</span></div><div class="line"><a name="l00093"></a><span class="lineno"> 93</span>  <span class="comment">// Computes the base C++ implementation, if there are no partial_funcs_.</span></div><div class="line"><a name="l00094"></a><span class="lineno"> 94</span>  <span class="comment">// NOTE: The size of the input vector (u) must be padded using</span></div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  <span class="comment">// RoundInputs above.</span></div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span>  <span class="comment">// The input will be over-read to the extent of the padding. There are no</span></div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span>  <span class="comment">// alignment requirements.</span></div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>  <span class="keywordtype">void</span> <a class="code" href="a02226.html#a442dc784affc6b5b9c62d03997fee9e2">MatrixDotVector</a>(<span class="keyword">const</span> <a class="code" href="a02222.html">GENERIC_2D_ARRAY<int8_t></a>& w,</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span>  <span class="keyword">const</span> <a class="code" href="a02182.html">GenericVector<double></a>& scales, <span class="keyword">const</span> int8_t* u,</div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  <span class="keywordtype">double</span>* v) <span class="keyword">const</span>;</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="keyword">protected</span>:</div><div class="line"><a name="l00103"></a><span class="lineno"> 103</span>  <span class="comment">// Function to compute part of a matrix.vector multiplication. The weights</span></div><div class="line"><a name="l00104"></a><span class="lineno"> 104</span>  <span class="comment">// are in a very specific order (see above) in w, which is multiplied by</span></div><div class="line"><a name="l00105"></a><span class="lineno"> 105</span>  <span class="comment">// u of length num_in, to produce output v after scaling the integer results</span></div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span>  <span class="comment">// by the corresponding member of scales.</span></div><div class="line"><a name="l00107"></a><span class="lineno"> 107</span>  <span class="comment">// The amount of w and scales consumed is fixed and not available to the</span></div><div class="line"><a name="l00108"></a><span class="lineno"> 108</span>  <span class="comment">// caller. The number of outputs written to v will be at most num_out.</span></div><div class="line"><a name="l00109"></a><span class="lineno"><a class="line" href="a02226.html#ad2e7073d382c84f09f72621188e62456"> 109</a></span>  <span class="keyword">typedef</span> void (*<a class="code" href="a02226.html#ad2e7073d382c84f09f72621188e62456">PartialFunc</a>)(<span class="keyword">const</span> int8_t* w, <span class="keyword">const</span> <span class="keywordtype">double</span>* scales,</div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span>  <span class="keyword">const</span> int8_t* u, <span class="keywordtype">int</span> num_in, <span class="keywordtype">int</span> num_out,</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>  <span class="keywordtype">double</span>* v);</div><div class="line"><a name="l00112"></a><span class="lineno"> 112</span> </div><div class="line"><a name="l00113"></a><span class="lineno"> 113</span>  <span class="comment">// Rounds the input up to a multiple of the given factor.</span></div><div class="line"><a name="l00114"></a><span class="lineno"><a class="line" href="a02226.html#acf9f1a5093e7fa9de9c107b543404d0f"> 114</a></span>  <span class="keyword">static</span> <span class="keywordtype">int</span> <a class="code" href="a02226.html#acf9f1a5093e7fa9de9c107b543404d0f">Roundup</a>(<span class="keywordtype">int</span> input, <span class="keywordtype">int</span> factor) {</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  <span class="keywordflow">return</span> (input + factor - 1) / factor * factor;</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> </div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  <span class="comment">// Number of 32 bit outputs held in each register.</span></div><div class="line"><a name="l00119"></a><span class="lineno"><a class="line" href="a02226.html#a8760aa1d64a18e46cc77698f2381177c"> 119</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#a8760aa1d64a18e46cc77698f2381177c">num_outputs_per_register_</a>;</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  <span class="comment">// Maximum number of registers that we will use to hold outputs.</span></div><div class="line"><a name="l00121"></a><span class="lineno"><a class="line" href="a02226.html#a6c071f8e8104b7f4ff744622fa21edc1"> 121</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#a6c071f8e8104b7f4ff744622fa21edc1">max_output_registers_</a>;</div><div class="line"><a name="l00122"></a><span class="lineno"> 122</span>  <span class="comment">// Number of 8 bit inputs in the inputs register.</span></div><div class="line"><a name="l00123"></a><span class="lineno"><a class="line" href="a02226.html#ad17ed5688109967a28d46853f5babbab"> 123</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#ad17ed5688109967a28d46853f5babbab">num_inputs_per_register_</a>;</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>  <span class="comment">// Number of inputs in each weight group.</span></div><div class="line"><a name="l00125"></a><span class="lineno"><a class="line" href="a02226.html#a23256c8836310012b8a80e61f9dd5a35"> 125</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#a23256c8836310012b8a80e61f9dd5a35">num_inputs_per_group_</a>;</div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <span class="comment">// Number of groups of inputs to be broadcast.</span></div><div class="line"><a name="l00127"></a><span class="lineno"><a class="line" href="a02226.html#a42ae28ff1dbacd64628913946684ca30"> 127</a></span>  <span class="keywordtype">int</span> <a class="code" href="a02226.html#a42ae28ff1dbacd64628913946684ca30">num_input_groups_</a>;</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>  <span class="comment">// The weights matrix reorganized in whatever way suits this instance.</span></div><div class="line"><a name="l00129"></a><span class="lineno"><a class="line" href="a02226.html#a6de2b2f60bec1231ebebee433ed900f7"> 129</a></span>  std::vector<int8_t> <a class="code" href="a02226.html#a6de2b2f60bec1231ebebee433ed900f7">shaped_w_</a>;</div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span>  <span class="comment">// A series of functions to compute a partial result.</span></div><div class="line"><a name="l00131"></a><span class="lineno"><a class="line" href="a02226.html#a1fcea4ce2c16453174e3386ebd81fdf1"> 131</a></span>  std::vector<PartialFunc> <a class="code" href="a02226.html#a1fcea4ce2c16453174e3386ebd81fdf1">partial_funcs_</a>;</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> </div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span> } <span class="comment">// namespace tesseract</span></div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span> </div><div class="line"><a name="l00136"></a><span class="lineno"> 136</span> <span class="preprocessor">#endif // TESSERACT_ARCH_INTSIMDMATRIX_H_</span></div><div class="ttc" id="a02226_html_a6c071f8e8104b7f4ff744622fa21edc1"><div class="ttname"><a href="a02226.html#a6c071f8e8104b7f4ff744622fa21edc1">tesseract::IntSimdMatrix::max_output_registers_</a></div><div class="ttdeci">int max_output_registers_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00121">intsimdmatrix.h:121</a></div></div>
<div class="ttc" id="a02226_html_a1fcea4ce2c16453174e3386ebd81fdf1"><div class="ttname"><a href="a02226.html#a1fcea4ce2c16453174e3386ebd81fdf1">tesseract::IntSimdMatrix::partial_funcs_</a></div><div class="ttdeci">std::vector< PartialFunc > partial_funcs_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00131">intsimdmatrix.h:131</a></div></div>
<div class="ttc" id="a02226_html_a8760aa1d64a18e46cc77698f2381177c"><div class="ttname"><a href="a02226.html#a8760aa1d64a18e46cc77698f2381177c">tesseract::IntSimdMatrix::num_outputs_per_register_</a></div><div class="ttdeci">int num_outputs_per_register_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00119">intsimdmatrix.h:119</a></div></div>
<div class="ttc" id="a02222_html"><div class="ttname"><a href="a02222.html">GENERIC_2D_ARRAY</a></div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00025">intsimdmatrix.h:25</a></div></div>
<div class="ttc" id="a02226_html_ad17ed5688109967a28d46853f5babbab"><div class="ttname"><a href="a02226.html#ad17ed5688109967a28d46853f5babbab">tesseract::IntSimdMatrix::num_inputs_per_register_</a></div><div class="ttdeci">int num_inputs_per_register_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00123">intsimdmatrix.h:123</a></div></div>
<div class="ttc" id="a02226_html_ad2e7073d382c84f09f72621188e62456"><div class="ttname"><a href="a02226.html#ad2e7073d382c84f09f72621188e62456">tesseract::IntSimdMatrix::PartialFunc</a></div><div class="ttdeci">void(* PartialFunc)(const int8_t *w, const double *scales, const int8_t *u, int num_in, int num_out, double *v)</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00109">intsimdmatrix.h:109</a></div></div>
<div class="ttc" id="a02226_html_a8b7dc48f425a7d108df0f5f489c92cb7"><div class="ttname"><a href="a02226.html#a8b7dc48f425a7d108df0f5f489c92cb7">tesseract::IntSimdMatrix::GetFastestMultiplier</a></div><div class="ttdeci">static IntSimdMatrix * GetFastestMultiplier()</div><div class="ttdef"><b>Definition:</b> <a href="a00041_source.html#l00031">intsimdmatrix.cpp:31</a></div></div>
<div class="ttc" id="a02226_html_a23256c8836310012b8a80e61f9dd5a35"><div class="ttname"><a href="a02226.html#a23256c8836310012b8a80e61f9dd5a35">tesseract::IntSimdMatrix::num_inputs_per_group_</a></div><div class="ttdeci">int num_inputs_per_group_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00125">intsimdmatrix.h:125</a></div></div>
<div class="ttc" id="a02226_html_aa1f04190c56fc632d8da1a35caf496a5"><div class="ttname"><a href="a02226.html#aa1f04190c56fc632d8da1a35caf496a5">tesseract::IntSimdMatrix::RoundInputs</a></div><div class="ttdeci">int RoundInputs(int size) const</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00081">intsimdmatrix.h:81</a></div></div>
<div class="ttc" id="a02226_html_a48e759a483e6ce9574f0e05d7a7eb4b6"><div class="ttname"><a href="a02226.html#a48e759a483e6ce9574f0e05d7a7eb4b6">tesseract::IntSimdMatrix::Init</a></div><div class="ttdeci">void Init(const GENERIC_2D_ARRAY< int8_t > &w)</div><div class="ttdef"><b>Definition:</b> <a href="a00041_source.html#l00046">intsimdmatrix.cpp:46</a></div></div>
<div class="ttc" id="a02226_html_a42ae28ff1dbacd64628913946684ca30"><div class="ttname"><a href="a02226.html#a42ae28ff1dbacd64628913946684ca30">tesseract::IntSimdMatrix::num_input_groups_</a></div><div class="ttdeci">int num_input_groups_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00127">intsimdmatrix.h:127</a></div></div>
<div class="ttc" id="a02226_html_a442dc784affc6b5b9c62d03997fee9e2"><div class="ttname"><a href="a02226.html#a442dc784affc6b5b9c62d03997fee9e2">tesseract::IntSimdMatrix::MatrixDotVector</a></div><div class="ttdeci">void MatrixDotVector(const GENERIC_2D_ARRAY< int8_t > &w, const GenericVector< double > &scales, const int8_t *u, double *v) const</div><div class="ttdef"><b>Definition:</b> <a href="a00041_source.html#l00096">intsimdmatrix.cpp:96</a></div></div>
<div class="ttc" id="a02226_html_a74a99972e32fe1659f839791bb7dde49"><div class="ttname"><a href="a02226.html#a74a99972e32fe1659f839791bb7dde49">tesseract::IntSimdMatrix::RoundOutputs</a></div><div class="ttdeci">int RoundOutputs(int size) const</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00085">intsimdmatrix.h:85</a></div></div>
<div class="ttc" id="a02182_html"><div class="ttname"><a href="a02182.html">GenericVector</a></div><div class="ttdef"><b>Definition:</b> <a href="a00008_source.html#l00037">baseapi.h:37</a></div></div>
<div class="ttc" id="a02226_html_acf9f1a5093e7fa9de9c107b543404d0f"><div class="ttname"><a href="a02226.html#acf9f1a5093e7fa9de9c107b543404d0f">tesseract::IntSimdMatrix::Roundup</a></div><div class="ttdeci">static int Roundup(int input, int factor)</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00114">intsimdmatrix.h:114</a></div></div>
<div class="ttc" id="a02226_html_a6de2b2f60bec1231ebebee433ed900f7"><div class="ttname"><a href="a02226.html#a6de2b2f60bec1231ebebee433ed900f7">tesseract::IntSimdMatrix::shaped_w_</a></div><div class="ttdeci">std::vector< int8_t > shaped_w_</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00129">intsimdmatrix.h:129</a></div></div>
<div class="ttc" id="a02226_html"><div class="ttname"><a href="a02226.html">tesseract::IntSimdMatrix</a></div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00061">intsimdmatrix.h:61</a></div></div>
<div class="ttc" id="a01629_html"><div class="ttname"><a href="a01629.html">tesseract</a></div><div class="ttdef"><b>Definition:</b> <a href="a00005_source.html#l00094">baseapi.cpp:94</a></div></div>
<div class="ttc" id="a02226_html_a12c832ad24fd4c1581259b7a2743625d"><div class="ttname"><a href="a02226.html#a12c832ad24fd4c1581259b7a2743625d">tesseract::IntSimdMatrix::IntSimdMatrix</a></div><div class="ttdeci">IntSimdMatrix()</div><div class="ttdef"><b>Definition:</b> <a href="a00044_source.html#l00065">intsimdmatrix.h:65</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_fce9a394c0d2c636e610a0a34fe30580.html">tesseract-ocr.master</a></li><li class="navelem"><a class="el" href="dir_fd2783e80b2d56815818e17a68fc4d98.html">src</a></li><li class="navelem"><a class="el" href="dir_112254772d44b3d8efe7783321f69b12.html">arch</a></li><li class="navelem"><a class="el" href="a00044.html">intsimdmatrix.h</a></li>
<li class="footer">Generated on Mon Oct 29 2018 11:03:42 for tesseract by
<a href="http://www.doxygen.org/index.html">
<img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.14 </li>
</ul>
</div>
</body>
</html>