arm_compute v19.08
diff --git a/documentation/gemm__hybrid_8hpp_source.xhtml b/documentation/gemm__hybrid_8hpp_source.xhtml
index 01aae0b..2e24cba 100644
--- a/documentation/gemm__hybrid_8hpp_source.xhtml
+++ b/documentation/gemm__hybrid_8hpp_source.xhtml
@@ -40,7 +40,7 @@
<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">19.05</span>
+  <span id="projectnumber">19.08</span>
</div>
</td>
</tr>
@@ -100,7 +100,7 @@
<div class="title">gemm_hybrid.hpp</div> </div>
</div><!--header-->
<div class="contents">
-<a href="gemm__hybrid_8hpp.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">#pragma once</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 <assert.h></span></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="preprocessor">#include <algorithm></span></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="preprocessor">#include "arm_gemm.hpp"</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span> <span class="preprocessor">#include "<a class="code" href="ndrange_8hpp.xhtml">ndrange.hpp</a>"</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span> <span class="preprocessor">#include "<a class="code" href="utils_8hpp.xhtml">utils.hpp</a>"</span></div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span> </div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span> <span class="preprocessor">#include "<a class="code" href="mergeresults_8hpp.xhtml">mergeresults.hpp</a>"</span></div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span> <span class="preprocessor">#include "<a class="code" href="transform_8hpp.xhtml">transform.hpp</a>"</span></div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span> </div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span> <span class="preprocessor">#include "<a class="code" href="profiler_8hpp.xhtml">profiler.hpp</a>"</span></div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span> <span class="preprocessor">#endif</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"> 41</span> <span class="keyword">namespace </span><a class="code" href="namespacearm__gemm.xhtml">arm_gemm</a> {</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="comment">// Implementation of the GemmCommon abstract class.</span></div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span> <span class="keyword">template</span><<span class="keyword">typename</span> strategy, <span class="keyword">typename</span> To, <span class="keyword">typename</span> Tr></div><div class="line"><a name="l00045"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml"> 45</a></span> <span class="keyword">class </span><a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> : <span class="keyword">public</span> GemmCommon<To, Tr> {</div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span>  <span class="keyword">typedef</span> <span class="keyword">typename</span> strategy::operand_type Toi;</div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span>  <span class="keyword">typedef</span> <span class="keyword">typename</span> strategy::result_type Tri;</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">/* const properties set by constructor */</span></div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span>  <span class="keyword">const</span> CPUInfo * <span class="keyword">const</span> _ci;</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="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Msize;</div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Nsize;</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Ksize;</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span> </div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _nbatches;</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _nmulti;</div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span> </div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> _trB;</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>  <span class="keyword">const</span> Tr _beta;</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span> </div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span>  <span class="comment">/* Blocking info */</span></div><div class="line"><a name="l00064"></a><span class="lineno"> 64</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _k_block;</div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _n_block;</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Mround;</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span> </div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  <span class="comment">/* Pretransposed buffer. */</span></div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>  <span class="keyword">const</span> Toi *_B_transposed=<span class="keyword">nullptr</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"> 71</span>  <span class="keyword">const</span> <a class="code" href="classarm__gemm_1_1_n_d_range.xhtml">NDRange<4></a> _window_range;</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span> </div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> compute_k_block(<span class="keyword">const</span> GemmArgs<Tr> &args) {</div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  <span class="keywordflow">if</span> (args._cfg && args._cfg->inner_block_size) {</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span>  <span class="keywordflow">return</span> args._cfg->inner_block_size;</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> </div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> L1_size = args._ci->get_L1_cache_size();</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">// k_block: Find out how much of the larger array can be loaded into half the cache.</span></div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>  <span class="comment">// This should account for associative caches.</span></div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_block = (L1_size / 2) / (<span class="keyword">sizeof</span>(Toi) * (std::max(strategy::out_width(), strategy::out_height())));</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">// Needs to be (at least a single) multiple of the K unroll level.</span></div><div class="line"><a name="l00085"></a><span class="lineno"> 85</span>  k_block /= strategy::k_unroll();</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  k_block = std::max(k_block, 1U) * strategy::k_unroll();</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="comment">// Now tune to presented problem size; this is how many blocks we need.</span></div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> numk_blocks = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Ksize, k_block);</div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span> </div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  <span class="comment">// So divide the space equally into that many blocks.</span></div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  k_block = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Ksize, numk_blocks);</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="comment">// And round UP to the K unroll level required.</span></div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  k_block = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(k_block, strategy::k_unroll());</div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span> </div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span>  <span class="keywordflow">return</span> k_block;</div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>  }</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span> </div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> compute_n_block(<span class="keyword">const</span> GemmArgs<Tr> &args) {</div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>  <span class="keywordflow">if</span> (args._cfg && args._cfg->outer_block_size) {</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>  <span class="keywordflow">return</span> args._cfg->outer_block_size;</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>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_block = compute_k_block(args);</div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> L2_size = args._ci->get_L2_cache_size();</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="comment">// n_block: Work out how many rows (of length k_block) will fit in the L2</span></div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>  <span class="comment">// Don't allocate more than 90% of the L2 to allow for overheads, and subtract off the L1 contents.</span></div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> n_block = (((L2_size * 9) / 10) - (k_block * <span class="keyword">sizeof</span>(Toi) * (strategy::out_width() + strategy::out_height()))) /</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>  (<span class="keyword">sizeof</span>(Toi) * k_block);</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">// Needs to be (at least a single) multiple of the kernel output width.</span></div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>  n_block /= strategy::out_width();</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  n_block = std::max(n_block, 1U) * strategy::out_width();</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">// And tune to the presented problem size.</span></div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> numblocks = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Nsize, n_block);</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  n_block = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Nsize, numblocks);</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  n_block = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(n_block, strategy::out_width());</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="keywordflow">return</span> n_block;</div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  }</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span> </div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span> <span class="keyword">public</span>:</div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a67f91182afb2b1092518637116562440">GemmHybrid</a>(<a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> &) = <span class="keyword">delete</span>;</div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> & <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a82ec71716669b6765670e0ed242f5a47">operator= </a>(<a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> &) = <span class="keyword">delete</span>;</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>  <span class="comment">/* Constructor */</span></div><div class="line"><a name="l00130"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9e9f8c1982f7144528bb6d01af482a13"> 130</a></span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9e9f8c1982f7144528bb6d01af482a13">GemmHybrid</a>(<span class="keyword">const</span> GemmArgs<Tr> &args)</div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>  : _ci(args._ci), _Msize(args._Msize), _Nsize(args._Nsize), _Ksize(args._Ksize),</div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  _nbatches(args._nbatches), _nmulti(args._nmulti), _trB(args._trB), _beta(args._beta),</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  _k_block(compute_k_block(args)), _n_block(compute_n_block(args)),</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span>  _Mround(<a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(args._Msize, strategy::out_height())),</div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>  _window_range(<a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Msize, strategy::out_height()), _nbatches, <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(_Nsize, _n_block), _nmulti) { }</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">// Interface implementation - Compulsory functions</span></div><div class="line"><a name="l00138"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9fe814165d032363ea4f9cfdd36dd5e2"> 138</a></span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9fe814165d032363ea4f9cfdd36dd5e2">get_window_size</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span>  <span class="keywordflow">return</span> _window_range.<a class="code" href="classarm__gemm_1_1_n_d_range.xhtml#a2ee55bf0e2d8324a1360f6994a762e28">total_size</a>();</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> </div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span>  <span class="comment">// This kernel can always be dynamically scheduled.</span></div><div class="line"><a name="l00143"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a1bb76b8a32e09fb5ffffd66e04d7f741"> 143</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a1bb76b8a32e09fb5ffffd66e04d7f741">supports_dynamic_scheduling</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</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> </div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  <span class="comment">// Execute</span></div><div class="line"><a name="l00148"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560"> 148</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560">execute</a>(<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> start, <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> end, <span class="keywordtype">int</span> threadid)<span class="keyword"> override </span>{</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span>  profiler prof;</div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span>  strategy strat(_ci);</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>  <span class="comment">/* Make sure we've been set up correctly. */</span></div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span>  assert(_B_transposed);</div><div class="line"><a name="l00156"></a><span class="lineno"> 156</span>  static_assert(std::is_same<To, Toi>::value, <span class="stringliteral">"gemm_native: Operand types must be the same."</span>);</div><div class="line"><a name="l00157"></a><span class="lineno"> 157</span>  static_assert(std::is_same<Tr, Tri>::value, <span class="stringliteral">"gemm_native: Result types must be the same."</span>);</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="comment">/* For now, each work item implies all the K for a given output</span></div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span> <span class="comment"> * pixel (so we don't need to synchronize access to the output</span></div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span> <span class="comment"> * array). So separate the loop over K blocks here. */</span></div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k0=0; k0<_Ksize; k0+=_k_block) {</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kmax = std::min(k0 + _k_block, _Ksize);</div><div class="line"><a name="l00164"></a><span class="lineno"> 164</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kern_k = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(kmax-k0, strategy::k_unroll());</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">auto</span> p = _window_range.<a class="code" href="classarm__gemm_1_1_n_d_range.xhtml#a78a5c697e8717cb27954513e8ad2a508">iterator</a>(start, end);</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>  <span class="keywordflow">if</span> (p.done()) {</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>  <span class="keywordflow">return</span>;</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> </div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  <span class="keywordflow">do</span> {</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> m_start = p.dim(0) * strategy::out_height();</div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> m_end = std::min(p.dim0_max() * strategy::out_height(), _Msize);</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> batch = p.dim(1);</div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> n0 = p.dim(2) * _n_block;</div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> nmax = std::min(n0 + _n_block, _Nsize);</div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> multi = p.dim(3);</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>  <span class="keyword">const</span> Toi *b_panel = _B_transposed +</div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  (multi * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width()) * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Ksize, strategy::k_unroll())) +</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>  (k0 * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width())) +</div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>  (n0 * kern_k);</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> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>  <span class="keyword">auto</span> p = prof.ScopedProfiler(<a class="code" href="profiler_8hpp.xhtml#a97f4e398174b5e094f5262e5d785ac20">PROFILE_KERNEL</a>, (m_end - m_start) * kern_k * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(nmax-n0, strategy::out_width()));</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span> </div><div class="line"><a name="l00189"></a><span class="lineno"> 189</span>  strat.kernel(this->_Aptr + (multi * this->_A_multi_stride) + (batch * this->_A_batch_stride) + (m_start * this->_lda) + k0, this->_lda,</div><div class="line"><a name="l00190"></a><span class="lineno"> 190</span>  b_panel,</div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  this->_Cptr + (multi * this->_C_multi_stride) + (batch * this->_C_batch_stride) + (m_start * this->_ldc) + n0, this->_ldc,</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  (k0 == 0) ? _beta : static_cast<Tr>(1),</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  (m_end - m_start), (nmax - n0), kern_k);</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span>  } <span class="keywordflow">while</span> (p.next_dim1());</div><div class="line"><a name="l00195"></a><span class="lineno"> 195</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> </div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  <span class="comment">// Interface implementation - pretransposed</span></div><div class="line"><a name="l00199"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ac5dec5063a75604fdc60555f15577b94"> 199</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ac5dec5063a75604fdc60555f15577b94">B_is_pretransposed</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  }</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span> </div><div class="line"><a name="l00203"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#aa7cfff39cbb2be65cd40042a75e5ae1c"> 203</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#aa7cfff39cbb2be65cd40042a75e5ae1c">B_pretranspose_required</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  <span class="keywordflow">return</span> (_B_transposed==<span class="keyword">nullptr</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> </div><div class="line"><a name="l00207"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae9fc7380e533c43a1d91e48a7578e412"> 207</a></span>  <span class="keywordtype">size_t</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae9fc7380e533c43a1d91e48a7578e412">get_B_pretransposed_array_size</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>  <span class="keywordflow">return</span> <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width()) * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Ksize, strategy::k_unroll()) * _nmulti * <span class="keyword">sizeof</span>(Toi);</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"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae385dce1a98e03f0a3ca311f031e5110"> 211</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae385dce1a98e03f0a3ca311f031e5110">pretranspose_B_array</a>(<span class="keywordtype">void</span> *in_buffer, <span class="keyword">const</span> To *B, <span class="keyword">const</span> <span class="keywordtype">int</span> ldb, <span class="keyword">const</span> <span class="keywordtype">int</span> B_multi_stride)<span class="keyword"> override </span>{</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>  Toi *buffer = reinterpret_cast<Toi *>(in_buffer);</div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span>  _B_transposed = buffer;</div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span>  strategy strat(_ci);</div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span> </div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> multi=0; multi<_nmulti; multi++) {</div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k0=0; k0<_Ksize; k0+=_k_block) {</div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kmax = std::min(k0 + _k_block, _Ksize);</div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_size = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(kmax-k0, strategy::k_unroll());</div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span> </div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> x0=0; x0<_Nsize; x0+=_n_block) {</div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xmax = std::min(x0+_n_block, _Nsize);</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>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> size = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(xmax-x0, strategy::out_width()) * k_size;</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>  strat.transforms.PrepareB( buffer, B + (multi * B_multi_stride), ldb,</div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  x0, xmax, k0, kmax, _trB);</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span> </div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  buffer += size;</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span>  }</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>  }</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"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a6e76d099f51fc0ef7f48dbd7d00214ac"> 235</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a6e76d099f51fc0ef7f48dbd7d00214ac">set_pretransposed_B_data</a>(<span class="keywordtype">void</span> *in_buffer)<span class="keyword"> override </span>{</div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span>  _B_transposed = reinterpret_cast<Toi *>(in_buffer);</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> };</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="comment">// namespace arm_gemm</span></div><div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_a0b21ce5febc3f7eed9c7b37c0b9a3560"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560">arm_gemm::GemmHybrid::execute</a></div><div class="ttdeci">void execute(unsigned int start, unsigned int end, int threadid) override</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00148">gemm_hybrid.hpp:148</a></div></div>
+<a href="gemm__hybrid_8hpp.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">#pragma once</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 <assert.h></span></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="preprocessor">#include <algorithm></span></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="preprocessor">#include "arm_gemm.hpp"</span></div><div class="line"><a name="l00031"></a><span class="lineno"> 31</span> <span class="preprocessor">#include "<a class="code" href="ndrange_8hpp.xhtml">ndrange.hpp</a>"</span></div><div class="line"><a name="l00032"></a><span class="lineno"> 32</span> <span class="preprocessor">#include "<a class="code" href="utils_8hpp.xhtml">utils.hpp</a>"</span></div><div class="line"><a name="l00033"></a><span class="lineno"> 33</span> </div><div class="line"><a name="l00034"></a><span class="lineno"> 34</span> <span class="preprocessor">#include "<a class="code" href="mergeresults_8hpp.xhtml">mergeresults.hpp</a>"</span></div><div class="line"><a name="l00035"></a><span class="lineno"> 35</span> <span class="preprocessor">#include "<a class="code" href="transform_8hpp.xhtml">transform.hpp</a>"</span></div><div class="line"><a name="l00036"></a><span class="lineno"> 36</span> </div><div class="line"><a name="l00037"></a><span class="lineno"> 37</span> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00038"></a><span class="lineno"> 38</span> <span class="preprocessor">#include "<a class="code" href="profiler_8hpp.xhtml">profiler.hpp</a>"</span></div><div class="line"><a name="l00039"></a><span class="lineno"> 39</span> <span class="preprocessor">#endif</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"> 41</span> <span class="keyword">namespace </span><a class="code" href="namespacearm__gemm.xhtml">arm_gemm</a> {</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="comment">// Implementation of the GemmCommon abstract class.</span></div><div class="line"><a name="l00044"></a><span class="lineno"> 44</span> <span class="keyword">template</span><<span class="keyword">typename</span> strategy, <span class="keyword">typename</span> To, <span class="keyword">typename</span> Tr></div><div class="line"><a name="l00045"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml"> 45</a></span> <span class="keyword">class </span><a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> : <span class="keyword">public</span> GemmCommon<To, Tr> {</div><div class="line"><a name="l00046"></a><span class="lineno"> 46</span>  <span class="keyword">typedef</span> <span class="keyword">typename</span> strategy::operand_type Toi;</div><div class="line"><a name="l00047"></a><span class="lineno"> 47</span>  <span class="keyword">typedef</span> <span class="keyword">typename</span> strategy::result_type Tri;</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">/* const properties set by constructor */</span></div><div class="line"><a name="l00050"></a><span class="lineno"> 50</span>  <span class="keyword">const</span> CPUInfo * <span class="keyword">const</span> _ci;</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="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Msize;</div><div class="line"><a name="l00053"></a><span class="lineno"> 53</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Nsize;</div><div class="line"><a name="l00054"></a><span class="lineno"> 54</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Ksize;</div><div class="line"><a name="l00055"></a><span class="lineno"> 55</span> </div><div class="line"><a name="l00056"></a><span class="lineno"> 56</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _nbatches;</div><div class="line"><a name="l00057"></a><span class="lineno"> 57</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _nmulti;</div><div class="line"><a name="l00058"></a><span class="lineno"> 58</span> </div><div class="line"><a name="l00059"></a><span class="lineno"> 59</span>  <span class="keyword">const</span> <span class="keywordtype">bool</span> _trB;</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>  <span class="keyword">const</span> Tr _beta;</div><div class="line"><a name="l00062"></a><span class="lineno"> 62</span> </div><div class="line"><a name="l00063"></a><span class="lineno"> 63</span>  <span class="comment">/* Blocking info */</span></div><div class="line"><a name="l00064"></a><span class="lineno"> 64</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _k_block;</div><div class="line"><a name="l00065"></a><span class="lineno"> 65</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _n_block;</div><div class="line"><a name="l00066"></a><span class="lineno"> 66</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> _Mround;</div><div class="line"><a name="l00067"></a><span class="lineno"> 67</span> </div><div class="line"><a name="l00068"></a><span class="lineno"> 68</span>  <span class="comment">/* Pretransposed buffer. */</span></div><div class="line"><a name="l00069"></a><span class="lineno"> 69</span>  <span class="keyword">const</span> Toi *_B_transposed=<span class="keyword">nullptr</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"> 71</span>  <span class="keyword">const</span> <a class="code" href="classarm__gemm_1_1_n_d_range.xhtml">NDRange<4></a> _window_range;</div><div class="line"><a name="l00072"></a><span class="lineno"> 72</span> </div><div class="line"><a name="l00073"></a><span class="lineno"> 73</span>  <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> compute_k_block(<span class="keyword">const</span> GemmArgs<Tr> &args) {</div><div class="line"><a name="l00074"></a><span class="lineno"> 74</span>  <span class="keywordflow">if</span> (args._cfg && args._cfg->inner_block_size) {</div><div class="line"><a name="l00075"></a><span class="lineno"> 75</span>  <span class="keywordflow">return</span> args._cfg->inner_block_size;</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> </div><div class="line"><a name="l00078"></a><span class="lineno"> 78</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> L1_size = args._ci->get_L1_cache_size();</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">// k_block: Find out how much of the larger array can be loaded into half the cache.</span></div><div class="line"><a name="l00081"></a><span class="lineno"> 81</span>  <span class="comment">// This should account for associative caches.</span></div><div class="line"><a name="l00082"></a><span class="lineno"> 82</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_block = (L1_size / 2) / (<span class="keyword">sizeof</span>(Toi) * (std::max(strategy::out_width(), strategy::out_height())));</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">// Needs to be (at least a single) multiple of the K unroll level.</span></div><div class="line"><a name="l00085"></a><span class="lineno"> 85</span>  k_block /= strategy::k_unroll();</div><div class="line"><a name="l00086"></a><span class="lineno"> 86</span>  k_block = std::max(k_block, 1U) * strategy::k_unroll();</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="comment">// Now tune to presented problem size; this is how many blocks we need.</span></div><div class="line"><a name="l00089"></a><span class="lineno"> 89</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> numk_blocks = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Ksize, k_block);</div><div class="line"><a name="l00090"></a><span class="lineno"> 90</span> </div><div class="line"><a name="l00091"></a><span class="lineno"> 91</span>  <span class="comment">// So divide the space equally into that many blocks.</span></div><div class="line"><a name="l00092"></a><span class="lineno"> 92</span>  k_block = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Ksize, numk_blocks);</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="comment">// And round UP to the K unroll level required.</span></div><div class="line"><a name="l00095"></a><span class="lineno"> 95</span>  k_block = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(k_block, strategy::k_unroll());</div><div class="line"><a name="l00096"></a><span class="lineno"> 96</span> </div><div class="line"><a name="l00097"></a><span class="lineno"> 97</span>  <span class="keywordflow">return</span> k_block;</div><div class="line"><a name="l00098"></a><span class="lineno"> 98</span>  }</div><div class="line"><a name="l00099"></a><span class="lineno"> 99</span> </div><div class="line"><a name="l00100"></a><span class="lineno"> 100</span>  <span class="keyword">static</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> compute_n_block(<span class="keyword">const</span> GemmArgs<Tr> &args) {</div><div class="line"><a name="l00101"></a><span class="lineno"> 101</span>  <span class="keywordflow">if</span> (args._cfg && args._cfg->outer_block_size) {</div><div class="line"><a name="l00102"></a><span class="lineno"> 102</span>  <span class="keywordflow">return</span> args._cfg->outer_block_size;</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>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_block = compute_k_block(args);</div><div class="line"><a name="l00106"></a><span class="lineno"> 106</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> L2_size = args._ci->get_L2_cache_size();</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="comment">// n_block: Work out how many rows (of length k_block) will fit in the L2</span></div><div class="line"><a name="l00109"></a><span class="lineno"> 109</span>  <span class="comment">// Don't allocate more than 90% of the L2 to allow for overheads, and subtract off the L1 contents.</span></div><div class="line"><a name="l00110"></a><span class="lineno"> 110</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> n_block = (((L2_size * 9) / 10) - (k_block * <span class="keyword">sizeof</span>(Toi) * (strategy::out_width() + strategy::out_height()))) /</div><div class="line"><a name="l00111"></a><span class="lineno"> 111</span>  (<span class="keyword">sizeof</span>(Toi) * k_block);</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">// Needs to be (at least a single) multiple of the kernel output width.</span></div><div class="line"><a name="l00114"></a><span class="lineno"> 114</span>  n_block /= strategy::out_width();</div><div class="line"><a name="l00115"></a><span class="lineno"> 115</span>  n_block = std::max(n_block, 1U) * strategy::out_width();</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">// And tune to the presented problem size.</span></div><div class="line"><a name="l00118"></a><span class="lineno"> 118</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> numblocks = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Nsize, n_block);</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  n_block = <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Nsize, numblocks);</div><div class="line"><a name="l00120"></a><span class="lineno"> 120</span>  n_block = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(n_block, strategy::out_width());</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="keywordflow">return</span> n_block;</div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  }</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span> </div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span> <span class="keyword">public</span>:</div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a67f91182afb2b1092518637116562440">GemmHybrid</a>(<a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> &) = <span class="keyword">delete</span>;</div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> & <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a82ec71716669b6765670e0ed242f5a47">operator= </a>(<a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml">GemmHybrid</a> &) = <span class="keyword">delete</span>;</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>  <span class="comment">/* Constructor */</span></div><div class="line"><a name="l00130"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9e9f8c1982f7144528bb6d01af482a13"> 130</a></span>  <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9e9f8c1982f7144528bb6d01af482a13">GemmHybrid</a>(<span class="keyword">const</span> GemmArgs<Tr> &args)</div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>  : _ci(args._ci), _Msize(args._Msize), _Nsize(args._Nsize), _Ksize(args._Ksize),</div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  _nbatches(args._nbatches), _nmulti(args._nmulti), _trB(args._trB), _beta(args._beta),</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  _k_block(compute_k_block(args)), _n_block(compute_n_block(args)),</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span>  _Mround(<a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(args._Msize, strategy::out_height())),</div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span>  _window_range(<a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(args._Msize, strategy::out_height()), _nbatches, <a class="code" href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a>(_Nsize, _n_block), _nmulti) { }</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">// Interface implementation - Compulsory functions</span></div><div class="line"><a name="l00138"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9fe814165d032363ea4f9cfdd36dd5e2"> 138</a></span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9fe814165d032363ea4f9cfdd36dd5e2">get_window_size</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span>  <span class="keywordflow">return</span> _window_range.<a class="code" href="classarm__gemm_1_1_n_d_range.xhtml#a2ee55bf0e2d8324a1360f6994a762e28">total_size</a>();</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> </div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span>  <span class="comment">// This kernel can always be dynamically scheduled.</span></div><div class="line"><a name="l00143"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a1bb76b8a32e09fb5ffffd66e04d7f741"> 143</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a1bb76b8a32e09fb5ffffd66e04d7f741">supports_dynamic_scheduling</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</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> </div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  <span class="comment">// Execute</span></div><div class="line"><a name="l00148"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560"> 148</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560">execute</a>(<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> start, <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> end, <span class="keywordtype">int</span> threadid)<span class="keyword"> override </span>{</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span>  profiler prof;</div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span>  strategy strat(_ci);</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>  <span class="comment">/* Make sure we've been set up correctly. */</span></div><div class="line"><a name="l00155"></a><span class="lineno"> 155</span>  assert(_B_transposed);</div><div class="line"><a name="l00156"></a><span class="lineno"> 156</span>  static_assert(std::is_same<To, Toi>::value, <span class="stringliteral">"gemm_native: Operand types must be the same."</span>);</div><div class="line"><a name="l00157"></a><span class="lineno"> 157</span>  static_assert(std::is_same<Tr, Tri>::value, <span class="stringliteral">"gemm_native: Result types must be the same."</span>);</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="comment">/* For now, each work item implies all the K for a given output</span></div><div class="line"><a name="l00160"></a><span class="lineno"> 160</span> <span class="comment"> * pixel (so we don't need to synchronize access to the output</span></div><div class="line"><a name="l00161"></a><span class="lineno"> 161</span> <span class="comment"> * array). So separate the loop over K blocks here. */</span></div><div class="line"><a name="l00162"></a><span class="lineno"> 162</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k0=0; k0<_Ksize; k0+=_k_block) {</div><div class="line"><a name="l00163"></a><span class="lineno"> 163</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kmax = std::min(k0 + _k_block, _Ksize);</div><div class="line"><a name="l00164"></a><span class="lineno"> 164</span>  <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kern_k = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(kmax-k0, strategy::k_unroll());</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">auto</span> p = _window_range.<a class="code" href="classarm__gemm_1_1_n_d_range.xhtml#a78a5c697e8717cb27954513e8ad2a508">iterator</a>(start, end);</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>  <span class="keywordflow">if</span> (p.done()) {</div><div class="line"><a name="l00169"></a><span class="lineno"> 169</span>  <span class="keywordflow">return</span>;</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> </div><div class="line"><a name="l00172"></a><span class="lineno"> 172</span>  <span class="keywordflow">do</span> {</div><div class="line"><a name="l00173"></a><span class="lineno"> 173</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> m_start = p.dim(0) * strategy::out_height();</div><div class="line"><a name="l00174"></a><span class="lineno"> 174</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> m_end = std::min(p.dim0_max() * strategy::out_height(), _Msize);</div><div class="line"><a name="l00175"></a><span class="lineno"> 175</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> batch = p.dim(1);</div><div class="line"><a name="l00176"></a><span class="lineno"> 176</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> n0 = p.dim(2) * _n_block;</div><div class="line"><a name="l00177"></a><span class="lineno"> 177</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> nmax = std::min(n0 + _n_block, _Nsize);</div><div class="line"><a name="l00178"></a><span class="lineno"> 178</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> multi = p.dim(3);</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>  <span class="keyword">const</span> Toi *b_panel = _B_transposed +</div><div class="line"><a name="l00181"></a><span class="lineno"> 181</span>  (multi * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width()) * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Ksize, strategy::k_unroll())) +</div><div class="line"><a name="l00182"></a><span class="lineno"> 182</span>  (k0 * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width())) +</div><div class="line"><a name="l00183"></a><span class="lineno"> 183</span>  (n0 * kern_k);</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> <span class="preprocessor">#ifdef CYCLE_PROFILING</span></div><div class="line"><a name="l00186"></a><span class="lineno"> 186</span>  <span class="keyword">auto</span> p = prof.ScopedProfiler(<a class="code" href="profiler_8hpp.xhtml#a97f4e398174b5e094f5262e5d785ac20">PROFILE_KERNEL</a>, (m_end - m_start) * kern_k * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(nmax-n0, strategy::out_width()));</div><div class="line"><a name="l00187"></a><span class="lineno"> 187</span> <span class="preprocessor">#endif</span></div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span> </div><div class="line"><a name="l00189"></a><span class="lineno"> 189</span>  strat.kernel(this->_Aptr + (multi * this->_A_multi_stride) + (batch * this->_A_batch_stride) + (m_start * this->_lda) + k0, this->_lda,</div><div class="line"><a name="l00190"></a><span class="lineno"> 190</span>  b_panel,</div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  this->_Cptr + (multi * this->_C_multi_stride) + (batch * this->_C_batch_stride) + (m_start * this->_ldc) + n0, this->_ldc,</div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  (k0 == 0) ? _beta : static_cast<Tr>(1),</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  (m_end - m_start), (nmax - n0), kmax-k0);</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span>  } <span class="keywordflow">while</span> (p.next_dim1());</div><div class="line"><a name="l00195"></a><span class="lineno"> 195</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> </div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  <span class="comment">// Interface implementation - pretransposed</span></div><div class="line"><a name="l00199"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ac5dec5063a75604fdc60555f15577b94"> 199</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ac5dec5063a75604fdc60555f15577b94">B_is_pretransposed</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  <span class="keywordflow">return</span> <span class="keyword">true</span>;</div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  }</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span> </div><div class="line"><a name="l00203"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#aa7cfff39cbb2be65cd40042a75e5ae1c"> 203</a></span>  <span class="keywordtype">bool</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#aa7cfff39cbb2be65cd40042a75e5ae1c">B_pretranspose_required</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  <span class="keywordflow">return</span> (_B_transposed==<span class="keyword">nullptr</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> </div><div class="line"><a name="l00207"></a><span class="lineno"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae9fc7380e533c43a1d91e48a7578e412"> 207</a></span>  <span class="keywordtype">size_t</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae9fc7380e533c43a1d91e48a7578e412">get_B_pretransposed_array_size</a>()<span class="keyword"> const override </span>{</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span>  <span class="keywordflow">return</span> <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Nsize, strategy::out_width()) * <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(_Ksize, strategy::k_unroll()) * _nmulti * <span class="keyword">sizeof</span>(Toi);</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"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae385dce1a98e03f0a3ca311f031e5110"> 211</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae385dce1a98e03f0a3ca311f031e5110">pretranspose_B_array</a>(<span class="keywordtype">void</span> *in_buffer, <span class="keyword">const</span> To *B, <span class="keyword">const</span> <span class="keywordtype">int</span> ldb, <span class="keyword">const</span> <span class="keywordtype">int</span> B_multi_stride)<span class="keyword"> override </span>{</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>  Toi *buffer = reinterpret_cast<Toi *>(in_buffer);</div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span>  _B_transposed = buffer;</div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span>  strategy strat(_ci);</div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span> </div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> multi=0; multi<_nmulti; multi++) {</div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k0=0; k0<_Ksize; k0+=_k_block) {</div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> kmax = std::min(k0 + _k_block, _Ksize);</div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> k_size = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(kmax-k0, strategy::k_unroll());</div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span> </div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  <span class="keywordflow">for</span> (<span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> x0=0; x0<_Nsize; x0+=_n_block) {</div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> xmax = std::min(x0+_n_block, _Nsize);</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>  <span class="keyword">const</span> <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> size = <a class="code" href="utils_8hpp.xhtml#a8f6fbf8b243a10af40ce8d47a1013384">roundup</a>(xmax-x0, strategy::out_width()) * k_size;</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>  strat.transforms.PrepareB( buffer, B + (multi * B_multi_stride), ldb,</div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  x0, xmax, k0, kmax, _trB);</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span> </div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  buffer += size;</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span>  }</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>  }</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"><a class="line" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a6e76d099f51fc0ef7f48dbd7d00214ac"> 235</a></span>  <span class="keywordtype">void</span> <a class="code" href="classarm__gemm_1_1_gemm_hybrid.xhtml#a6e76d099f51fc0ef7f48dbd7d00214ac">set_pretransposed_B_data</a>(<span class="keywordtype">void</span> *in_buffer)<span class="keyword"> override </span>{</div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span>  _B_transposed = reinterpret_cast<Toi *>(in_buffer);</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> };</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="comment">// namespace arm_gemm</span></div><div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_a0b21ce5febc3f7eed9c7b37c0b9a3560"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#a0b21ce5febc3f7eed9c7b37c0b9a3560">arm_gemm::GemmHybrid::execute</a></div><div class="ttdeci">void execute(unsigned int start, unsigned int end, int threadid) override</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00148">gemm_hybrid.hpp:148</a></div></div>
<div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_a82ec71716669b6765670e0ed242f5a47"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#a82ec71716669b6765670e0ed242f5a47">arm_gemm::GemmHybrid::operator=</a></div><div class="ttdeci">GemmHybrid & operator=(GemmHybrid &)=delete</div></div>
<div class="ttc" id="utils_8hpp_xhtml_aa4508679e1d089c2bdcf000f72357683"><div class="ttname"><a href="utils_8hpp.xhtml#aa4508679e1d089c2bdcf000f72357683">iceildiv</a></div><div class="ttdeci">T iceildiv(const T a, const T b)</div><div class="ttdef"><b>Definition:</b> <a href="utils_8hpp_source.xhtml#l00036">utils.hpp:36</a></div></div>
<div class="ttc" id="classarm__gemm_1_1_n_d_range_xhtml_a78a5c697e8717cb27954513e8ad2a508"><div class="ttname"><a href="classarm__gemm_1_1_n_d_range.xhtml#a78a5c697e8717cb27954513e8ad2a508">arm_gemm::NDRange::iterator</a></div><div class="ttdeci">NDRangeIterator iterator(unsigned int start, unsigned int end) const</div><div class="ttdef"><b>Definition:</b> <a href="ndrange_8hpp_source.xhtml#l00095">ndrange.hpp:95</a></div></div>
@@ -108,7 +108,7 @@
<div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_a9e9f8c1982f7144528bb6d01af482a13"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#a9e9f8c1982f7144528bb6d01af482a13">arm_gemm::GemmHybrid::GemmHybrid</a></div><div class="ttdeci">GemmHybrid(const GemmArgs< Tr > &args)</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00130">gemm_hybrid.hpp:130</a></div></div>
<div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_ac5dec5063a75604fdc60555f15577b94"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#ac5dec5063a75604fdc60555f15577b94">arm_gemm::GemmHybrid::B_is_pretransposed</a></div><div class="ttdeci">bool B_is_pretransposed() const override</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00199">gemm_hybrid.hpp:199</a></div></div>
<div class="ttc" id="mergeresults_8hpp_xhtml"><div class="ttname"><a href="mergeresults_8hpp.xhtml">mergeresults.hpp</a></div></div>
-<div class="ttc" id="namespacearm__gemm_xhtml"><div class="ttname"><a href="namespacearm__gemm.xhtml">arm_gemm</a></div><div class="ttdef"><b>Definition:</b> <a href="buffer__manager_8hpp_source.xhtml#l00041">buffer_manager.hpp:41</a></div></div>
+<div class="ttc" id="namespacearm__gemm_xhtml"><div class="ttname"><a href="namespacearm__gemm.xhtml">arm_gemm</a></div><div class="ttdef"><b>Definition:</b> <a href="barrier_8hpp_source.xhtml#l00030">barrier.hpp:30</a></div></div>
<div class="ttc" id="profiler_8hpp_xhtml"><div class="ttname"><a href="profiler_8hpp.xhtml">profiler.hpp</a></div></div>
<div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_ae9fc7380e533c43a1d91e48a7578e412"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae9fc7380e533c43a1d91e48a7578e412">arm_gemm::GemmHybrid::get_B_pretransposed_array_size</a></div><div class="ttdeci">size_t get_B_pretransposed_array_size() const override</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00207">gemm_hybrid.hpp:207</a></div></div>
<div class="ttc" id="classarm__gemm_1_1_gemm_hybrid_xhtml_ae385dce1a98e03f0a3ca311f031e5110"><div class="ttname"><a href="classarm__gemm_1_1_gemm_hybrid.xhtml#ae385dce1a98e03f0a3ca311f031e5110">arm_gemm::GemmHybrid::pretranspose_B_array</a></div><div class="ttdeci">void pretranspose_B_array(void *in_buffer, const To *B, const int ldb, const int B_multi_stride) override</div><div class="ttdef"><b>Definition:</b> <a href="gemm__hybrid_8hpp_source.xhtml#l00211">gemm_hybrid.hpp:211</a></div></div>
@@ -130,7 +130,7 @@
<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="dir_9847dd9cb7a28e46201717fa7336cf62.xhtml">arm_gemm</a></li><li class="navelem"><a class="el" href="gemm__hybrid_8hpp.xhtml">gemm_hybrid.hpp</a></li>
- <li class="footer">Generated on Thu May 23 2019 17:11:20 for Compute Library by
+ <li class="footer">Generated on Mon Sep 2 2019 11:47:22 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>