Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / ComputeLibrary / b3a371bc429d2ba45e56baaf239d8200c2662a74 / . / documentation / index.xhtml
blob: 659b00714b08ad16777e7132bc020ed36ec5934d [file] [log] [blame]
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1<!-- HTML header for doxygen 1.8.9.1-->
2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
3<html xmlns="http://www.w3.org/1999/xhtml">
4<head>
5<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
6<meta http-equiv="X-UA-Compatible" content="IE=9"/>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]7<meta name="generator" content="Doxygen 1.8.11"/>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]8<meta name="robots" content="NOINDEX, NOFOLLOW" /> <!-- Prevent indexing by search engines -->
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]9<title>Compute Library: Introduction</title>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]10<link href="tabs.css" rel="stylesheet" type="text/css"/>
11<script type="text/javascript" src="jquery.js"></script>
12<script type="text/javascript" src="dynsections.js"></script>
13<link href="navtree.css" rel="stylesheet" type="text/css"/>
14<script type="text/javascript" src="resize.js"></script>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]15<script type="text/javascript" src="navtreedata.js"></script>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]16<script type="text/javascript" src="navtree.js"></script>
17<script type="text/javascript">
18 $(document).ready(initResizable);
19 $(window).load(resizeHeight);
20</script>
21<link href="search/search.css" rel="stylesheet" type="text/css"/>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]22<script type="text/javascript" src="search/searchdata.js"></script>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]23<script type="text/javascript" src="search/search.js"></script>
24<script type="text/javascript">
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]25 $(document).ready(function() { init_search(); });
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]26</script>
27<script type="text/x-mathjax-config">
28 MathJax.Hub.Config({
29 extensions: ["tex2jax.js"],
30 jax: ["input/TeX","output/HTML-CSS"],
31});
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]32</script><script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]33<link href="doxygen.css" rel="stylesheet" type="text/css" />
34</head>
35<body>
36<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
37<div id="titlearea">
38<table cellspacing="0" cellpadding="0">
39 <tbody>
40 <tr style="height: 56px;">
41 <td style="padding-left: 0.5em;">
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]42 <div id="projectname">Compute Library
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]43 &#160;<span id="projectnumber">18.05</span>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]44 </div>
45 </td>
46 </tr>
47 </tbody>
48</table>
49</div>
50<!-- end header part -->
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]51<!-- Generated by Doxygen 1.8.11 -->
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]52<script type="text/javascript">
53var searchBox = new SearchBox("searchBox", "search",false,'Search');
54</script>
55 <div id="navrow1" class="tabs">
56 <ul class="tablist">
57 <li class="current"><a href="index.xhtml"><span>Main&#160;Page</span></a></li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]58 <li><a href="pages.xhtml"><span>Related&#160;Pages</span></a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]59 <li><a href="namespaces.xhtml"><span>Namespaces</span></a></li>
60 <li><a href="annotated.xhtml"><span>Data&#160;Structures</span></a></li>
61 <li><a href="files.xhtml"><span>Files</span></a></li>
62 <li>
63 <div id="MSearchBox" class="MSearchBoxInactive">
64 <span class="left">
65 <img id="MSearchSelect" src="search/mag_sel.png"
66 onmouseover="return searchBox.OnSearchSelectShow()"
67 onmouseout="return searchBox.OnSearchSelectHide()"
68 alt=""/>
69 <input type="text" id="MSearchField" value="Search" accesskey="S"
70 onfocus="searchBox.OnSearchFieldFocus(true)"
71 onblur="searchBox.OnSearchFieldFocus(false)"
72 onkeyup="searchBox.OnSearchFieldChange(event)"/>
73 </span><span class="right">
74 <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
75 </span>
76 </div>
77 </li>
78 </ul>
79 </div>
80</div><!-- top -->
81<div id="side-nav" class="ui-resizable side-nav-resizable">
82 <div id="nav-tree">
83 <div id="nav-tree-contents">
84 <div id="nav-sync" class="sync"></div>
85 </div>
86 </div>
87 <div id="splitbar" style="-moz-user-select:none;"
88 class="ui-resizable-handle">
89 </div>
90</div>
91<script type="text/javascript">
92$(document).ready(function(){initNavTree('index.xhtml','');});
93</script>
94<div id="doc-content">
95<!-- window showing the filter options -->
96<div id="MSearchSelectWindow"
97 onmouseover="return searchBox.OnSearchSelectShow()"
98 onmouseout="return searchBox.OnSearchSelectHide()"
99 onkeydown="return searchBox.OnSearchSelectKey(event)">
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]100</div>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]101
102<!-- iframe showing the search results (closed by default) -->
103<div id="MSearchResultsWindow">
104<iframe src="javascript:void(0)" frameborder="0"
105 name="MSearchResults" id="MSearchResults">
106</iframe>
107</div>
108
109<div class="header">
110 <div class="headertitle">
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]111<div class="title">Introduction </div> </div>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]112</div><!--header-->
113<div class="contents">
114<div class="toc"><h3>Table of Contents</h3>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]115<ul><li class="level1"><a href="#S0_1_contact">Contact / Support</a></li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]116<li class="level1"><a href="#S0_2_prebuilt_binaries">Pre-built binaries</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]117<li class="level1"><a href="#S1_file_organisation">File organisation</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]118<li class="level1"><a href="#S2_versions_changelog">Release versions and changelog</a><ul><li class="level2"><a href="#S2_1_versions">Release versions</a></li>
119<li class="level2"><a href="#S2_2_changelog">Changelog</a></li>
120</ul>
121</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]122<li class="level1"><a href="#S3_how_to_build">How to build the library and the examples</a><ul><li class="level2"><a href="#S3_1_build_options">Build options</a></li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]123<li class="level2"><a href="#S3_2_linux">Building for Linux</a><ul><li class="level3"><a href="#S3_2_1_library">How to build the library ?</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]124<li class="level3"><a href="#S3_2_2_examples">How to manually build the examples ?</a></li>
125</ul>
126</li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]127<li class="level2"><a href="#S3_3_android">Building for Android</a><ul><li class="level3"><a href="#S3_3_1_library">How to build the library ?</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]128<li class="level3"><a href="#S3_3_2_examples">How to manually build the examples ?</a></li>
129</ul>
130</li>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]131<li class="level2"><a href="#S3_4_bare_metal">Building for bare metal</a><ul><li class="level3"><a href="#S3_4_1_library">How to build the library ?</a></li>
132<li class="level3"><a href="#S3_4_2_examples">How to manually build the examples ?</a></li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]133</ul>
134</li>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]135<li class="level2"><a href="#S3_5_windows_host">Building on a Windows host system</a><ul><li class="level3"><a href="#S3_5_1_ubuntu_on_windows">Bash on Ubuntu on Windows</a></li>
136<li class="level3"><a href="#S3_5_2_cygwin">Cygwin</a></li>
137</ul>
138</li>
139<li class="level2"><a href="#S3_6_cl_stub_library">The OpenCL stub library</a></li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]140<li class="level2"><a href="#S3_7_gles_stub_library">The Linux OpenGLES and EGL stub libraries</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]141</ul>
142</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]143</ul>
144</div>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]145<div class="textblock"><p>The Computer Vision and Machine Learning library is a set of functions optimised for both ARM CPUs and GPUs using SIMD technologies.Several builds of the library are available using various configurations:</p><ul>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]146<li>OS: Linux, Android or bare metal.</li>
147<li>Architecture: armv7a (32bit) or arm64-v8a (64bit)</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]148<li>Technology: NEON / OpenCL / GLES_COMPUTE / NEON and OpenCL and GLES_COMPUTE</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]149<li>Debug / Asserts / Release: Use a build with asserts enabled to debug your application and enable extra validation. Once you are sure your application works as expected you can switch to a release build of the library for maximum performance.</li>
150</ul>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]151<h1><a class="anchor" id="S0_1_contact"></a>
152Contact / Support</h1>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]153<p>Please email <a href="#" onclick="location.href='mai'+'lto:'+'dev'+'el'+'ope'+'r@'+'arm'+'.c'+'om'; return false;">devel<span style="display: none;">.nosp@m.</span>oper<span style="display: none;">.nosp@m.</span>@arm.<span style="display: none;">.nosp@m.</span>com</a></p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]154<p>In order to facilitate the work of the support team please provide the build information of the library you are using. To get the version of the library you are using simply run: </p><pre class="fragment">$ strings android-armv7a-cl-asserts/libarm_compute.so | grep arm_compute_version
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]155arm_compute_version=v16.12 Build options: {'embed_kernels': '1', 'opencl': '1', 'arch': 'armv7a', 'neon': '0', 'asserts': '1', 'debug': '0', 'os': 'android', 'Werror': '1'} Git hash=f51a545d4ea12a9059fe4e598a092f1fd06dc858
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]156</pre><h1><a class="anchor" id="S0_2_prebuilt_binaries"></a>
157Pre-built binaries</h1>
158<p>For each release we provide some pre-built binaries of the library <a href="https://github.com/ARM-software/ComputeLibrary/releases">here</a></p>
159<p>These binaries have been built using the following toolchains:</p><ul>
160<li>Linux armv7a: gcc-linaro-arm-linux-gnueabihf-4.9-2014.07_linux</li>
161<li>Linux arm64-v8a: gcc-linaro-4.9-2016.02-x86_64_aarch64-linux-gnu</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]162<li>Android armv7a: clang++ / gnustl NDK r16b</li>
163<li>Android am64-v8a: clang++ / gnustl NDK r16b</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]164</ul>
165<dl class="section warning"><dt>Warning</dt><dd>Make sure to use a compatible toolchain to build your application or you will get some std::bad_alloc errors at runtime.</dd></dl>
166<h1><a class="anchor" id="S1_file_organisation"></a>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]167File organisation</h1>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]168<p>This archive contains:</p><ul>
169<li>The <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> header and source files</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]170<li>The latest Khronos OpenCL 1.2 C headers from the <a href="https://www.khronos.org/registry/cl/">Khronos OpenCL registry</a></li>
171<li>The latest Khronos cl2.hpp from the <a href="https://www.khronos.org/registry/cl/">Khronos OpenCL registry</a> (API version 2.1 when this document was written)</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]172<li>The latest Khronos OpenGL ES 3.1 C headers from the <a href="https://www.khronos.org/registry/gles/">Khronos OpenGL ES registry</a></li>
173<li>The latest Khronos EGL 1.5 C headers from the <a href="https://www.khronos.org/registry/gles/">Khronos EGL registry</a></li>
174<li>The sources for a stub version of libOpenCL.so, libGLESv1_CM.so, libGLESv2.so and libEGL.so to help you build your application.</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]175<li>An examples folder containing a few examples to compile and link against the library.</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]176<li>A <a class="el" href="dir_cbdb8362360e11eafe2fa3bc74cf0ffd.xhtml">utils</a> folder containing headers with some boiler plate code used by the examples.</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]177<li>This documentation.</li>
178</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]179<p>You should have the following file organisation: </p><pre class="fragment">.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]180├── arm_compute --&gt; All the arm_compute headers
181│   ├── core
182│   │   ├── CL
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]183│   │   │   ├── CLKernelLibrary.h --&gt; Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]184│   │   │   ├── CLKernels.h --&gt; Includes all the OpenCL kernels at once
185│   │   │   ├── CL specialisation of all the generic objects interfaces (ICLTensor, ICLImage, etc.)
186│   │   │   ├── kernels --&gt; Folder containing all the OpenCL kernels
187│   │   │   │   └── CL*Kernel.h
188│   │   │   └── OpenCL.h --&gt; Wrapper to configure the Khronos OpenCL C++ header
189│   │ ├── CPP
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]190│   │   │   ├── CPPKernels.h --&gt; Includes all the CPP kernels at once
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]191│   │ │   └── kernels --&gt; Folder containing all the CPP kernels
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]192│   │   │      └── CPP*Kernel.h
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]193│   │   ├── GLES_COMPUTE
194│   │   │   ├── GCKernelLibrary.h --&gt; Manages all the GLES kernels compilation and caching, provides accessors for the GLES Context.
195│   │   │   ├── GCKernels.h --&gt; Includes all the GLES kernels at once
196│   │   │   ├── GLES specialisation of all the generic objects interfaces (IGCTensor, IGCImage, etc.)
197│   │   │   ├── kernels --&gt; Folder containing all the GLES kernels
198│   │   │   │   └── GC*Kernel.h
199│   │   │   └── OpenGLES.h --&gt; Wrapper to configure the Khronos EGL and OpenGL ES C header
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]200│   │   ├── NEON
201│   │   │   ├── kernels --&gt; Folder containing all the NEON kernels
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]202│   │   │   │ ├── assembly --&gt; headers for assembly optimised NEON kernels.
203│   │   │   │ ├── convolution --&gt; headers for convolution assembly optimised NEON kernels.
204│   │   │   │   │   ├── common --&gt; headers for code which is common to several convolution implementations.
205│   │   │   │   │   ├── depthwise --&gt; headers for Depthwise convolultion assembly implementation
206│   │   │   │   │   └── winograd --&gt; headers for Winograd convolution assembly implementation
207│   │   │   │ ├── detail --&gt; Common code for several intrinsics implementations.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]208│   │   │   │   └── NE*Kernel.h
209│   │   │   └── NEKernels.h --&gt; Includes all the NEON kernels at once
210│   │   ├── All common basic types (Types.h, Window, Coordinates, Iterator, etc.)
211│   │   ├── All generic objects interfaces (ITensor, IImage, etc.)
212│   │   └── Objects metadata classes (ImageInfo, TensorInfo, MultiImageInfo)
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]213│   ├── graph
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]214│   │   ├── algorithms
215│   │   │   └── Generic algorithms used by the graph backend (e.g Order of traversal)
216│   │   ├── backends --&gt; The backend specific code
217│   │   │   ├── CL --&gt; OpenCL specific operations
218│   │   │   ├── GLES --&gt; OpenGLES Compute Shaders specific operations
219│   │   │   └── NEON --&gt; NEON specific operations
220│   │   ├── detail
221│   │   │   └── Collection of internal utilities.
222│   │   ├── frontend
223│   │   │   └── Code related to the stream frontend interface.
224│   │   ├── mutators
225│   │   │   └── Used to modify / optimise the Graph intermediate representation(Operator fusion, in place operations, etc.)
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]226│   │   ├── nodes
227│   │   │   └── The various nodes supported by the graph API
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]228│   │   ├── printers
229│   │   │   └── Debug printers
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]230│   │   └── Graph objects ( INode, ITensorAccessor, Graph, etc.)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]231│   └── runtime
232│   ├── CL
233│   │   ├── CL objects &amp; allocators (CLArray, CLImage, CLTensor, etc.)
234│   │   ├── functions --&gt; Folder containing all the OpenCL functions
235│   │   │   └── CL*.h
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]236│   │   ├── CLScheduler.h --&gt; Interface to enqueue OpenCL kernels and get/set the OpenCL CommandQueue and ICLTuner.
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]237│   │   ├── CLFunctions.h --&gt; Includes all the OpenCL functions at once
238│   │   └── tuners
239│   │      └── Local workgroup size tuners for specific architectures / GPUs
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]240│   ├── CPP
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]241│      │   ├── CPPKernels.h --&gt; Includes all the CPP functions at once.
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]242│   │   ├── CPPScheduler.h --&gt; Basic pool of threads to execute CPP/NEON code on several cores in parallel
243│   │   └── functions --&gt; Folder containing all the CPP functions
244│   │      └── CPP*.h
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]245│   ├── GLES_COMPUTE
246│   │   ├── GLES objects &amp; allocators (GCArray, GCImage, GCTensor, etc.)
247│   │   ├── functions --&gt; Folder containing all the GLES functions
248│   │   │   └── GC*.h
249│   │   ├── GCScheduler.h --&gt; Interface to enqueue GLES kernels and get/set the GLES CommandQueue.
250│   │   └── GCFunctions.h --&gt; Includes all the GLES functions at once
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]251│   ├── NEON
252│   │ ├── functions --&gt; Folder containing all the NEON functions
253│   │ │   └── NE*.h
254│   │ └── NEFunctions.h --&gt; Includes all the NEON functions at once
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]255│   ├── OMP
256│   │   └── OMPScheduler.h --&gt; OpenMP scheduler (Alternative to the CPPScheduler)
257│ ├── Memory manager files (LifetimeManager, PoolManager, etc.)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]258│   └── Basic implementations of the generic object interfaces (Array, Image, Tensor, etc.)
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]259├── data -&gt; Contains test images and reference data dumps used by validation tests
260├── docs -&gt; Contains Doxyfile and Doxygen sources used to generate the HTML pages in the documentation folder.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]261├── documentation
262│   ├── index.xhtml
263│   └── ...
264├── documentation.xhtml -&gt; documentation/index.xhtml
265├── examples
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]266│   ├── cl_*.cpp --&gt; OpenCL examples
267│   ├── gc_*.cpp --&gt; GLES compute shaders examples
268│   ├── graph_*.cpp --&gt; Graph examples
269│   ├── neoncl_*.cpp --&gt; NEON / OpenCL interoperability examples
270│   └── neon_*.cpp --&gt; NEON examples
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]271├── graph.h --&gt; Includes all the Graph headers at once.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]272├── include
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]273│   ├── CL
274│   │ └── Khronos OpenCL C headers and C++ wrapper
275│   ├── half --&gt; FP16 library available from http://half.sourceforge.net
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]276│   ├── libnpy --&gt; Library to load / write npy buffers, available from https://github.com/llohse/libnpy
277│  └── linux --&gt; Headers only needed for Linux builds
278│   └── Khronos EGL and OpenGLES headers
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]279├── opencl-1.2-stubs
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]280│ └── opencl_stubs.c --&gt; OpenCL stubs implementation
281├── opengles-3.1-stubs
282│   ├── EGL.c --&gt; EGL stubs implementation
283│   └── GLESv2.c --&gt; GLESv2 stubs implementation
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]284├── scripts
285│   ├── caffe_data_extractor.py --&gt; Basic script to export weights from Caffe to npy files
286│   └── tensorflow_data_extractor.py --&gt; Basic script to export weights from Tensor Flow to npy files
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]287├── src
288│   ├── core
289│ │ └── ... (Same structure as headers)
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]290│   │ ├── CL
291│   │ │ └── cl_kernels --&gt; All the OpenCL kernels
292│   │ └── GLES_COMPUTE
293│   │ └── cs_shaders --&gt; All the OpenGL ES Compute Shaders
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]294│   ├── graph
295│ │ └── ... (Same structure as headers)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]296│ └── runtime
297│ └── ... (Same structure as headers)
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]298├── support
299│ └── Various headers to work around toolchains / platform issues.
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]300├── tests
301│   ├── All test related files shared between validation and benchmark
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]302│   ├── benchmark --&gt; Sources for benchmarking
303│ │ ├── Benchmark specific files
304│   │ ├── fixtures
305│ │ │ └── Backend agnostic fixtures to initialise and run the functions to test.
306│ │ ├── CL --&gt; OpenCL benchmarking tests
307│ │ ├── GLES_COMPUTE --&gt; GLES benchmarking tests
308│ │ └── NEON --&gt; NEON benchmarking tests
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]309│   ├── CL --&gt; OpenCL accessors
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]310│   ├── GLES_COMPUTE --&gt; GLES accessors
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]311│   ├── NEON --&gt; NEON accessors
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]312│   ├── datasets
313│ │ └── Datasets for all the validation / benchmark tests, layer configurations for various networks, etc.
314│   ├── framework
315│ │ └── Boiler plate code for both validation and benchmark test suites (Command line parsers, instruments, output loggers, etc.)
316│   ├── networks
317│ │ └── Examples of how to instantiate networks.
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]318│   └── validation --&gt; Sources for validation
319│ ├── Validation specific files
320│   ├── fixtures
321│ │ └── Backend agnostic fixtures to initialise and run the functions to test.
322│   ├── reference
323│ │ └── Reference implementation used to validate the results of the various backends.
324│ ├── CL --&gt; OpenCL validation tests
325│ ├── GLES_COMPUTE --&gt; GLES validation tests
326│ ├── CPP --&gt; C++ reference implementations
327│ └── NEON --&gt; NEON validation tests
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]328└── utils --&gt; Boiler plate code used by examples
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]329 └── Various utilities to print types, load / store assets, etc.
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]330</pre><h1><a class="anchor" id="S2_versions_changelog"></a>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]331Release versions and changelog</h1>
332<h2><a class="anchor" id="S2_1_versions"></a>
333Release versions</h2>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]334<p>All releases are numbered vYY.MM Where YY are the last two digits of the year, and MM the month number. If there is more than one release in a month then an extra sequential number is appended at the end: </p><pre class="fragment">v17.03 (First release of March 2017)
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]335v17.03.1 (Second release of March 2017)
336v17.04 (First release of April 2017)
337</pre><dl class="section note"><dt>Note</dt><dd>We're aiming at releasing one major public release with new features per quarter. All releases in between will only contain bug fixes.</dd></dl>
338<h2><a class="anchor" id="S2_2_changelog"></a>
339Changelog</h2>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]340<p>v18.05 Public maintenance release</p><ul>
341<li>Various bug fixes.</li>
342<li>Various optimisations.</li>
343<li>Major redesign in the interface for the neon kernels implemented in assembly.</li>
344<li>Removed arm_compute::NEGEMMLowpAArch64A53Kernel / arm_compute::NEGEMMLowpAArch64Kernel / arm_compute::NEGEMMLowpAArch64V8P4Kernel / arm_compute::NEGEMMInterleavedBlockedKernel / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_assembly_matrix_multiply_core.xhtml" title="Basic function to execute matrix multiply assembly kernels. ">arm_compute::NEGEMMLowpAssemblyMatrixMultiplyCore</a> / arm_compute::NEHGEMMAArch64FP16Kernel</li>
345<li>Added NEGEMMAssemblyWrapper and <a class="el" href="classarm__compute_1_1_assembly_kernel_glue.xhtml" title="Assembly kernel glue. ">AssemblyKernelGlue</a> which are used to execute assembly kernels in neon functions.</li>
346<li>Minor changes to the <a class="el" href="classarm__compute_1_1_c_p_u_info.xhtml">CPUInfo</a> type to make it compatible with the new assembly gemm interface.</li>
347<li>Moved neon assembly kernels to the folder src/core/NEON/kernels/arm_gemm.</li>
348<li>Improved doxygen documentation.</li>
349<li>Improved memory management for layer's transitions.</li>
350<li>Added support for NHWC data layout in tensors.</li>
351<li>Added NHWC data layout support to:<ul>
352<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_convolution_layer.xhtml">NEGEMMConvolutionLayer</a></li>
353<li><a class="el" href="classarm__compute_1_1_n_e_direct_convolution_layer.xhtml">NEDirectConvolutionLayer</a></li>
354<li><a class="el" href="classarm__compute_1_1_n_e_pooling_layer.xhtml">NEPoolingLayer</a> / <a class="el" href="classarm__compute_1_1_c_l_pooling_layer.xhtml">CLPoolingLayer</a></li>
355<li><a class="el" href="classarm__compute_1_1_n_e_batch_normalization_layer.xhtml">NEBatchNormalizationLayer</a> / <a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer.xhtml">CLBatchNormalizationLayer</a></li>
356<li><a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer.xhtml">NEDepthwiseConvolutionLayer</a></li>
357<li><a class="el" href="classarm__compute_1_1_n_e_scale.xhtml">NEScale</a></li>
358<li><a class="el" href="classarm__compute_1_1_n_e_im2_col.xhtml">NEIm2Col</a></li>
359</ul>
360</li>
361<li>Added support for dilated convolutions in <a class="el" href="classarm__compute_1_1_n_e_convolution_layer.xhtml">NEConvolutionLayer</a> and <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a>.</li>
362<li>New OpenCL kernels / functions:<ul>
363<li><a class="el" href="classarm__compute_1_1_c_l_channel_shuffle_layer.xhtml">CLChannelShuffleLayer</a> / <a class="el" href="classarm__compute_1_1_c_l_channel_shuffle_layer_kernel.xhtml">CLChannelShuffleLayerKernel</a></li>
364<li><a class="el" href="classarm__compute_1_1_c_l_convert_fully_connected_weights_kernel.xhtml">CLConvertFullyConnectedWeightsKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_convert_fully_connected_weights.xhtml">CLConvertFullyConnectedWeights</a></li>
365<li><a class="el" href="classarm__compute_1_1_c_l_copy.xhtml">CLCopy</a> / <a class="el" href="classarm__compute_1_1_c_l_copy_kernel.xhtml">CLCopyKernel</a></li>
366<li><a class="el" href="classarm__compute_1_1_c_l_l_s_t_m_layer.xhtml">CLLSTMLayer</a></li>
367<li><a class="el" href="classarm__compute_1_1_c_l_r_n_n_layer.xhtml">CLRNNLayer</a></li>
368<li><a class="el" href="classarm__compute_1_1_c_l_width_concatenate_layer.xhtml">CLWidthConcatenateLayer</a> / <a class="el" href="classarm__compute_1_1_c_l_width_concatenate_layer_kernel.xhtml">CLWidthConcatenateLayerKernel</a></li>
369<li><a class="el" href="classarm__compute_1_1_c_l_winograd_filter_transform_kernel.xhtml">CLWinogradFilterTransformKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_winograd_input_transform_kernel.xhtml">CLWinogradInputTransformKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_winograd_convolution_layer.xhtml">CLWinogradConvolutionLayer</a></li>
370<li><a class="el" href="classarm__compute_1_1_c_l_winograd_input_transform_kernel.xhtml">CLWinogradInputTransformKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_winograd_input_transform.xhtml">CLWinogradInputTransform</a></li>
371</ul>
372</li>
373<li>New Neon kernels / functions:<ul>
374<li><a class="el" href="classarm__compute_1_1_c_l_r_n_n_layer.xhtml">CLRNNLayer</a></li>
375<li><a class="el" href="classarm__compute_1_1_n_e_convert_fully_connected_weights_kernel.xhtml">NEConvertFullyConnectedWeightsKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_convert_fully_connected_weights.xhtml">NEConvertFullyConnectedWeights</a>.</li>
376</ul>
377</li>
378<li>Created the validate method in <a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer.xhtml">CLDepthwiseConvolutionLayer</a>.</li>
379<li>Beta and gamma are no longer mandatory arguments in <a class="el" href="classarm__compute_1_1_n_e_batch_normalization_layer.xhtml">NEBatchNormalizationLayer</a> and <a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer.xhtml">CLBatchNormalizationLayer</a>.</li>
380<li>Added depth multiplier support in <a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer.xhtml">NEDepthwiseConvolutionLayer</a> and <a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer.xhtml">CLDepthwiseConvolutionLayer</a>.</li>
381<li>Added broadcast multiply support in <a class="el" href="classarm__compute_1_1_n_e_pixel_wise_multiplication.xhtml">NEPixelWiseMultiplication</a> / <a class="el" href="classarm__compute_1_1_n_e_pixel_wise_multiplication_kernel.xhtml">NEPixelWiseMultiplicationKernel</a>.</li>
382<li>Port mobilenet example to NHWC data layout.</li>
383<li>Enabled Winograd method in <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a>.</li>
384<li>Renamed NEWinogradLayer to <a class="el" href="classarm__compute_1_1_n_e_winograd_convolution_layer.xhtml">NEWinogradConvolutionLayer</a>.</li>
385<li>Updated <a class="el" href="classarm__compute_1_1_n_e_winograd_convolution_layer.xhtml">NEWinogradConvolutionLayer</a> to use highly optimised assembly kernels in src/core/NEON/kernels/arm_gemm.</li>
386<li>Added memory manager support in GLES functions.</li>
387<li>Major refactoring of the graph API.</li>
388<li>Added GLES backend in the graph API.</li>
389<li>Added support for the memory manager in the graph API.</li>
390<li>Enabled Winograd Convolution method in the graph API.</li>
391<li>Added support for grouped convolutions in the graph API.</li>
392<li>Replaced NEDeconvolutionLayerUpsampleKernel with <a class="el" href="classarm__compute_1_1_n_e_scale_kernel.xhtml">NEScaleKernel</a> in <a class="el" href="classarm__compute_1_1_n_e_deconvolution_layer.xhtml">NEDeconvolutionLayer</a>.</li>
393<li>Added fast maths flag in <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a>.</li>
394<li>Added new tests and benchmarks in validation and benchmark frameworks</li>
395<li>Merge Activation layer with Convolution Layer (NEON. CL, GLES)</li>
396<li>Added support to OpenCL 2.0 SVM</li>
397<li>Added support to import memory in OpenCL tensors.</li>
398<li>Added the prepare() method to perform any one off pre-processing before running the function.</li>
399<li>Added new examples:<ul>
400<li><a class="el" href="graph__inception__v4_8cpp.xhtml">graph_inception_v4.cpp</a></li>
401<li><a class="el" href="graph__resnext50_8cpp.xhtml">graph_resnext50.cpp</a></li>
402</ul>
403</li>
404<li>Added memory measurement instrument for CL.</li>
405</ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]406<p>v18.03 Public maintenance release</p><ul>
407<li>Various bug fixes.</li>
408<li>Fixed bug in <a class="el" href="classarm__compute_1_1_n_e_activation_layer.xhtml">NEActivationLayer</a></li>
409<li>Fix in <a class="el" href="classarm__compute_1_1_c_l_tuner.xhtml">CLTuner</a> when using batches.</li>
410<li>Updated recommended NDK version to r16b (And fixed warnings).</li>
411<li>Fixed bug in validation code.</li>
412<li>Added Inception v4 graph example.</li>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]413<li>Renamed NEWinogradLayer.cpp to <a class="el" href="classarm__compute_1_1_n_e_winograd_convolution_layer.xhtml">NEWinogradConvolutionLayer</a></li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]414</ul>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]415<p>v18.02 Public major release</p><ul>
416<li>Various NEON / OpenCL / GLES optimisations.</li>
417<li>Various bug fixes.</li>
418<li>Changed default number of threads on big LITTLE systems.</li>
419<li>Refactored examples and added:<ul>
420<li>graph_mobilenet_qassym8</li>
421<li>graph_resnet</li>
422<li>graph_squeezenet_v1_1</li>
423</ul>
424</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]425<li>Renamed <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a> into <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_convolution_layer.xhtml">CLGEMMConvolutionLayer</a> and created a new <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a> to select the fastest convolution method.</li>
426<li>Renamed <a class="el" href="classarm__compute_1_1_n_e_convolution_layer.xhtml">NEConvolutionLayer</a> into <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_convolution_layer.xhtml">NEGEMMConvolutionLayer</a> and created a new <a class="el" href="classarm__compute_1_1_n_e_convolution_layer.xhtml">NEConvolutionLayer</a> to select the fastest convolution method.</li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]427<li>Added in place support to:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]428<li><a class="el" href="classarm__compute_1_1_c_l_activation_layer.xhtml">CLActivationLayer</a></li>
429<li><a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer.xhtml">CLBatchNormalizationLayer</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]430</ul>
431</li>
432<li>Added QASYMM8 support to:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]433<li><a class="el" href="classarm__compute_1_1_c_l_activation_layer.xhtml">CLActivationLayer</a></li>
434<li><a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer.xhtml">CLDepthwiseConvolutionLayer</a></li>
435<li><a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer.xhtml">NEDepthwiseConvolutionLayer</a></li>
436<li><a class="el" href="classarm__compute_1_1_n_e_softmax_layer.xhtml">NESoftmaxLayer</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]437</ul>
438</li>
439<li>Added FP16 support to:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]440<li><a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer3x3.xhtml">CLDepthwiseConvolutionLayer3x3</a></li>
441<li><a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer.xhtml">CLDepthwiseConvolutionLayer</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]442</ul>
443</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]444<li>Added broadcasting support to <a class="el" href="classarm__compute_1_1_n_e_arithmetic_addition.xhtml">NEArithmeticAddition</a> / <a class="el" href="classarm__compute_1_1_c_l_arithmetic_addition.xhtml">CLArithmeticAddition</a> / <a class="el" href="classarm__compute_1_1_c_l_pixel_wise_multiplication.xhtml">CLPixelWiseMultiplication</a></li>
445<li>Added fused batched normalization and activation to <a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer.xhtml">CLBatchNormalizationLayer</a> and <a class="el" href="classarm__compute_1_1_n_e_batch_normalization_layer.xhtml">NEBatchNormalizationLayer</a></li>
446<li>Added support for non-square pooling to <a class="el" href="classarm__compute_1_1_n_e_pooling_layer.xhtml">NEPoolingLayer</a> and <a class="el" href="classarm__compute_1_1_c_l_pooling_layer.xhtml">CLPoolingLayer</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]447<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]448<li><a class="el" href="classarm__compute_1_1_c_l_direct_convolution_layer_output_stage_kernel.xhtml">CLDirectConvolutionLayerOutputStageKernel</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]449</ul>
450</li>
451<li>New NEON kernels / functions<ul>
452<li>Added name() method to all kernels.</li>
453<li>Added support for Winograd 5x5.</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]454<li><a class="el" href="classarm__compute_1_1_n_e_permute_kernel.xhtml">NEPermuteKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_permute.xhtml">NEPermute</a></li>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]455<li><a class="el" href="classarm__compute_1_1_n_e_winograd_layer_transform_input_kernel.xhtml">NEWinogradLayerTransformInputKernel</a> / NEWinogradLayer</li>
456<li><a class="el" href="classarm__compute_1_1_n_e_winograd_layer_transform_output_kernel.xhtml">NEWinogradLayerTransformOutputKernel</a> / NEWinogradLayer</li>
457<li><a class="el" href="classarm__compute_1_1_n_e_winograd_layer_transform_weights_kernel.xhtml">NEWinogradLayerTransformWeightsKernel</a> / NEWinogradLayer</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]458<li>Renamed NEWinogradLayerKernel into <a class="el" href="classarm__compute_1_1_n_e_winograd_layer_batched_g_e_m_m_kernel.xhtml">NEWinogradLayerBatchedGEMMKernel</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]459</ul>
460</li>
461<li>New GLES kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]462<li><a class="el" href="classarm__compute_1_1_g_c_tensor_shift_kernel.xhtml">GCTensorShiftKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_tensor_shift.xhtml">GCTensorShift</a></li>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]463</ul>
464</li>
465</ul>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]466<p>v18.01 Public maintenance release</p><ul>
467<li>Various bug fixes</li>
468<li>Added some of the missing <a class="el" href="namespacearm__compute_1_1test_1_1validation.xhtml#a6813132c943295888972727864ea5c2f">validate()</a> methods</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]469<li>Added <a class="el" href="classarm__compute_1_1_c_l_deconvolution_layer_upsample_kernel.xhtml">CLDeconvolutionLayerUpsampleKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_deconvolution_layer.xhtml">CLDeconvolutionLayer</a> <a class="el" href="classarm__compute_1_1_c_l_deconvolution_layer_upsample.xhtml">CLDeconvolutionLayerUpsample</a></li>
470<li>Added <a class="el" href="classarm__compute_1_1_c_l_permute_kernel.xhtml">CLPermuteKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_permute.xhtml">CLPermute</a></li>
471<li>Added method to clean the programs cache in the CL <a class="el" href="classarm__compute_1_1_kernel.xhtml" title="Kernel class. ">Kernel</a> library.</li>
472<li>Added <a class="el" href="classarm__compute_1_1_g_c_arithmetic_addition_kernel.xhtml">GCArithmeticAdditionKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_arithmetic_addition.xhtml">GCArithmeticAddition</a></li>
473<li>Added <a class="el" href="classarm__compute_1_1_g_c_depthwise_convolution_layer3x3_kernel.xhtml">GCDepthwiseConvolutionLayer3x3Kernel</a> / <a class="el" href="classarm__compute_1_1_g_c_depthwise_convolution_layer3x3.xhtml">GCDepthwiseConvolutionLayer3x3</a></li>
474<li>Added <a class="el" href="classarm__compute_1_1_g_c_normalize_planar_y_u_v_layer_kernel.xhtml">GCNormalizePlanarYUVLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_normalize_planar_y_u_v_layer.xhtml">GCNormalizePlanarYUVLayer</a></li>
475<li>Added <a class="el" href="classarm__compute_1_1_g_c_scale_kernel.xhtml">GCScaleKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_scale.xhtml">GCScale</a></li>
476<li>Added <a class="el" href="classarm__compute_1_1_g_c_weights_reshape_kernel.xhtml">GCWeightsReshapeKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_convolution_layer.xhtml">GCConvolutionLayer</a></li>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]477<li>Added FP16 support to the following GLES compute kernels:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]478<li><a class="el" href="classarm__compute_1_1_g_c_col2_im_kernel.xhtml">GCCol2ImKernel</a></li>
479<li><a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_interleave4x4_kernel.xhtml">GCGEMMInterleave4x4Kernel</a></li>
480<li><a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_transpose1x_w_kernel.xhtml">GCGEMMTranspose1xWKernel</a></li>
481<li><a class="el" href="classarm__compute_1_1_g_c_im2_col_kernel.xhtml">GCIm2ColKernel</a></li>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]482</ul>
483</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]484<li>Refactored NEON Winograd (NEWinogradLayerKernel)</li>
485<li>Added <a class="el" href="classarm__compute_1_1_n_e_direct_convolution_layer_output_stage_kernel.xhtml">NEDirectConvolutionLayerOutputStageKernel</a></li>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]486<li>Added QASYMM8 support to the following NEON kernels:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]487<li><a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer3x3_kernel.xhtml">NEDepthwiseConvolutionLayer3x3Kernel</a></li>
488<li><a class="el" href="classarm__compute_1_1_n_e_fill_border_kernel.xhtml">NEFillBorderKernel</a></li>
489<li><a class="el" href="classarm__compute_1_1_n_e_pooling_layer_kernel.xhtml">NEPoolingLayerKernel</a></li>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]490</ul>
491</li>
492<li>Added new examples:<ul>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]493<li>graph_cl_mobilenet_qasymm8.cpp</li>
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]494<li><a class="el" href="graph__inception__v3_8cpp.xhtml">graph_inception_v3.cpp</a></li>
495<li><a class="el" href="gc__dc_8cpp.xhtml">gc_dc.cpp</a></li>
496</ul>
497</li>
498<li>More tests added to both validation and benchmarking suites.</li>
499</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]500<p>v17.12 Public major release</p><ul>
501<li>Most machine learning functions on OpenCL support the new data type QASYMM8</li>
502<li>Introduced logging interface</li>
503<li>Introduced opencl timer</li>
504<li>Reworked GEMMLowp interface</li>
505<li>Added new NEON assembly kernels for GEMMLowp, SGEMM and HGEMM</li>
506<li>Added validation method for most Machine Learning kernels / functions</li>
507<li>Added new graph examples such as googlenet, mobilenet, squeezenet, vgg16 and vgg19</li>
508<li>Added sgemm example for OpenCL</li>
509<li>Added absolute difference example for GLES compute</li>
510<li>Added new tests and benchmarks in validation and benchmark frameworks</li>
511<li>Added new kernels / functions for GLES compute</li>
512<li>New OpenGL ES kernels / functions<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]513<li><a class="el" href="classarm__compute_1_1_g_c_absolute_difference_kernel.xhtml">GCAbsoluteDifferenceKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_absolute_difference.xhtml">GCAbsoluteDifference</a></li>
514<li><a class="el" href="classarm__compute_1_1_g_c_activation_layer_kernel.xhtml">GCActivationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_activation_layer.xhtml">GCActivationLayer</a></li>
515<li><a class="el" href="classarm__compute_1_1_g_c_batch_normalization_layer_kernel.xhtml">GCBatchNormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_batch_normalization_layer.xhtml">GCBatchNormalizationLayer</a></li>
516<li><a class="el" href="classarm__compute_1_1_g_c_col2_im_kernel.xhtml">GCCol2ImKernel</a></li>
517<li><a class="el" href="classarm__compute_1_1_g_c_depth_concatenate_layer_kernel.xhtml">GCDepthConcatenateLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_depth_concatenate_layer.xhtml">GCDepthConcatenateLayer</a></li>
518<li><a class="el" href="classarm__compute_1_1_g_c_direct_convolution_layer_kernel.xhtml">GCDirectConvolutionLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_direct_convolution_layer.xhtml">GCDirectConvolutionLayer</a></li>
519<li><a class="el" href="classarm__compute_1_1_g_c_dropout_layer_kernel.xhtml">GCDropoutLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_dropout_layer.xhtml">GCDropoutLayer</a></li>
520<li><a class="el" href="classarm__compute_1_1_g_c_fill_border_kernel.xhtml">GCFillBorderKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_fill_border.xhtml">GCFillBorder</a></li>
521<li><a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_interleave4x4_kernel.xhtml">GCGEMMInterleave4x4Kernel</a> / <a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_interleave4x4.xhtml">GCGEMMInterleave4x4</a></li>
522<li><a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_matrix_accumulate_biases_kernel.xhtml">GCGEMMMatrixAccumulateBiasesKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_matrix_addition_kernel.xhtml">GCGEMMMatrixAdditionKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_matrix_multiply_kernel.xhtml">GCGEMMMatrixMultiplyKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_g_e_m_m.xhtml">GCGEMM</a></li>
523<li><a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_transpose1x_w_kernel.xhtml">GCGEMMTranspose1xWKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_g_e_m_m_transpose1x_w.xhtml">GCGEMMTranspose1xW</a></li>
524<li><a class="el" href="classarm__compute_1_1_g_c_im2_col_kernel.xhtml">GCIm2ColKernel</a></li>
525<li><a class="el" href="classarm__compute_1_1_g_c_normalization_layer_kernel.xhtml">GCNormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_normalization_layer.xhtml">GCNormalizationLayer</a></li>
526<li><a class="el" href="classarm__compute_1_1_g_c_pixel_wise_multiplication_kernel.xhtml">GCPixelWiseMultiplicationKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_pixel_wise_multiplication.xhtml">GCPixelWiseMultiplication</a></li>
527<li><a class="el" href="classarm__compute_1_1_g_c_pooling_layer_kernel.xhtml">GCPoolingLayerKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_pooling_layer.xhtml">GCPoolingLayer</a></li>
528<li><a class="el" href="classarm__compute_1_1_g_c_logits1_d_max_kernel.xhtml">GCLogits1DMaxKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_logits1_d_shift_exp_sum_kernel.xhtml">GCLogits1DShiftExpSumKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_logits1_d_norm_kernel.xhtml">GCLogits1DNormKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_softmax_layer.xhtml">GCSoftmaxLayer</a></li>
529<li><a class="el" href="classarm__compute_1_1_g_c_transpose_kernel.xhtml">GCTransposeKernel</a> / <a class="el" href="classarm__compute_1_1_g_c_transpose.xhtml">GCTranspose</a></li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]530</ul>
531</li>
532<li>New NEON kernels / functions<ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]533<li>arm_compute::NEGEMMLowpAArch64A53Kernel / arm_compute::NEGEMMLowpAArch64Kernel / arm_compute::NEGEMMLowpAArch64V8P4Kernel / arm_compute::NEGEMMInterleavedBlockedKernel / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_assembly_matrix_multiply_core.xhtml" title="Basic function to execute matrix multiply assembly kernels. ">arm_compute::NEGEMMLowpAssemblyMatrixMultiplyCore</a></li>
534<li>arm_compute::NEHGEMMAArch64FP16Kernel</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]535<li><a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer3x3_kernel.xhtml">NEDepthwiseConvolutionLayer3x3Kernel</a> / <a class="el" href="classarm__compute_1_1_n_e_depthwise_im2_col_kernel.xhtml">NEDepthwiseIm2ColKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_matrix_vector_multiply_kernel.xhtml">NEGEMMMatrixVectorMultiplyKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_depthwise_vector_to_tensor_kernel.xhtml">NEDepthwiseVectorToTensorKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_depthwise_convolution_layer.xhtml">NEDepthwiseConvolutionLayer</a></li>
536<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_offset_contribution_kernel.xhtml">NEGEMMLowpOffsetContributionKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_matrix_a_reduction_kernel.xhtml">NEGEMMLowpMatrixAReductionKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_matrix_b_reduction_kernel.xhtml">NEGEMMLowpMatrixBReductionKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_matrix_multiply_core.xhtml">NEGEMMLowpMatrixMultiplyCore</a></li>
537<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_by_fixed_point_kernel.xhtml">NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_by_fixed_point.xhtml">NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint</a></li>
538<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_kernel.xhtml">NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale.xhtml">NEGEMMLowpQuantizeDownInt32ToUint8Scale</a></li>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]539<li>NEWinogradLayer / NEWinogradLayerKernel</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]540</ul>
541</li>
542<li>New OpenCL kernels / functions<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]543<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_offset_contribution_kernel.xhtml">CLGEMMLowpOffsetContributionKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_matrix_a_reduction_kernel.xhtml">CLGEMMLowpMatrixAReductionKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_matrix_b_reduction_kernel.xhtml">CLGEMMLowpMatrixBReductionKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_matrix_multiply_core.xhtml">CLGEMMLowpMatrixMultiplyCore</a></li>
544<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_by_fixed_point_kernel.xhtml">CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_by_fixed_point.xhtml">CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint</a></li>
545<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale_kernel.xhtml">CLGEMMLowpQuantizeDownInt32ToUint8ScaleKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_quantize_down_int32_to_uint8_scale.xhtml">CLGEMMLowpQuantizeDownInt32ToUint8Scale</a></li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]546</ul>
547</li>
548<li>New graph nodes for NEON and OpenCL<ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]549<li>graph::BranchLayer</li>
550<li>graph::DepthConvertLayer</li>
551<li>graph::DepthwiseConvolutionLayer</li>
552<li>graph::DequantizationLayer</li>
553<li>graph::FlattenLayer</li>
554<li>graph::QuantizationLayer</li>
555<li>graph::ReshapeLayer</li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]556</ul>
557</li>
558</ul>
559<p>v17.10 Public maintenance release</p><ul>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]560<li>Bug fixes:<ul>
561<li>Check the maximum local workgroup size supported by OpenCL devices</li>
562<li>Minor documentation updates (Fixed instructions to build the examples)</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]563<li>Introduced a <a class="el" href="classarm__compute_1_1graph_1_1_graph_context.xhtml" title="Graph context. ">graph::GraphContext</a></li>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]564<li>Added a few new Graph nodes, support for branches and grouping.</li>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]565<li>Automatically enable cl_printf in debug builds</li>
566<li>Fixed bare metal builds for armv7a</li>
567<li>Added AlexNet and cartoon effect examples</li>
568<li>Fixed library builds: libraries are no longer built as supersets of each other.(It means application using the Runtime part of the library now need to link against both libarm_compute_core and libarm_compute)</li>
569</ul>
570</li>
571</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]572<p>v17.09 Public major release</p><ul>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]573<li>Experimental Graph support: initial implementation of a simple stream API to easily chain machine learning layers.</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]574<li><a class="el" href="classarm__compute_1_1_memory.xhtml" title="CPU implementation of memory object. ">Memory</a> Manager (<a class="el" href="classarm__compute_1_1_blob_lifetime_manager.xhtml">BlobLifetimeManager</a>, <a class="el" href="classarm__compute_1_1_blob_memory_pool.xhtml">BlobMemoryPool</a>, <a class="el" href="classarm__compute_1_1_i_lifetime_manager.xhtml">ILifetimeManager</a>, <a class="el" href="classarm__compute_1_1_i_memory_group.xhtml">IMemoryGroup</a>, <a class="el" href="classarm__compute_1_1_i_memory_manager.xhtml">IMemoryManager</a>, <a class="el" href="classarm__compute_1_1_i_memory_pool.xhtml">IMemoryPool</a>, <a class="el" href="classarm__compute_1_1_i_pool_manager.xhtml">IPoolManager</a>, <a class="el" href="classarm__compute_1_1_memory_manager_on_demand.xhtml">MemoryManagerOnDemand</a>, <a class="el" href="classarm__compute_1_1_pool_manager.xhtml">PoolManager</a>)</li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]575<li>New validation and benchmark frameworks (Boost and Google frameworks replaced by homemade framework).</li>
576<li>Most machine learning functions support both fixed point 8 and 16 bit (QS8, QS16) for both NEON and OpenCL.</li>
577<li>New NEON kernels / functions:<ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]578<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_assembly_base_kernel.xhtml" title="Base class for GEMM NEON kernels implemented in Assembly. ">arm_compute::NEGEMMAssemblyBaseKernel</a> arm_compute::NEGEMMAArch64Kernel</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]579<li><a class="el" href="classarm__compute_1_1_n_e_dequantization_layer_kernel.xhtml">NEDequantizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_dequantization_layer.xhtml">NEDequantizationLayer</a></li>
580<li><a class="el" href="classarm__compute_1_1_n_e_floor_kernel.xhtml">NEFloorKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_floor.xhtml">NEFloor</a></li>
581<li><a class="el" href="classarm__compute_1_1_n_e_l2_normalize_layer_kernel.xhtml">NEL2NormalizeLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_l2_normalize_layer.xhtml">NEL2NormalizeLayer</a></li>
582<li><a class="el" href="classarm__compute_1_1_n_e_quantization_layer_kernel.xhtml">NEQuantizationLayerKernel</a> <a class="el" href="classarm__compute_1_1_n_e_min_max_layer_kernel.xhtml">NEMinMaxLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_quantization_layer.xhtml">NEQuantizationLayer</a></li>
583<li><a class="el" href="classarm__compute_1_1_n_e_r_o_i_pooling_layer_kernel.xhtml">NEROIPoolingLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_r_o_i_pooling_layer.xhtml">NEROIPoolingLayer</a></li>
584<li><a class="el" href="classarm__compute_1_1_n_e_reduction_operation_kernel.xhtml">NEReductionOperationKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_reduction_operation.xhtml">NEReductionOperation</a></li>
585<li><a class="el" href="classarm__compute_1_1_n_e_reshape_layer_kernel.xhtml">NEReshapeLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_reshape_layer.xhtml">NEReshapeLayer</a></li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]586</ul>
587</li>
588<li>New OpenCL kernels / functions:<ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]589<li><a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer3x3_n_c_h_w_kernel.xhtml">CLDepthwiseConvolutionLayer3x3NCHWKernel</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer3x3_n_h_w_c_kernel.xhtml">CLDepthwiseConvolutionLayer3x3NHWCKernel</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_im2_col_kernel.xhtml">CLDepthwiseIm2ColKernel</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_vector_to_tensor_kernel.xhtml">CLDepthwiseVectorToTensorKernel</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_weights_reshape_kernel.xhtml">CLDepthwiseWeightsReshapeKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer3x3.xhtml">CLDepthwiseConvolutionLayer3x3</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_convolution_layer.xhtml">CLDepthwiseConvolutionLayer</a> <a class="el" href="classarm__compute_1_1_c_l_depthwise_separable_convolution_layer.xhtml">CLDepthwiseSeparableConvolutionLayer</a></li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]590<li><a class="el" href="classarm__compute_1_1_c_l_dequantization_layer_kernel.xhtml">CLDequantizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_dequantization_layer.xhtml">CLDequantizationLayer</a></li>
591<li><a class="el" href="classarm__compute_1_1_c_l_direct_convolution_layer_kernel.xhtml">CLDirectConvolutionLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_direct_convolution_layer.xhtml">CLDirectConvolutionLayer</a></li>
592<li><a class="el" href="classarm__compute_1_1_c_l_flatten_layer.xhtml">CLFlattenLayer</a></li>
593<li><a class="el" href="classarm__compute_1_1_c_l_floor_kernel.xhtml">CLFloorKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_floor.xhtml">CLFloor</a></li>
594<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_transpose1x_w.xhtml">CLGEMMTranspose1xW</a></li>
595<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_matrix_vector_multiply_kernel.xhtml">CLGEMMMatrixVectorMultiplyKernel</a></li>
596<li><a class="el" href="classarm__compute_1_1_c_l_l2_normalize_layer_kernel.xhtml">CLL2NormalizeLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_l2_normalize_layer.xhtml">CLL2NormalizeLayer</a></li>
597<li><a class="el" href="classarm__compute_1_1_c_l_quantization_layer_kernel.xhtml">CLQuantizationLayerKernel</a> <a class="el" href="classarm__compute_1_1_c_l_min_max_layer_kernel.xhtml">CLMinMaxLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_quantization_layer.xhtml">CLQuantizationLayer</a></li>
598<li><a class="el" href="classarm__compute_1_1_c_l_r_o_i_pooling_layer_kernel.xhtml">CLROIPoolingLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_r_o_i_pooling_layer.xhtml">CLROIPoolingLayer</a></li>
599<li><a class="el" href="classarm__compute_1_1_c_l_reduction_operation_kernel.xhtml">CLReductionOperationKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_reduction_operation.xhtml">CLReductionOperation</a></li>
600<li><a class="el" href="classarm__compute_1_1_c_l_reshape_layer_kernel.xhtml">CLReshapeLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_reshape_layer.xhtml">CLReshapeLayer</a></li>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]601</ul>
602</li>
603</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]604<p>v17.06 Public major release</p><ul>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]605<li>Various bug fixes</li>
606<li>Added support for fixed point 8 bit (QS8) to the various NEON machine learning kernels.</li>
607<li>Added unit tests and benchmarks (AlexNet, LeNet)</li>
608<li>Added support for sub tensors.</li>
609<li>Added infrastructure to provide GPU specific optimisation for some OpenCL kernels.</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]610<li>Added <a class="el" href="classarm__compute_1_1_o_m_p_scheduler.xhtml">OMPScheduler</a> (OpenMP) scheduler for NEON</li>
611<li>Added <a class="el" href="classarm__compute_1_1_single_thread_scheduler.xhtml">SingleThreadScheduler</a> scheduler for NEON (For bare metal)</li>
612<li>User can specify his own scheduler by implementing the <a class="el" href="classarm__compute_1_1_i_scheduler.xhtml">IScheduler</a> interface.</li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]613<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]614<li><a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer_kernel.xhtml">CLBatchNormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_batch_normalization_layer.xhtml">CLBatchNormalizationLayer</a></li>
615<li><a class="el" href="classarm__compute_1_1_c_l_depth_concatenate_layer_kernel.xhtml">CLDepthConcatenateLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_depth_concatenate_layer.xhtml">CLDepthConcatenateLayer</a></li>
616<li><a class="el" href="classarm__compute_1_1_c_l_h_o_g_orientation_binning_kernel.xhtml">CLHOGOrientationBinningKernel</a> <a class="el" href="classarm__compute_1_1_c_l_h_o_g_block_normalization_kernel.xhtml">CLHOGBlockNormalizationKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_h_o_g_detector_kernel.xhtml">CLHOGDetectorKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_h_o_g_descriptor.xhtml">CLHOGDescriptor</a> <a class="el" href="classarm__compute_1_1_c_l_h_o_g_detector.xhtml">CLHOGDetector</a> <a class="el" href="classarm__compute_1_1_c_l_h_o_g_gradient.xhtml">CLHOGGradient</a> <a class="el" href="classarm__compute_1_1_c_l_h_o_g_multi_detection.xhtml">CLHOGMultiDetection</a></li>
617<li><a class="el" href="classarm__compute_1_1_c_l_locally_connected_matrix_multiply_kernel.xhtml">CLLocallyConnectedMatrixMultiplyKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_locally_connected_layer.xhtml">CLLocallyConnectedLayer</a></li>
618<li><a class="el" href="classarm__compute_1_1_c_l_weights_reshape_kernel.xhtml">CLWeightsReshapeKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_convolution_layer_reshape_weights.xhtml">CLConvolutionLayerReshapeWeights</a></li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]619</ul>
620</li>
621<li>New C++ kernels:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]622<li><a class="el" href="classarm__compute_1_1_c_p_p_detection_window_non_maxima_suppression_kernel.xhtml">CPPDetectionWindowNonMaximaSuppressionKernel</a></li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]623</ul>
624</li>
625<li>New NEON kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]626<li><a class="el" href="classarm__compute_1_1_n_e_batch_normalization_layer_kernel.xhtml">NEBatchNormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_batch_normalization_layer.xhtml">NEBatchNormalizationLayer</a></li>
627<li><a class="el" href="classarm__compute_1_1_n_e_depth_concatenate_layer_kernel.xhtml">NEDepthConcatenateLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_depth_concatenate_layer.xhtml">NEDepthConcatenateLayer</a></li>
628<li><a class="el" href="classarm__compute_1_1_n_e_direct_convolution_layer_kernel.xhtml">NEDirectConvolutionLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_direct_convolution_layer.xhtml">NEDirectConvolutionLayer</a></li>
629<li><a class="el" href="classarm__compute_1_1_n_e_locally_connected_matrix_multiply_kernel.xhtml">NELocallyConnectedMatrixMultiplyKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_locally_connected_layer.xhtml">NELocallyConnectedLayer</a></li>
630<li><a class="el" href="classarm__compute_1_1_n_e_weights_reshape_kernel.xhtml">NEWeightsReshapeKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_convolution_layer_reshape_weights.xhtml">NEConvolutionLayerReshapeWeights</a></li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]631</ul>
632</li>
633</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]634<p>v17.05 Public bug fixes release</p><ul>
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]635<li>Various bug fixes</li>
636<li>Remaining of the functions ported to use accurate padding.</li>
637<li>Library does not link against OpenCL anymore (It uses dlopen / dlsym at runtime instead to determine whether or not OpenCL is available).</li>
638<li>Added "free" method to allocator.</li>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]639<li>Minimum version of g++ required for armv7 Linux changed from 4.8 to 4.9</li>
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]640</ul>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]641<p>v17.04 Public bug fixes release</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]642<p>The following functions have been ported to use the new accurate padding:</p><ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]643<li><a class="el" href="classarm__compute_1_1_c_l_color_convert_kernel.xhtml">CLColorConvertKernel</a></li>
644<li><a class="el" href="classarm__compute_1_1_c_l_edge_non_max_suppression_kernel.xhtml">CLEdgeNonMaxSuppressionKernel</a></li>
645<li><a class="el" href="classarm__compute_1_1_c_l_edge_trace_kernel.xhtml">CLEdgeTraceKernel</a></li>
646<li><a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_hor_kernel.xhtml">CLGaussianPyramidHorKernel</a></li>
647<li><a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_vert_kernel.xhtml">CLGaussianPyramidVertKernel</a></li>
648<li><a class="el" href="classarm__compute_1_1_c_l_gradient_kernel.xhtml">CLGradientKernel</a></li>
649<li><a class="el" href="classarm__compute_1_1_n_e_channel_combine_kernel.xhtml">NEChannelCombineKernel</a></li>
650<li><a class="el" href="classarm__compute_1_1_n_e_fill_array_kernel.xhtml">NEFillArrayKernel</a></li>
651<li><a class="el" href="classarm__compute_1_1_n_e_gaussian_pyramid_hor_kernel.xhtml">NEGaussianPyramidHorKernel</a></li>
652<li><a class="el" href="classarm__compute_1_1_n_e_gaussian_pyramid_vert_kernel.xhtml">NEGaussianPyramidVertKernel</a></li>
653<li><a class="el" href="namespacearm__compute.xhtml#a0b6679b5d5c7f7dc527258181b04cf35">NEHarrisScoreFP16Kernel</a></li>
654<li><a class="el" href="classarm__compute_1_1_n_e_harris_score_kernel.xhtml">NEHarrisScoreKernel</a></li>
655<li><a class="el" href="classarm__compute_1_1_n_e_h_o_g_detector_kernel.xhtml">NEHOGDetectorKernel</a></li>
656<li><a class="el" href="classarm__compute_1_1_n_e_logits1_d_max_kernel.xhtml">NELogits1DMaxKernel</a></li>
657<li>NELogits1DShiftExpSumKernel</li>
658<li>NELogits1DNormKernel</li>
659<li><a class="el" href="namespacearm__compute.xhtml#a38cad49e6beaef76bc1ec5064c9e9dba">NENonMaximaSuppression3x3FP16Kernel</a></li>
660<li><a class="el" href="classarm__compute_1_1_n_e_non_maxima_suppression3x3_kernel.xhtml">NENonMaximaSuppression3x3Kernel</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]661</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]662<p>v17.03.1 First Major public release of the sources</p><ul>
663<li>Renamed the library to <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]664<li>New CPP target introduced for C++ kernels shared between NEON and CL functions.</li>
665<li>New padding calculation interface introduced and ported most kernels / functions to use it.</li>
666<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]667<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_lowp_matrix_multiply_kernel.xhtml">CLGEMMLowpMatrixMultiplyKernel</a> / CLGEMMLowp</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]668</ul>
669</li>
670<li>New NEON kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]671<li><a class="el" href="classarm__compute_1_1_n_e_normalization_layer_kernel.xhtml">NENormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_normalization_layer.xhtml">NENormalizationLayer</a></li>
672<li><a class="el" href="classarm__compute_1_1_n_e_transpose_kernel.xhtml">NETransposeKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_transpose.xhtml">NETranspose</a></li>
673<li><a class="el" href="classarm__compute_1_1_n_e_logits1_d_max_kernel.xhtml">NELogits1DMaxKernel</a>, NELogits1DShiftExpSumKernel, NELogits1DNormKernel / <a class="el" href="classarm__compute_1_1_n_e_softmax_layer.xhtml">NESoftmaxLayer</a></li>
674<li><a class="el" href="classarm__compute_1_1_n_e_im2_col_kernel.xhtml">NEIm2ColKernel</a>, <a class="el" href="classarm__compute_1_1_n_e_col2_im_kernel.xhtml">NECol2ImKernel</a>, NEConvolutionLayerWeightsReshapeKernel / <a class="el" href="classarm__compute_1_1_n_e_convolution_layer.xhtml">NEConvolutionLayer</a></li>
675<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_matrix_accumulate_biases_kernel.xhtml">NEGEMMMatrixAccumulateBiasesKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_fully_connected_layer.xhtml">NEFullyConnectedLayer</a></li>
676<li><a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_lowp_matrix_multiply_kernel.xhtml">NEGEMMLowpMatrixMultiplyKernel</a> / NEGEMMLowp</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]677</ul>
678</li>
679</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]680<p>v17.03 Sources preview</p><ul>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]681<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]682<li><a class="el" href="classarm__compute_1_1_c_l_gradient_kernel.xhtml">CLGradientKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_edge_non_max_suppression_kernel.xhtml">CLEdgeNonMaxSuppressionKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_edge_trace_kernel.xhtml">CLEdgeTraceKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_canny_edge.xhtml">CLCannyEdge</a></li>
683<li>GEMM refactoring + FP16 support: <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_interleave4x4_kernel.xhtml">CLGEMMInterleave4x4Kernel</a>, <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_transpose1x_w_kernel.xhtml">CLGEMMTranspose1xWKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_matrix_multiply_kernel.xhtml">CLGEMMMatrixMultiplyKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_matrix_addition_kernel.xhtml">CLGEMMMatrixAdditionKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_g_e_m_m.xhtml">CLGEMM</a></li>
684<li><a class="el" href="classarm__compute_1_1_c_l_g_e_m_m_matrix_accumulate_biases_kernel.xhtml">CLGEMMMatrixAccumulateBiasesKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_fully_connected_layer.xhtml">CLFullyConnectedLayer</a></li>
685<li><a class="el" href="classarm__compute_1_1_c_l_transpose_kernel.xhtml">CLTransposeKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_transpose.xhtml">CLTranspose</a></li>
686<li><a class="el" href="classarm__compute_1_1_c_l_l_k_tracker_init_kernel.xhtml">CLLKTrackerInitKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_l_k_tracker_stage0_kernel.xhtml">CLLKTrackerStage0Kernel</a>, <a class="el" href="classarm__compute_1_1_c_l_l_k_tracker_stage1_kernel.xhtml">CLLKTrackerStage1Kernel</a>, <a class="el" href="classarm__compute_1_1_c_l_l_k_tracker_finalize_kernel.xhtml">CLLKTrackerFinalizeKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_optical_flow.xhtml">CLOpticalFlow</a></li>
687<li><a class="el" href="classarm__compute_1_1_c_l_normalization_layer_kernel.xhtml">CLNormalizationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_normalization_layer.xhtml">CLNormalizationLayer</a></li>
688<li><a class="el" href="classarm__compute_1_1_c_l_laplacian_pyramid.xhtml">CLLaplacianPyramid</a>, <a class="el" href="classarm__compute_1_1_c_l_laplacian_reconstruct.xhtml">CLLaplacianReconstruct</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]689</ul>
690</li>
691<li>New NEON kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]692<li><a class="el" href="classarm__compute_1_1_n_e_activation_layer_kernel.xhtml">NEActivationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_activation_layer.xhtml">NEActivationLayer</a></li>
693<li>GEMM refactoring + FP16 support (Requires armv8.2 CPU): <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_interleave4x4_kernel.xhtml">NEGEMMInterleave4x4Kernel</a>, <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_transpose1x_w_kernel.xhtml">NEGEMMTranspose1xWKernel</a>, <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_matrix_multiply_kernel.xhtml">NEGEMMMatrixMultiplyKernel</a>, <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m_matrix_addition_kernel.xhtml">NEGEMMMatrixAdditionKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_g_e_m_m.xhtml">NEGEMM</a></li>
694<li><a class="el" href="classarm__compute_1_1_n_e_pooling_layer_kernel.xhtml">NEPoolingLayerKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_pooling_layer.xhtml">NEPoolingLayer</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]695</ul>
696</li>
697</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]698<p>v17.02.1 Sources preview</p><ul>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]699<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]700<li><a class="el" href="classarm__compute_1_1_c_l_logits1_d_max_kernel.xhtml">CLLogits1DMaxKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_logits1_d_shift_exp_sum_kernel.xhtml">CLLogits1DShiftExpSumKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_logits1_d_norm_kernel.xhtml">CLLogits1DNormKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_softmax_layer.xhtml">CLSoftmaxLayer</a></li>
701<li><a class="el" href="classarm__compute_1_1_c_l_pooling_layer_kernel.xhtml">CLPoolingLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_pooling_layer.xhtml">CLPoolingLayer</a></li>
702<li><a class="el" href="classarm__compute_1_1_c_l_im2_col_kernel.xhtml">CLIm2ColKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_col2_im_kernel.xhtml">CLCol2ImKernel</a>, CLConvolutionLayerWeightsReshapeKernel / <a class="el" href="classarm__compute_1_1_c_l_convolution_layer.xhtml">CLConvolutionLayer</a></li>
703<li><a class="el" href="classarm__compute_1_1_c_l_remap_kernel.xhtml">CLRemapKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_remap.xhtml">CLRemap</a></li>
704<li><a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_hor_kernel.xhtml">CLGaussianPyramidHorKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_vert_kernel.xhtml">CLGaussianPyramidVertKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid.xhtml">CLGaussianPyramid</a>, <a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_half.xhtml">CLGaussianPyramidHalf</a>, <a class="el" href="classarm__compute_1_1_c_l_gaussian_pyramid_orb.xhtml">CLGaussianPyramidOrb</a></li>
705<li><a class="el" href="classarm__compute_1_1_c_l_min_max_kernel.xhtml">CLMinMaxKernel</a>, <a class="el" href="classarm__compute_1_1_c_l_min_max_location_kernel.xhtml">CLMinMaxLocationKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_min_max_location.xhtml">CLMinMaxLocation</a></li>
706<li><a class="el" href="classarm__compute_1_1_c_l_non_linear_filter_kernel.xhtml">CLNonLinearFilterKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_non_linear_filter.xhtml">CLNonLinearFilter</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]707</ul>
708</li>
709<li>New NEON FP16 kernels (Requires armv8.2 CPU)<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]710<li><a class="el" href="namespacearm__compute.xhtml#aff99c045b07329b332b1cb97a2dd1518">NEAccumulateWeightedFP16Kernel</a></li>
711<li><a class="el" href="namespacearm__compute.xhtml#a96f7f6f98dc47e0dc3b928bf87397ebf">NEBox3x3FP16Kernel</a></li>
712<li><a class="el" href="namespacearm__compute.xhtml#a38cad49e6beaef76bc1ec5064c9e9dba">NENonMaximaSuppression3x3FP16Kernel</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]713</ul>
714</li>
715</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]716<p>v17.02 Sources preview</p><ul>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]717<li>New OpenCL kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]718<li><a class="el" href="classarm__compute_1_1_c_l_activation_layer_kernel.xhtml">CLActivationLayerKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_activation_layer.xhtml">CLActivationLayer</a></li>
719<li><a class="el" href="classarm__compute_1_1_c_l_channel_combine_kernel.xhtml">CLChannelCombineKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_channel_combine.xhtml">CLChannelCombine</a></li>
720<li><a class="el" href="classarm__compute_1_1_c_l_derivative_kernel.xhtml">CLDerivativeKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_channel_extract.xhtml">CLChannelExtract</a></li>
721<li><a class="el" href="classarm__compute_1_1_c_l_fast_corners_kernel.xhtml">CLFastCornersKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_fast_corners.xhtml">CLFastCorners</a></li>
722<li><a class="el" href="classarm__compute_1_1_c_l_mean_std_dev_kernel.xhtml">CLMeanStdDevKernel</a> / <a class="el" href="classarm__compute_1_1_c_l_mean_std_dev.xhtml">CLMeanStdDev</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]723</ul>
724</li>
725<li>New NEON kernels / functions:<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]726<li><a class="el" href="classarm__compute_1_1_h_o_g.xhtml" title="CPU implementation of HOG data-object. ">HOG</a> / SVM: <a class="el" href="classarm__compute_1_1_n_e_h_o_g_orientation_binning_kernel.xhtml">NEHOGOrientationBinningKernel</a>, <a class="el" href="classarm__compute_1_1_n_e_h_o_g_block_normalization_kernel.xhtml">NEHOGBlockNormalizationKernel</a>, <a class="el" href="classarm__compute_1_1_n_e_h_o_g_detector_kernel.xhtml">NEHOGDetectorKernel</a>, NEHOGNonMaximaSuppressionKernel / <a class="el" href="classarm__compute_1_1_n_e_h_o_g_descriptor.xhtml">NEHOGDescriptor</a>, <a class="el" href="classarm__compute_1_1_n_e_h_o_g_detector.xhtml">NEHOGDetector</a>, <a class="el" href="classarm__compute_1_1_n_e_h_o_g_gradient.xhtml">NEHOGGradient</a>, <a class="el" href="classarm__compute_1_1_n_e_h_o_g_multi_detection.xhtml">NEHOGMultiDetection</a></li>
727<li><a class="el" href="classarm__compute_1_1_n_e_non_linear_filter_kernel.xhtml">NENonLinearFilterKernel</a> / <a class="el" href="classarm__compute_1_1_n_e_non_linear_filter.xhtml">NENonLinearFilter</a></li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]728</ul>
729</li>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]730<li>Introduced a <a class="el" href="classarm__compute_1_1_c_l_scheduler.xhtml" title="Provides global access to a CL context and command queue. ">CLScheduler</a> to manage the default context and command queue used by the runtime library and create synchronisation events.</li>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]731<li>Switched all the kernels / functions to use tensors instead of images.</li>
732<li>Updated documentation to include instructions to build the library from sources.</li>
733</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]734<p>v16.12 Binary preview release</p><ul>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]735<li>Original release</li>
736</ul>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]737<h1><a class="anchor" id="S3_how_to_build"></a>
738How to build the library and the examples</h1>
739<h2><a class="anchor" id="S3_1_build_options"></a>
740Build options</h2>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]741<p>scons 2.3 or above is required to build the library. To see the build options available simply run <code>scons -h</code>: </p><pre class="fragment">debug: Debug (yes|no)
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]742 default: False
743 actual: False
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]744
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]745asserts: Enable asserts (this flag is forced to 1 for debug=1) (yes|no)
746 default: False
747 actual: False
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]748
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]749arch: Target Architecture (armv7a|arm64-v8a|arm64-v8.2-a|x86_32|x86_64)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]750 default: armv7a
751 actual: armv7a
752
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]753os: Target OS (linux|android|bare_metal)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]754 default: linux
755 actual: linux
756
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]757build: Build type (native|cross_compile|embed_only)
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]758 default: cross_compile
759 actual: cross_compile
760
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]761examples: Build example programs (yes|no)
762 default: True
763 actual: True
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]764
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]765Werror: Enable/disable the -Werror compilation flag (yes|no)
766 default: True
767 actual: True
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]768
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]769opencl: Enable OpenCL support (yes|no)
770 default: True
771 actual: True
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]772
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]773neon: Enable Neon support (yes|no)
774 default: False
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]775 actual: False
776
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]777gles_compute: Enable OpenGL ES Compute Shader support (yes|no)
778 default: False
779 actual: False
780
781embed_kernels: Embed OpenCL kernels and OpenGL ES compute shader in library binary (yes|no)
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]782 default: True
783 actual: True
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]784
785set_soname: Set the library's soname and shlibversion (requires SCons 2.4 or above) (yes|no)
786 default: False
787 actual: False
788
789openmp: Enable OpenMP backend (yes|no)
790 default: False
791 actual: False
792
793cppthreads: Enable C++11 threads backend (yes|no)
794 default: True
795 actual: True
796
797build_dir: Specify sub-folder for the build ( /path/to/build_dir )
798 default: .
799 actual: .
800
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]801extra_cxx_flags: Extra CXX flags to be appended to the build command
802 default:
803 actual:
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]804
805pmu: Enable PMU counters (yes|no)
806 default: False
807 actual: False
808
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]809mali: Enable Mali hardware counters (yes|no)
810 default: False
811 actual: False
812
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]813validation_tests: Build validation test programs (yes|no)
814 default: False
815 actual: False
816
817benchmark_tests: Build benchmark test programs (yes|no)
818 default: False
819 actual: False
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]820</pre><p><b>debug</b> / <b>asserts:</b> </p><ul>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]821<li>With debug=1 asserts are enabled, and the library is built with symbols and no optimisations enabled.</li>
822<li>With debug=0 and asserts=1: Optimisations are enabled and symbols are removed, however all the asserts are still present (This is about 20% slower than the release build)</li>
823<li>With debug=0 and asserts=0: All optimisations are enable and no validation is performed, if the application misuses the library it is likely to result in a crash. (Only use this mode once you are sure your application is working as expected).</li>
824</ul>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]825<p><b>arch:</b> The x86_32 and x86_64 targets can only be used with neon=0 and opencl=1.</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]826<p><b>os:</b> Choose the operating system you are targeting: Linux, Android or bare metal. </p><dl class="section note"><dt>Note</dt><dd>bare metal can only be used for NEON (not OpenCL), only static libraries get built and NEON's multi-threading support is disabled.</dd></dl>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]827<p><b>build:</b> you can either build directly on your device (native) or cross compile from your desktop machine (cross-compile). In both cases make sure the compiler is available in your path.</p>
828<dl class="section note"><dt>Note</dt><dd>If you want to natively compile for 32bit on a 64bit ARM device running a 64bit OS then you will have to use cross-compile too.</dd></dl>
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]829<p>There is also an 'embed_only' option which will generate all the .embed files for the OpenCL kernels and / or OpenGLES compute shaders. This might be useful if using a different build system to compile the library.</p>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]830<p><b>Werror:</b> If you are compiling using the same toolchains as the ones used in this guide then there shouldn't be any warning and therefore you should be able to keep Werror=1. If with a different compiler version the library fails to build because of warnings interpreted as errors then, if you are sure the warnings are not important, you might want to try to build with Werror=0 (But please do report the issue either on Github or by an email to <a href="#" onclick="location.href='mai'+'lto:'+'dev'+'el'+'ope'+'r@'+'arm'+'.c'+'om'; return false;">devel<span style="display: none;">.nosp@m.</span>oper<span style="display: none;">.nosp@m.</span>@arm.<span style="display: none;">.nosp@m.</span>com</a> so that the issue can be addressed).</p>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]831<p><b>opencl</b> / <b>neon</b> / <b>gles_compute:</b> Choose which SIMD technology you want to target. (NEON for ARM Cortex-A CPUs or OpenCL / GLES_COMPUTE for ARM Mali GPUs)</p>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]832<p><b>embed_kernels:</b> For OpenCL / GLES_COMPUTE only: set embed_kernels=1 if you want the OpenCL / GLES_COMPUTE kernels to be built in the library's binaries instead of being read from separate ".cl" / ".cs" files. If embed_kernels is set to 0 then the application can set the path to the folder containing the OpenCL / GLES_COMPUTE kernel files by calling <a class="el" href="classarm__compute_1_1_c_l_kernel_library.xhtml#af353532ea782387df6bcb6d01894f4ae" title="Initialises the kernel library. ">CLKernelLibrary::init()</a> / <a class="el" href="classarm__compute_1_1_g_c_kernel_library.xhtml#abe24625d55f2fb35da7e293e5e28d483" title="Initialises the kernel library. ">GCKernelLibrary::init()</a>. By default the path is set to "./cl_kernels" / "./cs_shaders".</p>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]833<p><b>set_soname:</b> Do you want to build the versioned version of the library ?</p>
834<p>If enabled the library will contain a SONAME and SHLIBVERSION and some symlinks will automatically be created between the objects. Example: libarm_compute_core.so -&gt; libarm_compute_core.so.1.0.0 libarm_compute_core.so.1 -&gt; libarm_compute_core.so.1.0.0 libarm_compute_core.so.1.0.0</p>
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]835<dl class="section note"><dt>Note</dt><dd>This options is disabled by default as it requires SCons version 2.4 or above.</dd></dl>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]836<p><b>extra_cxx_flags:</b> Custom CXX flags which will be appended to the end of the build command.</p>
837<p><b>build_dir:</b> Build the library in a subfolder of the "build" folder. (Allows to build several configurations in parallel).</p>
838<p><b>examples:</b> Build or not the examples</p>
839<p><b>validation_tests:</b> Enable the build of the validation suite.</p>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]840<p><b>benchmark_tests:</b> Enable the build of the benchmark tests</p>
841<p><b>pmu:</b> Enable the PMU cycle counter to measure execution time in benchmark tests. (Your device needs to support it)</p>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]842<p><b>mali:</b> Enable the collection of Mali hardware counters to measure execution time in benchmark tests. (Your device needs to have a Mali driver that supports it)</p>
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]843<p><b>openmp</b> Build in the OpenMP scheduler for NEON.</p>
844<dl class="section note"><dt>Note</dt><dd>Only works when building with g++ not clang++</dd></dl>
845<p><b>cppthreads</b> Build in the C++11 scheduler for NEON.</p>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]846<dl class="section see"><dt>See also</dt><dd><a class="el" href="classarm__compute_1_1_scheduler.xhtml#a12775a7fbfa126fa4f9f06f8e02d9a8e" title="Sets the user defined scheduler and makes it the active scheduler. ">Scheduler::set</a></dd></dl>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]847<h2><a class="anchor" id="S3_2_linux"></a>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]848Building for Linux</h2>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]849<h3><a class="anchor" id="S3_2_1_library"></a>
850How to build the library ?</h3>
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]851<p>For Linux, the library was successfully built and tested using the following Linaro GCC toolchain:</p>
852<ul>
853<li>gcc-linaro-arm-linux-gnueabihf-4.9-2014.07_linux</li>
854<li>gcc-linaro-4.9-2016.02-x86_64_aarch64-linux-gnu</li>
855<li>gcc-linaro-6.3.1-2017.02-i686_aarch64-linux-gnu</li>
856</ul>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]857<p>To cross-compile the library in debug mode, with NEON only support, for Linux 32bit: </p><pre class="fragment">scons Werror=1 -j8 debug=1 neon=1 opencl=0 os=linux arch=armv7a
858</pre><p>To cross-compile the library in asserts mode, with OpenCL only support, for Linux 64bit: </p><pre class="fragment">scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=1 embed_kernels=1 os=linux arch=arm64-v8a
859</pre><p>To cross-compile the library in asserts mode, with GLES_COMPUTE only support, for Linux 64bit: </p><pre class="fragment">scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=0 gles_compute=1 embed_kernels=1 os=linux arch=arm64-v8a
860</pre><p>You can also compile the library natively on an ARM device by using <b>build=native</b>: </p><pre class="fragment">scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=arm64-v8a build=native
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]861scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a build=native
Anthony Barbierdbdab852017-06-23 15:42:00 +0100[diff] [blame]862</pre><dl class="section note"><dt>Note</dt><dd>g++ for ARM is mono-arch, therefore if you want to compile for Linux 32bit on a Linux 64bit platform you will have to use a cross compiler.</dd></dl>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]863<p>For example on a 64bit Debian based system you would have to install <b>g++-arm-linux-gnueabihf</b> </p><pre class="fragment">apt-get install g++-arm-linux-gnueabihf
864</pre><p>Then run </p><pre class="fragment">scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a build=cross_compile
865</pre><p>or simply remove the build parameter as build=cross_compile is the default value: </p><pre class="fragment">scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a
Anthony Barbiera4376382017-04-12 15:12:46 +0100[diff] [blame]866</pre><dl class="section attention"><dt>Attention</dt><dd>To cross compile with opencl=1 you need to make sure to have a version of libOpenCL matching your target architecture.</dd></dl>
867<h3><a class="anchor" id="S3_2_2_examples"></a>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]868How to manually build the examples ?</h3>
869<p>The examples get automatically built by scons as part of the build process of the library described above. This section just describes how you can build and link your own application against our library.</p>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]870<dl class="section note"><dt>Note</dt><dd>The following command lines assume the <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> binaries are present in the current directory or in the system library path. If this is not the case you can specify the location of the pre-built library with the compiler option -L. When building the OpenCL example the commands below assume that the CL headers are located in the include folder where the command is executed.</dd></dl>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]871<p>To cross compile a NEON example for Linux 32bit: </p><pre class="fragment">arm-linux-gnueabihf-g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute -larm_compute_core -o neon_convolution
872</pre><p>To cross compile a NEON example for Linux 64bit: </p><pre class="fragment">aarch64-linux-gnu-g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]873</pre><p>(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)</p>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]874<p>To cross compile an OpenCL example for Linux 32bit: </p><pre class="fragment">arm-linux-gnueabihf-g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
875</pre><p>To cross compile an OpenCL example for Linux 64bit: </p><pre class="fragment">aarch64-linux-gnu-g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]876</pre><p>To cross compile a GLES example for Linux 32bit: </p><pre class="fragment">arm-linux-gnueabihf-g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -mfpu=neon -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
877</pre><p>To cross compile a GLES example for Linux 64bit: </p><pre class="fragment">aarch64-linux-gnu-g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]878</pre><p>(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]879<p>To cross compile the examples with the Graph API, such as <a class="el" href="graph__lenet_8cpp.xhtml">graph_lenet.cpp</a>, you need to link the examples against arm_compute_graph.so too.</p>
880<dl class="section note"><dt>Note</dt><dd>The compute library must currently be built with both neon and opencl enabled - neon=1 and opencl=1</dd></dl>
881<p>i.e. to cross compile the "graph_lenet" example for Linux 32bit: </p><pre class="fragment">arm-linux-gnueabihf-g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
882</pre><p>i.e. to cross compile the "graph_lenet" example for Linux 64bit: </p><pre class="fragment">aarch64-linux-gnu-g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]883</pre><p>(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]884<dl class="section note"><dt>Note</dt><dd>If compiling using static libraries, this order must be followed when linking: arm_compute_graph_static, <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a>, arm_compute_core</dd></dl>
885<p>To compile natively (i.e directly on an ARM device) for NEON for Linux 32bit: </p><pre class="fragment">g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -larm_compute -larm_compute_core -o neon_convolution
886</pre><p>To compile natively (i.e directly on an ARM device) for NEON for Linux 64bit: </p><pre class="fragment">g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier46d59272017-05-04 09:15:15 +0100[diff] [blame]887</pre><p>(notice the only difference with the 32 bit command is that we don't need the -mfpu option)</p>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]888<p>To compile natively (i.e directly on an ARM device) for OpenCL for Linux 32bit or Linux 64bit: </p><pre class="fragment">g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]889</pre><p>To compile natively (i.e directly on an ARM device) for GLES for Linux 32bit or Linux 64bit: </p><pre class="fragment">g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
890</pre><p>To compile natively the examples with the Graph API, such as <a class="el" href="graph__lenet_8cpp.xhtml">graph_lenet.cpp</a>, you need to link the examples against arm_compute_graph.so too. </p><dl class="section note"><dt>Note</dt><dd>The compute library must currently be built with both neon and opencl enabled - neon=1 and opencl=1</dd></dl>
891<p>i.e. to natively compile the "graph_lenet" example for Linux 32bit: </p><pre class="fragment">g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
892</pre><p>i.e. to natively compile the "graph_lenet" example for Linux 64bit: </p><pre class="fragment">g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]893</pre><p>(notice the only difference with the 32 bit command is that we don't need the -mfpu option)</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]894<dl class="section note"><dt>Note</dt><dd>If compiling using static libraries, this order must be followed when linking: arm_compute_graph_static, <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a>, arm_compute_core</dd>
895<dd>
896These two commands assume libarm_compute.so is available in your library path, if not add the path to it using -L</dd></dl>
897<p>To run the built executable simply run: </p><pre class="fragment">LD_LIBRARY_PATH=build ./neon_convolution
898</pre><p>or </p><pre class="fragment">LD_LIBRARY_PATH=build ./cl_convolution
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]899</pre><dl class="section note"><dt>Note</dt><dd>Examples accept different types of arguments, to find out what they are run the example without any argument and the help will be displayed at the beginning of the run.</dd></dl>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]900<p>For example: </p><pre class="fragment">LD_LIBRARY_PATH=. ./graph_lenet
901
902./graph_lenet
903
904Usage: ./graph_lenet [target] [path_to_data] [batches]
905
906No data folder provided: using random values
907
908Test passed
909</pre><p>In this case the first argument of LeNet (like all the graph examples) is the target (i.e 0 to run on NEON, 1 to run on OpenCL if available, 2 to run on OpenCL using the <a class="el" href="classarm__compute_1_1_c_l_tuner.xhtml" title="Basic implementation of the OpenCL tuner interface. ">CLTuner</a>), the second argument is the path to the folder containing the npy files for the weights and finally the third argument is the number of batches to run.</p>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]910<h2><a class="anchor" id="S3_3_android"></a>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]911Building for Android</h2>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]912<p>For Android, the library was successfully built and tested using Google's standalone toolchains:</p><ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]913<li>clang++ from NDK r16b for armv7a</li>
914<li>clang++ from NDK r16b for arm64-v8a</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]915</ul>
916<p>Here is a guide to <a href="https://developer.android.com/ndk/guides/standalone_toolchain.html">create your Android standalone toolchains from the NDK</a></p>
917<ul>
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]918<li>Download the NDK r16b from here: <a href="https://developer.android.com/ndk/downloads/index.html">https://developer.android.com/ndk/downloads/index.html</a></li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]919<li>Make sure you have Python 2 installed on your machine.</li>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]920<li>Generate the 32 and/or 64 toolchains by running the following commands:</li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]921</ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]922<pre class="fragment">$NDK/build/tools/make_standalone_toolchain.py --arch arm64 --install-dir $MY_TOOLCHAINS/aarch64-linux-android-ndk-r16b --stl gnustl --api 21
923$NDK/build/tools/make_standalone_toolchain.py --arch arm --install-dir $MY_TOOLCHAINS/arm-linux-android-ndk-r16b --stl gnustl --api 21
924</pre><dl class="section attention"><dt>Attention</dt><dd>Due to some NDK issues make sure you use clang++ &amp; gnustl</dd></dl>
925<dl class="section note"><dt>Note</dt><dd>Make sure to add the toolchains to your PATH: <pre class="fragment">export PATH=$PATH:$MY_TOOLCHAINS/aarch64-linux-android-ndk-r16b/bin:$MY_TOOLCHAINS/arm-linux-android-ndk-r16b/bin
926</pre></dd></dl>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]927<h3><a class="anchor" id="S3_3_1_library"></a>
928How to build the library ?</h3>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]929<p>To cross-compile the library in debug mode, with NEON only support, for Android 32bit: </p><pre class="fragment">CXX=clang++ CC=clang scons Werror=1 -j8 debug=1 neon=1 opencl=0 os=android arch=armv7a
930</pre><p>To cross-compile the library in asserts mode, with OpenCL only support, for Android 64bit: </p><pre class="fragment">CXX=clang++ CC=clang scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=1 embed_kernels=1 os=android arch=arm64-v8a
931</pre><p>To cross-compile the library in asserts mode, with GLES_COMPUTE only support, for Android 64bit: </p><pre class="fragment">CXX=clang++ CC=clang scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=0 gles_compute=1 embed_kernels=1 os=android arch=arm64-v8a
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]932</pre><h3><a class="anchor" id="S3_3_2_examples"></a>
933How to manually build the examples ?</h3>
934<p>The examples get automatically built by scons as part of the build process of the library described above. This section just describes how you can build and link your own application against our library.</p>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]935<dl class="section note"><dt>Note</dt><dd>The following command lines assume the <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> binaries are present in the current directory or in the system library path. If this is not the case you can specify the location of the pre-built library with the compiler option -L. When building the OpenCL example the commands below assume that the CL headers are located in the include folder where the command is executed.</dd></dl>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]936<p>Once you've got your Android standalone toolchain built and added to your path you can do the following:</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]937<p>To cross compile a NEON example: </p><pre class="fragment">#32 bit:
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]938arm-linux-androideabi-clang++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o neon_convolution_arm -static-libstdc++ -pie
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]939#64 bit:
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]940aarch64-linux-android-clang++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o neon_convolution_aarch64 -static-libstdc++ -pie
941</pre><p>To cross compile an OpenCL example: </p><pre class="fragment">#32 bit:
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]942arm-linux-androideabi-clang++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o cl_convolution_arm -static-libstdc++ -pie -DARM_COMPUTE_CL
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]943#64 bit:
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]944aarch64-linux-android-clang++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o cl_convolution_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_CL
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]945</pre><p>To cross compile a GLES example: </p><pre class="fragment">#32 bit:
946arm-linux-androideabi-clang++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o gc_absdiff_arm -static-libstdc++ -pie -DARM_COMPUTE_GC
947#64 bit:
948aarch64-linux-android-clang++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o gc_absdiff_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_GC
949</pre><p>To cross compile the examples with the Graph API, such as <a class="el" href="graph__lenet_8cpp.xhtml">graph_lenet.cpp</a>, you need to link the library arm_compute_graph also. (notice the compute library has to be built with both neon and opencl enabled - neon=1 and opencl=1) </p><pre class="fragment">#32 bit:
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]950arm-linux-androideabi-clang++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -Wl,--whole-archive -larm_compute_graph-static -Wl,--no-whole-archive -larm_compute-static -larm_compute_core-static -L. -o graph_lenet_arm -static-libstdc++ -pie -DARM_COMPUTE_CL
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]951#64 bit:
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]952aarch64-linux-android-clang++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -Wl,--whole-archive -larm_compute_graph-static -Wl,--no-whole-archive -larm_compute-static -larm_compute_core-static -L. -o graph_lenet_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_CL
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]953</pre><dl class="section note"><dt>Note</dt><dd>Due to some issues in older versions of the Mali OpenCL DDK (&lt;= r13p0), we recommend to link <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> statically on Android. </dd>
954<dd>
955When linked statically the arm_compute_graph library currently needs the &ndash;whole-archive linker flag in order to work properly</dd></dl>
956<p>Then you need to do is upload the executable and the shared library to the device using ADB: </p><pre class="fragment">adb push neon_convolution_arm /data/local/tmp/
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]957adb push cl_convolution_arm /data/local/tmp/
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]958adb push gc_absdiff_arm /data/local/tmp/
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]959adb shell chmod 777 -R /data/local/tmp/
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]960</pre><p>And finally to run the example: </p><pre class="fragment">adb shell /data/local/tmp/neon_convolution_arm
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]961adb shell /data/local/tmp/cl_convolution_arm
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]962adb shell /data/local/tmp/gc_absdiff_arm
963</pre><p>For 64bit: </p><pre class="fragment">adb push neon_convolution_aarch64 /data/local/tmp/
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]964adb push cl_convolution_aarch64 /data/local/tmp/
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]965adb push gc_absdiff_aarch64 /data/local/tmp/
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]966adb shell chmod 777 -R /data/local/tmp/
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]967</pre><p>And finally to run the example: </p><pre class="fragment">adb shell /data/local/tmp/neon_convolution_aarch64
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]968adb shell /data/local/tmp/cl_convolution_aarch64
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]969adb shell /data/local/tmp/gc_absdiff_aarch64
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]970</pre><dl class="section note"><dt>Note</dt><dd>Examples accept different types of arguments, to find out what they are run the example without any argument and the help will be displayed at the beginning of the run.</dd></dl>
971<p>For example: adb shell /data/local/tmp/graph_lenet</p>
972<p>/data/local/tmp/graph_lenet</p>
973<p>Usage: /data/local/tmp/graph_lenet [target] [path_to_data] [batches]</p>
974<p>No data folder provided: using random values</p>
975<p>Test passed</p>
976<p>In this case the first argument of LeNet (like all the graph examples) is the target (i.e 0 to run on NEON, 1 to run on OpenCL if available, 2 to run on OpenCL using the <a class="el" href="classarm__compute_1_1_c_l_tuner.xhtml" title="Basic implementation of the OpenCL tuner interface. ">CLTuner</a>), the second argument is the path to the folder containing the npy files for the weights and finally the third argument is the number of batches to run.</p>
977<h2><a class="anchor" id="S3_4_bare_metal"></a>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]978Building for bare metal</h2>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]979<p>For bare metal, the library was successfully built using linaros's latest (gcc-linaro-6.3.1-2017.05) bare metal toolchains:</p><ul>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]980<li>arm-eabi for armv7a</li>
981<li>aarch64-elf for arm64-v8a</li>
982</ul>
983<p>Download linaro for <a href="https://releases.linaro.org/components/toolchain/binaries/6.3-2017.05/arm-eabi/">armv7a</a> and <a href="https://releases.linaro.org/components/toolchain/binaries/6.3-2017.05/aarch64-elf/">arm64-v8a</a>.</p>
984<dl class="section note"><dt>Note</dt><dd>Make sure to add the toolchains to your PATH: export PATH=$PATH:$MY_TOOLCHAINS/gcc-linaro-6.3.1-2017.05-x86_64_aarch64-elf/bin:$MY_TOOLCHAINS/gcc-linaro-6.3.1-2017.05-x86_64_arm-eabi/bin</dd></dl>
985<h3><a class="anchor" id="S3_4_1_library"></a>
986How to build the library ?</h3>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]987<p>To cross-compile the library with NEON support for baremetal arm64-v8a: </p><pre class="fragment">scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=bare_metal arch=arm64-v8a build=cross_compile cppthreads=0 openmp=0 standalone=1
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]988</pre><h3><a class="anchor" id="S3_4_2_examples"></a>
989How to manually build the examples ?</h3>
990<p>Examples are disabled when building for bare metal. If you want to build the examples you need to provide a custom bootcode depending on the target architecture and link against the compute library. More information about bare metal bootcode can be found <a href="http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dai0527a/index.html">here</a>.</p>
991<h2><a class="anchor" id="S3_5_windows_host"></a>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]992Building on a Windows host system</h2>
993<p>Using <code>scons</code> directly from the Windows command line is known to cause problems. The reason seems to be that if <code>scons</code> is setup for cross-compilation it gets confused about Windows style paths (using backslashes). Thus it is recommended to follow one of the options outlined below.</p>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]994<h3><a class="anchor" id="S3_5_1_ubuntu_on_windows"></a>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]995Bash on Ubuntu on Windows</h3>
996<p>The best and easiest option is to use <a href="https://msdn.microsoft.com/en-gb/commandline/wsl/about">Ubuntu on Windows</a>. This feature is still marked as <em>beta</em> and thus might not be available. However, if it is building the library is as simple as opening a <em>Bash on Ubuntu on Windows</em> shell and following the general guidelines given above.</p>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]997<h3><a class="anchor" id="S3_5_2_cygwin"></a>
Kaizen8938bd32017-09-28 14:38:23 +0100[diff] [blame]998Cygwin</h3>
999<p>If the Windows subsystem for Linux is not available <a href="https://www.cygwin.com/">Cygwin</a> can be used to install and run <code>scons</code>. In addition to the default packages installed by Cygwin <code>scons</code> has to be selected in the installer. (<code>git</code> might also be useful but is not strictly required if you already have got the source code of the library.) Linaro provides pre-built versions of <a href="http://releases.linaro.org/components/toolchain/binaries/">GCC cross-compilers</a> that can be used from the Cygwin terminal. When building for Android the compiler is included in the Android standalone toolchain. After everything has been set up in the Cygwin terminal the general guide on building the library can be followed.</p>
Kaizenbf8b01d2017-10-12 14:26:51 +0100[diff] [blame]1000<h2><a class="anchor" id="S3_6_cl_stub_library"></a>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1001The OpenCL stub library</h2>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]1002<p>In the opencl-1.2-stubs folder you will find the sources to build a stub OpenCL library which then can be used to link your application or <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> against.</p>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1003<p>If you preferred you could retrieve the OpenCL library from your device and link against this one but often this library will have dependencies on a range of system libraries forcing you to link your application against those too even though it is not using them.</p>
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]1004<dl class="section warning"><dt>Warning</dt><dd>This OpenCL library provided is a stub and <em>not</em> a real implementation. You can use it to resolve OpenCL's symbols in <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> while building the example but you must make sure the real libOpenCL.so is in your PATH when running the example or it will not work.</dd></dl>
1005<p>To cross-compile the stub OpenCL library simply run: </p><pre class="fragment">&lt;target-prefix&gt;-gcc -o libOpenCL.so -Iinclude opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
1006</pre><p>For example: </p><pre class="fragment">#Linux 32bit
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1007arm-linux-gnueabihf-gcc -o libOpenCL.so -Iinclude opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
1008#Linux 64bit
1009aarch64-linux-gnu-gcc -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC
1010#Android 32bit
1011arm-linux-androideabi-clang -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
1012#Android 64bit
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]1013aarch64-linux-android-clang -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
1014</pre><h2><a class="anchor" id="S3_7_gles_stub_library"></a>
1015The Linux OpenGLES and EGL stub libraries</h2>
1016<p>In the opengles-3.1-stubs folder you will find the sources to build stub EGL and OpenGLES libraries which then can be used to link your Linux application of <a class="el" href="namespacearm__compute.xhtml" title="This file contains all available output stages for GEMMLowp on OpenCL. ">arm_compute</a> against.</p>
1017<dl class="section note"><dt>Note</dt><dd>The stub libraries are only needed on Linux. For Android, the NDK toolchains already provide the meta-EGL and meta-GLES libraries.</dd></dl>
1018<p>To cross-compile the stub OpenGLES and EGL libraries simply run: </p><pre class="fragment">&lt;target-prefix&gt;-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
1019&lt;target-prefix&gt;-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared
1020
1021#Linux 32bit
1022arm-linux-gnueabihf-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
1023arm-linux-gnueabihf-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared
1024
1025#Linux 64bit
1026aarch64-linux-gnu-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
1027aarch64-linux-gnu-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared</pre> </div></div><!-- contents -->
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1028</div><!-- doc-content -->
1029<!-- start footer part -->
1030<div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
1031 <ul>
Jenkinsb3a371b2018-05-23 11:36:53 +0100[diff] [blame^]1032 <li class="footer">Generated on Wed May 23 2018 11:36:45 for Compute Library by
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1033 <a href="http://www.doxygen.org/index.html">
Anthony Barbier8140e1e2017-12-14 23:48:46 +0000[diff] [blame]1034 <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.11 </li>
Anthony Barbier871448e2017-03-24 14:54:29 +0000[diff] [blame]1035 </ul>
1036</div>
1037</body>
1038</html>