1 # How to Cross-Compile ArmNN on x86_64 for arm64
3 * [Introduction](#introduction)
4 * [Cross-compiling ToolChain](#installCCT)
5 * [Build and install Google's Protobuf library](#buildProtobuf)
6 * [Build Caffe for x86_64](#buildCaffe)
7 * [Build Boost library for arm64](#installBaarch)
8 * [Build Compute Library](#buildCL)
9 * [Build ArmNN](#buildANN)
10 * [Run Unit Tests](#unittests)
11 * [Troubleshooting and Errors](#troubleshooting)
14 #### <a name="introduction">Introduction</a>
15 These are the step by step instructions on Cross-Compiling ArmNN under an x86_64 system to target an Arm64 system. This build flow has been tested with Ubuntu 16.04.
16 The instructions show how to build the ArmNN core library and the Boost, Protobuf, Caffe and Compute Libraries necessary for compilation.
18 #### <a name="installCCT">Cross-compiling ToolChain</a>
19 * Install the standard cross-compilation libraries for arm64:
21 sudo apt install crossbuild-essential-arm64
24 #### <a name="buildProtobuf">Build and install Google's Protobuf library</a>
26 * Get protobuf-all-3.5.1.tar.gz from here: https://github.com/protocolbuffers/protobuf/releases/tag/v3.5.1
29 tar -zxvf protobuf-all-3.5.1.tar.gz
32 * Build a native (x86_64) version of the protobuf libraries and compiler (protoc):
33 (Requires cUrl, autoconf, llibtool, and other build dependencies if not previously installed: sudo apt install curl autoconf libtool build-essential g++)
37 ../configure --prefix=$HOME/armnn-devenv/google/x86_64_pb_install
41 * Build the arm64 version of the protobuf libraries:
45 CC=aarch64-linux-gnu-gcc \
46 CXX=aarch64-linux-gnu-g++ \
47 ../configure --host=aarch64-linux \
48 --prefix=$HOME/armnn-devenv/google/arm64_pb_install \
49 --with-protoc=$HOME/armnn-devenv/google/x86_64_pb_install/bin/protoc
54 #### <a name="buildCaffe">Build Caffe for x86_64</a>
55 * Ubuntu 16.04 installation. These steps are taken from the full Caffe installation documentation at: http://caffe.berkeleyvision.org/install_apt.html
56 * Install dependencies:
58 sudo apt-get install libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev
59 sudo apt-get install --no-install-recommends libboost-all-dev
60 sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev
61 sudo apt-get install libopenblas-dev
62 sudo apt-get install libatlas-base-dev
64 * Download Caffe-Master from: https://github.com/BVLC/caffe
66 git clone https://github.com/BVLC/caffe.git
68 cp Makefile.config.example Makefile.config
70 * Adjust Makefile.config as necessary for your environment, for example:
75 #Add hdf5 and protobuf include and library directories (Replace $HOME with explicit /home/username dir):
76 INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/ $HOME/armnn-devenv/google/x86_64_pb_install/include/
77 LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/ $HOME/armnn-devenv/google/x86_64_pb_install/lib/
81 export PATH=$HOME/armnn-devenv/google/x86_64_pb_install/bin/:$PATH
82 export LD_LIBRARY_PATH=$HOME/armnn-devenv/google/x86_64_pb_install/lib/:$LD_LIBRARY_PATH
84 * Compilation with Make:
90 These should all run without errors
91 * caffe.pb.h and caffe.pb.cc will be needed when building ArmNN's Caffe Parser
93 #### <a name="installBaarch">Build Boost library for arm64</a>
94 * Build Boost library for arm64
95 Download Boost version 1.64 from http://www.boost.org/doc/libs/1_64_0/more/getting_started/unix-variants.html
96 Using any version of Boost greater than 1.64 will fail to build ArmNN, due to different dependency issues.
98 tar -zxvf boost_1_64_0.tar.gz
100 echo "using gcc : arm : aarch64-linux-gnu-g++ ;" > user_config.jam
101 ./bootstrap.sh --prefix=$HOME/armnn-devenv/boost_arm64_install
102 ./b2 install toolset=gcc-arm link=static cxxflags=-fPIC --with-filesystem --with-test --with-log --with-program_options -j32 --user-config=user_config.jam
105 #### <a name="buildCL">Build Compute Library</a>
106 * Building the Arm Compute Library:
108 git clone https://github.com/ARM-software/ComputeLibrary.git
110 scons arch=arm64-v8a neon=1 opencl=1 embed_kernels=1 extra_cxx_flags="-fPIC" -j8 internal_only=0
113 #### <a name="buildANN">Build ArmNN</a>
114 * Compile ArmNN for arm64:
116 git clone https://github.com/ARM-software/armnn.git
122 * Use CMake to configure your build environment, update the following script and run it from the armnn/build directory to set up the armNN build:
125 CXX=aarch64-linux-gnu-g++ \
126 CC=aarch64-linux-gnu-gcc \
128 -DARMCOMPUTE_ROOT=$HOME/armnn-devenv/ComputeLibrary \
129 -DARMCOMPUTE_BUILD_DIR=$HOME/armnn-devenv/ComputeLibrary/build/ \
130 -DBOOST_ROOT=$HOME/armnn-devenv/boost_arm64_install/ \
131 -DARMCOMPUTENEON=1 -DARMCOMPUTECL=1 -DARMNNREF=1 \
132 -DCAFFE_GENERATED_SOURCES=$HOME/armnn-devenv/caffe/build/src \
133 -DBUILD_CAFFE_PARSER=1 \
134 -DPROTOBUF_ROOT=$HOME/armnn-devenv/google/x86_64_pb_install/ \
135 -DPROTOBUF_LIBRARY_DEBUG=$HOME/armnn-devenv/google/arm64_pb_install/lib/libprotobuf.so.15.0.1 \
136 -DPROTOBUF_LIBRARY_RELEASE=$HOME/armnn-devenv/google/arm64_pb_install/lib/libprotobuf.so.15.0.1
139 * If you want to include standalone sample dynamic backend tests, add the argument to enable the tests and the dynamic backend path to the CMake command:
141 -DSAMPLE_DYNAMIC_BACKEND=1 \
142 -DDYNAMIC_BACKEND_PATHS=$SAMPLE_DYNAMIC_BACKEND_PATH
149 #### <a name="buildStandaloneBackend">Build Standalone Sample Dynamic Backend</a>
150 * The sample dynamic backend is located in armnn/src/dynamic/sample
156 * Use CMake to configure your build environment, update the following script and run it from the armnn/src/dynamic/sample/build directory to set up the armNN build:
159 CXX=aarch64-linux-gnu-g++ \
160 CC=aarch64-linux-gnu-gcc \
162 -DCMAKE_CXX_FLAGS=--std=c++14 \
163 -DBOOST_ROOT=$HOME/armnn-devenv/boost_arm64_install/ \
164 -DBoost_SYSTEM_LIBRARY=$HOME/armnn-devenv/boost_arm64_install/lib/libboost_system.a \
165 -DBoost_FILESYSTEM_LIBRARY=$HOME/armnn-devenv/boost_arm64_install/lib/libboost_filesystem.a \
166 -DARMNN_PATH=$HOME/armnn-devenv/armnn/build/libarmnn.so
174 #### <a name="unittests">Run Unit Tests</a>
175 * Copy the build folder to an arm64 linux machine
176 * Copy the libprotobuf.so.15.0.1 library file to the build folder
177 * If you enable the standalone sample dynamic tests, also copy libArm_SampleDynamic_backend.so library file to the folder specified as $SAMPLE_DYNAMIC_BACKEND_PATH when you build ArmNN
178 * cd to the build folder on your arm64 machine and set your LD_LIBRARY_PATH to its current location:
181 export LD_LIBRARY_PATH=`pwd`
183 * Create a symbolic link to libprotobuf.so.15.0.1:
185 ln -s libprotobuf.so.15.0.1 ./libprotobuf.so.15
190 Running 567 test cases...
192 *** No errors detected
194 #### <a name="troubleshooting">Troubleshooting and Errors:</a>
195 #### Error adding symbols: File in wrong format
196 * When building armNN:
198 /usr/local/lib/libboost_log.a: error adding symbols: File in wrong format
199 collect2: error: ld returned 1 exit status
200 CMakeFiles/armnn.dir/build.make:4028: recipe for target 'libarmnn.so' failed
201 make[2]: *** [libarmnn.so] Error 1
202 CMakeFiles/Makefile2:105: recipe for target 'CMakeFiles/armnn.dir/all' failed
203 make[1]: *** [CMakeFiles/armnn.dir/all] Error 2
204 Makefile:127: recipe for target 'all' failed
205 make: *** [all] Error 2
207 * Boost libraries are not compiled for the correct architecture, try recompiling for arm64
209 #### Virtual memory exhausted
210 * When compiling the boost libraries:
212 virtual memory exhausted: Cannot allocate memory
214 * Not enough memory available to compile. Increase the amount of RAM or swap space available.
217 #### Unrecognized command line option '-m64'
218 * When compiling the boost libraries:
220 aarch64-linux-gnu-g++: error: unrecognized command line option ‘-m64’
222 * Clean the boost library directory before trying to build with a different architecture:
226 * It should show the following for arm64:
234 #### Missing libz.so.1
235 * When compiling armNN:
237 /usr/lib/gcc-cross/aarch64-linux-gnu/5/../../../../aarch64-linux-gnu/bin/ld: warning: libz.so.1, needed by /home/<username>/armNN/usr/lib64/libprotobuf.so.15.0.0, not found (try using -rpath or -rpath-link)
240 * Missing arm64 libraries for libz.so.1, these can be added by adding a second architecture to dpkg and explicitly installing them:
242 sudo dpkg --add-architecture arm64
243 sudo apt-get install zlib1g:arm64
247 * If apt-get update returns 404 errors for arm64 repos refer to section 5 below.
248 * Alternatively the missing arm64 version of libz.so.1 can be downloaded and installed from a .deb package here:
249 https://launchpad.net/ubuntu/wily/arm64/zlib1g/1:1.2.8.dfsg-2ubuntu4
251 sudo dpkg -i zlib1g_1.2.8.dfsg-2ubuntu4_arm64.deb
254 #### Unable to install arm64 packages after adding arm64 architecture
255 * Using sudo apt-get update should add all of the required repos for arm64 but if it does not or you are getting 404 errors the following instructions can be used to add the repos manually:
256 * From stackoverflow:
257 https://askubuntu.com/questions/430705/how-to-use-apt-get-to-download-multi-arch-library/430718
258 * Open /etc/apt/sources.list with your preferred text editor.
260 * Mark all the current (default) repos as \[arch=<current_os_arch>], e.g.
262 deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ xenial main restricted
264 * Then add the following:
266 deb [arch=arm64] http://ports.ubuntu.com/ xenial main restricted
267 deb [arch=arm64] http://ports.ubuntu.com/ xenial-updates main restricted
268 deb [arch=arm64] http://ports.ubuntu.com/ xenial universe
269 deb [arch=arm64] http://ports.ubuntu.com/ xenial-updates universe
270 deb [arch=arm64] http://ports.ubuntu.com/ xenial multiverse
271 deb [arch=arm64] http://ports.ubuntu.com/ xenial-updates multiverse
272 deb [arch=arm64] http://ports.ubuntu.com/ xenial-backports main restricted universe multiverse
274 * Update and install again:
276 sudo apt-get install zlib1g:arm64
281 #### Undefined references to google::protobuf:: functions
282 * When compiling armNN there are multiple errors of the following type:
284 libarmnnCaffeParser.so: undefined reference to `google::protobuf:*
286 * Missing or out of date protobuf compilation libraries.
287 Use the command 'protoc --version' to check which version of protobuf is available (version 3.5.1 is required).
288 Follow the instructions above to install protobuf 3.5.1
289 Note this will require you to recompile Caffe for x86_64
292 #### Errors on strict-aliasing rules when compiling the Compute Library
293 * When compiling the Compute Library there are multiple errors on strict-aliasing rules:
295 cc1plus: error: unrecognized command line option ‘-Wno-implicit-fallthrough’ [-Werror]
297 * Add Werror=0 to the scons command:
299 scons arch=arm64-v8a neon=1 opencl=1 embed_kernels=1 extra_cxx_flags="-fPIC" -j8 internal_only=0 Werror=0