aarch64 kylin CMAKE_STATICBUILD=1 tools/staticbuild/build.sh cpu > build.log 2>&1

[Duanexiao]

os-release info:

NAME="Kylin Linux Advanced Server"
VERSION="V10 (Tercel)"
ID="kylin"
VERSION_ID="V10"
PRETTY_NAME="Kylin Linux Advanced Server V10 (Tercel)"
ANSI_COLOR="0;31"

incubator-mxnet branch: 1.9.x

execute command: CMAKE_STATICBUILD=1 tools/staticbuild/build.sh cpu > build.log 2>&1

error info:
/usr/bin/ld: ../staticdeps/lib64/libpng.a(pngrutil.c.o): in function png_inflate_claim': pngrutil.c:(.text+0xff8): undefined reference to inflateValidate'

[Duanexiao]

build.log

[Duanexiao]

for build log detail info,you can download the build.log file above

[Duanexiao]

google the error info,returns me the answer to install zlib.Exactly, the system already exists zlib with 1.2.11,and you can find the zlib with 1.2.6 in the staticdeps.I have no idea to continue solve the problem,so sos for help.Many tks!

[leezu]

Please follow the build from source instructions https://mxnet.apache.org/get_started/build_from_source The tools/staticbuild/build.sh is a special script used to statically link dependencies for publication on https://pypi.org/project/mxnet. Do you need that?

[Duanexiao]

Thank you for you reply.I want to build mxnet for python on aarch64.When I pip install mxnet, the libmxnet.so file is not avaliable.So I need to build the libmxnet.so

[leezu]

Great. For that you should not use tools/staticbuild/build.sh but rather follow the instructions in https://mxnet.apache.org/get_started/build_from_source (which also links to https://github.com/apache/incubator-mxnet/blob/9ebe801e14dd657ee9f410f934b9e2d94416ea29/config/linux.cmake#L18-L36)

[Duanexiao]

Thank you.I am newbee for this operation.So I am not sure about your answer.DId you mean that:
cd mxnet/config;mkdir build; cd build;cmake ..;cmake --build .
I did like this,but there is no cmakelist.txt.I should copy the mxnet/CMakeList.txt here?Looking for detail build operation on the get_started.

[leezu]

No, you shouldn't do cd mxnet/config. Just follow the commands listed above in sequence from the mxnet root directory. For example:

cd /data/mxnet
cp config/linux.cmake config.cmake 
mkdir build; cd build 
cmake .. 
cmake --build . 

[Duanexiao]

cmake --build . error:
/data/mxnet/3rdparty/onednn/src/cpu/gemm_convolution.cpp: In lambda function:
/data/mxnet/3rdparty/onednn/src/cpu/gemm_convolution.cpp:321:29: sorry, unimplemented: unexpected AST of kind omp_simd
});
^
/data/mxnet/3rdparty/onednn/src/cpu/gemm_convolution.cpp:321: confused by earlier errors, bailing out
gmake[2]: *** [3rdparty/onednn/src/cpu/CMakeFiles/dnnl_cpu.dir/build.make:336: 3rdparty/onednn/src/cpu/CMakeFiles/dnnl_cpu.dir/gemm_convolution.cpp.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:1338: 3rdparty/onednn/src/cpu/CMakeFiles/dnnl_cpu.dir/all] Error 2
gmake: *** [Makefile:141: all] Error 2

gcc version: 9.4.0
cmake version:3.15.0

[Duanexiao]

I've solved this problem.So here I want to share my answer.Below as my whole reproduciable operation.All my operations in docker
arch: arm64
os: centos7

step1:

centos7 arm64 dockerfile:
FROM centos:centos7
ENV TZ=Asia/Shanghai
RUN yum install kde-l10n-Chinese -y && \
yum install glibc-common -y && \
yum install cronie -y && \
localedef -c -f UTF-8 -i zh_CN zh_CN.utf8 && \
export LANG=zh_CN.UTF-8 && \
echo "export LANG=zh_CN.UTF-8" >> /etc/locale.conf && \
ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone && \
yum clean all
ENV LANG zh_CN.UTF-8
ENV LC_ALL zh_CN.UTF-8
ENTRYPOINT /bin/bash

use this dockerfile to build image, command is below:
docker build -t shakespeare-base-centos7-arm:0.1 .

bash into one container based on docker image above
docker run -it --entrypoint /bin/bash 6cb7b6f6cd18

step 2:

install high version gcc
yum install centos-release-scl -y
yum install devtoolset-8 -y
#activate devtoolset-8 env
source /opt/rh/devtoolset-8/enable

step 3:

install high version cmake
rpm -ivh https://mirrors.tuna.tsinghua.edu.cn/epel/epel-release-latest-7.noarch.rpm
yum install cmake3 -y
ln -s /usr/bin/cmake3 /usr/bin/make

step 4:

source code compile high version opencv
mkdir ~/opencv_build && cd ~/opencv_build
git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git
cd ~/opencv_build/opencv && make build && cd build
cmake3 -D CMAKE_BUILD_TYPE=RELEASE
-D CMAKE_INSTALL_PREFIX=/usr/local
-D INSTALL_C_EXAMPLES=ON
-D INSTALL_PYTHON_EXAMPLES=ON
-D OPENCV_GENERATE_PKGCONFIG=ON
-D OPENCV_EXTRA_MODULES_PATH=~/opencv_build/opencv_contrib/modules
-D BUILD_EXAMPLES=ON ..

#notice that successfully installed display like these：
#-- Configuring done
#-- Generating done
#-- Build files have been written to: /root/opencv_build/opencv/build

#next step
make -j $(cat /proc/cpuinfo | grep processor | wc -l)

#next step
make install

#soft link
ln -s /usr/local/lib64/pkgconfig/opencv4.pc /usr/share/pkgconfig/opencv4.pc
ldconfig

#check opencv version
pkg-config --modversion opencv4

step 5:

source code compile openblas

git clone https://github.com/xianyi/OpenBLAS.git
cd OpenBLAS
make -j $(cat /proc/cpuinfo | grep processor | wc -l)
#make successfully display like this：
make[1]: Leaving directory '/root/OpenBLAS/exports'

OpenBLAS build complete. (BLAS CBLAS LAPACK LAPACKE)

OS ... Linux
Architecture ... arm64
BINARY ... 64bit
C compiler ... GCC (cmd & version : cc (GCC) 8.3.1 20190311 (Red Hat 8.3.1-3))
Fortran compiler ... GFORTRAN (cmd & version : GNU Fortran (GCC) 8.3.1 20190311 (Red Hat 8.3.1-3))
Library Name ... libopenblas_tsv110p-r0.3.15.dev.a (Multi-threading; Max num-threads is 128)

To install the library, you can run "make PREFIX=/path/to/your/installation install".

#make install set prefix so that you donot need to change findOpenBLAS file content
make PREFIX=/opt/OpenBLAS install

step 6:

source code compile mxnet

git clone --recursive https://github.com/apache/incubator-mxnet mxnet
cd mxnet
mkdir build
cp config/linux.cmake config.cmake
cd build
cmake ..
#cmake ok display like these:
-- Using unsigned short
-- Check if the system is big endian - little endian
-- /root/mxnet/3rdparty/dmlc-core/cmake/build_config.h.in -> include/dmlc/build_config.h
-- Performing Test SUPPORT_MSSE2
-- Performing Test SUPPORT_MSSE2 - Failed
-- Using 64-bit integer for tensor size
-- Found Python3: /usr/bin/python3.6 (found version "3.6.8") found components: Interpreter
-- Configuring done
-- Generating done
-- Build files have been written to: /root/mxnet/build

#next step
cmake --build .

#cmake build ok,you'll find libmxnet.so file

step 7:

python install mxnet

#after compile, we should change directory to root of mxnet !!!
#download Archiconda3-0.2.3-Linux-aarch64.sh file and install that, you should pip install numpy opencv and many other related modules

mkdir /opt/wuzhi_ai
tar -xf anaconda3_aarch64.tar.gz -C /opt/wuzhi_ai
echo "export PATH=/opt/wuzhi_ai/anaconda3/bin:$PATH" >> /etc/profile
source /etc/profile
#last steps are install python environment, you can change these operation as you like
python -m pip install --user -e ./python

That's all