Ubuntu20.04交叉编译OpenCV 4.5.2, 运行在Tina Linux

[aldfaaa]

本着尽可能快完成编译和能用则用的原则, 谈谈编译目前最新的OpenCV 4.5.2编译过程.

工具链

• Tina Linux工具链

▲ 用平头哥的工具链(riscv64-linux-x86_64-20210512)编译出来:

• 在Tina Linux能跑, 但会段错误,
• 在RVBoard debian运行报运行库版本缺失(不匹配), 替换的话, 分分钟要重新烧录系统

背景

Linux tao 5.8.0-59-generic #66~20.04.1-Ubuntu SMP Thu Jun 17 11:14:10 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

cmake: 3.16.3

工具链选用: (Tina SDK 自带)
tina-d1-open/out/d1-nezha/staging_dir/toolchain/bin/riscv64-unknown-linux-gnu-gcc等...
版本: gcc version 8.1.0 (C-SKY RISCV Tools V1.8.4 B20200702)

要编译的OpenCV版本: 4.5.2

开始编译

这里先提几个tips:

1. ippicv库是给Intel CPU加速用的, 哪吒表示无感. 可以窜改:opencv-4.5.2/3rdparty/ippicv/ippicv.cmake中的URI导致其下载失败, 从而跳过cmake过程的下载.(亦可断网)
2. 所有与python的相关库都可跳过(本人主要涉猎C/C++)

编译的指令在此. CV目录下的cmake我可能窜改了一下, 在下文补充.

opencv-4.5.2 mkdir build-tina && cd build-tina

# 能编尽编原则, 可自行删减. 
cmake \ 
-DCMAKE_C_COMPILER=riscv64-unknown-linux-gnu-gcc \
-DCMAKE_CXX_COMPILER=riscv64-unknown-linux-gnu-g++ \
-DOPENCV_FORCE_3RDPARTY_BUILD=ON \
-DBUILD_ZLIB=ON \
-DBUILD_SHARED_LIBS=ON \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_PREFIX=./output ..

# 嗯, 是Xeon.
make -j12

CMakeLists.txt的diff:

#diff tmp/opencv-4.5.2/CMakeLists.txt opencv-4.5.2/CMakeLists.txt
223,227c223,224
< OCV_OPTION(BUILD_IPP_IW             "Build IPP IW from source"           (NOT MINGW OR OPENCV_FORCE_3RDPARTY_BUILD) IF (X86_64 OR X86) AND NOT WINRT )
< OCV_OPTION(BUILD_ITT                "Build Intel ITT from source"
<     (NOT MINGW OR OPENCV_FORCE_3RDPARTY_BUILD)
<     IF (X86_64 OR X86 OR ARM OR AARCH64 OR PPC64 OR PPC64LE) AND NOT WINRT AND NOT APPLE_FRAMEWORK
< )
---
> OCV_OPTION(BUILD_IPP_IW             "Build IPP IW from source"           ON)
> OCV_OPTION(BUILD_ITT                "Build Intel ITT from source" OFF)
416c413
< OCV_OPTION(WITH_ITT "Include Intel ITT support" ON
---
> OCV_OPTION(WITH_ITT "Include Intel ITT support" OFF
614c611
<   include(cmake/OpenCVDetectPython.cmake)
---
>     #  include(cmake/OpenCVDetectPython.cmake)
731a729,730
> set(PYTHON_DEFAULT_AVAILABLE 0)
> set(ENABLE_PYLINT 0)
1454a1454,1456
> set(WITH_IPP 0)
> set(HAVE_IPP 0)
> set(HAVE_IPP_IW 0)
1608a1611,1612
> set(BUILD_opencv_python2 0)
> set(BUILD_opencv_python3 0)

完整的CMakeLists.txt.bmp如下图:

base64 -d CMakeLists.txt.bmp > CMakeLists.txt


[aldfaaa]

不皮了. 补充一点:

1. Tina Linux中运行test会遇到libatomic.so库缺失.
   应从SDK目录中复制到板子上:

# -d 是复制快捷方式. *号会把实体也一起复制.
cp -d tina-d1-open/prebuilt/gcc/linux-x86/riscv/toolchain-thead-glibc/riscv64-glibc-gcc-thead_20200702/riscv64-unknown-linux-gnu/lib64/lp64d/libatomic.so* (target_path)

2. diff如何应用?

• 百度一下: linux patch用法