Ubuntu18.04下pybind11的安装

东方耀

Ubuntu18.04下pybind11的安装
https://github.com/pybind/pybind11
C ++ 11和Python之间的无缝可操作性


首先安装cmake：http://www.ai111.vip/thread-977-1-1.html


查看版本：cmake --version
jiang@jiang-Ubuntu:~/GNU_Radio_works/gr-adapt-master$ cmake --version
cmake version 3.19.4


a.先下载pybind11的源代码：https://github.com/pybind/pybind11


b.进入pybind源代码，并建立新文件build，进行build，cmake,make..,具体如下：


cd 到pybind11目录
mkdir build
cd build
cmake ..
sudo make
sudo make install


如果不通过源码编译的方式安装：
sudo apt-get install python-pybind11

1. jiang@jiang-Ubuntu:~/downloads/pybind11-master/build$ sudo make install
   
2. [  4%] Built target pybind11_cross_module_tests
   
3. [ 95%] Built target pybind11_tests
   
4. [100%] Built target cross_module_gil_utils
   
5. Install the project...
   
6. -- Install configuration: "MinSizeRel"
   
7. -- Installing: /usr/local/include/pybind11
   
8. -- Installing: /usr/local/include/pybind11/cast.h
   
9. -- Installing: /usr/local/include/pybind11/numpy.h
   
10. -- Installing: /usr/local/include/pybind11/gil.h
    
11. -- Installing: /usr/local/include/pybind11/buffer_info.h
    
12. -- Installing: /usr/local/include/pybind11/common.h
    
13. -- Installing: /usr/local/include/pybind11/eval.h
    
14. -- Installing: /usr/local/include/pybind11/iostream.h
    
15. -- Installing: /usr/local/include/pybind11/complex.h
    
16. -- Installing: /usr/local/include/pybind11/functional.h
    
17. -- Installing: /usr/local/include/pybind11/pytypes.h
    
18. -- Installing: /usr/local/include/pybind11/chrono.h
    
19. -- Installing: /usr/local/include/pybind11/attr.h
    
20. -- Installing: /usr/local/include/pybind11/stl.h
    
21. -- Installing: /usr/local/include/pybind11/detail
    
22. -- Installing: /usr/local/include/pybind11/detail/class.h
    
23. -- Installing: /usr/local/include/pybind11/detail/type_caster_base.h
    
24. -- Installing: /usr/local/include/pybind11/detail/common.h
    
25. -- Installing: /usr/local/include/pybind11/detail/init.h
    
26. -- Installing: /usr/local/include/pybind11/detail/internals.h
    
27. -- Installing: /usr/local/include/pybind11/detail/descr.h
    
28. -- Installing: /usr/local/include/pybind11/detail/typeid.h
    
29. -- Installing: /usr/local/include/pybind11/eigen.h
    
30. -- Installing: /usr/local/include/pybind11/options.h
    
31. -- Installing: /usr/local/include/pybind11/pybind11.h
    
32. -- Installing: /usr/local/include/pybind11/embed.h
    
33. -- Installing: /usr/local/include/pybind11/stl_bind.h
    
34. -- Installing: /usr/local/include/pybind11/operators.h
    
35. -- Installing: /usr/local/share/cmake/pybind11/pybind11Config.cmake
    
36. -- Installing: /usr/local/share/cmake/pybind11/pybind11ConfigVersion.cmake
    
37. -- Installing: /usr/local/share/cmake/pybind11/FindPythonLibsNew.cmake
    
38. -- Installing: /usr/local/share/cmake/pybind11/pybind11Common.cmake
    
39. -- Installing: /usr/local/share/cmake/pybind11/pybind11Tools.cmake
    
40. -- Installing: /usr/local/share/cmake/pybind11/pybind11NewTools.cmake
    
41. -- Installing: /usr/local/share/cmake/pybind11/pybind11Targets.cmake
    
42. jiang@jiang-Ubuntu:~/downloads/pybind11-master/build$
    

复制代码

东方耀

Tensorflow已经从SWIG切到pybind11了

个人比较喜欢/偏向于pybind11，有以下几个原因1. 灵活性：pybind11本质上还是在C API外面包了一层C++（或者说利用C++的元编程能力批量产生binding）所以可定制性很好，尤其是面向我这种有非主流需求的用户。2. 抽象能力：cython这种Python的补丁抽象能力没有完整的C++好，对于一个倾向于只让Python成为傻瓜式接口的人，我更希望能够同时在C++层面有丰富的抽象来方便developer。3. 生态：pybind11现在有很多基于它的生态，例如xtensor，Python/C++里最好的多维数组没有之一。然后还有pythran这个transpiler，可以把Python的prototype转到C++，一定程度上减轻编写C++的蛋疼点。PS. 我觉得有了pythran以后cython已经没有任何优势了