python性能分析

Python的性能分析工具cProfile使用起来比较简单，如下：

python -m cProfile -o profile.txt xxx.py arg1 arg2 ...

其中，xxx.py是要分析的python程序入口函数，后面跟的就是对应的参数了。

生成的profile.txt是存储的profile信息，可以用下面的python脚本查询。
check_profile.py的内容如下：

#coding:utf-8
import pstats
import sys

p=pstats.Stats(sys.argv[1])
p.sort_stats('cumulative').print_stats()

用法很简单：

$ python3 check_profile.py profile.txt

然后，就可以看到下面的类似的信息了：

Fri Sep 20 13:37:54 2019    profile.txt

         14927559 function calls (14919846 primitive calls) in 26.559 seconds

   Ordered by: cumulative time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    410/1    0.004    0.000   26.559   26.559 {built-in method builtins.exec}
        1    0.000    0.000   26.559   26.559 generate_glasses.py:2(<module>)
        1    0.000    0.000   24.723   24.723 generate_glasses.py:35(main)
        1    0.000    0.000   24.465   24.465 /usr/local/lib/python3.7/site-packages/psopy/minimize.py:216(minimize)
        1    0.000    0.000   24.463   24.463 /usr/local/lib/python3.7/site-packages/psopy/minimize.py:44(_minimize_pso)
        1    0.000    0.000   24.451   24.451 /usr/local/lib/python3.7/site-packages/psopy/minimize.py:433(<lambda>)
        1    0.002    0.002   24.451   24.451 /usr/local/lib/python3.7/site-packages/numpy/lib/shape_base.py:23(apply_along_axis)
       78    0.008    0.000   24.448    0.313 generate_glasses.py:15(compare_generated_face_with_target)
       78    0.001    0.000   17.289    0.222 /usr/local/lib/python3.7/site-packages/face_recognition/api.py:190(face_encodings)
       78    0.002    0.000   15.405    0.198 /usr/local/lib/python3.7/site-packages/face_recognition/api.py:151(_raw_face_landmarks)
       78   15.280    0.196   15.280    0.196 /usr/local/lib/python3.7/site-packages/face_recognition/api.py:89(_raw_face_locations)