首页 > 解决方案 > 如何计算几个 cProfile 结果的平均结果?

问题描述

而不是像这样只运行一次配置文件:

import cProfile

def do_heavy_lifting():
    for i in range(100):
        print('hello')

profiller = cProfile.Profile()
profiller.enable()

do_heavy_lifting()

profiller.disable()
profiller.print_stats(sort='time')

配置文件结果如下所示:

      502 function calls in 0.000 seconds

Ordered by: internal time

ncalls  tottime  percall  cumtime  percall filename:lineno(function)
   100    0.000    0.000    0.000    0.000 {built-in method builtins.print}
   200    0.000    0.000    0.000    0.000 cp1252.py:18(encode)
   200    0.000    0.000    0.000    0.000 {built-in method _codecs.charmap_encode}
     1    0.000    0.000    0.000    0.000 test.py:2(do_heavy_lifting)
     1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}

我想运行几次并打印平均结果,以获得更好的精度。

这是我想到的初始脚本配方:

import cProfile

def do_heavy_lifting():
    for i in range(100):
        print('hello')

def best_of_profillings(target_profile_function, count):
    profile_results = []

    for index in range(count):
        profiller = cProfile.Profile()
        profiller.enable()

        target_profile_function()

        profiller.disable()
        profile_results.append(profiller)

    profile_results /= count
    return profile_results

heavy_lifting_result = best_of_profillings(do_heavy_lifting, 10)
heavy_lifting_result.print_stats(sort='time')

运行后,它应该像第一个版本一样显示结果,但不同之处在于它们是几次运行的平均值,而不是运行一次。

草稿脚本仍然缺少profile_results /= count在所有迭代之后,我将获得所有计算结果并创建平均结果并将其始终显示在屏幕上的部分:

      502 function calls in 0.000 seconds

Ordered by: internal time

ncalls  tottime  percall  cumtime  percall filename:lineno(function)
   100    0.000    0.000    0.000    0.000 {built-in method builtins.print}
   200    0.000    0.000    0.000    0.000 cp1252.py:18(encode)
   200    0.000    0.000    0.000    0.000 {built-in method _codecs.charmap_encode}
     1    0.000    0.000    0.000    0.000 test.py:2(do_heavy_lifting)
     1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}

标签: pythonpython-3.xprofilercprofile

解决方案


我设法使用函数创建了以下代码average()。我打开了实现pstats并观察到有一个函数调用Stats.add(),它似乎只是将结果连接到当前对象中: https ://docs.python.org/3.7/library/profile.html#pstats.Stats.add

import io
import pstats
import cProfile

def do_heavy_lifting():
    for i in range(100):
        print('hello')

def average(stats, count):
    stats.total_calls /= count
    stats.prim_calls /= count
    stats.total_tt /= count

    for func, source in stats.stats.items():
        cc, nc, tt, ct, callers = source
        stats.stats[func] = ( cc/count, nc/count, tt/count, ct/count, callers )

    return stats

def best_of_profillings(target_profile_function, count):
    output_stream = io.StringIO()
    profiller_status = pstats.Stats( stream=output_stream )

    for index in range(count):
        profiller = cProfile.Profile()
        profiller.enable()

        target_profile_function()

        profiller.disable()
        profiller_status.add( profiller )

        print( 'Profiled', '%.3f' % profiller_status.total_tt, 'seconds at', index,
                'for', target_profile_function.__name__, flush=True )

    average( profiller_status, count )
    profiller_status.sort_stats( "time" )
    profiller_status.print_stats()

    return "\nProfile results for %s\n%s" % ( 
           target_profile_function.__name__, output_stream.getvalue() )

heavy_lifting_result = best_of_profillings( do_heavy_lifting, 10 )
print( heavy_lifting_result )

结果:

Profile results for do_heavy_lifting
         102.0 function calls in 0.001 seconds

   Ordered by: internal time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    100.0    0.001    0.000    0.001    0.000 {built-in method builtins.print}
      1.0    0.000    0.000    0.001    0.001 D:\test.py:5(do_heavy_lifting)
      1.0    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler' objects}

推荐阅读