Question

该测试是单线程的，它调用函数“计算”的1000倍。此函数在约0.6秒内将0到10,000,000之间的所有整数相加。它返回执行此计算所需的实际墙时间。该测试会跟踪最短时间，时间总和和时间总和的平方。调用1000次后，将计算并打印均值，标准差和忙于其他操作的值。

忙于其他事情是：sum_time-最短时间*循环数

忙于在MacOS，Linux / Ubuntu和iPad上进行操作的价值从0.493％到4.050％。这是操作系统的预期费用。

Windows上的结果不断提高（最佳和最差从10开始）：

>python test.py 
Busy with something else 9.723% or 58.734 from 604.051 seconds 
min = 0.545 mean = 0.604 stdev = 0.133 

>python test.py 
Busy with something else 16.230% or 105.709 from 651.320 seconds 
min = 0.546 mean = 0.651 stdev = 0.194

我的Windows计算机信息：

2.4 GHz Intel Core i5-6300U, Memory 4 GB 
Microsoft Windows [Version 10.0.17763.379] 
Python version 3.7.2, Executing in 64bit

测试：

import time
import sys


def calculate():
    start = time.perf_counter()
    result = 0
    for i in range(10000000):
        result += i
    stop = time.perf_counter()
    return stop - start


number_of_loops = 1000

min_time = sys.maxsize
sum_time = 0.0
sum_time_square = 0.0
for _ in range(number_of_loops):
    t = calculate()
    min_time = min(min_time, t)
    sum_time += t
    sum_time_square += t ** 2

mean = sum_time / number_of_loops
stdev = ((sum_time_square + number_of_loops * mean**2 - 2.0 * mean * sum_time) /
         (number_of_loops - 1.0)) ** 0.5
busy_with_else = sum_time - min_time * number_of_loops
print('Busy with something else %.3f%% or %.3f from %.3f seconds' %
      (100.0 * busy_with_else / sum_time, busy_with_else, sum_time))
print('min = %.3f mean = %.3f stdev = %.3f' % (min_time, mean, stdev))

在测试之前处于空闲状态的系统上执行的所有测试。在Windows上进行类似的C ++测试的结果是0.626％至2.059％。因此，这不是Windows的责任。

这仅仅是我的软件/硬件，还是可以复制和/或解释？

Python忙于Windows计算机上的其他事情

0 个答案: