使用multiprocessing.Manager.list而不是真实列表会使计算耗时 [英] Using multiprocessing.Manager.list instead of a real list makes the calculation take ages

问题描述

我想从本示例开始尝试使用multiprocessing的不同方式:

I wanted to try different ways of using multiprocessing starting with this example:

$ cat multi_bad.py 
import multiprocessing as mp
from time import sleep
from random import randint

def f(l, t):
#   sleep(30)
    return sum(x < t for x in l)

if __name__ == '__main__':
    l = [randint(1, 1000) for _ in range(25000)]
    t = [randint(1, 1000) for _ in range(4)]
#   sleep(15)
    pool = mp.Pool(processes=4)
    result = pool.starmap_async(f, [(l, x) for x in t])
    print(result.get())

在这里，l是当产生4个进程时被复制4次的列表.为了避免这种情况，文档页面提供了使用队列，共享数组或使用multiprocessing.Manager创建的代理对象的信息.对于最后一个，我更改了l的定义:

Here, l is a list that gets copied 4 times when 4 processes are spawned. To avoid that, the documentation page offers using queues, shared arrays or proxy objects created using multiprocessing.Manager. For the last one, I changed the definition of l:

$ diff multi_bad.py multi_good.py 
10c10,11
<     l = [randint(1, 1000) for _ in range(25000)]
---
>     man = mp.Manager()
>     l = man.list([randint(1, 1000) for _ in range(25000)])

结果看起来仍然正确，但是执行时间却大大增加，以至于我做错了事情:

The results still look correct, but the execution time has increased so dramatically that I think I'm doing something wrong:

$ time python multi_bad.py 
[17867, 11103, 2021, 17918]

real    0m0.247s
user    0m0.183s
sys 0m0.010s

$ time python multi_good.py 
[3609, 20277, 7799, 24262]

real    0m15.108s
user    0m28.092s
sys 0m6.320s

文档确实说这种方法比共享数组要慢，但是这感觉很不对.我也不确定如何才能对此进行分析，以获取有关正在发生的事情的更多信息.我想念什么吗?

The docs do say that this way is slower than shared arrays, but this just feels wrong. I'm also not sure how I can profile this to get more information on what's going on. Am I missing something?

P.S.使用共享数组，我得到的时间低于0.25s.

P.S. With shared arrays I get times below 0.25s.

P.P.S.这是在Linux和Python 3.3上.

P.P.S. This is on Linux and Python 3.3.

推荐答案

Linux使用复制- os.fork子进程时，写时.演示:

Linux uses copy-on-write when subprocesses are os.forked. To demonstrate:

import multiprocessing as mp
import numpy as np
import logging
import os

logger = mp.log_to_stderr(logging.WARNING)

def free_memory():
    total = 0
    with open('/proc/meminfo', 'r') as f:
        for line in f:
            line = line.strip()
            if any(line.startswith(field) for field in ('MemFree', 'Buffers', 'Cached')):
                field, amount, unit = line.split()
                amount = int(amount)
                if unit != 'kB':
                    raise ValueError(
                        'Unknown unit {u!r} in /proc/meminfo'.format(u = unit))
                total += amount
    return total

def worker(i):
    x = data[i,:].sum()    # Exercise access to data
    logger.warn('Free memory: {m}'.format(m = free_memory()))

def main():
    procs = [mp.Process(target = worker, args = (i, )) for i in range(4)]
    for proc in procs:
        proc.start()
    for proc in procs:
        proc.join()

logger.warn('Initial free: {m}'.format(m = free_memory()))
N = 15000
data = np.ones((N,N))
logger.warn('After allocating data: {m}'.format(m = free_memory()))

if __name__ == '__main__':
    main()

产生了

[WARNING/MainProcess] Initial free: 2522340
[WARNING/MainProcess] After allocating data: 763248
[WARNING/Process-1] Free memory: 760852
[WARNING/Process-2] Free memory: 757652
[WARNING/Process-3] Free memory: 757264
[WARNING/Process-4] Free memory: 756760

这表明最初大约有2.5GB的可用内存. 分配float64的15000x15000数组后，有763248 KB可用空间.因为15000 ** 2 * 8个字节= 1.8GB，而内存的下降2.5GB-0.763248GB也大约是1.8GB，所以这大致是合理的.

This shows that initially there was roughly 2.5GB of free memory. After allocating a 15000x15000 array of float64s, there was 763248 KB free. This roughly makes sense since 15000**2*8 bytes = 1.8GB and the drop in memory, 2.5GB - 0.763248GB is also roughly 1.8GB.

现在，在生成每个进程之后，再次报告空闲内存为〜750MB.可用内存没有明显减少，因此我得出结论，系统必须使用写时复制.

Now after each process is spawned, the free memory is again reported to be ~750MB. There is no significant decrease in free memory, so I conclude the system must be using copy-on-write.

结论:如果不需要修改数据，则在__main__模块的全局级别上定义数据是一种方便的(至少在Linux上是)内存友好的方式，可以在子进程之间共享数据.

Conclusion: If you do not need to modify the data, defining it at the global level of the __main__ module is a convenient and (at least on Linux) memory-friendly way to share it among subprocesses.

这篇关于使用multiprocessing.Manager.list而不是真实列表会使计算耗时的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！