码迷,mamicode.com
首页 > 编程语言 > 详细

Python中的多进程

时间:2018-11-13 02:59:11      阅读:198      评论:0      收藏:0      [点我收藏+]

标签:tco   RKE   工作   __name__   方式   import   导致   OLE   elf   

由于cPython的gill,多线程未必是CPU密集型程序的好的选择。

多线程可以完全独立的进程环境中运行程序,可以充分利用多处理器。

但是进程本身的隔离带来的数据不共享也是一种问题,线程比进程轻量级。

 

1、Multiprocessing


import multiprocessing
import datetime

def calc(i):
    sum = 0
    for _ in range(10000000):
        sum += 1
    print(i,sum)

if __name__ == ‘__main__‘:
    start = datetime.datetime.now()

    ps =[]
    for i in range(5):
        p = multiprocessing.Process(target=calc,args=(i,),name=‘calc-{}‘.format(i))

        ps.append(p)
        p.start()

    for p in ps:
        p.join()

    delta = (datetime.datetime.now()-start).total_seconds()
    print(delta)
    print(‘end==‘)
   

 

0 10000000

1 10000000

2 10000000

3 10000000

4 10000000

2.738819

end==

 

多进程,是真的并行。

名称

说明

Pid

进程id

Exitcode

进程的退出状态码

Terminate()

终止指定的进程

 

 

进程间同步:同步提供了和线程同步一样的类,使用的方法一样,使用效果也类似。

不过,进程间代价要高于线程间,而且底层实现是不同的,只不过Python屏蔽了这些不同的地方,让用户简单使用多进程。

 

Multiprocessing还提供了共享内存,服务器进程来共享数据,还提供了queue队列,pipe管道用于进程间通信。

 

 

通信方式不同:

 多进程就是启动多个解释器进程,进程间通信必须序列化和反序列化。

 数据的线程安全性问题。由于每个进程之间没有实现度贤臣,gil可以说没有什么用了。

 

2、进程池

Multiprocessing。Pool是进程池类;

名称

说明

Apply(self,func,args=(),kwds={})

 

阻塞执行,导致主进程执行其他子进程一个个执行

Apply_async(self,func,args=(),kwargs={},callback=None,error_callback=None)

与apply方法用法一直,非阻塞执行,得到结果后会执行回调

Close()

 

关闭池,池不能再接受新的任务

Terminate()

 

结束工作进程,不在处理为处理的任务

Join()

 

主进程阻塞等待子进程的退出,join方法要在close或terminate之后使用

 

3、多进程、多线程的选择

CPU密集型:cPython中使用到了gill,多线程的时候锁相互竞争,且多核优势不能发挥,Python多进程效率更高。

IO密集型:适合多线程,可以减少多进程之间的IO序列化开销,且在IO等待的时候,切换到其他线程继续执行,效率不错。

 

4、应用场景

请求/应答模型:web应用中常见的处理模型。

Master启动多个worker工作进程,一般和CPU数目相同。发挥多核优势。

Worker工作进程中,往往需要操作网络IO和磁盘IO,启动多线程,提高高并发处理能力,worker处理用户的请求,往往需要等待数据,处理完请求还需要通过网络IO响应返回,这个就是nginx工作模式。

 

 

 

、concurrent包

 

1、concurrent.futures

 

异步并行任务编程模块,提供一个高级的异步可执行的便利接口。

 

提供了两个池执行器

Threadpoolexecutor异步调用的线程池的executor

Processpoolexecutor异步调用的进程池的executor。

2、threadpoolexecutor对象

 

首先需要定义一个池的执行器对象,executor类子类对象。

方法

含义

Threadpoolexecutor(man_worker=1)

池中至多创建max_wokers个线程的池来同时异步执行,返回exector实例

Submit(fn,*args,**kwargs)

提交执行的函数及其参数,返回future实例

Shutdown(wait=True)

清理池

 

Future类

方法

含义

Done()

如果调用被成功的取消或者执行完成,返回true

Cancelled()

如果调用被成功的取消,返回true

Running()

如果正在运行且不能被取消,返回true

Cancel()

尝试取消调用,如果已经执行且不能取消返回False,否则返回true

Result(timeout=none)

取返回的结果,timeout为none,一直等待返回,timeout设置到期,抛出concurrent.futures.timeouterror异常

Exception(timeout=none)

取返回的结果,timeout为none,一直等待返回,timeout设置到期,抛出concurrent.futures.timeouterror异常

 

import threading
from concurrent import futures
import logging
import time


FORMAT = ‘%(asctime)s %(threadName)s %(thread)d %(message)s‘
logging.basicConfig(format=FORMAT,level=logging.INFO)

def worker(n):
    logging.info(‘start to work{]‘.format(n))
    time.sleep(4)
    logging.info(‘stop{}‘.format(n))


exector = futures.ThreadPoolExecutor(max_workers=3)
fs = []
for i in range(3):
    futures = exector.submit(worker,i)
    fs.append(futures)

for i in range(3,6):
    futures = exector.submit(worker,i)
    fs.append(futures)

while True:
    time.sleep(5)
    logging.info(threading.enumerate())

    flag = True
    for f in fs:
        logging.info(f.done())
        flag = flag and f.done()
    print(‘-‘*30)

    if flag:
        exector.shutdown()
        logging.info(threading.enumerate())
        break

 

------------------------------

2018-06-13 09:57:35,049 MainThread 8376 [<_MainThread(MainThread, started 8376)>, <Thread(Thread-1, started daemon 7480)>, <Thread(Thread-3, started daemon 7368)>, <Thread(Thread-2, started daemon 7464)>]

2018-06-13 09:57:35,049 MainThread 8376 True

2018-06-13 09:57:35,050 MainThread 8376 True

2018-06-13 09:57:35,050 MainThread 8376 True

2018-06-13 09:57:35,050 MainThread 8376 True

2018-06-13 09:57:35,050 MainThread 8376 True

2018-06-13 09:57:35,051 MainThread 8376 True

2018-06-13 09:57:35,051 MainThread 8376 [<_MainThread(MainThread, started 8376)>]

 

3、processpoolexector对象

import threading
from concurrent import futures
import logging
import time


FORMAT = ‘%(asctime)s %(threadName)s %(thread)d %(message)s‘
logging.basicConfig(format=FORMAT,level=logging.INFO)

def worker(n):
    logging.info(‘start to work{]‘.format(n))
    time.sleep(4)
    logging.info(‘stop{}‘.format(n))


if __name__ == ‘__main__‘:
    exector = futures.ThreadPoolExecutor(max_workers=3)
    fs = []
    for i in range(3):
        futures = exector.submit(worker,i)
        fs.append(futures)

    for i in range(3,6):
        futures = exector.submit(worker,i)
        fs.append(futures)

    while True:
        time.sleep(5)
        logging.info(threading.enumerate())

        flag = True
        for f in fs:
            logging.info(f.done())
            flag = flag and f.done()
        print(‘-‘*30)

        if flag:
            exector.shutdown()
            logging.info(threading.enumerate())
            break

 

------------------------------

2018-06-13 10:01:18,076 MainThread 6436 [<Thread(Thread-3, started daemon 4188)>, <Thread(Thread-1, started daemon 7284)>, <_MainThread(MainThread, started 6436)>, <Thread(Thread-2, started daemon 6164)>]

2018-06-13 10:01:18,076 MainThread 6436 True

2018-06-13 10:01:18,076 MainThread 6436 True

2018-06-13 10:01:18,077 MainThread 6436 True

2018-06-13 10:01:18,077 MainThread 6436 True

2018-06-13 10:01:18,077 MainThread 6436 True

2018-06-13 10:01:18,077 MainThread 6436 True

2018-06-13 10:01:18,078 MainThread 6436 [<_MainThread(MainThread, started 6436)>]

 

 

进程代码的执行过程中,必须要加上if __name__ == ‘__main__’

 

 

4、支持上下文管理的调用

Concurrent.futures.processpoolexecutor继承自concurrent.futures.base.executor,而父类有__enter__、__exit__、方法,支持上下文管理,可以使用with语句。

 

__exit__方法本质上还是调用shutdown(wait=True),就会一直阻塞到所有运行的任务完成。

 

import threading
from concurrent import futures
import logging
import time


FORMAT = ‘%(asctime)s %(threadName)s %(thread)d %(message)s‘
logging.basicConfig(format=FORMAT,level=logging.INFO)

def worker(n):
    logging.info(‘start to work{}‘.format(n))
    time.sleep(5)
    logging.info(‘stop{}‘.format(n))
    return n + 100

if __name__ == ‘__main__‘:
    executor = futures.ProcessPoolExecutor(max_workers=3)

    with executor:
        fs = []
        for i in range(3):
            futures = executor.submit(worker,i)
            fs.append(futures)

        for i in range(3,6):
            futures = executor.submit(worker,i)
            fs.append(futures)

        while True:
            time.sleep(2)
            logging.info(threading.enumerate())

            flag = True
            for f in fs:
                logging.info(f.done())
                flag = flag and f.done()
                if f.done():
                    logging.info(‘result={}‘.format(f.result()))

            print(‘-‘*30)
            if flag:break

    logging.info(‘-------end--------‘)
    logging.info(threading.enumerate())

 

2018-06-13 10:18:35,744 MainThread 5468 start to work0

2018-06-13 10:18:35,751 MainThread 7936 start to work1

2018-06-13 10:18:35,763 MainThread 7020 start to work2

2018-06-13 10:18:37,528 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:37,528 MainThread 7976 False

2018-06-13 10:18:37,529 MainThread 7976 False

2018-06-13 10:18:37,529 MainThread 7976 False

2018-06-13 10:18:37,529 MainThread 7976 False

2018-06-13 10:18:37,529 MainThread 7976 False

2018-06-13 10:18:37,529 MainThread 7976 False

------------------------------

------------------------------

2018-06-13 10:18:39,530 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:39,530 MainThread 7976 False

2018-06-13 10:18:39,531 MainThread 7976 False

2018-06-13 10:18:39,531 MainThread 7976 False

2018-06-13 10:18:39,531 MainThread 7976 False

2018-06-13 10:18:39,532 MainThread 7976 False

2018-06-13 10:18:39,532 MainThread 7976 False

2018-06-13 10:18:40,744 MainThread 5468 stop0

2018-06-13 10:18:40,745 MainThread 5468 start to work3

2018-06-13 10:18:40,751 MainThread 7936 stop1

2018-06-13 10:18:40,753 MainThread 7936 start to work4

2018-06-13 10:18:40,764 MainThread 7020 stop2

2018-06-13 10:18:40,764 MainThread 7020 start to work5

2018-06-13 10:18:41,533 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:41,533 MainThread 7976 True

2018-06-13 10:18:41,534 MainThread 7976 result=100

2018-06-13 10:18:41,534 MainThread 7976 True

2018-06-13 10:18:41,535 MainThread 7976 result=101

2018-06-13 10:18:41,535 MainThread 7976 True

2018-06-13 10:18:41,536 MainThread 7976 result=102

2018-06-13 10:18:41,536 MainThread 7976 False

2018-06-13 10:18:41,536 MainThread 7976 False

2018-06-13 10:18:41,537 MainThread 7976 False

------------------------------

------------------------------

2018-06-13 10:18:43,537 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:43,537 MainThread 7976 True

2018-06-13 10:18:43,537 MainThread 7976 result=100

2018-06-13 10:18:43,538 MainThread 7976 True

2018-06-13 10:18:43,538 MainThread 7976 result=101

2018-06-13 10:18:43,538 MainThread 7976 True

2018-06-13 10:18:43,538 MainThread 7976 result=102

2018-06-13 10:18:43,538 MainThread 7976 False

2018-06-13 10:18:43,539 MainThread 7976 False

2018-06-13 10:18:43,539 MainThread 7976 False

------------------------------

2018-06-13 10:18:45,540 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:45,540 MainThread 7976 True

2018-06-13 10:18:45,540 MainThread 7976 result=100

2018-06-13 10:18:45,541 MainThread 7976 True

2018-06-13 10:18:45,541 MainThread 7976 result=101

2018-06-13 10:18:45,541 MainThread 7976 True

2018-06-13 10:18:45,541 MainThread 7976 result=102

2018-06-13 10:18:45,542 MainThread 7976 False

2018-06-13 10:18:45,542 MainThread 7976 False

2018-06-13 10:18:45,542 MainThread 7976 False

2018-06-13 10:18:45,746 MainThread 5468 stop3

2018-06-13 10:18:45,754 MainThread 7936 stop4

2018-06-13 10:18:45,765 MainThread 7020 stop5

------------------------------

2018-06-13 10:18:47,542 MainThread 7976 [<_MainThread(MainThread, started 7976)>, <Thread(QueueFeederThread, started daemon 8136)>, <Thread(Thread-1, started daemon 3932)>]

2018-06-13 10:18:47,542 MainThread 7976 True

2018-06-13 10:18:47,542 MainThread 7976 result=100

2018-06-13 10:18:47,543 MainThread 7976 True

2018-06-13 10:18:47,543 MainThread 7976 result=101

2018-06-13 10:18:47,544 MainThread 7976 True

2018-06-13 10:18:47,544 MainThread 7976 result=102

2018-06-13 10:18:47,544 MainThread 7976 True

2018-06-13 10:18:47,544 MainThread 7976 result=103

2018-06-13 10:18:47,544 MainThread 7976 True

2018-06-13 10:18:47,545 MainThread 7976 result=104

2018-06-13 10:18:47,545 MainThread 7976 True

2018-06-13 10:18:47,545 MainThread 7976 result=105

2018-06-13 10:18:47,587 MainThread 7976 -------end--------

2018-06-13 10:18:47,587 MainThread 7976 [<_MainThread(MainThread, started 7976)>]

 

总结,统一了线程池、进程池调用,简化了编程。

 

无法设置线程名称。

Python中的多进程

标签:tco   RKE   工作   __name__   方式   import   导致   OLE   elf   

原文地址:https://www.cnblogs.com/wangchunli-blogs/p/9949898.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!