�定义的c语言类型,如c_double

继承标准库ctypes其中的Structure对象创建的自定义结构体, c_double

elapsed time: 0.50s elapsed time: 1.02s OK 1

elapsed time: 0.50s elapsed time: 1.02s OK 2

为了更加方便的在主进程获取到子进程中的数据,我们可以,下例就是一个例子.

elapsed time: 0.50s elapsed time: 1.02s OK 3

elapsed time: 0.50s elapsed time: 1.02s OK 4

1.7.1. 高级共享multiprocessing.Manager

之前介绍了queue,pipe,array和value,这些都太具体底层,有没有什么方法可以像处理python容器一样简单地处理数据共享的问题呢?multiprocess提供一个manager模块.

Manager()返回的manager对象控制了一个server进程,此进程包含的python对象可以被其他的进程通过proxies来访问.从而达到多进程间数据通信且安全。

Manager支持的类型有

list
dict
Namespace
Lock
RLock
Semaphore
BoundedSemaphore
Condition
Event
Queue
Value
Array。

elapsed time: 0.50s elapsed time: 1.02s OK 5

elapsed time: 0.50s elapsed time: 1.02s OK 6

namespace对象没有公共的方法,但是有可写的属性

elapsed time: 0.50s elapsed time: 1.02s OK 7

elapsed time: 0.50s elapsed time: 1.02s OK 8

1.8. 使用Unix信号管理子进程

类似在多线程中控制子线程,我们使用os.kill(pid,sig)来控制子进程.

elapsed time: 0.50s elapsed time: 1.02s OK 9

def crypto_process(size, key):     in_text = bytearray(randrange(256) for i in range(size))     cypher_text = arcfour(key, in_text)     out_text = arcfour(key, cypher_text)     assert in_text == out_text, 'Failed arcfour_test'     return size  def main(workers=None):     JOBS = 12     SIZE = 2**18      KEY = b"'Twas brillig, and the slithy toves\nDid gyre"     STATUS = '{} workers, elapsed time: {:.2f}s'     if workers:         workers = int(workers)     t0 = time()      with ProcessPoolExecutor(workers) as executor:         actual_workers = executor._max_workers         to_do = []         for i in range(JOBS, 0, -1):             size = SIZE + int(SIZE / JOBS * (i - JOBS/2))             job = executor.submit(crypto_process, size, KEY)             to_do.append(job)          for future in as_completed(to_do):             res = future.result()             print('{:.1f} KB'.format(res/2**10))      print(STATUS.format(actual_workers, time() - t0)) 0