Question

我正在尝试使用多线程加速磁盘上的写入时间。为了看到有效性，我编写了以下代码，测量两种方法之间的时间：

from __future__ import print_function
import random, time
import numpy as np
import threading


def file_write(filepath, V):
    filepath = open(filepath, 'w')
    np.savetxt(filepath, V, fmt='%i')
    filepath.close()


if __name__ == "__main__":
    filepath = 'output.txt'
    V = 10000
    dim = 128

    print('Starting generating ...')
    V_ = [random.randint(0, 2 ** dim) for _ in range(V)]
    V = list(set(V_))
    V = [('{0:0' + str(dim) + 'b}').format(x) for x in V]
    V = np.asarray([list(map(int, list(x))) for x in V], dtype=np.uint8)

    # This is Non-Threading write
    t0 = time.time()
    file_write(filepath, V)
    t1 = time.time()
    print('\n\n>> No-Thread Writing Time= %.2f '% (t1-t0))

    # Generating 10 threads
    thread_list = []
    for i in range(1, 10):
        t = threading.Thread(target=file_write, args=(filepath, V))
        thread_list.append(t)

    # Starts threads
    t0 = time.time()
    for thread in thread_list:
        thread.start()

    # Closing threads
    for thread in thread_list:
        thread.join()

    t1 = time.time()
    print('\n\n>> 10-Threads Writing Time= %.2f ' % (t1 - t0))

结果令人惊讶如下。线程比非线程工作需要更长的时间。不应该是相反的吗？我认为线程更快！

Starting generating ...
>> No-Thread Writing Time= 0.26 
>> 10-Threads Writing Time= 2.44

线程工作会导致意外的执行时间延长

0 个答案: