Python是运行在解释器中的语言,查找资料知道,python中有一个全局锁(GIL),在使用多进程(Thread)的情况下,不能发挥多核的优势。而使用多进程(Multiprocess),则可以发挥多核的优势真正地提高效率。
对比实验
资料显示,如果多线程的进程是CPU密集型的,那多线程并不能有多少效率上的提升,相反还可能会因为线程的频繁切换,导致效率下降,推荐使用多进程;如果是IO密集型,多线程进程可以利用IO阻塞等待时的空闲时间执行其他线程,提升效率。所以我们根据实验对比不同场景的效率
操作系统 | CPU | 内存 | 硬盘 |
---|---|---|---|
Windows 10 | 双核 | 8GB | 机械硬盘 |
(1)引入所需要的模块
import requestsimport timefrom threading import Threadfrom multiprocessing import Process
(2)定义CPU密集的计算函数
def count(x, y): # 使程序完成50万计算 c = 0 while c < 500000: c += 1 x += x y += y
(3)定义IO密集的文件读写函数
def write(): f = open("test.txt", "w") for x in range(5000000): f.write("testwrite\n") f.close()def read(): f = open("test.txt", "r") lines = f.readlines() f.close()
(4) 定义网络请求函数
_head = { 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/48.0.2564.116 Safari/537.36'}url = "http://www.tieba.com"def http_request(): try: webPage = requests.get(url, headers=_head) html = webPage.text return {"context": html} except Exception as e: return {"error": e}
(5)测试线性执行IO密集操作、CPU密集操作所需时间、网络请求密集型操作所需时间
# CPU密集操作t = time.time()for x in range(10): count(1, 1)print("Line cpu", time.time() - t)# IO密集操作t = time.time()for x in range(10): write() read()print("Line IO", time.time() - t)# 网络请求密集型操作t = time.time()for x in range(10): http_request()print("Line Http Request", time.time() - t)
输出
CPU密集:95.6059999466、91.57099986076355 92.52800011634827、 99.96799993515015
IO密集:24.25、21.76699995994568、21.769999980926514、22.060999870300293
网络请求密集型: 4.519999980926514、8.563999891281128、4.371000051498413、4.522000074386597、14.671000003814697
(6)测试多线程并发执行CPU密集操作所需时间
counts = []t = time.time()for x in range(10): thread = Thread(target=count, args=(1,1)) counts.append(thread) thread.start()e = counts.__len__()while True: for th in counts: if not th.is_alive(): e -= 1 if e <= 0: breakprint(time.time() - t)
Output: 25.69700002670288、24.02400016784668
(7)测试多线程并发执行IO密集操作所需时间
def io(): write() read()t = time.time()ios = []t = time.time()for x in range(10): thread = Thread(target=count, args=(1,1)) ios.append(thread) thread.start()e = ios.__len__()while True: for th in ios: if not th.is_alive(): e -= 1 if e <= 0: breakprint(time.time() - t)
Output: 99.9240000248 、101.26400017738342、102.32200002670288
(8)测试多线程并发执行网络密集操作所需时间
t = time.time()ios = []t = time.time()for x in range(10): thread = Thread(target=http_request) ios.append(thread) thread.start()e = ios.__len__()while True: for th in ios: if not th.is_alive(): e -= 1 if e <= 0: breakprint("Thread Http Request", time.time() - t)
Output: 0.7419998645782471、0.3839998245239258、0.3900001049041748
(9)测试多进程并发执行CPU密集操作所需时间
counts = []t = time.time()for x in range(10): process = Process(target=count, args=(1,1)) counts.append(process) process.start()e = counts.__len__()while True: for th in counts: if not th.is_alive(): e -= 1 if e <= 0: breakprint("Multiprocess cpu", time.time() - t)
Output: 54.342000007629395、53.437999963760376
(10)测试多进程并发执行IO密集型操作
t = time.time()ios = []t = time.time()for x in range(10): process = Process(target=io) ios.append(process) process.start()e = ios.__len__()while True: for th in ios: if not th.is_alive(): e -= 1 if e <= 0: breakprint("Multiprocess IO", time.time() - t)
Output: 12.509000062942505、13.059000015258789
(11)测试多进程并发执行Http请求密集型操作
t = time.time()httprs = []t = time.time()for x in range(10): process = Process(target=http_request) ios.append(process) process.start()e = httprs.__len__()while True: for th in httprs: if not th.is_alive(): e -= 1 if e <= 0: breakprint("Multiprocess Http Request", time.time() - t)
Output: 0.5329999923706055、0.4760000705718994
实验结果
CPU密集型操作 | IO密集型操作 | 网络请求密集型操作 | |
---|---|---|---|
线性操作 | 94.91824996469 | 22.46199995279 | 7.3296000004 |
多线程操作 | 101.1700000762 | 24.8605000973 | 0.5053332647 |
多进程操作 | 53.8899999857 | 12.7840000391 | 0.5045000315 |
通过上面的结果,我们可以看到:
多线程在IO密集型的操作下似乎也没有很大的优势(也许IO操作的任务再繁重一些就能体现出优势),在CPU密集型的操作下明显地比单线程线性执行性能更差,但是对于网络请求这种忙等阻塞线程的操作,多线程的优势便非常显著了
多进程无论是在CPU密集型还是IO密集型以及网络请求密集型(经常发生线程阻塞的操作)中,都能体现出性能的优势。不过在类似网络请求密集型的操作上,与多线程相差无几,但却更占用CPU等资源,所以对于这种情况下,我们可以选择多线程来执行
原文地址:http://blog.atomicer.cn/2016/09/30/Python%E4%B8%AD%E5%A4%9A%E7%BA%BF%E7%A8%8B%E5%92%8C%E5%A4%9A%E8%BF%9B%E7%A8%8B%E7%9A%84%E5%AF%B9%E6%AF%94/