Threading 串行VS并发

Threading 串行VS并发

IO 密集型

结果

>>> 串行开始...
下载https://www.baidu.com/成功，状态码为200
下载https://www.sina.com.cn/成功，状态码为200
下载https://www.bilibili.com/成功，状态码为412
耗时：0.42061281204223633
>>> 串行结束...
>>> 并发开始...
下载https://www.baidu.com/成功，状态码为200
下载https://www.sina.com.cn/成功，状态码为200
下载https://www.bilibili.com/成功，状态码为412
耗时：0.1446540355682373
>>> 并发结束...

示例代码

# -*- coding: utf-8 -*-
# I/O密集时多线程例子：一边打印一边下载网页内容 串行VS并发
import requests
import threading
import time# 用几个响应速度较快的真实网页
urls = ["https://www.baidu.com/","https://www.sina.com.cn/","https://www.bilibili.com/"
]def download(url):try:r = requests.get(url, timeout=3)print(f"下载{url}成功，状态码为{r.status_code}")except requests.RequestException as e:print(e)# ===== 串行
def run_serial():print(">>> 串行开始...")start_time = time.time()for url in urls:download(url)end_time = time.time()print(f"耗时：{end_time - start_time}")print(">>> 串行结束...")# ==== 并发（多线程）
def run_threading():print(">>> 并发开始...")start_time = time.time()threads = []for url in urls:t = threading.Thread(target=download, args=(url,))  # target=函数名，args=函数参数（元组）threads.append(t)t.start()  # 线程开始执行for t in threads:t.join()  # join 主线程等待子线程执行完end_time = time.time()print(f"耗时：{end_time - start_time}")print(">>> 并发结束...")if __name__ == '__main__':run_serial()run_threading()

计算密集型

结果

CPU密集任务运行时，并行并不会比串行快多少，达不到加速的效果

>>> 开始串行...
>>> 串行耗时： 73.33004999160767 秒
>>> 开始并发...
>>> 并发耗时： 73.11574292182922 秒

示例代码

# -*- coding: utf-8 -*-
import mathimport requests
import threading
import time# “烧 CPU”的函数：对一大批数字做平方根运算（math.sqrt()）N = 500000000def cpu_task():total = 0for i in range(N):total += math.sqrt(i)return total# ==== 串行
def run_serial():print(">>> 开始串行...")start_time = time.time()cpu_task()cpu_task()cpu_task()print(">>> 串行耗时：", time.time() - start_time, "秒")# ==== 并发
def run_threaded():print(">>> 开始并发...")start_time = time.time()threas = []for _ in range(3):t = threading.Thread(target=cpu_task)t.start()threas.append(t)for t in threas:t.join()print(">>> 并发耗时：", time.time() - start_time, "秒")if __name__ == '__main__':run_serial()run_threaded()

总结：多线程的优势与局限

I/O 密集	请求网页、读写文件、等待数据库	✅ 非常适合
CPU 密集	图像处理、大量计算、加密解密	❌ 会被 GIL 限制，不如用多进程
少量任务	任务不多，时间也短	⚠️ 线程切换反而是开销
并发数量极大	上千并发、强隔离	⚠️ 推荐协程/进程池，线程管理成本高

参考来源

https://zhuanlan.zhihu.com/p/1923003098605033459