python多线程与c++多线程对比

查了一下说是python对于线程有个全局的大锁GIL。这个互斥锁保护了内存安全，但是将多线程在多核上的并行运算功能也干掉了。

也就是说python的多线程是多个线程交替运行，同一时间只会有一个线程在运行，可以同步推进各项工作，但不能利用多核实现并行运算提高速度。

参考链接： https://www.cnblogs.com/xhnxhnu/p/9728239.html

https://encyclopedia.thefreedictionary.com/Global+interpreter+lock

实际测试：

from threading import Thread
import time

def my_counter():
    i = 0
    inf = 2147483647
    for _ in range(inf):
        i = i + 1
    return True

def single():
    print("顺序运算")
    n=6
    thread_array = {}
    start_time = time.time()
    for tid in range(n):
        t = Thread(target=my_counter)
        print("start")
        t.start()
        print("join")
        t.join()
    end_time = time.time()
    print("Total time: {}".format(end_time - start_time))

def multi():
    print("同步运算")
    n=6
    thread_array = {}
    start_time = time.time()
    for tid in range(n):
        t = Thread(target=my_counter)
        print("start")
        t.start()
        thread_array[tid] = t
    for i in range(n):
        print("join")
        thread_array[i].join()
    end_time = time.time()
    print("Total time: {}".format(end_time - start_time))

if __name__ == '__main__':
    single()
    multi()

运行结果：

D:\python\python_install\python.exe D:/python/python_data/multi_thread_test/main.py
顺序运算
start
join
start
join
start
join
start
join
start
join
start
join
Total time: 496.3798580169678
同步运算
start
start
start
start
start
start
join
join
join
join
join
join
Total time: 496.61106872558594

Process finished with exit code 0

然后我们再来看一下c++的多线程，运行同样功能的程序

// thread.cpp : 此文件包含 "main" 函数。程序执行将在此处开始并结束。
//
#include<iostream>
#include<thread>
#include<future>
#include<ctime>
using namespace std;

int n = 12;

bool func()
{
    int t = 0;
    int inf = ~(1<<31);
    for (int i = 0; i != inf; i++)
    {
        t++;
    }
    return true;
}

void single()
{
    cout << "顺序执行" << endl;
    clock_t start_t, end_t;
    start_t = clock();
    vector<future<bool>> v;
    for (int i = 0; i != n; i++)
    {
        v.push_back(async(func));
        bool ans = v[i].get();
        cout << ans << endl;
    }
    end_t = clock();
    cout << "cost time: " <<double(end_t-start_t)/1000 <<"s"<< endl;
}

void multi()
{
    cout << "同时执行" << endl;
    clock_t start_t, end_t;
    start_t = clock();
    vector<future<bool>> v;
    for (int i = 0; i != n; i++)
    {
        v.push_back(async(func));
    }

    for (int i = 0; i != n; i++)
    {
        bool ans = v[i].get();
        cout << ans << endl;
    }

    end_t = clock();
    cout << "cost time: " << double(end_t-start_t) / 1000 << "s" << endl;
}

int main()
{
    single();
    multi();
    return 0;
}

运行结果:

1
1
1
1
1
1
1
1
1
1
cost time: 20.581s
同时执行
1
1
1
1
1
1
1
1
1
1
1
1
cost time: 3.219s

哎，这天上差地下啊，跟c++一比这python就跟蜗牛一样，艹

估算一下 （496.3798580169678/6） / （20.581/12）

=82.7299763361613/1.715083333333333

=48.23670939390045

也就是说单线程c++的速度要比python快约48倍

（496.61106872558594/6）/（3.219/12）

= 82.76851145426432/0.26825

=308.5499029049928

多线程c++比python快约308倍

下面是我的cpu