本文使用独占锁、原子类型、条件变量,实现线程间同步,API接口模仿Go语言中的 WaitGroup
#ifndef __WAITGROUP_H_
#define __WAITGROUP_H_
#include <mutex>
#include <atomic>
#include <condition_variable>
namespace algo {
class WaitGroup {
public:
WaitGroup() {
m_counter = 0;
}
void Add(int count = 1) {
m_counter += count;
}
void Done() {
m_counter--;
if (m_counter.load() <= 0) {
m_cond.notify_all();
}
}
int GetCount() {
return m_counter.load();
}
void Wait() {
if (m_counter.load() <= 0) {
return;
}
std::unique_lock<std::mutex> lock(m_mutex);
m_cond.wait(lock, [&]() {
return this->m_counter.load() <= 0;
});
}
private:
std::mutex m_mutex;
std::atomic<int> m_counter;
std::condition_variable m_cond;
};
}
#endif
1、mutex
独占锁,用于保证同一时刻只有一个线程进入临界区
2、m_counter
原子类型,用于计数。根据 m_counter 的变化,实现多个线程间的同步
3、m_cond
条件变量,持有条件变量的线程会一直阻塞,直到被通知 或者 某一个注册的条件满足,才会解除阻塞状态
测试代码如下:
启动多个线程,并进行计数,每一个线程设置为分离状态,主线程无需等待子线程退出。
使用 waitgroup 计数后,在最外层等待所有线程结束
#include <iostream>
#include <thread>
#include <chrono>
#include "waitgroup.h"
using namespace algo;
using namespace std;
int main() {
algo::WaitGroup wg;
std::thread thread_list[10];
for (int i = 0; i < 10; ++i) {
wg.Add(1);
thread_list[i] = std::thread([&wg, i]() {
std::this_thread::sleep_for(std::chrono::milliseconds(10));
std::cout << "[DEBUG] thread[" << i << "] exit" << std::endl;
wg.Done();
});
thread_list[i].detach();
}
wg.Wait();
std::cout << "[INFO] all thread to level" << std::endl;
return 0;
}
备注:因为所有子线程是同步运行的,所以打印输出是无序的
结果(每次都是随机的):
[DEBUG] thread[3] exit
[DEBUG] thread[4] exit
[DEBUG] thread[8] exit
[DEBUG] thread[5] exit
[DEBUG] thread[7] exit
[DEBUG] thread[0] exit
[DEBUG] thread[1] exit
[DEBUG] thread[2] exit
[DEBUG] thread[6] exit
[DEBUG] thread[9] exit
[INFO] all thread to level
谢谢阅读~