std::condition_variable

条件变量std::condition_variable有wait和notify接口用于线程间的同步。如下图所示，Thread 2阻塞在wait接口，Thread 1通过notify接口通知Thread 2继续执行。

具体参见示例代码：

<code>#include<iostream>

#include<mutex>

#include<thread>

#include<queue>

std::mutex mt;

std::queue<int> data;

std::condition_variable cv;

auto start=std::chrono::high_resolution_clock::now();

void logCurrentTime()

{

auto end = std::chrono::high_resolution_clock::now();

auto elapsed = std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count();

std::cout << elapsed << ":";

}

void prepare_data()

{

logCurrentTime();

std::cout << "this is " << __FUNCTION__ << " thread:" << std::this_thread::get_id() << std::endl;

for (int i = 0; i < 10; i++)

{

data.push(i);

logCurrentTime();

std::cout << "data OK:" << i << std::endl;

}

//start to notify consume_data thread data is OK!

cv.notify_one();

}

void consume_data()

{

logCurrentTime();

std::cout << "this is: " << __FUNCTION__ << " thread:" << std::this_thread::get_id() << std::endl;

std::unique_lock<std::mutex> lk(mt);

//wait first for notification

cv.wait(lk); //it must accept a unique_lock parameter to wait

while (!data.empty())

{

logCurrentTime();

std::cout << "data consumed: " << data.front() << std::endl;

data.pop();

}

}

int main()

{

std::thread t2(consume_data);

//wait for a while to wait first then prepare data,otherwise stuck on wait

std::this_thread::sleep_for(std::chrono::milliseconds(10));

std::thread t1(prepare_data);

t1.join();

t2.join();

return 0;

}

输出结果

分析

主线程中另启两个线程，分别执行consume_data和prepare_data，其中consume_data要先执行，以保证先等待再通知，否则若先通知再等待就死锁了。首先consume_data线程在从wait 处阻塞等待。后prepare_data线程中依次向队列写入0-10，写完之后通过notify_one 通知consume_data线程解除阻塞，依次读取0-10。

std::future

std::future与std::async配合异步执行代码，再通过wait或get接口阻塞当前线程等待结果。如下图所示，Thread 2中future接口的get或wait接口会阻塞当前线程，std::async异步开启的新线程Thread1执行结束后，将结果存于std::future后通知Thread 1获取结果后继续执行.

具体参见如下代码：

<code>#include <iostream>

#include <future>

#include<thread>

int test()

{

std::cout << "this is " << __FUNCTION__ << " thread:" << std::this_thread::get_id() << std::endl;;

std::this_thread::sleep_for(std::chrono::microseconds(1000));

return 10;

}

int main()

{

std::cout << "this is " <<__FUNCTION__<<" thread:" << std::this_thread::get_id() << std::endl;;

//this will lanuch on another thread

std::future<int> result = std::async(test);

std::cout << "After lanuch a thread: "<< std::this_thread::get_id() << std::endl;

//block the thread and wait for the result

std::cout << "result is: " <<result.get()<< std::endl;

std::cout << "After get result "<< std::endl;

return 0;

}

输出结果

分析

主程序中调用std::async异步调用test函数，可以看到main函数的线程ID 27428与test函数执行的线程ID 9704并不一样，说明std::async另起了一个新的线程。在test线程中，先sleep 1000ms，所以可以看到"After lanuch a thread:"先输出，说明主线程异步执行，不受子线程影响。而"After get result "最后输出，说明get()方法会阻塞主线程，直到获取结果。