前言

• 在阅读Apollo Planning代码的dp_st_graph部分的时候，遇到future这个多线程接口，特地来记录一下。

用法一：等待task执行完毕

#include <iostream>
#include <future>
#include <vector>

using namespace std;
void task(string c){
  for (int i = 0; i < 10; i++){
    cout<<c;
  }
}

int main(){
  vector<future<void>> results;
  results.emplace_back(std::async(std::launch::async, task, "A"));
  results.emplace_back(std::async(std::launch::async, task, "B"));
  results.emplace_back(std::async(std::launch::async, task, "C"));
  
  for (int i = 0; i < 10; i++){
    cout<<"D";
  }

  for (auto& result : results){
    result.get();//强制阻塞main线程，直到对应的task线程执行完毕
  }
}

用法二：同步获取异步结果

#include <iostream>
#include <future>
#include <vector>


using namespace std;

int task(string c){
  if (c == "A"){
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(3));
    cout<<c<<endl;
    return 1;
  }

  else if (c == "B"){
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(2));
    cout<<c<<endl;
    return 2;
  }

  else{
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    std::this_thread::sleep_for(chrono::seconds(1));
    cout<<c<<endl;
    return 3;
  }

}

int main(){
  vector<future<int>> results;
  results.emplace_back(std::async(std::launch::async, task, "A"));
  results.emplace_back(std::async(std::launch::async, task, "B"));
  results.emplace_back(std::async(std::launch::async, task, "C"));

  cout<<"D"<<endl;

  for (auto& result : results){
    int tmp = result.get();//异步结束的task，同步获取返回值
    cout<<"get "<<tmp<<endl;
  }
}

• 三个task结束的时间不一样，但是程序获取task返回值是一起获取的，