C++ 以最优雅的方式同步线程

Question

我正在尝试解决以下问题，我知道有多种解决方案，但我正在寻找最优雅的方法（更少的代码）来解决它。

我有 4 个线程，其中 3 个尝试在无限循环中将唯一值（0、1 或 2）写入易失性整数变量，第四个线程尝试读取此变量的值并将值打印到标准输出也处于无限循环中。

我想在线程之间同步，这样写 0 的线程将被运行，然后是“打印”线程，然后是写 1 的线程，然后是打印线程，依此类推......所以最后我期望在“打印”线程的输出中看到一个零序列，然后是 1 序列，然后是 2，然后是 0 等等......

在这些线程之间同步的最优雅和最简单的方法是什么？

这是程序代码：

volatile int value;
int thid[4];

int main() {
    HANDLE handle[4];
    for (int ii=0;ii<4;ii++) {
        thid[ii]=ii;
        handle[ii] = (HANDLE) CreateThread( NULL, 0, (LPTHREAD_START_ROUTINE)                 ThreadProc, &thid[ii], 0, NULL);
    }
    return 0;
}

void WINAPI ThreadProc( LPVOID param ) {
    int h=*((int*)param);

    switch (h) {
        case 3:
            while(true) {
                cout << value << endl;
            }
            break;
        default:
            while(true) {
                // setting a unique value to the volatile variable
                value=h;
            }
            break;
    }
}

Answer 1

您的问题可以通过生产者消费者模式解决。我的灵感来自维基百科，所以如果你想了解更多细节，这里是链接。

https://en.wikipedia.org/wiki/Producer%E2%80%93consumer_problem

我使用随机数生成器来生成 volatile 变量，但您可以更改该部分。

这是代码：它可以在样式方面进行改进（对随机数使用 C++11），但它产生了你所期望的。

#include <iostream>
#include <sstream>
#include <vector>
#include <stack>
#include <thread>
#include <mutex>
#include <atomic>
#include <condition_variable>
#include <chrono>
#include <stdlib.h>     /* srand, rand */
using namespace std;

//random number generation
std::mutex mutRand;//mutex for random number generation (given that the random generator is not thread safe).
int GenerateNumber()
{
    std::lock_guard<std::mutex> lk(mutRand);
    return rand() % 3;
}

// print function for "thread safe" printing using a stringstream
void print(ostream& s) { cout << s.rdbuf(); cout.flush(); s.clear(); }

//      Constants
//
const int num_producers = 3;                //the three producers of random numbers
const int num_consumers = 1;                //the only consumer
const int producer_delay_to_produce = 10;   // in miliseconds
const int consumer_delay_to_consume = 30;   // in miliseconds

const int consumer_max_wait_time = 200;     // in miliseconds - max time that a consumer can wait for a product to be produced.

const int max_production = 1;              // When producers has produced this quantity they will stop to produce
const int max_products = 1;                // Maximum number of products that can be stored

//
//      Variables
//
atomic<int> num_producers_working(0);       // When there's no producer working the consumers will stop, and the program will stop.
stack<int> products;                        // The products stack, here we will store our products
mutex xmutex;                               // Our mutex, without this mutex our program will cry

condition_variable is_not_full;             // to indicate that our stack is not full between the thread operations
condition_variable is_not_empty;            // to indicate that our stack is not empty between the thread operations

//
//      Functions
//

//      Produce function, producer_id will produce a product
void produce(int producer_id)
{
    while (true)
    {
        unique_lock<mutex> lock(xmutex);
        int product;

        is_not_full.wait(lock, [] { return products.size() != max_products; });
        product = GenerateNumber();
        products.push(product);

        print(stringstream() << "Producer " << producer_id << " produced " << product << "\n");
        is_not_empty.notify_all();
    }

}

//      Consume function, consumer_id will consume a product
void consume(int consumer_id)
{
    while (true)
    {
        unique_lock<mutex> lock(xmutex);
        int product;

        if(is_not_empty.wait_for(lock, chrono::milliseconds(consumer_max_wait_time),
                [] { return products.size() > 0; }))
        {
                product = products.top();
                products.pop();

                print(stringstream() << "Consumer " << consumer_id << " consumed " << product << "\n");
                is_not_full.notify_all();
        }
    }

}

//      Producer function, this is the body of a producer thread
void producer(int id)
{
        ++num_producers_working;
        for(int i = 0; i < max_production; ++i)
        {
                produce(id);
                this_thread::sleep_for(chrono::milliseconds(producer_delay_to_produce));
        }

        print(stringstream() << "Producer " << id << " has exited\n");
        --num_producers_working;
}

//      Consumer function, this is the body of a consumer thread
void consumer(int id)
{
        // Wait until there is any producer working
        while(num_producers_working == 0) this_thread::yield();

        while(num_producers_working != 0 || products.size() > 0)
        {
                consume(id);
                this_thread::sleep_for(chrono::milliseconds(consumer_delay_to_consume));
        }

        print(stringstream() << "Consumer " << id << " has exited\n");
}

//
//      Main
//

int main()
{
        vector<thread> producers_and_consumers;

        // Create producers
        for(int i = 0; i < num_producers; ++i)
                producers_and_consumers.push_back(thread(producer, i));

        // Create consumers
        for(int i = 0; i < num_consumers; ++i)
                producers_and_consumers.push_back(thread(consumer, i));

        // Wait for consumers and producers to finish
        for(auto& t : producers_and_consumers)
                t.join();
        return 0;
}

希望有帮助，如果您需要更多信息或者您不同意某事，请告诉我:-)

祝所有法国人巴士底日好！

Answer 2

如果要同步线程，则使用同步对象以“乒乓”或“滴答”模式保持每个线程。在 C++ 11 中，您可以使用条件变量，这里的示例显示了与您所要求的类似的内容。