C++11 并发指南九(综合运用: C++11 多线程下生产者消费者模型详解)

前面八章介绍了 C++11 并发编程的基础(抱歉哈，第五章-第八章还在草稿中)，本文将综合运用 C++11 中的新的基础设施(主要是多线程、锁、条件变量)来阐述一个经典问题——生产者消费者模型，并给出完整的解决方案。

生产者消费者问题是多线程并发中一个非常经典的问题，相信学过操作系统课程的同学都清楚这个问题的根源。本文将就四种情况分析并介绍生产者和消费者问题，它们分别是：单生产者-单消费者模型，单生产者-多消费者模型，多生产者-单消费者模型，多生产者-多消费者模型，我会给出四种情况下的 C++11 并发解决方案，如果文中出现了错误或者你对代码有异议，欢迎交流 ;-)。

单生产者-单消费者模型

顾名思义，单生产者-单消费者模型中只有一个生产者和一个消费者，生产者不停地往产品库中放入产品，消费者则从产品库中取走产品，产品库容积有限制，只能容纳一定数目的产品，如果生产者生产产品的速度过快，则需要等待消费者取走产品之后，产品库不为空才能继续往产品库中放置新的产品，相反，如果消费者取走产品的速度过快，则可能面临产品库中没有产品可使用的情况，此时需要等待生产者放入一个产品后，消费者才能继续工作。C++11实现单生产者单消费者模型的代码如下：

#include <unistd.h>

#include <cstdlib>
#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>

static const int kItemRepositorySize  = ; // Item buffer size.
static const int kItemsToProduce  = ;   // How many items we plan to produce.

struct ItemRepository {
    int item_buffer[kItemRepositorySize]; // 产品缓冲区, 配合 read_position 和 write_position 模型环形队列.
    size_t read_position; // 消费者读取产品位置.
    size_t write_position; // 生产者写入产品位置.
    std::mutex mtx; // 互斥量,保护产品缓冲区
    std::condition_variable repo_not_full; // 条件变量, 指示产品缓冲区不为满.
    std::condition_variable repo_not_empty; // 条件变量, 指示产品缓冲区不为空.
} gItemRepository; // 产品库全局变量, 生产者和消费者操作该变量.

typedef struct ItemRepository ItemRepository;

void ProduceItem(ItemRepository *ir, int item)
{
    std::unique_lock<std::mutex> lock(ir->mtx);
    while(((ir->write_position + ) % kItemRepositorySize)
        == ir->read_position) { // item buffer is full, just wait here.
        std::cout << "Producer is waiting for an empty slot...\n";
        (ir->repo_not_full).wait(lock); // 生产者等待"产品库缓冲区不为满"这一条件发生.
    }

    (ir->item_buffer)[ir->write_position] = item; // 写入产品.
    (ir->write_position)++; // 写入位置后移.

    if (ir->write_position == kItemRepositorySize) // 写入位置若是在队列最后则重新设置为初始位置.
        ir->write_position = ;

    (ir->repo_not_empty).notify_all(); // 通知消费者产品库不为空.
    lock.unlock(); // 解锁.
}

int ConsumeItem(ItemRepository *ir)
{
    int data;
    std::unique_lock<std::mutex> lock(ir->mtx);
    // item buffer is empty, just wait here.
    while(ir->write_position == ir->read_position) {
        std::cout << "Consumer is waiting for items...\n";
        (ir->repo_not_empty).wait(lock); // 消费者等待"产品库缓冲区不为空"这一条件发生.
    }

    data = (ir->item_buffer)[ir->read_position]; // 读取某一产品
    (ir->read_position)++; // 读取位置后移

    if (ir->read_position >= kItemRepositorySize) // 读取位置若移到最后，则重新置位.
        ir->read_position = ;

    (ir->repo_not_full).notify_all(); // 通知消费者产品库不为满.
    lock.unlock(); // 解锁.

    return data; // 返回产品.
}

void ProducerTask() // 生产者任务
{
    for (int i = ; i <= kItemsToProduce; ++i) {
        // sleep(1);
        std::cout << "Produce the " << i << "^th item..." << std::endl;
        ProduceItem(&gItemRepository, i); // 循环生产 kItemsToProduce 个产品.
    }
}

void ConsumerTask() // 消费者任务
{
    static int cnt = ;
    while() {
        sleep();
        int item = ConsumeItem(&gItemRepository); // 消费一个产品.
        std::cout << "Consume the " << item << "^th item" << std::endl;
        if (++cnt == kItemsToProduce) break; // 如果产品消费个数为 kItemsToProduce, 则退出.
    }
}

void InitItemRepository(ItemRepository *ir)
{
    ir->write_position = ; // 初始化产品写入位置.
    ir->read_position = ; // 初始化产品读取位置.
}

int main()
{
    InitItemRepository(&gItemRepository);
    std::thread producer(ProducerTask); // 创建生产者线程.
    std::thread consumer(ConsumerTask); // 创建消费之线程.
    producer.join();
    consumer.join();
}

单生产者-多消费者模型

与单生产者和单消费者模型不同的是，单生产者-多消费者模型中可以允许多个消费者同时从产品库中取走产品。所以除了保护产品库在多个读写线程下互斥之外，还需要维护消费者取走产品的计数器，代码如下:

#include <unistd.h>

#include <cstdlib>
#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>

static const int kItemRepositorySize  = ; // Item buffer size.
static const int kItemsToProduce  = ;   // How many items we plan to produce.

struct ItemRepository {
    int item_buffer[kItemRepositorySize];
    size_t read_position;
    size_t write_position;
    size_t item_counter;
    std::mutex mtx;
    std::mutex item_counter_mtx;
    std::condition_variable repo_not_full;
    std::condition_variable repo_not_empty;
} gItemRepository;

typedef struct ItemRepository ItemRepository;

void ProduceItem(ItemRepository *ir, int item)
{
    std::unique_lock<std::mutex> lock(ir->mtx);
    while(((ir->write_position + ) % kItemRepositorySize)
        == ir->read_position) { // item buffer is full, just wait here.
        std::cout << "Producer is waiting for an empty slot...\n";
        (ir->repo_not_full).wait(lock);
    }

    (ir->item_buffer)[ir->write_position] = item;
    (ir->write_position)++;

    if (ir->write_position == kItemRepositorySize)
        ir->write_position = ;

    (ir->repo_not_empty).notify_all();
    lock.unlock();
}

int ConsumeItem(ItemRepository *ir)
{
    int data;
    std::unique_lock<std::mutex> lock(ir->mtx);
    // item buffer is empty, just wait here.
    while(ir->write_position == ir->read_position) {
        std::cout << "Consumer is waiting for items...\n";
        (ir->repo_not_empty).wait(lock);
    }

    data = (ir->item_buffer)[ir->read_position];
    (ir->read_position)++;

    if (ir->read_position >= kItemRepositorySize)
        ir->read_position = ;

    (ir->repo_not_full).notify_all();
    lock.unlock();

    return data;
}

void ProducerTask()
{
    for (int i = ; i <= kItemsToProduce; ++i) {
        // sleep(1);
        std::cout << "Producer thread " << std::this_thread::get_id()
            << " producing the " << i << "^th item..." << std::endl;
        ProduceItem(&gItemRepository, i);
    }
    std::cout << "Producer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void ConsumerTask()
{
    bool ready_to_exit = false;
    while() {
        sleep();
        std::unique_lock<std::mutex> lock(gItemRepository.item_counter_mtx);
        if (gItemRepository.item_counter < kItemsToProduce) {
            int item = ConsumeItem(&gItemRepository);
            ++(gItemRepository.item_counter);
            std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is consuming the " << item << "^th item" << std::endl;
        } else ready_to_exit = true;
        lock.unlock();
        if (ready_to_exit == true) break;
    }
    std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void InitItemRepository(ItemRepository *ir)
{
    ir->write_position = ;
    ir->read_position = ;
    ir->item_counter = ;
}

int main()
{
    InitItemRepository(&gItemRepository);
    std::thread producer(ProducerTask);
    std::thread consumer1(ConsumerTask);
    std::thread consumer2(ConsumerTask);
    std::thread consumer3(ConsumerTask);
    std::thread consumer4(ConsumerTask);

    producer.join();
    consumer1.join();
    consumer2.join();
    consumer3.join();
    consumer4.join();
}

多生产者-单消费者模型

与单生产者和单消费者模型不同的是，多生产者-单消费者模型中可以允许多个生产者同时向产品库中放入产品。所以除了保护产品库在多个读写线程下互斥之外，还需要维护生产者放入产品的计数器，代码如下:

#include <unistd.h>

#include <cstdlib>
#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>

static const int kItemRepositorySize  = ; // Item buffer size.
static const int kItemsToProduce  = ;   // How many items we plan to produce.

struct ItemRepository {
    int item_buffer[kItemRepositorySize];
    size_t read_position;
    size_t write_position;
    size_t item_counter;
    std::mutex mtx;
    std::mutex item_counter_mtx;
    std::condition_variable repo_not_full;
    std::condition_variable repo_not_empty;
} gItemRepository;

typedef struct ItemRepository ItemRepository;

void ProduceItem(ItemRepository *ir, int item)
{
    std::unique_lock<std::mutex> lock(ir->mtx);
    while(((ir->write_position + ) % kItemRepositorySize)
        == ir->read_position) { // item buffer is full, just wait here.
        std::cout << "Producer is waiting for an empty slot...\n";
        (ir->repo_not_full).wait(lock);
    }

    (ir->item_buffer)[ir->write_position] = item;
    (ir->write_position)++;

    if (ir->write_position == kItemRepositorySize)
        ir->write_position = ;

    (ir->repo_not_empty).notify_all();
    lock.unlock();
}

int ConsumeItem(ItemRepository *ir)
{
    int data;
    std::unique_lock<std::mutex> lock(ir->mtx);
    // item buffer is empty, just wait here.
    while(ir->write_position == ir->read_position) {
        std::cout << "Consumer is waiting for items...\n";
        (ir->repo_not_empty).wait(lock);
    }

    data = (ir->item_buffer)[ir->read_position];
    (ir->read_position)++;

    if (ir->read_position >= kItemRepositorySize)
        ir->read_position = ;

    (ir->repo_not_full).notify_all();
    lock.unlock();

    return data;
}

void ProducerTask()
{
    bool ready_to_exit = false;
    while() {
        sleep();
        std::unique_lock<std::mutex> lock(gItemRepository.item_counter_mtx);
        if (gItemRepository.item_counter < kItemsToProduce) {
            ++(gItemRepository.item_counter);
            ProduceItem(&gItemRepository, gItemRepository.item_counter);
            std::cout << "Producer thread " << std::this_thread::get_id()
                << " is producing the " << gItemRepository.item_counter
                << "^th item" << std::endl;
        } else ready_to_exit = true;
        lock.unlock();
        if (ready_to_exit == true) break;
    }
    std::cout << "Producer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void ConsumerTask()
{
    static int item_consumed = ;
    while() {
        sleep();
        ++item_consumed;
        if (item_consumed <= kItemsToProduce) {
            int item = ConsumeItem(&gItemRepository);
            std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is consuming the " << item << "^th item" << std::endl;
        } else break;
    }
    std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void InitItemRepository(ItemRepository *ir)
{
    ir->write_position = ;
    ir->read_position = ;
    ir->item_counter = ;
}

int main()
{
    InitItemRepository(&gItemRepository);
    std::thread producer1(ProducerTask);
    std::thread producer2(ProducerTask);
    std::thread producer3(ProducerTask);
    std::thread producer4(ProducerTask);
    std::thread consumer(ConsumerTask);

    producer1.join();
    producer2.join();
    producer3.join();
    producer4.join();
    consumer.join();
}

多生产者-多消费者模型

该模型可以说是前面两种模型的综合，程序需要维护两个计数器，分别是生产者已生产产品的数目和消费者已取走产品的数目。另外也需要保护产品库在多个生产者和多个消费者互斥地访问。

代码如下：

#include <unistd.h>

#include <cstdlib>
#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>

static const int kItemRepositorySize  = ; // Item buffer size.
static const int kItemsToProduce  = ;   // How many items we plan to produce.

struct ItemRepository {
    int item_buffer[kItemRepositorySize];
    size_t read_position;
    size_t write_position;
    size_t produced_item_counter;
    size_t consumed_item_counter;
    std::mutex mtx;
    std::mutex produced_item_counter_mtx;
    std::mutex consumed_item_counter_mtx;
    std::condition_variable repo_not_full;
    std::condition_variable repo_not_empty;
} gItemRepository;

typedef struct ItemRepository ItemRepository;

void ProduceItem(ItemRepository *ir, int item)
{
    std::unique_lock<std::mutex> lock(ir->mtx);
    while(((ir->write_position + ) % kItemRepositorySize)
        == ir->read_position) { // item buffer is full, just wait here.
        std::cout << "Producer is waiting for an empty slot...\n";
        (ir->repo_not_full).wait(lock);
    }

    (ir->item_buffer)[ir->write_position] = item;
    (ir->write_position)++;

    if (ir->write_position == kItemRepositorySize)
        ir->write_position = ;

    (ir->repo_not_empty).notify_all();
    lock.unlock();
}

int ConsumeItem(ItemRepository *ir)
{
    int data;
    std::unique_lock<std::mutex> lock(ir->mtx);
    // item buffer is empty, just wait here.
    while(ir->write_position == ir->read_position) {
        std::cout << "Consumer is waiting for items...\n";
        (ir->repo_not_empty).wait(lock);
    }

    data = (ir->item_buffer)[ir->read_position];
    (ir->read_position)++;

    if (ir->read_position >= kItemRepositorySize)
        ir->read_position = ;

    (ir->repo_not_full).notify_all();
    lock.unlock();

    return data;
}

void ProducerTask()
{
    bool ready_to_exit = false;
    while() {
        sleep();
        std::unique_lock<std::mutex> lock(gItemRepository.produced_item_counter_mtx);
        if (gItemRepository.produced_item_counter < kItemsToProduce) {
            ++(gItemRepository.produced_item_counter);
            ProduceItem(&gItemRepository, gItemRepository.produced_item_counter);
            std::cout << "Producer thread " << std::this_thread::get_id()
                << " is producing the " << gItemRepository.produced_item_counter
                << "^th item" << std::endl;
        } else ready_to_exit = true;
        lock.unlock();
        if (ready_to_exit == true) break;
    }
    std::cout << "Producer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void ConsumerTask()
{
    bool ready_to_exit = false;
    while() {
        sleep();
        std::unique_lock<std::mutex> lock(gItemRepository.consumed_item_counter_mtx);
        if (gItemRepository.consumed_item_counter < kItemsToProduce) {
            int item = ConsumeItem(&gItemRepository);
            ++(gItemRepository.consumed_item_counter);
            std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is consuming the " << item << "^th item" << std::endl;
        } else ready_to_exit = true;
        lock.unlock();
        if (ready_to_exit == true) break;
    }
    std::cout << "Consumer thread " << std::this_thread::get_id()
                << " is exiting..." << std::endl;
}

void InitItemRepository(ItemRepository *ir)
{
    ir->write_position = ;
    ir->read_position = ;
    ir->produced_item_counter = ;
    ir->consumed_item_counter = ;
}

int main()
{
    InitItemRepository(&gItemRepository);
    std::thread producer1(ProducerTask);
    std::thread producer2(ProducerTask);
    std::thread producer3(ProducerTask);
    std::thread producer4(ProducerTask);

    std::thread consumer1(ConsumerTask);
    std::thread consumer2(ConsumerTask);
    std::thread consumer3(ConsumerTask);
    std::thread consumer4(ConsumerTask);

    producer1.join();
    producer2.join();
    producer3.join();
    producer4.join();

    consumer1.join();
    consumer2.join();
    consumer3.join();
    consumer4.join();
}

另外，所有例子的代码(包括前面一些指南的代码均放在github上)，希望对大家学习 C++11 多线程并发有所帮助。