C++ 实现生产者和消费者（并发）（多线程）

C++ 实现生产者和消费者（并发）（多线程）

一、解释生产者和消费者模型

就是生产者生产产品，放到缓冲区里面，消费者从缓冲区里面取出产品进行消费。 （假设一次只能生产和消费一个产品。）

主要需要实现的功能是：

1. 生产者生产内容。

2. 消费者消费内容。

3. （阻塞）缓冲区的大小有限制（超过限制，生产者暂停生产），反之就一直等待。

4. （阻塞）只有缓冲区有内容的时候，消费者才可以进行消费，反之就一直等待。

5. 缓冲区需要加锁，才能保证线程安全。

6. 有结束条件 —— 生产n份产品并消费n份产品之后就停止。

生产者消费者模型有啥用（真实应用场景举例）？？

——我本科的时候，感觉这种有点像是工厂，操作系统那么底层的内容会使用吗？
——其实它的应用场景比我本科的时候想象的更加底层。
比如，操作屏幕输出，可能一个CPU读取数据，另一个CPU操作uart（连接了console和keyboard）的buffer，这时候我们希望过程是串行的，因此需要进行同步。uart和它对应的buffer只有一个。这就是一个典型的生产者消费者模型。生产者可能是一个CPU处理得到的输入，队列是uart对应的buffer(存放待输出数据)，消费者是另一个CPU或者说是uart（将CPU传送过来的数据输出到屏幕上）——具体例子可以看uart.c，中的uartputc()函数。（或者看我的博客 6.S081-7中断 - Interrupts ）

二、分析核心代码流程

0. 定义需要的变量

uint64_t num = 0;											// 产品编号

std::queue<uint64_t> buffer_q{};							// 缓冲区
uint64_t max_buffer_q_size = 10;							// 缓冲区队列的最大容量

uint64_t total_production_count = 100;						// 生产多少产品可以结束
bool produce_is_not_finished_flag = true;					// 标记生产者是否还没结束
bool consume_is_not_finished_flag = true;					// 标记消费者是否还没结束

std::mutex mutexLock{};             						// 互斥锁
std::condition_variable condition{};						// 条件变量

1. 生产者生产内容（暂时不考虑锁）

1.1 模拟生产，就是往队列里面添加产品号。

uint64_t data = num++; 			// 生产一个产品（同时产品号++）
buffer_q.push(data);			// 产品添加到缓冲区队列

1.2 生产者生产内容，就是不停生产（不考虑消息队列大小，也不考虑线程安全）

void produce(){
        int produced_production_count = 0;
        while(true){
            uint64_t data = num++;
            buffer_q.push(data);
            std::cout << "task data = " << data << " produced" << std::endl;
            produced_production_count++;
            if(produced_production_count > total_production_count){
                break;    
            }
        }   
    }

1.3 阻塞—— 生产者需要考虑消费者也需要访问缓冲区队列—— 加锁。

void produce(){
    int produced_production_count = 0;
    while(true){
        std::unique_lock<std::mutex> lockGuard(mutexLock);
        uint64_t data = num++;
        buffer_q.push(data);
        std::cout << "task data = " << data << " produced" << std::endl;
        produced_production_count++;
        if(produced_production_count > total_production_count){
            break;       
        }
        lockGuard.unlock();                   
        condition.notify_all();              
    }   
}

注意，这里在过程中加入了锁，并在结束前解锁，并在最后唤醒所有等待线程

std::unique_lock<std::mutex> lockGuard(mutexLock);			//	加锁（不一定能加上，如果信号量当前正在被锁，那么会阻塞）

lockGuard.unlock();											//  解锁
condition.notify_all();										//  唤醒所有等待线程

• 不加condition.notify_all();这句也可以运行，但是过程结束就唤醒其他线程，显然是效率更高的（不然其他进程可能还要“睡一段时间”才会“醒来”）。

• 这里的**lockGuard有一个特性 —— 可以不进行解锁操作（当跳出作用域的时候lockGuard会自动调用析构函数进行解锁操作）**

• 但是这里我还是加了解锁操作。这是因为，lockGuard的析构解锁，是在跳出while的时候执行的（跳出作用域），但是我想通过condition.notify_all(); 唤醒所有进等待的线程，因此我要在此之前就释放锁，而不是在此之后才释放锁。

1.4 阻塞 —— 生产过程中超过缓冲区大小，需要阻塞 —— 通过锁进行wait

void produce(){
    int produced_production_count = 0;
    while(true){
        std::unique_lock<std::mutex> lockGuard(mutexLock);
        while(buffer_q.size() >= max_buffer_q_size){
            condition.wait(lockGuard);      // 等待，并暂时释放锁。（wait的时候就会释放）
        }
        uint64_t data = num++;
        buffer_q.push(data);
        std::cout << "task data = " << data << " produced" << std::endl;
        produced_production_count++;
        if(produced_production_count > total_production_count){
            break;       
        }
        lockGuard.unlock();                   
        condition.notify_all();              
    }   
}

如果当前缓冲区大小大于最大缓冲区大小，那么需要阻塞

while(buffer_q.size() >= max_buffer_q_size){
	condition.wait(lockGuard);      // 等待，并暂时释放锁。（wait的时候就会释放）
}

解释一下上面代码是如何解决“ 生产过程中超过缓冲区大小，需要阻塞 ”的情况的：

• 显然，当生产过程中超过缓冲区大小的时候，我们将会进入while循环，然后执行condition.wait(lockGuard);， 它所进行的操作是，如果当前加锁加上了，那就暂时释放这个锁，并等待（锁可以被加上）。如果当前锁没有加上，那就等待（锁可以被加上）。

• 当消费者消费了物品后，将会让该过程跳出。

wait会将当前在运行的进程存放于sleep数据中 (我猜应该指的就是阻塞队列中)。sleep传入的参数是需要等待的信号，在这个例子中传入的是lockGuard对应的mutexLock的地址。在while过程中，一旦buffer_q中有了空间，会调用与sleep对应的函数wakeup，传入的也是mutexLock的地址。任何等待在这个地址的进程都会被唤醒。有时候这种机制被称为conditional synchronization。

1.5 锁的经典要求 —— 锁的内容的内部，尽量不要出现跳转命令，比如break， return， goto

考虑前面的代码，有个明显的bug：如果我们执行了break那么我们就没有释放锁——会导致死锁~！

while (true) {
    std::unique_lock<std::mutex> lockGuard(mutexLock);
    ... 
    if(produced_production_count > total_production_count){
         break;  
    }
    lockGuard.unlock();  
}

修改方法非常简单，我们将break;跳出改成

while (produce_is_not_finished_flag) {
    std::unique_lock<std::mutex> lockGuard(mutexLock);
    ... 
    if(produced_production_count > total_production_count){
         produce_is_not_finished_flag = false;    
    }
    lockGuard.unlock();  
}

1.6 生产者完整代码

void produce(){
    int produced_production_count = 0;
    while(produce_is_not_finished_flag){
        std::unique_lock<std::mutex> lockGuard(mutexLock);
        while(buffer_q.size() >= max_buffer_q_size){
            condition.wait(lockGuard);      // 等待，并暂时释放锁。（wait的时候就会释放）
        }
        uint64_t data = num++;
        buffer_q.push(data);
        std::cout << "task data = " << data << " produced" << std::endl;
        produced_production_count++;
        if(produced_production_count > total_production_count){
            // break;
            produce_is_not_finished_flag = false;       
        }
        lockGuard.unlock();                   
        condition.notify_all();              
    }   
}

2. 消费者

按照上面流程，可以写出消费者消费代码如下

void consume(){
    while(consume_is_not_finished_flag){
        std::unique_lock<std::mutex> lockGuard(mutexLock);
        while(buffer_q.empty()){
            condition.wait(lockGuard);        
        }
        uint64_t data = buffer_q.front();
        buffer_q.pop();
        std::cout << "task data = " << data << " has been consumed" << std::endl;
        if(!produce_is_not_finished_flag && buffer_q.empty()){
            consume_is_not_finished_flag = false;
        }
        lockGuard.unlock();                   
        condition.notify_all();              
    }  
}

三、多线程调用

1. 开启多线程bind

bind 是用来绑定函数调用的某些参数，可以将bind函数看作一个通用的函数适配器，它接受一个可调用对象，生成新的可调用对象来适应原对象的参数列表。

在构造函数里面使用进行绑定（将参数绑定到可执行的produce / consume函数）。并开启多线程。

producer_consumer(){
    producer = new std::thread(std::bind(&producer_consumer::produce, this));
    consumer = new std::thread(std::bind(&producer_consumer::consume, this));
}

2. 多线程结束：join（析构中join）

通常，我们进行多线程的时候，需要等待所有线程执行完毕了，再进行退出。

~producer_consumer(){
    if(producer -> joinable()){
        producer -> join();
    }
    if(consumer -> joinable()){
        consumer -> join();
    }
    produce_is_not_finished_flag = false;            
    std::cout << "producer_consumer end" << std::endl;
}

四、完整代码

1. 完整producer_consumer.cpp程序

//  g++ producer_consumer.cpp -o producer_consumer -l pthread
#include 
#include                // 锁
#include                 // 任务队列
#include               // 创建线程
#include           // bind
#include   // 条件变量


uint64_t num = 0;
class producer_consumer{
public:
    producer_consumer(){
        producer = new std::thread(std::bind(&producer_consumer::produce, this));
        consumer = new std::thread(std::bind(&producer_consumer::consume, this));
    }
    ~producer_consumer(){
        if(producer -> joinable()){
            producer -> join();
        }
        if(consumer -> joinable()){
            consumer -> join();
        }
        produce_is_not_finished_flag = false;            
        std::cout << "producer_consumer end" << std::endl;
    }

    void produce(){
        int produced_production_count = 0;
        while(produce_is_not_finished_flag){
            std::unique_lock<std::mutex> lockGuard(mutexLock);
            while(buffer_q.size() >= max_buffer_q_size){
                condition.wait(lockGuard);      
            }
            uint64_t data = num++;
            buffer_q.push(data);
            std::cout << "task data = " << data << " produced" << std::endl;
            produced_production_count++;
            if(produced_production_count > total_production_count){
                produce_is_not_finished_flag = false;     
            }
            lockGuard.unlock();                   
            condition.notify_all();               
        }   
    }
    void consume(){
        while(consume_is_not_finished_flag){
            std::unique_lock<std::mutex> lockGuard(mutexLock);
            while(buffer_q.empty()){
                condition.wait(lockGuard);        
            }
            uint64_t data = buffer_q.front();
            buffer_q.pop();
            std::cout << "task data = " << data << " has been consumed" << std::endl;
            if(!produce_is_not_finished_flag && buffer_q.empty()){
                consume_is_not_finished_flag = false;
            }
            lockGuard.unlock();                   
            condition.notify_all();               
        }   
    }

private:
    // uint64_t num = 0;
    std::queue<uint64_t> buffer_q{};
    std::mutex mutexLock{};             
    std::condition_variable condition{};
    std::thread * producer;
    std::thread * consumer;
    bool produce_is_not_finished_flag = true;
    bool consume_is_not_finished_flag = true;
    uint64_t total_production_count = 100;
    uint64_t max_buffer_q_size = 10;
};

static producer_consumer * global_PC;

int main() {
    global_PC = new producer_consumer();
    delete global_PC;
    return 0;
}

2. 编译运行命令

编译的时候需要说明使用多线程的动态库pthread（链接阶段链接这个库）

g++ producer_consumer.cpp -o producer_consumer -l pthread

运行命令

./producer_consumer

3. 运行结果示例

levi@LEVI1:~/code$ g++ producer_consumer.cpp -o producer_consumer -l pthread
levi@LEVI1:~/code$ ./producer_consumer 
task data = 0 produced
task data = 1 produced
task data = 2 produced
task data = 3 produced
task data = 4 produced
task data = 5 produced
task data = 6 produced
task data = 7 produced
task data = 8 produced
task data = 9 produced
task data = 0 has been consumed
task data = 1 has been consumed
task data = 2 has been consumed
task data = 3 has been consumed
task data = 4 has been consumed
task data = 5 has been consumed
task data = 6 has been consumed
task data = 7 has been consumed
task data = 8 has been consumed
task data = 9 has been consumed
task data = 10 produced
task data = 11 produced
task data = 12 produced
task data = 13 produced
task data = 14 produced
task data = 15 produced
task data = 16 produced
task data = 17 produced
task data = 18 produced
task data = 19 produced
task data = 10 has been consumed
task data = 11 has been consumed
task data = 12 has been consumed
task data = 13 has been consumed
task data = 14 has been consumed
task data = 15 has been consumed
task data = 16 has been consumed
task data = 17 has been consumed
task data = 18 has been consumed
task data = 19 has been consumed
task data = 20 produced
task data = 21 produced
task data = 22 produced
task data = 23 produced
task data = 24 produced
task data = 25 produced
task data = 26 produced
task data = 27 produced
task data = 28 produced
task data = 29 produced
task data = 20 has been consumed
task data = 21 has been consumed
task data = 22 has been consumed
task data = 23 has been consumed
task data = 24 has been consumed
task data = 25 has been consumed
task data = 26 has been consumed
task data = 27 has been consumed
task data = 28 has been consumed
task data = 29 has been consumed
task data = 30 produced
task data = 31 produced
task data = 32 produced
task data = 33 produced
task data = 34 produced
task data = 35 produced
task data = 36 produced
task data = 37 produced
task data = 38 produced
task data = 39 produced
task data = 30 has been consumed
task data = 31 has been consumed
task data = 32 has been consumed
task data = 33 has been consumed
task data = 34 has been consumed
task data = 35 has been consumed
task data = 36 has been consumed
task data = 37 has been consumed
task data = 38 has been consumed
task data = 39 has been consumed
task data = 40 produced
task data = 41 produced
task data = 42 produced
task data = 43 produced
task data = 44 produced
task data = 45 produced
task data = 46 produced
task data = 47 produced
task data = 48 produced
task data = 49 produced
task data = 40 has been consumed
task data = 41 has been consumed
task data = 42 has been consumed
task data = 43 has been consumed
task data = 44 has been consumed
task data = 45 has been consumed
task data = 46 has been consumed
task data = 47 has been consumed
task data = 48 has been consumed
task data = 49 has been consumed
task data = 50 produced
task data = 51 produced
task data = 52 produced
task data = 53 produced
task data = 54 produced
task data = 55 produced
task data = 56 produced
task data = 57 produced
task data = 58 produced
task data = 59 produced
task data = 50 has been consumed
task data = 51 has been consumed
task data = 52 has been consumed
task data = 53 has been consumed
task data = 54 has been consumed
task data = 55 has been consumed
task data = 56 has been consumed
task data = 57 has been consumed
task data = 58 has been consumed
task data = 59 has been consumed
task data = 60 produced
task data = 61 produced
task data = 62 produced
task data = 63 produced
task data = 64 produced
task data = 65 produced
task data = 66 produced
task data = 67 produced
task data = 68 produced
task data = 69 produced
task data = 60 has been consumed
task data = 61 has been consumed
task data = 62 has been consumed
task data = 63 has been consumed
task data = 64 has been consumed
task data = 65 has been consumed
task data = 66 has been consumed
task data = 67 has been consumed
task data = 68 has been consumed
task data = 69 has been consumed
task data = 70 produced
task data = 71 produced
task data = 72 produced
task data = 73 produced
task data = 74 produced
task data = 75 produced
task data = 76 produced
task data = 77 produced
task data = 78 produced
task data = 79 produced
task data = 70 has been consumed
task data = 71 has been consumed
task data = 72 has been consumed
task data = 73 has been consumed
task data = 74 has been consumed
task data = 75 has been consumed
task data = 76 has been consumed
task data = 77 has been consumed
task data = 78 has been consumed
task data = 79 has been consumed
task data = 80 produced
task data = 81 produced
task data = 82 produced
task data = 83 produced
task data = 84 produced
task data = 85 produced
task data = 86 produced
task data = 87 produced
task data = 88 produced
task data = 89 produced
task data = 80 has been consumed
task data = 81 has been consumed
task data = 82 has been consumed
task data = 83 has been consumed
task data = 84 has been consumed
task data = 85 has been consumed
task data = 86 has been consumed
task data = 87 has been consumed
task data = 88 has been consumed
task data = 89 has been consumed
task data = 90 produced
task data = 91 produced
task data = 92 produced
task data = 93 produced
task data = 94 produced
task data = 95 produced
task data = 96 produced
task data = 97 produced
task data = 98 produced
task data = 99 produced
task data = 90 has been consumed
task data = 91 has been consumed
task data = 92 has been consumed
task data = 93 has been consumed
task data = 94 has been consumed
task data = 95 has been consumed
task data = 96 has been consumed
task data = 97 has been consumed
task data = 98 has been consumed
task data = 99 has been consumed
task data = 100 produced
task data = 100 has been consumed
producer_consumer end
levi@LEVI1:~/code$