抄自维基百科 :
生产者消费者问题(英语:Producer-consumer problem),也称有限缓冲问题(英语:Bounded-buffer problem),是一个多线程同步问题的经典案例。该问题描述了共享固定大小缓冲区的两个线程——即所谓的“生产者”和“消费者”——在实际运行时会发生的问题。生产者的主要作用是生成一定量的数据放到缓冲区中,然后重复此过程。与此同时,消费者也在缓冲区消耗这些数据。该问题的关键就是要保证生产者不会在缓冲区满时加入数据,消费者也不会在缓冲区中空时消耗数据。
要解决该问题,就必须让生产者在缓冲区满时休眠(要么干脆就放弃数据),等到下次消费者消耗缓冲区中的数据的时候,生产者才能被唤醒,开始往缓冲区添加数据。同样,也可以让消费者在缓冲区空时进入休眠,等到生产者往缓冲区添加数据之后,再唤醒消费者。
本文用一个ItemRepository类表示产品仓库,其中包含一个数组和两个坐标表示的环形队列、一个std::mutex成员、用来保证每次只被一个线程读写操作 (为了保证打印出来的消息是一行一行的,在它空闲的时候也借用的这个互斥量╮(╯▽╰)╭)、两个std::condition_variable表示队列不满和不空的状态,进而保证生产的时候不满,消耗的时候不空。
#pragma once #include <chrono>//std::chrono #include <mutex>//std::mutex,std::unique_lock,std::lock_guard #include <thread>//std::thread #include <condition_variable>//std::condition_variable #include <iostream>//std::cout,std::endl #include <map>//std::map namespace MyProducerToConsumer { static const int gRepositorySize = 10;//total size of the repository static const int gItemNum = 97;//number of products to produce std::mutex produce_mtx, consume_mtx;//mutex for all the producer thread or consumer thread std::map<std::thread::id, int> threadPerformance;//records of every thread's producing/consuming number struct ItemRepository {//repository class int m_ItemBuffer[gRepositorySize];//Repository itself (as a circular queue) int m_ProducePos;//rear position of circular queue int m_ConsumePos;//head position of circular queue std::mutex m_mtx;//mutex for operating the repository std::condition_variable m_RepoUnfull;//indicating that this repository is unfull(then producers can produce items) std::condition_variable m_RepoUnempty;//indicating that this repository is unempty(then consumers can produce items) }gItemRepo; void ProduceItem(ItemRepository *ir, int item) { std::unique_lock <std::mutex>ulk(ir->m_mtx); while ((ir->m_ProducePos + 1) % gRepositorySize == ir->m_ConsumePos) {//full(spare one slot for indicating) std::cout << "Reposity is full. Waiting for consumers..." << std::endl; ir->m_RepoUnfull.wait(ulk);//unlocking ulk and waiting for unfull condition } //when unfull ir->m_ItemBuffer[ir->m_ProducePos++] = item;//procude and shift std::cout << "Item No." << item << " produced successfully by " <<std::this_thread::get_id()<<"!" << std::endl; threadPerformance[std::this_thread::get_id()]++; if (ir->m_ProducePos == gRepositorySize)//loop ir->m_ProducePos = 0; ir->m_RepoUnempty.notify_all();//item produced, so it's unempty; notify all consumers } int ConsumeItem(ItemRepository *ir) { std::unique_lock<std::mutex>ulk(ir->m_mtx); while (ir->m_ConsumePos == ir->m_ProducePos) {//empty std::cout << "Repository is empty.Waiting for producing..." << std::endl; ir->m_RepoUnempty.wait(ulk); } int item = ir->m_ItemBuffer[ir->m_ConsumePos++]; std::cout << "Item No." << item << " consumed successfully by " <<std::this_thread::get_id()<<"!" << std::endl; threadPerformance[std::this_thread::get_id()]++; if (ir->m_ConsumePos == gRepositorySize) ir->m_ConsumePos = 0; ir->m_RepoUnfull.notify_all();//item consumed, so it's unempty; notify all consumers return item; } void ProducerThread() { static int produced = 0;//static variable to indicate the number of produced items while (1) { std::this_thread::sleep_for(std::chrono::milliseconds(10));//sleep long enough in case it runs too fast for other threads to procude std::lock_guard<std::mutex>lck(produce_mtx);//auto unlock when break produced++; if (produced > gItemNum)break; gItemRepo.m_mtx.lock(); std::cout << "Producing item No." << produced << "..." << std::endl; gItemRepo.m_mtx.unlock(); ProduceItem(&gItemRepo, produced); } gItemRepo.m_mtx.lock(); std::cout << "Producer thread " << std::this_thread::get_id() << " exited." << std::endl; gItemRepo.m_mtx.unlock(); } void ConsumerThread() { static int consumed = 0; while (1) { std::this_thread::sleep_for(std::chrono::milliseconds(10)); std::lock_guard<std::mutex>lck(consume_mtx); consumed++; if (consumed > gItemNum)break; gItemRepo.m_mtx.lock(); std::cout << "Consuming item available..." << std::endl; gItemRepo.m_mtx.unlock(); ConsumeItem(&gItemRepo); } gItemRepo.m_mtx.lock(); std::cout << "Consumer thread " << std::this_thread::get_id() << " exited." << std::endl; gItemRepo.m_mtx.unlock(); } void InitItemRepository(ItemRepository* ir) { ir->m_ConsumePos = 0; ir->m_ProducePos = 0; } void Run() { InitItemRepository(&gItemRepo); std::thread thdConsume[11]; std::thread thdProduce[11]; for (auto& t : thdConsume)t = std::thread(ConsumerThread); for (auto& t : thdProduce)t = std::thread(ProducerThread); for (auto& t : thdConsume)t.join(); for (auto& t : thdProduce)t.join(); for (auto& iter : threadPerformance)cout << iter.first << ":" << iter.second << endl; } }
相关文章:
Atas ialah kandungan terperinci C++解决方法:多线程同步经典案例之生产者消费者问题. Untuk maklumat lanjut, sila ikut artikel berkaitan lain di laman web China PHP!