1、std::thread
在C++11之前,C++语言层面是不支持多线程的,想利用C++实现并发程序,借助操作系统的API实现跨平台的并发程序存在着诸多不便,当C++11在语言层面支持多线程后,编写跨平台的多线程代码就方便了许多。
C++11提供的std::thread在开发多线程方面带来了便捷。
#include <iostream>
#include <thread>
void threadfunc()
{
std::cout << "thread func" << std::endl;
}
int main()
{
std::thread t1(threadfunc);
t1.join(); //等待threadfunc运行结束
return 0;
}
首先定义线程对象t1,线程函数threadfunc运行在线程对象t1中,当线程创建成功并执行线程函数后,一定要保证线程函数运行结束才能退出,这里调用了join()函数阻塞线程,直到threadfunc()运行结束,回收对应创建线程的资源。如果不阻塞线程,就不能保证线程对象t1在threadfunc()运行期间有效,下面不调用join()阻塞线程。
#include <iostream>
#include <thread>
void threadfunc()
{
std::cout << "thread func" << std::endl;
}
int main()
{
std::thread t1(threadfunc);
//t1.join(); //等待threadfunc运行结束
return 0;
}
在运行时引起了程序崩溃。
除了调用join()阻塞线程,保证线程对象在线程函数运行期间的有效性,还可以通过线程分离的手段实现,调用detach()函数使得线程对象与线程函数分离,这样,在线程函数运行期间,线程对象与线程函数就没有联系了,此时的线程是作为后台线程去执行,detach()后就无法再和线程发生联系,也不能通过join()来等待线程执行完毕,线程何时执行完无法控制,它的资源会被init进程回收,所以,通常不采用detach()方法。
#include <iostream>
#include <thread>
void threadfunc()
{
std::cout << " detach thread func" << std::endl;
}
int main()
{
std::thread t1(threadfunc);
t1.detach(); //线程分离
return 0;
}
这里调用detach()实现线程分离,但是运行后,主线程退出的时候threadfunc()还没有输出“detach thread func”,threadfunc()什么时候运行结束也无法确定,为了看到所创建的线程运行结果,在主线程等待一下再退出。
#include <iostream>
#include <thread>
#include <chrono> //时间
void threadfunc()
{
std::cout << "detach thread func" << std::endl;
}
int main()
{
std::thread t1(threadfunc);
t1.detach();
while (true)
{
std::this_thread::sleep_for(std::chrono::milliseconds(1000));//睡眠1000毫秒
break;
}
return 0;
}
此时运行结果:
detach thread func
通过std::thread创建的线程是不可以复制的,但是可以移动。
#include <iostream>
#include <thread>
#include <chrono>
void threadfunc()
{
std::cout << "move thread func" << std::endl;
}
int main()
{
std::thread t1(threadfunc);
std::thread t2(std::move(t1));
t2.join();
while (true)
{
std::this_thread::sleep_for(std::chrono::milliseconds(1000));//睡眠1000毫秒
break;
}
return 0;
}
输出结果:
move thread func
移动后t1就不代表任何线程了,t2对象代表着线程threadfunc()。另外,还可以通过std::bind来创建线程函数。
#include <iostream>
#include <thread>
#include <chrono> //时间
#include <functional> //std::bind
class A {
public:
void threadfunc()
{
std::cout << "bind thread func" << std::endl;
}
};
int main()
{
A a;
std::thread t1(std::bind(&A::threadfunc,&a));
t1.join();
while (true)
{
std::this_thread::sleep_for(std::chrono::milliseconds(1000));//睡眠1000毫秒
break;
}
return 0;
}
创建一个类A,然后再main函数中将类A中的成员函数绑定到线程对象t1上,运行结果:
bind thread func
每个线程都有自己的线程标识,也就是线程ID,当线程创建成功后,可以通过get_id()来获取线程的ID。
#include <iostream>
#include <thread>
#include <chrono>
#include <functional>
class A {
public:
void threadfunc()
{
std::cout << "bind thread func" << std::endl;
}
};
int main()
{
A a;
std::thread t1(std::bind(&A::threadfunc,&a));
std::cout << "main thread ID is : " << std::this_thread::get_id() << std::endl;
std::cout << "t1 thread ID is : " << t1.get_id() << std::endl;
t1.join();
while (true)
{
std::this_thread::sleep_for(std::chrono::milliseconds(1000));//睡眠1000毫秒
break;
}
return 0;
}
std::this_thread::get_id()获取的是当前线程的ID,t1.get_id()获取的是所创建的t1对象中运行的线程ID,对应的ID分别为:
main thread ID is : 11932
t1 thread ID is : 12076
bind thread func
虽然get_id()可以获取线程的ID,但是其返回类型是thread::id,通过std::cout可以输出线程ID,但是这样使用似乎不太方面,要是能转换为整形就好了。其实可以将得到的线程ID写入到ostreamstring流中,转换成string类型,再转换成整形。
#include <iostream>
#include <thread>
#include <chrono>
#include <functional>
#include <sstream>
class A {
public:
void threadfunc()
{
std::cout << "bind thread func" << std::endl;
}
};
int main()
{
A a;
std::thread t1(std::bind(&A::threadfunc, &a));
std::ostringstream os1;
os1 << t1.get_id() << std::endl;
std::string strID = os1.str(); //转换成string类型
int threadID = atoi(strID.c_str()); //转换成int类型
std::cout << "t1 thread ID is : " << threadID << std::endl;
t1.join();
while (true)
{
std::this_thread::sleep_for(std::chrono::milliseconds(1000));//睡眠1000毫秒
break;
}
return 0;
}
输出结果:
t1 thread ID is : 6956
bind thread func
2、std::mutex
进入多线程编程的世界,除了要牢牢掌握std::thread使用方法,还要掌握互斥量(锁)的使用,这是一种线程同步机制,在C++11中提供了4中互斥量。
std::mutex; //非递归的互斥量
std::timed_mutex; //带超时的非递归互斥量
std::recursive_mutex; //递归互斥量
std::recursive_timed_mutex; //带超时的递归互斥量
从各种互斥量的名字可以看出其具有的特性,在实际开发中,常用就是std::mutex,它就像是一把锁,我们需要做的就是对它进行加锁与解锁。
#include <iostream>
#include <thread>
#include <mutex>
#include <chrono>
std::mutex g_mutex;
void func()
{
std::cout << "entry func test thread ID is : " << std::this_thread::get_id() << std::endl;
std::this_thread::sleep_for(std::chrono::microseconds(1000));
std::cout << "leave func test thread ID is : " << std::this_thread::get_id() << std::endl;
}
int main()
{
std::thread t1(func);
std::thread t2(func);
std::thread t3(func);
std::thread t4(func);
std::thread t5(func);
t1.join();
t2.join();
t3.join();
t4.join();
t5.join();
return 0;
}
创建了5个线程,然后分别调用func()函数,得到结果:
entry func test thread ID is : entry func test thread ID is : 19180
entry func test thread ID is : 3596
13632
entry func test thread ID is : 9520
entry func test thread ID is : 4460
leave func test thread ID is : 13632
leave func test thread ID is : 19180
leave func test thread ID is : leave func test thread ID is : 9520
3596
leave func test thread ID is : 4460
可以看出,并没有按顺序去执行线程函数,后面创建的线程并没有等待前面的线程执行完毕,导致结果混乱,下面用std::mutex进行控制:
#include <iostream>
#include <thread>
#include <mutex>
#include <chrono>
std::mutex g_mutex;
void func()
{
g_mutex.lock();
std::cout << "entry func test thread ID is : " << std::this_thread::get_id() << std::endl;
std::this_thread::sleep_for(std::chrono::microseconds(1000));
std::cout << "leave func test thread ID is : " << std::this_thread::get_id() << std::endl;
g_mutex.unlock();
}
int main()
{
std::thread t1(func);
std::thread t2(func);
std::thread t3(func);
std::thread t4(func);
std::thread t5(func);
t1.join();
t2.join();
t3.join();
t4.join();
t5.join();
return 0;
}
只要线程进入func()函数就进行加锁处理,当线程执行完毕后进行解锁,保证每个线程都能按顺序执行,输出结果:
entry func test thread ID is : 8852
leave func test thread ID is : 8852
entry func test thread ID is : 15464
leave func test thread ID is : 15464
entry func test thread ID is : 17600
leave func test thread ID is : 17600
entry func test thread ID is : 16084
leave func test thread ID is : 16084
entry func test thread ID is : 4156
leave func test thread ID is : 4156
虽然通过lock()与unlock()可以解决线程之间的资源竞争问题,但是这里也存在不足。
func()
{
//加锁
执行逻辑处理; //如果该过程抛出异常导致程序退出了,就没法unlock
//解锁
}
int main()
{
......
}
func()中再执行逻辑处理中程序因为某些原因退出了,此时就无法unlock()了,这样其他线程也就无法获取std::mutex,造成死锁现象,其实在加锁之前可以通过trylock()尝试一下能不能加锁。实际开发中,通常也不会这样写代码,而是采用lock_guard来控制std::mutex。
template <class _Mutex>
class lock_guard {
public:
using mutex_type = _Mutex;
explicit lock_guard(_Mutex& _Mtx) : _MyMutex(_Mtx)
{
_MyMutex.lock(); //构造函数加锁
}
lock_guard(_Mutex& _Mtx, adopt_lock_t) : _MyMutex(_Mtx)
{
}
~lock_guard() noexcept
{
_MyMutex.unlock(); //析构函数解锁
}
lock_guard(const lock_guard&) = delete;
lock_guard& operator=(const lock_guard&) = delete;
private:
_Mutex& _MyMutex;
};
lock_guard是类模板,在其构造函数中自动给std::mutex加锁,在退出作用域的时候自动解锁,这样就可以保证std::mutex的正确操作,这也是RAII(获取资源便初始化)技术的体现。
#include <iostream>
#include <thread>
#include <mutex>
#include <chrono>
std::mutex g_mutex;
void func()
{
std::lock_guard<std::mutex> lock(g_mutex); //加锁
std::cout << "entry func test thread ID is : " << std::this_thread::get_id() << std::endl;
std::this_thread::sleep_for(std::chrono::microseconds(1000));
std::cout << "leave func test thread ID is : " << std::this_thread::get_id() << std::endl;
//退出作用域后,lock_guard对象析构就自动解锁
}
int main()
{
std::thread t1(func);
std::thread t2(func);
std::thread t3(func);
std::thread t4(func);
std::thread t5(func);
t1.join();
t2.join();
t3.join();
t4.join();
t5.join();
return 0;
}
运行结果:
entry func test thread ID is : 19164
leave func test thread ID is : 19164
entry func test thread ID is : 15124
leave func test thread ID is : 15124
entry func test thread ID is : 2816
leave func test thread ID is : 2816
entry func test thread ID is : 17584
leave func test thread ID is : 17584
entry func test thread ID is : 15792
leave func test thread ID is : 15792
3、std::condition_variable
条件变量是C++11提供的另外一种线程同步机制,通过判断条件是否满足,决定是否阻塞线程,当线程执行条件满足的时候就会唤醒阻塞的线程,常与std::mutex配合使用,C++11提供了两种条件变量。
- std::condition_variable,配合std::unique_lockstd::mutex使用,通过wait()函数阻塞线程;
- std::condition_variable_any,可以和任意带有lock()、unlock()语义的std::mutex搭配使用,比较灵活,但是其效率不及std::condition_variable;
std::unique_lock:C++11提供的 std::unique_lock 是通用互斥包装器,允许延迟锁定、锁定的有时限尝试、递归锁定、所有权转移和与条件变量一同使用。std::unique_lock比std::lock_guard使用更加灵活,功能更加强大。使用std::unique_lock需要付出更多的时间、性能成本。
下面利用std::mutex与std::condition_variable实现生产者与消费者模式。
#include <iostream>
#include <condition_variable>
#include <thread>
#include <list>
#include <mutex>
#include <chrono>
class CTask {
public:
CTask(int taskID)
{
this->taskId = taskID;
}
void dotask()
{
std::cout << "consumer a task Id is " << taskId << std::endl;
}
private:
int taskId;
};
std::list<std::shared_ptr<CTask>> g_task;
std::mutex g_mutex;
std::condition_variable g_conv;
//生产者线程
void ProdecerFunc()
{
int n_taskId = 0;
std::shared_ptr<CTask> ptask = nullptr;
while (true)
{
ptask = std::make_shared<CTask >(n_taskId); //创建任务
{
std::lock_guard<std::mutex> lock(g_mutex);
g_task.push_back(ptask);
std::cout << "produce a task Id is " << n_taskId << std::endl;
}
//唤醒线程
g_conv.notify_one();
n_taskId++;
std::this_thread::sleep_for(std::chrono::milliseconds(1000));
}
}
//消费者线程
void ConsumerFunc()
{
std::shared_ptr<CTask> ptask = nullptr;
while (true)
{
std::unique_lock<std::mutex> lock(g_mutex);
while (g_task.empty()) //即使被唤醒还要循环判断一次,防止虚假唤醒
{
g_conv.wait(lock);
}
ptask = g_task.front(); //取出任务
g_task.pop_front();
if (ptask == nullptr)
{
continue;
}
ptask->dotask(); //执行任务
}
}
int main()
{
std::thread t1(ConsumerFunc);
std::thread t2(ConsumerFunc);
std::thread t3(ConsumerFunc);
std::thread t4(ProdecerFunc);
t1.join();
t2.join();
t3.join();
t4.join();
return 0;
}
创建3个消费者线程,一个生产者线程,当存放任务的std::list为空时,消费者线程阻塞,当生产者线程生产一个任务放入std::list中时候,此时满足条件,条件变量就可以唤醒阻塞的线程去执行任务。
produce a task Id is 0
consumer a task Id is 0
produce a task Id is 1
consumer a task Id is 1
produce a task Id is 2
consumer a task Id is 2
produce a task Id is 3
consumer a task Id is 3
produce a task Id is 4
consumer a task Id is 4
produce a task Id is 5
consumer a task Id is 5
produce a task Id is 6
consumer a task Id is 6
produce a task Id is 7
consumer a task Id is 7
......
条件变量的使用过程可以归纳如下:
- 拥有条件变量的线消费者程获取互斥锁;
- 消费者线程循环检查条件是否满足,不满足则阻塞等待,此时释放互斥锁;
- 当生产者线程产生任务后,调用notify_one()或者notify_all()唤醒阻塞的消费者线程;
- 当消费者线程被唤醒后再次获得互斥锁去执行任务;
4、thread_local
C++11中提供了thread_local,thread_local定义的变量在每个线程都保存一份副本,而且互不干扰,在线程退出的时候自动销毁。
#include <iostream>
#include <thread>
#include <chrono>
thread_local int g_k = 0;
void func1()
{
while (true)
{
++g_k;
}
}
void func2()
{
while (true)
{
std::cout << "func2 thread ID is : " << std::this_thread::get_id() << std::endl;
std::cout << "func2 g_k = " << g_k << std::endl;
std::this_thread::sleep_for(std::chrono::milliseconds(1000));
}
}
int main()
{
std::thread t1(func1);
std::thread t2(func2);
t1.join();
t2.join();
return 0;
}
在func1()对g_k循环加1操作,在func2()每个1000毫秒输出一次g_k的值:
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
func2 thread ID is : 15312
func2 g_k = 0
......
可以看出func2()中的g_k始终保持不变。