C ++中的非阻塞线程安全队列？ [英] non-blocking thread-safe queue in C++?

问题描述

在C ++中有一个线程安全的，非阻塞的队列类吗？

可能是一个基本的问题，但我一直没有做C ++很长时间

解决方案

/ div>

假设你的CPU有一个双指针宽的比较和交换（compxchg8b在486或更高，compxchg16b在大多数amd64机器[不存在于一些早期的型号由英特尔]）有一个算法此处。

更新：如果你不害怕做一点工作，不难把它翻译成C ++。 ：P

此算法假设一个带标签的指针结构，如下所示：

  //注意，复制此结构必须以原子方式执行... 
 
 template< class T> 
 struct pointer 
 {
 T * ptr; 
 uintptr_t tag; 
}; 
  

然后你想包装指令 lock cmpxchg {8 | 16} b

也许你可以这样写队列节点：

 模板< class T> 
 struct queue_node 
 {
 T value; 
指针< queue_node< T> >下一个; 
}; 
  

其余或多或少是我链接到...的算法的转录

Is there a thread-safe, non-blocking queue class in the C++?

Probably a basic question but I haven't been doing C++ for a long time...

EDIT: removed STL requirement.

解决方案

Assuming your CPU has a double-pointer-wide compare-and-swap (compxchg8b on 486 or higher, compxchg16b on most amd64 machines [not present on some early models by Intel])... There is an algorithm here.

Update: It's not hard to translate this to C++ if you aren't afraid of doing a bit of work. :P

This algorithm assumes a "pointer with tag" structure which looks like this:

// Be aware that copying this structure has to be done atomically...

template <class T>
struct pointer
{
   T *ptr;
   uintptr_t tag;
};

Then you want to wrap the instructions lock cmpxchg{8|16}b with some inline asm...

Maybe then you can write the queue node like this:

template <class T>
struct queue_node
{
    T value;
    pointer<queue_node<T> > next;
};

The rest is more or less a transcription of the algorithm I linked to...

这篇关于C ++中的非阻塞线程安全队列？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！