c++11 有没有办法防止“正常”在原子操作之前或之后滑动的操作
我有兴趣做类似的事情（单线程更新，多线程读取 banneedURLs）：
atomic<bannedURLList*> bannedURLs;//global variable pointing to the currently used instance of struct
void updateList()
{
//no need for mutex because only 1 thread updates
bannedURLList* newList= new bannedURLList();
bannedURLList* oldList=bannedURLs;
newList->initialize();
bannedURLs=newList;// line must be after previous line, because list must be initialized before it is ready to be used
//while refcnt on the oldList >0 wait, then delete oldList;
}
读取器线程做这样的事情：
{
bannedURLs->refCnt++;
//use bannedURLs
bannedURLs->refCnt--;
}
struct memeber refCnt 也是原子整数 我的问题是如何防止这两行重新排序：
newList->initialize();
bannedURLs=newList;
可以以 std:: 方式完成吗？
I'm interested in doing something like(single thread update, multiple threads read banneedURLs):
atomic<bannedURLList*> bannedURLs;//global variable pointing to the currently used instance of struct
void updateList()
{
//no need for mutex because only 1 thread updates
bannedURLList* newList= new bannedURLList();
bannedURLList* oldList=bannedURLs;
newList->initialize();
bannedURLs=newList;// line must be after previous line, because list must be initialized before it is ready to be used
//while refcnt on the oldList >0 wait, then delete oldList;
}
reader threads do something like this:
{
bannedURLs->refCnt++;
//use bannedURLs
bannedURLs->refCnt--;
}
struct memeber refCnt is also atomic integer
My question is how to prevent reordering of this 2 lines:
newList->initialize();
bannedURLs=newList;
Can it be done in std:: way?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
使用
bannedURLs.store(newList);而不是bannedURLs=newList;。由于您没有传递弱排序说明符，因此这会强制在商店中进行完全排序。Use
bannedURLs.store(newList);instead ofbannedURLs=newList;. Since you didn't pass a weak ordering specifier, this forces full ordering in the store.