问题描述
根据此主题，，其中的数据结构是在并行for循环中的 shared std :: vector？主要方面是速度，向量可能需要在循环期间调整大小。
Based on this thread, OpenMP and STL vector, which data structures are good alternatives for a shared std::vector in a parallel for loop? The main aspect is speed, and the vector might require resizing during the loop.
推荐答案
在多个线程写入单个容器的情况下，STL向量容器不是线程安全的。这只是真的，如正确地说，如果你调用的方法，可以导致重新分配的底层数组 std :: vector 成立。 push_back（）， pop_back（）和 insert（）是这些危险方法的示例。
The question you link was talking about the fact that "that STL vector container is not thread-safe in the situation where multiple threads write to a single container" . This is only true, as stated correctly there, if you call methods that can cause reallocation of the underlying array that std::vector holds. push_back(), pop_back() and insert() are examples of these dangerous methods.
如果您需要线程安全重新分配，则为您提供。你不应该在单线程程序中使用tbb :: concurrent_vector，因为访问随机元素所花费的时间比std :: vector需要做的时间更长（这是O（1））。然而，并发向量调用 push_back（）， pop_back（）， insert
If you need thread safe reallocation, then the library intel thread building block offers you concurrent vector containers . You should not use tbb::concurrent_vector in single thread programs because the time it takes to access random elements is higher than the time std::vector takes to do the same (which is O(1)). However, concurrent vector calls push_back(), pop_back(), insert() in a thread safe way, even when reallocation happens.
编辑1：介绍了使用tbb :: concurrent_vector
EDIT 1: The slides 46 and 47 of the following Intel presentation give an illustrative example of concurrent reallocation using tbb::concurrent_vector
编辑2：顺便说一下，如果你开始使用英特尔Tread Building Block（它是开源的，它适用于大多数编译器，它是更好的集成C ++ / C ++ 11功能比openmp），那么你不需要使用openmp创建一个parallel_for，是一个很好的例子，parallel_for使用tbb 。
EDIT 2: By the way, if you start using Intel Tread Building Block (it is open source, it works with most compilers and it is much better integrated with C++/C++11 features than openmp), then you don't need to use openmp to create a parallel_for, Here is a nice example of parallel_for using tbb.
这篇关于C ++ OpenMP Parallel For Loop - std :: vector的替代方法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！