volatile in C++

理解volatile变量的内存特性及使用注意事项

最新推荐文章于 2023-06-23 19:40:06 发布

原创最新推荐文章于 2023-06-23 19:40:06 发布 · 890 阅读

0 ·

CC 4.0 BY-SA版权

C++ 专栏收录该内容

49 篇文章

订阅专栏

本文深入探讨了volatile变量在多线程环境下的内存特性，包括它与原子操作的区别、编译器优化限制、单线程环境中的行为以及如何正确使用以避免常见陷阱。

1) "volatile" has nothing to do with multithreading. It is not atomic.
2) The compiler will not reorder the sequence of "volatile" variables, but may reorder the sequence of "volatile" and non-volatile. CPU may reorder the execution of "volatile" variables (CPU memory order, depending on the CPU vendor).

3) In a single threaded env, "volatile int vi = 0", vi = 1; int read_back = vi; assert (read_back == 1); the assertion may fail. WHY ? (think of "vi" is mapping to an hardware port. Memory mapped I/O).
4) The compiler will not do any optimization for volatile variables. It generates code which always read the value from the memory.
5) "volatile" doesn't guarantee Acquire-Release semantics. "Acquire": read an atomic, the code after the "read of an atomic" will NOT be reordered before the read. "Release" : write an atomic, the code before the "write of an atomic" will NOT be reordered after the write (everything will be committed to memory, and seen by other CPUs).