内存障碍解决iOS ARM弱内存写入排序？

Question

The ARM processors in iOS devices are reported to not to enforce the ordering of memory writes. This means that if one processor writes a data struct or array and then updates a valid flag, another processor could see the updated valid flag before the data gets to memory, and thus read garbage. This is true even if the flag variable is small (atomic sized) and declared volatile.

Is there any way (in Swift or Objective C) to tell Xcode to compile in some ARM memory barrier instructions, intrinsics, or equivalent calls, to guarantee that data is written to memory (as visible to other processors on the same chip) before subsequent valid flag updates are issued?

This is for real-time code where taking locks in real-time callbacks is not permitted.

Answer 1

Look at libkern/OSAtomic.h which should include everything you need.

Or look at whatever the latest C and C++ Standard support; that will work on iOS as well.