📝Atomic ordering
- tags
- Relaxed
No order imposed (i.e., this load/store can be reordered with other loads/stores). Only atomicity and modification order consistency is guaranteed.
- Release-Acquire
All memory writes (non-atomic and relaxed atomic) before the atomic (release) store become visible to a thread that uses (acquire) load.
- Release-Consume
Similar to Release-Acquire but the order is only established between dependencies. : this options requires compiler to track dependencies and doesn’t seem to be implemented widely.
Note that currently (2/2015) no known production compilers track dependency chains: consume operations are lifted to acquire operations.
The specification of release-consume ordering is being revised, and the use of memoryorderconsume is temporarily discouraged. (since C++17)
- Sequentially-consistent
Similar to Release-Acquire but additionally establishes a single global order of all atomic operations that are so tagged.
This is the strongest guarantee and is used by default in C++. It is a safe default choice but carries performance penalty.
Total sequential ordering requires a full memory fence CPU instruction on all multi-core systems. This may become a performance bottleneck since it forces the affected memory accesses to propagate to every core.