This is not simply a cosmetic issue. ink_atomic.h effectively assumes there is no need for memory fence instructions on any architecture that TS runs on:
|
/* not used for Intel Processors which have sequential(esque) consistency */ |
For x86 architectures for example, this is not true for any version after the 386. It may also be true, for some architectures (or compiler options), that the alignment requirements when doing atomic access are stricter than those for non-atomic access.