https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104480
--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> --- (In reply to Jonathan Wakely from comment #4) > Nothing in your code imposes an ordering on those stores. Another thread > cannot observe it, because you are using non-atomic operations. To be more precise, another thread cannot even observe *one* of those stores without a data race. It certainly can't observe both and conclude they happened out of order. To avoid a data race the other thread must ensure that the entire call to f happens-before any read of either memory location. And if the entire call happens-before the read, you can't see which order the stores happened in.
