I think the clarification is good, probably the amount of optimizations that are prevented by treating volatile and atomics as UB barriers is limited, but as yo

I think the clarification is good, probably the amount of optimizations that are prevented by treating volatile and atomics as UB barriers is limited, but as your example show, a lot of very surprising transformations are still allowed. Unfortunately I don't think there is a good fix for that.