[ekiwi]

> Access to any global variable should always occur direct from memory.

What if your function takes a pointer that might be pointing to a global variable? Does that mean that all accesses through a pointer are now excempt from optimization unless the compiler can prove that the pointer will never point to a global variable?

[cbsmith]

The great thing about simple memory models is they work really well until you think about it. ;-)

[jqpabc123]

What if your function takes a pointer that might be pointing to an "atomic" variable?

Pointers can be used to circumvent most safety measures. If you obscure the access, you should assume responsibility for the result.

[gpderetta]

At least in C++, atomicity is part of the type system, and you would have to explicitly reinterpret cast it away.