The typical definition of an RMW instruction with e.g. sequential consistency is an instruction where no memory operation can happen between the load and the store in the global memory order.

Is there a reason to use this definition? Why is it not sufficient to just require that no operation to the same memory address can happen between the load and the store? Are these definitions equivalent? What about if you relax the memory model to something like x86 TSO?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.