How could relaxing R->R result in something different from SC (sequential consistency)? It seems like the order of reads wouldn't affect a program unless they were intermixed with other writes.
Olorin
@regi I think this can happen in program 2 on slide 15. If we allow the read of A in thread 2 to move before the read of B, then I believe that we might see "1 0" be output (because the read of A could well happen before the write of A, then we switch to thread 1, which runs to completion, then the read of B completes).
How could relaxing R->R result in something different from SC (sequential consistency)? It seems like the order of reads wouldn't affect a program unless they were intermixed with other writes.
@regi I think this can happen in program 2 on slide 15. If we allow the read of A in thread 2 to move before the read of B, then I believe that we might see "1 0" be output (because the read of A could well happen before the write of A, then we switch to thread 1, which runs to completion, then the read of B completes).