Let's take the situation where two processors have the same cache line in their caches, and after they both modify it once, one processor leaves the data alone (I call this processor "idle") while the other (I call this processor "active") continues to work on it. In an invalidation-based protocol, the "idle" processor would simply invalidate the line and not have to worry about it any more, but in an update protocol, there'd be a lot of traffic as the "active" processor must keep sending line updates to the "idle" processor, which in this case would be useless updates.
Let's take the situation where two processors have the same cache line in their caches, and after they both modify it once, one processor leaves the data alone (I call this processor "idle") while the other (I call this processor "active") continues to work on it. In an invalidation-based protocol, the "idle" processor would simply invalidate the line and not have to worry about it any more, but in an update protocol, there'd be a lot of traffic as the "active" processor must keep sending line updates to the "idle" processor, which in this case would be useless updates.
This comment was marked helpful 2 times.