If I recall correctly, the IGM state means that the first-level directory has issued a request to the second-level directory. But since the second-level request has not yet been ordered at the second-level directory, it is still possible to receive invalidation request ( an L2 S state transitions to IGM when the first-level directory sees a GETX). The first-level directory just acknowledges the invalidate, but stays in IGM/IGS because eventually its request will be handled. If you set IGM/IGS to I, then the protocol will not work.
--Mike On Thu, Jul 10, 2008 at 11:19 PM, wei wang < apollo.mobility@xxxxxxxxx> wrote:
i spotted the following state transitions related to handling an INV message in MOESI_CMP_directory
transition({IGM, IGS}, Inv) transition({I,NP}, Inv) transition(S, Inv, I)
all the transitions will send an ACK message to global directory, which is simply ignored in the directory. While
I -> I, NP -> NP and S->I set the data in one of the stable state, IGM -> IGM and IGS ->IGS will leave data in a transient states. Is this right? should we set IGM and IGS to I as well? thanks.
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
|