James,
this is an optimization to improve lock handoffs and other migratory
sharing. The assumption is that blocks that are modified by one
processor are likely to be modified by another. So, the protocol hands
the data off in exclusive mode.
--Kevin
On Feb 23, 2006, at 4:39 PM, James Wang wrote:
Hi Guys:
I am working on some modification of the
MOESI_SMP_directory-cache.sm file in the protocol folder. I notice
that the
following transition
transition(MM, Fwd_GETS, I) {
ee_sendDataExclusive;
l_popForwardQueue;
}
is kind of weird because I would expect it to go to O or S state
rather
than I sate. Could anyone explain to me why?
Regards
James Wang
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
|