Re: [Gems-users] question on time in eventqueue

["hongxia sun" <sunhx.seraph@xxxxxxxxx>]

Hi, Mike,

 

Sorry for sending the wrong mail.

 

If the protocol allow more thant one transitions taken in the same cycle, could the second transtion see the new state of the cacheline changed by the first transition?

If so, the protocol need to take many special conditions into account.

 

If a new event putting into the EventQueue with zero delay and zero latency, does it means it will happen in the same cycle which generate it ?

 

Hongxia sun

 

 

2007/1/26, Mike Marty <mikem@xxxxxxxxxxx>:

> SLICC defines a state machine -- no state can change more than once per cycle, so all future transitions,
> output changes, etc, must have at least delay 1.
>

Thats not really true.  A state can change more than once in a cycle.
For example, a Directory controller could receive two GETS messages in the
same cycle that are both handled (and both potentially state changing..for
example the first GETS could transition to S where the second might
transition to SS indicating multiple outstanding GETS requests).

The EventQueue is a priority heap.  Adding a new entry that is the minimum
could be dicey...for example if could change the current root of the tree.
We've always enforced that adding something to the EventQueue must have a
wakeup at least 1 cycle in the future.  If you try modifying this
behavior, I'd be interested to hear your experiences.
_______________________________________________
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.