Re: [Gems-users] A question in LogTM

[郭锐 <timmyguo@xxxxxxxxxxxxxxxx>]

Thanks for your explanation, Mike!

> -----Original Message-----
> From: gems-users-bounces@xxxxxxxxxxx
[mailto:gems-users-bounces@xxxxxxxxxxx]
> On Behalf Of Mike Marty
> Sent: Sunday, January 21, 2007 11:48 PM
> To: 'Gems Users'
> Subject: Re: [Gems-users] A question in LogTM
> 
> > Yes, I also think the cache structure modeled by ruby is quite confusing
> > at
> > first.
> > In addition to a shared controller, the L1 & L2 are exclusive rather
than
> > inclusive.
> > It's quite strange to me. Sorry for my ignorant.
> > One more question, why did you say that CMP don't need fast path?
> 
> The CMP protocols have separate L1 and L2 controllers.  If you disable
fast
> path, the L1 controller sees every hit but there is still a distinction
> between L1 and L2 hits (because the controllers are separate).  For one,
> this allows more flexibility in doing actions on hits (like setting bits).
> This also allows an arbitrary L1 hit latency to be used by setting the
> SEQUENCER_TO_CONTROLLER latency to whatever you want.  And the L1 hits
show
> up in the transition counts in the Ruby stats file.
> 
> 
> 
> _______________________________________________
> 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.