Re: [Gems-users] MOESI_CMP_* protocols

[Dan Gibson <degibson@xxxxxxxx>]

On Fri, May 29, 2009 at 7:04 PM, Edward Lee <edwl202@xxxxxxxxx> wrote:

I have couple of potentially trivial questions. I have been working on MOESI_SMP_directory protocol and I want to start doing some tests on MOESI_CMP protocols as well. Here are my questions:

1- What is the difference between MOESI_CMP_directory and MOESI_CMP_directory_m?

The protocols ending in _m use DDR2-like memory controllers. The non-_m protocols use a simpler delay model for memory latency.
 

2- As I looked at the html files, the title seems to be Token Protocol for MOESI_CMP_directory. So, is this just a typo or if not then, what is the difference between MOESI_CMP_token?

I have no idea.
 

3- And as I checked the MOESI_CMP_* protocol files, and generated html tables I failed to see the "E" state. The CMP protocols seem to be more complex and I couldn't really follow them from the html table so I might be missing something trivial here but I would like to know the reason of not having the exclusive state explicity in these protocols if possible.

Unlike the SMP protocols, the CMP protocols treat the L1s and L2s as separate, inter-communicating state machines. Keep staring at the state transitions, say, for the L1. I find that it helps to look at those not as one unified protocol but as multiple discrete entities with communication between L1/L2 and L2/directory.
 

Regards,

Ed

_______________________________________________
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.

-- 
http://www.cs.wisc.edu/~gibson [esc]:wq!