Yes, you can count off-chip accesses like you suggest.
--Mike
pana wrote:
Hello all,
I'm using simics + ruby to simulate a CMP system with: 8 CPUs, 16MB
shared L2 cache, MOESI_CMP_NUCA protocol.
I need to count the off-chip accesses from my L2 cache to RAM memory,
(memory controller is called Directory)
I found in the Slicc file "MOESI_CMP_NUCA-dir.sm <http://dir.sm>" the
TRANSITIONs which perform ACTIONs with "latency = MEMORY_LATENCY" and
I have counted the their total number showed in ruby stats file.
I found a very low number of off-chip accesses in comparison with the
number of misses.
actions:
- action(d_sendDataWithAllTokens, "d", desc="Send data and tokens to
requestor")
- action(dd_sendDataWithAllTokensToStarver, "\d", desc="Send data and
tokens to starver")
transactions:
- transition(M, GETX, NO)
- transition(M, GETS, NO)
- transition(M, Lockdown, L)
- transition(O, GETX, NO)
- transition(O, GETS, NO)
Is this procedure correct to calculate off-chip accesses?
francesco
------------------------------------------------------------------------
_______________________________________________
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.
|