On Wed, Mar 17, 2010 at 2:49 PM, <ubaid001@xxxxxxx> wrote:
1. each link has a width of 32 bits.
2. yes, the utilization numbers are different for different links.
3. We are using Simics + Ruby. Can you explain why the utilization is low in
this case ?
You are using an in-order processor model. Each processor has at most one request outstanding. I would be worried if utilization wasn't very low.
4. the topology is mesh.
Suhail
Mar 15 2010, Dan Gibson wrote:
1. How much bandwidth have you configured for your links?
2. Do some links show more utilization than others?
3. Are you using Opal, or just Simics+Ruby. I would expect very low
utilizations with the latter combination only.
4. What topology are you using?
Regards,
Dan
On Mon, Mar 15, 2010 at 3:14 PM, <ubaid001@xxxxxxx> wrote:
Hi,
I am running some traffic simulations in "ruby". I found that the link
utilization numbers (in flits/cycle) reported by ruby were less than 5% for
the seven SPEC OMP2001 benchmarks (applu, apsi, art, equake, mgrid, swim,
wupwise).
I think a link utilization number around 20% to 30% should be reasonable.
These benchmarks were simulated on a CMP system configured with 4
processors and 4 shared L2 banks. In ruby, the "Garnet fixed 5-stage
pipeline"
interconnection network model was used. Besides, the parameter
"OPAL_RUBY_MULTIPLIER" was set to be 1.
Suhail
_______________________________________________
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.
_______________________________________________
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!
|
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}