Re: [Gems-users] Memory Request Latency vs. CPI

["Carole-Jean Wu" <carolewu@xxxxxxxxxxxxx>]

Hello Dan,

From my understanding of issueRequest, only cache misses go there right? 
Do you have suggestions how to tackle this problem? Why would my added latency be canceled out by Ruby's event queue? I tried to increase the latency to e.g. 40000 something big, and I will start seeing some variation in CPI. 

Thanks!
Carole

On Thu, Apr 10, 2008 at 9:39 AM, Dan Gibson <degibson@xxxxxxxx> wrote:

My first guess is that your latency is cancelled by Ruby's periodic processing of its event queue. E.g. if you stall a request by 10 cycles in the sequencer, then Ruby's event queue runs (to simulate the cache hierarchy) 2 cycles later, decides the request was a hit, and unstalls the processor, with a net stall time of just 2 cycles.

Regards,
Dan

Carole-Jean Wu wrote:

Hello GEMS-users,

I am running into a strange situation when simulating a multiprogrammed environment. To make the long story short,

in Sequencer.C's issueRequest, I am adding additional latency to "latency" to mimic e.g. a cache hit, latency+=10 and a cache miss, latency+=100. However, when running two applications with great differences on miss ratio, I've find the same CPI in the ruby0.dumpstat file.

Can someone explain to me what is going wrong? Am I inserting this additional latency in the wrong place?


Any help is greatly appreciated!

Thanks a lot,
Carole

_______________________________________________
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!

_______________________________________________
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.