Re: [Gems-users] Number of instruction executed in GEMS for CMPs

[Dan Gibson <degibson@xxxxxxxx>]

Check to see how many instructions each processor executes. I wonder if perhaps c X runs, e.g., Proc0 for X instructions?

Also try cc instead of c -- I forget which one means 'instructions' and which one means 'Simics cycles'.

In general, we don't use 'c X' directly to control the simulation. Instead, we insert magic breakpoints to stop simulation when the correct amount of work has been done. That helps us to get rid of artifacts from spin locks, etc., as well.

Regards,
Dan

On Fri, Apr 30, 2010 at 12:03 AM, Shoaib Altaf <shoaibbinalt@xxxxxxxx> wrote:

List,

 

I was comparing some results between a single processor and a CMP (say 4 processors), and the criteria depends on the number of cycles. I am running for 350 M instructions like this for both cases

c 350000000

 

Now on the output stat file I got these results.....

 

1 Processor:

 

instructions executed =350000000 

Ruby cycles= 1435639304

 

 

4 Processors:

 

instructions executed =1449213046 (roughly 1500 M)

Ruby cycles = 1861225208

 

 

So I have these questions :

 

1) I was expecting the instructions executed to be the same why they are different?? in the present scenario the comparison will  not matter....so how should I make sure that the CMP will also run for 350 M instructions

 

2) I was also expecting that the ruby cycles will also reduce in case of 4 processors, but they have increased since more instructions are executed ..Why??

 

3) My protocol is MSI_MOSI_CMP  and workload is OLTP.

 

Shoaib

_______________________________________________
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!