Re: [Gems-users] Different processor frequencies

[Philip Garcia <pcgarcia@xxxxxxxx>]

I'm pretty sure this can be done by modifying the simulate loop in  
opal/system/system.C.  You can modify which sequencers are "advanced"  
on a given cycle.  I am not 100% sure what would happen if you did  
this though, as there might be some unexpected consequences.  In  
particular, many things in opal get thrown onto wait queues.  However,  
if you change how often you advance each processor, the wait states  
might end up becoming invalid.  For instance, telling a pipeline to  
stall for 2 cycles might end up stalling only one cycle if it's  
stalling "real" cycles, and not that individual processor's cycles.   
Your best bet might be to modify these loops, and see the impact on  
actual performance on the two cores.  Worse case scenario, you cause  
gems to crash, and best case is that the two processors behave as  
expected.  Of course, delays issued to ruby will still be counted in  
"ruby" cycles separately from the processor pipeline.
Phil
On Oct 28, 2008, at 1:39 PM, Niket Agarwal wrote:

> Hi all,
>
> 	I am trying to model DVFS inside GEMS. The way I was thinking of
> doing this was by changing the Opal and Simics interface to fetch
> instructions at different rates. My understanding of the interface is
> that Opal simulates a multiprocessor system by fetching one
> instruction at a time, in a round-robin manner, from the Simics
> interface from the processors that have active instructions. I wish to
> change this interface.
>
> For e.g. if a processor is running at half the frequency of the other,
> in 2 iterations I will fetch instructions twice from the faster
> processor and once from the slower processor.
>
> Can someone let me know if that sounds reasonable?
>
> Cheers,
> Niket
> _______________________________________________
> 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.