Re: [Gems-users] An old question about Ruby Cycles and Simics Cycles

[Aoxiang Tang <soartang@xxxxxxxxx>]

thank you Philips. That helps a lot.

On Sat, Jun 18, 2011 at 12:47 AM, Philip Garcia <pcgarcia@xxxxxxxx> wrote:

On Jun 17, 2011, at 4:31 PM, Aoxiang Tang wrote:

Hi Philips,

    Thanks for your answer. Let me rephrase and correct me if I'm wrong. Thank you again.

    The ratio between cpu cycle time (or rather simics cycle time) and ruby's cycle time is SIMICS_RUBY_MILTIPLIER right?  Suppose it's 2. In your case, the executed time should be "ruby's cycles * 2 * 1/950M".

that is correct

    You say that we need to make sure the memory latency parameters are set with respect to the chosen clock speed. Does the clock mean the cpu clock? Usually we use cpu clock to represent the latency. For example, 6 cycles to access L2. But in ruby's default configuration, all the time related parameters are specified in the unit of ruby cycles. For example, still suppose the SIMICS_RUBY_MULTIPLIER is 2. In the case, I should specify 6/2 for the L2 latency, right?

I personally always have multiplier set to 1, but if you are targeting an aggressively clocked system, you might want to set the ruby multiplier to a different value.  If you want a latency to be 6 CPU cycles, and your multiplier is set to 2, than you would set a latency to 3.  However, beware that many of the latencies in Ruby's default configuration don't apply to all CC protocols, and that the total latency can vary quite a bit based on the cache network chosen etc.

Phil

best
Aoxiang

On Fri, Jun 17, 2011 at 5:03 PM, Philip Garcia <pcgarcia@xxxxxxxx> wrote:

GEMS doesn't really run at a set frequency.  If you're just using simics+ruby (no opal) the only associated parameter is the ratio between the cpu cycle time, and ruby's cycle time.  For the CPU cycle time, choose the appropriate clock speed that you're targeting.  In my work, I target embedded devices, and use a 950 MHz cpu.   Just make sure your memory latency parameters are set with respect to your chosen clock speed.

Phil

On Jun 17, 2011, at 2:18 PM, Aoxiang Tang wrote:

Hi,

I'm new to GEMS. I know this questions has been discussed in the email list many times but I'm still confused. Could anybody help me with it? I think lots of experienced people know them clearly. Thanks!

I'm going to calculate the power of the cache. I manage to get the ENERGY/access from cacti and the number of accesses to the L1/L2 cache from Ruby, so the total ENERGY can be attained. Still execution time in the target architecture is needed to calculate the power consumption. How to get it? In the Ruby dump report, there are Ruby cycles and Simics Cycles. The FAQ says that Ruby Cycle is the right value to calculate time but how about the frequency? Should I use the frequency specified in Simics together with Ruby cycles to get the time?

Could anybody give me advice on how to get the simulated time? Great thanks!
(PS: I use Simics+RUBY without Opal to simualte a single core processor)

Aoxiang

_______________________________________________
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.

_______________________________________________
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.