Re: [Gems-users] Problem running a workload

[Dan Gibson <degibson@xxxxxxxx>]

I think what you are asking is:
"How do I know I am simulating long enough to cover the entire run of the workload, but not so long that I'm simulating well after its completion?"

If you are NOT asking that question, re-word your question so I understand it.

Here is the answer to the question above. It has two parts.
1. It is a common practice to not simulate workloads to their completion -- it simply takes too long.
2. To ensure you do not over-simulate, we recommend you insert magic breakpoints after your workloads, which will stop the simulation automatically if reached.

You will find plenty of information on magic breakpoints if you search this list. (e.g., from http://www.cs.wisc.edu/gems/doc.html)

Regards,
Dan

On Wed, Apr 7, 2010 at 1:36 AM, Mahmood Naderan <mahmood.nt@xxxxxxxxx> wrote:

Although the previous question has not been answered yet, one more important question is now raised.
Consider that running the barnes workload take 2 second on simics as warm-up. Since I don't know how many cycles it took, it is ambiguous for me how to determine the number of cycles in "opal0.sim-step N". Two possible cases are:
1- if (N < number of actual cycles in 2 seconds run), then simulation result is invalid because the rest of warm-up cycles won't be simulated with Opal.
2- if ( N > number of a actual cycles in 2 second run), then simulation will be valid for M cycles where M is the true number of cycles in 2 seconds run, and the remaining N-M cycles will be devoted to OS background instruction. But the result is for N cycles which means a summation of 2 seconds run and remaining OS background instruction.

Am I right?

// Naderan *Mahmood;

On Tue, Mar 30, 2010 at 1:33 PM, Mahmood Naderan <mahmood.nt@xxxxxxxxx> wrote:

Finay I run barnes workload. One question still present: what is the
difference between cold and warm input. How can one determine a
boundary for cold and warm input?

On 3/29/10, Dan Gibson <degibson@xxxxxxxx> wrote:
> On Mon, Mar 29, 2010 at 9:29 AM, Mahmood Naderan
> <mahmood.nt@xxxxxxxxx>wrote:
>
>> I did that and then some lines inserted in the file. Then I made
>> another checkpoint but every time I load from the new checkpoint, I
>> saw a message that is:
>>
>>
>> /***************************************************************************\
>>  > Physical Memory object cannot be found. If you are NOT compiling Ruby
>> and <
>>  > you see this message, something is wrong.
>>   <
>>  > This message is part of the normal compilation process.
>>   <
>>
>>
>> \***************************************************************************/
>>
>> I don't know what is he saying. Either it is "normal" or "wrong"?
>>
>
> It is wrong. That message comes from Ruby. You took a checkpoint after
> loading Ruby, therefore, Ruby is part of the .check file. This is bad,
> because ruby is then re-loaded out of order with respect to the physical
> memory object -- I.e., Ruby can't find the physical memory, because it
> doesn't exist at ruby load-time.
>
> Solution: Manually edit your .check file and remove all mention of Ruby.
> Make a backup before you begin.
>
>
>>
>> I have to say, running a workload is a very complicated task for
>> beginners, with lots of .py file that I don't understand what are the
>> steps to run them in order.
>>
>
> Agreed -- it ain't easy. It turns out, running entire operating systems in
> simulation is tricky.
>
>
>>
>> Before reading that mailing list thread, I did what is stated in the
>> "Configure Simics for Ruby/Opal" section of the Ubuntu PDF guide.
>> There are some commands that shows the order of laoding the Ruby and
>> Opal. One more question is, by running the command "opal0.sim-step
>> 1000000" which workload is actually running because in the prior steps
>> no workload is selected!
>>
>
> If you don't start a workload on the target os, then you just run whatever
> background processes the OS wants to schedule (and/or the idle loop).
>
> Regards,
> Dan
>
>
>>
>>
>> On 3/29/10, Dan Gibson <degibson@xxxxxxxx> wrote:
>> > It looks like you never actually ran the simulation, therefore Ruby's
>> caches
>> > are empty.
>> >
>> > Try inserting this:
>> > simics> c 100
>> > After ruby0.init.
>> >
>> > Regards,
>> > Dan
>> >
>> > On Mon, Mar 29, 2010 at 4:31 AM, Mahmood Naderan
>> > <mahmood.nt@xxxxxxxxx>wrote:
>> >
>> >> I am trying to run the barnes workload based on a disscussion at
>> >> https://lists.cs.wisc.edu/archive/gems-users/2006-March/msg00143.shtml.
>> >> I ran the ./barnes < input-warm in the taret machine and after the
>> >> run, I made a checkpoint. Then I exited and loaded the checkpoint. Now
>> >> when I run "ruby0.savecache filename", it says that 0 lines written
>> >> and the file actually has zero size.
>> >>
>> >> mahmood@magma:sarek$ ./simics -stall -c
>> >> ../../../checkpoints/barnes-warm.check
>> >> simics> istc-disable
>> >> Turning I-STC off and flushing old data
>> >> simics> dstc-disable
>> >> Turning D-STC off and flushing old data
>> >> simics> cpu-switch-time 1
>> >> The switch time will change to 1 cycles (for CPU-0) once all
>> >> processors have synchronized.
>> >> simics> load-module ruby
>> >> successful installation of the ruby timing model.
>> >> simics> ruby0.init
>> >> Ruby Timing Mode
>> >> Creating event queue...
>> >> Creating event queue done
>> >> Creating system...
>> >>  Processors: 1
>> >> Creating system done
>> >> Ruby initialization complete
>> >> simics> ruby0.save-caches barnes.cache.gz
>> >> Writing cache contents to 'barnes.cache.gz'...done. (0 cache lines
>> >> written)
>> >> simics>
>> >>
>> >> Where did I make a mistake?
>> >>
>> >> --
>> >> // Naderan *Mahmood;
>> >> _______________________________________________
>> >> 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

>> > <http://www.cs.wisc.edu/%7Egibson>[esc]:wq!

>> >
>>
>>
>> --
>> // Naderan *Mahmood;
>> _______________________________________________
>> 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!
>

--
// Naderan *Mahmood;

_______________________________________________
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!