What you'd probably want to do in that case is modify the ROI hooks to call
ptlcall_switch_to_native() at the end of the ROI. Then, you'll just have to
write script on the guest that will launch the benchmarks one by one (so
each program will call ptlcall_switch_to_sim() at the start and
ptlcall_switch_to_native() at the end).

I don't know how marss handles log files in this case, but if it always
goes to the same log file, you could always just use the ptlcall_log() to
put in a marker each time one program ends and another begins and just
split them afterwards.




On Fri, May 10, 2013 at 12:53 PM, Zoran Jaksic <[email protected]> wrote:

> Hi Paul,
>
> Well I would actually would like to run them one after other with shutting
> down virtual machine between each benchmark run.
>
> is that posible?
>
> what I would like actually is this
>
> 1)boot up virtual machine
> 2)set simulation parameters (set simconfig stats file)
> 3)run benchmark
> 4)kill simulation (shut down virtual machine)
> 5)return to step 1 and pass everything for second benchmark and log file,
>
> What I want to do is actually to make a script that will run 10 different
> benchmarks that would generate me 10 different log files (log file per
> benchmark) with shutting down virtual machine after every benchmark run?
>
> Thanks for everything Paul,
>
> regards
> Zoran
>
>
> On Fri, May 10, 2013 at 6:40 PM, Paul Rosenfeld <[email protected]>wrote:
>
>> Do you want to run them one after the other without shutting off the
>> virtual machine? If so, is the only reason to save yourself bootup time?
>>
>> If you just want to run many different benchmarks (not one after the
>> other without shutting off the VM), I think the standard methodology to run
>> many benchmarks is this:
>>
>> 1. Make a checkpoint for each benchmark you want to run using the
>> create_checkpoint.py script (this will create a checkpoint right at the ROI
>> start marker)
>> 2. Run each of the checkpoints making sure to use the -snapshot flag (so
>> as not to corrupt the disk image). This can be done in parallel or serially
>> (use run_bench.py).
>>
>>
>>
>>
>> On Fri, May 10, 2013 at 8:50 AM, Zoran Jaksic <[email protected]> wrote:
>>
>>> Hello guys,
>>>
>>> Again I am in problem with marss simulation. If you can help... I would
>>> really appreciate that.
>>>
>>> Is it possible to run different benchmarks one after another and later
>>> collect simulation results. What would be the easiest way to solve that.
>>> Basically what I would like to do is:
>>>
>>> 1)set simulation parameters (set simconfig stats file)
>>> 2)run benchmark
>>> 3)kill simulation
>>> 4) return to step 1 for different benchmark (and log file).
>>>
>>> I mean, if I use kill_sim it kills my image OS but I prefer using this
>>> instead of stop_sim. Is it possible to somehow run simulation from host OS
>>> lets say?
>>> Or is it only way to run different benchmarks in the procedure  I
>>> describe by make some script on image OS and use stop_sim between them? Is
>>> there any other solution for this problem?
>>>
>>> I prefer using kill_sim instead stop_sim since I am not completely sure
>>> what happens with the image when I want to start different simulation after
>>> stop_sim (but many times it generally works slower or it crashes with
>>> message that order of using run, stop, and kill is not good (although I
>>> think that I use them regularly run, stop, run...))
>>>
>>>
>>> I currently use your parsec image for my work but lets say that I am
>>> interested in general for this procedure for any kind of image.
>>>
>>> thanks for all help, you constantly provide.
>>>
>>> Best regards
>>> --
>>> Zoran
>>>
>>> _______________________________________________
>>> http://www.marss86.org
>>> Marss86-Devel mailing list
>>> [email protected]
>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>>>
>>>
>>
>
>
> --
> Zoran Jakšić
> ------------------------------------------------------
> Research Assistant
> Computer Architecture Department
> Universitat Politecnica de Catalunya
> -------------------------------------------------------
> Address: Campus Nord UPC C6-E208
> CR. Jordi Girona, 1-3
> 08034 Barcelona, Spain
> -------------------------------------------------------
> Tel:     +34648025808
> e-mail: [email protected]
>            [email protected]
>  <[email protected]>
> -------------------------------------------------------
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to