Re: [gem5-users] Lower bound on interval size for dumping statistics

[Andreas Hansson]

Hi Shervin,

Of all the things that could be going wrong, I'm tempted to think it's more 
likely in either McPAT or the integration.

I'd suggest you run some simple control experiments (without McPAT) and verify 
that the gem5 stats do indeed make sense when using different dump frequencies. 
If not, then perhaps do some digging into what stats differ and how/why.

Thanks,

Andreas

From: shervin hajiamini 
<hajamini_2...@yahoo.com<mailto:hajamini_2...@yahoo.com>>
Reply-To: gem5 users mailing list 
<gem5-users@gem5.org<mailto:gem5-users@gem5.org>>
Date: Monday, 15 April 2013 09:49
To: "gem5-users@gem5.org<mailto:gem5-users@gem5.org>" 
<gem5-users@gem5.org<mailto:gem5-users@gem5.org>>
Subject: Re: [gem5-users] Lower bound on interval size for dumping statistics

Hi Andreas,

Thanks for the confirmation. When I feed the statistics for small interval size 
(e.g. 1M cycles) to McPAT, I get high power usage (around 3 Watts per core) 
whereas for large interval size(e.g. 10M cycles), the power usage is low 
(around 0.6 Watts per core) which is a normal power usage . I do not know yet 
why McPAT generates high power usage for the small interval size. That's why I 
am wondering whether gem5 generates correct statistics for the small intervals. 
Do you have an idea?

Thanks,
Shervin

Message: 1
Date: Mon, 15 Apr 2013 08:33:45 +0100
From: Andreas Hansson 
<andreas.hans...@arm.com</mc/compose?to=andreas.hans...@arm.com>>
To: gem5 users mailing list 
<gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org>>
Subject: Re: [gem5-users] Lower bound on interval size for dumping
    statistics
Message-ID: 
<cd916ba9.9abc%andreas.hans...@arm.com</mc/compose?to=andreas.hans...@arm.com>>
Content-Type: text/plain; charset="windows-1252"

Hi Shervin,

It should work without any issues, so if you stumble on anything that looks 
suspicious I'd say it's most likely a bug and we should fix it.

Go ahead and give it a spin. Just make sure you've got plenty disk space.

Andreas

From: shervin hajiamini 
<hajamini_2...@yahoo.com</mc/compose?to=hajamini_2...@yahoo.com><mailto:hajamini_2...@yahoo.com</mc/compose?to=hajamini_2...@yahoo.com>>>
Reply-To: gem5 users mailing list 
<gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org><mailto:gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org>>>
Date: Sunday, 14 April 2013 23:36
To: 
"gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org><mailto:gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org>>"
 
<gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org><mailto:gem5-users@gem5.org</mc/compose?to=gem5-users@gem5.org>>>
Subject: [gem5-users] Lower bound on interval size for dumping statistics

Hi,

Using statsDump(ticks) function in gem5, I dump statistics of a benchmark 
periodically and I am wondering whether there is a lower bound on the 
interval(period) size. I mean whether I can dump the statistics with an 
arbitrary size for the interval or there is a threshold for the interval size? 
Does gem5 dump stats correctly for intervals of small size?

Thanks,
Shervin


-- IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users