From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============5705992003754973792==" MIME-Version: 1.0 From: Arjan van de Ven Subject: Re: [Powertop] Simple Question Date: Mon, 07 Jul 2014 18:24:58 -0700 Message-ID: <53BB486A.5060804@linux.intel.com> In-Reply-To: COL129-W476C4ACB661CB7F869CA41C60D0@phx.gbl To: powertop@lists.01.org List-ID: --===============5705992003754973792== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On 7/7/2014 4:53 PM, Christopher Mansour wrote: > > Hello all, > > I was wondering if I can use the PowerTop tool in order to observe the po= wer consumption of a thread. I am doing some experiments which would run on= 7 threads at the same > time, I need to know the power consumption of each thread. Is this feasib= le using the power top tool and how. I would appreciate your help and feedb= ack. Powertop, with its internal power model properly trained on the hardware in= question, will attribute power consumption to individual processes. This is not an exact science; the machine learning model needs a whole bunc= h of samples to learn, and the power attribution logic is slightly rudiment= ary, but it's also not unreasonable at the same time. For example a thread gets accounted for the (modelled) CPU power proportion= al to the time it consumes, but also for the power of the audio chip if its= playing audio and if you have the right kernel patches, even for the IO it does. --===============5705992003754973792==--