From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============7272875308006166632==" MIME-Version: 1.0 From: Rajagopal Venkat Subject: Re: [Powertop] [PATCH] Fix timer and work perf events timestamp tracing Date: Wed, 05 Sep 2012 22:49:32 +0530 Message-ID: In-Reply-To: 5047872C.1090608@linux.intel.com To: powertop@lists.01.org List-ID: --===============7272875308006166632== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On 5 September 2012 22:39, Arjan van de Ven wrote: > On 9/5/2012 9:56 AM, Rajagopal Venkat wrote: >>> measure1: >>> ev3.start >>> ev1.end <<<<< >> >> evX.end <<<<< >> These events are causing numbers to go wrong. > > but out of a 20 second window.. this is a tiny tiny window... > if you see 100.1% I'd buy this reasoning. > but you're seeing much more than that. How about generating a report for 1sec duration? Since timestamp itself is added to accumulated_runtime, the usage percentage is really dependent on event end timestamp value. > > >>> >>> if so, then we're loosing events, which is no good. reporting less than= 100% >>> is ok, but reporting less than real is not. >> >> I did thought of it. Yes, agree that, we are loosing events for which >> start timestamp > > we can't lose those! > those are the events that give us the initial CPU frequency in the window= etc.... > Yes agree. I will submit the next version patch considering those events end timestamp relative to first_stamp(src/process/do_process.cpp). -- = Regards, Rajagopal --===============7272875308006166632==--