[mythtv-users] mythfrontend vs mplayer cpu usage

[Thomas Fjellstrom]

On Monday 10 November 2008, Al McIntosh wrote:
> On Mon, Nov 10, 2008 at 8:22 PM, Johnny Russ <jruss at mit.edu> wrote:
> > On Mon, Nov 10, 2008 at 8:12 PM, Michael T. Dean 
<mtdean at thirdcontact.com>wrote:
> >> On 11/10/2008 08:03 PM, Al McIntosh wrote:
> >> > The card does not support XVMC, so we can rule that out.
> >> >
> >> > I have the following in my frontend output:
> >> >
> >> > 2008-11-10 19:15:58.470 TV: Changing from None to WatchingPreRecorded
> >> > Initialize Yadif Deinterlacer. In-Pixformat = 1 Out-Pixformat=1
> >> > yadifdeint: size changed from 0 x 0 -> 720 x 480
> >>
> >> Yadif = chew up CPU
> >>
> >> Set your playback profile group to the default configuration of Slim and
> >> you'll get a more realistic comparison (although chances are MPlayer is
> >> not doing any deinterlacing at all, so really to compare, you'd have to
> >> enable a similar-quality deinterlacer).
> >
> > I totally agree if you are only interested in making a comparison with
> > mplayer. However, since your CPU can handle it (you indicated 50%
> > utilization), you may find it is worth the extra CPU cycles to use one of
> > the better deinterlacers. Whether it is worth it or not would depend on
> > your setup and your personal preferences.
>
> I was really just curious as to what the difference was. Deinterlacing did
> not cross my mind to be the culprit.
>
> Thanks this has been helpful.

Just in case, I've found that newer (.26/27) kernels seem to have a bug that 
causes some processes to have their usage calculated wrong.

I managed to see the global cpu usage hover around 20%, while X and 
mythfrontend would total 60% or more. Going back to .25 fixed it for me.

-- 
Thomas Fjellstrom
tfjellstrom at shaw.ca