[mythtv-users] Mythfrontend idle cpu consumption help

Michael T. Dean mtdean at thirdcontact.com
Mon Feb 1 18:41:44 UTC 2010

On 02/01/2010 08:37 AM, jansenj wrote:
>> In any case, this has been pretty much put to rest by
>> some recent checkins by danielk.
> Yep, I started the original thread, and if you have this issue, you have 3
> options as far as I can tell:
> 1. Move to the trunk, as it has now been fixed in the trunk.
> 2. Apply the patch identified in this thread which replaces Qtimer (which,
> the use of it was the root cause).

Actually, the root cause is a "bug" in Qt that's supposedly fixed in Qt
4.6.x.  http://bugreports.qt.nokia.com/browse/QTBUG-7618 , specifically

> 3. Live with it, and maybe change your frontend isn't setup as a realtime
> thread so other threads can use the "busy wait" time.

4. Upgrade to Qt 4.6.0 and see if it's fixed.  If so, enjoy, *and tell
people on the list*.

I would gladly test with Qt 4.6.0, but since I don't experience the
issue even with Qt 4.4.x or Qt 4.6.x (i.e. it's an issue that's only
triggered on some systems--either with newer libc/kernel versions than
I'm using or when using eglibc rather than glibc), I can /not/ determine
whether it's fixed with the newer Qt.

I would very much appreciate some user who's affected by the issue
testing 0.22-fixes (or trunk with the changes reverted--feel free to ask
me for a patch) with Qt 4.6.0 to see if it fixes the issue. 
Unfortunately, the Qt devs didn't say that the fix is in 4.6.0, so it
may require 4.6.1 (which is not yet released), but regardless, we need
data points.

Note, also, that the new code on my system (which was not affected by
the 7%/14% CPU usage issue) results in mythfrontend always consuming 10%
to 50% CPU usage on an Athlon X2 5000+ running in Performance (full
2.6GHz) mode--thus the reason I already have a patch to revert the changes.


More information about the mythtv-users mailing list