[mythtv-users] Long delays with remote occasionally

[Tom Dexter]

On Thu, Jul 23, 2009 at 4:57 PM, Richard Shaw<hobbes1069 at gmail.com> wrote:
> [SNIP]
>> As I mentioned here:
>>
>> http://www.gossamer-threads.com/lists/mythtv/users/386493#386493
>>
>> ...I'm not convinced that's the case for everyone.  I had this issue
>> big time, and as I mention several times in the original thread, it
>> started like flicking a switch when I upgraded to xorg-server-1.5.3-r5
>> under gentoo (from 1.3.x I believe), and for others it seems to have
>> started after upgrading Fedora to a version with that as well.  for
>> me, it seems as though either subsequent kernel upgrades or maybe the
>> upgrade from Gentoo xorg-server-1.5.3-r5 to xorg-server-1.5.3-r6 has
>> fixed the issue, and I've been running perfectly without Josh's patch
>> for over a month now.
>>
>> It's a real tough one to nail down, but the bug addressed in Josh's
>> patch was there before I had the problem, and without the patch is
>> there now and simply isn't causing the issue is my case.
>>
>> Tom
>
> I agree, my problems seemed to begin with the Xorg update. To date,
> the only solution I found since I don't want to patch anything, is to
> use the VDPAU backport of myth from atrpms. I'm not sure if this works
> because it bypasses ffmpeg or it it works because it offloads the
> processor enough that it fumbles through the errors fast enough it
> doesn't cause input delays anymore. If I don't use VDPAU I get 100%
> load on one processor and even though the other core is not fully
> utilized, I still get input delays.
>
> Richard

Someone can correct me if I'm wrong, but I actually don't think
running that is by-passing that part of ffmpeg.  The patch is in
libavformat...the ffmpeg code that handles file formats and
demuxing..._not_ the actual codec decoding.  I think you're probably
correct that it's related to offloading stuff that otherwise might be
handles by the CPU.

I tried to look into the patches Gentoo added between
xorg-server-1.5.3-r5 and xorg-server-1.5.3-r6, but there were so many
I couldn't tell if any might be related to this, but yea...I'm fairly
convinced that the core of the issue was with xorg.

Tom