[mythtv-users] Playback performance [all platforms]
David Abrahams
dave at boost-consulting.com
Mon Jan 9 16:16:23 UTC 2006
Jens Baumeister <jens.baumeister at gmail.com> writes:
> On 1/9/06, David Abrahams <dave at boost-consulting.com> wrote:
>>
>> I notice that the performance of playback seems to be suboptimal. For
>> example, I captured some HDTV video over firewire using MythTV. When
>> I try to play it back using the frontend, it is totally unwatchable.
>>
> [...]
>> However, when I use VLC to play the same file, it looks almost
>> perfect
> [...]
>> And if Darwin-specific playback
>> code turns out to help, doesn't it indicate that the playback code
>> MythTV is using for all other platforms today could be a whole lot
>> more efficient?
>
>>From what I've read, there isn't much PPC or Darwin optimization in
> MythTV - at least not in the releases. I don't know about SVN.
AFAICT, there's no processor-specific optimization in playback at all.
So if performance is bad on PPC/Darwin, it's probably bad (compared to
what's possible) on x86/Linux as well.
> I use 0.18.1 on an old 550 MHz G4 PowerBook with 768 MB RAM as my
> kitchen frontend with nothing else running on it, and it just barely
> handles the task. (E.g. it starts to stutter when mythtvosd scrolls
> something through the frame.) OTOH my master backend ist a Gentoo
> PIII 600 MHz with only 256 MB RAM (I know, I should buy some more...)
> which also doubles as my living room frontend and usually has other
> stuff (server apps, iPod transcoding, commercial flagging, etc.)
> running in the background - yet it feels a lot more responsive than
> the OS X frontend and hardly ever stutters.
Ouch, really? Not good news for the MacOS folks. I wonder what about
the MacOS frontend is sucking so badly? I am running a
frontend/backend combo on a 1.42 GHz Mac Mini (admittedly it could
stand a RAM upgrade from 512M to 1G, but I don't think that's the
issue) and I see a *lot* of beachballs. Maybe I'd be better off
buying a cheap linux box, but I really like the "just-works appliance"
nature of the Mac for this purpose.
Any idea how I can find out what the problem is? I don't imagine
profiling will make a huge difference in such a highly threaded
application.
> I've spent some time setting up setuid permissions and process
> priorities on the Linux box, but I don't think I'll manage to squeeze
> much more performance out of the OS X frontend doing that.
Why not?
> So I' can't vouch for possible x86 optimization, but I'm pretty sure
> the code could be tweaked to run faster on Darwin/PPC.
Can you explain why you are pretty sure of that?
Thanks,
--
Dave Abrahams
Boost Consulting
www.boost-consulting.com
More information about the mythtv-users
mailing list