[mythtv-users] OpenGL Support and PiP

[Brian Wood]

On Jan 28, 2006, at 2:00 PM, Jarod Wilson wrote:

> On Saturday 28 January 2006 11:19, Brian Wood wrote:
>> With my previous incarnation of MythTV PiP worked just fine. Due to
>> hardware problems I have had to re-build and things are mostly
>> working now, but PiP is not.
>>
>> The only differences (that I'm aware of) are a newer kernel, newer
>> version of LIRC (shouldn't matter), and the fact that I have compiled
>> MythTV without OpenGL support. This was due to high CPU load when
>> watching live TV. Originally I solved this by commenting out "load
>> glx" in my X config, but this time I took some "advice" and simply
>> compiled without OpenGL. My CPU load is now reasonable even with glx
>> loaded.
>
> Uh, why recompile to disable OpenGL, when you can simply toggle it  
> off in the
> settings?

Ah, that's true of the SVN versions but not the plain-vanilla 0.18.1,  
which I am trying to stay with for several reasons.

>
>> Before I re-compile MythTV with OpenGL support in an effort to
>> resolve this problem, does it make sense that this could have
>> anything to do with it? Itdoesn't make real sense to me but I'm
>> running out of ideas :-)
>>
>> Any other ideas as to what might make PiP mal-function?
>
> nVidia driver version change, possibly.
>
>> Gentoo linux, 2.6.15, amd64, plenty of storage and RAM
>> MythTV 0.18.1
>> IVTV 0.4.2
>> LIRC 0.8.0_pre3
>> PVR-150, PVR-350, HD-3000
>> nVidia 5700 video card
>
> What nVidia driver are you using? The latest one results in pegged  
> cpus on
> playback for some folks, while 7676 works just fine. Though I dunno  
> if 7676
> compiles against 2.6.15 w/o patching...

I'm using 8178, but PiP was working fine with this same nVidia driver  
in my previous installation.

BTW - I've seen people having different results with the "same"  
version of nVidia drivers, and I now realize that I am using the 64- 
bit drivers, 32-bit versions with the same release number might  
behave differently. This might explain some of the inconsistent  
reports about nVidia drivers. We should all report our architectures  
when reporting problems (I'm as guilty as anyone here), as well as  
the kernel version as the nVidia modules compiles against it and its  
sources.

I've also discovered that trying to engage PiP (by typing "V") locks  
up my frontend in that "ESC" does not get me out of it. Killing the  
frontend from an ssh login allows me to see a "waiting for input"  
message repeated many times.

It's just frustrating because I *did* have it working :-)