On Sun, 18 Mar 2007 15:16:26 -0400
Nicholas S-A <[EMAIL PROTECTED]> wrote:


> The initial tests are not good. I don't have a G550, just an Nvidia  
> 6800,
> but I disabled OpenGL to test. I first tested with the mplayer that  
> came from
> the Ubuntu repositories. With only 1 core enabled, the performance
> was miserable. With both cores, it can play with some shearing (with  
> scene
> changes, it is quite distracting) except that every once in a while  
> it stops,

This sounds like you were using -vo x11 instead of -vo xv.
Which has a very poor performance (multiple copies within
the x server performed, has to change color format at least
two times etc pp).

Using two cores together with x has the same effect as with
using DMA. The second core performs the copying to the graphics
card while the first is busy decoding.

> switching between two frames (I would say every 20-30 seconds). Audio  
> is fine.

Audio is always fine...unless you have a broken sound card driver :)

> Just for kicks, I turned on OpenGL direct rendering with binary  
> drivers. The video
> was perfectly smooth. 

Cool... so that works. Good. This means that software
decoding is fast enough for 1080p content in MPEG4.

> How do I find the number of dropped frames? I have run
> 'mplayer -nosound -vo null -benchmark Elephants_Dream_HD.avi' with  
> various debugging
> output options (up to all=4) without success. 

Play it normaly, without -benchmark, but with -framedrop.
Then you have it in the status line it's iirc the second
last number. (just as you did below)

But the -benchmark -nosound output would be interesting too.
Just run it a few times and use the best result

> It complains that
> "aspect: Warning: no suitable new res found!"

This comes from the aspect code that tries to scale
the window to fix aspect while still fitting on the screen.
As it only tries to scale up, it cannot get any resolution
that would fit on the screen as your window is already
bigger. You can safely ignore this message.


> I then thought the res might be off and ran "mplayer -framedrop -x  
> 1920 -y 1080 Elephants_Dream_HD.avi".
> 
> At the end it printed:
> 
> "A: 653.8 V: 653.7 A-V: 0.028 ct: -0.037 15691/15691 27% 55% 0.7% 58 0"

Which would mean it would have dropped 58 frames out of 16k.
Ie, it would work without dropping any frames (-framedrop is
a bit too agressive if the cpu is fast enough but not too fast)

> I decided that I should test the SVN version at this point, so I  
> compiled mplayer from svn:
> 
> There is still a lot, but noticeably less shearing. There were  
> absolutely no stalls.
> This is with "./mplayer -framedrop Elephants_Dream_HD.avi" and two  
> cores enabled.
> 
> At the end it displayed this:
> "A: 653.8 V: 653.7 A-V: 0.075 ct:  0.063 15691/15691 39% 58%  1.6%  
> 10931 0"

Now, this is worse. Strange....What -vo did it use?
 
> I just ran a few completely subjective tests:
> 1) with -vo xv and direct rendering, -framedrop was needed to give  
> decent sync of audio
> and video after the first few seconds in the Ubuntu version. The SVN  
> said there were errors
> opening the selected vo device.
> 
> Just let me know if you want me to run more tests - once I get some  
> with good results,
> I can post them to the list.

Thanks. I think this is enough.

I think we can live with software decoding for the moment.
(Given that the player does DMA)

                        Attila Kinali

-- 
Linux ist... wenn man einfache Dinge auch mit einer kryptischen
post-fix Sprache loesen kann
                        -- Daniel Hottinger
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to