[MPlayer-dev-eng] patch for quadbuffered stereo using "-vo gl2:stereo"

Alexander Strange astrange at ithinksw.com
Thu Nov 29 19:54:56 CET 2007


On Nov 29, 2007, at 12:35 PM, Stuart Levy wrote:

> On Thu, Nov 29, 2007 at 06:22:53PM +0100, Guillaume Poirier wrote:
>> Hello,
>>
>> Stuart Levy wrote:
>>
>>> But across the board, the texture loading performance is a good  
>>> deal faster for
>>> medium-sized textures than large ones, on both nVidia Linux  
>>> drivers and for
>>> Apple's driver.  And to be fair, the improvement isn't always  
>>> 30-40%;
>>> it's often more like 15%.  Still it's significant.
>>
>> Mmmm... does that mean that you managed get a working -vo gl / -vo  
>> gl2
>> on OSX?
>
> No, I didn't try that.  I was trying to make an uncompressed  
> movieplayer
> and made a bunch of tests to see where the bottlenecks are.
> So I have some other programs running on OS X but didn't attempt
> to build mplayer there!  Just mentioned Apple to point out that,
> if texture performance is a driver issue, it's an issue with
> drivers on at least two unrelated platforms.

Aside from strange things like not exposing direct YUV 4:2:0, Apple's
OpenGL is designed for nothing but uploading big textures really fast.
That's what the window system is based on, after all.

You might want to check some of the extensions they support, like
GL_APPLE_client_storage, GL_APPLE_texture_range(?), and especially
GL_ARB_non_power_of_two, although it's probably not any faster than
the older EXT_rectangle. I'm not sure what extensions Linux usually  
provides,
but maybe there's something similar?

By the way, vo_macosx probably uses these implicitly by calling  
CoreVideo.

(other things to note here are that OS X's Nvidia drivers are much  
slower than ATi,
and that using immediate mode via glBegin/glFinish is deprecated)



More information about the MPlayer-dev-eng mailing list