Optimization idea: soft XvPutImage

Adam Richter adam_richter2004 at yahoo.com
Wed Sep 17 10:43:55 PDT 2008


On Wed, 9/17/08, Roland Scheidegger <sroland at tungstengraphics.com> wrote:
> On 17.09.2008 14:22, Adam Richter wrote:
> > 	The memory bus utilization would also be reduced (but
> never
> > more than that factor 3) as the ratio of the size of
> the unstretched
> > video to stretched video increases, such as when
> playing a 720x480 video
> > on a newer 2560x1600 display.
> It works the other way around too however, think Full HD
> playback on a
> (mobile) device with a (say) 800x480 screen (not that
> I'd say this is
> very common, but it's a case which could definitely
> happen). Of course,
> "normal" Xv needs to transfer the full resolution
> image too (albeit only
> as packed or more common planar yuv which is a bit
> smaller).

a) Even in the case of shrinking the video, soft XvPutImage should save memory bandwidth when running locally, although not as much (since it should be just eliminating two memory transfers of the stretched or shrunken RGB image).  For the unusual case of video player running across the network, any use of XvPutImage  (hardware or software) should use less network bandwidth than XPutImage when stretching the video and more network bandwidth when the video is being shrunken.

b) The availability of soft XvPutImage or Xv in general does not prevent a video player from using any of the other display mechanisms it has available.

In any case, thanks for taking the time to consider my suggestion.

Adam Richter



      



More information about the xorg mailing list