Linux PVR

[Jason Clinton]

Brian Densmore wrote:

>>I think we've heard from some people who have done 
>>full-screen, full quality 
>>capture and playback - with processors available that are 
>>three times as fast, 
>>is a single-CPU unit enough these days?  How much CPU do you need for 
>>full-speed, full-screen video processing? 
>>    
>>
>
>Doesn't DVD become grainy at full screen? I mean how many pixels are in
>a normal DVD picture? I play mine at about half screen, and it is very 
>watchable. In fact it's better quality than on a TV and I see no delay.
>I run on a 750MHz Duron with a 32 MB video card. Once I got the settings
>figured out,
>with some help from the list, it was very usable.
>  
>

I'll answer both questions at the same time. Regarding full screen 
video, there are a number of issues that come in to play.

   1. does the card and driver support accelerated video scaling?
   2. what about the IDE bus?
   3. implementation of decss

Generally speaking, the CPU load is greatly reduced if you have:

   1. A DVD drive that supports DMA access.
   2. A video card and driver that support the Render and Xv V4L
      extentions -- the i810, 3DFX, NVidia and ATI drivers all do.
   3. An implementation of libdecss that is fast. (rarely a problem)
   4. A player applications that correctly understands all of the above
      and is correctly attempting to use them. (Xine, Mplayer, Ogle, etc.)

Also, since the scalling from 740x568 to full screen such as 1024x768 is 
handled in hardware on the video card, there's lots of room there to do 
what's called "ansiotropic texture filtering" which is difficult to 
explain but basically means that the grainy stuff is smoothed out. Some 
argue that it just makes it worse but I disagree. I would google search 
for ansiotropic.