Re: [linuxdc-dev] RedBoot CD booting and MPEG playback

[David W. <dav...@ia...>]

> I think to get that sort of throughput you'd certainly need accelerated 2-D 
> video or else be prepared to accept frames being dropped. I and others have 
> compiled mplayer but there are two eseential problems:
> 
> 1. Sound driver design (needs dma) eats up processor cycles for high sample 
> rate sounds meaning that the pressure is really on the video

Does the DMA 'steal' cycles from the CPU?  Normally, isn't DMA supposed to 
help use less processor time?  This sounds backwards.

> 2. Alternative sound driver design (not with dma, but less processor hogging) 
> gives better video performance, but sound clicks and repeats - an accelerated 
> 2d video driver might fix that.

Ick.

Are there sufficient spec's on the video to write a DRI driver or even a good
accelerated 2D driver?  I've been following the DRI development for about a
year, now, and they are always busy with things, but they're also very helpful
to new people.  So, getting questions answered is easy.  They also have weekly
meetings on Monday afternoons at 2100UTC (5:00EDT) on irc.freenode.net #dri-devel.
Anyone interested in developing DRI or related software is invited to participate
and all are invited to listen in. :)

The big thing for video playback would be what colorspaces and scaling ability
the video chip has.  If it has a nicely scaled YUV colorspace ability, video
playback gets a lot easier.  The DC has pretty good floating point, so that
should help.

Cheers,
David