Neo1973 video (was Re: What would be a realistic but challenging level for Bryce announced trophy money for video playback on the Neo1973? Re: h.264 format is now open?)
laforge at openmoko.org
Mon Feb 5 11:52:05 CET 2007
On Mon, Feb 05, 2007 at 02:36:04AM +0200, Siarhei Siamashka wrote:
> On Saturday 03 February 2007 15:47, you wrote:
> > Also, the 770 has a landscape display. We have a portrait display. The
> > S3C2410 cannot rotate the image, so you would have to rotate every frame
> > in software, too!
> That's a good point. Anyway, rotation can be combined with scaling or color
> format conversion and done in a single pass, so overhead should not be too
> big. One more solution is rotation at video transcode stage as Mikko
> > > Just for some experiment, I compiled mplayer for arm920t (not using
> > > armv5te instructions), and benchmarked it with sdl video output (software
> > > YUV->RGB conversion, generic nonoptimized scaling 320x240 => 640x480) and
> > > libmad mp3 audio decoder.
> > Please note that the LCM we use in the Neo1973 can do hardware scaling,
> > e.g. theoretically you can software-reconfigure the LCM to behave as
> > QVGA 240x320, and then change the s3c2410_fb kernel driver timings
> > accordingly.
> If I understand that correctly, it is not arbitrary scaling but support for
> 240x320 resolution?
yes, you understand correctly.
> > This has not been tested or implemented by us, since we're mainly
> > interested in getting a high-res phone UI working right now :)
> I clearly understand that :) I guess it is one of the reasons, why you
> announced early access to the device for open source developers. I hope
> that some of them would try implementing some video support.
exactly. I just wanted to make sure that this is not on our agenda, at
least with any significant priority.
> As it seems to be not quite trivial to do, this part of work can wait a bit
> until some initial video benchmarks are available (centered nonscaled
> 320x240 video playback vs. software scaled to fullscreen).
Yes, obviously those benchmarks would help. Anyway, the "hardware
scaling" is also interesting for non-video applicaitions such as many
old video games (ScummVM and the like).
As for the benchmarks: Obviously you consume only a quarter of memory
bandwidth when you use 320x240 than 640x480. So native 320x240 playback
should at least theoretically be faster than centered nonscaled 320x240
> I'm just interested in improving video support for ARM based devices,
> that's why I posted to this openmoko mailing list . I'm currently trying to
> integrate a fast scaler for ARM into ffmpeg library (the engine used
> by mplayer, vlc and the other video players for linux ):
good luck! I've done some ffmpeg hacking before, and it was ...
- Harald Welte <laforge at openmoko.org> http://openmoko.org/
Software for the world's first truly open Free Software mobile phone
More information about the community