FFmpeg Gains VDPAU MPEG-4 ASP Acceleration

[phoronix]

Phoronix: FFmpeg Gains VDPAU MPEG-4 ASP Acceleration

What we were in the process of writing about when we discovered MPlayer's support for most Blu-ray and HD-DVD codecs was that there is now support for MPEG-4 ASP decoding with VDPAU (NVIDIA's Video Decode and Presentation API for Unix) in the mainline FFmpeg tree. Support for acceleration MPEG-4 ASP was introduced in NVIDIA's most recent revision to their PureVideo technology and now the software support is properly in place. The current NVIDIA graphics cards that support MPEG-4 ASP decoding are the GeForce G210, G210M, GT 220 (we previously provided a NVIDIA GeForce GT 220 review under Linux), GT 230M, GT 240M, GT 240, GTS 250M, and GTS 260M...

http://www.phoronix.com/vr.php?view=NzcwMg

[sylware]

code? Hardware Programming Manual??

Still no code neither programming manual.
Better push radeon open source support.

[Kano]

I don't get your point. ffmpeg (and mplayer) got patches directly by Nvidia for vdpau support in first place. This was just an addon with an extra codec. You don't need to know how the chip works to use it, also other vendors are free to implement vdpau too. s3 already did - just for a chip which is very rarely used - maybe they want to sell it to oems like ati

If you own a supported card already feel free to try my script on Debian/Ubuntu:

http://kanotix.com/files/fix/mplayer-svn-snapshot.txt

If you prefer VAAPI you have to use this script:

http://kanotix.com/files/fix/mplayer-vaapi-latest.txt

As the mplayer VAAPI patch is not yet agaist that new ffmpeg/mplayer code - svn up will fail then. The -1 script just had an extra "svn up" command.

[sylware]

freedom to computer...

I don't get your point. ffmpeg (and mplayer) got patches directly by Nvidia for vdpau support in first place. This was just an addon with an extra codec. You don't need to know how the chip works to use it...

I'm a GPL fan, my goal is to have optimal GPL code installed by default on systems.
Kernel blobs are irrelevant for people like me. In order to start coding an optimal GPL driver for Linux, we need the hardware programming manual.
The current target is to have an optimal BSD driver for radeon chips in order to make the professionnal market switch on radeon chips intead of nvidia chips. Indeed according to the nvidia driver developer interviewed here, the professionnal market share would be made mainly of GNU/Linux systems.
This is a naughty way to try to obtain hardware programming specifications. There is also a project of making distribution of hardware illegal without a complete hardware programming manual.

[Kano]

Then write a vdpau backend for radeon or intel. Nobody holds you back!

[sylware]

freedom to computers...

Then write a vdpau backend for radeon or intel. Nobody holds you back!

BTW, how does compare VAAPI to VDPAU? And your are almost right, I have the plan to write a *GPL* radeon driver (currently those drivers are dangerously BSD). But I have something holding me back: not enough available brain time.

[droidhacker]

There is also a project of making distribution of hardware illegal without a complete hardware programming manual.

Link? I'd like to voice my support for such initiatives.

[gururise]

Why is it that Nvidia's VDPAU patches made it into mainline ffmpeg/mplayer and the VA-API patches have not? Is somebody trying to push them into mainline? Are they of high enough quality (w/good enough documentation) to make it into the mainline ffmpeg?

[dashcloud]

VAAPI patches are in, but as far as I know, they only cover Poulsbo chips right now. There's patches in the works to allow reading the decoded data off the card (so you can do hardware-accelerated video editing), but no other VAAPI patches that I am aware of.

As for why the VDPAU patches went in: the MPEG-4 ASP patches were quite simple compared to the original VDPAU or VAAPI patches, so very few changes were needed to it before it was judged acceptable.
In general, patches that add new features or change large sections of code take many revisions before they meet the exacting standards of the dev team. If the person behind the patch isn't motivated and driven to get the patch into the main tree, it could very well get left behind.