Does anyone know when we will get the 11.04 image?

We are working on it, but we do not have an ETA yet.

_________________
Steev Klimaszewski, Genesi USA Inc.
Senior Software Engineer

Just for curiosity's sake, do you plan to include (better) 2d & video drivers with the natty release?

"Better" is too subjective. Better to some people might mean playing videos, or using (proprietary) software like flash; better to me means I can use a terminal.

We always do our best to move forward rather than backward, so when the natty image is released, it will include the latest updates and bugfixes that we have available.

_________________
Steev Klimaszewski, Genesi USA Inc.
Senior Software Engineer

I meant the drivers that provided usable video playback through hardware acceleration that have been talked before on this forum.

When it comes to video drivers, I'd say better is quite objective.

But they're not 'video drivers'. They're graphics drivers. And like what was already stated, there are different kinds of graphics. I'm sure that since this is an ARM platform the graphics driver is proprietary. The devs likely can't do very much with them in terms of optimizations. That said, what they can do is see which version of the driver gives them the best general results. They'll probably go with that one.

If the driver is opensource of course, then that means everything I said doesn't apply in the least.

_________________
\"But I want Wandows!\"

We actually have full control over the graphics stack - IPU (framebuffer), 3D GPU (OpenGLES2), 2D GPU (OpenVG).

The "Natty" Xorg driver is vastly improved technologically, as it does far more to get the best performance out of the chips and uses a different acceleration method which allows us to switch between Z430 (3D) and Z160 (2D) acceleration units "on the fly" (actually it's limited to doing it on driver start but, it is better than it was).

The problem is, while "better", it is entirely subjective as Steve mentioned. Benchmarking it shows some extremely good performance benefits, especially with the Z430 driver, but where the driver operates faster in Xv support, you would never notice (since a 24fps video will play at 24fps regardless of driver) and where on single op benchmarks it is several times faster, it is a rare day indeed that an application other than a benchmark pipelines single operations on single pixmap to see the same performances. So far the benefits are restricted to "it feels like it drags windows faster" and "X-Chat seems not to be so sticky". They are essentially not quantifiable improvements.

Assuming you could look inside the new driver (it's not released yet but it will probably be available as source when it is) and you will notice it does eviction of pixmaps from GPU to system memory, bilinear filtering on scaling, and uses a lot less CPU in the best cases but a lot more in others. It doesn't pipeline GPU accesses as well (as the acceleration API is basically stateless, where the libz160 API used currently has full knowledge of the current GPU context and writes a lot less to the GPU) but the way EXA works plays to our favor in that it doesn't really matter as there is a lot going on behind the scenes that is soaking up resources more than we lose. There is a balance to achieve which we believe we are actually achieving. A lot of work needs to get done though..

~

What is not quite as well controlled is the multimedia stack. It's really not ready to go out in our opinion.

The wait is entirely down to quality control right now. It would be stupid to ship it if it only played half the videos people want to play. We've had some successes, but also some dismal failures (playing an Ogg that came as standard with older Ubuntus, and several of the encodings available from BigBuckBunny.org act strangely).

Considering these are the first things people will test (I am not going to acknowledge the existence of anything but free Creative Commons-licensed media :) if some don't work then it's tantamount to it not working at all.

_________________
Matt Sealey

I'm curious to know how much work you need to put into making this hardware video decoding acceleration work. I mean, I thought this was already taken care of by Freescale?

There's a demo of an i.MX53 Quick Start Board playing 1080p AVC video here, running the default Ubuntu installation that comes with this board.
Haven't Freescale released the source for this Ubuntu build, so you can just use that code?

Yes we could, if we wanted to release Ubuntu 10.04 for the MX51, which is what all of their (released) BSPs are against, so we have to forward port the patches and work out the upstream changes that have happened. So for example, their gstreamer patches are for gstreamer 0.10.28, and natty has gstreamer 0.10.32; and the changes between .28 and .32 are substantial. And then we have to go through testing, QA, and so on and so forth, before we can finally release.

It's not trivial, but if enterprising users want to help out, that's what open source is all about :)

_________________
Steev Klimaszewski, Genesi USA Inc.
Senior Software Engineer

Ah, I see. I hadn't thought of that.
One could wish that these things would be easier, like a standard gstreamer interface attaching to the interface that the BSP (or something on top of the BSP) provides, without patching the software (gstreamer).

What about supporting an interface like VDPAU? Both gstreamer and programs like VLC and mplayer supports it out of the box, so implementing this would give support to a large number of media players as far as I can see. Also, no patching of the programs themselves would be necessary. Is this an option?