What Do You Want From Linux GPU Drivers In 2010?

[lmurray]

Something realistic taking into account current FOSS development:

Stable and full OpenGL 2 support for 99%+ of systems less than five years old so we can finally drop the old and clunky OpenGL 1 code from KWin.

I am not expecting this until 2011 but if it can be done this year I will be a very happy person.

[Yfrwlf]

Gallium3D, as writing drivers needs to be easier to help the Linux ecosystem as a whole. Core changes like that are the most important, along with the fact that closed drivers need to be ditched any way. Sure, improvements to the old are fine for now, but that's the future it seems.

Then, 2D and 3D tearless support.

[dashcloud]

I'll throw my list in here:

Better 945 Intel graphics (maybe even OpenGL 2.x support?)
Maybe a little boost for R100 radeon cards? (My other laptop has a Radeon 345M in it- this one has an Intel 945)
FGLRX to do well enough to start changing it's perception

Wine resolution switches to be perfect & flicker free (it's a lot better now than when it would just dump me out to the login screen).

[dashcloud]

The only thing I really want is the open source drivers to get their ass in gear with openCL support so that the multimedia programmers have motivation to get gpu assisted encoding going in linux. It's so sad to use a open API you really have to use a closed system.

Considering the x264 developers don't care at all about GPU encoding (none of the core developers are working it- it's been judged too much effort for too little return, and none of the folks who've popped up to say they would work on it have ever shown anything), I'm kind of curious if you'd use the GPU for other codecs.

[V!NCENT]

Restricted driver = no influence whatsoever

So going for open source...

The future is Gallium3D. Period.

So I would like to have a fully working Gallium3D driver first (for a capable graphics card like the r600 and not the ancient OpenGL 1.5 crap r300's).
Secondly I'd like the Gallium drivers, once fully working, to focus on speed.

Then I'd like to see the Mesa state tracker reach OpenGL 3.x compatibility, fully I might add.

This is the basic functionality that is what graphics cards are made for so that needs to be adressed first.

When this is all addressed the 2D+3D drivers for compositing performance mess should be resolved. This is performance hell that Mac OS X and Windows Vista and above not suffer from at all... Is this resolved with an X.org state tracker? Then that.

Then it's time for some horizontal and futuristic stuff like a fully OpenCL supported state tracker.

Then it's the details and the polish like video acceleration state trackers, etc.

After that I do not give a shit, realy... I first want my graphics card to support its primary purpose and get some speed out of it while at it.


I know my post sounds harsh, but I've been asked what I wanted,,,

[ad_267]

I just want to install Kubuntu and have suspend, Kwin effects, VLC, Google Earth, and Stellarium to work. I don't want to fiddle, I don't want to install.

Heh, now that's something that's more possible in Linux than Windows.

[dotancohen]

I just want to install Kubuntu and have suspend, Kwin effects, VLC, Google Earth, and Stellarium to work. I don't want to fiddle, I don't want to install.

[Apopas]

*Accelerated THEORA and DIRAC (and H264 ) playback.

As far as I know Theora doesn't support accelerated decoding in general. Am I mistaken?

[ckain]

Freaking awesome Intel Clarkdale GPU support. With all juices like GPU assisted video decoding, rock solid performance, and shit like that. That's what I would like get this year.

[cician]

*Open source drivers for PowerVR graphics. I see more and more ARM adoption, but always bundled with PowerVR and its closed driver
*Accelerated THEORA and DIRAC (and H264 ) playback.
*OpenGL[ES] 2.0 on i945. I know the hardware doesn't support all features, but soon we'll need it even for WEB! This crappy underdeveloped gfx is REALLY widespread and with Intel actually putting this crap into the CPUs...