What Do You Want In Linux Drivers This Year?

[hubick]

I want Intel to move their Poulsbo (xf86-video-psb) driver from Moblin to x.org and get the same team Keith Packard is on to actually maintain it like their other drivers - starting with porting it to work with a recent x server.

[deanjo]

OpenCL for sure, improvements to alsa (instead of that God aweful pulse audio), and quickly realizing that linux needs much better voice navigation and dictation (been testing alot the last few days and quite honestly they all suck in linux). Profile per application support in the nvidia driverswould be welcomed as well.

[energyman]

why are people using pulseaudio anyway?

[deanjo]

To ATI/AMD:
- Quit being Microsoft's b!tch and open up your blob driver.


To Nvidia:
- Quit being Microsoft's b!tch and open up your blob driver.

First of all you think it's M$'s doing. That isn't the case at all. If you were to strip out all the licensed 3rd party tech (very little of it owned by MS) your left with basically with what you already presently have in the opensource drivers. To have fully featured opensource drivers you would have to have all the owners of that IP agree. One hold back and you lose their technology where alternatives may not exist or have no satisfactory alternative. Just like when a project wants to change licenses, you have to have all contributing parties are on board with the change or you have to lose their contributions. It's a far more complex issue then "stop being MS's bitch".

[bugmenot]

On the open source side (radeonhd/nouveau/openchrome):
Powermanagement, OpenCL support, video decode support (over OpenCL?), Gallium3d working (3D), KMS and DRI2 working well.

Radeonhd/Nouveau/openchrome should be rock stable. And in order to avoid having two drivers for the ATI radeonhd GPUs, radeon should stop supporting chips from r700 on. 3D will be in Gallium3d anyway.

I wish documentation for nouveau developers and in general more people getting involved with the development of the open drivers, also for ati/openchrome drivers. Now we have the docs for ati chips, but you can't ever have enough developers.

I don't expect/demand high performance of the open drivers (which is no problem for me, you can buy a better card if it is too slow, 70% of fglrx speed would be nice.) Open drivers should 'just work'.
If you really only want to play, you'd probably use fglrx/proprietary nvidia driver anyway.

For fglrx I wish it to be more stable in general. But from the moment on radeonhd supports 3D and powermanagement and maybe video decode (even if only in a very simple way) I don't see fglrx to be still important for most users.

Thanks for reading!


Edit: And of course I want a really good solution for the GEM/TTM problem. TTM in the kernel? Openchrome/radeon(hd)/nouveau should use the same interface. And GEM isn't ready for dedicated cards.

ATI should go on designing its products open source friendly and there should be a better timings for the releasing of X (and also more devs working on it)
To quote Steve Ballmer: "Developers! Developers! Developers!"
(Although this is no proof to get a good product (look at windows))
And also I wish good support for old products (yes I know, AGP is ugly, hopefully it dies soon) and want better support for great projects like coreboot.

[Craig73]

Intel Driver

Keep up the great work - you have been implementing some amazing new stuff... but if I had to pick something additional, it would be to enable TV-out on my 855GM card because my computer is still quite capable (and I can't drop Windows comfortably without this)

Of course
- Tear free video when compositing is turned on
- TV-out support for older cards (855GM) in the intel driver (only available in i810 right now)

[another_sam]

world domination or gtfo!

radeonhd R500 2D and 3D graphics working, including full hd dvi monitor and s-video old-tv outs, shipped with ubuntu (9.04 or 9.10).

ah, and... the destruction of all the linux sound drivers except one. no matter which; only one that works (16 bit stereo is enough), please.

[DanL]

First of all you think it's M$'s doing. That isn't the case at all.

Didn't ATI state that they didn't want to open fglrx for fear of losing their WHQL certification or giving away DRM secrets?

It's a far more complex issue then "stop being MS's bitch".

Maybe, but the whole Linux video driver situation just reeks of rampant corporate greed.

[bridgman]

WHQL certification is only an issue to the extent that certification requires DRM compliance. DRM is really a Hollywood thing, not a Microsoft thing.

[oblivious_maximus]

First of all you think it's M$'s doing. That isn't the case at all.

Didn't ATI state that they didn't want to open fglrx for fear of losing their WHQL certification or giving away DRM secrets?

It's a far more complex issue then "stop being MS's bitch".

Maybe, but the whole Linux video driver situation just reeks of rampant corporate greed.

Yeah, who are these 3rd parties(other than Macrovision and AACS-LA) anyway? Did they contribute code to the driver? If so then why doesn't $MEGA_CHIP_CORP just contract them and keep the rights for themselves, and why can't $MEGA_CHIP_CORP just write it themselves in the first place? If not then why do they get a say in what $MEGA_CHIP_CORP can or can't let be known WRT the driver? Something has run amok there for sure.
-----
What I'd like to see:
from Nvidia

• A tvout overscan control (a position control would be nice too - you know, like the one that Windows users (with 7x00 and earlier model cards that is) get?)! It works fine up to the 7x00 cards. Make it work for subsequent hardware also! The fact that it just stopped working when you came out with a new line is absolutely ridiculous! It's the kind of thing that makes a person want to hurl epithets!
• A compelling reason to stay with Nvidia once there's a remotely stable open ATI driver with basic 3D and TVout support. Even with an overscan fix I'd still need something else to keep me from switching.
• de-obfuscate nv. Why even bother if you're just going to be dicks?

from AMD

• to be able to look at the threads about fglrx releases and not cringe at how everything just seems to stay broken.
• Not saddling GNU/Linux with Windows paradigms, certainly not if it means the answer is "that's not supported" when someone asks why they can't do it the UNIX way.
  eg: X :1
  eg: VT switching should not be a source of trepidation! See first point.
• the light at the end of the tunnel? the pot of gold at the end of the rainbow?
• more of the same on every front that isn't fglrx. 0-day GPU docs/xorg drivers?

from AMD and Nvidia:

• reiterating other posters, I think there should be a lot less "$PIECE_OF_LINUX_ECOSYSTEM isn't good enough, let's do it ourselves and put it in our BLOB" and more "$PIECE_OF_LINUX_ECOSYSTEM isn't good enough, let's do it ourselves and make $PIECE_OF_LINUX_ECOSYSTEM better."
• No more 3rd party dicks. You cats should be able to do whatever the hell you want with your driver, end of story.