kernel based modesetting

[martini]

@nvidia: are there any plans to support the upcoming kernel based modesetting?
would that be open or closed source?

[Mr.A]

Originally Posted by martini

@nvidia: are there any plans to support the upcoming kernel based modesetting?
would that be open or closed source?

That is a really good question!
In addition, I would also like to know what plans NVIDIA has relating to DRI2, Gallium3D, TTM/GEM, XVMC (for newer Chips).

But as far as I know NVIDIA, they don't like to answer these questions in detail.
Who knows, maybe I am wrong and NVIDIA would really like to answer these questions and show up what the future will bring to us customers.

best wishes
Mr. A

[AaronP]

Unfortunately, we can't. Kernel modesetting was specifically designed to be incompatible with the NVIDIA driver.

There are no plans to use DRI, TTM, or Gallium3D since the driver already has its own code for what those components provide.

[Kirurgs]

Originally Posted by AaronP

Kernel modesetting was specifically designed to be incompatible with the NVIDIA driver.

Is this really true? So this is like they specifically planned everything to be incomplete with nVidia? I'm just trying to understand why the heck they would like to do that...

[AaronP]

Well, let me rephrase that... it was specifically designed to be incompatible with non-GPL drivers, at least according to Dave Airlie when I asked him about it a couple of months ago. I haven't actually looked at the code, myself.

[Mr.A]

Originally Posted by AaronP

Well, let me rephrase that... it was specifically designed to be incompatible with non-GPL drivers, at least according to Dave Airlie when I asked him about it a couple of months ago. I haven't actually looked at the code, myself.

That is sad.

I really hoped that NVIDIA would release GPL drivers.
So in the future it looks like I have to switch to another graphics solution, were everything works out of the box.

At the moment it looks like that AMD and Intel will give use (the end user) a better Linux Desktop experience in the near future. They will provide kernel based modesetting, DRI2, XVMC (Geforce 8+ users are waiting for it for a ling time now), Gallium3D (could make your live easier, take a look at this GSOC Project. http://code.google.com/soc/2008/xorg...D6AA025594454A)

So it is on you NVIDIA to take the chance.
Everybody could win from a GPL driver, YOU and the User.

best wishes
Mr. A

[ledoc]

Originally Posted by AaronP

There are no plans to use DRI, TTM, or Gallium3D since the driver already has its own code for what those components provide.

Ok, but that would also mean that in the long run, once those things have settled, you could toss that code from your driver, reducing the amount of legacy closed source code and make your life (and ours, too) a lot easier...

[AaronP]

Originally Posted by ledoc

Ok, but that would also mean that in the long run, once those things have settled, you could toss that code from your driver, reducing the amount of legacy closed source code and make your life (and ours, too) a lot easier...

Actually, that would make things a lot more difficult for us. Right now, the NVIDIA drivers use the same core code across all platforms, which means that we can build the OpenGL driver pretty much unmodified on Linux, FreeBSD, Windows, etc. Using DRI, TTM, and Gallium3D would mean that the Linux OpenGL implementation would diverge significantly from the Windows one, and the end result would be a poorer user experience and more work for everybody.

[Mr.A]

Originally Posted by AaronP

Actually, that would make things a lot more difficult for us. Right now, the NVIDIA drivers use the same core code across all platforms, which means that we can build the OpenGL driver pretty much unmodified on Linux, FreeBSD, Windows, etc. Using DRI, TTM, and Gallium3D would mean that the Linux OpenGL implementation would diverge significantly from the Windows one, and the end result would be a poorer user experience and more work for everybody.

So as I see it, the solution would be Gallium3D.

http://www.tungstengraphics.com/tech...gallium3d.html

Then NVIDIA would be able to support more platforms with a simpler driver.
Something must happen in the near future. The situation like it is right now isn't really good.

As time passes by NVIDIA does support less linux features like XRAND 1.2+ or xvmc.

I know OpenGL support is important but that is not everything.

Regards
Mr. A

[gbil]

Originally Posted by Mr.A

So as I see it, the solution would be Gallium3D.

http://www.tungstengraphics.com/tech...gallium3d.html

Then NVIDIA would be able to support more platforms with a simpler driver.
Something must happen in the near future. The situation like it is right now isn't really good.

As time passes by NVIDIA does support less linux features like XRAND 1.2+ or xvmc.

I know OpenGL support is important but that is not everything.

Regards
Mr. A

I really doubt NVidia will change their driver policy and design just to accommodate better linux compatibility. Don't take me wrong, I would love such a development, but market share doesn't allow this - at least for now.

[Henning Rogge]

Originally Posted by Mr.A

That is sad.

I really hoped that NVIDIA would release GPL drivers.
So in the future it looks like I have to switch to another graphics solution, were everything works out of the box.

If Nvidia gets lucky, the Nouveau driver will be good enough to replace the Nvidia binary driver in a year or two... we still have no support for the recent card generation, but at least there will be hope.

If not AMD/ATI and Intel will grab the market share of Nvidia for the linux market.

The recent driver discussions and comments like "we don't plan to support feature X for anything else than Windows" from Nvidia are the reasons I still work with my 7900GS... I got a larger monitor and some of my games are not as "smooth" as before, but at the moment it seems to be a bad decission to buy a new nvidia card for linux users...

Just my 2 cents.

[Mr.A]

Originally Posted by gbil

I really doubt NVidia will change their driver policy and design just to accommodate better linux compatibility. Don't take me wrong, I would love such a development, but market share doesn't allow this - at least for now.

I don't think linux market share is as small as everybody says.
It is hard to measure how big the market share really is.
If I take a look at MacOS, the market share isn't really big compared to WinXP.

But one thing I know for sure, proprietary drivers don't help to spread linux, quite the reverse.

regards
Mr. A