NVIDIA's Looking To Expand Its Linux Team

[TheBlackCat]

That isn't the case at all however. Wayland is still a proof of concept more then anything else and there are still tons of plans and releases ahead for X.

GTK, Qt, and KDE are all in the process of getting wayland support working.

[TheBlackCat]

All depends on the implementation of the BIOS.

I was under the impression it had more to do with how the hardware was built, with proper support requiring a building an extra, separate hardware pathway for the Nvidia card which very few companies appear to be willing to do (with such implementations being mostly limited to enterprise-level laptops).

[deanjo]

There is a difference between code to support the blob and code to give preference to the blob.

There is no difference. It was a simple situation where if the blob was installed it would use it otherwise utilize the free drivers if no blob was installed.

Do you have examples where they turned down patches that were needed to support features on anything other than its technical merits?

Pretty much every feature that the blob provides. By turning down the above mentioned patch they essentially knocked out all the features of it as well. Don't forget that kernel guys also turfed out parts of the kernel in the past until they realized that virtualbox was also using them. Then all of a sudden they decided they would keep those in the kernel until vb had a chance to adjust.

And I am not exactly clear on how GPL makes optimus impossible.

The greatest difficulty would more then likely having to interface with KMS (which as we know is a big no no because of the GPL licensing). You are also dealing with multiple memory manager between the two video devices.

GTK, Qt, and KDE are all in the process of getting wayland support working.

That still doesn't mean much. Until the day that Wayland is poised to replace X it isn't a pressing matter.

I was under the impression it had more to do with how the hardware was built, with proper support requiring a building an extra, separate hardware pathway for the Nvidia card which very few companies appear to be willing to do (with such implementations being mostly limited to enterprise-level laptops).

I see what your getting at, you are referring to the crappy implementations that vendors like Asus and Sony use where they hardwire the IGP to the display. This is true however not ever vendor of optimus graphics out there does this.

[TheBlackCat]

There is no difference. It was a simple situation where if the blob was installed it would use it otherwise utilize the free drivers if no blob was installed.

Of course there is a difference. You were talking about "Xorg's reluctance to accept any code required for nvidia blob use". But the example you gave was not required for use. It made things slightly more convenient, but it wasn't by any stretch of the imagination required in order to use the blob. It also had no impact whatsoever on what features the blob could or could not use.

Pretty much every feature that the blob provides. By turning down the above mentioned patch they essentially knocked out all the features of it as well.

If people are still able to use the blob just fine without it, then by definition it was no required. It was a convenience function, it did not provide any new functionality, nor did it limit the functionality of the blob when it was used.

Are you seriously arguing that picking a default value is somehow equivalent to eliminating features? You do realize that people can still use Nvidia cards just fine even if they aren't the default, right?

The greatest difficulty would more then likely having to interface with KMS (which as we know is a big no no because of the GPL licensing). You are also dealing with multiple memory manager between the two video devices.

Why would it have to interface with KMS? And why would dealing with multiple memory managers cause licensing problems?

That still doesn't mean much.

Wait, the fact that three of the biggest players in the entire open-source development world are all working on wayland support "doesn't mean much"? If that doesn't mean much I can't imagine what would.

Until the day that Wayland is poised to replace X it isn't a pressing matter.

That is my whole point. By the time Wayland is poised to replace X it will be too late. They need to start working on it well beforehand if they are going to be ready by the time Wayland is ready. That is exactly why groups like GTK, Qt, and KDE are working on it now. These sorts of things don't happen overnight.

Nvidia always waits until the last second on these things, and then it takes months or even years to get caught up, leaving users stranded in the meantime. It is exactly the philosophy you are espousing here that Nvidia needs to change. They do this every time, and every time it causes no end of problems to users.

I see what your getting at, you are referring to the crappy implementations that vendors like Asus and Sony use where they hardwire the IGP to the display. This is true however not ever vendor of optimus graphics out there does this.

I thought this was the recommended layout, so that the IGP could handle simple tasks and off-load more difficult ones to the Nvidia card. Laptops that don't do these either aren't taking advantage of optimus's capabilities, or require setting up a much more complicated system with several different pathways for graphics data.

[smitty3268]

The greatest difficulty would more then likely having to interface with KMS (which as we know is a big no no because of the GPL licensing). You are also dealing with multiple memory manager between the two video devices.

Their driver already has to interface with the GPL licensed kernel. They work around this by providing a compatibility layer with a GPL license, that they can safely call from their proprietary driver because the abstractions mean the driver isn't "derived" from the kernel. I see no reason they couldn't do the exact same thing with Optimus. It's likely a bit more complicated, but not fundamentally different as far as i can tell. They just don't want to go to the effort to get a consumer-oriented feature working on linux, when they are mostly focused on more business-oriented features for that market. And I'm not sure what you mean by multiple memory managers - Windows drivers don't share memory managers between each other either, they are highly hardware specific and hardware vendors spent a lot of time coming up with optimal solutions for their hardware.

[t.s.]

my english really fails here...sorry. amd really hiring more opensource devs really!
but yes if i need the functionality i use the amd cloused source driver to.
means i'm not against the the nvidia cloused source driver i'm just against there Politic about: not give the opensource guys a chance by support with spec's.

Yep, that's what I mean. Well, glad to hear that [they hire more devs].

really Fusion beats tegra in all ways...

Hm.. how about power consumption?

[DeepDayze]

I don't think that this would be a good idea, for several reasons:

- 3rd party code
- patented software
- it would be an island separate from everything else in Linux land like it is now. Much of the nouveau core is well designed and efficient, it's just missing the countless optimisations across the board. A HUGE driver which reimplements all of Mesa, most of X, and the entire linux kernel would be impossible for most OSS hackers to understand, let alone improve on.

The best way would be to support nouveau with documentation (most important) and developers.

Exactly in much the same fashion AMD has done with supporting the FOSS radeon driver...all nvidia has to do is release docs that has all the info that devs need in order to write a usable driver..of course removing anything referring to 3rd party stuff and getting such docs cleared by their legal department before releasing.