DMA-BUF PRIME Coming Together For Linux 3.5

[danger89]

Good news, very nice work David Arlie and Alex Deucher! I hope to see it soon, I can't wait

We clearly need to clone Dave and Alex a couple of times each and then we'd have the linux graphics stack done in a whiff.

HAHA

[soupbowl]

When you buy a video card, you ARE free to use it as you like. Feel free to stab your wrists with it or cook on it.

[scjet]

If Bumblebee team needs testers, use case reports, bug reports or anything else during Ubuntu 12.10 alpha and beta, I'm in.
I know your project shares the same goal as PRIME, but the current backend is a hack, and if you replace VGL with PRIME, I'll be really interested in following the developement and testing. I can't code, but I'm experienced with basic debugging tools, so I think I could give a hand with that.

In other words, this is just yet another hopeful disgusting lie to appease the peasants. !
omfg, it really never ends, does it ?.
Here's what it is dude.
M$ has enough money to even buy the most stalwart tree-huggin' Linux dev(s)/programmer(s), even you, even if you're NOT a tree-jerker, who cares abyway right ?
So why are you being the worser fo those evils by pretending that maybe there is a chance that Corporations are Comsumer-tree-huggers too ?
Why are you, and your stoopid articles being the cruelst false-hope derivitive.
We start a petition and then we die under it, knowing no one gives a crap. That would save all these so-called authors a lot of wasted energy ?.
But hey, I'll clear it all up for you ALL.
If I PAY for an nvidia/ATI GPU then i want to use it, whenever I want. forever I want !!!
Thats all those lying-greedy-corporate popsicle-suckers need to know. Anything else is just illegally "stealing" from the Consumer.
allrighty then.

[Nepenthes]

If Bumblebee team needs testers, use case reports, bug reports or anything else during Ubuntu 12.10 alpha and beta, I'm in.
I know your project shares the same goal as PRIME, but the current backend is a hack, and if you replace VGL with PRIME, I'll be really interested in following the developement and testing. I can't code, but I'm experienced with basic debugging tools, so I think I could give a hand with that.

[LLStarks]

Does this mean Optimus support could land in Ubuntu 12.10 (I don't expect it in the "official" distribution, but with xorg-edgers, or any state of the art X server, kernel, and drivers repository)?

Judging by current progress, I'd say yes.

Kernel: Landing in 3.5
Xserver: Might land for 1.13, but likely Edgers
Intel DDX: Edgers eventually
Nouveau DDX: Edgers eventually
Mesa and glproto: Unknown if needed

[agd5f]

What else need to be done to support Optimus and PowerXpress in Mesa drivers? At least at the same level like in Catalyst (with Xorg Server restart).

Nothing really on the mesa side. There are basically two things needed. Define a way for kernel drivers to share buffers (DMA BUF), and fix X to allow decoupled display and rendering (Dave's X server work).

Any chance for MUXless notebooks? Still hope...

That is one of the main goals of this work. Other goals include support for acceleration on non-accelerated devices like USB displays using the GPU and zero copy buffer sharing between subsystems (e.g., V4L and KMS).

But AFAIK there is no documentation published about Crossfire. And since there is many use cases (different Crossfire generations and 2, 3 or 4 GPU configuration), I guess this is a huge task!

There is nothing really secret about crossfire or any additional magic bits that need to be documented. It's mostly all done in the driver with software. The driver just has to add the necessary infrastructure to proper split jobs between multiple GPUs. You are correct that that is a big job however.

[Nepenthes]

Does this mean Optimus support could land in Ubuntu 12.10 (I don't expect it in the "official" distribution, but with xorg-edgers, or any state of the art X server, kernel, and drivers repository)?

[gilboa]

Is it still EXPORT_SYMBOL_GPL?

- Gilboa

[zeealpal]

Also CrossFire is only useful for gaming... compute operations do not require it (although Catalyst is still a better choice for compute) and most people will not need CrossFire for linux gaming.

[whitecat]

Why? Is there some reason to think that it will take long time?

Because Crossfire is a useless technology in most cases and there are tasks very more useful to finish before (Power-management, OpenGL implementation, performance[*], OpenCL, Video decoding, 100% piglit match, etc.).

I don't think people hired by AMD, Red Hat, VmWare or Intel will work one day on multicard support. Perhaps a independant developer will do it, who knows. But AFAIK there is no documentation published about Crossfire. And since there is many use cases (different Crossfire generations and 2, 3 or 4 GPU configuration), I guess this is a huge task!

[*] Currently the free 3D drivers are much slower than the proprietary ones, so perhaps when the drivers for mono-card's configuration will be on par with proprietary ones, there will be an interest in multicard's support.