Slashdot Mirror
NVIDIA Open-Sources 3D Driver For Tegra SoCs
An anonymous reader writes "Linux developers are now working on open-source 3D support for NVIDIA's Tegra in cooperation with NVIDIA and months after the company published open-source 2D driver code. There are early patches for the Linux kernel along with a Gallium3D driver. The Tegra Gallium3D driver isn't too far along yet but is enough to run Wayland with Weston."
Linus made it happen!
Will this benefit OUYA?
He who laughs lasts.
So phoronix had the news yesterday about an open-source wrapper to AMD's concession to open-source "AMD Releases UVD Engine Source Code" for the kernel-level wrapper code, and today, NVidia open sources the 3-d driver for Tegra. That's progress. I use the Nvidia binary blob on my debian distro hardware, and the Nvidia blob with the knoppix live-boot system, as Nouveau does not work well enough on my hardware. I hope this will help Nouveau a lot.
April Fool's Day was MONDAY, you morons! And this one isn't even that good! Nvidia? Open sourcing their 3D hardware? Come ON, who do you expect to fool with THAT?
Unfuck you, NVidia!!!
I was promised a flying car. Where is my flying car?
What NVidia did was document a very small and specific part of the chipset. They previously opened documentation of 2D accelleration, now the 3D part. The part that accelerates media playback is still closed. Given the fact that this is a SoC that will most likely be used for media playback just as much as for gaming and it's not their own driver code they have released, I'd not consider this open sourcing. They are merely releasing part of the specifications so third parties can develop drivers. Yes, they are actively helping one company, but there is no actual working code available as open source yet. Not from the 3rd party company, nor from NVidia themselves.
I was promised a flying car. Where is my flying car?
Hacking benefits maybe. But running a partially implemented and slower open source driver isn't going to attract too many additional customers to OUYA.
It's been possible for *years* now. You just use their L4T/Linux-for-Tegra/Vibrante product. It comes with ordinary glibc, a sort of hacked up mini distro based on Ubuntu tweaked to load their modified kernel and userspace. You have to use EGL instead of GLX, but otherwise everything works. If you're building embedded apps on Tegra things like Clutter work with EGL now. So very pretty custom user interfaces are possible and fully accelerated.
Looks like they(nvidia) has some support for OpenKODE as well on Linux for Tegra.
So, completely useless...
[The Tegra-based OUYA console] is dead before it's even hit the gates.
Then which platform for playing indie games should people be backing instead?
The driver the article refers to is NOT developed by NVIDIA, it's reverse-engineered. And it's also in very early development, so it's not very news-worthy yet.
they magically discover the advantages of becoming more customer and developer friendly- well at least until their fortunes reverse.
Nvidia is barely managing to keep their ARM project alive. The first Tegras were bombs. Tegra2 was late, had to be heavily discounted, and couldn't decode H264 HD content. Tegra3 was late, expensive, and is no more powerful than the better parts from the Chinese competition of Rockchip and Allwinner. Tegra4 is so late and expensive, it has possibly no current design wins, and will not appear in this year's refresh of Google's massively successful tablet.
Nvidia struggles to have first quality GPU performance (it is currently slaughtered by Qualcomm's Adreno- and old ATI mobile design- and the best PowerVR options). Nvidia struggle to have decent low power operation. Nvidia struggles to get its parts on the latest process nodes. Nvidia struggles to have a family that contains even one current SoC, although it is promising for the first time to have two parts in its current line-up by years end.
Nvidia needs friends, and badly. The GPU (up to and including the Tegra 4) is now officially obsolete (Tegra 5 gets a PC class GPU) so Nvidia can 'generously' reveal its inner working to those that wish to create open-source driver alternatives to the usual proprietary 'binary-blob' drivers.
It should be noted that open-source drivers always run like crap- buggy and with a fraction of the performance of the binary blob ones. A modern GPU system is an horrifically complicated state-machine. If open-source bods were not driven by zealots, they would demand access to a hyper-thin, 'to the metal' 'binary blob' driver that would provide the most primitive and direct API for accessing the GPU hardware and video/image processing functions. Wanting to write code that directly talks to the hardware is just plain daft. Nvidia understand their hardware in ways no sets of documentation can ever compensate for. What developers need is that thin universal driver that works on all Tegra platforms.
Dude, TNT2 didn't even have hardware TNL. There were no shader support whatsoever on it.