Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Open Source Programming

NVIDIA Begins Supplying Open-Source Register Header Files 77

An anonymous reader writes: NVIDIA's latest mark of their newly discovered open-source kindness is beginning to provide open-source hardware reference headers for their latest GK20A/GM20B Tegra GPUs while they are working to also provide hardware header files on their older GPUs. These programming header files in turn will help the development of the open-source Nouveau driver as up to this point they have had to do much of the development via reverse-engineering. Perhaps most interesting is that moving forward they would like to use the Nouveau kernel driver code-base as the primary development environment for new hardware.
This discussion has been archived. No new comments can be posted.

NVIDIA Begins Supplying Open-Source Register Header Files

Comments Filter:
  • by Bruce Perens ( 3872 ) <bruce@perens.com> on Thursday June 25, 2015 @01:01AM (#49983223) Homepage Journal

    We have nVidia helping but not making their own Open Source driver. Intel, after a long period of Open Drivers, said it would require BLOBs for future graphical interfaces. AMD helps with Open Drivers more than nVidia so far but doesn't support them.

    • by Pi1grim ( 1956208 ) on Thursday June 25, 2015 @02:26AM (#49983421)

      Blob used to be stored in ROM part on the die, upgraded from time to time. Now they'll store in drivers directly and load it into the hardware on initialization. Intel didn't close anything, they just revealed the same blob that used to be hidden from the eyes and included it into driver. So, I'd say that's not Intel creating a problem, merely exposing it. Also, perhaps now that everyone has suddenly got their panties in a bunch over this issue, Intel might consider opensourcing the blob as well (in case publicity benefits will outweigh the work needed).

      • by Bruce Perens ( 3872 ) <bruce@perens.com> on Thursday June 25, 2015 @02:53AM (#49983451) Homepage Journal

        If they actually told us how to program their microengines, something good might come of it. But they'll probably just BSD-license a list of numbers, as others have.

        I liked writing bit-slice microcode at Pixar. I really could get every last bit of power out of the hardware.

        • One of the hardest lessons I ever learned in this business is that if you tell a customer (often any customer) about it, you support it. It does not matter in the slightest whether it's officially documented or not.

          I learned this lesson long before I started reading Raymond Chen's blog, which is full of examples.

        • Coding on the metal seems to be a dying art these days. If resources are plentiful, it's likely faster to implement a solution in a higher-level abstracted method. (And if resoruces aren't plentiful, it's easier to tell your boss "the hardware can't handle it" and get the objective changed to something less agressive.)
        • What type of GPUs does Pixar use? Is it still the A113?

          • I didn't actually work on GPUs very much at Pixar, the image computer I worked on was the grandfather of the SIMD image processing instructions on modern CPUs. What would become a GPU later on was a very expensive box from Silicon Graphics, I had one that cost at least a quarter Million dollars.

    • Comment removed (Score:4, Informative)

      by account_deleted ( 4530225 ) on Thursday June 25, 2015 @04:12AM (#49983623)
      Comment removed based on user account deletion
      • by Kjella ( 173770 )

        Citation please? Because the last press release I saw from AMD on the subject said they were releasing docs as fast as the lawyer could sign off on them and that they hoped to replace their binary blob in the future with the FOSS driver. To that end they had gone so far as to hire a couple extra devs to work on the FOSS drivers to help them get closer to release parity.

        Well no, not really. They're hoping to replace the base infrastructure on Linux with the open source "amdgpu" driver, but with separate open and closed source user mode drivers for OpenGL, multimedia and so on. It would get you up to release parity if all you need is a framebuffer and software fallback, but they are still planning to develop Mesa and Catalyst separately. They don't have any plans to give Intel a free high performance OpenGL engine or to let nVidia look at their game-specific optimizations.

    • by ledow ( 319597 )

      I'm an Open-Source advocate, don't get me wrong.

      However, they are under no obligation whatsoever, so why should they? What advantage do they get from opening them? What's going to be the thing that will make them want to open their drivers? What's going to outweigh potential patent etc. risks?

      Because, as far as I can see, they gain basically nothing. They might get a "good news" article or two but it won't increase their sales significantly at all.

      Are we still in the era of hoping that huge multinationa

      • The advantage? Their hardware could be used on more operating systems, without them having to do anything, and so they could sell more cards for more computers. Their business is hardware, after all, not software.
        • by dave420 ( 699308 )
          How many more OSs, and with what market share? You also didn't mention the many disadvantages, which if the numbers from the first question are too low, will make this a complete non-starter.
        • by ledow ( 319597 )

          Could be used on more operating systems doesn't translate to more sales.

          And it's not "without them having to do anything". Just patent/copyright-auditing the proprietary driver they had in order to open-source would probably wipe out any extra sales they gained alone. Let alone ongoing maintenance, catering for all the Linux kernel changes as they try to get it accepted into the kernel for several years, bus-changes, new versions of CUDA / OpenGL translations etc.

          Honestly, look into the costs. They would

          • Maintenance? Publishing free software themselves? Why? They could simply release the specs needed to interface with their hardware, and whoever wants would write its driver, much like the nouveau team is doing, but with much more difficulty. It would not be their business. And even if this means only 10 more users, it would be a gain anyway. But I know, I'm speaking too idealistically.
  • Oh, sure, Tegra (Score:2, Insightful)

    by Anonymous Coward

    Approximately zero people actually use Tegra in real life, which is probably the whole reason that this was authorized. Every generation they make huge noise about how awesome the new Tegra is, then it ships in maybe 5 or 6 devices, half of which can't actually be bought anywhere.

    • by Whiteox ( 919863 )

      Not so.
      There are more devices out there than you think:
      http://www.nvidia.com/object/t... [nvidia.com]

      • Re:Oh, sure, Tegra (Score:4, Informative)

        by CurryCamel ( 2265886 ) on Thursday June 25, 2015 @03:29AM (#49983513) Journal

        Sampling a few of the device's in your link, GP's claim that the devices can't be bought still seems to hold true, mostly.
        For example, of the devices on the first page you link to:

        Can't ship out of the US:
        -Acer Chromebook 13 CB5-311-T1UU
        -Acer Chromebook 13 CB5-311-T7NN
        -NVIDIA SHIELD Tablet
        -EVGA Tegra Note 7
        -Nabi Big Tab (24")

        Can be bought:
        -NVIDIA SHIELD Portable
        -Google Nexus 9

        Dead link:
        -HP Chromebook 14 G3

        No wonder nVidia can't sell them, if people cannot buy them... I wonder if there is something in the Tegra itself that causes this, or if the link provided was just a happy coincidence to confirm GP's gut feel.

        • by Whiteox ( 919863 )

          Hmmm... Maybe you/GP is right.

        • Sampling a few of the device's in your link, GP's claim that the devices can't be bought still seems to hold true, mostly. For example, of the devices on the first page you link to:

          Can't ship out of the US: -Acer Chromebook 13 CB5-311-T1UU -Acer Chromebook 13 CB5-311-T7NN

          Right, those are likely American models. My model in Germany has a different model number (it's a Full HD, 4 GB RAM model, BTW). I have it since October 2, 2014.

    • This is a gimmick to try to get Tegra to see an uptick in adoption. If Tegra actually starts getting into everything, everywhere, Nvidia will yank the carrot away as fast as they can, because Nvidia doesn't like sharing.
  • by Lisias ( 447563 ) on Thursday June 25, 2015 @02:38AM (#49983439) Homepage Journal

    A well said "fuck you" does wonders! :-)

  • In all these years I've been wondering why they are so jealous about their drivers. I know, it's a very complicated matter of APIs, exposing internal details, etc. But it's not that someone can suddenly copy their silicon by knowing how it exchanges data with the computer, is it? And who exactly would be able to copy their silicon with their current technology?

    Or are they afraid that a cheap manufacturer could use their precious drivers with a cheap, compatible card? But if their hardware is better than th

    • by Actually, I do RTFA ( 1058596 ) on Thursday June 25, 2015 @06:26AM (#49983893)

      ATI and nVidia try to compete for share. They have high-payed repstrying to convince companies making the games used in the benchmarks to use features that favor their cards over their competitions'. I can see publicizing the drivers leading to the discovery of new holes that screw up a specific card getting pushed.

      Security by obscurity is not a replacement for real security, but it helps in this narrow case.

    • by dave420 ( 699308 )
      They license a lot of technologies from other companies, and the conditions of those licenses usually specifically mention not being allowed to publish details of its operation, and especially not precise details. That alone means their hands are tied in this respect. It's the price we pay for the performance we get.
    • In all these years I've been wondering why they are so jealous about their drivers. I know, it's a very complicated matter of APIs, exposing internal details, etc.

      nVidia doesn't own a lot of the IP in their mainstream graphics cards. Tesla is a separate development, and they do own most of the GPU IP in there, so they can release the specs. But nVidia got deeply into bed with Microsoft in the NV2x era. They got insider information on Direct3D, which they used to guide geforce development, and they got their chip into the original Xbox, but they also wound up beholden to Microsoft. They have never outright come out and said that, but they've strongly implied it, and i

    • 1) They license other companies intellectual property. They can't just willy-nilly make all of their drivers' source code public.

      2) A decade ago, NVIDIA other companies were so competitive, they were concerned that revealing their driver source code could reveal their hardware tricks to speed up graphics performance. Then their competitors could reverse-engineer the detail in their next generation card, and "catch up" to NVIDIA's performance. Luckily, they may be finally realizing that they get more of

  • A fellow can dream.

  • by Anonymous Coward
    Haven't used Linux for ages. Under Windows I can have a deluxe NVIDIA driver installed right now.
    • Haven't used Linux for ages. Under Windows I can have a deluxe NVIDIA driver installed right now.

      The blob has the same performance in linux as in windows, games generally yield the same performance too, except for some that use a wrapper without decent optimization. See here for a januari test - http://www.phoronix.com/scan.p... [phoronix.com] The nvidia header file releases discussed in this article and their future plans might boost the opensource driver linux though, which performs badly.

If you think the system is working, ask someone who's waiting for a prompt.

Working...