Slashdot Mirror

← Back to Stories (view on slashdot.org)

NVIDIA Engineers On The Realities Of Linux Drivers

Posted by timothy on from the not-at-this-time dept.

linuxquestions writes "LinuxQuestions.org recently interviewed members of the NVIDIA Linux team. The interview covers the internal use of Linux at NVIDIA, the current demand NVIDIA is seeing for Linux drivers, the biggest perceived obstacle in Linux becoming a mainstream gaming platform and the decision to maintain both an Open Source and closed source Linux driver."

21 comments

  1. cool by Lith+Maethor · · Score: 0, Redundant

    oh yeah, first post and since more games come penguin supported,the future looks bright

  2. Unified Driver Infrastructure by crow · · Score: 5, Interesting

    Much of the interview is the standard optimistic corporate smiley-face stuff you would expect. What I found interesting is the reference to a unified driver infrastructure. Apparently the bulk of their driver code is identical across platforms, so mostly what they need to maintain for Linux is a compatibility layer.

    This is what they cite in not open sourcing the driver--too much of the unified code is licensed by them from third parties. (Now why don't they ask their sources about a dual GPL/proprietary license?)

    The followup question that this raises is: Given that the base driver code is the same across platforms, are there any particular aspects of X or Linux that reduce performance?

    1. Re:Unified Driver Infrastructure by Stevyn · · Score: 2, Interesting

      "Now why don't they ask their sources about a dual GPL/proprietary license?"

      You think this never crossed their mind?

    2. Re:Unified Driver Infrastructure by Espectr0 · · Score: 1

      If the driver is so unified, i wonder why mac users don't get the same linux-x86 nvidia drivers.

      --
      Open Source Java Web Forum with LDAP authentication
    3. Re:Unified Driver Infrastructure by abdulla · · Score: 1

      Because they're not x86? Or Linux. Mac's driver infrastructure is quite different from what I've read, and is infact embeded C++ based.

    4. Re:Unified Driver Infrastructure by Creepy+Crawler · · Score: 2, Informative

      WHat? Third party stuff? I call bullshit. Nvidia just didnt want their stuff to be GPL'ed.. Source..(cannot find original MS article)

      http://news.zdnet.co.uk/software/0,39020381,210103 7,00.htm

      (QUOTE) An OEM is required to sort through a potential legal morass of licensing issues around the GPL if it wants to protect its intellectual property rights. This creates extra costs from both a development and legal perspective. An example of this risk can be taken from nVidia. An nVidia programmer, in the course of developing a driver for one of its products, used a portion of code from a freely available video driver. The developer failed to realise the code was licensed under the GPL and would therefore require nVidia to release the source code for its entire driver. Because nVidia did not want to release the source code to its commercial software, the company incurred substantial costs to develop a new driver that did not contain the GPL code. (/QUOTE)

      --
    5. Re:Unified Driver Infrastructure by Gogo+Dodo · · Score: 2, Insightful
      You didn't read the article carefully. The nVidia programmer added GPL code to the driver, however doing so triggered the GPL clause where the entire driver needed to be released. Since third party licensed code is included in the driver that could not be released, the GPL code had to be removed.

      The only people that can truly comment on why the nVidia driver can or can not be released are internal to nVidia. So you have to take whatever nVidia says at face value. If they say there is licensed code in their drivers that they can't release, then you have to accept that, as you are in no position to say "Bullshit! There is no licensed code." You don't have access to the source, so you wouldn't know.

    6. Re:Unified Driver Infrastructure by 4of12 · · Score: 4, Interesting

      are there any particular aspects of X or Linux that reduce performance?

      Probably more so for X11, given its age.

      I'd be interested in having Someone That Really Knows tell me

      "Given the current trends in GPU speed, memory, system bus increases, 3D, scalable graphics and fonts, a clean sheet of paper, would it be possible to create a high-performance graphics subsystem that would last as long as X11 has?"

      "Could X11 be layered over the of the Y.NOT graphics system to speed its adoption?"
      --
      "Provided by the management for your protection."
    7. Re:Unified Driver Infrastructure by Creepy+Crawler · · Score: 1

      """Because nVidia did not want to release the source code to its commercial software"""

      "Did not want to" implies that they CAN. Can not would infer third party reasons why they cant.

      SO.. YOu just didnt read it carefully enough. I know what Im talking about.

      --
    8. Re:Unified Driver Infrastructure by cronius · · Score: 0

      Given that the base driver code is the same across platforms, are there any particular aspects of X or Linux that reduce performance?

      I'm not thinking this at all, NVIDIA has previously stated that creating drivers for Linux is easier than for Windows, so I'm thinking their Linux code is simply not up to par. Everyone knows drivers are everything when it comes to performance, they're simply not spending enough resources on the linux version.

      --
      Life is Reality
    9. Re:Unified Driver Infrastructure by Smallpond · · Score: 2, Insightful

      From the nVidia license agreement:

      "All title and copyrights in and to the SOFTWARE ... are owned by NVIDIA, or its suppliers."

      This might be standard legalese, but it certanly states that code isn't necessarily all nVidia's.

      By the way, "yum update" is not a good idea on Fedora Core 2 if you have the nVidia driver installed.

    10. Re:Unified Driver Infrastructure by boots@work · · Score: 1

      I suggest you go to Keith Packard's talk at the upcoming linux.conf.au: The (re)architecture of the X Window System, etc.

      The X.org people seem to be making good progress on updating X11 to suit modern hardware, while maintaining backward compatibility. If we threw out all our software every time the hardware changed we'd never get anywhere.

    11. Re:Unified Driver Infrastructure by c0p0n · · Score: 1

      I think that it was the driver itself. Recently, nvidia got out their new linux driver and, man, they got a perfomance boost so awesome that I cannot yet believe it. I am getting like 7fps more on Doom3 (from 21avg->28avg on demo1, 640x480), UT2004 goes way smooth and some IO hogging by the card was resolved (noticeable by clippy sound when playing games on certain versions of the kernel).

      In fact they're so awesome that I began to wonder if the previous driver releases were a bit fucked up. Still dunno, but the perfomance gain has been brutal.

      --

      Your head a splode
  3. already /.ed :( by Anonymous Coward · · Score: 0

    new record?

  4. Re:Article text incase of slashdotting by Anonymous Coward · · Score: 0

    Who modded this up? Look at the user who posted it, then think about how unlikely LQ is to get slashdotted. Don't feed the trolls karma.

  5. The obvious follow-up question by ArmorFiend · · Score: 1

    To me the obvious follow up question is: Okay, so you can't free the driver software. So be it. Can you at least open up the hardware specs, so that open-source people like the DRI boys can have a shot to compete with you?

    1. Re:The obvious follow-up question by Anonymous Coward · · Score: 0

      A couple follow up questions, including this one, have been posted.

  6. You'll get the same idiot reply out of all of them by leonbrooks · · Score: 2, Interesting
    They'll claim that their competitors will get a leg-up from it. My answer?
    "Said competitors have the labs, people and equipment to do stuff like electron microscopy if they feel the urge, and probably know more about how your product actually works than you do. Competitors can't steal ideas that are patented anyway. Not that unstable drivers are such a patentable idea. Your suppliers would probably be delighted if a competitor started licencing their technology, so they also have positive motivation to publish, which conflicts with yours. The only people you're really hurting are your customers, and consequently your own sales."
    They'll ignore that too, but the first one to go truly open (XGI did that, but only with their 2D stuff, ATI did too but only with limited amounts of their older stuff) will see people doing useful tricks and performance enhancements with their cards that they didn't think was possible. Once that becomes common knowledge, the people with the wallets will follow too. By that time, the damage will be done and the other manufacturers will be chasing tail-lights.
    --
    Got time? Spend some of it coding or testing
  7. Re:You'll get the same idiot reply out of all of t by Anonymous Coward · · Score: 0

    I would be suprised if the different manufacturers cards are so similar that they could copy parts of the design and splice it into their cards.

    Even though they generally do the same thing they must do it in ways that aren't 100% compatible, otherwise each card wouldn't have different drivers.

    The cards being designed today will be more advanced than those out now, so I don't see how opening the details would give the competitors much useful information, they would still have a lot of development to catch up - even if they completely copied one of today's cards.

    Hopefully the opensource graphics card mentioned on Kerneltrap will see the light of day and prove that the arguments given are just excuses.