Announcement

Collapse
No announcement yet.

NVIDIA Reportedly Working On A Unified EGL Driver

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Also I don't get what this hype about OpenGL ES is about, it is a tuned down spec for Mobile/Embedded Devices, there is no need to move all software on the desktop to it. Why limit ourselves when there is an open standart aviable, OpenGL?
    Because ES clears out a lot of the cruft from earlier openGL versions that is mainly preserved for CAD software. And more importantly, it is the version of GL that every device has, not just desktop. Even if an app / game you write now is never meant to hit mobile now, it probably will in a few years when phones are as beefy as modern laptops, but these mobile OSes got a clean graphics API slate and probably would (for good reason) trying to force adoption of full class openGL.

    Comment


    • #22
      Originally posted by AJSB View Post
      That possibly will never happen because of patents....i heard that both NVIDIA and AMD use stuff not-in-house so they could not fully ever open it....dunno if it's true....
      Also add the complex licensing behind video processing. DRM requires some component never to be allowed to be openly revealed. And saddly, on some engines (AMD's UVD) they are too much interlinked and indissociable with the parts useful for the community (the video codec hardware decompression). At least AMD has promised to address this in the future and make future UVD more open-source friendly.

      Originally posted by dee. View Post
      Yet ANOTHER problem that could be solved by simply abolishing software patents.
      The problem is that software patents aren't the only obstacle. you also have problems regarding DRM (not only DMCA, but the consortium controlling a given scheme can decide to revoke the key present in AMD hardware if they are un happy), and plain simple copyright (Nvidia could decide to opensource code that they own - basically: that they wrote themselves. They have no control over code that they don't own - basically: code that they bought from 3rd parties).

      You'll need to abolish not only software patents, but also DCMA, copyright on software, and outlaw the activities of organisation like those handing the keys for HDCP encryption.
      I'm all for it, but its a much more difficult task than what even you suggested before.


      Originally posted by zanny View Post
      but the current lack of documentation about how Nvidia / AMD gpus work from the manufacturers means its hard to implement key components.
      AMD *do* document their hardware. It's only slow, because any release (documentation or code) has to trickle through their legal department to make sure none of the opponent i cited before feels violated. But their are doing it. And promising (and actually showing signs or) stream-lining the whole process, and integrating the taking of account of FOSS in their hardware development life cycle. Just remember that H/W is a very long process. As we are talking right now, somewhere in their R&D department, the first steps of hardware 2 generations away from now are already beginning. That means, we still have to wait a few generation between an announcement (we're going to play nicely with open-source) and actual hardware designed under this policy hitting the shelves.
      I think the current generation of Radeon is the first whose entire design dates after the beginning of AMD's opensource policy right after their acquisition of ATI. Now factor in that they have to get used to the process, fix the problems, iron-out the bump... it's getting to take quite a lot of time before they are a fully open-source supporting house like Intel. (As an indication, look at all the time it took google to transition to an FOSS model of development with Android and integrate their mods into main-line kernel, *even* if they've been pro open-source from the begining, and *even* if it's software only with way much more shorter cycles than hardware).

      Nvidia indeed *does not* document their desktop hardware. But regarding to Tegra and their mobile hardware, they are warming up to the idea of opening and documenting their stuff (probably because Linux *is* a very strong player in the embed world). But then again (specially taking into account the transition of AMD mentioned above) it's going to take quite some time until their opensource offering rivals what AMD (or even Intel) are doing.


      Originally posted by cirk2 View Post
      But wouldn't that be be a horendus step backwards?
      Isn't OpenGL Es only supporting a subset of OpenGL 3 features? {...}
      Also I don't get what this hype about OpenGL ES is about, it is a tuned down spec for Mobile/Embedded Devices, there is no need to move all software on the desktop to it. Why limit ourselves when there is an open standart aviable, OpenGL?
      OpenGL ES 2.x = subset of OpenGL 3.x
      Current OpenGL ES 3.x = subset of the current OpenGL 4.x
      (That's why Mesa can claim Open GL ES 3.x support sooner than OpenGL 4.x. : less things to support).

      Now you have to look closely:
      - what OpenGL ES and OpenGL have in common is all basic graphic functionality needed to draw things on a screen. (Triangles, shaders, etc.)
      - what OpenGL has in addition is a lot of complex arcane stuff which currently is only used on high end workstation running CAD software. Things that don't actually make sense for any everyday use of graphic hardware (desktop compositing, gaming, etc) but only make sense for this specific hardware.
      - also OpenGL ES keeps a lot of old almost 'deprecated' functionality because of backward compatibility to please the CAD-on-workstation crowd. For example you'll find a lot of API for fixed pippeline handling, even if most of the modern hardware use programmable shaders. OpenGL ES tries to throw away some of this (almost unused) legacy cruft.

      So:
      - Open GL = every single API call dating all the way back from when it was a library running on expensive SGI workstations.
      - Open GL ES = only what modern hardware does and what modern applications needs.

      Believe it or not, under the hood, there's a lot more in common between what latest generation embed hardware is doing and your desktop, than there is between an old dusty Silicon Graphics machine and your desktop.

      So the hype is that it is a thiner and lighter layer that does exactly what Open GL is currently used for on desktops, making it easier to develop for and making it easier to add support into FOSS drivers.
      The functionnality we're throwing away between OpenGL ES 3.x and OpenGL 4.x mostly isn't that much relevant to us.

      *BUT* this functionnality is relevant to a very specific crowd, which happens to be the most profitable market segment for hardware makres (the CAD-on-workstation crowd) so you can expect that vendors will continue to cater to them and produce OpenGL in addition to OpenGL ES support.

      Just don't expect regular dekstop application to use much of OpenGL beyond what's available in OpenGL ES. Maybe expect for a few specific stuff. Which never the less can be incorporated into OpenGL ES as specific official extension. (Which aren't supported on some mobile devices).

      Comment


      • #23
        Originally posted by zanny View Post
        Because ES clears out a lot of the cruft from earlier openGL versions that is mainly preserved for CAD software. And more importantly, it is the version of GL that every device has, not just desktop. Even if an app / game you write now is never meant to hit mobile now, it probably will in a few years when phones are as beefy as modern laptops, but these mobile OSes got a clean graphics API slate and probably would (for good reason) trying to force adoption of full class openGL.
        So they're using OpenGl ES as an "workaround" to introduce necessary but possible compatibility breaking cleanups? Well thats ok then... But still it is quite a way behind of what is possible with OGL4.3, and I dislike the concept to limit everyone to the smallest denominator. That caused near-stalling of Graphical development in Games because the Consoles stuck at 7 year old HW, and 'I hope the don't repeat that process with OpenGL ES

        Originally posted by DrYak
        OpenGL ES 2.x = subset of OpenGL 3.x
        Current OpenGL ES 3.x = subset of the current OpenGL 4.x
        wait GL ES 3 is OGL 4? now that's confusing.
        edit: Ok but it seems to leave out the Tessellation. Understandable for their target but Tessellation is one of the most interesting Technics introduced in OGL 4/DX 11.
        Last edited by cirk2; 08 March 2013, 10:28 AM.

        Comment


        • #24
          GL ES versions also tend to align better with HW generations.

          GL versions sometimes use "mostly functions from DX(n), with a few from DX(n+1)", which ends up fitting some chips but not others.
          Test signature

          Comment


          • #25
            THANK YOU, NVIDIA

            I want Wayland on my desktop, your finally looking at producing an EGL driver.

            great frickin' news

            Comment


            • #26
              Originally posted by ninez View Post
              THANK YOU, NVIDIA

              I want Wayland on my desktop, your finally looking at producing an EGL driver.

              great frickin' news
              Yeah seriously, I have no idea what these Zealots are screaming about. Ubuntu brought Nvidia to EGL (Mir & Wayland) single handedly. They are acticly working with Nvidia to go this. NO ONE in wayland seems to have even got Nvidia to entertain the idea. Seems like a success story to me.

              Comment


              • #27
                Originally posted by uid313 View Post
                Great that Nvidia is developing a EGL driver, then we can use proprietary Nvidia driver on Wayland which use OpenGL ES.

                However, to get full OpenGL support, and not just OpenGL ES support, we need to split out X support from libgl.

                libgl is needed for full OpenGL support, but unfortunately it is tied to X.
                There's a proposal for exactly that: https://github.com/aritger/linux-opengl-abi-proposal

                Comment


                • #28
                  Originally posted by dh04000 View Post
                  Yeah seriously, I have no idea what these Zealots are screaming about. Ubuntu brought Nvidia to EGL (Mir & Wayland) single handedly. They are acticly working with Nvidia to go this. NO ONE in wayland seems to have even got Nvidia to entertain the idea. Seems like a success story to me.
                  1) There is exactly as much official commitment from Nvidia to Mir and Wayland (hint: 0)
                  2) There is approximately as much expected commitment from Nvidia to Mir and Wayland ("maybe" and "someone says so")
                  3) Both need EGL + extensions. Sadly, not the same extensions.

                  Comment


                  • #29
                    Originally posted by spstarr View Post
                    For all those who don't understand, I don't want to be *forced* to use binary drivers because vendors will be happy to close up their driver code again, putting me into a black box. I *DON'T* want to live in a Windows world of no control over what runs on my systems!
                    You won't be forced. You can use open-source drivers in addition to closed-source binary drivers. You absolutely have control over your system - you can uninstall the binary drivers if you do not want them.

                    Originally posted by spstarr View Post
                    Mir will undermine *ALL THE WORK* that's been done to get vendors to open source their drivers and allow for both Linux and *BSD communities the ability to use open source drivers.

                    The only thing Mir undermines is collaboration between two sets of devs working on the same (or if you'd believe Canonical, not-quite-the-same) display server. In my view, it duplicates effort needlessly and Canonical has yet to come up with a single valid technical reason as to why Wayland could not be made to support their paradigm of thinking. The cherry on top is that Mark Shuttleworth reneged on Wayland after committing to it publicly.

                    Any GPU vendor can still choose to open-source their drivers, regardless of what happens with Mir. Wayland already runs on open-source GPU drivers.

                    Lastly, please stop trolling.

                    Comment


                    • #30
                      Originally posted by dh04000 View Post
                      Yeah seriously, I have no idea what these Zealots are screaming about. Ubuntu brought Nvidia to EGL (Mir & Wayland) single handedly. They are acticly working with Nvidia to go this. NO ONE in wayland seems to have even got Nvidia to entertain the idea. Seems like a success story to me.
                      No. I highly doubt Canonical 'single-handily' did anything. :\

                      EGL is becoming very popular and this is a natural progression of standards, that nvidia will need to support, regardless. + (as others have told you) their hasn't actually been any clarification directly from Nvidia on their motivations or anything like that...

                      I'd also like some citations on this idea that 'NO ONE' from the Wayland camp has contacted nVidia - so please provide citations, thanks. (I didn't realize that you were privy to every email that is received by nVidia ~ which is what you are claiming, essentially).

                      ..and to clarify, i am in the same camp as people annoyed with the Mir news and canonical, in general. (i've never really liked canonical to begin with).

                      Comment

                      Working...
                      X