Announcement

Collapse
No announcement yet.

NVIDIA Wants Feedback On Its Device Memory Allocator Project

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

  • #11
    Originally posted by sebastianlacuesta View Post

    Nvidia wants to help Nouveau... Good... Tell them that proper reclocking is needed in order to get the latest cards working at acceptable speed using nouveau.
    Not just the latest cards. 400 and 500 series need it badly too to be usable, but they don't depend on signed blobs.

    Comment


    • #12
      Originally posted by sebastianlacuesta View Post
      Nvidia wants to help Nouveau...
      In the past, NVidia just ignored requests from Nouveau and then proceeded to implement what was requested in their own open source nvgpu (limited hw support/functionality, targeted at Android) driver without telling anyone... I mean how can you show the finger more obviously to the Nouveau project?


      NVidia needs to get support code into the open source drivers in order for their EGLstreams successor to fly, and so far there is no reason for the community to write the code themselves.

      Comment


      • #13
        Interesting. What is the source and date of that slide?

        Comment


        • #14
          Originally posted by Veto View Post
          Interesting. What is the source and date of that slide?
          If I'm not mistaken it was in the presentation made by nouveau guys at XDC 2017.

          Comment


          • #15
            Originally posted by Veto View Post
            Interesting. What is the source and date of that slide?
            Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
            Michael Larabel
            https://www.michaellarabel.com/

            Comment


            • #16
              i have a theory that nvidia finally realized that doing things their own way is going to be maintenance nightmare for them pretty soon, since graphics stack in linux is increasing in complexity.

              amd has drivers that leverage llvm or other opensource code for various things, although i am not sure how exactly their amdgpu-pro variant exactly works. majority of the code is upstreamed into respective projects, with active maintainers (some of them employed by amd) and that also allows other platforms and operating systems (aros on x86/ppc, android on x86, non-x86 plaforms where one can use amd cards, hobbyist operating systems like kolibri, etc) to use their hardware.

              i think nvidia will have to do what amd did - opensource kernel and xorg part. maybe basic 3d driver and optional proprietary 3d driver with max performance. i don't believe they will ever go fully opening their code, they might simply not be able to do so legally. i am pretty certain the dev team is starting to feel the maintenance burden.

              Comment


              • #17
                Originally posted by schmidtbag View Post
                But now, AMD+Intel are almost fully caught-up with Nvidia (in terms of performance and features), and Wayland is starting to be more widely adopted. Ignoring hardware limitations, we're very close to a point where AMD will be a resounding better option for Linux (regardless of your principles) and where Nvidia will be holding back development to the point where they're resented rather than "begrudgingly accepted".
                Really? Then why did Intel find it necessary to hire the GPU guy from AMD to restart their GPU division? I mean, if Intel was "fully caught-up" then Intel wouldn't be panicking about their lack of competitive GPU for machine learning....

                As for AMD, they can't match the top of Nvidia's line from over a year ago with their latest "best release", so I don't think Nvidia is too worried - unlike Intel who does have competition from Zen.

                Comment


                • #18
                  Typo:

                  Originally posted by phoronix
                  developing a new library that sites besides/on-top-of GBM for in turn doing the low-level graphics buffer allocations.
                  That would be the best path.

                  Also, is it going to work on BSDs/macOS/Solaris/other Unixes? (or does NVIDIA merely hate saying "Linux"?)

                  Comment


                  • #19
                    sites beside
                    Do you mean sits?

                    Comment


                    • #20
                      Originally posted by Niarbeht View Post

                      Do you mean sits?
                      Yeah was fixed a minute ago after tildearrow reported it, thanks.
                      Michael Larabel
                      https://www.michaellarabel.com/

                      Comment

                      Working...
                      X