Announcement

Collapse
No announcement yet.

NVIDIA's New Memory Allocator Project To Be Standalone, Undecided On Name

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

  • #11
    libgendevbuf

    Comment


    • #12
      libspecialsnowflake

      more serious now

      libvidgbuff
      Last edited by WorBlux; 30 November 2017, 07:52 PM.

      Comment


      • #13
        Originally posted by FishPls View Post
        standardized solutions such as this must happen.
        Ummm... like GBM?

        Comment


        • #14
          Anyone knows how this one compares against GBM? Why couldn't nVidia simply implement GBM? They must have a good reason for proposing another open source library instead of simply using GBM.

          Comment


          • #15
            Originally posted by sarmad View Post
            Anyone knows how this one compares against GBM? Why couldn't nVidia simply implement GBM? They must have a good reason for proposing another open source library instead of simply using GBM.
            GBM's API didn't match the semantics of their hardware for optimal performance and IIRC other upstream developers also acknowledged improvements could be made to GBM / new API.
            Michael Larabel
            https://www.michaellarabel.com/

            Comment


            • #16
              Why is allocating API so hard?

              Comment


              • #17
                Is it generic right? So it's nVidia allocator serve any purpose.
                Let's call it libnalloc and call it a day.

                Comment


                • #18
                  Originally posted by FishPls View Post
                  Good. Even though there may be lots of people here who hate Nvidia for various reasons (and I'm not saying they're wrong), but for the overall benefit of the desktop Linux ecosystem, standardized solutions such as this must happen. Nvidia handled this horribly in the beginning, but it looks like this time this might end up working out for everyone.

                  So, sure, feel free to hate on Nvidia, just give them the props they deserve when appropriate.
                  Well people would care a lot more if they actually put the work into the open source stack for their own hardware, actually maintain this instead of just making it work well with their proprietary driver like they usually do with things.....etc.

                  Comment


                  • #19
                    Originally posted by Michael View Post

                    GBM's API didn't match the semantics of their hardware for optimal performance and IIRC other upstream developers also acknowledged improvements could be made to GBM / new API.
                    In that case, why not using the existing code base from GBM to submit the improvement rather than reinvented the wheel? It is one of biggest complain for the lack of active cooperation to the existing community and other companies.

                    Comment


                    • #20
                      Originally posted by DanL View Post

                      I ran out of breath reading that, and then I had to read it again because it didn't make sense at first. "more on their next steps is being formulated" seems like awkward language to me.
                      Agreed, it was a bit difficult to read, the first comma was way further than I expected and I also was confused when I read what was after.

                      Comment

                      Working...
                      X