Announcement

Collapse
No announcement yet.

Zink GL-On-Vulkan Driver Approaching OpenGL 3.1 Support

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

  • Zink GL-On-Vulkan Driver Approaching OpenGL 3.1 Support

    Phoronix: Zink GL-On-Vulkan Driver Approaching OpenGL 3.1 Support

    Zink is the generic OpenGL over Vulkan driver that has been in development as part of Mesa's Gallium3D code. It was just earlier this month that Zink achieved OpenGL 3.0 support and now it looks like OpenGL 3.1 will soon be flipped on...

    http://www.phoronix.com/scan.php?pag...ing-OpenGL-3.1

  • #2
    I, for one, am looking forward to Zink being generally available and usable. I am sure that there are many GPU reverse engineering efforts that will start now by implementing a Vulkan driver to start rendering triangles

    Comment


    • #3
      When Vallium gets upstreamed, I'm expecting a war over the best way round to do things.

      Do you write a Vulkan driver, and get a free GL driver, or write a Gallium (i.e. GL) driver, and get a free Vulkan one?



      I've been looking into Vallium for Panfrost, and have already got vkcube sending GPU jobs without crashing.

      Comment


      • #4
        Originally posted by archsway View Post
        When Vallium gets upstreamed, I'm expecting a war over the best way round to do things.

        Do you write a Vulkan driver, and get a free GL driver, or write a Gallium (i.e. GL) driver, and get a free Vulkan one?



        I've been looking into Vallium for Panfrost, and have already got vkcube sending GPU jobs without crashing.
        Isn't Vallium a Vulkan software renderer? Couldn't you run Zink on top of it for instance?

        Comment


        • #5
          Originally posted by geearf View Post

          Isn't Vallium a Vulkan software renderer? Couldn't you run Zink on top of it for instance?
          Vallium is a generic Gallium frontend so can potentially work with any Gallium driver (most GL drivers in Mesa except for the "classic" ones like i965), but for now the focus is just on making it work with llvmpipe.


          Vallium can be used as both a frontend and (with another driver) backend for Zink, so looping between them could theoretically be done an arbitrary number of times.

          Comment


          • #6
            Originally posted by archsway View Post

            Vallium is a generic Gallium frontend so can potentially work with any Gallium driver (most GL drivers in Mesa except for the "classic" ones like i965), but for now the focus is just on making it work with llvmpipe.


            Vallium can be used as both a frontend and (with another driver) backend for Zink, so looping between them could theoretically be done an arbitrary number of times.
            Oh I didn't realize Vallium was a frontend for drivers. Could it be run on top of radeonsi in the future then?

            Comment


            • #7
              Given that older AMD GPUs tend to work well with Vulkan, is there any chance at all that Zink over RADV will eventually become competitive with the RadeonSI Mesa OpenGL driver?

              If not, what are the biggest hurdles re. performance?

              Comment


              • #8
                Originally posted by archsway View Post
                When Vallium gets upstreamed, I'm expecting a war over the best way round to do things.

                Do you write a Vulkan driver, and get a free GL driver, or write a Gallium (i.e. GL) driver, and get a free Vulkan one?



                I've been looking into Vallium for Panfrost, and have already got vkcube sending GPU jobs without crashing.
                Isn't Vulkan much easier to implement than GL? Kind one of it's biggest selling points: less cruft, low abstractions, closer to the metal. Harder to get wrong.

                Comment


                • #9
                  Originally posted by archsway View Post
                  Do you write a Vulkan driver, and get a free GL driver, or write a Gallium (i.e. GL) driver, and get a free Vulkan one?
                  obviously vulkan backend for gallium will be faster than gallium backend for vulkan. but two separate drivers will be even faster. and most of the code will be shared anyway, so you'll get second driver almost for free
                  Last edited by pal666; 07-01-2020, 02:43 PM.

                  Comment


                  • #10
                    Originally posted by flashmozzg View Post
                    Isn't Vulkan much easier to implement than GL? Kind one of it's biggest selling points: less cruft, low abstractions, closer to the metal. Harder to get wrong.
                    opengl is already implemented. what you have to implement is gallium backend

                    Comment

                    Working...
                    X