Announcement

Collapse
No announcement yet.

AMD Has A Vulkan Linux Driver, But Will Be Closed-Source At First

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

  • #21
    Originally posted by Prescience500 View Post
    So...does this mean that the new OpenCL and Vulcan won't be using Gallium and the mesa stack like the current one does? That's what the slide seems to imply...though it may not matter much since no one else seems interesting in using Gallium for OpenCL.
    For Vulkan, Gallium makes no sense because they're both low level GPU abstraction interfaces, and I'd say Vulkan is even lower level than Gallium, which doesn't expose explicit command buffer building and submitting at all.

    Comment


    • #22
      Well, seriously, AMD just confirms Vulkan driver, and everyone just continues program as usual - complain about something This is very very welcome news, also Future in open source - YES! That's news I wanted to hear. Please fix your drivers AMD, and you will be our friend again.

      Comment


      • #23
        Originally posted by Ancurio View Post

        For Vulkan, Gallium makes no sense because they're both low level GPU abstraction interfaces, and I'd say Vulkan is even lower level than Gallium, which doesn't expose explicit command buffer building and submitting at all.

        Comment


        • #24
          Originally posted by haagch View Post
          And I guess that the PRIME synchronization issues will still not be solved then, making it basically unusable on intel+amd systems. Supposedly there should be no problems when the radeon gpu renders "fast enough", but even tesseract (just an example) at 100+fps produces very annoying graphical hickups with DRI3.
          The solution is dma-buf fences, and the amd kernel does support it since quite some time (no idea for amdgpu), problem is it needs to be done intel side too.

          Comment


          • #25
            Originally posted by andre30correia View Post
            easy to understand, big names like nvidia, amd, intel, apple, m$ steal code and use others code and they simply can open the source code without having legal problems
            It works to everyone's benefit, BUT in case they won't attempt to take advantage of other's work. For that open source is not enough, must be totally free (see GPL). Like in science and technology, there are some legal issues with patents in certain field but in others all humanity benefits (we don't pay to use the wheel, right?).

            Comment


            • #26
              Originally posted by mannerov View Post
              The solution is dma-buf fences, and the amd kernel does support it since quite some time (no idea for amdgpu), problem is it needs to be done intel side too.
              Yes, and they know it too... We should send them https://www.youtube.com/watch?v=ZXsQAXx_ao0.

              On the other hand, what nine did with thread_submit works a lot better. Maybe it won't be such a big problem with a new driver.

              Comment


              • #27
                Originally posted by Pecisk View Post
                Well, seriously, AMD just confirms Vulkan driver, and everyone just continues program as usual - complain about something This is very very welcome news, also Future in open source - YES! That's news I wanted to hear. Please fix your drivers AMD, and you will be our friend again.
                I was excited for the news, until I read the initial closed-source note and questioned 7xxx support. So basically, without either of those, I can't say I care about this news for now.

                Comment


                • #28
                  If my Radeon 7950 turns out to be not supported because the outlined approach requires AMDGPU, I'm seriously p*****!

                  Comment


                  • #29
                    Tonga 285 here hopefully I will be alive to see an opensource full power vulkan and openGL 4.5 driver! Cause Catalyst is like our economics here in Greece... A total mess.
                    But please AMD show as some actual stuff, I mean AMDGPU does not have reclocking yet, Catalyst performs awfully on VI cards and you are talking about full Vulkan closed then open?
                    Seems like a midnight summer dream to me... I pray to be wrong here.

                    Comment


                    • #30
                      Originally posted by haagch View Post
                      Yes, and they know it too... We should send them https://www.youtube.com/watch?v=ZXsQAXx_ao0.

                      On the other hand, what nine did with thread_submit works a lot better. Maybe it won't be such a big problem with a new driver.
                      thread_submit is not really adapted to GL because apps are allowed to do more things with its windows and pixmaps, and thread_submit would have wrong behaviours for these cases.

                      Comment

                      Working...
                      X