Announcement

Collapse
No announcement yet.

Microsoft Working On Direct3D 12 Video Acceleration For Mesa

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

  • #51
    Originally posted by smitty3268 View Post

    Well, yeah, people obviously can complain about anything. But i don't find it valid, personally. It came out first, and at the time Vulkan was seemingly mired in quicksand and no one knew if it would be successful or not.


    I didn't know that Valve was such a important force behind Vulkan, but well so they released a early working version in the beginning of 2015, the full final specs were released and they waited till release at dez. 2016 till drivers implemented it.

    DX12 materialized July 29 2015 with Windows 10, well maybe you could argue that it was before in some beta programs, but I think half a year for a standardisation of a industry standard is that really long time? And worth having 2 different standards and therefor reinventing the wheel incompatibility and so on? Well only if you want to use it to keep your market dominating position.

    And why would you suggest it would be in quicksand Chronos did well with Opengl kept that standard for ever... so there is no base in doubting them for doing the same. Now opengl was maybe a less ambitious project and it was probably more a different aproach of the same problem but direct3d and vulkan are both as far as I understand basically 1:1 Mantle clones, so both are the same kind of wheel that got invented twice. Heck what would have stopped them from pushing vulkan and make vulkan a part of directx, so that you have sdl + vulcan vs DX/Vulkan?

    Also Opengl was released many years before DX so this who was first argument seemed to not mattered back then either.
    Last edited by blackiwid; 22 November 2021, 09:13 PM.

    Comment


    • #52
      Originally posted by geearf View Post
      Why does Mesa accept these contributions? What's the benefit of accepting something that does not benefit a single nix OS?
      because it is open source, mesa also has drivers for windows too. mesa is not solely a *nix project. not to mention it is beneficial, like it or not, wsl2 is just as good as getting people to use linux, even as a vm, as it is at stopping people from swapping. many people have been introduced to an actual linux enviroment through wsl2, instead of using the god awful mingw or msys enviroments.

      Comment


      • #53
        We're talking about Microsoft here. If they are contributing, it's for their benefit. That, or to sneak and destroy.

        Comment


        • #54
          Originally posted by Monsterovich View Post
          Can someone please explain why does D3D12 even exist? I know that it was released earlier than Vulkan. But what's a benefit from using it instead of Vulkan? Or it's just an instrument of Microsoft supremacy?

          Vulkan works everywhere, and literally any graphical API was implemented on top of it.

          D3D9 -> dxvk (formelly d9vk) -> Vulkan
          D3D10-11 -> dxvk -> Vulkan
          D3D12 -> vk3d -> Vulkan
          OpenGL -> Zink -> Vulkan
          D3D12 on Xbox One family. Windows 10 and 11. Since this family together makes the largest market for games on earth, and Vulkan is essentially a response to the early prototypes of D3D12 on Xbox One, we wouldn't have Vulkan without it.

          It is a supported API, meaning if you have issues there is one place you can contact and get support and fixes, as well as additions. Vulkan isn't native to gaming consoles, of which most games target, so it increases the complexity of 2 or 3 APIs to one more.

          There are lots of other reasons people may give as well, but they come down to support and numbers.

          Comment


          • #55
            Originally posted by smitty3268 View Post

            Well, yeah, people obviously can complain about anything. But i don't find it valid, personally. It came out first, and at the time Vulkan was seemingly mired in quicksand and no one knew if it would be successful or not.

            It would be nice if companies were altruistic and wanted to help out the linux community, but i don't hold them to that high a standard. They're always going to look out for themselves first.
            Vulkan (or something like Vulkan) would have always succeed because it solved a real problem that existed in the game engine space, i.e. all other API's being high level black box single threaded implementations with design woes inherited from decades ago and filled with lots of quirks.

            Vulkan solved all of these issues and this is why both game developers and creators of game engines love it. OpenGL/ DX11 or lower was also one of the primary reasons why game engines didn't usually end up using more than 4 cores because it was ridiculously difficult to multithread the rendering pipeline in those API's. If there was any multithread/multicore being done, it was other parts of the game (i.e. AI/physics).

            Comment


            • #56
              Originally posted by pete910 View Post

              Thats surprised me TBH, I have looked at a few games this year and all have been DX so not sure how up to date that is.
              High chance that those games you looked at where DX11 and not DX12. Release as DX12 and you lock out all those Windows 7 users so most release as DX11, having both DX11 and DX12 is a huge investment so there you really have to get a benefit from DX12 in order to make it worth it.

              Comment


              • #57
                Originally posted by F.Ultra View Post

                High chance that those games you looked at where DX11 and not DX12. Release as DX12 and you lock out all those Windows 7 users so most release as DX11, having both DX11 and DX12 is a huge investment so there you really have to get a benefit from DX12 in order to make it worth it.
                I think MS solved it by providing a DX12 on Win7 library for developers (mostly likely it's not widely available but may be developers get it on demand). I know CP2077 ships such.

                Comment


                • #58
                  Originally posted by mdedetrich View Post
                  Vulkan (or something like Vulkan) would have always succeed because it solved a real problem that existed in the game engine space, i.e. all other API's being high level black box single threaded implementations with design woes inherited from decades ago and filled with lots of quirks.
                  I think it would have failed if Valve had decided to abandon Linux. They were the driving force to actually get something out to replace OpenGL. OpenGL 4 was supposed to fix a bunch of issues but then that got abandoned and watered down to just another set of extensions on top of old GL because nobody cared enough about fixing the API. Companies just wanted something backwards compatible that would keep running their old CAD applications.

                  And on non-linux, there was D3D12 out for game engines to use.
                  Last edited by smitty3268; 23 November 2021, 08:21 PM.

                  Comment


                  • #59
                    More interested in LSW...

                    Comment


                    • #60
                      Originally posted by smitty3268 View Post

                      I think it would have failed if Valve had decided to abandon Linux. They were the driving force to actually get something out to replace OpenGL. OpenGL 4 was supposed to fix a bunch of issues but then that got abandoned and watered down to just another set of extensions on top of old GL because nobody cared enough about fixing the API. Companies just wanted something backwards compatible that would keep running their old CAD applications.

                      And on non-linux, there was D3D12 out for game engines to use.
                      Firstly its not just Valve, there were also heavyweights like John Carmack

                      Secondly even the original OpenGL 4 didn't come close to fixing everything that Vulkan. The bigger problems with OpenGL are fundamental to its design (i.e state being global and hidden is a fundamental part of their API) so any kind of solution would have required a new API anyways or a complete rewrite.

                      Its not surprising that OpenGL instead opted for more incremental changes instead but thats indicative of the general problem. Also the existence of Metal has the exact same story, Apple was fed up with OpenGL and all of the issues it had and for obvious reasons couldn't just implement DSD12.

                      Comment

                      Working...
                      X