Announcement

Collapse
No announcement yet.

Wine-Staging 1.9.21 Improves Its Vulkan Wrapper

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

  • #11
    I'm compiling wine-gaming-nine right now to go test some stuff. @Pontostroy Can you include some DX11 tests too? While compiling, I saw a flag for patching DX11 support in and maybe its worth a try to have a look into them. Thx for testing btw!

    Comment


    • #12
      Are there any news about getting DOOM (2016) up and running? ATM the Denuvo copy protection did make the game impossible to run on linux.

      Comment


      • #13
        Originally posted by Pontostroy View Post

        i did some test with radv and amdhpu-pro and wine vulkan and native vulkan are very close,


        http://www.gearsongallium.com/?p=3405
        Thanks so much for your efforts. I've started to use gearsongallium recently, I find it extremely useful for my curiosity!

        Comment


        • #14
          fresh linux\wine\windows benchmark in talos
          API Windows Linux Wine
          DX11 65.4
          OpenGL 41.5 36.8 33.7
          Vulkan 60.1 46.4 45.7
          DX9 58.6 28.8
          DX9-csmt 29.3
          DX9-gallium-nine 50.2

          Comment


          • #15
            So vulkan is slower than opengl.. even though its supposed to have less overhead? And even slower an Linux :/

            Comment


            • #16
              Originally posted by cj.wijtmans View Post
              So vulkan is slower than opengl.. even though its supposed to have less overhead? And even slower an Linux :/
              You're comparing an optimized engine with optimized drivers to a new experimental test engine. Remember Vulkan shifts the optimization from the drivers to the engine developers. So you can't really expect an experiment for developers to learn how to use vulkan to immediately outperform mature drivers that have been highly optimized.

              Comment


              • #17
                Originally posted by slacka View Post

                You're comparing an optimized engine with optimized drivers to a new experimental test engine. Remember Vulkan shifts the optimization from the drivers to the engine developers. So you can't really expect an experiment for developers to learn how to use vulkan to immediately outperform mature drivers that have been highly optimized.
                What makes you think developers can do it at all? Especially when the driver devs know the running hardware better.

                Comment


                • #18
                  Originally posted by starshipeleven View Post
                  Apart from the fact that you are converting system calls win->linux and dx->opengl (vulkan is basically passthrough apart for windows-specific extensions).

                  Totally not going to impact performance to add a layer of redirection and conversion.
                  There's not even any layer of redirection and conversion.

                  On windows, the app calls into a win32 library that natively implements the calls.

                  On linux, the app calls into a wine library that natively implements the calls.

                  The only time there should be conversions is when the low-level underlying calls are different (such as system calls, OpenGL, etc.) With Vulkan none of that applies, it's just the same thing.

                  Comment


                  • #19
                    Originally posted by cj.wijtmans View Post

                    What makes you think developers can do it at all? Especially when the driver devs know the running hardware better.
                    It's not me it's developers like Johan Andersson and Tim Sweeney who say they can push their engines more with Vukan then Nvidia/AMD driver devs can after the fact. Not to mention that the Developer Relation Programs aren't going away.

                    Comment


                    • #20
                      Thank you for putting in the DX11 as anchor-point.

                      I think its a very good result for the early stage we are in. I'm using as a second game Mechwarrior online under Linux. The fact, that DX9-gallium nine is slower in Talos than the Windows DX9 is one of the more interesting datapoints. I MWO, gallium-nine (Linux) runs faster with than the Windows version with both DX9 and DX11. So here we see that its the other way around. So its game-dependent (who had guessed, right?)

                      MWO-test 64-bit (FPS by eye averaging the FPS-counter - yes I know its rather rough)
                      [Arch Linux - latest stable
                      Radeon HD 7970 - latest stable mesa (non git)
                      32 GB RAM
                      Intel Core I7 4930K @4.2 GHz]
                      Windows 10 aver. FPS Linux (Wine gaming nine) aver. FPS
                      DX 9 80 - 100 DX 9 csmt 25-40
                      DX 11 110 - 130 DX 9 gallium nine 130 - 150
                      MWO is very CPU demanding and not very optimized, thus we should see very good improvements by reducing overhead.

                      The next thing: DX11 is still the best result of all in Talos principle, while in MWO DX9 gallium-nine already beats it with the current mesa drivers. While its still a very rough comparison (with different mesa drivers), I get the impression, that the game-engine is the primary source of possible performance gains by implementing Vulkan.

                      Comment

                      Working...
                      X