Announcement

Collapse
No announcement yet.

GTK4's Vulkan Renderer Is Close To Complete

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

  • GTK4's Vulkan Renderer Is Close To Complete

    Phoronix: GTK4's Vulkan Renderer Is Close To Complete

    Red Hat's Matthias Clasen has written a blog post concerning the changes found in the big GTK+ 3.92 development release that is pushing towards the GTK4 tool-kit release...

    http://www.phoronix.com/scan.php?pag...se-To-Complete

  • #2
    I would be glad if anyone knowledgeable could help me out on this: what is the benefit of this compared to the opengl renderer? more performance, stability,...? thank you!

    Comment


    • #3
      Originally posted by froyo View Post
      I would be glad if anyone knowledgeable could help me out on this: what is the benefit of this compared to the opengl renderer? more performance, stability,...? thank you!
      Much better performance. The ogl render for GTK4 is actually really shit, the cairo backend performs better, due to them focusing on the Vulkan render to start with. Plan seems to be to finish the Vulkan render and then improve the ogl renderer by backporting the stuff they can. Once that happens we shall be able to see if there is enough of a worth while improvement

      Comment


      • #4
        Originally posted by froyo View Post
        I would be glad if anyone knowledgeable could help me out on this: what is the benefit of this compared to the opengl renderer? more performance, stability,...? thank you!
        I'm not an expert, developer or whatever, so my opinion is just that, therefore, take it with a grain of salt:​​​​​​
        The benefit of Vulkan is supposedly better performance. I say supposedly, because we're talking about something where performance shouldn't even be an issue.
        Other benefits could be the ability of rewriting the code cleaner, more modern, and taking advantage of Vulkan's features that otherwise might not be possible.

        Bottom line, the end user might not notice any difference.
        That i said, i still like this, am very excited about it, and hope to see it enabled by default ASAP!!!

        Comment


        • #5
          This will be awsome when XFCE gets gtk4 support

          Comment


          • #6
            Originally posted by Kendji View Post
            This will be awesome when XFCE gets gtk4 support
            That won’t happen in less than 5 years, I’m afraid. That’s why I’m enjoying Gnome, it’s cool to use something that changes at a reasonably fast pace, and I don’t just mean the looks. I was 25 when I started using Xfce, I’m turning 38 in a few weeks. I mean, Xfce hasn’t changed that much in the meanwhile, and many of the very old themes still work.. don’t get me wrong, for somebody else that’s totally awesome, but for me.. life is too short to wait ages for small incremental upgrades, I believe. Also KDE/Plasma is getting exciting.. but for reasons I can’t explain, I can’t get myself to like it, and I mean purely aesthetically this time.
            Last edited by GdeR; 10-23-2017, 11:55 AM.

            Comment


            • #7
              Originally posted by froyo View Post
              I would be glad if anyone knowledgeable could help me out on this: what is the benefit of this compared to the opengl renderer? more performance, stability,...? thank you!
              Currently, the fastest renderer for GTK is the Cairo renderer. It is difficult to implement an OpenGL renderer which can compete with the Cairo (CPU) renderer in terms of raw performance. Often the overhead involved in an OpenGL backend defeats the purpose by adding more latency (worse user experience) to operations, or using almost as much CPU power as a CPU-based renderer (defeating the main purpose: power consumption). One major benefit of Vulkan is lower basic overhead at the outset, which means that it may finally have better latency (better UX) than the CPU renderer, and take considerably less CPU time (usually leading to better power consumption). You can get close with OpenGL, but it involves almost the same amount of effort. Another major benefit of using Vulkan is that it has straightforward windowing system integration, and the drivers are likely to be more consistent and stable than OpenGL drivers (by virtue of being much simpler).

              All that being said, the major factor driving the quality of this renderer is Matthias having the will, patience, and intellect to implement it. Vulkan is more the cherry on top than the Grund zu sein. :- )
              Last edited by microcode; 10-23-2017, 12:10 PM.

              Comment


              • #8
                Originally posted by froyo View Post
                I would be glad if anyone knowledgeable could help me out on this: what is the benefit of this compared to the opengl renderer? more performance, stability,...? thank you!
                The OpenGL renderer was the first implementation, but it falls back to Cairo for a lot of operations because that was the baseline. While the GSK API for rendering widgets was being developed, Benjamin Otte started writing the Vulkan renderer; he implemented a bunch of operations on it to avoid Cairo fallbacks, and Matthias has been working on things like texture atlasing and glyph caches, as well as improvements in the shaders used to implement the rendering operations. Vulkan gives us a better idea of where caching should live, and how to emit geometry and operations in the most efficient way, because we don't have to fight the OpenGL state machine.

                Ideally, once the Vulkan renderer is complete, we'll have a baseline for the GPU-based rendering, and we'll port the OpenGL renderer to the new architecture, for platforms that do not support Vulkan.

                Comment


                • #9
                  I hope it will be smooth to use GTK4 applications.

                  Look at macOS how smooth it is. Everything is animated with smooth animations. Very smooth. It makes GTK, Qt and Windows look very jerky.

                  Comment


                  • #10
                    People in Linux world need to realize that macOS [OS X/NeXTSTEP] the WindowServer and Display Postscript were highly optimized and worked with no middleware. We ported it, dropped DPS due to Adobe's retarded $10/unit fee and created Display PDF; an end around seeing as us at NeXT co-developed DPS with them. When Quart/Quartz Extreme came along and later Metal API all communicates from top to bottom with as seamless a messaging system as possible, and as low level as possible.

                    With Vulkan you'd be an idiot to build a middle layer between Wayland and Vulkan and then add GNOME/GTK+ on top of that. Cut the fat, unify the messaging system and move as much of the processing to the GPU as you can. Cairo won't compete. It will lose and badly once Wayland/Vulkan/GTK+4 are mature and on the same page.

                    The backport for OpenGL seems like a legacy target to transition everyone later to Vulkan, and wisely so. That includes Blender who will benefit greatly and allow them to make the port to OS X with Metal much more coherent and less redundant coding.

                    Comment

                    Working...
                    X