My opinion is different. The timing of eglSwapBuffers IS consistent under GNOME. It is ~20ms or so which is a good number for about 1/60 FPS playback with HDR or tonemapping enabled. Kwin seems to batch process/queue the OpenGL drawing, as that one IS indeed inconsistant (it sometimes returns immediately and sometimes takes longer).
I've ruled out the FLTK toolkit as: - Same binary under GNOME49 did not exhibit the issue. - Same binary under Kwin does not exhibit the issue. I've added debugging of Vsync into my application, just in case and while I did find a bug in Vsync handling in fullscreen mode, that's not the issue. To be quite honest, I've been performing benchmarks under GNOME 50.1 and Kwin 6.6.4 both under OpenGL and Vulkan for my application and GNOME performs badly in comparison (unlike previous releases, which was the opposite). Kwin works flawlessly without a problem now barely skipping frames under 4K with a 60 FPS playback rythm under any backend. GNOME has issues keeping the frame rate and it also suffers more when turning HDR on or fractional scaling. As it stands now, while I like GNOME's minimalistic approach better as my WM, I will have to start recommending Kwin/Plasma for my application. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2153492 Title: Regression in OpenGL Wayland performance (not NVidia issue) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+bug/2153492/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
