X.Org Server Development Hits A Nearly Two Decade Low

[Zan Lynx]

Ok, but games themselves are getting more parallel, and up until recently most laptops had relatively weak GPU's and only two cores. Add to that that most games are designed with the philosophy of "if it's there, consume it", most games are designed to ping the CPU at 100% and they can do that on up to 4 cores and can almost do it on 8 cores. Which is exactly the scenario wayland compositors experience input lag and which is exactly the worst case scenario for games. You can see how on the vast majority of all computers games -do- lag terribly on wayland compositors.

So you have a 24 core processor, your computer may not experience that much lag, but most computers will. Just because you have a high end desktop and don't have any experience with lower end hardware doesn't mean that everyone else fits in your same boat.

OK but I also do builds on a Dell laptop with a 4 core i7 and Intel iGPU and I don't get Wayland lag there either. I actually credit it to Fedora's systemd cgroup setup, which nicely isolates system daemons, user sessions, and individual user applications from each other. Wayland isn't fighting to run in a 1/10th share of the CPU while I compile. It's got a 1/2 share.

Now, I agree that software like a Wayland compositor should be running as a memory locked real-time application that never does disk IO. But gnome-shell doesn't do that. Still, with a nice cgroup setup it gets plenty of CPU time.

[duby229]

OK but I also do builds on a Dell laptop with a 4 core i7 and Intel iGPU and I don't get Wayland lag there either. I actually credit it to Fedora's systemd cgroup setup, which nicely isolates system daemons, user sessions, and individual user applications from each other. Wayland isn't fighting to run in a 1/10th share of the CPU while I compile. It's got a 1/2 share.

Now, I agree that software like a Wayland compositor should be running as a memory locked real-time application that never does disk IO. But gnome-shell doesn't do that. Still, with a nice cgroup setup it gets plenty of CPU time.

I'm not so sure giving wayland half your CPU share is the correct solution, though that'll probably reduce input lag. The correct solution is to extend the protocol so it can sync output on input. But at this stage a decade later, that's not likely to happen.