02:34 <icecream95> Hmm.. this is odd, Darkplaces runs at about 2 fps in the "Introduction" level but everywhere else is fine

02:41 <icecream95> Even odder.. going into the menu makes the frame-rate improve again

05:13 <icecream95> I get the feeling that there's something very wrong with this kernel (5.10.63), even with an idle desktop the kernel uses about 50% CPU on all cores

05:53 <suihkulokki> icecream95: see what perf top says

07:16 <icecream95> The problem seems to be fixed with a newer version of DarkPlaces

07:58 <icecream95> The commit fixing it was "gl_rmain: Perform lightmap updates only on visible surfaces"

08:27 <icecream95> When AFBC is enabled, updating the lightmap involves a number of blits

08:28 <icecream95> If I had BPF working on this machine, I would be able to quickly find out exactly how many updates there are per frame, but I assume that it is "a lot"

09:25 <icecream95> Neverball controlled by duet's accelerometer is fun :)

09:47 <macc24> icecream95: still not using cadmium on it? xD

09:47 <macc24> ,_,

10:02 <icecream95> macc24: It has a broken screen and keyboard, so by now it's just an SBC with a battery

10:03 <icecream95> (And an accelerometer, of course)

10:12 <macc24> ouch

10:27 <icecream95> Even Quake works surprisingly well. Just point the "gun" at enemies and shoot!

10:28 <macc24> icecream95: does external display work on it?

10:29 <icecream95> macc24: Haven't tried, currently I'm sending accelerometer readings over SSH and game is running on an RK3399 board

10:29 <macc24> how the-

10:30 <icecream95> os.system(f"DISPLAY=:0 xdotool mousemove_relative -- {x - curx} {y - cury}")

10:30 <icecream95> I'm planning to switch that to using Wayland protocols natively

10:31 <macc24> i don't think you can

10:31 <macc24> you'd need to directly use uinput

10:32 <icecream95> Oh, just like the VNC server I'm running that is only used for input devices needs uinput? /s

10:32 <icecream95> (The board has no keyboard or mouse attached, so keyboard input comes from speedy)

10:32 <macc24> so

10:33 <macc24> your keyboard is on speedy, "mouse" on krane, and gpu on a rk3399 machine?

10:33 <macc24> damn you're gaming on 3 different computers at once

10:33 <icecream95> The protocol is "wlr_virtual_pointer_unstable_v1"

10:34 <icecream95> macc24: No, more than that

16:07 <ndufresne> alyssa: does this partial renderer and resume mechanism on Apple m1 also imply that a proper premption model can be implemented? Something that prevents unfair use of the GPU time?

18:45 <alyssa> ndufresne: probably, but that's not self-evident / immediate

21:40 <ndufresne> You'd basically need a reliable way to trigger a fault every now and then ...