No pageflipping on Ubuntu 24.04-LTS with Ubuntu/Gnome desktop GUI

I have a newly built PC with an Intel chip and integrated GPU (i7-13700K) running PTB (3.0.19) on Octave (8.4.0) on Ubuntu (24.04) with the low-latency kernel and a single monitor. Totally fresh install of OS and everything else, and VBLSyncTest hasn’t missed a single flip across many tests. However, VBLSyncTest reports that flip didn’t use pageflipping, for each of 667 flips. Same warnings are thrown by other calls to flip the screen. This continues to happen when I run it repeatedly, and through reboots. PC has 16G of RAM, and notifications are disabled.

In trying to find the problem I ran XOrgConfCreator, thinking maybe a good X config might help. However, it reported that “This seems to be a hybrid graphics laptop…” and explained all the pitfalls with that, and concludes that I have no need for an X config. But it’s a desktop PC with integrated Intel GPU! So, I thought this was worth sharing before I try a different windowing system or just put a graphics card in there.

Some details below. Thanks for your thoughts!
Mike

Output of XOrgConfCreator

octave:1> XOrgConfCreator

Detected X-Server version 21.1.11
Creating configuration for a modern X-Server 21+, assuming Mesa 22+ and Linux 5.15+ …
Detecting type of primary display graphics card (GPU) and driver to use…
Intel GPU detected. Primary display gpu type: intel
Uses the xf86-video-modesetting DDX video driver.
Found a total of 1 video output displays on 1 X-Screens.

This seems to be a hybrid graphics laptop.
[truncated]

Output of lshw

*-display
description: VGA compatible controller
product: Raptor Lake-S GT1 [UHD Graphics 770]
vendor: Intel Corporation
physical id: 2
bus info: pci@0000:00:02.0
logical name: /dev/fb0
version: 04
width: 64 bits
clock: 33MHz
capabilities: vga_controller bus_master cap_list rom fb
configuration: depth=32 driver=i915 latency=0 resolution=1920,1080

I don’t see errors in OpenWindow, just the following:

PTB-INFO: Will try to use OS-Builtin OpenML sync control support for accurate Flip timestamping.
PTB-INFO: Measured monitor refresh interval from VBLsync = 16.666660 ms [60.000023 Hz]. (50 valid samples taken, stddev=0.001412 ms.)
PTB-INFO: Reported monitor refresh interval from operating system = 16.666667 ms [60.000000 Hz].

Try to click on the notification area in the top-right corner of your Ubuntu desktop GUI, where the various controls, battery / network etc. indicators are.

I think GNOME desktop, and thereby its derivative the standard Ubuntu desktop, has a bug on Ubuntu 24.04-LTS that prevents page-flipping, unless you click there, which somehow magically fixes the problem until the next logout + login.

Or install a different desktop GUI like KDE and see how that goes. I wouldn’t know, but assuming a GNOME bug, it would help.

The lack of financial support by the vast majority of over 99% of our users and the resulting almost catastrophic lack of funding meant that I only could spend less than 10 hours of very light touch and go testing of Ubuntu 24.04-LTS on one single loaned laptop so far, sometime in April. But I also noticed that problem immediately with the Ubuntu GUI - the only GUI installed on that machine which had AMD graphcs. Ofc. lack of funding also means there ain’t no time to fix any found bugs for the foreseeable future, and by now this bug is likely unfixable for the whole Ubuntu 24.04 cycle with GNOME anyway. I don’t think there will be time to test Ubuntu 24.04-LTS anytime soon. Right now I don’t even have one machine to test it, let alone the time. In the past, weeks or months of testing and improvement work went into making sure that Ubuntu LTS releases are top notch for neuroscience applications, not this year though.

Good luck,
-mario

Thank you Mario,

I can confirm on multiple machines that clicking the notification area, even once, seems to fix the issue. I wonder if this is related to the slew of bug reports (e.g. 3134) suggesting that full screen windows aren’t being “unredirected” (as in they’re still being pushed through the compositor). I haven’t found any easy fixes for that bug, but it’s killing frame rates for games so will hopefully be fixed soon. That said, it sounds like they’re really focusing on Wayland for gnome.

I’ll just revert to 22.04, anything else is too risky. But I’ll keep an eye on the compositor issue and test 24.04 again in the future.

Best,
Mike

Good.

I had a look at that issue as well a couple of months ago, and it is unclear that would cause it. That issue suggests some applications aren’t affected, and nothing about a simple mouse click somewhere fixing it. Could be involved, or not. No time to investigate anytime soon.

Wrt. Wayland, yes. All major desktop GUI’s and distributions like Fedora or Ubuntu are moving to Wayland by default, with an increasingly strong push to “Wayland only” at some future point in time. Native XOrg X-Server is more or less in a more or less maintenance mode, only kept alive by a few volunteers like myself. X-Server 21 and all the improvements it had in store for neuroscience, starting with Ubuntu 22.04-LTS and later, was basically only volunteer work by myself (multiple unpaid months of work) and a couple of others. Due to the lack of funding by our users, I don’t have any meaningful time to contribute to or test the development branch of it atm., so bugs creeping in over time is unavoidable. The server however is very stable and bug-free for neuroscience purposes. The big worry are desktop environments like GNOME and KDE which are no longer very motivated to test and improve their XOrg paths, as evident in this specific Ubuntu 24.04 bug on Ubuntu/GNOME desktop. E.g., “Fedora 40 Asahi Linux remix” for use on Apple Silicon Mac machines already dropped any meaningful support for XOrg, and the XOrg X-Server on those machines is already completely broken and unusable.

Wayland itself is not yet ready for more complex or demanding vision science / neuroscience applications, although some minor improvements have happened throughout the last 2 years. Neither is Psychtoolbox ready, apart from some experimental backend for basic use cases - and other toolkits are not even aware of the ice-bergs ahead, nor are their developers even remotely equipped to deal with them if they even were aware. The severe lack of funding ofc. means I don’t and didn’t have the time to deal with the evolving Wayland situation appropriately over the last few years either, only squeezing out a few days of work here and there, by far not an appropriate amount.

Switching to another non-GNOME shell / Mutter powered desktop GUI like KDE should also work, although that is also completely untested by myself.