[Bug 254240] xorg/drm-kmod/modesetting driver: XResetScreenSaver() very slow

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=254240

            Bug ID: 254240
           Summary: xorg/drm-kmod/modesetting driver: XResetScreenSaver()
                    very slow
           Product: Base System
           Version: 12.2-RELEASE
          Hardware: amd64
                OS: Any
            Status: New
          Severity: Affects Some People
          Priority: ---
         Component: kern
          Assignee: bugs at FreeBSD.org
          Reporter: tphilipp at potion-studios.com

I apologize for filing this bug here, as I don't know if this is the correct
place. However, given the issue at hand I have a hard time even figuring out
who to report it to:

After an upgrade to 12.2, I noticed heavy stutters in a game I work on. I
traced them down to xlib's XResetScreenSaver() function, which the game loop
calls on input events (so fairly regularly, often every frame), to avoid the
screen saver from triggering. Measuring this, calls to this function sometimes
take sometimes even up to a second to complete, nearly always at least 40-50ms,
though, which for a real time game creates a heavy stutter.

system info:

- uname -a: FreeBSD x210 12.2-RELEASE-p4 FreeBSD 12.2-RELEASE-p4 GENERIC  amd64
- GPU in question is Intel UHD graphics 620 (KBL GT2)
- drm-fbsd12.0-kmod-4.16.g20201016_1
- gpu-firmware-kmod-g20201213
- libdrm-2.4.104,1
- my application is using OpenGL 4.6 via mesa 20.2.3, however given the xlib
call I think this is unrelated to mesa

Trying to isolate the issue, here's what I found:

- this *only* happens on FreeBSD 12.2-RELEASE (and not on 12.1-RELEASE, haven't
tested 13-CURRENT, though)
- this *only* happens with the modesetting driver (w/ AccelMethod "glamor",
haven't tested "none" b/c the game wouldn't be accelerated to begin with)
- with the intel driver (using either "sna" or "uxa" as AccelMethod), this does
*not* happen

I decided to report this here, for now, b/c of this behaviour only being
present on 12.2. Please let me know if this should go somewhere else, e.g. the
xorg people, etc.. Also, sorry for the brevity, I'm a bit lost where to and
what to report, let me know if I can debug this further, what logs might be
helpful, etc..

Thanks!

-- 
You are receiving this mail because:
You are the assignee for the bug.