Bug 96783

Summary: Intel 945 GM is very slow on rendering with 4.5 kernel
Product: Mesa Reporter: Michal Nowak <mnowak>
Component: Drivers/DRI/i915Assignee: Default DRI bug account <dri-devel>
Status: RESOLVED MOVED QA Contact: Default DRI bug account <dri-devel>
Severity: normal    
Priority: medium CC: grawity, intel-gfx-bugs, leho
Version: unspecified   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: I945GM i915 features: display/Other
Attachments: journalctl with drm.debug=14
xrandr
lspci
journal

Description Michal Nowak 2016-07-02 15:11:01 UTC
Created attachment 124860 [details]
journalctl with drm.debug=14

I am running Fedora 24 on Dell Inspiron 6400 a 32-bit CPU with Intel Corporation Mobile 945GM/GMS graphics from GM installation media with 4.5.5 kernel.

After login to GNOME Shell/X11 the gnome-shell process takes 50-100 % of CPU, rendering is slow. If I run `xrandr --output TV1 --off` as suggested here: https://bugzilla.redhat.com/show_bug.cgi?id=1350133 the CPU utilisation drops but rendering a 720p video on YouTube in Firefox is impossible, frames are dropped CPU utilisation spikes to 180 % for firefox process. If I run a 144p video, it renders fine but CPU utilisation is at 70 %.

xorg-x11-drv-intel-2.99.917-23.20160512.fc24.i686
kernel-4.5.5-300.fc24.i686
xorg-x11-server-Xorg-1.18.3-2.fc24.i686

Graphics used to be OK on Fedora 22 with all updates (a 4.4.14 kernel).

Logs attached.
Comment 1 Michal Nowak 2016-07-02 15:11:22 UTC
Created attachment 124861 [details]
xrandr
Comment 2 Michal Nowak 2016-07-02 15:11:40 UTC
Created attachment 124862 [details]
lspci
Comment 3 Michal Nowak 2016-07-02 15:14:04 UTC
Similar bug: https://bugs.freedesktop.org/show_bug.cgi?id=93263.
Comment 4 Michal Nowak 2016-07-02 15:21:41 UTC
Fedora 22 packages:

kernel-4.4.13-200.fc22.i686
xorg-x11-server-common-1.17.4-1.fc22.i686
xorg-x11-drv-intel-2.99.917-15.20150729.fc22.i686
Comment 5 Chris Wilson 2016-07-02 15:47:44 UTC
Adding to xorg.conf

Section "Device"
  Identifier "igfx"
  Option "DRI" "2"
EndSection

should fix the gnome-shell issue (due to fedora's settings). Looks like mesa + dri3 is inoperable on 945gm.

Similarly look at disabling "hardware acceleration" in firefox? Not sure what firefox is doing with youtube, but if it is suffering from GL swrast that will help.
Comment 6 Michal Nowak 2016-07-03 11:54:35 UTC
(In reply to Chris Wilson from comment #5)
> Adding to xorg.conf
> 
> Section "Device"
>   Identifier "igfx"
>   Option "DRI" "2"
> EndSection
> 
> should fix the gnome-shell issue (due to fedora's settings). Looks like mesa
> + dri3 is inoperable on 945gm.
> 
> Similarly look at disabling "hardware acceleration" in firefox? Not sure
> what firefox is doing with youtube, but if it is suffering from GL swrast
> that will help.

I noticed that 720p videos run quite poorly even on F22 but your config snippet helps with greatly with Totem. With DRI3 (the F24 default) glamor can't be enabled and gstreamer renders via slow swrast.
Comment 7 Jari Tahvanainen 2017-04-24 13:03:42 UTC
Hello Michal, it seem so that this bug has been neglected wayyyyy too long, I apologize that. Can you please retest with the latest kernel (preferable from drm-tip) and X-components, and change status to REOPENED if problem still persist.
Chris - is the component (DRM/Intel) right place for this? You said in comment 5 that " mesa + dri3 is inoperable on 945gm."
Comment 8 Chris Wilson 2017-04-24 13:09:50 UTC
As far as I know, this is just a mesa bug. I left it here as I don't know what gave the impression that the kernel was a critical factor in the bug.
Comment 9 Michal Nowak 2017-04-24 20:46:40 UTC
Created attachment 131005 [details]
journal

Running Fedora 25 with all updates and the original problem with high utilisation after login is gone (DRI2 seems to be enable by default in Wayland) but YouTube video playback performance in Firefox is still the same, barely plays 360p.

kernel-4.10.10-200.fc25.i686
mesa 13.0.4-3.fc25.i686
wayland 1.12.0-1.fc25.i686

Just for the record: I don't use the system any more, just for testing from time to time.
Comment 10 Ricardo 2017-07-20 22:35:14 UTC
Chris should this bug be moved to Mesa?
Comment 11 Ricardo 2017-08-15 16:24:17 UTC
this issue seems to be out side of the i915 and should be more involved with Mesa. Closing issue also for the lack of response...
Comment 12 Chris Wilson 2017-08-15 16:32:05 UTC
The beauty of having a unified bug tracker is that if you think there is no kernel aspect to the bug, you can just reassign it to the mesa component. And vice versa.
Comment 13 GitLab Migration User 2019-09-18 19:39:37 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/765.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.