Bug 111558 - Performance issues on Iris 550/Skylake
Summary: Performance issues on Iris 550/Skylake
Status: RESOLVED MOVED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/Iris (show other bugs)
Version: git
Hardware: Other All
: not set not set
Assignee: Intel 3D Bugs Mailing List
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-09-04 15:09 UTC by Cedric Sodhi
Modified: 2019-09-25 18:47 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Strace as root (119.51 KB, text/plain)
2019-09-05 13:20 UTC, Cedric Sodhi
Details
Sway log as root (6.83 KB, text/plain)
2019-09-05 13:20 UTC, Cedric Sodhi
Details

Description Cedric Sodhi 2019-09-04 15:09:21 UTC
Unfortunally, I don't have much more to report than I'm seing performance issues under Wayland/Sway. Starting the compositor, spawning a terminal, operating in the terminal, all these actions are significantly (usually a couple of seconds) delayed. No crashes occur and nothing is brought up in dmesg.

Mesa was compiled by portage from master@04.09.2019 with:

meson --buildtype plain --wrap-mode nodownload -Dplatforms=surfaceless,wayland,drm -Dllvm=false -Dlmsensors=false -Dlibunwind=false -Dgallium-nine=false -Dgallium-va=false -Dgallium-vdpau=false -Dgallium-xa=false -Dgallium-xvmc=false -Dgallium-opencl=disabled -Dosmesa=none -Dbuild-tests=false -Dglx=disabled -Dshared-glapi=true -Ddri3=true -Degl=true
-Dgbm=true -Dgles1=false -Dgles2=true -Dglvnd=false -Dselinux=false -Dvalgrind=false -Ddri-drivers= -Dgallium-drivers=iris,swrast -Dvulkan-drivers=intel -Dvulkan-overlay-layer=false --buildtype plain -Db_ndebug=true
Comment 1 Tapani Pälli 2019-09-05 11:38:29 UTC
Just as a first thought, this sounds very much like how the experience is when one runs gnome-shell with a software rasterizer. Is there some log output from sway that would prove it is HW accelerated? You could use strace to see which dri module it opens (iris_dri.so, not swrast_dri.so) or for example gdb the process and set break to iris_draw_vbo that should get called on each draw.
Comment 2 Cedric Sodhi 2019-09-05 12:45:25 UTC
True, everything points to SW being used. I will attach output of sway and strace. I had to run it as root, because strace would't get the correct caps, so sway terminates prematurely, but late enough to get to the point.
Comment 3 Cedric Sodhi 2019-09-05 13:20:26 UTC
Created attachment 145268 [details]
Strace as root

There is new evidence, suggesting that with MESA_LOADER_DRIVER_OVERRIDE=iris the driver is actually loaded, but only if run by root (as seen in the attached files). If run as non-root, however, mesa attempts (and fails) to load i965 rather than iris. There might be a permission/caps problem.
Comment 4 Cedric Sodhi 2019-09-05 13:20:54 UTC
Created attachment 145269 [details]
Sway log as root
Comment 5 Tapani Pälli 2019-09-06 11:27:36 UTC
(In reply to Cedric Sodhi from comment #3)
> Created attachment 145268 [details]
> Strace as root
> 
> There is new evidence, suggesting that with MESA_LOADER_DRIVER_OVERRIDE=iris
> the driver is actually loaded, but only if run by root (as seen in the
> attached files). If run as non-root, however, mesa attempts (and fails) to
> load i965 rather than iris. There might be a permission/caps problem.

Oh that is strange :/ Maybe try installing iris as regular user (as example under /home/user/iris, just use --prefix= with meson), then use LD_LIBRARY_PATH so that sway finds it there?
Comment 6 GitLab Migration User 2019-09-25 18:47:46 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/1362.


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.