Bug 89228 - Regression with left for dead 2
Summary: Regression with left for dead 2
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/radeonsi (show other bugs)
Version: git
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact: Default DRI bug account
Depends on:
Reported: 2015-02-19 14:12 UTC by kdj0c
Modified: 2015-08-14 21:52 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Description kdj0c 2015-02-19 14:12:55 UTC
L4d2 used to work with mesa 10.3, I now have graphical glitches, and it ends up in GPU hang after ~5min of game.

might be related to https://bugs.freedesktop.org/show_bug.cgi?id=88561

Hardware is Radeon HD7950

llvm 3.7svn, Linux 3.19.0-g18320f2

OpenGL vendor string: X.Org
OpenGL renderer string: Gallium 0.4 on AMD TAHITI
OpenGL core profile version string: 3.3 (Core Profile) Mesa 10.6.0-devel (git-9216348)
OpenGL core profile shading language version string: 3.30

I'm uploading an apitrace, will add a comment with the link when it's done.
Comment 1 kdj0c 2015-02-19 17:22:55 UTC
apitrace available at

Comment 2 Daniel Scharrer 2015-02-19 19:54:17 UTC
Did you try the patch form 88561 comment 6. I still see some glitches in your trace with that patch, although it may be worse without it (didn't test). Didn't get any GPU lockups or errors in dmesg with the trace, even after running it a couple of times.

GPU: Radeon HD 7950 (OpenGL renderer string: Gallium 0.4 on AMD TAHITI)
Mesa git-6c34fd2 + "radeonsi: don't use SQC_CACHES to flush ICACHE and KCACHE on SI"
LLVM r229671
Kernel 3.19.0-gentoo
Comment 3 kdj0c 2015-02-21 15:00:03 UTC
the patch is now merged on upstream, so I just had to rebuild lib32mesa-git.

also I had a build error due to commit, so I checkout a commit or two before:
41b1882ed45f0fcb827a178cb3f7143fcaa8c4d9 mesa: Use u_math.h from macros.h

When replaying the trace, the glitches are still there, but when playing the game, most of the glitches are gone. (there are still a few, but I will need to take long apitrace to capture it).

also gpu hang seems to be gone too.
Comment 4 Daniel Scharrer 2015-05-21 15:21:55 UTC
Are you still experiencing lockups?

If so, can you
a) run the game with R600_DEBUG=switch_on_eop
b) compile LLVM with the patch from bug #90378 comment 2 and run without R600_DEBUG

This should at least confirm if we are seeing the same issue.
Comment 5 Marek Olšák 2015-07-03 15:40:14 UTC

Could you please give us feedback on the current state of this bug.

Comment 6 kdj0c 2015-08-14 21:52:47 UTC
no more issue on Mesa 10.6.4.

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.