Bug 75794 - Accumulation of black squares with OpenGL applications at high resolutions
Summary: Accumulation of black squares with OpenGL applications at high resolutions
Status: RESOLVED MOVED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/r300 (show other bugs)
Version: 9.1
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-03-05 11:47 UTC by l.mollari
Modified: 2019-09-18 18:52 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
1. glxgears with artifacts (51.09 KB, image/png)
2014-03-05 11:47 UTC, l.mollari
Details
2. glxgears without artifacts at lower resolution (57.53 KB, image/png)
2014-03-05 11:49 UTC, l.mollari
Details
3. 0 A. D. with artifacts (1.39 MB, image/png)
2014-03-05 11:50 UTC, l.mollari
Details
4. 0 A. D. with silhouette artifacts (1.06 MB, image/png)
2014-03-05 11:51 UTC, l.mollari
Details
5. Latest Xorg.0.log from the affected system (44.92 KB, text/plain)
2014-03-05 11:52 UTC, l.mollari
Details
6. List of tested versions (734 bytes, text/plain)
2014-03-05 11:53 UTC, l.mollari
Details
7. Output from lspci (excerpt) (204 bytes, text/plain)
2014-03-05 11:55 UTC, l.mollari
Details
8. Output from xrandr (1.43 KB, text/plain)
2014-03-05 11:55 UTC, l.mollari
Details

Description l.mollari 2014-03-05 11:47:24 UTC
Created attachment 95154 [details]
1. glxgears with artifacts

I am having trouble with black areas made up by 8 × 8 pixel squares that occur in 3D enabled applications when using the build-in TFTs native resolution of 1920 × 1200. Lower resolutions are not affected (see attached screenshots [1] and [2]). The shape of these areas is sensitive to the actual content drawn in such a way that it acts like a shadow or an impression of the actual 3D geometry and even of 2D HUDs in 3D enabled games. These artifacts are persistent across different applications i. e. the current application inherits the artifacts from the previous one and vice versa [4]. I have uploaded a video showing the artifacts described in 0 A. D.:

    https://www.youtube.com/watch?v=Rpu-s8DowFw

I narrowed the glitches down to Mesa 3D version 9.1 that seemingly introduced the bug. The bug seems to be still present in the most recent Mesa 3D and Linux kernel versions [6].

Damaged hardware is unlikely to be the cause for these glitches since I was able to reproduce the same artifacts on three different laptops of the same product line, all equipped with the same GPU.

All tests were performed with unaltered default settings of the respective distributions. No manually touched or customized xorg.conf files or kernel mode settings were in place.
Comment 1 l.mollari 2014-03-05 11:49:12 UTC
Created attachment 95155 [details]
2. glxgears without artifacts at lower resolution
Comment 2 l.mollari 2014-03-05 11:50:04 UTC
Created attachment 95156 [details]
3. 0 A. D. with artifacts
Comment 3 l.mollari 2014-03-05 11:51:30 UTC
Created attachment 95157 [details]
4. 0 A. D. with silhouette artifacts
Comment 4 l.mollari 2014-03-05 11:52:47 UTC
Created attachment 95158 [details]
5. Latest Xorg.0.log from the affected system
Comment 5 l.mollari 2014-03-05 11:53:40 UTC
Created attachment 95159 [details]
6. List of tested versions
Comment 6 l.mollari 2014-03-05 11:55:02 UTC
Created attachment 95160 [details]
7. Output from lspci (excerpt)
Comment 7 l.mollari 2014-03-05 11:55:47 UTC
Created attachment 95161 [details]
8. Output from xrandr
Comment 8 Alex Deucher 2014-03-05 14:13:37 UTC
Can you use git to bisect between the 9.0 and 9.1 releases to see what commit caused the regression?
Comment 9 l.mollari 2014-03-05 20:34:01 UTC
Sure, I am going to give it a try.
Comment 10 l.mollari 2014-03-11 07:45:06 UTC
My attempt to track down the commit that introduced the glitches ended with the following error from the configure script, using revisions "mesa-9.1" and "mesa-9.0.3":

configure: error: DRI driver directory 'r300' does not exist

I followed the instructions given at the Ubuntu wiki site, using the configure command with the parameters as listed below:

https://wiki.ubuntu.com/X/BisectingMesa

git clone git://anongit.freedesktop.org/git/mesa/mesa

...

./configure --prefix=/usr --mandir=\${prefix}/share/man \
    --infodir=\${prefix}/share/info --sysconfdir=/etc \
    --localstatedir=/var --build=i486-linux-gnu --disable-gallium --with-driver=dri \
    --with-dri-drivers="r300" --with-demos=xdemos --libdir=/usr/lib/glx \
    --enable-glx-tls --enable-driglx-direct --disable-egl \
    --disable-glu --disable-glut --disable-glw CFLAGS="-Wall -g -O2"

To my understanding the particular driver "r300" is not (any longer?) part of what I have cloned so far from the git repository, hence the configure script wants to include a precompiled version. Could you please give me a hint how I can get past this missing driver obstacle?

However opting out the parameter --with-dri-drivers="r300" enables the configure script and subsequently the make command to succeed even though glxgears, now running without artifacts at decent fps, is stuttering noticeably. This applies to 9.0.3 and 9.1 likewise.
Comment 11 Michel Dänzer 2014-03-11 08:29:06 UTC
First of all, have you confirmed that the problem is still present in Mesa 10.x?

Mesa 9.0 already didn't have the classic r300 driver anymore. You need to use --with-gallium-drivers=r300 instead of --with-dri-drivers="r300" (and maybe also drop --disable-gallium just in case, although AFAICT that shouldn't have any effect).
Comment 12 l.mollari 2014-03-11 13:00:51 UTC
Mesa 10.2.0-devel is still affected by this bug.
Comment 13 Brais 2018-10-28 16:26:06 UTC
Opening Google Chrome seems to "fix" the bug. Thats weird, my computer is a HP Compaq nw8440 with a FireGL V5200 GPU, only resolutions over 1920x1080 trigger the bug. It's fine at 1920x1080 but not at 1920x1081 (In Window mode).
Comment 14 GitLab Migration User 2019-09-18 18:52:35 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/360.


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.