Bug 36410 - [SNB] Rendering errors in 3DMMES subtest taiji
[SNB] Rendering errors in 3DMMES subtest taiji
Status: VERIFIED FIXED
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965
7.10
Other Linux (All)
: medium major
Assigned To: Kenneth Graunke
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2011-04-19 23:12 UTC by meng
Modified: 2011-04-21 18:29 UTC (History)
1 user (show)

See Also:


Attachments
Xorg.0.log (25.82 KB, text/plain)
2011-04-19 23:13 UTC, meng
Details
Photo for this screen mess (541.48 KB, image/png)
2011-04-19 23:14 UTC, meng
Details

Note You need to log in before you can comment on or make changes to this bug.
Description meng 2011-04-19 23:12:45 UTC
System Environment:
--------------------------
Platform:        SNB
Libdrm:          (master)2.4.25
Mesa:            (7.10)1f8c5611a2575e7c660fd10d322cebdafa41e29f
Xserver:         (server-1.10-branch)xorg-server-1.10.1
Xf86_video_intel:(master)2.15.0-3-gc9fb69cb2502917dfb2828c90802de7766072899
Cairo:           (master)b377380b7a5fe0982d057975cd87f6b652d3b0ff
Libva:           (master)bcb8dc4ab8a5da0a815211aa54aa383853fd92c2
Kernel:(drm-intel-fixes) 5c72d064f7ead1126bed6faab0c2bfb7418036e2

Bug detailed description:
-------------------------
Screen mess when running 3DMMES/taiji.It contains a series of demoes when runnig taiji.Screen mess only exists on the first demo(taiji.fms),but it works fine after demoes(eg.hover.fms,feature_tests.fms).Pls see attached photo.
Especifically:
1.It exists on Sugarbay and HuronRiver.
2.It's mesa regression.It works fine on master branch.

Reproduce steps:
-------------------------
1.xinit&
2.sh run.sh(3DMMES/taiji/bin)
Comment 1 meng 2011-04-19 23:13:23 UTC
Created attachment 45840 [details]
Xorg.0.log
Comment 2 meng 2011-04-19 23:14:19 UTC
Created attachment 45841 [details]
Photo for this screen mess
Comment 3 meng 2011-04-20 00:02:31 UTC
It also exists when running glbenchmark.
Comment 4 Ian Romanick 2011-04-20 14:29:27 UTC
(In reply to comment #0)
> System Environment:
> --------------------------
> Platform:        SNB
> Libdrm:          (master)2.4.25
> Mesa:            (7.10)1f8c5611a2575e7c660fd10d322cebdafa41e29f
> Xserver:         (server-1.10-branch)xorg-server-1.10.1
> Xf86_video_intel:(master)2.15.0-3-gc9fb69cb2502917dfb2828c90802de7766072899
> Cairo:           (master)b377380b7a5fe0982d057975cd87f6b652d3b0ff
> Libva:           (master)bcb8dc4ab8a5da0a815211aa54aa383853fd92c2
> Kernel:(drm-intel-fixes) 5c72d064f7ead1126bed6faab0c2bfb7418036e2
> 
> Bug detailed description:
> -------------------------
> Screen mess when running 3DMMES/taiji.It contains a series of demoes when
> runnig taiji.Screen mess only exists on the first demo(taiji.fms),but it works
> fine after demoes(eg.hover.fms,feature_tests.fms).Pls see attached photo.
> Especifically:
> 1.It exists on Sugarbay and HuronRiver.
> 2.It's mesa regression.It works fine on master branch.

That's not what "regression" means.  If it used to work on 7.10 but fails on 7.10 now, it would be a regression.  This is just a bug that has been fixed on master and the fix isn't in 7.10 yet.

Could you bisect to the commit (or commits) in master that fix it?
Comment 5 meng 2011-04-21 00:06:40 UTC
In Mesa's master branch,by bisected shows that:
9a21bc640188e4078075b9f8e6701853a4f0bbe4 bad
9d60a7ce08a67eb8b79c60f829d090ba4a37ed7e good
The commit 9d60a7c fixed it.

commit 9d60a7ce08a67eb8b79c60f829d090ba4a37ed7e
Author: Kenneth Graunke <kenneth@whitecape.org>
Date:   Tue Mar 29 05:49:40 2011 -0700

    i965: Resolve implied moves in brw_dp_READ_4_vs_relative.

    Fixes piglit test glsl-vs-arrays-3 on Sandybridge, as well as garbage
    rendering in 3DMarkMobileES 2.0's Taiji demo and GLBenchmark 2.0's
    Egypt and PRO demos.

    NOTE: This a candidate for stable release branches.  It depends on
    commit 9a21bc640188e4078075b9f8e6701853a4f0bbe4
Comment 6 Gordon Jin 2011-04-21 00:12:46 UTC
Ken, can you cherry pick it?
Comment 7 meng 2011-04-21 01:28:08 UTC
It works fine with 7.10-snb-vs-threads)0144f94c6b85b61f25dd8906041555e52c7370e1.
Comment 8 Ian Romanick 2011-04-21 12:52:50 UTC
(In reply to comment #5)
> In Mesa's master branch,by bisected shows that:
> 9a21bc640188e4078075b9f8e6701853a4f0bbe4 bad
> 9d60a7ce08a67eb8b79c60f829d090ba4a37ed7e good
> The commit 9d60a7c fixed it.
> 
> commit 9d60a7ce08a67eb8b79c60f829d090ba4a37ed7e
> Author: Kenneth Graunke <kenneth@whitecape.org>
> Date:   Tue Mar 29 05:49:40 2011 -0700
> 
>     i965: Resolve implied moves in brw_dp_READ_4_vs_relative.
> 
>     Fixes piglit test glsl-vs-arrays-3 on Sandybridge, as well as garbage
>     rendering in 3DMarkMobileES 2.0's Taiji demo and GLBenchmark 2.0's
>     Egypt and PRO demos.
> 
>     NOTE: This a candidate for stable release branches.  It depends on
>     commit 9a21bc640188e4078075b9f8e6701853a4f0bbe4

Great!  This was one of the commits in my 7.10-snb-vs-threads branch.  I merged that branch to 7.10 just today.  On 7.10, this is commit 0f02b4253d5b3a3bbdfc72711bb9a7c0781061f4.
Comment 9 meng 2011-04-21 18:29:00 UTC
Verified with commit (7.10)0f02b4253d5b3a3bbdfc72711bb9a7c0781061f4, it works fine.