Bug 68344

Summary: [NV50 gallium] [piglit] shaders/glsl-fs-texture2d-dependent-4 randomly passes or fails on NVAA/NV50
Product: xorg Reporter: dirkneukirchen
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: NEW --- QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium    
Version: git   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
normal bootlog none

Description dirkneukirchen 2013-08-20 16:38:58 UTC
Created attachment 84349 [details]
normal bootlog

./piglit-run.py -t '/shaders/glsl-fs-texture2d-dependent-4' tests/quick.tests results/quick-shaders-single-test.tests 

passes or fails seemingly random
20 passes - 9 fails

Result of fail:
Probe color at (25,224)
  Expected: 0.000000 0.000000 1.000000
  Observed: 1.000000 0.000000 0.000000
Probe color at (224,224)
  Expected: 1.000000 1.000000 1.000000
  Observed: 0.000000 1.000000 0.000000

Hardware:
Nvidia 8200 Onboard NVAA/NV50
nouveau  [  DEVICE][0000:02:00.0] Chipset: MCP77/MCP78 (NVAA)
nouveau  [  DEVICE][0000:02:00.0] Family : NV50

Software:
Linux Mint 15 + xorg-edgers + 3.11-rc5
Kernel: 3.11.0-rc5-vanillaice (compiled)
libdrm-nouveau: 2.4.46+git20130702.c6d73cfe-0ubuntu0sarvatt~raring
xserver-xorg-video-nouveau : 1:1.0.9+git20130730.300c5a32-0ubuntu0sarvatt~raring
mesa: 9.2.0~git20130729+9.2.9b8ad643-0ubuntu0sarvatt~raring
Comment 1 Ilia Mirkin 2013-08-20 17:10:32 UTC
As a datapoint, I also got it to fail on NV96, similar failure. However for me it was more like 1/100 times. And I never saw the failure with mesa 9.1.2, only mesa-git (from a few days ago).

This is how I ran it:

for i in `seq 100`; do bin/shader_runner tests/shaders/glsl-fs-texture2d-dependent-4.shader_test -auto; done

Here was the failure:

Probe color at (25,225)
  Expected: 0.000000 0.000000 1.000000
  Observed: 1.000000 0.000000 0.000000
Probe color at (225,225)
  Expected: 1.000000 1.000000 1.000000
  Observed: 0.000000 1.000000 0.000000

Note how for me it says ,225 and not ,224. Might be a different piglit test version... I'm at latest. (Or could be a difference between nv84 3d and nva0 3d.)

This feels like a timing issue -- e.g. we return too early without having flushed something.
Comment 2 Ilia Mirkin 2013-12-08 08:25:51 UTC
Hmmm... I just went and retested this with latest mesa, and I can't get it to fail on NV98. There have been a few patches that went in that may affect this type of issue.

Could you verify that this still happens with mesa-git?
Comment 3 dirkneukirchen 2013-12-10 20:14:33 UTC
It will take a while to set up because several other things are not working with NVAA too and temporary "removal" of the driver in 3.13-rc1 - 3.13-rc3 is not helping either

see (similar?) NVAA / NVAC:
https://bugs.freedesktop.org/show_bug.cgi?id=72399
and on the mailing list:
http://lists.freedesktop.org/archives/dri-devel/2013-December/050037.html


building with Ubuntu wasnt as nice ... now trying to set up with Fedora
Comment 4 Ilia Mirkin 2013-12-10 20:32:12 UTC
This could be tested with 3.12 just fine. The commits I was referencing were to mesa. The important part would be to have a fresh mesa git checkout. (And use it to run the tests :) )

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.