Bug 8312 - Bad video quality during playback on media player
Summary: Bad video quality during playback on media player
Status: RESOLVED FIXED
Alias: None
Product: BSPs
Classification: Build System, Metadata & Runtime
Component: bsps-runtime (show other bugs)
Version: 1.9
Hardware: NUC x86_64
: Medium normal
Target Milestone: 2.1
Assignee: Saul Wold
QA Contact: Cristina Agurida
URL:
Whiteboard:
Depends on:
Blocks:
 
Reported: 2015-09-16 15:40 UTC by Cristina Agurida
Modified: 2016-04-06 15:13 UTC (History)
7 users (show)

See Also:
OS type for building Yocto: ---
Type of Regression: Regression (Used to work)
Verified:
Documentation change: No (bug/feature does not impact docs)


Attachments
screenshot (2.66 MB, image/jpeg)
2015-09-16 15:40 UTC, Cristina Agurida
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Cristina Agurida 2015-09-16 15:40:10 UTC
Created attachment 2739 [details]
screenshot

STATUS: Failed
BUILD: 1.9_M3.rc1: ea562de57590c966cd5a75fda8defecd397e6436
ENVIRONMENT: genericx86-64 on NUC
IMAGE: core-image-sato-sdk
HARDWARE: NUC - model: NUCC5i5RYH
              - processor: Intel(R) Core(TM) i5-5250U CPU @ 1.60GHz

STEPS TO REPRODUCE: 
1. copy sample ogv file to system
2. launch media player and make sure it can play the ogv file

EXPECTED OUTCOME: 
ogv file can be played without problem

Actual outcome:
Artifacts appears on screen during video playback.
Check the attachment for details.
Comment 1 Bruce Ashfield 2015-09-16 16:34:27 UTC
Saul, can you redirect this one ?
Comment 2 Jussi Kukkonen 2015-09-17 08:17:55 UTC
I'm assuming the test video is this https://ia600404.us.archive.org/11/items/VideoSample-Video3/ArchitectVideo.ogv

Note that it's very low resolution and seems to be quite badly encoded: so the smudginess and the flicker are there in every player. Those pixels at suspiciously exact intervals should be a bug though.
Comment 3 Jussi Kukkonen 2015-10-05 12:46:41 UTC
I believe I was able to reproduce this originally but the video plays fine on my NUC now. I have a D54250WYK, it should have the HD graphics 5000 in it. I'm currently running intel-corei7-64 but I'll test the genericx86-64 too, just in case.

Cristina, can you re-test if you are still able to reproduce this?
Comment 4 Stephen K Jolley 2015-10-05 21:02:57 UTC
I belive he has answered the question.  Please put it back into NEEDINFO if that is not the case.
Comment 5 Jussi Kukkonen 2015-10-06 07:33:04 UTC
(In reply to comment #4)
> I belive he has answered the question.  Please put it back into NEEDINFO if
> that is not the case.

I marked it as needing info with the hope that this would be re-tested on the original device. I'm setting it to NEEDINFO again: please tell me what I should be doing instead if that's still not correct. 

To recap: I've tested this on my NUC (haswell, HD 5000 I think): the video plays as it should, no artifacts. I was hoping Cristina could re-test as the original test machine seems to have newer GPU in it.
Comment 6 Cristina Agurida 2015-10-06 08:36:37 UTC
It is still reproducible on  NUC model: NUCC5i5RYH with genericx86-64
2.0 release: eac61f37e36099f74485dab398b57f3812826d17
Comment 7 Jussi Kukkonen 2015-10-16 11:03:40 UTC
(In reply to comment #6)
> It is still reproducible on  NUC model: NUCC5i5RYH with genericx86-64
> 2.0 release: eac61f37e36099f74485dab398b57f3812826d17

It looks like the NUCC5i5RYH has HD6000 graphics in it.  Video played fine on my HD5000 (note that the naming is confusing: HD6000 is not actually skylake as one might imagine).

I'll try to get my hands on something with HD6000.
Comment 8 Stephen K Jolley 2015-11-19 16:01:45 UTC
This was REOPENED
Comment 9 Saul Wold 2015-12-15 21:14:10 UTC
Is this still a problem with the newer 2.1M1? I tried this on my NUC5i5RYH and 
it seems to play correctly..
Comment 10 Alexandru Roman 2015-12-22 06:56:43 UTC
I've tested again with:

BUILD: 2.1_M1.RC1
ENVIRONMENT: corei7-64 on NUC
IMAGE: core-image-sato-sdk -  downloaded from http://yocto-ab-master.jf.intel.com/pub/releases/yocto-2.0_M1.rc1/machines/intel-corei7-64/
HARDWARE: NUC - model: NUCC5i5RYH
              - processor: Intel(R) Core(TM) i5-5250U CPU @ 1.60GHz

and I wasn't able to reproduce it.
Comment 11 Stephen K Jolley 2016-01-04 19:39:50 UTC
Info provided
Comment 12 Saul Wold 2016-01-07 16:04:34 UTC
As reported by QA, it appears to be playing correctly, closing this again, 
please reopen if this issue re-appears again.
Comment 13 Alexandru Roman 2016-02-18 14:17:17 UTC
I tested for 2.1_M2.rc3 poky : 152914f2983c5d69001de1d46ce99547fa1e75fe and the bug is still reproductible.

ENVIRONMENT: genericx86-64 on NUC
IMAGE: core-image-sato-sdk
HARDWARE: NUC - model: NUCC5i5RYH
              - processor: Intel(R) Core(TM) i5-5250U CPU @ 1.60GHz

This bug appears only on this Hardware.
Put this on reopened.
Comment 14 Saul Wold 2016-02-18 22:53:31 UTC
I just tested this on a NUC5 and it seems to play correctly for me. Can you 
please include a screenshot (take picture ) or something or better define what 
bad video quality is.

This might be a driver issue in genericx86-64 vs bad hardware, so the generic 
driver is just not the same as the intel graphics driver.
Comment 15 Alexandru Roman 2016-02-22 13:50:31 UTC
An attachment is uploaded. See screenshot.
Comment 16 Saul Wold 2016-03-28 17:47:29 UTC
This seems to be specific to genericx86-64 build due to it missing the graphics 
acceleration code from the Gstreamer VAAPI libraries, it also appears to be 
specific to the NUC5i5 hardware, which is the Intel HD Graphics 5000. 

The fix is to enable the older UXA Acceleration code in the xf86-video-intel 
driver (intel_drv.so), this can be done by enabling uxa via a PACKAGECONFIG 
option in genericx86-64 since it only affects generic and not the meta-intel 
core machines.
Comment 17 Saul Wold 2016-03-31 22:01:13 UTC
Once the UXA acceleration is available in the xf86-video-intel driver, it needs 
to be enabled via the /etc/xorg.conf.d/20-intel.conf settings

Section "Device"
    Identifier  "Intel Graphics"
    Driver      "intel"
    Option      "AccelMethod"  "uxa"
EndSection


The AccelMethod needs to be set to uxa to override the default, the failure 
here can not be automagically detected and overridden it has to be done 
manually.
Comment 18 Saul Wold 2016-04-06 15:13:06 UTC
Found that master for the xf86-video-intel driver had a fix to address this, 
after doing some bisecting found the associated patch resolved this issue on 
the NUC5i5

http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=2ed2089c9b2a9a680434a6d29c312b347604fec5