Artifacts in the decoded frames in chromeOS
Reported by
sean.v.k...@intel.com,
Jan 25 2017
|
|||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/56.0.2924.67 Safari/537.36 Example URL: Android CTS test file Steps to reproduce the problem: There are human visible artifacts in the decoded frames of the attached videos. This happens for all format: vp8, vp9, and h264, and also of different resolutions. The issue can be reproduced on both Cyan and Samus. The screenshots of hardware and software output are also attached. What is the expected behavior? Colors are consistent with no artifacts What went wrong? The root cause is that our AVS (Advanced Video Scalar) firmware in the driver is not really as well optimized for AVS 8x8 coefficients. It needs to sample the neighboring 8x8 pixels to get the result. When the default flag is passed, the AVS only samples the neighboring 2x2 pixel to get the result. When the HQ flag is passed, it needs to use 8x8 pixels. Separately, I've taken a look at video output by setting the scalar back to default and it should be sufficient for now. Did this work before? N/A Is it a problem with Flash or HTML5? N/A Does this work in other browsers? N/A Chrome version: 56.0.2924.67 Channel: n/a OS Version: Flash Version: Contents of chrome://gpu: This not a web browser issue. It's a vaapi wrapper issue. https://code.google.com/p/chrome-os-partner/issues/detail?id=61609 https://code.google.com/p/chrome-os-partner/issues/detail?id=61609 Creating this bug so I can submit a patch
,
Jan 26 2017
clips and screen shots
,
Jan 26 2017
Added code for review: https://codereview.chromium.org/2656073003
,
Feb 8 2017
does the videos work fine on previous version of Chrome and other browsers? can you rule out the artifact is because of the video quality of itself?
,
Feb 8 2017
Re#4, see my comments in the Google issue, https://code.google.com/p/chrome-os-partner/issues/detail?id=61609. Issue is quality and the fact that the AVS engine does not currently support 8x8 coefficients in the HQ mode and I've moved it to the default mode in the patch. This CTS test was not previously tested and is a part of the requirements for ARC++
,
Feb 8 2017
Updated md5 sums: https://codereview.chromium.org/2683953003
,
Feb 16 2017
Thank you for providing more feedback. Adding requester "yiningc@chromium.org" for another review and adding "Needs-Review" label for tracking. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Feb 16 2017
jrummell@, since you are reviewer in CL as per c#8, I assign this bug to you. Please feel free to resolve it when the CL is checked in.
,
Feb 17 2017
,
Feb 20 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/4fc5997a7496bc2b803d049ab0218c385f6d6427 commit 4fc5997a7496bc2b803d049ab0218c385f6d6427 Author: sean.v.kelley <sean.v.kelley@intel.com> Date: Mon Feb 20 12:03:01 2017 media/gpu: restore VPP scaling to default AVS (Advanced Video Scalar) support in the Intel vaapi driver is not really as well optimized for AVS 8x8 coefficients. It needs to sample the neighboring 8x8 pixels to get the result. When the default flag is passed, the AVS only samples the neighboring 2x2 pixel to get the result. When the HQ flag is passed, it needs to use 8x8 pixels. Separately, I've taken a look at video output by setting the scalar back to default and it should be sufficient for now. Changing scaling property to VA_FILTER_SCALING_DEFAULT in media/gpu/vaapi_wrapper.c requires an update to checksums due to the differing filter properties in AVS Scaling options. BUG= 685432 TEST=video_decode_accelerator_unittest, visual decode of samples; see crbug.com/685432#c2 for samples; Thumbnail.* test R=posciak@chromium.org,jrummell Signed-off-by: Sean V Kelley <seanvk@posteo.de> CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel Review-Url: https://codereview.chromium.org/2656073003 Cr-Commit-Position: refs/heads/master@{#451601} [modify] https://crrev.com/4fc5997a7496bc2b803d049ab0218c385f6d6427/media/gpu/vaapi_wrapper.cc [modify] https://crrev.com/4fc5997a7496bc2b803d049ab0218c385f6d6427/media/test/data/test-25fps.h264.md5 [modify] https://crrev.com/4fc5997a7496bc2b803d049ab0218c385f6d6427/media/test/data/test-25fps.vp8.md5 [modify] https://crrev.com/4fc5997a7496bc2b803d049ab0218c385f6d6427/media/test/data/test-25fps.vp9.md5
,
Feb 21 2017
,
Apr 17 2017
,
May 30 2017
,
Jun 16 2017
Verified on 60.0.3112.30/9592.18.0 and 9656.0.0, 61.0.3132.0 on cyan device |
|||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||
Comment 1 by dalecur...@chromium.org
, Jan 26 2017