Environment: -------------- Platform:HSW/IVB/SGB Libva: (staging) b8d3cf092c9b07cfd909552a3c160b7db3b5a91d Intel-driver: (staging) a6dfb8f7fb601f102b414ac0c9088b5b8e88060f Bug Info: -------------- Poor quality when decoding SML0014.rcv Steps: -------------- 1.xinit& 2.mplayer -fps 30 -va vaapi -vo vaapi SML0014.rcv Comments: -------------- Issue can be reproduce with below files: SA10107.vc1 SA10109.vc1 sa10111.vc1 SA10116.vc1 SA10117.vc1 sa10119.vc1 SA10127.vc1 SA10133.vc1 SA10136.vc1 SA10140.vc1 SA10141.vc1 SA10144.vc1 SA10147.vc1 SA10148.vc1 SA10149.vc1 SA10153.vc1 SA10154.vc1 SA10155.vc1 SA10157.vc1 SA10167.vc1 SA10168.vc1 SA10169.vc1 SA10171.vc1 SA10172.vc1 SA10179.vc1 SA30034.vc1 SA30036.vc1 SA30038.vc1 sa30039.vc1 SA30042.vc1 SA30044.vc1 SA30048.vc1 SA30049.vc1 SA30051.vc1 SA30052.vc1 SA30054.vc1 SA30055.vc1 SA40017.vc1
update comments: Issue can be reproduced with below files. SML0014.rcv SML0015.rcv SML0016.rcv SML0017.rcv SSL0018.rcv SSL0024.rcv SSM0013.rcv SSM0015.rcv SSM0016.rcv SA00040.vc1 SA00046.vc1 SA00048.vc1 SA00049.vc1 SA00052.vc1 SA00058.vc1 SA00059.vc1 SA00067.vc1 SA00068.vc1 SA00070.vc1 SA00072.vc1 SA00073.vc1 SA00074.vc1 SA00075.vc1 SA00078.vc1 SA00079.vc1 SA10090.vc1 SA10094.vc1 SA10097.vc1 SA10131.vc1 SA10161.vc1 SA10163.vc1 SA10165.vc1 SA20021.vc1 SA20029.vc1 SA20030.vc1 SA30030.vc1 SA30031.vc1 SA40013.vc1
(In reply to comment #0) > Environment: > -------------- > Platform:HSW/IVB/SGB > Libva: (staging) b8d3cf092c9b07cfd909552a3c160b7db3b5a91d > Intel-driver: (staging) a6dfb8f7fb601f102b414ac0c9088b5b8e88060f > > > Bug Info: > -------------- > Poor quality when decoding SML0014.rcv > > > Steps: > -------------- > 1.xinit& > 2.mplayer -fps 30 -va vaapi -vo vaapi SML0014.rcv I can't reproduce the issue on my HSW machine, could you attach an image to show what is the issue ? > > > Comments: > -------------- > Issue can be reproduce with below files: > > SA10107.vc1 Are you sure it can be reproduced with this file ? SA10107.vc1 is an interlaced file. > SA10109.vc1 > sa10111.vc1 > SA10116.vc1 > SA10117.vc1 > sa10119.vc1 > SA10127.vc1 > SA10133.vc1 > SA10136.vc1 > SA10140.vc1 > SA10141.vc1 > SA10144.vc1 > SA10147.vc1 > SA10148.vc1 > SA10149.vc1 > SA10153.vc1 > SA10154.vc1 > SA10155.vc1 > SA10157.vc1 > SA10167.vc1 > SA10168.vc1 > SA10169.vc1 > SA10171.vc1 > SA10172.vc1 > SA10179.vc1 > SA30034.vc1 > SA30036.vc1 > SA30038.vc1 > sa30039.vc1 > SA30042.vc1 > SA30044.vc1 > SA30048.vc1 > SA30049.vc1 > SA30051.vc1 > SA30052.vc1 > SA30054.vc1 > SA30055.vc1 > SA40017.vc1
SSIM value is too low. eg: SML0014.rcv Y SSIM extreme 0.336215 U SSIM average 0.986056 V SSIM extreme 0.538501 Y SSIM average 0.994315 U SSIM extreme 0.536182 V SSIM average 0.993605
The SSIM is low from frame 669 and every two frames. 2005 Y_SSIM = 0.337364 2006 U_SSIM = 0.777391 2007 V_SSIM = 0.739837 2008 Y_SSIM = 1.000000 2009 U_SSIM = 1.000000 2010 V_SSIM = 1.000000 2011 Y_SSIM = 1.000000 2012 U_SSIM = 1.000000 2013 V_SSIM = 1.000000 2014 Y_SSIM = 0.408797 2015 U_SSIM = 0.817545 2016 V_SSIM = 0.770707 2017 Y_SSIM = 1.000000 2018 U_SSIM = 1.000000 2019 V_SSIM = 1.000000 2020 Y_SSIM = 1.000000 2021 U_SSIM = 1.000000 2022 V_SSIM = 1.000000 2023 Y_SSIM = 0.361894 2024 U_SSIM = 0.821343 2025 V_SSIM = 0.784269 2026 Y_SSIM = 1.000000 2027 U_SSIM = 1.000000 2028 V_SSIM = 1.000000 2029 Y_SSIM = 1.000000 2030 U_SSIM = 1.000000 2031 V_SSIM = 1.000000 2032 Y_SSIM = 0.387867 2033 U_SSIM = 0.737673 2034 V_SSIM = 0.731032
Do we have further update on this issue?
I can reproduce this issue with SML0014.rcv, but fail to reproduce this issue with SML0015.rcv, could you also list the SSIM for SML0015.rcv ?
The issue for SML0014.rcv has been fixed, please retest with other clips commit f0523c58e304bd8e7b2a470c4cb4e06057a0a4bf Author: Xiang, Haihao <haihao.xiang@intel.com> Date: Wed Jan 9 15:44:29 2013 +0800 No overlap smoothing for B frame in Main and Simple profiles
(In reply to comment #6) > I can reproduce this issue with SML0014.rcv, but fail to reproduce this > issue with SML0015.rcv, could you also list the SSIM for SML0015.rcv ? (In reply to comment #6) > I can reproduce this issue with SML0014.rcv, but fail to reproduce this > issue with SML0015.rcv, could you also list the SSIM for SML0015.rcv ? SML0015.rcv Y SSIM extreme 0.996214 U SSIM average 0.998777 V SSIM extreme 0.999285 Y SSIM average 0.999712 U SSIM extreme 0.999403 V SSIM average 0.999823
OK, they should be different issues. The extreme SSIM of SML0014.rcv is too low, however the extreme SSIM of ML0015.rcv is closed to 1. Could you file a new bug and attach the whole SSIM values for SML0015.rcv so I know which frame has the extreme SSIM ?
Filename, Y extreme,Y average,U extreme,U average,V extreme, V agerage, SA10165.vc1, 0.005538, 0.067395, 0.042607, 0.04688, 0.000823, 0.02186, SA20029.vc1, 0.081796, 0.145497, 0.054147, 0.083552, 0.072035, 0.107538, SA20030.vc1, 0.081796, 0.145497, 0.054147, 0.083552, 0.072035, 0.107538, SA30031.vc1, 0.281476, 0.993156, 0.685092, 0.995448, 0.722642, 0.995745, above 4 files SSIM value is too low. maybe same as SML0014.rcv.
And for the SML0015.rcv and other clips that SSIM value is closed to 1, i have reported a new bug : #bug 59228
for SML0014.rcv, this issue has been fix. but for the 4 files below, SSIM value is too low. SA10165.vc1, 0.005538, 0.067395, 0.042607, 0.04688, 0.000823, 0.02186, SA20029.vc1, 0.081796, 0.145497, 0.054147, 0.083552, 0.072035, 0.107538, SA20030.vc1, 0.081796, 0.145497, 0.054147, 0.083552, 0.072035, 0.107538, SA30031.vc1, 0.281476, 0.993156, 0.685092, 0.995448, 0.722642, 0.995745,
SA10165.vc1 SA20029.vc1 SA20030.vc1 FFmpeg doesn't parse PANSCAN section in picture header, so it passes the wrong parameters to driver.
mplayer log: ... VO: [vaapi] 720x576 => 720x576 VC-1 VA-API Acceleration [VD_FFMPEG] XVMC-accelerated MPEG-2. [vc1 @ 0x8b50ec0]Pan-scan not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented. ...
SML0014.rcv SSIM value still too low on SNB plaform. SML0014.rcv 0.508544 0.992074 0.876745 0.997556 0.809875 0.996775 Platform: SNB libva: staging 21649988d6b532cc96f633db017d1e4369f640e9 inter-driver: staging d206b47a6ac86c089149ecd71b01eea6ebda5796
(In reply to comment #15) > SML0014.rcv SSIM value still too low on SNB plaform. > > SML0014.rcv 0.508544 0.992074 0.876745 0.997556 0.809875 0.996775 > > Platform: SNB > libva: staging 21649988d6b532cc96f633db017d1e4369f640e9 > inter-driver: staging d206b47a6ac86c089149ecd71b01eea6ebda5796 Another commit is needed for this issue on SNB :( commit c650dab1c8e22fbe17f2ad2f8f26c24a12cb60fd Author: Xiang, Haihao <haihao.xiang@intel.com> Date: Wed Jan 23 09:59:17 2013 +0800 Enlarge deblocking filter row store for VC-1 decoding
SA30031.vc1 has multi slices, so patch your mplayer-vaapi (FFmpeg) firstly.
(In reply to comment #16) > (In reply to comment #15) > > SML0014.rcv SSIM value still too low on SNB plaform. > > > > SML0014.rcv 0.508544 0.992074 0.876745 0.997556 0.809875 0.996775 > > > > Platform: SNB > > libva: staging 21649988d6b532cc96f633db017d1e4369f640e9 > > inter-driver: staging d206b47a6ac86c089149ecd71b01eea6ebda5796 > > Another commit is needed for this issue on SNB :( > > commit c650dab1c8e22fbe17f2ad2f8f26c24a12cb60fd > Author: Xiang, Haihao <haihao.xiang@intel.com> > Date: Wed Jan 23 09:59:17 2013 +0800 > > Enlarge deblocking filter row store for VC-1 decoding IIRC, the spec indeed mentioned that we need twice the vertical resolution, though initially that was implied for interlaced contents only.
(In reply to comment #18) > (In reply to comment #16) > > (In reply to comment #15) > > > SML0014.rcv SSIM value still too low on SNB plaform. > > > > > > SML0014.rcv 0.508544 0.992074 0.876745 0.997556 0.809875 0.996775 > > > > > > Platform: SNB > > > libva: staging 21649988d6b532cc96f633db017d1e4369f640e9 > > > inter-driver: staging d206b47a6ac86c089149ecd71b01eea6ebda5796 > > > > Another commit is needed for this issue on SNB :( > > > > commit c650dab1c8e22fbe17f2ad2f8f26c24a12cb60fd > > Author: Xiang, Haihao <haihao.xiang@intel.com> > > Date: Wed Jan 23 09:59:17 2013 +0800 > > > > Enlarge deblocking filter row store for VC-1 decoding > > IIRC, the spec indeed mentioned that we need twice the vertical resolution, > though initially that was implied for interlaced contents only. No, this buffer is not scalable with picture height
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.