tree 13261dad739908c712bae9837c0aa4cafcdda0b7
parent 44276df755930ac8bfebbb260d6b3de8067c8fc2
author Jingning Han <jingning@google.com> 1573080907 -0800
committer Jingning Han <jingning@google.com> 1573169355 +0000

Remove unneeded constraints in tpl inference model

Remove the additional constraint that multiplies 7/8 to the block
qp offset in the tpl model inference. The new tpl kernel allows
more precise track of the information propagation flow, hence no
need to add in this heuristic. It improves the coding performance.
In speed 1 vbr 150 frames setting, the changes are
          avg PSNR    ovr PSNR    SSIM
lowres    -0.01%      -0.02%      -0.39%
midres    -0.04%      -0.06%      -0.38%
ugc360p   -0.01%      0.0%        -0.16%

STATS_CHANGED

Change-Id: Ie951d50a15e640cff1996b3d8a844e3d0829ea01