)]}'
{
  "commit": "763ccd8c0cd710b74022acc2da7380d79f73dcb7",
  "tree": "0df297258902fef3ea95eceaa6cfc5dd6b5e1260",
  "parents": [
    "70539b10d90d6dfd8898605e68c3faffe1975120"
  ],
  "author": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Fri Dec 01 17:07:20 2017 -0800"
  },
  "committer": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Wed Dec 06 20:44:24 2017 +0000"
  },
  "message": "Simplify warped motion parameter estimation\n\nThe purpose of this change is to reduce the cycles needed for warped\nmotion parameter estimation.\n\nMethod 1:\nIf we remove the 2-bit bit-depth reduction(as in patch set 2), the\ndownshifting of A, Bx, By is also removed. The borg test result(over\nthe baseline) is:\n             avg_psnr ovr_psnr  ssim\nlowres:      0.023     0.020    0.071\ncam_lowres: -0.009    -0.017   -0.031\n\nMethod 2:\nIn theory, the above change uses 2 more bits for elements of A, Bx,\nBy. In patchset 3, we modified LS_STEP to be 8(1 full pixel), and now,\nthe least 2 bits in A, Bx, By elements are always 0. Namely, 2-bit\nbit-depth reduction are achieved without extra operations. The borg\ntest result(over the baseline) is:\nlowres:     -0.004    -0.007   -0.023\ncam_lowres: -0.031    -0.033   -0.045\nThis is a little better than patch set 2 result.\n\nMethod 2 is the final choice.\n\nChange-Id: I945aaba412e2ea86b7d67e8a90741fdf395b94cd\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "f3bc84957e44070ba15f53fe22d6d34f57c21fbe",
      "old_mode": 33188,
      "old_path": "av1/common/warped_motion.c",
      "new_id": "1a9c6dcfdc492b16e4deadb745fcc29b26005e6c",
      "new_mode": 33188,
      "new_path": "av1/common/warped_motion.c"
    }
  ]
}