)]}'
{
  "commit": "320c39905652d24aa44a13f995bbcd83967eeb5d",
  "tree": "bc8970e9062d85d3767764eb7989572af65c7d4f",
  "parents": [
    "7c26f2e9c8d0d7b3d56a2730858b21afc396c8e9"
  ],
  "author": {
    "name": "Paul Wilkins",
    "email": "paulwilkins@google.com",
    "time": "Mon May 18 15:27:37 2020 +0100"
  },
  "committer": {
    "name": "Paul Wilkins",
    "email": "paulwilkins@google.com",
    "time": "Thu May 21 16:23:53 2020 +0000"
  },
  "message": "Change to STRICT_RC calculation of active best Q.\n\nThis on average improves rate accuracy and metrics when this flag\nis set.\n\nLow Res: -6.683\t-9.350\t-4.811\t-7.751\t-9.960\nUGC 360: -4.748\t-8.286\t-3.154\t-5.952\t-7.703\nMid Res: -3.122\t-6.052\t-0.598\t-4.436\t-5.823\nHd Res:  -5.381\t-6.944\t-2.542\t-5.649\t-7.736\n\nNote however that there are a few clips where metrics are much worse\nall be it with big improvements to rate control accuracy across the\nclip.\n\nThis is exemplified by the screen res set where undershoot is\ngenerally reduced but metrics are down.  This can be followed up on\na separate patch and if need be screen content can be handled\ndifferently.\n\nChange-Id: I704f4712fdacc09289a3f5fec00da48edb4906b9\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "2ee6b466449e5845a78460795111c99850f32fc0",
      "old_mode": 33188,
      "old_path": "av1/encoder/pass2_strategy.c",
      "new_id": "1dd121776d7d80434c3c10c137809a6d55a09d06",
      "new_mode": 33188,
      "new_path": "av1/encoder/pass2_strategy.c"
    },
    {
      "type": "modify",
      "old_id": "de63029aeafef5a4cb51d93c75e8ae64b7a60be9",
      "old_mode": 33188,
      "old_path": "av1/encoder/ratectrl.c",
      "new_id": "cf49da2aaf19e57230f8895ebcb24a80c97aa82d",
      "new_mode": 33188,
      "new_path": "av1/encoder/ratectrl.c"
    }
  ]
}