)]}'
{
  "commit": "9f59323c1225454aecf7748739e9fcaba88cc753",
  "tree": "d1ccfe0bc4e1d9a3183441017385e1a11be1a8c0",
  "parents": [
    "52bd722bfa59d7df80a994bf0cdeb1f30119398e"
  ],
  "author": {
    "name": "Rachel Barker",
    "email": "rachelbarker@google.com",
    "time": "Fri Feb 09 14:32:38 2024 +0000"
  },
  "committer": {
    "name": "Rachel Barker",
    "email": "rachelbarker@google.com",
    "time": "Mon Feb 12 17:50:41 2024 +0000"
  },
  "message": "Optimize RANSAC code\n\nEliminate a lot of intermediate arrays and copies by\nmaking the following changes:\n\n* Combine point projection and scoring into one function,\n  which removes the need for the projected_points array\n\n* Pass correspondence lists directly into the model fitting\n  and scoring functions\n\n* Pass the selected indices into model fitting functions,\n  instead of copying the selected correspondences into a temporary\n  array. Note that this is more efficient because the selected\n  points are only used once; if they were used more times, then\n  there may be a benefit to explicitly copying them.\n\nChange to encoder speed:\n\nSpeed 1: +0.024% encode time\nSpeed 2: -0.028% encode time\nSpeed 3: -0.017% encode time\nSpeed 4: -0.126% encode time\n\nChange-Id: I5fbfb653c0083f9854a402910028a5bedfdcb40f\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "b622ff2254748d2f66ac212f45e60ac4c0b5ca45",
      "old_mode": 33188,
      "old_path": "aom_dsp/flow_estimation/ransac.c",
      "new_id": "7c7bebdda46aea136d7f09b674ff5c9a7dedf495",
      "new_mode": 33188,
      "new_path": "aom_dsp/flow_estimation/ransac.c"
    }
  ]
}