)]}'
{
  "commit": "e4e586d769ad1f11f07d67fe56187b4ee11939d2",
  "tree": "b36553742294df1863558abc033b75a93aa83be5",
  "parents": [
    "a94d4d759b03ae70a93efaadbcb1684518fa0f29"
  ],
  "author": {
    "name": "Yue Chen",
    "email": "yuec@google.com",
    "time": "Thu Apr 26 16:40:27 2018 -0700"
  },
  "committer": {
    "name": "Yue Chen",
    "email": "yuec@google.com",
    "time": "Fri Apr 27 23:07:56 2018 +0000"
  },
  "message": "Fix sse overflow in highbd_*_obmc_variance sse4 version\n\nAlso the final sse will be stored as an unsigned 64-bit integer,\nwhich is enough. The cumulation process is done in 4 x int32.\nTherefore, for 10-bit and 12-bit video, int32 is not enough to\nhold 1/4 of the total sse of large blocks.\nIn this patch, max_pel_allowed_per_ovf is set up for 10-bit and\n12-bit cases, hbd_obmc_variance_w8n is applied to patches sizes at\n                   bw * (max_allowed_per_ovf / bw)\nto avoid overflow.\n\nBUG\u003daomedia:1772\n\nChange-Id: Iaae2e1403baa93c4ba454d6760281f95428f437a\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "147476ad4bfaaa277d52b3898681a77032125a94",
      "old_mode": 33188,
      "old_path": "aom_dsp/x86/obmc_variance_sse4.c",
      "new_id": "b44aa0bac090b058b996f8aaf22f2e3beebb1c49",
      "new_mode": 33188,
      "new_path": "aom_dsp/x86/obmc_variance_sse4.c"
    }
  ]
}