)]}'
{
  "commit": "6de4a77df349bd9072d2bf8f257602eb819f5e61",
  "tree": "4d17e57385470b4add47a2cfa588e82a140c742c",
  "parents": [
    "716f0ea3cf9efc3ca4ce53dc9f9b08e9234cd367"
  ],
  "author": {
    "name": "Angie Chiang",
    "email": "angiebird@google.com",
    "time": "Thu Apr 14 15:16:08 2016 -0700"
  },
  "committer": {
    "name": "Angie Chiang",
    "email": "angiebird@google.com",
    "time": "Mon Apr 18 11:00:13 2016 -0700"
  },
  "message": "Remove double operation from tx_size selection\n\nThis CL fix the bug\nrdopt.c:1687: choose_tx_size_from_rd: Assertion\n`mbmi-\u003etx_type \u003d\u003d DCT_DCT\u0027 failed\n\nIt is caused by\n1) mms register access before double operation\n2) different compiler behaviors\ncode:\n  int64_t a \u003d INT64_MAX;\n  double b \u003d 1. * INT64_MAX;\n  printf(\"a \u003c b: %d\\n\", a \u003c b);\nresult:\n  a \u003c b: 0\n\ncode:\n  --target\u003dx86-linux-gcc\n  int64_t a \u003d INT64_MAX;\n  double b \u003d 1. * INT64_MAX;\n  printf(\"a \u003c b: %d\\n\", a \u003c b);\nresult:\n  a \u003c b: 1\n\nI remove the double operation and test it with EXT_TX experiment.\nThe psnr change is around 0.05%, which is considered as noise level.\n\nChange-Id: If8935c70c8603617fcfa8571accd30ccdda786a0\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "b3f833663ef54425f7e3f92c00393ce89efa0b60",
      "old_mode": 33188,
      "old_path": "vp10/encoder/rdopt.c",
      "new_id": "63b6a68093f42a4a0f54bc84cc68f91dfb1f5fb1",
      "new_mode": 33188,
      "new_path": "vp10/encoder/rdopt.c"
    }
  ]
}