)]}'
{
  "commit": "dc89ee09f12c5541f391fc89db8458db6cae0950",
  "tree": "a328159773ffc0427aa4238c1d2b56213f123408",
  "parents": [
    "9cd4a15c8a4b716bfa31f889455197e5bffbe7ab"
  ],
  "author": {
    "name": "DRC",
    "email": "information@libjpeg-turbo.org",
    "time": "Mon Feb 19 14:16:46 2018 -0600"
  },
  "committer": {
    "name": "DRC",
    "email": "information@libjpeg-turbo.org",
    "time": "Tue Feb 27 10:40:32 2018 -0600"
  },
  "message": "64-bit AVX2 implementation of fast integer FDCT\n\nStill not faster than SSE2.  Improving upon SSE2 performance will\nprobably require restructuring the algorithm to combine the various\nmultiply/add operations, but I\u0027m not sure how to do that without\nintroducing further roundoff error.  Left as an exercise for the reader.\n\nThe IFAST FDCT algorithm is sort of a legacy feature anyhow.  Even with\nSSE2 instructions, the ISLOW FDCT was almost as fast as the IFAST FDCT.\nSince the ISLOW FDCT has been accelerated with AVX2 instructions, it is\nnow about the same speed as the IFAST FDCT on AVX2-equipped CPUs.\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "3b881ead3767bca14c1f3008dbd9899185f1f16f",
      "old_mode": 33188,
      "old_path": "simd/x86_64/jfdctfst-avx2.asm",
      "new_id": "d472ee5eb1354aad54af6904730c9a61da31b87c",
      "new_mode": 33188,
      "new_path": "simd/x86_64/jfdctfst-avx2.asm"
    }
  ]
}