)]}'
{
  "commit": "641ec17a1d9c7fb648b089161e12c3d972d2cb04",
  "tree": "144705c4ca0e87475b56857044b3f8dac1681d52",
  "parents": [
    "a10fad7cc233e70ac647081eaeb9e0c4ecbb1e1d"
  ],
  "author": {
    "name": "Behdad Esfahbod",
    "email": "behdad@behdad.org",
    "time": "Thu May 04 11:58:39 2023 -0600"
  },
  "committer": {
    "name": "Behdad Esfahbod",
    "email": "behdad@behdad.org",
    "time": "Thu May 04 11:58:39 2023 -0600"
  },
  "message": "[layout] Use NEON SIMD instructions for apply_forward()\n\nTo skip four glyphs at a time.\n\nUnfortunately slows down benchmark-shape by up to 25% for Amiri, instead of\nspeeding anything up. This is due to the overhead of switching back and forth\nbetween the ARM and NEON processors.\n\nTo be experimented with: Same architecture but with AVX on x86_64 to skip 8\nglyphs at a time.\n\nhttps://github.com/harfbuzz/harfbuzz/issues/566\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "aae2bf9cea172b77149a3199e78fd9ef94aeead5",
      "old_mode": 33188,
      "old_path": "src/hb-ot-layout-gsubgpos.hh",
      "new_id": "7717cadb3525f208459311cd3e00709e03a69a80",
      "new_mode": 33188,
      "new_path": "src/hb-ot-layout-gsubgpos.hh"
    },
    {
      "type": "modify",
      "old_id": "c66ee8cfd06be1577f8b7df28d259a0edcae7140",
      "old_mode": 33188,
      "old_path": "src/hb-ot-layout.cc",
      "new_id": "97f3473afb1b02ec41a585d6a531b41ec6e71e77",
      "new_mode": 33188,
      "new_path": "src/hb-ot-layout.cc"
    },
    {
      "type": "modify",
      "old_id": "dab713729b7d43ffda87a4f536ee0c23138d6f2a",
      "old_mode": 33188,
      "old_path": "src/hb-set-digest.hh",
      "new_id": "0dcdc3e56c1692cf5474497d3a7ce5cacb78fa50",
      "new_mode": 33188,
      "new_path": "src/hb-set-digest.hh"
    }
  ]
}