)]}'
{
  "commit": "f28e0bbefa41fe643cce2f107e868abff312ced9",
  "tree": "933db7fedccb1c2590441909271db03ff8cba52f",
  "parents": [
    "93a6085618f16fb2cd316d1e84f1a638b7e2d8ff"
  ],
  "author": {
    "name": "Alexander Monakov",
    "email": "amonakov@ispras.ru",
    "time": "Tue Feb 06 23:48:08 2024 +0300"
  },
  "committer": {
    "name": "Richard Henderson",
    "email": "richard.henderson@linaro.org",
    "time": "Fri May 03 08:03:05 2024 -0700"
  },
  "message": "util/bufferiszero: Optimize SSE2 and AVX2 variants\n\nIncrease unroll factor in SIMD loops from 4x to 8x in order to move\ntheir bottlenecks from ALU port contention to load issue rate (two loads\nper cycle on popular x86 implementations).\n\nAvoid using out-of-bounds pointers in loop boundary conditions.\n\nFollow SSE2 implementation strategy in the AVX2 variant. Avoid use of\nPTEST, which is not profitable there (like in the removed SSE4 variant).\n\nSigned-off-by: Alexander Monakov \u003camonakov@ispras.ru\u003e\nSigned-off-by: Mikhail Romanov \u003cmmromanov@ispras.ru\u003e\nReviewed-by: Richard Henderson \u003crichard.henderson@linaro.org\u003e\nMessage-Id: \u003c20240206204809.9859-6-amonakov@ispras.ru\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "00118d649ef0f1e14dd31ca7514a85c0fb97ede7",
      "old_mode": 33188,
      "old_path": "util/bufferiszero.c",
      "new_id": "02df82b4ff88fe4ca3945ca679c0e046fc6c0ea8",
      "new_mode": 33188,
      "new_path": "util/bufferiszero.c"
    }
  ]
}