)]}'
{
  "commit": "40356038df5efd4e9a40d18f6e2b9219566e1a65",
  "tree": "90f9d63fd8a6ee7ffa933bfaccbaf9db23ef8fae",
  "parents": [
    "02c2000795e157e4cf63fbac89d21a1ed886a7b0"
  ],
  "author": {
    "name": "Stanley Winata",
    "email": "68087699+raikonenfnu@users.noreply.github.com",
    "time": "Fri Jul 12 15:36:42 2024 -0700"
  },
  "committer": {
    "name": "GitHub",
    "email": "noreply@github.com",
    "time": "Fri Jul 12 15:36:42 2024 -0700"
  },
  "message": "[LLVMGPU][VectorDist] Enable support to distribute vector.transfer_write with non-contiguous dims (#17895)\n\nThis commit adds support to distribute vector.transfer_write with\r\nnon-contiguous dims. It does so by adding better support to detect and\r\nhandle \"projected\"/unused dimensions.\r\n\r\nThis is especially useful for attention-transpose fusion, where it\r\nintroduces a vector.transfer_write with indexing_map and types that\r\nlooks like:\r\n```\r\npermutation_map \u003d affine_map\u003c(d0, d1, d2, d3) -\u003e (d1, d3)\u003e} : vector\u003c32x64xf16\u003e, memref\u003c2x4096x10x64xf16\u003e\r\n```\r\n\r\nPreviously it used to crash on an assertion error which stated that it\u0027s\r\nan invalid indexing_map. This fixes that.\r\n\r\n---------\r\n\r\nCo-authored-by: Kunwar Grover \u003cgroverkss@gmail.com\u003e",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "8837891699f7cbe358f1829e05e9a670ce5ddb00",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/Common/GPU/GPUDistributionPatterns.cpp",
      "new_id": "4b3ae90362b9c58bfa1a8cb133e313eb94607c4f",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/Common/GPU/GPUDistributionPatterns.cpp"
    },
    {
      "type": "modify",
      "old_id": "ebeca29d4bbf8c52d0d2cba3416f185e79da6ef8",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/Common/GPU/test/gpu_vector_distribution.mlir",
      "new_id": "e5ecf598359b2c0eb2d39e59fca3fcadd4297c35",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/Common/GPU/test/gpu_vector_distribution.mlir"
    }
  ]
}