)]}'
{
  "commit": "efadc51d46f5581ee15190b23c449dbce941f390",
  "tree": "3abeca4c3a37ccf00081bb748364e02ccd8d62f3",
  "parents": [
    "fd3db2d0ed7157bc44adf94ece9a35ca6b2ff27f"
  ],
  "author": {
    "name": "Kunwar Grover",
    "email": "groverkss@gmail.com",
    "time": "Wed Feb 25 11:42:16 2026 +0000"
  },
  "committer": {
    "name": "GitHub",
    "email": "noreply@github.com",
    "time": "Wed Feb 25 11:42:16 2026 +0000"
  },
  "message": "[LLVMGPU] Remove fold unit extent dims from vector distribute (#22906)\n\nThis unit dim folding was added early on when the pipeline was created.\nIt helps reduce some cognitive load in tests, but adds a lot of\nimplementation overhead for each new codegen op to implement unit dim\nextent folding. Vectorization completly fails if this unit dim folding\ndoesn\u0027t work.\n\nRemove unit dim folding from the codegen pipeline and handle it in\nvector distribute properly. We can drop a lot of code for droppping unit\ndims for codegen ops after this patch.\n\nEventually, this work will lead to dropping `outer_tile` from the\nlayout, which will eventually reduce the number of final vector\ndimensions.\n\nci-extra: test_torch",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "b93814925b5c8c43ea2f3d4e3f6b59b0f5847e9a",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/LLVMGPUVectorDistribute.cpp",
      "new_id": "ff05c335022434ae8afbeae1c879a571bed712da",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/LLVMGPUVectorDistribute.cpp"
    },
    {
      "type": "modify",
      "old_id": "5afb04d5b85d5c1ce01d1237523026d5e82b36a7",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/Passes.cpp",
      "new_id": "a481e0dc45febee26549ba400828879fd2a85ea6",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/Passes.cpp"
    },
    {
      "type": "modify",
      "old_id": "8bc6b878c74fc8c466eb73a46891acc08062f6c0",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_gfx942.mlir",
      "new_id": "d2b536dfca753a70c756954a59ad1da300c7da9f",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_gfx942.mlir"
    },
    {
      "type": "modify",
      "old_id": "e8fca6bcbd845b0ca0404de7d3392b892750e963",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_gfx950.mlir",
      "new_id": "5eaebabb37262977934079b0674161f46d2a7d0c",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_gfx950.mlir"
    },
    {
      "type": "modify",
      "old_id": "edf7648249c004298b7afc7bb7d294280968f58a",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_reduction_gfx942.mlir",
      "new_id": "a74ca5e8bf16ba18703324e08dbdffcfafc17834",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/pipeline_vector_distribute_reduction_gfx942.mlir"
    },
    {
      "type": "modify",
      "old_id": "87738c69638027f12bdde489740a4c2c77bbaff1",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/reduction_pipeline_cuda.mlir",
      "new_id": "89f15a16dec089d09bc86d78fc7b6ad47337376e",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/reduction_pipeline_cuda.mlir"
    },
    {
      "type": "modify",
      "old_id": "d48dc93ec9476b5f43ceb418e1c04f9de5204446",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/reduction_pipeline_rocm.mlir",
      "new_id": "7d7d4feac4c5b45b3034830a98a7dd30289d47d4",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMGPU/test/reduction_pipeline_rocm.mlir"
    }
  ]
}