)]}'
{
  "commit": "a0f36282a8e6b517891ce15f89d6ee914875c5f2",
  "tree": "702d715bcc7dc5280a56eb33fb56092bb0d866e5",
  "parents": [
    "f4f3451e0b947b628b8de6eba4c19473cced1ebb"
  ],
  "author": {
    "name": "Diego Caballero",
    "email": "diegocaballero@google.com",
    "time": "Tue Nov 07 10:37:42 2023 -0800"
  },
  "committer": {
    "name": "GitHub",
    "email": "noreply@github.com",
    "time": "Tue Nov 07 10:37:42 2023 -0800"
  },
  "message": "[CPU][SVE] Fix peeling for scalable-tiled loops (#15460)\n\nLLVMCPUPeel in IREE tried to peel the same loop multiple times, and for\r\nscalable vectors it succeeded :). This PR changes a bit the algorithm to\r\nensure that loops are only visited once while preserving the existing\r\nbehavior. I think we should eventually use the loop specialization pass\r\nfrom upstream. It looks really robust and has flag to also peel loops\r\nnested in partial iterations.",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "4d74dfee49f9fac2e82a30f79d8f1984afeba426",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMCPU/LLVMCPUPeel.cpp",
      "new_id": "7ac35284dd041a2db48800799471ec92a41ea85f",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMCPU/LLVMCPUPeel.cpp"
    },
    {
      "type": "modify",
      "old_id": "decd671b5865b6f6c6f4eaa99c87302a8011ed99",
      "old_mode": 33188,
      "old_path": "compiler/src/iree/compiler/Codegen/LLVMCPU/test/peel_and_vectorize.mlir",
      "new_id": "a768109bb04a2f4d5ea163faba393272f2f03b12",
      "new_mode": 33188,
      "new_path": "compiler/src/iree/compiler/Codegen/LLVMCPU/test/peel_and_vectorize.mlir"
    }
  ]
}