[all-commits] [llvm/llvm-project] 3324f4: [mlir][sparse] avoid incompatible linalg fuse-into...

[Aart Bik via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 3324f4d4f4bd82bc9fd43062d21a450671a3531b
      https://github.com/llvm/llvm-project/commit/3324f4d4f4bd82bc9fd43062d21a450671a3531b
  Author: Aart Bik <ajcbik at google.com>
  Date:   2024-03-26 (Tue, 26 Mar 2024)

  Changed paths:
    M mlir/lib/Dialect/Tensor/IR/TensorOps.cpp
    A mlir/test/Dialect/SparseTensor/no_fold_into_consumer.mlir

  Log Message:
  -----------
  [mlir][sparse] avoid incompatible linalg fuse-into-consumer (#86752)

This fixes an "infinite" loop bug, where the incoming IR was repeatedly
rewritten while adding identical cast operations. The test for
compatible types should include the notion of an encoding. If it
differs, then a
naive fusion into the consumer is invalid.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications