[all-commits] [llvm/llvm-project] b533b0: Define a DataFlowSolver helper that loads sensible...

[Jeremy Kun via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: b533b0ec34ac36d8a6af406d1fb046e07f95f717
      https://github.com/llvm/llvm-project/commit/b533b0ec34ac36d8a6af406d1fb046e07f95f717
  Author: Jeremy Kun <jkun at google.com>
  Date:   2025-06-20 (Fri, 20 Jun 2025)

  Changed paths:
    A mlir/include/mlir/Analysis/DataFlow/Utils.h
    M mlir/lib/Analysis/DataFlow/LivenessAnalysis.cpp
    M mlir/lib/Dialect/XeGPU/Transforms/XeGPUSubgroupDistribute.cpp
    M mlir/test/lib/Analysis/DataFlow/TestDenseBackwardDataFlowAnalysis.cpp
    M mlir/test/lib/Analysis/DataFlow/TestDenseForwardDataFlowAnalysis.cpp
    M mlir/test/lib/Analysis/DataFlow/TestSparseBackwardDataFlowAnalysis.cpp

  Log Message:
  -----------
  Define a DataFlowSolver helper that loads sensible default analyses (#143415)

Cf. https://discourse.llvm.org/t/mlir-dead-code-analysis/67568/10

Custom analysis passes will not work properly unless both
DeadCodeAnalysis and SparseConstantPropagation are loaded to the
DataFlowSolver. This is intended behavior, but surprising to many users
as shown in the thread. In lieu of a longer-term fix (which I am not
knowledgeable enough to implement myself, yet), this commit adds a
helper function that loads these two analyses, as well as providing
breadcrumbs for an explanation of the problem. The existing places in
the codebase where these two analyses are loaded for the purpose of
running other unrelated analyses are replaced by the use of the helper.

---------

Co-authored-by: Jeremy Kun <j2kun at users.noreply.github.com>
Co-authored-by: Oleksandr "Alex" Zinenko <azinenko at amd.com>



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