[all-commits] [llvm/llvm-project] 674ee4: [LoopSink] Use MemorySSA with legacy pass manager

[Nikita Popov via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 674ee4d353ede9bbcb233aeb5e641ef4e17c7897
      https://github.com/llvm/llvm-project/commit/674ee4d353ede9bbcb233aeb5e641ef4e17c7897
  Author: Nikita Popov <npopov at redhat.com>
  Date:   2022-04-07 (Thu, 07 Apr 2022)

  Changed paths:
    M llvm/lib/Transforms/Scalar/LoopSink.cpp
    M llvm/test/CodeGen/AMDGPU/opt-pipeline.ll

  Log Message:
  -----------
  [LoopSink] Use MemorySSA with legacy pass manager

LoopSink with the legacy pass manager still uses AST, because we
can't compute MemorySSA conditionally. I think now that the legacy
pass manager will be removed soon(TM) we don't need to care about
compile-time impact here anymore. Additionally, since MemorySSA is
no longer eagerly optimized, the impact is actually not that high
anymore (~0.2% geomean regression on CTMark).

This just makes legacy PM and new PM behavior line up -- as a
followup I'll drop these options entirely and make MemorySSA use
mandatory.

Differential Revision: https://reviews.llvm.org/D123216