[all-commits] [llvm/llvm-project] f61612: [SeparateConstOffsetFromGEP] Remove support for ar...

[Nikita Popov via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: f6161271a369f278dfad8529348a122f309315d6
      https://github.com/llvm/llvm-project/commit/f6161271a369f278dfad8529348a122f309315d6
  Author: Nikita Popov <npopov at redhat.com>
  Date:   2025-08-01 (Fri, 01 Aug 2025)

  Changed paths:
    M llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp
    M llvm/test/CodeGen/AArch64/aarch64-gep-opt.ll
    M llvm/test/Transforms/SeparateConstOffsetFromGEP/split-gep-or-as-add.ll
    M llvm/test/Transforms/SeparateConstOffsetFromGEP/split-gep-sub.ll

  Log Message:
  -----------
  [SeparateConstOffsetFromGEP] Remove support for arithmetic lowering (#151477)

I don't think there is any benefit to lowering to ptrtoint + arithmetic
+ inttoptr over the newer ptradd lowering. Even if a target does not use
codegen AA, it probably still has IR passes that benefit from correct
representation.

As far as I can tell, no targets actually use this configuration anymore
(they either don't use the LowerGEP option, or they they UseAA and thus
the ptradd lowering).



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