[PATCH] D119013: [ArgPromotion][AMDGPU] New MSSA-based function argument promotion pass with input/output argument support

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.
Herald added subscribers: nlopes, kosarev.
Herald added a project: All.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp:746
+        // Add pass to promote arguments passed by reference
+        if (Level.getSpeedupLevel() >= OptimizationLevel::O3.getSpeedupLevel())
+          PM.addPass(MSSAArgPromotionPass());
----------------
vpykhtin wrote:
> arsenm wrote:
> > Why require all the way to -O3? -O2?
> I thought its quite "agressive" kind of opt, may be 02 should suite better.
I think it should be in the default, which is -O2


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D119013/new/

https://reviews.llvm.org/D119013