[PATCH] D64596: [AMDGPU] Skip calculating callee saved registers for entry function.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

What is this supposed to solve?



================
Comment at: llvm/lib/Target/AMDGPU/SIFrameLowering.cpp:985-986
   SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  if (MFI->isEntryFunction())
+    return;
+
----------------
Why wouldn't this skip the base call?


================
Comment at: llvm/lib/Target/AMDGPU/SIFrameLowering.cpp:1055-1057
 void SIFrameLowering::determineCalleeSavesSGPR(MachineFunction &MF,
                                                BitVector &SavedRegs,
                                                RegScavenger *RS) const {
----------------
SGPR version should also be skipped?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64596/new/

https://reviews.llvm.org/D64596