[PATCH] D97628: GlobalISel: Merge and cleanup more AMDGPU call lowering code

[Amara Emerson via Phabricator via llvm-commits]

aemerson added a comment.

Thanks for doing this. I think we should try to document things here extra carefully. I have to re-learn what this code does every few months.



================
Comment at: llvm/lib/CodeGen/GlobalISel/CallLowering.cpp:284
 
-  const int NumWide = LCMTy.getSizeInBits() / PartLLT.getSizeInBits();
-  Register Undef = B.buildUndef(PartLLT).getReg(0);
+  Register UnmergeSrcReg;
+  if (LCMTy != PartLLT) {
----------------
An example in a comment would be helpful here.


================
Comment at: llvm/lib/CodeGen/GlobalISel/CallLowering.cpp:398
 
+static void buildCopyToRegs(MachineIRBuilder &B, ArrayRef<Register> DstRegs,
+                            Register SrcReg, LLT SrcTy, LLT PartTy,
----------------
Document this?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D97628/new/

https://reviews.llvm.org/D97628