[PATCH] D93659: [AIX][XCOFF] emit vector info of traceback table.

Digger Lin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Jun 3 12:10:01 PDT 2021


DiggerLin marked 2 inline comments as done.
DiggerLin added inline comments.


================
Comment at: llvm/include/llvm/CodeGen/MachineRegisterInfo.h:857
+  bool isPhysRegModified(MCRegister PhysReg, bool SkipNoReturnDef = false,
+                         bool SkipTestUsedPhysRegMask = false) const;
 
----------------
jasonliu wrote:
> If this is necessary, could we add a test case for it? Right now, even setting the third parameter to always be false, I don't see any lit failing. 
I do not think we need to new test case ,
in the function MachineRegisterInfo::isPhysRegUsed , we change the 

if (UsedPhysRegMask.test(PhysReg)) 
  return true;

---> 

if (!SkipTestUsedPhysRegMask && UsedPhysRegMask.test(PhysReg))
    return true;

and the the default value of SkipTestUsedPhysRegMask  is false,  it means , all the behaviors not change.

the behaviors only changes when SkipTestUsedPhysRegMask   =true;

and we have invoked
MRI.isPhysRegModified(Reg, false, true) in several place in the function  PPCAIXAsmPrinter::emitTracebackTable. 
I think we have test  SkipTestUsedPhysRegMask   =true too.



================
Comment at: llvm/lib/Target/PowerPC/PPCMachineFunctionInfo.h:123
+  /// HasVectorParms - True if ther are vector parameters
+  bool HasVectorParms = false;
 
----------------
jasonliu wrote:
> Do you need to store this info? 
> For hasVectorParms() function, could you do `return VectorParmsNum;` instead? Or you could also get rid of that function and use `getVectorParmsNum() != 0`.
yes, I can delete the variable .


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93659/new/

https://reviews.llvm.org/D93659



More information about the llvm-commits mailing list