[llvm-branch-commits] [llvm] [AMDGPU] Make AMDGPURewriteAGPRCopyMFMA aware of subreg reload (PR #174998)

[Christudasan Devadasan via llvm-branch-commits]

================
@@ -546,6 +546,16 @@ TargetRegisterInfo::getRegSizeInBits(Register Reg,
   return getRegSizeInBits(*RC);
 }
 
+unsigned
+TargetRegisterInfo::getSubRegIdxFromLaneMask(LaneBitmask LaneMask) const {
----------------
cdevadas wrote:

`getCoveringSubRegIndexes` isn't useful at all. There is an early exit when the LankeMask exactly matches to a SubReg index.
https://github.com/llvm/llvm-project/blob/main/llvm/lib/CodeGen/TargetRegisterInfo.cpp#L562

https://github.com/llvm/llvm-project/pull/174998