[llvm] [AMDGPU] Add IR LiveReg type-based optimization (PR #66838)

[Matt Arsenault via llvm-commits]

================
@@ -81,6 +81,88 @@ class AMDGPULateCodeGenPrepare
   bool visitLoadInst(LoadInst &LI);
 };
 
+class ConversionCandidateInfo {
+private:
+  // The instruction which defined the original virtual register used across
+  // blocks
+  Instruction *LiveRegDef;
+  // The original type
+  Type *OriginalType;
+  // The desired type
+  Type *NewType;
+  // The instruction sequence that converts the virtual register, to be used
+  // instead of the original
+  Instruction *Converted = nullptr;
+  // The builder used to build the conversion instruction
+  IRBuilder<> ConvertBuilder;
+
+public:
+  // The instruction which defined the original virtual register used across
----------------
arsenm wrote:

/// for doxygen comments 

https://github.com/llvm/llvm-project/pull/66838