[llvm] [AMDGPU][True16][CodeGen] Support AND/OR/XOR and LDEXP True16 format (PR #102620)

[Matt Arsenault via llvm-commits]

================
@@ -236,5 +243,38 @@ bool GCNPreRAOptimizations::runOnMachineFunction(MachineFunction &MF) {
     Changed |= processReg(Reg);
   }
 
+  if (!ST.useRealTrue16Insts())
+    return Changed;
+
+  // Add RA hints to improve True16 COPY elimination.
----------------
arsenm wrote:

Is this really necessary? It adds an extra walk over the function, for something I assume RA would naturally do anyway. In any case, this should be done as a separate change 

https://github.com/llvm/llvm-project/pull/102620