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

[Matt Arsenault via llvm-commits]

================
@@ -369,9 +450,269 @@ bool AMDGPUCodeGenPrepareImpl::run(Function &F) {
       }
     }
   }
+
+  // GlobalISel should directly use the values, and do not need to emit
+  // CopyTo/CopyFrom Regs across blocks
+  if (UsesGlobalISel)
----------------
arsenm wrote:

This one isn't easy because of how the fallback works. The full pipeline always runs, the IR optimizations aren't redone after the fallback hits 

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