[llvm] [AMDGPU][DAG] Allow extract_subvector for free register copy (PR #214152)

[Matt Arsenault via llvm-commits]

================
@@ -17041,6 +17119,38 @@ SITargetLowering::performExtractVectorEltCombine(SDNode *N,
   return SDValue();
 }
 
+// extract_subvector (extract_subvector X, I), J -> extract_subvector X, I + J
----------------
arsenm wrote:

Generc combiner should do this (and I suspect it already does?)

https://github.com/llvm/llvm-project/pull/214152