[llvm] [AArch64] Optimize extending loads of small vectors (PR #163064)

[David Green via llvm-commits]

================
@@ -7188,12 +7230,89 @@ SDValue AArch64TargetLowering::LowerStore128(SDValue Op,
   return Result;
 }
 
+/// Helper function to optimize loads of extended small vectors.
+/// These patterns would otherwise get scalarized into inefficient sequences.
+static SDValue performSmallVectorLoadExtCombine(LoadSDNode *Load,
----------------
davemgreen wrote:

Perhaps performSmallVectorLoadExtCombine -> lowerSmallVectorExtLoad, as this is a lowering.

https://github.com/llvm/llvm-project/pull/163064