[llvm] [AArch64][Machine-Combiner] Split loads into lanes of neon vectors into multiple vectors when possible (PR #142941)

[Jonathan Cohen via llvm-commits]

================
@@ -7367,11 +7372,245 @@ static bool getMiscPatterns(MachineInstr &Root,
   return false;
 }
 
+static bool getGatherPattern(MachineInstr &Root,
+                             SmallVectorImpl<unsigned> &Patterns,
+                             unsigned LoadLaneOpCode, unsigned NumLanes) {
+  // Early exit if optimizing for size.
+  if (Root.getMF()->getFunction().hasMinSize())
----------------
jcohen-apple wrote:

Done

https://github.com/llvm/llvm-project/pull/142941