[llvm] [AArch64] Increase fixed-length SVE scalarization costs (PR #226120)

[Sander de Smalen via llvm-commits]

================
@@ -4897,17 +4897,16 @@ InstructionCost AArch64TTIImpl::getScalarizationOverhead(
   // TODO: Model the scalarization overhead of wide fixed-length SVE vectors
   // accurately.
   std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(Ty);
-  if (ST->useSVEForFixedLengthVectors(LT.second)) {
-    return BaseT::getScalarizationOverhead(Ty, DemandedElts, Insert, Extract,
-                                           CostKind) +
-           5;
-  }
+  unsigned FixedLengthSVECost =
----------------
sdesmalen-arm wrote:

nit: can you add a comment describing why we're adding an extra cost here?

https://github.com/llvm/llvm-project/pull/226120