[llvm] [AArch64] Add @llvm.experimental.vector.match (PR #101974)

[Paul Walker via llvm-commits]

================
@@ -1771,6 +1771,11 @@ class TargetTransformInfo {
   /// This should also apply to lowering for vector funnel shifts (rotates).
   bool isVectorShiftByScalarCheap(Type *Ty) const;
 
+  /// \returns True if the target has hardware support for vector match
+  /// operations between vectors of type `VT` and search vectors of `SearchSize`
+  /// elements, and false otherwise.
+  bool hasVectorMatch(VectorType *VT, unsigned SearchSize) const;
----------------
paulwalker-arm wrote:

For such situations the cost model should be used.  Perhaps this in itself might fall on to a TTI method but I think if we reach the point of agreeing the intrinsic is necessary then once things have settled I figure a proper ISD node would be the next step and then the cost model and talk purely in the context of whether the operation is legal and based the cost of that.

https://github.com/llvm/llvm-project/pull/101974