[AArch64] Add MATCH loops to LoopIdiomVectorizePass

This patch adds a new loop to LoopIdiomVectorizePass, enabling it to
recognise and use @llvm.experimental.vector.match to vectorise loops
such as:

    char* find_first_of(char *first, char *last,
                        char *s_first, char *s_last) {
      for (; first != last; ++first)
        for (char *it = s_first; it != s_last; ++it)
          if (*first == *it)
            return first;
      return last;
    }

These loops match the C++ standard library's std::find_first_of.

Author: rj-jesus

llvm/include/llvm/Analysis/TargetTransformInfoImpl.h

@@ -775,7 +775,9 @@ class TargetTransformInfoImplBase {
     default:
       break;
     case Intrinsic::experimental_vector_histogram_add:
-      // For now, we want explicit support from the target for histograms.
+    case Intrinsic::experimental_vector_match:
+      // For now, we want explicit support from the target for histograms and
+      // matches.
       return InstructionCost::getInvalid();
     case Intrinsic::allow_runtime_check:
     case Intrinsic::allow_ubsan_check:

llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

@@ -914,6 +914,23 @@ AArch64TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     }
     break;
   }
+  case Intrinsic::experimental_vector_match: {
+    EVT SearchVT = getTLI()->getValueType(DL, ICA.getArgTypes()[0]);
+    unsigned SearchSize =
+        cast<FixedVectorType>(ICA.getArgTypes()[1])->getNumElements();
+    // If we can't lower to MATCH, return an invalid cost.
+    if (getTLI()->shouldExpandVectorMatch(SearchVT, SearchSize))
+      return InstructionCost::getInvalid();
+    // Base cost for MATCH instructions. At least on the Neoverse V2 and
+    // Neoverse V3 these are cheap operations with the same latency as a vector
+    // ADD, though in most cases we also need to do an extra DUP.
+    InstructionCost Cost = 4;
+    // For fixed-length vectors we currently need an extra five--six
+    // instructions besides the MATCH.
+    if (isa<FixedVectorType>(RetTy))
+      Cost += 6;
+    return Cost;
+  }
   default:
     break;
   }